$121 m LIBRARY 3 ,a Michigan State 200 5 University This is to certify that the dissertation entitled A Systematic Evaluation of Culturally Sensitive HIV/AIDS Prevention Interventions in the US, 1996-2007 presented by Ya-Chien Wang has been accepted towards fulfillment of the requirements for the ' Ph.D. degree in - Sociology A” rffig Major Pro ess r’s Signature 2/ 8; [£003 Date MSU is an affirmative-action, equal-opportunity employer .—--.-----~c--.--------cc---n-o-c-n—-.-o--c-o---.----- PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 5/08 K1IProj/Acc8PreleIRCIDateDue indd A SYSTEMATIC EVALUATION OF CULTURALLY SENSITIVE HIV/AIDS PREVENTION INTERVENTIONS IN THE US, 1996-2007 By Ya-Chien Wang A DISSERTATION Submitted to Michigan State University In partial fulfillment of the requirements for the degree of DOCTOR OI" PHILOSOPHY Sociology 2008 ABSTRACT A SYSTEMATIC EVALUATION OF CULTURALLY SENSITIVE HIV/AIDS PREVENTION INTERVENTIONS IN THE US, 1996-2007 By Ya-Chien Wang This dissertation is based on 50 culturally sensitive HIV/AIDS prevention intervention (CSHAPI) studies in either randomized controlled trial or quasi-experiment design, consisting of 26.500 study participants. Five cultural groups were identified: drug users, men having sex with men, Afiican Americans, Hispanic Americans, and White Americans. This dissertation summarizes profiles of the CSHAPI studies, systematically examines the study qualities, and meta-analyzes the long-term effectiveness of CSHAPIS in comparison with that of non-CSHAPI. It was found that the CSHAPIS were more likely to target females, adults, single individuals, and people with low socioeconomic status. The most common intervention objective is improving behavioral skills. The culturally sensitive features were heterogeneous among the cultural groups under review. The most common is matching facilitators or film actors with the participants’ ethnic status or with the significant experience, such as drug use. This dissertation identified threats to internal validity including a lack of random sampling before intervention and inadequate attention to attrition analysis. The major threats to statistical conclusion validity included no reports on reliability of behavioral measures (98%) and a lack of intention-to-treat analysis (66%). The threats to construct validity included a lack of report of allocation concealment (76%) and a lack of report of double blinding (84%). The meta-analysis found that although non-CSHAPIS were more effective than CSHAPIs at post-tests, the advantages tended to attenuate over time. In contrast, the effectiveness of CSHAPIs either increased over time or at least remained at similar levels. It was found that CSHAPIs were more effective than non-CSHAPIS after a one- year period. However, homogeneity tests showed that cultural sensitivity was not a sufficient factor in explaining variation of intervention effectiveness between CSHAPIs and non- C SHAPIs. Among the cultural sensitive indicators, an intervention focus on (ethnic) pride was consistently related to less intervention effectiveness. Cultural intensity (number of cultural sensitive indicators) had a weak relationship with intervention effectiveness. Among the intervention objectives, interpersonal skills and behavioral skills training (such as condom use) were positively related to condom use. Suggestions for future research are provided given the findings discussed above. Copyright by Ya-Chien Wang 2008 ACKNOWLEDGEMENTS Throughout my. graduate training at Michigan State University, I have received great support from faculty in different disciplines. The faculty members in the department of Sociology have instilled in me knowledge and perspectives in Sociology, while those in other departments have broadened my perspectives and my skills. This dissertation is an interdisciplinary work built upon my prior training. Without the help and support, this dissertation would not have been possible. I especially thank Timothy Levine, Sabrina McCormick, Harry Perlstadt, John Schweitzer, and Toby Ten Eyck for their valuable contributions serving on my dissertation committee. My advisor, Dr. Ten Eyck has greatly influenced my dissertation by his constant and thorough support. Dr. Levine and Dr. Perlstadt have provided critical examinations of my dissertation. Dr. McCormick and Dr. Schweitzer have broadened the basis of my dissertation by their perspectives in theory and urban studies. I am also indebted to Stan Kaplowitz, Vijayan Pillai (the University of Texas at Arlington), and Maria Tatto. Dr. Kaplowitz and Dr. Pillai helped me initiate conceptual ideas in the early stage of the dissertation. I especially thank Dr. Pillai for his continual support throughout the process of completing my work. I learned a great deal of meta- analysis from him. I thank Dr. Tatto for the excellent program evaluation course she taught, which helped me finalize conceptual models of the work. Other faculty members, such as Dr. Steve Gold and Dr. Harriette McAdoo, have also greatly inspired me. I also thank Dr. Jack Colford (the University of California at Berkeley) for the meta-analysis course he taught at the University of Michigan at Ann Arbor, which provided a knowledge base of meta-analysis. I also wish to express my gratitude to the Graduate School for the financial support in the form of a MSU Research Enhancement Award I received in 2005. I am very thankful for the Center for Statistical Consultation and Research at the University of Michigan at Ann Arbor. I am especially indebted to Dr. Ananda Sen and Lingling Zhang for their help in polishing the meta-analysis I conducted for my dissertation. Finally, I would like to express my appreciation to my family members for their continual support and understanding throughout my graduate training. Without them, I cannot imagine completing my graduate training. Colleagues and friends also played important roles in my work. I thank my colleagues in Sociology for their graduate training and making it socially and personally meaningful. Thanks to Meng-Jia Wu for suggesting important works in meta-analysis. Thanks to Geoffrey Duh and Youfeng Zhu for smoothing out the process of finalizing my dissertation. I cannot express my gratitude enough and to all others who may have indirectly influenced my graduate training and my dissertation one way or another. Thank you all. vi TABLE OF CONTENTS LIST OF TABLES .............................................................. X LIST OF FIGURES ............................................................. Xii CHAPTER 1 INTRODUCTION Program Statement ........................................................................ 1 Definition of HIV/AIDS and Its Preventive Behavior ............................... 2 Significance and Purposes of this Dissertation ....................................... 4 Research Questions .................................................................. 5 Methodology .............................................................................. 7 Organization of this Dissertation ........................................................ 9 Reference .................................................................................. 10 CHAPTER 2 LITERATURE REVIEW Introduction ................................................................................ 1 2 Early HIV/AIDS Prevention Interventions ............................................ 13 Social and Political Aspects of the HIV/AIDS Epidemic ............................ 14 Social Structure and the Disadvantaged HIV/AIDS High-risk Groups ............ 17 Limited Applications of Structural Intervention and its Evaluation ................ 19 Culturally Sensitive HIV/AIDS Prevention Interventions ........................... 21 Culture .................................................................................. 22 Social Oppression and Sub-culture .................................................. 23 Racial/ethnic Culture and Gender Roles ........................................... 25 Theories from which Culturally Sensitive Interventions Are Derived ........ 26 Culturally Sensitive Frameworks of HIV/AIDS Prevention Intervention for 28 Racial/ethnic Groups ................................................................. Hypotheses ................................................................................. 30 Reference .................................................................................. 33 CHAPTER 3 A PROFILE of CULTURALLY SENSITIVE HIV/AIDS PREVENTION INTERVENTION Introduction ................................................................................ 38 Literature Review ......................................................................... 39 Culturally Sensitive Research for Different HIV/AIDS High-risk Groups .. 41 African American ................................................................ 41 Hispanic American .............................................................. 44 Injection Drug Users ............................................................ 46 Men Having Sex with Men .................................................... 49 vii Methodology .............................................................................. 50 Criteria of Eligibility ................................................................. 51 Criteria of Exclusion ................................................................. 55 Data Collection ....................................................................... 56 Findings .................................................................................... 59 Demographic Characteristics ....................................................... 59 Study Designs ........................................................................ 65 Contents of culturally sensitive HIV/AIDS prevention interventions ........ 67 General Features .................................................................. 67 Drug User Studies ................................................................. 71 Studies Targeting African Americans only ............................... 71 Studies Targeting Mixed Racial/ethnic Groups ......................... 73 Men Having Sex with Men Studies ........................................... 73 African American Studies ...................................................... 75 Hispanic American Studies ..................................................... 76 White American Studies ........................................................ 77 Conclusion ................................................................................. 78 Reference .................................................................................. 82 CHAPTER 4 A SYSTEMATIC QUALITY ASSESSMENT OF CULTURALLY SENSITIVE HIV/AIDS PREVENTION INTERVENTIONS IN THE US, 1996-2007 Introduction ................................................................................ 92 Literature Review .......................................................................... 94 Methodology ............................................................................... l 00 Samples ................................................................................. 100 Quality Measures Used ............................................................... 101 Data Analysis ......................................................................... 102 Findings .................................................................................... 104 Quality of Sampling and Study Designs .......................................... 106 Quality of Program Description .................................................... 107 Quality of Data Analysis ............................................................ 107 Quality of Study Results ............................................................. 109 Quality of Study Conclusions ....................................................... 110 Allocation, Concealment, Blinding, and Intention-to-treat Analysis .......... 112 Level of Study Quality by Subgroups .............................................. 114 Conclusion ................................................................................. 1 15 Reference ................................................................................... 124 viii CHAPTER 5 EFFECTIVENESS OF CULTURALLY SENSITIVE HIV/AIDS PREVENTION INTERVENTIONS IN THE US, 1996-2007: A META-ANALYSIS Introduction ................................................................................ 1 28 Methodology ............................................................................... 129 Samples ..................................................................................... 129 Dependent Variables ................................................................. 13 1 Validity Assessment .................................................................. 131 Data Abstraction ...................................................................... 133 Multi-Follow-ups ..................................................................... 133 Types of Effect Sizes ................................................................. 134 Missing Data ........................................................................... 135 Data Analysis .......................................................................... 135 Variation of Intervention Effectiveness ........................................ 136 Subgroup Analysis .............................................................. 138 Publication Bias .................................................................. 138 Findings .................................................................................... 139 Demographic Characteristics ........................................................ 139 Sample Differences between CSHAPI and Non-CSHAPI studies. . . . 139 Comparison of Effectiveness of CSHAPIS with Non-CSHAPIS .............. 142 Unprotected Sex .................................................................. 142 Condom Use ...................................................................... 147 The Effectiveness of CSHAPIs by Subgroup Differences ...................... 152 Publication Bias ...................................................................... 165 Conclusion ................................................................................. 169 Reference .................................................................................. 175 CHAPTER 6 CONCLUSION .............................................................................................. 178 APPENDICES Appendix A Coding Sheet ............................................................... 183 Appendix B Quality Scale ............................................................... 205 3.1 3.2 3.3 3.4 3.5 4.1 4.2 4.3 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 LIST OF TABLES Demographic Characteristics of the Participants in Culturally Sensitive HIV/AIDS Prevention Interventions .................................................. Aggregate Estimation of the socioeconomic and Marital status of the Participants in Culturally Sensitive HIV/AIDS Prevention Interventions ...... Theories Used in Culturally Sensitive HIV/AIDS Prevention Interventions... Description of Culturally Sensitive HIV/AIDS Prevention Interventions ...... Cultural Indicators and Culturally Sensitive HIV/AIDS Prevention Interventions ............................................................................. Threats to Internal Validity and External Validity ................................. Scoring Systems .......................................................................... Level of Study Quality by Domains in the Evaluations ........................... Demographic Characteristics of Participants by CSHAPI and non-CSHAPI studies ................................................................................... Aggregate Estimate of Age, Socioeconomic Status, and Marital Status of Participants by CSHAPI and Non-CSHAPI Studies .............................. Comparison of the Effectiveness of CSHAPIs with non-CSHAPIS on Reducing Unprotected Sex Behavior ................................................ Comparison of the Effectiveness of CSHAPIs with Non-CSHAPIS on Enhancing Condom Use Behavior ................................................. Effectiveness of CSHAPIs by Time Points and Dependent Variables .......... Effect Sizes (Cohen’s d) of Culturally Sensitive Interventions by Time Point CSHAPI Effectiveness on Enhancing Condom Use Behavior: a Subgroup Analysis .................................................................................. CSHAPI Effectiveness on Reducing Unprotected Sex Behavior: a Subgroup Analysis ................................................................................. 62 63 69 70 72 96 103 105 140 141 145 146 149 150 153 154 5.9 5.10 5.11 5.12 5.13 5.14 5.15 Meta-regression on Relationships between Age, SES, and CSHAPI Effectiveness on Enhancing Condom Use at Time 1 and Time 3 ............... Meta-regression on Relationships between Age, SES, and Effectiveness of Culturally Sensitive Interventions on Reducing Unprotected Sex Behavior at Time 1 and Time 3 ..................................................................... Culturally Sensitive Indicators and CSHAPI Effectiveness on Increasing Condom Use ............................................................................. Culturally Sensitive Indicator and CSHAPI Effectiveness on Reducing Unprotected Sex ........................................................................ Relationships between Intervention Components and CSHAPI Effectiveness on Increasing Condom Use ............................................................ Relationships between Study Quality and CSHAPI Effectiveness on Condom Use ............................................................................ F ail-Safe N of C SHAPI Effectiveness by Dependent Variables and Time Points ..................................................................................... xi 155 155 158 159 162 163 166 3.1 4.1 4.2 5.1 5.2 5.3 5.4 5.5 5.6 LIST OF FIGURES Flow Chart of Data Screening .......................................................... Distribution of Accumulative Study Quality ........................................ Level of Study Quality by Culturally Sensitive Interventions ..................... Intervention Effectiveness by Cultural Sensitivity and Time Points: Unprotected Sex ......................................................................... Intervention Effectiveness by Cultural Sensitivity and Time Points: Condom Use ........................................................................................ Funnel Plot of CSHAPI Effectiveness on Increasing Condom Use at Time 1.. Funnel Plot of CSHAPI Effectiveness on Increasing Condom Use at Time 3.. Funnel Plot of C SHAPI Effectiveness on Reducing Unprotected Sex at Time 1 ........................................................................................... Funnel Plot of CSHAPI Effectiveness on Reducing Unprotected Sex at Time 3 ........................................................................................... xii 58 105 115 151 152 166 167 167 CHAPTER 1 INTRODUCTION Problem Statement More than two decades after the outbreak of the HIV/AIDS epidemic in 1982, this disease is still a public health challenge in the US. In 2005, the estimated number of diagnoses of AIDS in the US. and dependent areas was 41,897, while the estimated cumulative number of AIDS patients was 984,155 (The Center for Disease Control and Prevention [CDC], 2007). However, the number of reported AIDS cases is disguised by the incubation from infection of HIV to onset of AIDS, which takes between two to fifteen years. It is estimated that at the end of 2003, about 24-27 % of 1,039,000 to 1,185,000 people living with HIV/AIDS were undiagnosed (CDC, 2007). The potential undiagnosed HIV/AIDS cases confound people’s risk of HIV infection as studies indicate that some people tend to estimate their risk of HIV infection by selection of sexual partners rather than by their actual safe sex behavior (Wight, 1999). While a variety of new scientific breakthroughs have been invented to alleviate the affliction of AIDS, scientific limits in developing effective medical vaccines and cures for HIV/AIDS have made its prevention imperative. Since 1992, the general public and the public health community have started to direct their attention to the importance of HIV/AIDS prevention targeted toward disadvantaged high-risk HIV groups. These groups often contain subcultures of their own, which are distinctively different from the mainstream culture. Consequently, culturally sensitive prevention interventions are believed to be effective because people often act in accord with the behavioral expectations which are culturally meaningful in their self-identified communities. Although HIV/AIDS prevention has recognized the importance of culture, evaluation of the effectiveness of cultural sensitivity in HIV/AIDS prevention interventions has rarely been examined (e.g., Dushay, Singer, Weeks, Rohena, and Gruber, 2001). Among the few quantitative systematic reviews with a focus on cultural sensitivity (Jemmott III and Jemmott, 2000; Herbst et a1., 2007), only Herbst and colleagues have investigated the relationships between the Hispanic cultural features of intervention and its effectiveness. There is still a lack of quantitative systematic reviews of culturally sensitive HIV/AIDS prevention interventions (CSHAPIs) toward other HIV/AIDS high-risk subpopulations. However, using evaluation information to plan a program is necessary to ensure the effectiveness and efficiency of any HIV/AIDS prevention efforts. More specifically, the questions of interest to program planners and key policy makers are: whether a particular intervention for a social group will work for other social groups? At what level may the intervention change participants’ behavior? And, what is the optimal mix of HIV prevention interventions in a given locale (Holtgrave, Gilliam, Gentry, and Sy, 2002)? Therefore, evaluations of how effective culturally sensitive interventions are to a specific high-risk HIV/AIDS group are greatly needed. Evaluations of those interventions across different HIV/AIDS high-risk groups may aid in generalization of the findings. This dissertation is a meta-analysis, addressing the above questions. Definition of HIV/AIDS and Its Preventive Behavior HIV is a virus that destroys the blood cells of CD4 positive T cells, which are crucial to the normal function of the human immune system. As defined by the Center for Disease Control and Prevention, HIV-infected people are those who have fewer than 200 CD4 positive T cells per cubic millimeter of blood, by contrast to healthy adults who usually have CD4 positive T-cell counts of 1,000 or more. When people first become infected with HIV, they often do not have any symptoms. However, AIDS, the advanced stage of HIV infection, is associated with 26 clinical conditions, most of which are opportunistic infections that generally do not affect healthy people. These infections are often severe and sometimes fatal in people with AIDS because their immune systems are so ravaged by HIV that their bodies cannot fight off certain bacteria, viruses, fungi, parasites, and other microbes. The time of progression from HIV to AIDS varies greatly from person to person. It depends on many factors, including a person's health status and their health-related behaviors (National Institute of Allergy and Infectious Diseases. 2003). HIV may be passed from one person to another via infected blood, semen, or vag inal secretions through an uninfected person’s broken skin or mucous membranes. In add i tion, infected pregnant women can pass HIV to their babies during pregnancy or deli Very, as well as through breast-feeding. Hence, individuals’ behaviors that put themselves at the risk of HIV infection may include 1) having unprotected sexual contact With an infected person or a person whose HIV status is unknown, 2) sharing needles and/or syringes (primarily for drug injection) with someone who is infected or whose HIV status is unknown, or 3) having transfusions of infected blood or blood clotting fact()rs (National Institute of Allergy and Infectious Diseases, 2003). Research of HI\f/AIDS prevention often defines an individual’s HIV/AIDS preventive behavior as using latex condoms, restraining from needle sharing, or reducing the number of sexual panil‘lers. This dissertation uses condom use or unprotected sex as a behavioral measure of CSHAPI effectiveness, as each of them is related to a level of risk for HIV infection. Condom use is protected sex. while unprotected sex is the frequency of sex deducted by protected sex. Significance and Purposes of this Dissertation In the field of CSHAPIs no theory has been developed to guide intervention design and implementation. The pilot efforts to develop a culturally sensitive framework and indicators in CSHAPIs (Bemal, Bonilla, and Bellido, 1995; Marr'n , 1993; Vinh- Thomas, Bunch, and Card, 2003) tended to contain a broad range of culturally sensitive factors and represented an ideal type of CSHAPI in terms of cultural sensitivity. These complete sets of frameworks and indicators have not been empirically tested in any single study or systematic review. On the other hand, the evidence of the effectiveness of CSHAPIS has been mixed part ly due to the heterogeneity of this research. The existent empirical CSHAPIs have been diverse in terms of conceptualization and operationalization of cultural sensitivity, Other intervention content, intervention strategies, subjects, and intended intervention outc omes. Different levels of study quality may also contribute to the heterogeneity of resStarch. This dissertation aims to fill this gap in terms of empirical evidence of the effectiveness of the CSHAPIs via a qualitative review and meta-analysis. It provides a qualitatively detailed account of the existent CSHAPIs, especially about the way cultural Ser‘lSitivity has been differently operationalized in empirical studies and the different ways that the C SHAPIs have been conducted with different subpopulation groups. In addition. this dissertation systematically evaluates collective quality of the CSHAPIs, using a quality scale, in order to assess causal inference and external validity that the primary studies can make. Via meta-analysis this dissertation estimates the importance of different culturally sensitive factors in facilitating behavior change. This is done through a comparison of studies containing different culturally sensitive contents and strategies. The findings about the impact of different culturally sensitive indicators on intervention effectiveness will facilitate program development of interventions. Given that resources available to interventions are usually restricted in terms of time and money, intervention planners may prioritize the incorporation of those culturally sensitive contents and strategies that show greater association with behavior change. In addition, this dissertation compares the studies with different designs and intervention content (such as education or communication skill training) to evaluate the extent to which these factors moderate the effectiveness of the CSHAPIs. An exploratory anal ysis about factors that may moderate the effectiveness of the CSHAPIs aids the understanding about intervention context under which culturally sensitive contents and Strategies may be more effective. This may be a step to build and test an overarching frarl‘lework that bridges culturally sensitive perspectives and other essential components of C SHAPIs. Res\earch Questions 1) Profiles of the Existent CSHAPIs o What are the general characteristics of the existent CSHAPIS, in terms of their participants, program contents, strategies. culturally sensitive contents and strategies, and study designs? How do the existent CSHAPIs conceptualize cultural sensitivity and operationalize the concept accordingly? Which culturally sensitive contents and strategies are widely used? 0 Do applications of CSHAPIs differ by subcultural groups? Do conceptualization and operationalization of CSHAPIs vary by cultural groups? Which subcultural groups are less likely to receive CSHAPIs? 2) Qualities of the Existent CSHAPI Studies 0 What are the study qualities of the existent CSHAPIS in terms of their designs, implementations, causal inference, and external validity? o Are the study qualities of interventions associated with the effectiveness of interventions? 3) Relationships between the Cultural Sensitivity and Effectiveness of CSHAPls 0 Do the existent CSHAPI studies provide evidence that CSHAPIs are more effective in promoting behavioral change, compared to interventions without cultural sensitivity? 0 Are other intervention contents and strategies or participants’ demographic characteristics associated with the effectiveness of CSHAPIS? Methodology This dissertation will conduct a systematic review of the existent empirical studies. It differs from a traditional literature review in that it searches literature, evaluates the quality of these studies, and extracts values of eligible studies with prior established criteria. In addition, the criteria and process of searching studies and analyzing data are explicitly documented to ensure that the review is reproducible. The systematic approach avoids the bias of traditional literature reviews induced by their reviewers’ biased idiosyncratic decisions in retrieving and interpreting literature. In addition, because the systematic review is explicitly detailed, examination of its objectivity and generalization is possible (Fink, 1998). This dissertation will utilize both a qualitative review (Patton, 2001) and meta- analysis. This parallels Cooper’s approach of integrated literature review in which two kinds of evidence are integrated. One is “synthesis-generated evidence,” which is provided by an analysis of relevant studies in a research body. The other is “study- generated evidence,” which is provided by a well-designed single study (Cooper, 1998). The use of meta—analysis in this dissertation is to quantitatively synthesize findings across stud ies to gather “synthesis-generated evidence” about relations contained across studies (COOper, 1998). Meta-analysis can estimate overall effectiveness of a specific intervention by synthesizing studies which use the same type of intervention. The r esearch questions dealing with effectiveness (and its magnitude) of different types of CSHAPIS to a particular subcultural group and its generalization within and acl‘Osssubcultural groups could be tested only by combining and comparing these studies in a synthesized review. Meta-analysis may handle such generalizations of the findings by formal statistical methods (Hedges, 1994). Because studies are heterogeneous in design or operationalization, a comparison to find the factors that contribute to the differences in the findings is no less important than estimating overall magnitude of the effectiveness of an intervention. The heterogeneity of the empirical studies on CSHAPIs necessitates a qualitative review of their general profiles in order to document the different ways that the empirical studies have been conceptualized and operationalized with regards to cultural sensitivity, to learn to what extent CSHAPls have been conducted with different subcultural groups, and to identify research gaps. In addition, a qualitative review may also complement a meta-analysis and gather “study-generated evidence” (Cooper, 1998). Study-generated evidence is a single study containing results that directly test the relationships being considered. A study that uses a randomized experiment may provide direct evidence of causes of behavioral or attitudinal change. In a randomized experiment, participants are randomly assigned into a control or experiment group so that factors other than experimental manipulation that may confound experimental outcomes are controlled. Therefore, behavioral or attitudinal change of people in the experimental group could be attributed only to the experimental manipulation. Causal inference of a synthesized review will depend on those underlying studies because a meta-analysis cannot randomly assign the studies to different subgroups. As a result, the boundaries to which causal inference could be drawn from a synthesized review will be defined by the underlying primary studies (Cooper 1998; Hall, Tickle- Degnen, Rosenthal, and Mosteller 1994). Organization of this Dissertation This dissertation consists of three independent manuscripts, each of which deals with a distinctive area of the research questions. Each manuscript is based on its preceding chapter and leads to its following chapter. I review relevant literature in Chapter 2. Chapter 3 summarizes and discusses the profile of the existent CSHAPIs. Chapter 4 is a systematic evaluation of the quality of the existent CSHAPI studies. Chapter 5 is an analysis of the relationships between cultural sensitivity and effectiveness of CSHAPIs, its generalization, and the tests of the hypotheses. In Chapter 6, I summarize findings from each chapter and suggest directions of theory development for CSHAPI. Based on these findings and ideas, I close this dissertation by presenting areas for fiiture research of theory and intervention. References Bemal, G., Bonilla, J., & Bellido, C. (1995). Ecological validity and cultural sensitivity for outcome research: Issues for the cultural adaptation and development of psychosocial treatments with Hispanics. Journal of A bnormal Child Psychology, 23, 67-82. Center for Disease Prevention and Control. (2007). Cases of HIV infection and AIDS in the United States and Dependent Areas, 2005. Retrieved December 15‘, 2008 from the World Wide Web: http://www.ch.gow’hiv/topics/surveillance/resourecs/reports/2005report/default.htm Cooper, Harris. 1998. Synthesizing research: A guidefor literature review (Vol. 2). Thousand Oaks, California: Sage Publications, Inc. Dushay, R. A., Singer, M., Weeks, M. R., Rohena, L., & Gruber, R. (2001). Lowering HIV risk among ethnic minority drug Users: Comparing culturally targeted intervention to a standard intervention. American Journal of Drug and Alcohol Abuse, 27, 501-524. Fink, Arlene. 1998. Conducting research literature reviews. Thousand Oaks, California: Sage Publications Ltds. Hall, J. A., Tickle-Degnen, L., Rosenthal, R., & Mosteller, F. (1994). Hypotheses and problems in research synthesis. In H. Cooper and & L. V. Hedges (Eds), The handbook of research synthesis (pp. 1 7-29). New York: Russell Sage Foundation. Hedges, L. V. (1994). Statistical considerations In H. Cooper & L. V. Hedges (Eds), The handbook of research synthesis (pp.29-40). New York: Russell Sage Foundation. Herbst, J. H., Kay, L. S., Passin, W. F., Lyles, C. M., Crepaz, N., & Marl'n, B.V. (2007). A systematic review and meta-analysis of behavioral interventions to reduce HIV risk behaviors of Hispanics in the United States and Puerto Rico. AIDS and Behavior, 11(1), 25-47. Holtgrave, D. R., Gilliam, A., Gentry, D., & Sy, F. S. (2002). Evaluating HIV prevention efforts to reduce new infections and ensure accountability. AIDS Education and Prevention, 14, 1-4. Jemmott III, .1. B., & Jemmott, L. S. (2000). HIV risk reduction behavioral interventions with heterosexual adolescents. AIDS, 14(Suppl. 2), 40-52. Man’n, B. V. (1993). Developing culture-specific interventions for Latinas to reduce HIV-Risk behaviors. Journal of the Association ofNurses in AIDS Care, 11(3), 70-76. 10 National Institution of Allergy and Infectious Diseases. (2003). HIV/AIDS fact sheets and brochures. Retrieved December I“, 2008 from the World Wide Web: http://www3.niaid.nih.gov/hcalthscience/healthtopics/HIVAIDS/default.htm Patton, M. Q. (2001). Qualitative research and evaluation methods (3rd ed.). California, Thousand Oaks: Sage Publications. Vinh-Thomas, P., Bunch, M. M., & Card, J. J. (2003). A research-based tool for identifying and strengthening culturally competent and evaluation-ready HIV/AIDS prevention programs. AIDS Education and Prevention 15, 481-498. Wight, D. (1999). Cultural factors in young heterosexual men’s perception of HIV risk. Sociology of Health and Illness, 21, 735-758. 11 CHAPTER 2 LITERATURE REVIEW Introduction I first reviewed the literature in a historical perspective from a sociological point of view, in order to shed light on various political, social, structural, and cultural factors that may have facilitated the spread of AIDS and have continually shaped individual HIV/AIDS preventive behavior. The impact of these factors may converge at a societal, institutional, or community level, and may condition intervention effectiveness on changing individual HIV/AIDS preventive behavior. HIV/AIDS prevention intervention is an organizational product shaped by various institutional forces. I reviewed different types of HIV/AIDS prevention interventions in accordance with their underlying social forces. By examining various factors at macro- and meso-levels, the effectiveness of culturally sensitive HIV/AIDS prevention interventions (CSHAPIs) could be evaluated with a sociological understanding of the importance of social contexts in which the CSHAPIs and individuals may be situated. In addition, I examined relevant sociological concepts, such as social oppression, subculture, and gender roles, which may be used to explain unequal demographic distribution of the HIV/AIDS epidemic and the varying effectiveness of CSHAPIs conducted with subpopulations in different social positions. I also investigated the development of program theory of CSHAPIs and its application to racial/ethnic groups having their own subcultures. Based on the literature review, I developed study hypotheses that would be tested in Chapter 5. 12 Early HIV/AIDS Prevention Interventions HIV/AIDS prevention interventions are an organized effort to reduce the spread of HIV/AIDS. In the US, earlier public health programs to stop HIV considered HIV/AIDS as preventable to a significant degree via the practice of individuals’ HIV/AIDS preventive behavior, similar to the efforts to fight other major diseases, such as cancer, heart disease, and stroke (Vinh-Thomas, Bunch, and Card, 2003). The design of HIV/AIDS prevention interventions has often utilized cognitive behavioral models, such as the Health Belief Model (Becker, 1974; Rosenstock, 1974; Rose, 1996), and the Theory of Reasoned Action (Fishbein and Ajzen, 1975; Ajzen and F ishbein, 1980; Kinsler, Sneed, Morisky, and Aug, 2004), which served as guides to intervention content. The components of these behavioral models in general focus on individuals’ health beliefs about causation of illness, attitude toward preventive behavior, perception of group norms, and/or self-perception of efficacy. HIV/AIDS prevention interventions usually incorporate at least one of the following intervention contents to enhance HIV/AIDS preventive behavior: increasing individuals’ HIV/AIDS-related knowledge, promoting individuals’ positive attitudes toward HIV/AIDS preventive behavior, promoting individuals’ skills of sexual communication, and enhancing individuals’ perception of self-efficacy and/or self-esteem (Card, Benner, Shields, and Feinstein, 2001) The early focus of HIV/AIDS public health campaigns primarily targeted individuals of certain HIV high-risk groups except in the case of gay communities in which grass-rooted community organizations successfully mobilized a change in the community norm regarding safe sex (Rushing, 1995). Even if the interventions were 13 conducted in a group format, such as a small group discussion and a large group lecture, the purpose of the intervention dealt with individuals, aiming at enhancing individual awareness of their risk and advocating individual preventive behavior. However, the fact that the HIV/AIDS epidemic has spread indicates the limit of effectiveness of these early educational HIV/AIDS prevention interventions with an individual focus. Syntheses of the intervention studies also found that evidence of effectiveness of the early HIV/AIDS prevention interventions is not consistent, especially for African American women (Logan, Cole, and Leukefeld, 2002; Mize, Robinson, Bockting, and Scheltema, 2002). The failure of earlier organizations to stop the HIV/AIDS epidemic necessitated the examination of factors other than individuals which may have facilitated the spread of the epidemic. Social and Political Aspects of the HIV/AIDS Epidemic Sociologists (e.g., Rushing, 1995) and anthropologists (e.g., Herdt and Lindenbaum, 1992; Setel, 1999; Parker, 2001) have examined social factors beyond the individual level which may have impacted the HIV/AIDS epidemic. From a sociological point of view, individuals’ problems are conditioned by the social contexts in which they are situated. TheHIV/AIDS epidemic was not caused singly by a failure of demoralized individuals’ self-control in sexual encounters. Examination of the HIV/AIDS epidemic requires more than an individual perspective. The early history of HIV/AIDS prevention proves that a variety of social and political factors have combined to contribute to the spread of HIV/AIDS. At the social level, the sexual liberation in the US. since the 19605 has provided a fertile field for sexually transmitted diseases, including syphilis. Since the 19605. the 14 population has increased sexual behavior due to the sexual revolution, as reflected in the increasing rates of syphilis during the 19603 (Kiefer and Hulley, 1990). Furthermore, a growing problem with crack cocaine and other recreational drugs has been associated with a sharp increase of syphilis since 1985 (Kiefer and Hulley, I990). The excessively high rate of HIV/AIDS in the gay communities at the beginning of the AIDS epidemic led to the early misrepresentation of AIDS by the medical community as a gay disease (Clatts, 1995), which has been stigmatized along with homosexuality and the sex positive culture of gay communities (Rushing, 1995). The use of sexuality categories in early HIV/AIDS prevention had often failed to differentiate between the sexual identity and the sexual behavior of individuals (Clatts, 1995). It had influenced the population to estimate their risk of HIV infection by their membership to the HIV high-risk groups. Biased estimates of HIV/AIDS risk are especially true among people with lower socioeconomic status. Studies show that people with lower socioeconomic status are more likely to estimate their HIV risk by their sexual relationships with people of HIV high-risk groups while people with higher socioeconomic status are more likely to estimate their HIV risk by their actual unsafe sexual behavior (Wight, 1999). Furthermore, HIV/AIDS preventive behaviors were stigmatized themselves. It was not until the 19905 that the idea of condom use as an effective way of HIV/AIDS prevention was adopted by the government at state and federal levels. In 1991, the New York City Board of Education approved a HIV/AIDS initiative, which included condom availability in high schools. In addition, programs which provided free needles to drug 15 users were likely to encounter political and community resistance due to the debate of the preference of drug treatment to HIV/AIDS prevention (Anderson, 1993). Political and financial factors also impacted the HIV/AIDS epidemic. The stigmatization of HIV/AIDS as “gay cancer” led directly to stalling funding for AIDS prevention in the early years of the epidemic. Due to a lack of media recognition of the HIV/AIDS epidemic, the US p0pulation was unaware of their risk of HIV infection. It was not until 1992 that the public became aware of the overwhelming effect of HIV/AIDS partly due to the fact that AIDS had become the number one cause of death for U. S. men aged 25 to 44 and remained so through 1995 (Public Broadcasting Service [PBS], 2005). When HIV/AIDS increasingly spread outside the gay communities, moral- attributing factors were no longer fit to address the cause of the continuing HIV/AIDS epidemic. It was not until the late 19803 that the political sectors started to allocate funds for HIV/AIDS social research and prevention. For example, in 1988 the US. Health Omnibus Programs Extension (HOPE) Act authorized the use of federal funds for AIDS prevention, education, and testing. In 1993, federal funding began assisting in community planning of HIV/AIDS prevention and in establishing women’s interagency HIV study (WIHS) and HIV Epidemiology Study (HERS). After Reagan’s administration’s attitude of denial toward the prevalence of HIV/AIDS, the new administration’s attitude toward HIV/AIDS was more open to the problem (PBS, 2005). The exacerbating and continuing AIDS epidemic has reflected the failure of political, social, and organizational factors in responding to the epidemic. After the endeavors of public health campaigns targeted towards individuals utilizing rational 16 behavioral models in the 19905, the AIDS epidemic has continued and increasingly resided in communities of disadvantaged minorities. Social Structure and the Disadvantaged HIV/AIDS High-risk Groups The pattern of HIV/AIDS distribution in the US. appears to reflect a pattern of social inequality in the US. population. While HIV/AIDS continued to take a high toll on men having sex with men (MSM) in 2005, racial/ethnic minorities and women have become the subpopulations hit hardest by the HIV/AIDS epidemic. In 2005, African Americans, which made up 13% of the US. population, accounted for about 49 % of new HIV/AIDS cases. In addition, in 2004, AIDS was the leading cause of death for African American women aged 25—34 years while it is the sixth leading cause of death for all women within the same age range (CDC, 2007b). In 2005, of the 126,964 women living with HIV/AIDS, 64% were African Americans (CDC, 2007a). HIV/AIDS is no longer limited to gay communities, but has become a disease which has lingered mostly among these disadvantaged due to a combination of political, social, cultural, and behavioral factors (Anderson, 1993). The highly concentrated rate of HIV/AIDS cases among these disadvantaged groups has complicated the issue of availability and quality of health care for AIDS patients. Because African Americans are more likely to be impoverished, this group has more cases, shorter survival times, and more deaths due to AIDS than any other racial/ethnic groups (CDC, 2007a). By comparing HIV prevalence among groups with different social positions, research reveals the effect of structure on the differential health status of people. Different forms of social discrimination ranging from social oppression of employment, 17 education, health care services, daily interaction, etc., manifest their effect through people’s social positions, such as race/ethnicity, sex, age, and sexual preference. People’s social positions, in turn, determine the ecological environments where people are situated and the resources available to them to buffer social adversity. The high concentration of the later HIV/AIDS epidemic among ethnic minorities, particularly African American women, reveals the relationship between social structure and the spread of HIV/AIDS, as this subpopulation is more likely to reside in communities with concentrated poverty. It indicates that the high prevalence of HIV/AIDS among ethnic minorities may be the consequence of systematic social oppression, instead of racial/ethnic, genetic, or moral factors. Poverty and ecological factors in these communities may have led to the spread of AIDS. Ethnographic studies in HIV/AIDS high-risk communities are of particular relevance to the influence of social structure on the epidemic as they examine the long term changes of a particular ethnic community for reasons other than HIV/AIDS-related interventions. Most anthropological ethnographic work on HIV/AIDS has been conducted in developing countries, such as Parker and Daniel’s work in Brazil (1993), Paul Farmer’s work in Rural Haiti (1993), Kammerer and his colleagues’ work in Northern Thailand (1995), and Obbo’s work in Uganda (1995). Domestic studies of HIV/AIDS high-risk groups with a political economic focus were usually conducted in drug user communities (Wallace, 1990; Anderson, 1993; Groseclose et al., 1995; Gostin, 1998), while domestic studies of gay communities have tended to have a focus on culture (Henriksson and Mannsson, 1995) and/or stress (Vincke and Bolton, 1995). These political economic and cultural studies of HIV/AIDS prevalence in the high-risk HIV/AIDS communities pointed out that multiple folds of structural factors, 18 such as poverty, class, political failure, and social oppression, converged in the ecology of high-risk HIV/AIDS communities which has exposed community members to the risk of HIV/AIDS. Another branch of the literature of social factors relevant to HIV/AIDS prevalence could be derived from studies of health beliefs on HIV/AIDS preventive behavior. Typically, these studies ask participants to identify factors that inhibited preventive behavior. Access to, and cost of, condoms are often the cited structural barriers to condom use among the populations of college students and inner city women. Limited Applications of Structural Intervention and its Evaluation The idea of structural change as a means of HIV/AIDS prevention intervention has been discussed in the literature, as introduced by the social studies of the HIV/AIDS epidemic (e.g., Rushing, 1995; Parker, 2002). However, it was not until recent years that public health professionals started to incorporate the findings about the effects of structural factors into prevention design. Structural intervention targets the immediate social context of sexual or injection behaviors by changing the physical or normative environments within which they occur. It may include programs that directly change legal environments or make safer behavior easier, such as allowing syringes to be sold over the counter or providing free drug needles to drug users (The Center for AIDS Preventidn and Studies at the University of California. 2003; Blankenship, Friedman, Dworkin, and Mantell, 2006). Structural interventions may also include programs that address factors that may have an indirect impact on HIV infection. such as income inequality. racism, and other 19 inequality and oppressions. These factors may create vulnerability to HIV/AIDS. For example, structural intervention that provided micro-financing programs to women may reduce income inequality and thus, increase women’s power in sexual communication. In Blankenship, Bray, and Merson’s study (2000), they classified structural interventions of HIV prevention by individual, organizational, and environmental levels, contingent on issues of availability, acceptability, and accessibility. Structural interventions have become popular with a particular emphasis on comnumity participation and ownership. That a high rate of HIV/AIDS in the inner-city communities in the US. created hardship to wellbeing of the communities has highlighted the needs of structural interventions in the inner cities. Because the HIV-infected individuals are mostly (male) adults, who are the main bread earners in their families. HIV/AIDS prevalence has led to disfunctioning of their families. Furthermore, establishing HIV/AIDS prevention initiatives is particularly difficult in the urban settings because the residents are preoccupied by other issues more urgent to their survival on a daily basis, such as food and safety. Long-term health is less a concern for those with low socioeconomic status. Therefore, structural interventions that improve the inner-city residents’ economic status and the urban environments may be conducive to success of HIV/AIDS prevention. However, empirical evaluation of structural interventions is still rare in the literature. Structural interventions alone tend to involve large amounts of manpower over a long period of time. As the use of structural interventions is new, systematic evaluation of the effectiveness of structural interventions awaits a large number of structural interventions to take place. 20 Culturally Sensitive HIV/AIDS Prevention Interventions In addition to structural interventions, culturally sensitive intervention has been considered effective in stopping the HIV/AIDS epidemic. As the disadvantaged HIV/AIDS high-risk groups tend to have subcultures distinct from mainstream culture, an intervention with a framework of White culture and language is more likely to yield limited effectiveness. A search of major databases (such as Medline, Sociological Abstract, and Psylnfo) and majorjournals (for example, AIDS Education and Prevention and Journal of Counseling and Clinical Psychology) shows that before 1992, there were only sporadic empirical studies on HIV/AIDS prevention interventions that explicitly claim that their interventions fit into ethnic cultural backgrounds (Schinke, Gordon, and Weston, 1990). After 1992, along with the availability of funding for community HIV/AIDS prevention, there was an increasing number of CSHAPIs conducted within HIV/AIDS high-risk communities (e.g., Kalichman, Kelly, Hunter, Murphy, and Tyler, 1993; DiClemente and Wingood, 1995; Choi et al., 1996; Dushay, Singer, Weeks, Rohena, and Gruber, 2001; Sterk, Theall, and Elifson, 2003a). In general, C SHAPIs are different from structural interventions in that most of the CSHAPIs still use a top-down approach and seek to change only individual behavior, instead of changing power relations and values within the targeted groups. CSHAPI is best considered as an organizational effort that communicates HIV/AIDS prevention messages in a culturally-acceptable way to persuade individuals to adopt desirable behaviors. CSHAPIs at most seek to change certain behavior norms that were directly related to the risk of HIV/AIDS infection while the cultural or structural factors behind the specifically targeted behavior norms were usually left intact. The purposes of 21 CSHAPIs may involve transforming a given meaning of undesirable behavior, instilling new meaning linked to desirable behavior, or even competing with other behaviors. For example, in terms of condom use, CSHAPIs may change its connotation from promiscuity to care for partners, and may advocate condoms as a contraceptive means instead of birth control pills. However, the CSHAPI studies have remained very diverse in terms of their definition of cultural sensitivity and its operationalization. The following sections will examine the concept of culture and its related theories that have guided the design of culturally sensitive HIV/AIDS prevention interventions. Sellers The concept of culture itself has remained a very ambiguous construct over time. In general, the term “culture” reflects features of shared characteristics, traditions, and belief systems, such as religious and medical belief systems (Slonim, 1991). It specifies what behaviors are acceptable or unacceptable for different roles in specific social places and specific times. The concept of culture contains an interconnection of the sub- concepts of person, time, place, and behavior (Trostle, 2005). From the cultural point of view, HIV/AIDS preventive behaviors, such as reducing the number of sexual partners or using condoms in sex, may manifest their cultural meanings in connection with specification of time (timing during the week or day), place (such as public or private space), and person (types of relationships with partners). For example, the cultural meaning of the behavior of condom use for women selling sex for money has to be understood in terms of where the sex is conducted (such as a street comer or home), 22 when (such as during the day or evening), and with whom (such as a sex customer or a primary sex partner). The cultural meaning of condom use could be desirable in the social context of selling sex, but undesirable in the social context with a primary sex partner. Instead of considering culture as an external force that restrains people’s behavior, Anderson and F enichel (1989) cautioned that a cultural framework is only a set of tendencies or possibilities for behavior. Individuals within any given society are essentially free to choose from all of the available possibilities. They may also appropriate and creatively interpret the culture in their own way within the frame of the mainstream culture or a subculture of any major social group (Wyn and White, 2003). Social Oppression and Subculture The view of individuals as users or appropriators of culture fails to recognize the effect of social oppression on people’s choice of behavioral options within given social contexts. The concept of subculture is particularly relevant to illustrate the effect of social oppression in cultural production. According to Sebald (1984), the necessary condition that propels a subculture to emerge is the existence of a number of persons in society who seek solutions to common problems that are often related to their social divisions. The sufficient condition is that there is acceptance of common norms and values in the group which specify the proper way of doing things among them. The system of norms and values of the subculture is saliently different from that of the mainstream culture. Therefore, from the point of view of a subculture, disadvantaged people, who are denied the access to social participation and/or economic production. are 23 likely to form their own subculture to satisfy their social and/or material needs. The subculture provides social support and status which the members of a subculture cannot find outside their group. The subcultures of inner city drug users, those living under poverty, MSM, and others who are socially oppressed because of their sexual preferences deviate from mainstream sexual culture, exemplify the relationship between social and economic oppression and cultural production. For example, Greenberg, in his work of the construction of homosexuality (1988), indicated that from Renaissance through the eighteenth century, social responses to homosexuality were characterized by extremely harsh legislation justified primarily on religious grounds, erratic enforcement, and popular indifference to homosexuality. He also linked more recent social repression of homosexuality to rise of competitive capitalism that encouraged austerity and self-restraint. In addition, development of bureaucracies has tended to suppress affective emotional responses towards males. It was during the 19703 that popular attitudes shifted in the direction of greater tolerance. Gay subculture symbolized by sex institutions has sustained sex positive norms in gay communities. Sex institutions provide access to social connections among homosexuals, which they cannot find outside the sex institution. In addition, sex positive norms in gay communities fimctioned to provide the homosexuals a source of gay pride. “Since male sexual preference for another male was the focus of homosexual oppression, when oppressive measures were lifted, sexual freedom and gay pride were joined.”(Rushing. 1995, p. 30) Similarly, shooting galleries (the institution of drug users) are the product of social, political, and economic oppression toward this group. Legal restrictions on the 24 sale of needles have indirectly promoted the existence of shooting galleries. Purchasing or exchanging drug use equipment in the shooting galleries alleviated the risk of carrying drugs and drug paraphernalia over long stretches. In addition, as drug use was considered deviant at a social level, drug use behavior tended to take place outside drug users’ communities of residence. In shooting galleries, drug users could access drug use equipment and socially connect with others. Furthermore, drug needle sharing may function as a symbol of friendship and trust. Mixing blood in drug users may symbolize brotherhood (Inciardi, 1992; Rushing, I995). Racial/ethnic Culture and Gender Roles The causes of the high concentration of HIV/AIDS among racial/ethnic minorities are also related to traditional gender roles and Catholic religious beliefs in certain racial/ethnic cultures. African and Hispanic cultures have held a conservative attitude toward gender roles that shows a great acceptance of sexual promiscuity for men and a restricted role for women. For example, Hispanic women are expected to be sexually inexperienced and passive, whereas Hispanic men are expected to be sexually dominant, knowledgeable about sex, and to have other sexual partners before marriage. However, the sexual promiscuity of men may put their sexual partners at risk of HIV/AIDS infection because women were unable to negotiate safe sex due to the lack of power in relationships. This lack of power also stemmed from women’s economic dependency on their sexual partners (Logan, Cole, and Leukefeld, 2002). Furthermore, the behavior of condom use contains culturally undesirable meanings. Condom use may imply lack of trust toward sexual partners. On the other 25 hand, it may also imply sexual promiscuity of the person who asked to use a condom. Either of the meanings may restrain women from acting on their intention to negotiate condom use with their partners when they do not have power in the relationship. Furthermore, Hispanic people tend to have a strong belief in Catholicism, which opposes the use of condoms due to its belief in life as a gift from God. Condom use in sex contradicts the Hispanic religious belief system, which is a strong social institution in their communities. In 2002, about 73% of Hispanics in the U. S. were Catholic (United States Conference of Catholic Bishops, 2007). Theories from which Culturally Sensitive Interventions Are Derived The concept of culturally appropriate intervention is largely derived from the theories of culturally competent services in health services. It has also been adapted in the fields of disability and mental illness. Culturally competent service is considered to be service provided within the cultural context of the consumers in health care (Stone, 2005). Cultural competency refers to a set of behaviors, attitudes, and policies that may enable a system, agency, or individual provider to function effectively with service recipients and communities with cultural diversity (Randall-David, 1989). In an effort to outline the major components of providing culturally competent care to a diverse group of patients, a panel of experts on culturally competent health care from the American Academy of Nursing concluded that there were no well-tested and tried models that could facilitate the provision of culturally competent care. However, it identified a number of relevant models that were useful. 26 Jezewski’s culture-brokering model was identified as one practical model that offered guidance for the delivery of culturally competent care (Jezewski and Sotnik, 2001 ). Culture brokering is the act of bridging, linking, or mediating between groups or persons of differing cultural backgrounds for the purpose of reducing conflict or producing change. The cross-cultural situation is assumed to be prone to conflicts and problems that inhibit acceptance of services. , Therefore, an effective culture broker needs knowledge, skills, sensitivity, and awareness of cross-cultural variables to solve the conflicts and problems. CSHAPIS could be considered as cultural brokers between public health, which intends to transmit medical and behavioral knowledge in mainstream culture, and HIV/AIDS high-risk groups, which generate and maintain unique racial/ethnic cultures or subcultures. Mari 'n (1993) has defined culturally appropriate interventions within a broader framework. The culturally appropriate strategies for behavioral change in her framework need to meet three basic criteria. The first is that the intervention or treatment is based on basic cultural values of the sub-groups. Ethnic culture differs in basic values and preferred behavioral patterns of social scripts. CSHAPIs based on such cultural basic values are expected to succeed and facilitate individual growth in their own communities or groups. For example, for Asian Americans and Hispanics, familialism is a strong cultural value. The involvement of relatives or the implication of familial responsibilities in an intervention can therefore be assumed to be more appropriate and more effective for Hispanics than for other social groups which may not assign familialism the same personal importance. On the other hand, White Americans value individualism and self- assertiveness. Therefore, interventions that frame behavioral change as the development 27 of individuals’ self-assertiveness and control in life may be culturally appropriate for White Americans. The second. is that the treatments be consonant with the subjective culture (such as attitudes, expectancies, norms) of the particular ethnic group, regarding the behavior of concern and illness. In terms of HIV/AIDS risk behavior, Wilson and Miller (2003) have identified two culturally bound factors as influential in HIV prevention initiatives. One is societal preferences for heterosexual relationships over homosexual ones, and the other is beliefs about gender roles that privilege men over women. Culturally appropriate strategies do not challenge the given significant cultural values, but work within the cultural framework. The third is that the components of the strategies be based on the expectations and behavioral preferences of the ethnic or minority groups. Thus, the intervention module should fit within the preferred behavioral repertoire of the targeted group. The components of the intervention must reflect patterns of use of the possible intervention channels (such as a local radio station or printed media) and are considered credible and usefirl (Mari 'n , 1993). Culturally Sensitive Frameworks of "HIV/AIDS Prevention Intervention for Racial/ethnic Groups In recent years researchers have developed frameworks of culturally sensitive intervention for specific racial/ethnic groups. For example, Bemal, Bonilla, and Bellido (1995) have attempted to construct a framework for culturally sensitive clinical research interventions with Hispanics. In this framework, culturally sensitive intervention is not necessarily intended to change the basic cultural value of the groups, but conveys the 28 desirable behavior in a positive way that is adaptive to the given culture and social milieu. In their framework, culturally sensitive research entails the consideration of the cultural contexts across several dimensions of the intervention, including language, persons, metaphors, content, concepts, goals, methods, and contexts. The culturally sensitive elements for Hispanics include 1) culturally ethnic language that has emotional appeal to them, 2) ethnic/racial similarities and differences between clients and therapists that shape therapy relationships, 3) symbols and concepts shared in the population’s cultural knowledge, 4) treatment concepts consonant with culture and context, 5) transmission of positive and adaptive cultural values, 6) support of adaptive values from the culture of origin, 7) development and/or cultural adaptation of treatment methods, and 8) consideration of the context of assessment during treatment or intervention, which is different from usual cultural contexts. However, the cultural elements suggested by this study remained abstract and did not specify content of intervention which may actualize the cultural elements. Vinh-Thomas, Bunch, and Card (2003) have developed a research-based tool for identifying culturally competent HIV/AIDS prevention interventions. They surveyed the various ways that the concept of cultural competence had been studied and have extended the concept to the field ofHIV/AIDS prevention. They also compiled a comprehensive list of cultural competence indicators in line with the lifecycle of an intervention that ranges from program development to implementation to sustainability. The culturally competent indicators are valuable as they provide a big picture about the way that an “ideal” culturally competent intervention should be conducted and the roles of different cultural competent factors at different stages of the lifecycle of an intervention. However, 29 HIV/AIDS prevention interventions usually have only limited resources over a short period of time. Therefore, effective selection of the salient culturally competent indicators that will sustain risk behavior change may precondition the success of HIV/AIDS prevention intervention. Hypotheses Hypotheses have been developed to examine the central research questions. These hypotheses examine the relationships between the effectiveness of CSHAPIs and participants’ social positions. Testing these hypotheses leads to a generalization of the findings across different subcultural groups. 1) People receiving CSHAPIs are more likely to increase condom use (or to reduce unprotected sex) than those receiving non-CSHAPIs. The main difference between CSHAPIS and standard interventions is that CSHAPIs reframe the behavior of condom use in a culturally desirable way. Because CSHAPIs often aim at enhancing knowledge about HIV/AIDS transmission, reframing cultural meanings of condom use, and/or promoting group norms of condom use, people who receive CSHAPIs may pay attention to the risk of HIV infection via unprotected sex and its social consequences. They may also perceive condom use as culturally acceptable. Therefore, I hypothesize that people who receive CSHAPIs will be more likely to increase condom use as a way to avoid HIV infection than those receiving non- CSHAPls. For similar reasons, I hypothesize that people who receive CSHAPIs will be more likely to reduce unprotected sex. 30 2) The CSHAPI participants with lower socioeconomic status (SES) are less likely to increase condom use (or to reduce unprotected sex) than those with higher SES. Because the cultural sensitivity of CSHAPIs is assumed to be tailored according to the subcultures that intervention participants have memberships with, the effectiveness of CSHAPIs may hold true across subcultural groups. However, lhypothesize that the participants of CSHAPIs with lower SES are less likely to increase their condom use or to reduce unprotected sex than those with higher SES because they have lower perceived behavioral control and more environmental barriers to condom use. As SES is one of the social structural factors that shape distribution of resources and opportunities, people with higher SES often enjoy a multitude of resources. They have a higher sense of personal control as a result of past successful experience and are more adaptive in solving their problem. Self-efficacy in condom use involves many skills, such as sexual communication and management of stress, which people with higher SES are more capable of maintaining. In contrast, people with low educational attainment, restricted employment opportunities, and poor economic circumstances often have learned helplessness and a lower sense of personal control. Furthermore, people with higher SES have less environmental barriers to condom use. The unavailability and cost of condoms, which those disadvantaged women often encountered as environmental barriers, may not be a problem for people with higher SES. Because CSHAPls do not aim at improving the lack of resources and learned helplessness experienced by 31 people with lower SES, I hypothesize that after receiving CSHAPIs, people with lower SES are less likely to increase condom use (or to reduce unprotected sex) than those with higher SES. 32 References Ajzen, I., & F ishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall. Anderson, R, & Fenichel, E. (1989). Serving culturally diverse families of infants and toddlers with disabilities. Monograph of the National Center for Clinical Infant Programs, Washington, DC. Anderson, W. (1993). The New York needle trial: the politics of public health in the age of AIDS. In V. Berridge & P. Strong (Eds), AIDS and contemporary history (pp. 157-181). Cambridge, UK: Cambridge University press. Becker, M. H. (1974). The health belief model and personal health behavior. Health Education Monographs, 2 (4): 324—473. Berkman, A., Garcia, J ., Mufioz-Laboy, M., Paiva, V., & Parker, R. (2005). A critical analysis of the Brazilian response to HIV/AIDS: Lessons learned for controlling and mitigating the epidemic in developing countries. American Journal of Public Health, 95(7), 162-1 172. Bemal, G., Bonilla, J ., & Bellido, C. (1995). Ecological validity and cultural sensitivity for outcome research: Issues for the cultural adaptation and development of psychosocial treatments with Hispanics. Journal of A bnormal Child Psychology, 23, 67-82. Blankenship, K. M., Bray, S. J ., & Merson, M. H. (2000). Structural interventions in public health. AIDS, 14(Suppl. 1), 11-21. Blankenship, K. M., Friedman, S. R., Dworkin, S., & Mantell, J. E. (2006). Structural interventions: concepts, challenges and opportunities for research. Journal of Urban Health, 83(1), 59-72. Card, J. J., Benner, T., Shields, J. P., & Feinstein, N. (2001). The HIV/AIDS prevention program archive (HAPPA): A collection of promising prevention programs in a box. AIDS Education and Prevention, 13, 1-28. The Center for AIDS Prevention and Studies at the University of California. (2003). What is the role of structural interventions in HIV prevention. Retrieved May I", 2008 from the World Wide Web: http://www.caps.ucsf.edu/pubs/FS/structuraling). Center for Disease Prevention and Control. (2007a). Cases of HIV infection and AIDS in the United States and Dependent Areas, 2005. Retrieved May I", 2008 from the World Wide Web: http://wwwcdcgw/hiv/topics/survcillance/resourccs/rcports/2005report/default.htm. 33 Center for Disease Prevention and Control. (2007b). HIV/AIDS among Women. Retrieved May I", 2008 from the World Wide Web: http://wuwcdcgov/hiv/topics/women/resources/factsheets/women.htm. Center for Disease Prevention and Control. (2008). AIDS cases by top 10 states/dependent areas. Retrieved May I", 2008 from the World Wide Web: http://www.cdc.gov/hiv/topics/surveillance/basic.htm#area. Choi, K., Lew, S., Vittinghoff, E., Catania, J. A., Barrett, D. C., & Coates, T. J. (1996). The efficacy of brief group counseling in HIV risk reduction among homosexual Asian and Pacific men. AIDS, 10, 81-87. Clatts, M. C. (1995). Disembodied acts: On the perverse use of sexual categories in the study of high-risk behaviour. In H. T. Brummelhuis & G. Herdt (Eds), Culture and sexual risk: Anthropological perspectives on AIDS (pp. 241-256). Amsterdam, Netherlands: Overseas Publishers. DiClemente, R. J ., & Wingood, G. M. (1995). A randomized controlled trial of an HIV sexual risk-reduction intervention for young African-American women. Journal of the American Medical Association, 274, 1271-1276. Dushay, R. A., Singer, M., Weeks, M. R., Rohena, L., & Gruber, R. (2001). Lowering HIV risk among ethnic minority drug users: comparing culturally targeted intervention to a standard intervention. American Journal of Drug and Alcohol Abuse, 27,501-524. Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley. Gostin, L. O. (1998). The legal environment impeding access to sterile syringe and needles: the conflict between law enforcement and public health. Journal of Acquired Immune Deficiency Syndromes and Human Retrovirology,18(suppl 1), 560-570. Greenberg, D. 1988. The construction ofmodern homosexuality. Chicago: University of Chicago Press. Groseclose, S.L., Weinstein, B., Jones, T.S., et al (1995). Impact of increased legal access to needles and syringes on practices of injecting-drug users and police officers- Connecticut, 1992-1993. Journal ofAcquired Immune Deficiency Syndromes and Human Retrovirology, 10, 82-89. Henriksson, & Mannsson. (1995). Sexual Negotiations: An Ethnographic Study of Men Who Have Sex with Men. In H. T. Brummelhuis & G. Herdt (Eds), Culture and sexual risk: Anthropological perspectives on AIDS (pp. 79-95). Amsterdam, Gordon and Breach Publishers. 34 Herdt, G., & Lindenbaum, S. (1992). The time ofAlDS: Social analysis, theory, and method. Sage Publications, Inc Inciardi, J. (1992). The War on Drugs 11: The continuing Epic of Heroin, Cocaine, Crack. Crimes, AIDS, and Public Policy. Mountain View, Calif: Mayfield. Jezewski, M.A., & Sotnik, P. (2001). Culture brokering: Providing culturally competent rehabilitation services to foreign-born persons. Center for International Rehabilitation Research Information and Exchange. CIRRIE Monograph Series, John Stone, Ed. Buffalo, NY: CIRRIE. Kalichman, S. C., Kelly, J. A., Hunter, T., Murphy, D., & Tyler, R. (1993). Culturally tailored HIV-AIDS risk-reduction messages targeted to African-American urban women: impact in risk sensitization and risk reduction. Journal of Consulting and Clinical Psychology, 61. 291-295. Kammerer, C. A., Hutheesing, O. K., Maneeprasert, R., & Symonds, P. V. (1995). Vulnerability to HIV infection among three hill tribes in Northern Thailand. In H. T. Brummelhuis & G. Herdt (Eds), Culture and sexual Risk: Anthropological perspectives on AIDS (pp. 53-78). Amsterdam, Netherlands: Overseas Publishers Association. Kiefer, R. G., & Hulley, S. B. (1990). A modern epidemic emerges: History and context. In S. Petrow. P. Franks, & T. R. Wolfred (Eds), Ending the HIV epidemic: Community strategies in disease prevention and health promotion (pp. 3-19). Santa Cruz, CA: Network Publications. Kinsler, J., Sneed, C. D., Morisky, D. E., & Ang, A. (2004). Evaluation of a school- based intervention for HIV/AIDS prevention among Belizean adolescents. Health Education Research, I 9(6):730-738. Logan, T. K., Cole, J ., & Leukefeld, C. (2002). Women, sex, and HIV: social and contextual factors, meta-analysis of published interventions, and implications for practice and research. Psychological Bulletin, 128(6), 851-885. Man'n, B. V. (1993). Developing culture-specific interventions for Latinas to reduce HIV-Risk behaviors. Journal of the Association of Nurses in AIDS Care, 1 I (3). 70-76. Mize, S. J. 8., Robinson, B. E., Bockting, W. 0., & Scheltema, K. E. (2002). Meta- analysis of the effectiveness of HIV prevention interventions for women. AIDS Care, 14(2), 163-180. Parker, R.G. & Daniel, H. (1993). Sexuality, politics and AIDS in Brazil: In another world? London: The Falmer Press. 35 Parker, R. (2001). Sexuality, culture, and power in HIV/AIDS research. Annual Review of Anthropology, 3 0, 63-179. Parker, R. (2002). The global HIV/AIDS pandemic. structural Inequalities, and the politics of international health. American Journal of Public Health, 93(3):343-346. Farmer, P. (1993). AIDS and accusation: Haiti and the geography of blame. University of California Press. Public Broadcast Services. (2005). Global Health: America’s Response: AIDS Policy Timeline. Retrieved December I”, 2008 from the World Wide Web: http://wwvv.pbs.org/now/science/aidstimeline.html Rose, M. A. (1996). Effect of an AIDS education program for older adults. Journal of Community Health Nursing, 13 (3), 141-148. Rosenstock, I. (1974). Historical origins of the Health Belief Model. Health Education Monographs, 2 (4): 328-335. Rushing, W. A. (1995). The AIDS epidemic: Social dimensions of an Infectious disease. Westview Press. Schinke, S. R. Gordon, A. N., & Weston, R. E. (1990). Self-instruction to prevent HIV infection among African-American and Hispanic-American adolescents. Journal of Consulting and Clinical Psychology, 58, 432-436. Obbo, C. (1995). Gender, age and class: Discourses on HIV transmission and control in Uganda. In H. T. Brummelhuis & G. H. Amsterdam (Eds), Culture and sexual risk: Anthropological perspectives on AIDS (pp.79-96). Netherlands: Overseas Publishers. Randall-David, E. (1989). Strategies for working with culturally diverse communities and clients. Comprehensive Hemophilia Program, Bowman Gray School of Medicine. Bethesda, MD: the Association for the Care of Children’s Health. (Monograph) Sebald, H. (1984). Etiology and dimensions of subculture. In Adolescence: a Social psychological analysis (pp. 206-225). Englewood Cliffs, New York: Prentice-Hall. Inc. Setel, P. W. (1999). A plague of paradoxes: AIDS, culture, and demography in North Tanzania. Chicago: University of Chicago Press. Stone, J. H. (2005). Culture and disability: Providing culturally competent services. Thousand Oaks, CA: Sage Publications. Slonim, M. (1991). Children. culture and ethnicity: Evaluating and understanding the impact. London: Garland Publishing Inc. 36 Sterk, C. E., Theall, K. P., & Elifson, K. W. (2003a). Effectiveness of a risk reduction intervention among African American women who use crack cocaine, AIDS Education and Prevention, 15, 15-32. Trostle, J. A. (2005). Epidemiology and culture. Cambridge: Cambridge University Press. United States Conference of Catholic Bishops, Hispanic Affairs. (2007). Demographics. Retrieved December I", 2008 from the World Wide Web: http://\n~w.usccb.org/hispanicaffairs/demo.shtml. Vincke, J ., & Bolton, R. (1995). Social stress and risky sex among gay men: An additional explanation for the persistence of unsafe sex. In H. T. Brummelhuis & G. Herd (Eds), Culture and sexual Risk: Anthropological perspectives on AIDS (pp. 183-204). Amsterdam, Netherlands: Overseas Publishers. Vinh-Thomas, P., Bunch, M. M., & Card, J. J. (2003). A research-based tool for identifying and strengthening culturally competent and evaluation-ready HIV/AIDS prevention programs. AIDS Education and Prevention, 15, 481-498. Wallace, R. (1990). Urban desertification, public health and public order: ‘Planned shrinkage,‘ violent death, substance abuse, and AIDS in the Bronx. Social Science and Medicine, 31, 801—813. Wight, D. (1999). Cultural factors in young heterosexual men’s perception of HIV risk, Sociology of Health and Illness, 21, 735-758. Wilson, B. D, & Miller, M. (2003). Examining strategies for culturally grounded HIV prevention: a review. AIDS Education and Prevention, 15,184-202. Wyn, J ., & White, R. (2003). Rethinking Youth. Sage Publications. 37 CHAPTER 3 A PROFILE OF U.S.-BASED CULTURALLY SENSITIVE HIV/AIDS PREVENTION INTERVENTIONS, 1996-2007 Introduction Culture is a term widely used, but poorly defined. When the term culture is used, it usually refers to some subculture that is distinct from mainstream culture. Culturally sensitive HIV/AIDS prevention interventions (C SHAPIs) are indeed an organizational effort to reduce the AIDS spread among subpopulations with distinct subcultures different from the mainstream. Non-culturally sensitive HIV/AIDS (non-CSHAPIs) are designed based on selected socio-demographic attributes of the mainstream culture. Culture in CSHAPIs, however, remains an umbrella term which covers all high- risk subcultural groups targeted by the interventions, while non-culture refers only to the White American culture. A comparison of CSHAPIs with non-CSHAPIs is often confounded by the ambiguous definition of cultural sensitivity and inconsistent operationalizations for different subcultural groups. In addition, the non-CSHAPIs are associated with White American culture. However, the varying level of cultural sensitivity in the CSHAPIs associated with different subcultural groups has rarely been explored in research. Moreover, the use of cultural sensitivity could be ideological based on White American researchers’ perspective on the subcultures rather than the subcultural members’ point of view. Cultural sensitivity could be political when it is used to switch the public’s attention from structural causes of the AIDS epidemic and to provide a basis on which subcultural groups are blamed for their insensitivity to the intervention 38 treatments especially tailored to their subcultures. Thus, an investigation of the cultural sensitive components in the CSHAPIs is the first step leading to a critical examination of the effectiveness of CSHAPIs. This study provides a detailed profile of the CSHAPIs in the U. S., published between 1996 and 2007. It investigated general characteristics of the existent CSHAPIs, in terms of their participants, program contents, strategies, culturally sensitive contents, and study designs. It also examined operationalizations of cultural sensitivity in the CSHAPIs targeting different subcultural groups and identified the gap of the coverage of CSHAPIS. Literature Review The rate of HIV/AIDS among racial/ethnic minorities, men having sex with men (MSM), and injection drug users is disproportionate to their numbers within the general population. However, it was not until recently that researchers have made an attempt to develop culturally sensitive theories that can practically guide HIV/AIDS prevention intervention targeting these subpopulations (Parker, 2001; Vinh-Thomas, Bunch, and Card, 2003; Wilson and Miller, 2003). In a qualitative review of culturally sensitive interventions between 1985 and 2001, Wilson and Miller (2003) examined 17 studies that explicitly incorporated cultural sensitivity in intervention activities. They found that few of the studies articulated how culture functioned within risk behavior models. Vinh-Thomas et al.’s study (2003) recently developed thirty-three culturally competent indicators for identifying culturally competent and evaluation-ready HIV/AIDS prevention programs in line with life cycle and quality of intervention. The culturally competent indicators range from the stages of program development to 39 program implementation to program sustainability to program validation. For example, at the stage of program development, intervention planners must be able to assess needs among the target population. At the stage of program implementation, intervention planners must be able to recruit and retain qualified, culturally competent and representative staff. They must also involve leaders, influential members, and individuals who are respected by the targeted subpopulation. In addition, the majority, about 70-75% of program participants. must belong to the targeted subpopulation. In terms of program sustainability, the intervention must be able to stabilize for at least a year and gain multiyear support for programs from outside sources. With respect to program validation, there must be evidence of participants’ and staff satisfaction with, interest in, or enthusiasm for the program, as well as evidence of intended intervention outcomes. Unfortunately, there is still a lack of research on culturally sensitive intervention theory in line with program lifecycles, specifically for each of the HIV high-risk subpopulations. Further research in this area is imperative. Each culture has different values and norms that guide behavior. Culturally tailored programs for one group may not be suitable for another. In addition, ecological differences (e.g., rural/urban or North/South residence) may diversify cultural variation and available resources (Beatty, Wheeler, and Gaiter, 2004). Therefore, a community—level investigation of the cultural meaning of HIV preventive behavior (e.g. perceived cost and benefit), needs of the targeted population (e.g., priority of behavior), and ecology of targeted communities (e.g., credible source of information) will help develop and implement culturally sensitive HIV/AIDS prevention interventions for major high-risk subcultural groups in community settings (Marr'n, 1993; DiClemente, Crosby, and Wingood, 2005). 40 Culturally sensitive research for differer_1t HIV/AIDS high-risk groups Research has explored a variety of social and cultural risk factors for high-risk subcultural groups which may be incorporated into the development of culturally sensitive HIV/AIDS theory for each high-risk group. Studies have focused on the subpopulations of African Americans, Hispanic Americans, injection drug users, and MSM. Below is a selective review of relevant work for each of the subpopulations with a focus on similarities and dissimilarities of risk factors between the subpopulations. African American Several researchers have suggested a number of social risk factors for African American communities exist, particularly for low-income African American women (Raj. Amaro, and Reed, 2001). The low socioeconomic status of this subpopulation has been associated with higher rates of injection drug use, sex trade. and imbalance of power in gender roles, all of which are high risk factors for HIV infection. Men with low socioeconomic status are more likely to define manhood in traditional gender roles due to their lack of success in society (Bowser, 1994). In addition, African American culture associates sexuality with manhood and womanhood. Being sexually active is considered entering adulthood for African American adolescents. The culture also encourages sexual adventurism, which may put people at risk for HIV. Compared with Whites, African Americans have sexual intercourse at an earlier age, more extramarital sex, and more sexual partners (Rushing, 1995). Bowser (1994) identified several reasons for the deep ambivalence among African Americans toward AIDS prevention. First of all, AIDS is just one of many crises that 41 African Americans with overall low socioeconomic status face. They may prioritize satisfaction of immediate survival needs over a distant threat of a chronic disease, such as AIDS. In addition, African Americans tend to distrust government interventions due to historical events, such as the Tuskegee experiment, which violated African Americans’ human rights in a trial for new drugs to combat Syphilis. Furthermore, the African American underclass in general is characterized by their invisibility and mobility within communities. As a result, the participation and completion rates of the African American underclass in prevention intervention tend to be low (Bowser, 1994). In addition, there is a lack of leadership in the community due to the urban sprawl of African American middle and working classes. As African American middle and working class women are the backbone of African American communities, exiting inner cities makes grass-root community intervention difficult (Bowser, 1994). This is in contrast to the gay communities in the early years of the HIV/AIDS epidemic. The grass- root efforts of the gay communities had effectively reduced new HIV infection cases (Rushing, 1995). Other researchers have indicated the need of cultivating minority leadership in research of and intervention in minority communities (Marin, 2002). Furthermore, African Americans with HIV/AIDS are marginalized people within African American communities. AIDS is considered by some fundamental Christians as a punishment of God to sinners. Researchers have indicated that the African American church is a credible institution to African Americans, which can be actively involved to advocate HIV prevention messages (Jarama, Belgrave, Bradford, Young, and Honnold , 2007). Strong religious beliefs in African American communities have been linked with a higher level of homophobia than that in the general society (e.g., Icard, 1985). Similarly. 42 because issues of drug addiction deprive viability of African American communities, there is an intense dislike for drug addicts. As a result, the distribution of limited community resources to these groups for HIV/AIDS prevention is a disputed issue (Bowser, 1994). African American communities may also prioritize the issue of drug treatment over HIV/AIDS prevention among drug users, as in the first case of free needle distribution in New York (Anderson, 1993). J arama et al.’s study (2007) explored cultural values in African Americans that AIDS prevention intervention can capitalize on. They found that African American women’s gender ideology is characterized by a relational orientation. They tend to hold strong attachments and responsibilities to not only partners, but also members of the immediate and extended family. Therefore. they may prioritize significant others’ needs before their own. Inclusion of family and community responsibilities in HIV/AIDS prevention may be more relevant than individual autonomy. In addition, Jarama et al.’s study (2007) suggests a relevance of the issues of trust and economic dependency in intervention targeted toward African American women. They found that these women tend to hold contradictory views about themselves in relationships. They were socialized to view themselves as sexual objects, especially in economic dependent relationships, while holding hope of building trust within relationships. These contradictory views may pose double risks of HIV infection for these women. 43 Hispanic American Similar to African Americans, low socioeconomic status, injection drug use, traditional cultural values, strong religious beliefs (in Roman Catholicism), and traditional gender roles have often been cited as factors affecting HIV infection among Hispanics (Perloff, 2001; Raj. Amaro, and Reed, 2001). Although researchers have argued that these characterizations typify only a small sector of Hispanic Americans, primarily derived from clinical samples of mostly Mexican American women (Fernandez, 1995), the characterizations highlight the risk factors of HIV infection within a larger portion of this population. More specifically. a cultural twist for Hispanic Americans’ risk for HIV infection centers on preferred sexual behavior and its meaning. Trickkett (2005) indicates a widespread belief among Hispanic Americans that men have little control over sexual impulses. Use of condoms is thought to hamper male autonomy. The social meaning of using condoms signifies the existence of disease, instead of preventing infection (Raj, Amaro, and Reed, 2001). A strand of research related to cultural variation among Hispanic Americans is the role of acculturation in HIV infection (Fernandez, 1995). For example, researchers investigated relationships between the level of acculturation and accessibility to condoms. In San Francisco, less aceulturated Latinas were found to be less likely to carry condoms (Man’n and Marl’n, 1992). Nevertheless, the study findings are usually confounded by the heterogeneity of the construct of acculturation (Fernandez, 1995). In addition, the emphasis on sexual penetration heightens the risk of HIV infection transmitted by body fluids. Exploitation of sexual behavior. such as dominance 44 in relationships, extramarital affairs, and multiple sexual partners, is related to the Hispanic definition of manhood (Perloff, 2001), which is grounded in the ideology of machismo (Logan, 2002). Penetrative sex rather than sex of partners defines men. Studies found that Hispanic American MSM are less likely to self-identify themselves as homosexual or bisexual (Perrow and Gullén, 1990; Trickkett, 2005). The fluidity of self- identification of sexuality complicates the issue of the categorization of HIV/AIDS high- risk groups. Parker (2001) suggested that a close examination of “local categories and classifications” (p. 167) is necessary to uncover the cultural meanings of sexuality which may guide HIV/AIDS prevention initiatives in any specific locale. Surveys found that Hispanic American MSM have a high incidence of unprotected anal intercourse (Perloff, 2001) In addition, compared with White Americans and Asian Americans, a lower percentage of Hispanic American men living with AIDS had been involved with other men (56% in Hispanic Americans vs. 75% in White Americans and 73 % in Asian/Pacific Islanders). However, there was a preportional increase in Hispanic American men living with AIDS, while their racial/ethnic counterparts witnessed a decrease (The Center for Disease Prevention and Control [CDC], 2007). Similar to African American culture, role constraints associated with cultural identification of family and gender roles are strong in Latino culture (Trickkett, 2005). The formation of self is integrated with self-identification of gender and family roles. A strong dislike for homosexuals as failed men in Latino culture (Perloff, 2001) is related to men’s failures in fulfilling their gender roles within family structures. Research indicates that strong family orientations are evidenced among all subgroups of Latinas. Within the 45 family structure, mothers play important roles in taking care of family members, which could be transformed into credible sources of HIV prevention within the family (Fernandez, 1995). Injection Drug Users Most injection drug users at risk of HIV infection are racial/ethnic minorities. As an underclass of the racial/ethnic minorities, injection drug users are marginalized even within their own communities. They do not participate in formal economic and social activities of the communities and are likely to survive within a criminal economy (Bowser, 1994). In addition, injection drug users develop unique social networks that may facilitate social and material exchanges of drug use, which is highly stigmatized in their own racial/ethnic communities (Watters, 1996). As a result, social networks of injection drug users may place more direct. important influence on individual HIV/AIDS risk behavior than racial/ethnic culture. Studies have investigated relationships between characteristics of the social networks of injection drug users and HIV/AIDS risk behavior. Structural characteristics of the social networks include network size, network density, and overlap between networks (Latkin et al., 1995; Neaigus et al., 1996). Cohesiveness of drug use networks may be influenced by drug type, local drug economies, population density, and wider drug trends. For example, heroin use networks tend to be more cohesive and interdependent due to daily transactions of the drug, typically taking place in private homes. Street drug scenes are more visible in large cities. Opiate injectors are more likely to report supportive relationships (Power, 1996). 46 Latkin, F orrnan, and Sherman (2003) have examined relationships between norms of the social networks of injection drug users and individual condom use. They found that condom use as a perceived norm (measured as friends talking about condoms, using condoms, or encouraging others’ use of condoms) are associated with having a larger financial support network or having a greater number of network members with whom one may talk about health matters. Perceived norms are strongly associated with condom use. Rhodes and Quirk’s study (1996) examined norms of drug use networks in light of HIV/AIDS infection. They found that HIV/AIDS risks are relatively less important concerns for regular drug users, compared with risks of overdose, vein damage, and addiction. In addition, drug users were more likely to assess their HIV/AIDS infection risks by their own or their partners’ injecting practices than by their own or their partners’ sexual practices. Everyday norms in regular drug users encourage safer drug use. Yet, unsafe sexual practice is a habitual behavior, not much different from the heterosexual population in general. Most cultural studies on male injection drug users at risk for HIV infection focus on drug use behavior. For example, cultural meanings of needle sharing connote brotherhood (Des Jarlais, Friedman, and Strug, 1986; Rushing, 1995). Very few cultural studies have focused on male injection drug users’ sexual behavior risks. In contrast. most studies on female injection drug users at risk for HIV/AIDS could be found in the literature of women selling sex for drugs, or money in order to buy drugs (e.g., Weissman and Brown, 1995). HIV/AIDS risk behavior varies not only by context, but also by gender. Studies found that there is a norm of condom use in the context of selling sex 47 due to awareness of HIV/AIDS. However, the same women are less likely to negotiate condom use with their primary partners due to an imbalance of power and traditional gender roles. Social network analysis and diffusion theory have suggested peer-oriented interventions which may incorporate indigenous leaders who have influence within the community and may exploit pre-existing channels of communication within the community (Rhodes and Quirk, 1996), such as shooting galleries (Rushing, 1995). Therefore, identifying key features in the structure of drug use networks may facilitate the development of intervention initiatives (Power, 1996). However, Rhodes, and Quirk (I996) cautioned that while there are functionary connections in drug use networks, there is a lack of strong social ties. They differentiated drug use networks from sexual networks. Drug user networks mainly function to facilitate drug use. There may not be influential people or appropriate channels of communication in drug use networks along which ideas of safe sex can be disseminated. This finding is in line with Latkin, F onnan, and Sherman (2003), who found the particular functions of networks, such as sharing health information or providing financial support, are associated with perceived norms of condom use. Rhodes and Quirk (1996) suggested that the changes of sexual norms of injection drug users may come from outside of their networks. A change of sexual norms in the heterosexual population as a whole may be necessary. Further research investigating structures and norms of the sexual networks of injection drug users by drug type is imperative. 48 Men Having Sex with Men Many cultural studies have characterized the subculture of MSM as a sex positive culture in which sex is viewed as pleasurable and recreational, in contrast to a sex negative culture in which sex is repressed. In the former, having multiple partners is the norm, while in the latter, having single partners is the ideal. Sex institutions, such as bathhouses, porno theaters, massage parlors, and tearooms, provide access to instant and impersonal sex with multiple partners. The cultural significance of sex institutions is that they symbolize gay liberation from social oppression due to sexual orientation (Rushing, 1995). Gay men’s engagement in a sex positive culture enhances their pride as homosexuals and their self-identification with the gay communities, which they cannot find outside their communities (Rushing, 1995). A study found that compared with bisexual men, gay men reported more tolerant attitudes toward homosexuality. In addition, they were more likely to perceive same-sex activity as accepted by others and were less likely to have self-homophobia (Stokes, Vanable, and McKirnan, 1997). Due to a high level of the HIV/AIDS awareness, condom use has become a normative sexual practice within this group, though in recent years condom use is fading away in some segments of the community as a result of the availability of drug treatment for HIV/AIDS (DiClemente et al., 2002). However, stress associated with self-denial of sexual orientation (Zamboni and Crawford, 2007), along with other factors, such as a lack of perceived norms, affectionate feelings, wishing to please a partner, and spontaneity of unsafe sex (Kelly, et al., 1991), have been related to a lapse of unsafe sex behavior. This is particularly relevant to racial/ethnic minority MSM because they are discriminated 49 both by the gay communities, which are composed of primarily White men, and by their own racial/ethnic communities, which maintain strong feelings of homophobia (Zamboni and Crawford, 2007). That they are invisible in racial/ethnic communities and are less integrated with the gay community makes them a challenge for HIV/AIDS outreach intervention. Intervention studies targeting MSM at the individual level have suggested that enhancing individual gay pride and strengthening self-identification with the gay community will facilitate self-acceptance of safe sex norms and self-efficacy of safe sex practices. At the level of community initiatives, norm changes for HIV/AIDS prevention in the gay community are made readily possible by the strength of social ties, which was related to early success of HIV/AIDS prevention (Rhodes and Quirk, 1996). Without much support from governments, gay communities were able to mobilize resources to stop the further spread of HIV (Rushing. 1995). Similar to interventions targeting injection drug users, peer-orientated intervention, as introduced by social diffusion theory, has played an important role in the design of interventions toward men having sex with men. The incorporation of key leaders and existing credible communication channels are essential to diffuse intervention information. Methodology This study uses existing culturally sensitive HIV/AIDS prevention intervention studies published between 1996 and 2007. The year of 1995 is the threshold of eligibility because it was not until about 1993 that the public became more aware of the HIV/AIDS pandemic. After 1993. the idea of condom use as an effective HIV/AIDS preventive 50 behavior has been commonly adopted in HIV/AIDS prevention initiatives due to the availability of federal funding. The process from designing an intervention to publishing its findings usually takes a couple of years. Criteria of Eligi_bili_tv_ Because culture is contextually embedded, the scope of eligible cultural groups was limited to those in the US. Only studies conducted in the U. S. and reported in English were eligible. Eligible interventions must include components of HIV sexual risk intervention and must have either HIV sexual risk behavior (unprotected sex) or HIV sexual preventive behavior (condom use) as a behavioral measure. They must be designed as randomized controlled trials or quasi-experiments with pre-tests and post- tests. In addition, in culturally sensitive interventions, at least 70% of the subjects in each eligible study must fall into at least one of the following social categories: MSM, drug users, or a single racial/ethnic subpopulation. For studies of MSM or drug users, ethnicity was not a criterion for inclusion as subcultures of homosexual men and drug users are more directly related to targets of interventions. For the same reason, this study prioritizes the social categories of MSM and drug users over racial/ethnic subpopulations in classifying dominant cultural groups in culturally sensitive HIV/AIDS prevention interventions. However, other eligible studies must include a single racial/ethnic group for at least 70% of their samples. A pilot study of twenty explicitly culturally sensitive HIV/AIDS prevention intervention studies was conducted to develop culturally sensitive themes that are inclusive to cultures of major HIV/AIDS high-risk groups. HIV/AIDS intervention studies are eligible if they explicitly claimed cultural sensitivity in intervention contents or strategies, or cultural sensitivity was implicated in the intervention contents or strategies. Based on the literature review in Chapter 2 and the result of the pilot study, eligible intervention studies of culturally sensitive contents or strategies include one or a combination of the following domains (but are not limited to them): Programs that involve community participation Bemal, Bonilla, and Bellido (1995) have indicated the importance of context in culturally sensitive interventions with Hispanics. Community participation may make interventions culturally sensitive because of local variations of behavioral norms and local resources inhibiting or facilitating HIV/AIDS prevention. Community participation may include activities to draw out a program theme, including pilot testing interventions in the targeted population and/or using focus groups before or in the middle of intervention. Program themes that address self-pride in relation to their social identities Self-pride is particularly important to socially-oppressed individuals as their social worth is often denied. Self-pride regarding people’s social status is considered as resources that may buffer stress due to social discrimination (Williams, Spencer, and Jackson, 1999). Stress has been linked to risky sexual behavior among gay men (Martin, Pryce, and Leeper, 2005). People’s racial/ethnic status, gender status, and sexual preferences are often sources of social oppression. This dissertation includes interventions addressing ethnic pride, women’s pride, sexuality pride, or a combination. 52 Program themes that address gender roles in a specific culture Behavioral norms in a culture are often contingent on gender roles. Wilson and Miller (2003) have also identified gender roles as influential in HIV prevention initiatives. This dissertation includes HIV/AIDS prevention interventions addressing gender roles in a specific culture. Program themes that address social responsibilities Responsibilities include family and community responsibilities that are important for individual integration with their communities. Man'n (1993) has indicated that a culturally appropriate intervention is based on cultural values of the subpopulation. Racial/ethnic minority cultures (such as African American, Hispanic American, and Asian American cultures) have a higher level of emphasis on individual obligations to families and/or communities than White American culture. Interventions addressing social responsibilities are culturally sensitive to the racial/ethnic minorities. Programs that address life experience of participants A subculture may emerge out of people’s collective attempts to solve their shared problems that are often associated with people’s social positions and geographical locations (Sebald, 1984). Life experience covers the social environment and significant experience associated with social categories of participants, for example, experience with urban settings, poverty, and drug use. Programs that use at least one facilitator matching participants’ social categories or having cultural competency 53 Jezewski’s culture-brokering model (Jezewski and Sotnik, 2001) suggests that an effective culture broker needs knowledge, skills, sensitivity, and awareness of cross-cultural variables to solve conflicts and problems due to cultural differences. Program facilitators may function as cultural brokers between intervention researchers and participants. In addition, facilitators matching participants’ social categories may be considered more credible source of information. Therefore, this dissertation includes interventions that used facilitators or film presenters matching participants’ significant social memberships, such as drug users, ethnic minorities, or gender. Or the facilitators were trained for cultural competency with the cultural group of interest 0 Programs that use ethnic language or music Bemal, Bonilla, and Bellido (1995) suggest use of culturally ethnic language that has emotional appeal to Hispanics. Similarly, Man'n (1993) has indicated that strategies of culturally appropriate interventions must fit within the preferred behavioral repertoire of the targeted group. This dissertation includes interventions using Black English, Spanish, and/or ethnic music, such as Rap. It is common that an intervention was reported in separate studies in light of the different segments of program profile, data analysis, and study findings. This dissertation reviews relevant studies of an intervention in order to cover essential information of an intervention which was reported in separate studies. The studies providing the most essential information of the empirical outcome data on individual culturally sensitive interventions are listed as eligible studies. 54 Criteria of Exclusion Multiple criteria of exclusion are applied in order to exclude studies with heterogeneous topics and with inadequate experimental or quasi-experimental designs. Drug treatment/prevention interventions without the inclusion of a sexual risk component in the intervention logic were excluded as were HIV/AIDS sexual risk interventions without either unprotected sex or condom use as a behavioral measure. Interventions which focused on preventing other sexually transmitted diseases or pregnancy were also excluded because other diseases or pregnancy may be related to different sets of cultural beliefs and behavioral expectancy. HIV/AIDS structural interventions targeting the economic or social structure of a community were excluded. Peer outreach was widely used in interventions. This dissertation assumes that interventions having peers as recruiters are not culturally sensitive, unless the peers also functioned as intervention facilitators. Peer facilitators may function as role models in the culturally sensitive interventions. In addition, interventions targeting HIV-positive patients, severely mentally ill patients, or inmates are excluded as this study assumes that the rationale underlying their targeted HIV/AIDS prevention interventions are different from the CSHAPIs examined in this dissertation. Freedom of these participants’ behaviors is often limited. Interventions reporting behavioral indexes combining measures of sexual communication, sexual behavior, and/or drug user behavior are not eligible as sexual communication and drug use behavior are not direct measures of sex behavior that may transmit HIV. This dissertation has a focus on risky sex behavior, instead of unsafe drug use that may also transmit HIV. Health beliefs and behavioral norms of drug use are different from those of sex behavior. In addition, interventions 55 with a design of one post-test group or one pre-test and post-test group without comparison groups are not eligible. Furthermore, studies reporting meta-analysis of several interventions are excluded because the studies did not provide sufficient information of each individual intervention. Studies reporting intervention profile, cost, or process without providing empirical data of sexual behavior are also excluded. Data Collection The first stage of data collection involved reviewing titles and abstracts of intervention studies to determine their potential eligibility. Databases and research registers were searched via a combination of the keywords of “HIV or AIDS, ” “prevention or intervention or program or training or education or clinical trial or randomized controlled trial,” and “behavior or condom. ” A list of 5889 HIV/AIDS prevention intervention studies published between 1996 and 2007 was obtained via electronic search of major databases], manual search of 32 major journals 2, manual search of the reference sections of eligible studies on HIV/AIDS prevention intervention, I IBSS, JSTOR, Medline, Proquest-Dissertation Theses, PsyARTICLES, , PsycINFO, Psychology, Social Sciences Abstracts, Social Sciences Citation Index, Social Services Abstracts, Social Work Abstracts, and Sociological Abstract 2 AIDS, AIDS and Behavior, AIDS Care, AIDS Education and Prevention, American Journal of Community Psychology, American Journal of Drug and Alcohol Abuse, American Journal of Epidemiology. American Journal of Health Behavior, American Journal of Health Promotion, American Journal of Preventive Medicine, American Journal of Public Health, Archives of Pediatrics and Adolescent Medicine. Contemporary Sociologr, Health Education and Behavior, Health Education Research, Health Psychology. International Journal of Health Education, Journal of A cquired Immune Deficiency Syndromes, Journal of Adolescent Health, Journal of American Medical Association, Journal of Behavioral Medicine, Journal of Community Health, Journal of Consulting and Clinical Psychology, Journal of Drug Issues, Journal of Health and Social Behavior Journal of School Health, Journal of Women ’s Health, Public Health Reports, Sexually Transmitted Diseases, Social Science and Medicine, and Women and Health 56 and The Cochrane Collaboration Register3. Screening the 5889 studies based on the above criteria yielded a list of 291 potentially eligible studies. A second stage of data collection involved reviewing abstracts and intervention descriptions of the potentially eligible 291 intervention studies based on the above criteria to determine their eligibility. The potentially eligible studies were screened in the order of cultural sensitivity in HIV prevention interventions, intervention outcomes, study designs, and intervention subjects. It resulted in a list of 50 studies with different culturally sensitive HIV/AIDS prevention interventions. The following figure is a flowchart of data screening. The Cochrane Collaboration Register 13 located at hum/clmucaltrralsgov. 57 Figure 3.1 Flow Chart of Data Screening 5889 studies from database, major journals. register 291 Potentially eligible studies T (289 journal articles, 2 books) ap—-—---————------a—-—---—-_---——--—-——q 9O Non-culturally sensitive HIV/AIDS- related intervention studies 51 intervention profiles, cost or process 12 culturally sensitive drug use or STD interventions, or structural interventions 3 studies with duplicate interventions 2 meta-analysis studies -------------------------------------- 133 Culturally sensitive HIV/AIDS-related I I I I I I I I O I --------- p: studies, or commentaries I l I I I I I intervention studies : r ----------------- >5 31 CSHAPI studies with outcomes as l I attitude, self-efficacy, intention, sexual l .. . 102 CSHAPI l communication, HIV-test, STD rate et al. 1 studies wrth outcomes as unprotected sex 1 _____________________________________ ; (UPS) and/or condom use i ----------------- >2 3 CSHAPI studies with one-group design 3 99 CSHAPI studies with UPS or condom use as outcome in randomized controlled trial or quasi-experimentpretest-posttestdesign r""“"""'""""“““"" ..... _, g 9 CSHAPI studies on HIV-positive patients I k _________________ 4 1 CSHAPI study on mentally ill patients 5 : 2 CSHAPI studies on inmates : 87 eligible CSHAPI studies 37 CSHAPI studies with mixed cultural groups 3 g (each cultural group <70%) : 50 studies with single cultural group: (%>=70) 14 drug user studies 5 MSM studies 18 African American studies 9 Hispanic American studies 4 White American studies 58 The entire set of the eligible studies was coded by the author on a coding sheet which had been developed based on the pilot study. It was constructed as an exhaustive inquiry to the existent culturally sensitive HIV/AIDS prevention interventions. Due to the heterogeneity of the eligible studies, semi-open coding was conducted to ensure inclusiveness of the characteristics of the eligible studies. Data on characteristics such as publication year, recruitment strategy, demographic characteristics of samples, intervention design, intervention content, intervention strategies, culturally sensitive content, culturally sensitive strategies, and intended intervention outcomes were gathered. Findings Demographic Cymteriptics The fifty studies included in this dissertation consisted of 26,500 study participants (19,088 in cohort studies and 7,412 in cross-sectional studies). Of these, 63.6% of the 22 intervention studies reporting the intervention completion years were completed before 1996 (k4=22). About 70 % of the intervention studies were funded by the National Institutes of Health or National Institute on Drug Abuse. The CSHAPI studies have been conducted in more than 10 states (k=44). Just over a quarter (27.2%) of the CSHAPI studies took place in California. Other centered states included Georgia (9%), Wisconsin (9%), and New York (6.8%). Not all the states where a large number of the CSHAPI studies were conducted had a large number of cumulative AIDS cases through 2006. New York had the largest number of cumulative AIDS cases, followed by California, Florida, Texas, New Jersey. and Illinois (CDC, 2008). In terms of the 4 . . . . . . k 15 the number of studies reporting the specrfic mfonnatron. 59 racial/ethnic status of the intervention subjects, over half of the CSHAPI studies focused on African Americans (consisting of more than 70% in each sample), followed by Hispanic Americans (18% of the CSHAPI studies) and White Americans (14 %), while Asians and Pacific Islanders were the least included subpopulation (2%) (k=50). In terms of cultural groups (defined as more than 70% of the participants in an intervention belonging to the subpopulations of MSM, drug users, or a single racial/ethnic subpopulation), African Americans were the subpopulation mostly used in the CSHAPIs as a major cultural group (36%), followed by drug users (28%), Hispanic Americans (18%), and MSM (8%), while White Americans were the least (8%) (k=50). As shown in Table 3.1, more CSHAPI studies targeted females (consisting of more than 70% in each sample) than males (39.6% vs. 27.1%, k=485). One-third (33.3 %) of the CSHAPI studies had mixed-gender samples, which consisted of more than 30 % of each gender. With a median age of 32.38 (k=48), a majority of the CSHAPI studies (75 %) focused on adults over the age of 20. Based on the aggregate of the reports on marital and socioeconomic status of the participants (Table 3.2), the targeted subpopulations of the CSHAPIs represented those who were single (never married, 55.7% or separated/divorced/widowed, 21.1%), poorly educated (10.93 years of education), and unemployed or employed part-time (80.8 %). An additional analysis shows that about one-tenth of the CSHAPI participants (k=9) were HIV positive. However, as showed in Table 3.1 and 3.2, the demographic characteristics of the CSHAPI participants were heterogeneous between the interventions targeting different social categories of subpopulations. Twenty-eight percent of the CSHAPIs were targeted 5 . . . . . . Two studies dld not report the gender composrtrons of the intervention samples. 60 toward adult drug users. They had an average of l 1.9 years of education (SD=0.88, k=4) and were mostly single (never married, 45.5%, SD=0.06, k=6; separated/divorced/widowed, 27.9%, SD=0.06, k=4). About seventy percent of the CSHAPI studies targeted African American as a cultural majority, while all other drug user studies had mixed racial/ethnic samples. No study targeted adolescent or youth drug users. Surprisingly, more CSHAPI studies focused primarily on female drug users than males (38.5% vs. 15.4%) while 46.2% of the interventions had mixed gender samples. Further analysis shows that over one-third (35.7 %) of the drug user studies targeted out- of-treatment drug users while 21.4% focused on in-treatment drug users. In addition, about one-tenth of the study participants were HIV positive (k=4). Seventeen percent (median) were homeless (k=7). Three of the five MSM studies (60%) targeted White Americans as the majority. Other racial/ethnic groups included African Americans (1 study) and Asian and Pacific Islanders (1 study). The socioeconomic characteristics of these samples represented the general population of MSM. Most of the White and Asian and Pacific Islander participants (more than 66%) had attended college or had a college degree or above. Over half of the African American participants (56%) reported an annual income of $15,000 or less. About 12 % of the participants in the two studies reporting HIV/AIDS prevalence rates were HIV positive. 61 Table 3.1 Demographic Characteristics of the Participants in Culturally Sensitive HIV/AIDS Prevention Interventions All Studies Drug MSM African Hispanic White (k=50) User Studies American American American Studies (k=5) Studies Studies Studies (k= 1 4) (k=1 8) (k=9) (k=4) Racial/ethnic majority6 White Americans7 7/50 0 3/5 0 0 4/4 14% 60% 100% African Americans 29/50 10/14 1/5 18/18 0 0 58% 71.400 20% 100% Hispanic 9/50 0 1/5 0 9/9 0 Americans 18% 20% 100% Asian and Pacific 1/50 0 O 0 0 0 Islanders 2% Mixed racial/ethnic 4/50 4 / l4 0 0 0 0 groups 8% 28.6% 9 . Agetmedian) 32.4 38.5 31 32.1 24.3 19.6 . . 10.1 1 Gender mgprfl Male 13148 2,1312 5/5 4118 2/9 0 27.1% ”4% 100% 22.2% 22.2% Female 19/48 5/13 0 9/18 5/9 0 39.6% 38.5% 50% 55.6% Mixed gender 16148 6/13 0 5/18 2/9 3/3'3 group 33.3% 46.2% 27.8% 22.2% [0000 Number and percentage of studies With their samples consrsting of more than 70% of a Single racial/ethnic group Number and percentage of studies with their samples consisting of more than 70% of White Americans Number and percentage of studies wrth multiple raCial/ethnic composrtion in the samples, each consrsting of less than 70% in the samples 9 Median of age was estimated based on aggregate data of available studies reporting mean age of the samples. 0 . . . . Number and percentage of studies With a gender group con3isting of more than 70% of the samples. Percentage was estimated based on available studies reporting gender composition in the samples. Two of the 50 studies did not report the gender compositions of the intervention samples. 12 . . . . . . One of the drug user studies did not report the gender composrtion of the intervention sample. 13 . . . . . . . . One of the White American studies did not report the gender composrtion of the intervention sample. 62 Table 3.2 Aggregate Estimation of the Socioeconomic and Marital Status of the Participants in Culturally Sensitive HIV/AIDS Prevention Interventions” Aill Studies Drug MSM African Hispanic White (k=50) User Studies American American American Studies (k=5) Studies Studies Studies (k= 1 4) (k=1 8) (k=9) (k=4) Employment Unemployed/ 80.8% 85.2% 56% 78.9% 78% - employed part-ti me kl 6: 1 7 k=9 k:] k: 5 k=2 (%)‘5 SD=0. ,22 SD=0.] SD=0. 132 SD=0. 085 Family income? 63.9% 55.9% 16% 73.2% 66.2% - $10,000/year (%) k=11 k=1 k=1 k=5 k=4 8010.221 SD=0. 11 SD=0.23 Education Years (mean) 10.9 I 1.9 13 l 1.3 9.4 - k=19 k=4 k=1 k=8 k=6 SD=1.87 SD=0.882 SD= 1.6 SD=2.001 High school or less 60.6% 68.6% 25.2% 67.8% 81.5% 0% ("/o) k=15 k=8 k=3 k=3 k=1 k =3 SD=0.244 SD=0. 149 SD=0.095 SD=0.277 Marital Status Single, never 55.7% 45.5% 88.0% 76.7% 44.3% - married (%) k= 13 k=6 k=1 k=3 k=3 SD: 0.224 SD=0.063 SD=0.204 SD=0.285 Married or steady 25.5% 21.1% - 8.1% 40.1% - relationships (9%) k: 1 7 k=8 k: 3 k=6 SD=0. 192 SD=0.1 SD=0.063 SD=0.237 Separated 21.1% 27.9% - 15% 12.3% - /divorced/widowed k=8 k=4 k: 3 k: 1 (”/01 SD=0. 117 SD=0. 059 SD=0. 154 - indicates that data were not obtained or are not reported. 4 . . . . Aggregate is estimated based on the number of available studies reporting raw information. In addition, studies may report more than one measure of employment, education, and/or marital status. Mean of the proportion reported by available studies. Number of the studies provides the information of unemployment or part-time employment. 63 Fifty percent of the African American studies predominantly involved women while 22.2% targeted males. Nearly 69% of these studies focused on adults. The participants tended to have low socioeconomic status (11.3 years of education) in the 8 African American studies that reported socioeconomic status. Over three-fourths (78.9%) of the study participants were unemployed or employed part-time (k=5). Thirteen percent of the participants were HIV positive (k=3), 45.7% ever engaged in selling sex for money or drugs (k=4), and 14.4% had or were having substance abuse problems (k=8). Just under 20% (1. 8.4 %) of the CSHAPI studies focused on Hispanic Americans with over half of the studies primarily targeting females (55.6%). About 55% of the Hispanic American studies were conducted with adults over the age of 20 (k=9). However, among all CSHAPI studies, the participants in the Hispanic American studies were young (24.3 vs. 32.4 of median age). About 40 % of the participants were married or had steady partners (k=6), compared with the participants in the other studies, who were mostly single. On average, the study participants had low socioeconomic status with 9.4 years of education (k=6) and family income less than $10,000 per year (66.2%, k=4). An additional analysis shows that a majority of them (74%, k=4) were born in foreign countries. Compared with other social categories, the participants in the White American studies were youngest (median age=19.7). Three of the 4 White American studies were conducted with mixed-gender heterosexual college samples. while 1 study targeted run away, homeless youth. 64 Study Designs Over 90% of the 44 CSHAPI studies reporting types of intervention incentives used money as an intervention incentive for the participants to complete the interventions and follow-up tests (mean=$54.10, mode=$60.00, SD=32.28). About half of the studies recruited the participants from the participants’ communities. The rest of the studies recruited the participants from specific locations frequented by the targeted subpopulations, such as health or STD clinics (20.4%), drug treatment centers (8.2%), or gay bars (4.1%) (k=49). The 50 studies had 120 intervention conditions, including 83 CSHAPIs, 18 non- CSHAPIs, 8 no—treatment groups, and 11 waiting-listed groups. Typically, these CSHAPI studies involved a comparison of a CSHAPI condition with a control/comparison condition or two CSHAPI conditions with a control/comparison condition. Among these CSHAPI studies, sampling bias was largely prevented by randomly assigning the participants to the interventions (68% at individual level, 24% at community/site/time block level, and 8% without any randomization assignment), instead of randomly selecting the participants from the targeted population (k=50). In addition, the number of randomized controlled trials (RCTs) was double that of quasi-experiments (66% vs. 34%). One-quarter of the quasi-experimental studies selected their participants based on matched comparison of communities. The common use of a RCT design in the CSHAPIs may generate study findings less confounded by factors others than the interventions. The 50 CSHAPI studies had been conducted at a variety of levels, meaning individual, group. community, or their combination, while a majority (74%) were 65 conducted at a group level. In addition, all except one study included local samples, involving a comparison of the participants within one community or a comparison of several communities. An overwhelming majority, 92%, used cohort data rather than cross-sectional data. Therefore, generalization of these CSHAPI studies is largely limited to individuals in similar locales at an aggregate level. Among the 50 studies, condom use was the most commonly used outcome measure (76%), followed by unprotected sex (56%) and number of sexual partners (54%). Previous research indicates that self-reported sex behavior is reliable within three months of the incurring behavior (Schroder, Carey, and Variable, 2003; Schrimshaw, Rosario, Meyer-Bahlburg, and ScharfMatlick, 2006). On average, these CSHAPI studies examined retrospective sex behavior prior to about 2.4 months from the time of any tests (median=3, SD=1.6, k=46). Therefore, the findings of the self-reported sex behavior in these studies are in general reliable. In addition, previous studies found that changes in sex behaviors after interventions differed by types of partners (for example, steady sex partners, casual sex partners, or sex workers) and types of sex behavior (such as oral sex or vaginal sex). Slightly over half of the CSHAPI studies (56%) focused on different types of sex, although some excluded certain types of sex in data analysis due to small sample sizes. In contrast, only about one-third of the studies (32%) examined sex behavior among different types of sex partners. Therefore, the findings of these CSHAPI studies may be. to some level, generalizable to general sex behavior, but very limited in application to specific types of sex behaviors and sex partners. 66 Furthermore, generalization of the CSHAPI studies is restricted in terms of the longevity of the intervention and change in behaviors. One-quarter of the studies had 3- month follow-ups as the last tests (k=48). The mean number of months of the last follow- up tests was 8.6 (median=6, SD=6.07). Few of the studies (12.5%) had their last follow- up tests conducted more than a year after the interventions were completed (k=48). Contents of Culturally Sensitive HIV/AIDS Prevention Interventions General Features Table 3.3 shows that a majority (86%) of the CSHAPI studies were based on theories while 38% used more than one theory. No single theory has dominated the CSHAPI studies in general while twenty-five different theories had been adopted in these studies. Social Cognitive Theory was found to be the most widely used theory (28% of the studies), followed by the Theory of Reasoned Action (22%), and the Information Motivation and Behavior Model (20%). Important theories used in the interventions by the participants’ cultural categories are heterogeneous. For example, 10 of the 14 drug user studies involved African Americans as a cultural majority (more than 70% of African Americans in the samples), while four of the drug user studies had mixed racial/ethnic groups in their samples (each consisting of less than 70% of the samples). In the former, Social Learning Theory and Theory of Reasoned Action were more widely used than other theories; in the latter, Transtheoretical Model of Change was most popular. Table 3.4 shows that the major intervention objectives in the C SHAPI studies focused on factors at the individual level, although 74% of the CSHAPI studies were 67 conducted at a group level. Group-level interventions typically involved a small group of intervention participants in each intervention session. The most common intervention objective of the CSHAPI studies was improving behavioral skills (74%) through skills demonstration of condom use and/or sexual communication. Enhancement of intrapersonal competence, such as behavioral self-management, coping enhancement, individual empowerment, and self-esteem, was the second most common intervention objective (50%). Enhancing awareness of the importance of interpersonal skills (46%), such as social skills and sexual communication, was also found in a number of the studies. The difference between the objectives of interpersonal skills and behavioral skills training is that the latter provided demonstration of specific behavioral skills while a few studies also offered opportunities for behavioral practice. The most common culturally sensitive feature in the CSHAPI studies (76%) was matching facilitators or film actors With the participants’ ethnic status or with their significant experience, such as drUg use, student, or sexual orientation. An important function of matching the facilitators/actors was role modeling of the intended intervention behavioral outcomes. The second most common culturally sensitive feature (62% of the studies) was community participation, which involved local subpopulations in designing the interventions through conducting formation studies or pilot testing the interventions. Culturally specific language or music was adopted by 30% of the studies. Among the culturally sensitive indicators, participants’ social identity as related to their self-pride of ethnic, gender, and /or sex orientation status, received the least attention in general (16% of the studies). 68 Table 3.3 Theories Used in Culturally Sensitive HIV/AIDS Prevention Interventions All DU MSM AA HA WA k=50 Ic=l4 k=5 k=18 k=9 k=4 AA 1 7 %ixerl WP %inority k=10 k=3 k=4 k=2 N umber of theory- 43 7 4 3 2 l8 7 2 b ased intervention 86% 70% I 00% 100% 100% 100% 77. 8% 5 0% studies21 Social Cognitive l4 2 0 0 I 5 5 l TheorL 28% 20% 50% 27.8% 55. 6% 2500 S ocial Learning 5 3 0 0 0 2 0 0 Theory 10% 30% 11. 1% T heory of l l 3 l O 1 2 4 0 _R easoned Action 22% 30% 25% 50% 1 1. 1% 44. 4% Theory of Planned 6 l O 0 O 2 2 I Qhavior 12% 10% 11. 1% 22. 2% 25% l hformation, 10 0 l 0 0 7 l 1 Motivation, and 20% 25% 38. 9% 11. 1% 25 % ghavior Model H ealth Belief 5 2 l 0 1 0 l 0 fiflodel 10% 2000 25% 50% 11.1% T he 4 2 2 O 0 0 0 0 Transtheoretical 8% 20% 5 0% _flodel of Change A IDS Risk 2 0 1 0 1 0 0 0 R eduction Model 4% 25 % 5000 S ocial Diffusion 2 0 0 l 0 0 l 0 x 4% 33. 3% I I. 1% S exological l 0 O l 0 0 0 0 %thodology 2% 33.3% Sexual Health 1 0 0 0 0 l 0 0 2% 5.6% Protection 2 0 l 0 0 l o o %tivation Model 4% 25% 5.6% G e rider and Power 4 I 0 o o 2 l 0 x 8% 10% 11.1% 11.1% E- mpowerment 2 l o 0 0 0 l 0 \ 4% 10% 11. 1% C3t her theories 1 l 4 l l 0 2 3 0 \ 22% 40% 25% 33.3% 11.1% 33.300 \ l7 l8 rug user studies involving more than 70% of African Americans in their samples. Drug user studies involving mixed racial/ethnic groups in their samples, each consisting of less than 70 0 1/ {3 0 f‘ the samples. 20 White American MSM studies. 2‘ 1r‘l.::luding one African American MSM study and one Asian American MSM study. N umber of studies involving use of any theory. Some studies were based on more than one theory. 69 Table 3.4 Description of Culturally Sensitive HIV/AIDS Prevention Interventions All DU MSM AA HA WA k=50 k= 1 4 k=5 k= 1 8 k=9 k=4 AA Mixed WA Minority k=10 k=4 k=3 k=2 Intervention objectives Perceived 9 2 l O 0 3 l 2 Susceptibility 18% 20% 25 ”o 16. 7% 1 1. 1% 50% Behavioral 21 2 2 O l l 1 5 O motivation 4200 20% 50% 50% 61.1% 55. 6% Interpersonal 23 3 2 l 2 l 1 3 1 skills 46% 30% 50% 33.3% 100% 61.100 33.3% 25% lntrapersonal 25 8 l l 2 l 0 3 O competence 50% 80% 25% 33.300 100% 55.6% 33.3% Interpersonal 3 0 0 2 0 l O 0 growth 6% 66. 7% 5. 6% Social support/ 19 5 3 3 2 3 1 2 community 38% 50% 75 % 100% 100% 16. 7% 11. 1% 50% norms Behavioral 37 8 2 l l l 5 7 3 skill training 74% 80% 5000 33.3% 50% 83.3% 77.8% 75% Intervention level Individual 5 l O 0 0 l l 2 10% 1000 5.6% 11. [”0 50% Group 37 7 2 l 2 17 8 0 74% 70% 50% 33. 3% 100% 94. 4 00 88. 9% Individual and 3 l l 0 0 0 0 1 group levels 6% 10% 25% _ 25 % Community 5 l 1 2 0 O 0 l 10% 10% 25% 66. 7% 25 % Total exposure [0 46322 672 520 1080 360 348 63024 14425 intervention by (531) (541) :1 (255) (258) . (379) (352) (187) m n tes SD = = = : __|__U_I_l k23=35 . 6 k 3 2 k 15 [(25 k=3 22 Median=360 23 . . . . . Number of studies prov1ding infomiation of total exposure 70 Drug User Studies Studies T meting Africgn Americans Only Table 3.3 shows that 70 percent of the 10 studies were based on theories, with Social Learning Theory (30%) and the Theory of Reasoned Action (3 0%) being the most commonly used. No study on African American drug users included the Information, Motivation and Behavior Model, which was used by 20% of the CSHAPI studies. Table 3.4 shows that intervention objectives of the studies on African American drug users were similar to the CSHAPI studies in general, focusing on intrapersonal competence (80%) and behavioral skill training (80%). Further analysis shows that ninety percent of the studies explicitly claimed cultural sensitivity of the interventions with African American culture, although 20% did not report the detail of the cultural sensitive intervention contents. The single study which did not explicitly claim cultural sensitivity focused on the life experience of drug users. As shown in Table 3.5, among the 50 CSHAPI studies, those targeting African American drug users had more emphasis on the salient social experience of the targeted subpopulation, such as experience in urban settings, of poverty, drug use, and/or selling sex (60% vs. 25%). Inclusion of peer facilitators/actors was common (60%). Other culturally sensitive features were evenly used in the studies: community participation (40%), gender norms (40%), social roles (40%). except for social identity (no studies) and culturally specific language/music (10%). 24 Median=480 25 Median=42 7I Table 3.5 Culturally Sensitive Indicators by Subcultural Groups All DU MSM AA HA WA k=50 k=l4 k=5 k=18 =9 k=4 ' AA Mixed WA Minority k=10 k=4 k=3 k=2 Community 31 4 2 1 2 12 7 1 participation 62% 40% 50% 33.3% 100% 66. 7% 77. 8% 25 % Formative 26 4 2 1 O 12 6 1 research 52% 40% 50% 33.3 % 66. 7% 66. 7% 25% Pilot test 15 2 O O 2 7 4 O 30% 20% 100% 38. 9% 44. 4% Social identig 8 O O l 2 6 0 0 16% 33.3% 100% 33.3% Ethnic pride 8 0 0 0 2 6 0 0 16% 100% 33.3% Women pride 2 0 O 0 0 2 0 0 4% 11. 1% Pride of sexual 3 O 0 I 2 0 0 0 orientation 6% 33.3% 100% Gender norms 12 4 l O 0 5 2 O 24% 40% 25 % 27.8% 22.2% Social roles 10 4 0 0 O 3 3 0 20% 40% 16. 7% 33.3% Responsibility of 8 3 0 0 O 2 3 0 family 16% 30% 11. 1% 33.3% Responsibility of 5 2 0 0 0 2 l 0 community 10% 20% 11.1% 11.1% Other social 2 2 O O 0 0 0 0 responsibility 4% 20% Social experience 14 6 3 O 0 2 3 O 28% 60% 75% 11. 1% 33.3% Urban settings 6 2 0 0 O 2 2 0 12% 20% 11. 1% 22.2% Poverty 2 1 0 O 0 0 1 0 4% 10% 11. 1% Drug use 1 1 6 3 O 0 0 2 0 22% 60% 75 % 22.2% Selling sex 1 l 0 0 0 0 0 0 2% 10% Matching 38 6 3 3 2 l6 4 4 fpcil itators/pctors 76% 60% 75 % 100% 100% 88.9% 44. 4% 100% Facilitators 31 5 2 3 2 l3 4 2 2% 50% 50% 100% 100% 72.2% 44.4% 50% Film/story actors l6 2 3 0 0 8 0 3 ‘ 32% 20% 75% 44.4% 75% Culturally specific 15 l 0 0 1 4 8 l language/musip 30% 10% 5 0% 22. 2% 88.9% 25 % Language 12 1 0 0 l 2 8 0 24% 10% 50% 11. 1% 88.9% Music 5 O 0 0 O 3 0 1 10% 16.7% 25 % 72 Studies Targeting Mixed Racial/ethnic Grow Theory and intervention objectives in the mixed-group drug user studies were different from the CSHAPI studies in general. Two of the four mixed-group studies were based on The Transtheoretical Model of Change. Enhancing social support and changing behavioral norms (or perceived norms) was the most common intervention objective (75% of the 4 studies). lntrapersonal competence was less focused in the mixed-group studies than African American drug user studies (25% vs. 80%). Further analysis shows that in contrast to the studies targeting African American drug users (90%), only 1 of the 4 mixed-group studies (25%) explicitly claimed that the interventions were culturally sensitive to different cultures of the participants. This study focused on the part of cultures which may have had an impact on gender norms while it did not specify the gender norms in each culture according to which the interventions had been tailored. In addition, 75% of the 4 mixed—group studies reported culturally sensitive contents without mentioning the term “culture” in the report. Table 3.5 shows that in the mixed-group drug user studies, cultural sensitivity was less likely to involve ethnic culture and focused more on life experience pertaining to drug use (75% of the 4 studies). Matching peer facilitators or actors was common (75%). Half of the 4 studies had involved local participants in the design of the interventions. In contrast to the 10 African American drug user studies, gender norms (25% vs. 40%) and social roles (0% vs. 40%) were less likely to be included in the 4 mixed-group studies. Men Having Sex with Men Studies Sixty percent of the MSM studies targeted White American MSM while 20% involved African American MSM, and 20% targeted Asian American MSM. All of the 73 CSHAPI studies targeting MSM were guided by specific theories which however, were more heterogeneous compared with other CSHAPI studies. The theories included the AIDS Risk Reduction Model (in the African American MSM study), Health Belief Model, Theory of Reasoned Action, Social Cognitive Theory (in the Asian and Pacific Islander MSM study). Social diffusion, Sexological Methodology, Homosexual Identity Formation, and Social Network Approach (in the White American MSM studies). Regardless of the racial/ethnic status of the participants, the MSM studies had distinct intervention objectives and culturally sensitive features from those of other CSHAPI studies. All of the MSM studies had an emphasis on improving social support and/or changing behavioral norms or perceived norms, while only 38% of all CSHAPI studies had a similar focus. In terms of culturally sensitive indicators, the MSM studies were more likely to match peer facilitators (100% vs. 62% of all CSHAPI studies) and had a focus on social identity (60% vs. 16%). Furthermore, in comparison to the CSHAPI studies in general, the MSM studies had a lack of focus on gender norms (0% vs. 24%), social roles (0% vs. 20%), and social experience, such as poverty, drug use, and experience in urban settings (0% vs. 28%). On the other hand, the MSM studies targeting racial/ethnic minorities appeared to be more similar to each other in terms of intervention objectives, compared with the White American MSM studies. In general, the White American MSM studies had more focus on the participants’ involvement in their gay communities while other MSM studies emphasized individuals’ self-identity to their racial/ethnic and sexual orientation status. For example, both of the Asian American and African American MSM studies aimed at changing HIV/AIDS risk behavior through improving the participants’ interpersonal 74 skills and intrapersonal competence. Individual empowerment of the racial/ethnic minority MSM was achieved through promoting self-pride in dual minority status being both racial/ethnic and sexual orientation minorities. Only 33.3% of the White American MSM studies had an emphasis on social identity. In contrast, none of the MSM studies for racial/ethnic minorities included interpersonal growth, such as gay community empowerment, in the intervention objectives, while 66.7% of the White American MSM studies had such a focus. Community empowerment was largely achieved through involvement of the community members in the process of intervention (33.3%) or through identifying the formation of individuals as homosexual in a group setting (33.3%). African American Studies Thirty-six percent of the 50 CSHAPI studies involved more than 70% of African Americans in the samples as a cultural majority. The prevailing theories, intervention objectives, and culturally sensitive indicators in the African American studies were similar to the CSHAPI studies in general. All of the African American studies were founded in specific theories. The Information, Motivation and Behavior Model was the most popular (38.9%), followed by the Social Cognitive Theory (27.8%). In terms of intervention objectives, an overall emphasis on behavioral skills training was found among 83.3% of the 18 African American studies. About 60% of the African American studies also included enhancement of behavioral motivation and/or interpersonal skills as intervention objectives. There was little emphasis on interpersonal growth (5.6%) and social support/behavioral norms (16.7%) in the African American studies. 75 With respect to culturally sensitive indicators, a variety of culturally sensitive features were included in the African American studies. Matching facilitators/actors was the most popular culturally sensitive feature in the African American studies (88.9%), followed by community participation (66.7%). Other culturally sensitive features were, however, incorporated by less than 40% of the studies. Hispanic American Studies All of the CSHAPI studies targeting Hispanic Americans were based on behavioral theories which were more homogeneous in comparison with other CSHAPI studies. Over half of the 9 Hispanic American studies were based on the Social Cognitive Theory while 44.4% of the Hispanic American studies used a combination of the Theory of Reasoned Action and the Social Cognitive Theory. In general, the Hispanic American studies were similar to the African American studies in terms of intervention objectives. The Hispanic American studies had an overall focus on behavioral skills training (83.3%) and behavioral motivation (55.6%). Few of the Hispanic American studies focused on social support/behavioral norms (11.1%) while no study promoted interpersonal growth. Similar to the African American studies, a variety of culturally sensitive features were used. Close to 80% of the Hispanic American studies involved community participation in designing the interventions. However, the Hispanic American studies were somewhat different from the African American studies in their emphasis on culturally specific language (88.9%) rather than on matching the facilitators with the participants’ ethnic status (44.4%). Furthermore, in comparison to the African American studies, the Hispanic American studies have more focus on social roles (33.3% vs. 16.7%) 76 and social experience (33.3% vs. 1 1.1%), but had a lack of focus on social identity (0% vs. 33.3%). White American Studies The White American studies were the least theoretically grounded among all of the CSHAPI studies (50% vs. 86%). The theories used in these studies were similar to the studies targeting African Americans and Hispanic Americans, such as the Social Cognitive Theory (25%), the Theory of Planned Behavior (25%), and the Information, Motivation and Behavior Model (25%). However, the White American studies were more likely to be conducted at an individual level (50% vs. 5.6% of the African American studies and 11.1% of the Hispanic American studies). In addition, the White American studies targeted the subpopulation of either college students or run-away youth. The most common intervention objective in the White American studies was behavioral skills training (75%), followed by improving social support/behavioral norms (50%) and enhancing perceived susceptibility of the participants (50%), and interpersonal skills (25%). Inter-personal growth and intra- personal competence were not included as the intervention objectives. In terms of culturally sensitive features, the White American studies incorporated the least features in the interventions among all of the CSHAPI studies. The most common culturally sensitive feature was matching peer facilitators/actors. The other culturally sensitive features were community participation (25%) and culturally specific language/music (25%). 77 Conclusion This study found that African Americans were the subpopulations most involved in the CSHAPI studies as either a cultural majority or a racial/ethnic majority. In addition, the CSHAPIs were more likely to target females, except for MSM studies. Socioeconomic and marital statuses of the samples were, however, heterogeneous. For example, White American studies and MSM studies were more likely to involve participants with a higher level of education. The participants in Hispanic American studies were less likely to be single or in separated/divorced/widowed situation than the other studies. This study found very few CSHAPI studies targeting adolescents. Men had not received adequate intervention attention as a gender majority in the intervention samples (15% of the drug user studies and 22.2% of either African American studies or Hispanic American studies). Adolescents and male drug users, however, represented those at high risk for HIV infection. Adolescents are at a stage of sexual exploration and are less skillful in sexual communication and in condom use. In addition, as reviewed earlier, safe sex, in general, is a relatively less important concern for regular drug users than is safe drug use. Further CSHAPI studies should be expanded to include these two subpopulations. In addition, the CSHAPI studies tend to have an individual focus on skills of condom use and intrapersonal competency, although 74% of the 50 studies were conducted at a group level. Eighty-six percent of the studies were based on at least one behavioral theory. Therefore, cultural sensitivity in the CSHAPI studies largely played a role in facilitating the variables in behavioral theories. 78 Similarly, this study found that while over 50% of the studies included involving the participants in formative research or matching facilitators with the participants’ status as culturally sensitive indicators, the indicators between studies targeting different subcultures were diverse. For example, matching facilitators was not as important as matching message and conducting formative research in the Hispanic American studies, as 74% of the Hispanic American participants were born in foreign countries (k=4), not fluent in English, and still maintained their culture of origin. In addition, racial/ethnic MSM studies (100%) and some African American studies (38.9%) include ethnic pride as a culturally sensitive indicator, while the same focus was lacking in drug user studies, Hispanic American studies, and White American studies. Ethnic pride may be an issue more for a racial/ethnic minority which has existed for a long time and has developed various social relationships with other racial/ethnic groups. Thus, they are more likely to collectively accumulate experience related to their racial/ethnic disadvantages in history and internalize the experience at an individual level. In a mixed-racial/ethnic context, social comparison may function to strengthen the importance of ethnic pride. Therefore, in comparison with African Americans, ethnic pride may be a less important issue for Hispanic Americans who were less integrated with the society. Furthermore, while none of the African American drug user studies included ethnic pride, drug user studies tend to include social experiences as a culturally sensitive indicator (64.3% vs. 28% in all studies). Therefore, the CSHAPI assumed that drug use experience is more important than the experience related to racial/ethnic status. This corresponds to W atters’ study (1996) that racial/ethnic minority drug users were 79 discriminated in their own racial/ethnic community. The drug use networks that racial/ethnic minorities frequented may represent a close community in which subcultural norms and economic activities were both sustainable. Therefore, for racial/ethnic minority drug users, ethnic pride is a less important issue than their social and economic relationships with other drug users However, what remains unclear are the characteristics of specific drug user networks with which the CSHAPI participants were involved. As mentioned earlier, the structure of drug user networks varies by network size, network density, and overlap between networks. Cohesiveness of the drug user networks varies by drug types. To what extent the structure of the drug user network conditions drug users’ safe sex behavior has not been examined in the drug user studies. All of the drug user studies did not provide the information of the social positions of drug users within the hierarchy of drug use communities for which they had memberships. Although drug users may represent those with low socioeconomic status in comparison with the general population, their status within the hierarchy of the drug user networks may impact their social relationships, including sex relationships. Development of the indicators of stratification within drug user networks may be the first step in facilitating an understanding between drug users’ social status within the drug user networks and their safe sex behavior. In general, White American MSM studies were more similar to White American studies, than mixed-raciaI/ethnic MSM studies, in that they incorporated relatively less culturally sensitive indicators than other studies, while they also had a general focus on community participation and matching facilitators. One White American MSM study included pride of sexual orientation as a culturally sensitive indicator. Another White 80 American study used culturally matched music. Because C SHAPIs targeting White Americans were partly based on the mainstream culture, the culturally sensitive indicators which were developed most for the subcultural groups socially oppressed by the society are not readily applicable. On the other hand, White American adolescents may represent the group of White Americans which are socially oppressed. However, as the C SHAPIs in general did not focus on adolescents, an examination of the impact of cultural sensitivity on the CSHAPI’s effectiveness targeting socially oppressed White Americans is limited by a lack of empirical studies. Similarly, there is a lack of CSHAPIs targeting White American drug users as a cultural majority in the samples. African American drug users were the only racial/ethnic majority in the drug user studies. Further research on socially oppressed White Americans other than MSM may aid in an understanding of the interplay of racial/ethnic majority status and social oppression in shaping people’s safe sex behavior. White American adolescents and White drug users may have lower socioeconomic status than White American MSM while they are not socially oppressed for their sexual orientation. Therefore, operationalizations of cultural sensitivity for White American MSM may not be applicable. 81 References *References marked with an asterisk indicate the CSHAPI studies included in the review Anderson, W. (1993). The New York needle trial: the politics of public health in the age of AIDS. In V. Berridge & P. Strong (Eds), AIDS and contemporary history (pp. 157-181). Cambridge, UK: Cambridge University Press. *Avants, S. K., Margolin. A., Usubiaga, M. H., & Doebrick, C. (2004). Targeting HIV- Related outcomes with intravenous drug users maintained on methadone: A randomized clinical trial of a harm reduction group therapy. Journal of Substance Abuse Treatment, 26, 67— 78. Beatty, L., Wheeler, D., & Gaiter, J. (2004). HIV prevention research for African Americans: current and future directions. Journal of Black Psychology, 30(1), 40- 58. Bemal, G., Bonilla, J., & Bellido, C. (1995). Ecological validity and cultural sensitivity for outcome research: Issues for the cultural adaptation and development of psychosocial treatments with Hispanics. Journal of A bnormal Child Psychology, 23(1), 67-82. Bowser, B. P. (1994). HIV prevention and Afiican Americans: a difference of class. In J. P. Van Vugt (Ed.), AIDS Prevention and Services: Community Based Research (pp. 93-108). CT: Westport: Bergin & Garvey.*Branson, B. M., Peterman, T. A., Cannon, R. O., et al. (1998). Group counseling to prevent STD and HIV: a randomized controlled trial. Sexual Transmitted Diseases, 25, 553-560. Card, J. J ., Benner, T., Shields, J. P, & Feinstein, N. (2001). The HIV/AIDS prevention program archive (HAPPA): a collection of promising prevention programs in a box. AIDS Education and Prevention, 13(1), 1-28. *Carey, M. P., Braaten. L. S.. Maisto, S. A., Gleason, J. R., Forsyth, A. D., Durant, L. E., & Jaworski, B. C. (2000). Using information, motivational enhancement, and skills training to reduce the risk of HIV infection for low-income urban women: a second randomized clinical trial. Health Psychology, 19, 3-11. *Carey, M. P., Maisto, S. A., Kalichman, S. C., Forsyth, A. D., Wright, E. M., & Johnson. B. T. (1997). Enhancing motivation to reduce the risk of HIV infection for economically disadvantaged urban women. Journal of Consulting and Clinical Psychology, 65(4), 53 1-541. Center for Disease Prevention and Control. (2007). Reports of cases of HIV/AIDS, AIDS, or HIV Infection: Table 11. Retrieved May 15’, 2008 from the World Wide Web: 82 littp://www.cdc. gov/hiv/topics/survci 1lance/resources/rcports/2005report/49ml table I 1.pdf on 12/16/2007. *Choi, K.-H., Lew, S., Vittinghoff, E., Catania, J. A., Barrett, D. C., & Coates, T. J. (1996). The efficacy of brief group counseling in HIV risk reduction among homosexual Asian» and Pacific Islander men. AIDS, 10, 81-87. *Collins, C., Kohler, C., DiClemente, R., & Wang, M. Q. (1999). Evaluation of the exposure effects of a theory-based street outreach HIV intervention on African- American drug users. Evaluation and Program Planning, 22(3), 279-293. *Cottler, L. B., Compton, W. M., Ben Abdallah, A., Cunningham-Williams, R., Abram, F ., F ichtenbaum, Cx., & Dotson, W. (1998). Peer-delivered interventions reduce HIV risk behaviors among out-of-treatment drug abusers. Public Health Reports, 113(Supplement 1), 331-341. *DeLamater, J ., Wagstaff, D. A., & Havens, K. K. (2000). The impact of a culturally appropriate STD/AIDS education intervention on Black male adolescents' sexual and condom use behavior. Health Education and Behavior, 27(4), 454-470. Des Jarlais, D. C., Friedman, S. R., & Strug, D. (1986). AIDS and needle sharing within the IV-drug use subculture. In D. A. Feldman & T. M. Johnson (Eds), The social dimensions of AIDS: Method and theory (pp. 11 1-125). NY: New York: Praeger Publishers. Des Jarlais, D. C., & Semaan, S. (2002). HIV prevention research: Culmulative knowledge or accumulating studies? Journal of Acquired Immune Deficiency Syndromes, 30(Suppl. I), 31-87. DiClemente, R. J ., Crosby, R. A., & Wingood, G. M. (2005). Community HIV prevention interventions: theoretical and methodological considerations. In E. J. Trickett & W. Pequegnat (Eds), Community interventions and AIDS (pp. 222- 248). New York: New York: Oxford University Press. DiClemente, R. J ., Funkhouser, E., Wingood, G. M., Fawal, H., Holmberg, S. D., & Verrnund, S. H. (2002). Protease Inhibitor combination therapy and decreased condom use among gay men. Southern Medical Journal, 95(4), 421-425. *DiClemente, R. J ., Wingood, G. M., Harrington, K. F., Lang, D., 1., Davies, S. L., Hook, E. W., et al. (2004). Efficacy of an HIV prevention intervention for African American adolescent girls: a randomized controlled trial. Journal of the American Medical Association, 292, 1 71 -1 79. Fernandez, M. l. (1995). Latinas and AIDS. In A. O'Leary & L. S. Jemmott (Eds). Women at Risk: Issues in the Primary Prevention of AIDS (pp. 159-174). NY: New York: Plenum Press. 83 *Fisher, J ., Fisher, W., Misovich, S., KImble, D., & Malloy, T. (1996). Changing AIDS risk behavior: effects of an intervention emphasizing AIDS risk reduction information, motivation, and behavioral skills in a college student population. Health Psychology, 15, 1 14-123. *Flaskerud, J. H., Nyamathi, A. M., & Uman, G. C. (1997). Longitudinal effects of an HIV testing and counseling programme for low-income Latina women. Ethnicity and Health, 2(1/2), 89-103. *Gleghom, A. A., Clements, K. D., Marx, R., Vittinghoff, E., Lee-Chu, P., & Katz, M. (1997). The impact of intensive outreach on HIV prevention activities of homeless, runaway, and street youth in San Francisco: The AIDS Evaluation of Street Outreach Project (AESOP). AIDS and Behavior, 1(4), 261-271. *Harris, R. M., Barker, B. R., Scott, D. E., Hetherington, S. E., & Kavanagh, K. H. (1998). An intervention for changing high-risk HIV behaviors of African- American drug-dependent women. Research in Nursing and Health, 21, 239-250. *Harvey, S. M., Henderson, J. T., Thorbum, S., & Beckman, L. J ., Casillas, A., Mendez, L., Cervantes, R. (2004). A randomized study of a pregnancy and disease prevention intervention for Hispanic couples. Perspectives on Sexual and Reproductive Health, 36(4), 162—169. Herbst, J. H., Kay, L. S., Passin, W. F., Lyles, C. M., Crepaz, N., & Marin, B. V. f. t. H. A. P. R. S. P. T. (2007). A systematic review and meta-analysis of behavioral interventions to reduce HIV risk behaviors of Hispanics in the United States and Puerto Rico. AIDS Behavior, 11, 25-47. *Hershberger, S. L., Wood. M. M., & Fisher, D. G. (2003). A Cognitive-behavioral intervention to reduce HIV risk behaviors in crack and injection drug users. AIDS and Behavior, 7(3), 229-243. *Hoffman, H. J. A., Klein, H., Crosby, H., & Clark, D. C. (1999). Project Neighborhoods in action: an HIV related intervention project targeting drug abusers in Washington, DC. Journal of Urban Health: Bulletin of the New York Academy of Medicine. 76, 419-434. ' Icard, L. (1985). Black gay men and conflicting social identities: sexual orientation versus racial identity. Journal of Social Work and Human Sexuality, 4(1-2), 83-93. *Jamner, M. S., Wolitski, R. J ., & Corby, N. H. (1997). Impact of a longitudinal community HIV intervention targeting injecting drug users‘ stage of change for condom and bleach use. American Journal of Health Promotion, 12(1), 15-24. 84 Jarama, S. L., Belgrave, F. 2., Bradford, J ., Young, M., & Honnold, J. A. (2007). Family, cultural and gender role aspects in the context of HIV risk among African American women of unidentified HIV status: an exploratory qualitative study. AIDS Care, 19(3), 307-317. *Jemmott, J. B., 111. Jemmott, L. S., & Fong, G. T. (1998). Abstinence and safer sex HIV risk-reduction interventions for African American adolescents. Journal of the American Medical Association, 2 79(19), 1529-1536. *Jemmott, J. B., 111, Jemmott, L. S., & Fong, G. T. (1999). Reducing HIV risk-associated sexual behavior among African American adolescents: testing the generality of intervention effects. American Journal of Community Psychology, 27(2), 161-187. *Jemmott, L. S., Jemmott, J. B., 111, & O'Leary, A. (2007). Effects on sexual risk behavior and STD rate of brief HIV/STD prevention interventions for African American women in primary care settings. American Journal of Public Health, 97(6), 1034-1040. Jezewski, M.A. & Sotnik, P. (2001). Culture brokering: Providing culturally competent rehabilitation services to foreign-born persons. Center for International Rehabilitation Research Information and Exchange. CIRRIE Monograph Series. John Stone, Ed. Buffalo, NY: CIRRIE. Johnson, W. D., Holtgrave, D. R., McClellan, W. M., Flanders, W. D., Hill, A. N., & Goodman, M. (2005). HIV intervention research for men who have sex with men: a 7-year update. AIDS Education and Prevention, 1 7(6), 568-589. *Kalichman, S. C.. Cain, D., Weinhardt, L., Benotsch, E., Presser, K., Zweben, A., Bjodstrup, B., & Swain. G. R. (2005). Experimental components analysis of brief theory-based HIV/AIDS risk-reduction counseling for sexually transmitted infection patients. Health Psychology, 24(2), 198-208. *Kalichman, S. C ., & Cherry, C. (1999). Male polyurethane condoms do not enhance brief HIV-STD risk reduction interventions for heterosexually active men: results from a randomized test of concept. International Journal of S TD and AIDS, 10. 548-553. *Kalichman, S. C., Cherry, C.. & Browne-Sperling, F. (1999). Effectiveness of a video- based motivational skills-building HIV risk-reduction intervention for inner-city African American men. Journal of Consulting and Clinical Psychology, 67(6), 959-966. *Kalichman, S. C.. Rompa, D., & Coley, B. (1996). Experimental component analysis of a behavioral HIV-AIDS prevention intervention for inner-city women. Journal of Consulting and Clinical Psychology, 64, 687-693. 85 *Kalichman, S. C., Rompa, D., & Coley, B. (1997). Lack of positive outcomes from a cognitive-behavioral HIV and AIDS prevention intervention for inner-city men: lessons from a controlled pilot study. AIDS Education and Prevention, 9(4), 299- 313. *Kalichman, S. C., Williams, E., & Nachimson, D. (1999). Brief behavioral skills building intervention for female controlled methods of STD-HIV prevention: outcomes of a randomized clinical field trial. International Journal of S TD AIDS, 10(3), 174-181. *Kegeles, S. M., Hays, R. B., & Coates, T. J. (1996). The Mpowerment Project: a community-level HIV prevention intervention for young gay men. American Journal of Public Health, 86(8), 1129—1136. Kelly, J. A., Kalichman, S. C., Kauth, M. R., Kilgore, H. G., Hood, H. V., Campos, P. E., et al. (1991). Situational factors associated with AIDS risk behavior lapses and coping strategies used by gay men who successfully avoid lapses. American Journal of Public Health, 81(10), 1335-1338. *Kelly, J. A., Murphy, D. A., & Sikkema, K. J ., et a1. (1997). Randomized, controlled, community-level HIV-prevention for sexual-risk behaviour among homosexual men in the US. cities. Lancet, 350, 1500-1505. *Koniak-Griffin, D., Lesser, J ., Nyamathi, A., Uman, G., Stein, J. A., & Cumberland, W. G. (2003). Project CHARM: An HIV prevention program for adolescent mothers. Family and Community Health, 26(2), 94-107. Latkin, C. A., Forman, V., Knowlton, A., & Sherman, S. (2003). Norms, social networks, and HIV-related risk behaviors among urban disadvantaged drug users. Social Science and Medicine, 56, 465-476. Latkin, C. A., Mandell, W., Vlahov, D., Knowlton, A. R., Oziemkowska, M., & Celentano, D. D. (1995). Personal network characteristics as antecedents to needle-sharing and shooting gallery attendance. Social Networks, 1 7(3-4), 219- 228. *Latkin, C. A., Sherman, S., & Knowlton, A. (2003). HIV prevention among drug users: outcome of a network-oriented peer outreach intervention. Health Psychology, 22. 332-339. *Lindenberg, C. S., Solorzano, R. M., Bear, D., Strickland, O., Galvis, C., & Pittman, K. (2002). Reducing substance use and risky sexual behavior among young, low- income, Mexican-American women: comparison of two interventions. Applied Nursing Research. 16. 137-148. 86 Logan, T. K., Cole, J ., & Leukefeld, C. (2002). Women, sex, and HIV: social and contextual factors, meta-analysis of published interventions, and implications for practice and research. Psychological Bulletin, 128(6), 851-885. Lyles, C. M., Crepaz, N., Herbst, J. H., Kay, L. S., & Team, f. t. H. A. P. R. S. (2006). Evidence-based HIV behavioral prevention from the perspective of the CDC's HIV/AIDS prevention research synthesis team. AIDS Education and Prevention, 18(Supplement A), 21-31. Lyles, C. M., Kay, L. S., Crepaz, N., Herbst, J. H., Passin, W. F ., Kim, A. S., et al. (2007). Best-evidence interventions: findings from a systematic review of HIV behavioral interventions for US populations at high risk, 2000-2004. American Journal of Public Health, 97(1), 133-143. Marin, B. V., & Diaz, R. M. (2002). Collaborative HIV prevention research in minority communities program: a model for developing investigators of color. Public Health Reports, 11 7(May-June), 218-230. Marin, B. V., & Marin, G. (1992). Predictors of condom accessibility among Hispanics in San Francisco. American Journal of Public Health, 82, 592-595. Marin, G. (1993). Defining culturally appropriate community interventions: Hispanics as a case study. Journal of Community Psychology, 2 1 (April). 149-161. Martin, J. 1., Pryce, J. G., & Leeper, J. D. (2005). Avoidance coping and HIV risk behavior among gay men. Health and Social Work, 30(3), 193-201. *McCoy, C. B., Weatherby, N. L., & Metseh, L. R., et al. (1996). Effectiveness of HIV interventions among cracker users. Drugs Society. 9(137-154). *McCoy, H. V., McCoy, C., & Lai, S. (1998). Effectiveness of HIV interventions among women drug users. Women and Health, 27(1/2), 49-66. *McMahon, R. C., Malow, R. M., Jenning, T. E.,& Gomez, C. J. (2001). Effects of a cognitive-behavioral HIV prevention intervention among HIV negative male substance abusers in VA residential treatment. AIDS Education and Prevention, 13(1), 91-107. *Mishra, S. 1., & Conner, R. F. (1996). Evaluation of an HIV prevention program among Latino farmWorkers. In S. I. Mishra & R. F. Conner & J. R. Magafia (Eds), Crossing borders: The spread of H1 V among migrant Latinos (pp. 157-181). CO: Boulder: Westview Press. *Mishra, S. I.. Sanudo. F., & Conner, R. F. (2004). Collaborative research toward HIV prevention among migrant farmworkers. In B. P. Bowser & S. I. Mishra & C. J. 87 Reback & G. F. Lemp (Eds), Preventing AIDS: community-science collaborations (pp. 69-95). NY: New York: Haworth Press. McKleroy, V. S., Galbraith, J. S., Cummings, B., Jones, R, Harshbarger, C., Collins, C., et al. (2006). Adapting evidence-based behavioral interventions for new settings and target populations. AIDS Education and Prevention, 18(Supplement A), SS9- S73. Neaigus, A., Friedman, S. R., Jose, B., Goldstein, M. F., Curtis. R., Ildefonso, G., et al. (1996). High-risk personal networks and syring sharing as risk factors for HIV infection among new drug injectors. Journal of Acquired Immune Deficiency Syndromes and Human Retrovirology, 11, 499-509: *Nyamathi, A., F laskerud, J., Keenan, C., & Leake, B. (1998). Effectiveness of a specialized vs. traditional AIDS education program attended by homeless and drug-addicted women alone or with supportive persons. AIDS Education and Prevention, 10(5), 433-446. *Nyamathi, A. M., & Stein, J. A. (1997). Assessing the impact of HIV risk reduction counseling in improvished African American women: A structural equations approach. AIDS Education and Prevention, 9(3), 253-273. Parker, R. (2001). Sexuality, culture, and power in HIV/AIDS research. Annual Review of Anthropology, 30, 163-179. *Peragallo, N., DeForge, B., O'Campo, R, Lee, S. M., Kim, Y. J ., Cianelli, R., et al. (2005). A randomized clinical trial of an HIV-risk reductiion intervention among low-income Latina women. Nursing Research, 54(2), 108-118. Perloff, R. M. (2001). Persuading People to Have Safer Sex: Applications for social science to the AIDS Crisis. NJ: Mahwah: Lawrence Erlbaum Association, Inc. Perrow, C., & Gullén, M. F. (1990). The AIDS Disaster: T hefailure of organizations in New York and the Nation. NY: Binghamton: Yale University. *Peterson, J. L., Coates. T., Catania, J ., Hauck, W., Acree, M., Daigle, D., et al. (1996). Evaluation of an HIV risk reduction intervention among African American homosexual and bisexual men. AIDS, 10, 319-325. Power, R. (1996). Promoting ri3k management among drug injectors. In T. Rhodes & R. Hartnoll (Eds), AIDS, Drugs and Prevention (pp. 149-160). NY: New York: Routledge. Raj, A., Amaro, H., & Reed, E. (2001). Culturally tailoring HIV/AIDS prevention programs: why, when, and how. In S. S. Kazarian & D. R. Evans (Eds), Handbook of Cultural Health Psychology (pp. 105-239): Academic Press. 88 *Raj, A., Amaro, H., Cranston, K.. Martin, B., Cabral, H., Navarro, A., et al. (2001). Is a general women's health promotion program as effective as an HIV-intensive prevention program in reducing HIV risk among Hispanic women? Public Health Reports, 116(6), 599-607. Rhodes, T., & Quirk, A. (1996). Heroin, risk and sexual safety: some problems for interventions encouraging community change. In T. Rhodes & R. Hartnoll (Eds), AIDS, Drugs and Prevention (pp. 161-185). NY: New York: Routledge. *Robinson, B. B. E., Uhl, G., Miner, M., Bockting, W. O., Scheltema, K. E., Rosser, B. R. et al. (2002). Evaluation of a sexual health approach to prevent HIV among low income, urban, primarily African American women: results of a randomized controlled trial. AIDS Education and Prevention, 14(SuppI.A), 81-96. *Rosser, S., B. R., Bockting, W. O., Rugg, D. L., Robinson, B. B. E., Ross, M. W., Bauer, G. R., et al. (2002). A randomized controlled intervention trial of a sexual health approach to long-term HIV risk reduction for men who have sex with men: effects of the intervention on unsafe sexual behavior. AIDS Education and Prevention, 14(Suppl. A), 59-71. Rushing, W. A. (1995). The AIDS epidemic: Social dimension ofan infectious disease. Colorado: Boulder: Westview Press. *Sanderson, C. A. (1999). Role of relationship cpntext in influencing college students' responsiveness to HIV prevention videos. Health Psychology, 18(3), 295-300. *Sanderson, C. A., & Yopyk, D. J. (2007). Improving condom use intentions and behavior by changing perceived partner norms: an evaluation of condom promotion videos for college students. Health Psychology, 26(4), 481-487. Sebald, H. (1984). Etiology and dimensions of subculture. In Adolescence: a Social Psychological Analysi (pp. 206-225). Englewood Cliffs, New York: Prentice-Hall, Inc. Semaan, S., Kay, L., Strouse, D., Sogolow, E., Mullen, P. D., Neumann, M. S., et al. (2002). A profile of U. S.-based trials of behavioral and social interventions for HIV risk reduction. Journal of Acquired Immune Deficiency Syndromes, 30(Suppl.), 30-50. Schrimshaw, E. W., Rosario, M., Meyer-Bahlburg, H. F. L., & ScharfMatlick, A. A. (2006). Test-retest reliability of self-reported sexual behavior, sexual orientation, and psychosexual milestones among, gay, lesbian, and bisexual youth. Archives of Sexual Behavior, 35(2), 225-234. 89 Schroder, K. E. E., Carey, M. P., & Vanable, P. A. (2003). Methodological challenges in research on sexual behavior: 11 Accuracy of self-reports. Annals of Behavioral Medicine, 26, 104-123. *St Lawrence, J. S., Wilson, T. E., Eldridge, G. D., Brasfield, T. L., & O'Bannon, R. E. I. (2001). Community-based interventions to reduce low income, African American women's risk of sexually transmitted diseases: a randomized controlled trial of three theoretical models. American Journal of Community Psychology, 29, 93 7- 964. *Stanton, B. F., Li, X., Ricardo, I., Galbraith, J ., F eigelman, S., & Kaljee, L. ( 1996). A randomized, controlled effectiveness trial of an AIDS prevention program for low-income African-American youths. Archives of Pediatrics and Adolescent Medicine, 150, 363-372. *Sterk, C. E., Theall, K. P., & Elifson, K. W. (2003). Effectiveness of a risk reduction intervention among African American women who use crack cocaine. AIDS Education and Prevention, 15(1), 15-32. Stokes, J. P., Vanable, P., & McKiman, D. J. (1997). Comparing gay and bisexual men on sexual behavior, condom use, and psychosocial variables related to HIV/AIDS. Archives of Sexual Behavior, 26(4), 383-397. Trickett, E. J. (2005). Community interventions and HIV/AIDS: affecting the community context. In E. J. Trickett & W. Pequegnat (Eds), Community Interventions and AIDS (pp. 3-27). New York: New York: Oxford University Press. Villarruel, A. M., Jemmott, L. S., & Jemmott, J. B. (2005). Designing a culturally based intervention to reduce HIV sexual risk for Latino adolescents. The Journal of the Association of Nurses in AIDS Care, 16, 23-31. *Villarruel, A. M., Jemmott, J. B., & Jemmott, L. S. (2006). A randomized controlled trial testing an HIV prevention intervention for Latino youth. Archives of Pediatrics and Adolescent Medicine, I 60, 772-777. Vinh-Thomas, P., Bunch, M. M., & Card, J. J. (2003). A research-based tool for identifying and strengthening culturally competent and evaluation-ready HIV/AIDS prevention programs. AIDS Education and Prevention, 15(6), 481-498. Watters, J. K. (1996). Americans and syring exchange: roots of resistance. In T. Rhodes & R. Hartnoll (Eds), AIDS, Drugs and Prevention (pp. 22-41). NY: New York: Routledge. 90 *Wechsberg, W. M., Lam, W. K., Zule, W. A., & Bobashev, G. (2004). Efficacy of a woman-focused intervention to reduce HIV risk and increase self-sufficiency among African American crack abusers. American Journal of Public Health, 94, 1165-1173. Weissman, G., & Brown, V. (1995). Drug-using women and HIV: risk-reduction and prevention issues. In A. O'Leary & L. S. Jemmott (Eds), Women at Risk: Issues in the Primary Prevention of AIDS (pp. 175—194). NY: New York: Plenum Press. Williams, D. R., Spencer, M. S., & Jackson, J. S. (1999). Race, stress, and physical health: the role of group identity. In R. J. Contrada & R. D. Ashmore (Eds), Self Social Identity, and Physical Health (pp. 71-100). New York: Oxford University Press. Wilson, B. D. M., & Miller, R. L. (2003). Examing strategies for culturally grounded HIV prevention: a review. AIDS Education and Prevention, 15(2), 184-202. Wong, F. Y., Chng, C. L., & Lo, W. (1998). A profile of six community-based HIV prevention programs targeting Asian and Pacific Islander Americans. AIDS Education and Prevention, 10(Suppl. A), 61-76. Zamboni, B. D., & Crawford, 1. (2007). Minority stress and sexual problems among African-American gay and bisexual men. Archives of Sexual Behavior, 36(4), 569-578. 91 CHAPTER 4 A SYSTEMATIC QUALITY ASSESSMENT OF CULTURALLY SENSITIVE HIV/AIDS PREVENTION INTERVENTIONS IN THE US, 1996-2007 Introduction Many empirical studies are a product of social interactions between the researchers and the participants. Perceptions and expectations of intervention participants about the research investigators and the studies will influence their interactions with the researchers. For example, when intervention behavioral outcomes are perceived as socially desirable behavior by intervention participants, their self- reported behaviors may inflate their actual behavioral changes. In addition, resources that the intervention participants have may be used to sustain their interactions in the studies. People with less resources may be less likely to be retained or to speak up in interventions. Therefore, social positions of the intervention participants may condition the process of interventions Furthermore, culturally sensitive HIV/AIDS preventions (CSHAPIs) targeting subcultural groups are more likely to have bias due to mismatches between the researchers’ and the participants’ racial/ethnic, cultural, socioeconomic status, or a combination thereof. For example, African Americans tend to have a distrust toward govemment-sponsored research. In addition, researchers may not easily access the potential intervention participants due to the closed nature of subcultural communities to outsiders. Therefore, intervention design and conduct that may prevent or minimize bias in intervention outcomes are especially important for CSHAPI studies. 92 Quality of intervention studies is broadly defined to include intervention design, conduct, data analysis, and presentation of study findings, which may minimize or prevent biases in intervention outcomes (Kjaergard, Villumsen, and Gluud, 2001). The importance of quality evaluation is twofold. First, quality may influence study outcomes. For example, studies with small sample sizes are less likely to have significant findings. Poorly conducted studies may inflate or underestimate intervention effectiveness. Second, study biases may reduce the generalizability of the findings and hamper the consistency of collective evidence (Lohr, 2004). However, quality evaluation is often neglected in meta-analysis studies. Moja and his colleagues (2005) indicated that more than 50% of the systematic reviews did not specify quality evaluation in data analysis and interpretation of study results. Similarly, systematic evaluation of study quality in the research of HIV/AIDS prevention intervention has been rare (Darbes, Kennedy, Peersman, Zohrabyan, and Rutherford, 2002; Herbst et al., 2007). In a review of meta-analyses on HIV/AIDS prevention interventions, Flores and C repaz (2004) found no consensus on a standard of study quality among the few meta- analyses involving any form of quality evaluation. The reviews often evaluated some aspects of study quality as inclusion criteria (Wingood and DiClemente, 1996; Lyles et al., 2007; Herbst et a1. , 2007). However, there is a lack of assessment on relationships between study quality and intervention effectiveness in research on HIV/AIDS prevention intervention. This study aims to summarize and examine the collective quality of CSHAPI studies in the US published between 1996 and 2007. It evaluates systematically to what 93 extent the collective quality may impact internal validity of the collective evidence and its generalizability. The relationships between study quality and effectiveness of the CSHAPIs will be tested in Chapter 5. Literature Review A traditional literature review is a narrative evaluation of the aggregation of study findings. A meta-analysis, on the other hand, explicitly lays out the process from data collection, to systematic coding, to quantitative data analysis (Lipsey and Wilson, 2001). Quality evaluation is a central part of meta-analysis, based on systematic coding and evaluation of methodological quality in primary studies. In a traditional narrative review, quality evaluation is less systematic, if not neglected. Study designs and their evaluations have been guided by many primary works (eg., Campbell and Stanley, 1963; Patton, 1978; Cook and Campbell, 1979; Shadish, Cook, and Campbell, 2002; Rossi, Lipsey, and Freeman, 2004). Chalmers and his colleague’s study (Chalmers, Smith, Blackburn et al., 1981) pioneered research of quality evaluation in intervention studies by using a scoring system. Since then, many quality scales have been developed to aid in the systematic evaluation of methodological quality in different study designs. Lohr (2004) evaluated 121 systems for grading the quality of articles assessing randomized controlled trials, observational studies, and diagnostic test studies. In terms of randomized controlled trials, the domains in the criteria of quality evaluation consist of study questions, study population, randomization, blinding, interventions, outcomes of statistical analysis, results, discussion, and funding or sponsorship. The major difference between randomized controlled trials and quasi- 94 experiments is that in the latter, the domain of comparability of subjects replaces the domains of randomization and blinding in randomized controlled trials. In quasi- experiments, there is no randomization involved, which may ensure random distribution of sampling errors between intervention and comparison/control groups in randomized experiments (Shadish etal., 2002). Dimensions of study quality consist of statistical conclusion validity, internal validity, construct validity, and external validity of the primary studies (Shadish et al., 2002). These dimensions may form important components of quality evaluation (Flores and Crepaz, 2004). In general, statistical conclusion validity is founded on the appropriate use of statistical methods to infer presumed co-variation of different variables. lntemal validity, in the context of intervention studies, refers to “inference about whether observed co-variation between A and B reflects a causal relationship from A to B in the form in which the variables were manipulated or measured” (Flores and Crepaz, 2004, p.53). The variables under manipulation may be associated with people, settings, - treatment variables, and measurement variables. Construct validity refers to generalizations of study operation to the constructs used to describe those operations. External validity is about generalizations of the findings (causal relationships in experiment studies) to and across populations, settings, treatment variables, and measurement variables (Shadish et al., 2002). The main purpose of quality evaluation is to evaluate whether any bias incurred in the process of intervention from design to data analysis may threaten the four kinds of validity. In studies of quality evaluation, internal validity has received more attention than the others partly due to its approximate inference to casual relationships (Flores and 95 Crepaz, 2004). External validity has not been examined as much, but has its essential implication on policy making. In a meta-analysis of collective quality, external validity may also be relevant as the primary purpose of meta-analysis is to generalize findings from the pool of primary studies. Table 4.1 Threats to Internal Validity and External Validity :55” Threats to Internal Validity Ambiguous temporal precedence: Lack of clarity about which variables occurred first may yield confusion about which variable is the cause and which is the effect. Selection: Systematic differences over conditions in respondent characteristics that could also cause the observed effect. History: Events occurring concurrently with treatment could cause the observed effect. Maturation: Naturally occurring changes over time could be confused with a treatment effect. Regression: when units are selected for their extreme scores, they will often have less extreme scores on other variables, an occurrence that can be confused with a treatment effect Attrition: Loss of respondents to treatment or to measurement can produce artificial effects if that loss is systematically correlated with conditions. Testing: Exposure to a test can affect scores on subsequent exposures to that test, an occurrence that can be confused with a treatment effect. Instrumentation: the nature of a measure may change over time or conditions in a way that could be confiised with a treatment effect. Additive and interactive effects of threats to internal validity: the impact of a threat can be added to that of another threat or any depend on the level of another threat. Threats to External Validity Interaction of the causal relationship with units: an effect found with certain kinds of units (people) might not hold if other kinds of units had been studied Interaction of the causal relationship over treatment variables: an effect found with one treatment variation might not hold with other variations of that treatment, or when that treatment is combined with other treatments, or when only part of that treatment is used. Interactions of the causal relationship with outcomes: an effect found on one kind of outcome observation may not hold if other outcome observations were used. Interactions of the causal relationship with settings: an effect found in one kind of setting may not hold if other kinds of settings were to be used. Context-dependent mediation: an explanatory mediator of a causal relationship in one context may not mediate in another context. Adapted from Shadish et al. (2002) Table 4.] lists threats26 to internal and external validity. Sources of threats to internal validity may vary by study design. For example, in randomized experiments, 6 . . see Shadish, C ook, and Campbell, 2002 for detailed discusswns on each threat 96 threats to internal validity are largely reduced by adequate randomization, which minimizes selection bias and pre-test differences between intervention and comparison/control groups as a result of history and maturation. Similarly, compared with observational studies, randomized experiments have fewer problems of confounded treatment effects by regression to mean, testing and instrumentation due to differences between intervention and comparison/controls. Under adequate randomization, the differences occur only by chance. Contrary to observational studies, ambiguous temporal precedence is not a threat to internal validity in a properly conducted randomized experiment because the treatment variables are manipulated in a controlled context. In randomized controlled trials, differential attrition by groups may cause inferential problems, like it does in observational studies (Shadish et al., 2002). Among the scoring systems and quality scales, the most frequently used tool is the Jadad scale (Jadad et al., 1996). However, it has been criticized for its lack of allocation concealment in evaluation, which was found to influence treatment effects (Moja et al., 2005). In HIV/AIDS intervention research, researchers have developed standards of study quality, which however, are descriptive, not a direct tool for assessment (Flores and Crepaz, 2004; Lyles, Crepaz, Herbst, and Kay, 2006). In addition, the standards of study quality are more applicable to individual- or group-level interventions (Flores and Crepaz, 2004). Researchers have adapted to the standards of study quality in various ways, such _ as by assigning a scoring system (Darbes et al., 2002; Herbst et al., 2007), or using them in inclusion criteria (Lyles et al., 2007). However, quality evaluation has not been widely used in evaluation research of HIV/AIDS prevention intervention. 97 In general, aggregating findings of quality evaluation in meta-analysis studies remains unstandardized as there is still no consensus on the ideal checklist and scale for assessing methodological quality. On the other hand, there is still no agreement whether a generic assessment tool could be developed for use in all cases (Moja et al., 2005). As each checklist and scale is somewhat different in their focuses and scoring systems, the same methodological quality may yield dissimilar quality scores using different quality scales (Jfini, Witschi, Bloch, and Egger, 1999). In addition, if one aspect of design is more importantly related to study outcomes, a quality score in total may risk loss of information (Greenland, 1994). The inconsistency in focus and scoring of quality evaluation makes its collective evidence less conclusive. However, a quality of study report is as important as methodological quality because sufficient information about the studies lays the groundwork for successful evaluation of these studies. Missing information has often been linked with deficiency in methodological quality. However, Soares et al.’s study (2004) yielded contradictory evidence that the methodological quality of randomized controlled trials in their review was better than reported. Huwiler-Mfintener, Jfini, Junker, and Egger’s (2002) study also found that randomized controlled trial studies with similar quality of reporting may differ in methodological quality. New guidelines from international groups have provided clear instructions on how studies with different study designs should be reported, such as randomized controlled trials in the CONSORT statement (Moher, Schulz, and Altman, 2001), observational studies in the MOOSE statement (Stroup et al., 2000), and systematic reviews in the QUORUM statement (Moher et al., 1999; Lohr, 2004). These guidelines are important 98 not only for dissemination of study findings, but also for aggregation of study findings which requires specific knowledge of each domain in the criteria of evaluation. In order to avoid the inconsistency in an aggregate of quality evaluation using different quality scales, the Cochrane Collaboration27 has encouraged a simple approach to quality assessment based on individual components. Among them, randomization, allocation concealment”, blinding”, and intention to treat 30 have received the most attention from researchers of randomized controlled trials (e.g., Kjaergard et al., 2001; J fini et al., 2001; Montenegro et al., 2002). Previous research suggests that biases in any of them may confound study outcomes (Shadish et al., 2002). For example, Ji'ini et al.’s study (2001) found that inadequate allocation concealment may impact the size of the treatment effect by an increase of 30%. Other similar studies also came to the same conclusion (Kjaergard et al., 2001; Schulz et al., 1995). The findings about the impact of inadequate double blinding on treatment effects vary (Montenegro et al., 2002). Some studies found larger treatment effects (Kjaergard, Villumsen, and Gluud, 2000; Schulz et al., 1995) while other studies point in the opposite direction (Moher et al., 1998; Als- Nielsen, Chen, Gluud, and Kjaergard, 2003). Schulz et al. (1995) found no association between intervention effects and a lack of intention-to-treat analysis. This study aims at summarizing and examining the collective qualities of 50 C SHAPI studies in the U. S. published between 1996 and 2007. It uses both approaches 27 http://www.cochrane.org/ 8 . . . . methods used to conceal random allocation sequence. Adequate methods include central randomization at a site remote from the intervention setting, sequentially numbered, sealed, opaque envelopes; etc (Jfini, Altman, and Egger, 200]) 29 . . . . . . , . . whether or not the partic1pants. intervention fac1litators, and/or outcome assessors were blinded to their assignments of intervention conditions (Moher, Schulz, and Altman, 2001). All participants assigned to different interventions were retained in data analysis (Schulz. Chalmers, Hayes et al., 1995). 99 of scoring and individual components in summarizing collective characteristics of quality among the studies. An examination of collective quality in the literature is based on aggregate quality scores/levels and percentage of the presence of important indicators to study quality in different study domains. The evaluation is conducted systematically by using a quality scale derived and modified from the existing quality scoring systems and scales. This study presents its findings by domains in the reports of primarily studies. It may unravel important differences in quality across domains, which proximate to different types of validity. lnferences to different kinds of validity are discussed in the conclusion. Methodology Samples Fifty C SHAPI studies were identified through extensive searches in a variety of sources, including databases31 and research registers”, 30 major journals”, and the reference sections of relevant studies. The search in databases and research registers used a combination of the following keywords “HIV or AIDS, prevention or intervention or program or training or education or clinical trial or randomized controlled trial,” and 3| . . . . IBSS, JSTOR, Medline, Proquest-Dissertation Theses, PsyARTICLES, , PsyclNFO, Psychology, Soc1al Sciences Abstracts, Social Sciences Citation Index, Social Services Abstracts, Social Work Abstracts, and Sociological Abstract The Cochrane Collaboration Register is located at httLfl/‘clinicaltrials.gov 33 AIDS, AIDS and Behavior, AIDS Care, AIDS Education and Prevention, American Journal of Community Psychology, American Journal of Drug and Alcohol Abuse, American Journal of Epidemiology, American Journal of Health Behavior. American Journal of Health Promotion, American Journal of Preventive Medicine, American Journal of Public Health, Archives of Pediatrics and Adolescent Medicine, Contemporary Sociology, Health Education and Behavior, Health Education Research, Health Psychologv, International Journal of Health Education, Journal of A cquired Immune Deficiency Syndromes, Journal of Adolescent Health, Journal of A merican Medical Association, Journal of Behavioral Medicine, Journal of Community Health, Journal ofConsulting and Clinical Psychologv. Journal of Drug Issues, Journal of Health and Social Behavior Journal of School Health, Journal of Women ’s Health, Public Health Reports, Sexually Transmitted Diseases, Social Science and Medicine, and Women and Health 100 “behavior or condom. ” Criteria of inclusion and exclusion of eligible studies were explained in Chapter 3. In general, the CSHAPI studies shared the following characteristics: they were published after 1995, conducted in the U. 8., included at least one culturally sensitive component in an intervention condition, were designed either as experiments with randomization or quasi-experiments with pre-tests and post-tests in both intervention and control/comparison conditions, and used either condom use or unprotected sex behavior as a behavioral measure. CSHAPI studies targeting HIV- positive patients, mentally-ill patients, or prison inmates were excluded. Quality measures used This study developed a quality scale34 based on previous work (Chalmers et al., 1981, Liberati, Himel, and Chalmers, 1986; Reisch, Tyson, and Mize, 1989; Sindhu, Carpenter, and Seers, 1997), standards of study report on systematic reviews and randomized controlled trials (Moher, Eastwood, Olkin, Rennie, and Stroup, 1999; Moher et al., 2001), suggestion from other studies of quality evaluation (Wortman, 1994; Flores and Crepaz, 2004; Lohr, 2004), research in meta-analysis (Lipsey and Wilson, 2001; Alderson, Green, and Higgins, 2004) and research methods (Campbell and Stanley, 1963; Cook and Campbell, 1979). A pilot test was conducted on ten CSHAPI studies first retrieved in the process of data collection. The scale was revised based on the pilot test. The quality scale consists of five domains of evaluation: quality of sampling and study designs, quality of program descriptions, quality of data analysis, quality of study results. and quality of study conclusion. These domains assess threats to four types of validity. 34 . Refer to Appendix B for the quality scale used in this study. 101 In general, questions in quality of sampling and study design assess threats to internal validity and construct validity. Refer to Appendix B for the items assessed for each domain. Questions in quality of analysis evaluate statistical conclusion validity. Questions in quality of study results and study conclusion focus on external validity. Quality of study reports by domains is assessed by confidence rating based on a 5- point scale (very low, low, moderate, high, and very high) with a higher score indicating a higher level of confidence. The confidence rating is defined as evaluators’ confidence in the judgment of quality based on the information provided in the report (Orwin and Cordray, 1985). Very low confidence refers to little basis for making a quality judgment. Low confidence means a judgment based on a best estimate. Moderate confidence refers to evaluation based on weak inference. A high level of confidence indicates strong inference from study reports. Finally, a very high level of confidence refers to judgment based on some explicit statements. Data on use of double blinding, allocation concealment, and intention to treat was also gathered. Data analysis The entire set of the eligible studies was coded by the author. Three kinds of quality indicators were estimated. One is the total of quality scores under each domain based on individual items, each of which was assigned equal weight under each domain. The quality scores by study domains were recorded to quality levels on a 5-point scale (poor/ marginal/moderate/solid/outstanding) with a higher score indicating a higher level of quality. The scoring system is shown in Table 4.2. 102 Table 4.2 Scoring System Domains Sampling Program Analysis Result Conclusion Accumulative / study description study quality designs Number of 10 4 8 9 7 38 Questions Individual 0-1 0-2 0-1 0-1 0-2 5.0213. Range of total 0-10 0-8 0-8 0-9 0-14 0-49 $.03 Level of gualig by range of scores Poor 0-2 0-l 0-l 0-1 0-2 0-7 Marginal 3-4 2-3 2-3 2-3 3-5 8-15 Moderate 5-6 4-5 4-5 4-5 6-8 16-29 Solid 7-8 6-7 6-7 6-7 9-1 I 3040 Outstanding 9- I 0 8 8 8-9 12-14 41-49 This scoring system by no means directly estimate a level of study quality, as each component under each domain was assigned equal weight. Some components may be more important than the others. Moreover, some domains may be more important than the others in terms of their contributions to study quality. However, quality scores by this scoring system may provide information of the number of quality issues being adequately addressed under each domain. The second quality indicator is an accumulative quality score that is an aggregate of total quality scores from five domains of evaluation. The accumulative quality scores were also recoded into quality levels on a 5-point scale, the same as the quality levels by study domain. The percentage of the presence of important individual components across studies is also reported as a quality indicator to complement the information provided by the quality scores/levels. 103 In addition, this study conducted subgroup analysis of study quality by interventions targeting different cultural groups and other subgroups in the samples. The classification of CSHAPI studies by cultural groups was documented in Chapter 3. Findings Figure 4.1 shows that the accumulative quality scores of the CSHAPI studies were normally distributed with a mean of 27.52 (SD= 4.358, Skewness=-0.029). By level of study quality, 68% of the studies were conducted with a moderate study quality, while 32% had solid study quality. In the selection of eligible study reports, this study excluded those reports with one-group design or without pretest and posttest applied to both intervention and control/comparison groups. Therefore, the focus was on studies with good overall designs, which may be related to quality of other study domains because relative fewer issues in these types of study designs will confound internal validity. 104 Figure 4.1 Distribution of Accumulative Study Quality 12- 10-1 Frequency 8— 6— 4.. 2—1 Mean =27.52 _ Std. Dev. =4.358 0 I ‘ I ' I I j j 'i' i N=50 15.00 20.00 25.00 30.00 35.00 40.00 Accumulative study quality scores Table 4.3 Level of Study Quality by Domains in the Evaluations Sampling/ Program Data analysis Study’s Study’s study designs description results conclusion Poor 4% 0 6% 28% 2% Marginal 46% 0 14% 36% 18% Moderate 38% 4% 36% 20% 36% Solid 10% 74% 40% 12% 34% Outstanding 2% 22% 4% 4% 10% Quality of Sampling and Stuih/ Designs Fifty percent of the studies were rated below moderate quality in sampling and study designs, partly due to non-random sample selection that may subject them to systematic sampling bias. None of the studies had conducted random sampling while only 12% explicitly mentioned that the participants were similar to the targeted population. In addition, only 14% of the studies provided information on the number and/or characteristics of subjects who were eligible but refused to participate. Instead, random assignment of the participants into different intervention conditions was practiced in 92% of the studies to ensure pre-intervention sample equivalency between different intervention conditions. Seventy-four percent of the random assignments were conducted at individual levels while 26% were at site/community/time block levels. Relatively little information was provided concerning the number and characteristics of the participants dropping out before completing interventions (3 0%), compared with information of the participants who were lost to follow-up tests (58%). The retention rate mean at final follow-ups was 76% (SD=0. I45). Implementation of the interventions was not generally described in the studies. About one-third of the studies reported monitoring mechanisms, among which 44.4% included participants’ rating of the interventions or of their satisfaction, and 23.5% had monitors checking implementation of the interventions, though not as facilitators. In addition, audio-taping interventions for review were described by 35.5% of the studies with monitoring mechanisms. The mean of confidence rating on the evaluations of sampling and study designs is 4.88 (SD=0.328) on a S-point scale, which is close to a very high level of confidence in this quality evaluation, based on the information 106 provided in the study reports. All of the evaluations on quality of the primary studies were based on strong inference or explicit reports. gLualitj of Program Description Quality of program descriptions among the studies was in general high, ranging from moderate to outstanding levels. However, the program descriptions tended to focus on experimental groups. More than 95% of the studies provided clear information of the objectives and contents of experimental groups, while only about 50% providing comparable information of control/comparison groups. The mean of the confidence rating was 4.94 (SD=0.314) on a 5-point scale, referring close to a very high level of confidence in the evaluation of program descriptions. Ninety-eight percent of the evaluations on the primary studies were based on strong inference or explicit statements. Quality of Data Analysis About three-quarters of the studies conducted data analysis in moderate (3 6%) or solid (40%) quality. Twenty percent of the studies were rated as either poor or marginal quality in data analysis. The high quality of data analysis mostly represented those with sufficient sample sizes. inclusion of baseline differences between intervention and control groups in the data analysis, control for potential confounders, and use of multivariate analysis. Over three-quarters of the studies had sufficient sample sizes, as indicated by the use of power analysis or assignment of more than 100 subjects in each intervention condition. Sixty-eight percent of the studies included pre-test scores of intervention and control groups. Eighty-four percent controlled for at least one potential confounder 107 through intervention effects. In addition to baseline scores, age and gender were the most commonly tested moderators. Socioeconomic status, such as education or poverty, was not generally tested mainly due to the homogeneous socioeconomic status within most of the samples. In addition, 40% of the studies reported repeated-measure analysis in order to test time effect on intervention effectiveness. Forty-eight percent of the studies reported using multivariate regression analysis. In terms of the reporting of study findings, about three-quarters of the studies described test statistics and observed probability levels. However, reports of confidence interval of intervention effects were not common (28%). Seventy-six percent of the studies had analytic specifications of behavioral measures consistent with the evaluation questions or hypotheses under study. The inconsistency in specification of variables was partly related to inadequacy of behavioral measures to quantify risks of HIV infection. On average, these CSHAPI studies examined retrospective sex behavior prior to about 2.44 months from the time of any tests (median=3, SD=1.583, k=46). However, some studies had tested self-reported behavior across a broad period of time, such as six months, while others did not specify a time range. Previous studies suggest that long-term recall of sex behavior of more than three months is not reliable (Schroder, Carey, and Variable, 2003; Schrimshaw, Rosario, Meyer-Bahlburg, and ScharfMatlick, 2006). Furthermore, some studies reported condom use during last sex encounter or frequency of risk behavior as behavioral measures. However, these did not take into consideration frequency of sex. Percentage of condom use, consistent condom use, or number/percentage of unprotected sex are more adequate behavior measures of risks to 108 HIV infection. The average confidence rating on the evaluations of data analysis is 4.82 (SD=0.482), involving 96% of the evaluations based on strong inference or explicit statements. In addition, previous studies have found that change of sex behavior after interventions differs by types of partners and types of sexual behavior. Slightly over half of the CSHAPI studies (56%) focused on different types of sex although some excluded certain types of sex in the analysis due to small sample sizes. In contrast, only about one- third of the studies (32%) examined sex behavior among different types of sex partners. Therefore, the findings of these CSHAPI studies may be, to some level, generalizable to general sex behavior, but is very limited in application to specific type of sex behaviors and sex partners. Furthermore, generalization of the CSHAPI studies is restricted in the sense of long-term prospects. One-quarter of the studies had 3-month follow-ups as the last tests (k=48). The mean number of months of the last follow-up tests was 8.6 (median=6, SD=6.07). Few of the studies (12.5%) had their last follow-up tests conducted more than a year after the interventions were completed (k=48). Quality of Study Results The quality of study results Was relatively low, compared with the qualities of other domains. Sixty-four percent of the studies were rated either poor or marginal in terms of study results. Twenty percent were of moderate quality. The overall low quality was mainly a result of lack of information on attrition rates for each intervention condition or equivalency of attrition between different intervention conditions. 109 Insufficient examination of the impact of attrition on intervention effectiveness was also common. Only 42% of the studies provided information for attrition in each intervention condition or tested whether there was differential attrition between different intervention conditions. Attrition analysis was usually conducted between those lost to follow-ups and others remaining throughout follow-ups (50%). Very few studies, 14%, analyzed whether those refusing to attend interventions were similar to others completing follow- ups. Slightly more studies (30%) analyzed whether those dropping out of the interventions were like those retained for follow-up tests. For studies immediately assigning participants into one-time treatment or involving long-term community-level interventions, the issue of dropping out of intervention was not applicable. Primarily due to lack of overall attention on attrition analysis, over 80% of the studies did not discuss limitations of their study findings due to differential attrition between intervention conditions or between completers and refusers or those failing to complete the interventions. Thirty percent of the studies addressed the limitations of their study findings due to differences between those completing and failing to complete follow-ups. Mean of confidence rating on the evaluations was 4.92 (SD=0.34). Ninety- eight percent of the evaluations on the primary studies were based on strong inference or explicit descriptions. Quality of Study Conclusions The quality of study conclusions was in general moderate or above. However, 20% made marginal or poor conclusions. The moderate to poor quality of conclusion 110 was largely due to a lack of discussion on limited generalization of the study findings due to factors in intervention implementation, sampling, and study designs. Ninety-eight percent of the studies discussed generalization of the study findings in terms of intervention programs. However, 44% did not discuss limitation of study designs. Attention on process evaluation was also largely lacking. While about two- thirds of the studies did not describe monitoring mechanisms, 88% did not discuss any limitations due to issues in intervention implementation. In addition, seventy percent of the studies discussed generalization of the representation of the samples to the targeted population, usually based on a match of racial/ethnic and sex status. However, none of the studies had conducted random sampling. Fifty-six percent did not address limitations for generalizing to the general public due to non-random sampling. Furthermore, as discussed earlier, attrition analysis about refusers and those failing to complete interventions or lost to follow-up tests were neglected by more than 80% of the studies. As a result, representation of the samples to the targeted population was largely confined to racial/ethnic and sex status of the participants in a given locale. Factors within the communities or the individuals that may confound participation in intervention were largely left unexplained among the studies. Fifty-eight percent of the studies did not test motivation of the participants in changing their HIV/AIDS sexual risk behavior. In terms of generalization of behavioral measures, limitation of self-reported behavior as a behavioral measure had been notified by two-thirds of the studies. Self- reported sex behavior was subject to bias due to social desirability. A variety of experimental manipulations were taken to reduce bias in self-reported sex behavior. The 111 low reliability of self-reported sex behavior over a lengthy period of time was also discussed. Limitation of data analysis was addressed by about half of the studies (46%). The most common problem in data analysis was small sample size due to high attrition after intervention or rarity of certain types of sex partners and sex practices. The mean confidence rating on evaluation of study conclusions was 4.94 (SD=0.24). All of the evaluations on the primary studies were based on strong inference or explicit statements. Allocation Concealment, Blinding, and Intention-to-treat Analysis Allocation concealment and double blinding are central procedures for appropriate randomization and manipulation in experiments. Ninety-two percent of the studies reported random assignment of participants into different intervention conditions. However, allocation concealment had not been used or reported in most of the studies. Only 24% of the studies reported adequate allocation concealment, among which 75% involved an on-site computer system or a staff keeping the randomization list private, while 25% used sequentially-numbered, opaque, sealed envelopes. Fourteen percent of the studies did not use any allocation concealment. Sixty-two percent of the studies did not provide information whether any allocation concealment had been used. In terms of blinding, only 16% of the studies described procedures involving double-blind assignments. Due to the nature of the HIV/AIDS prevention interventions, it may be difficult to blind intervention facilitators in their intervention assignments, and 14% of the double-blinding assignments were for the participants and behavioral 112 assessors or interviewers. In addition, over 80% of the studies were single-blinding trials. Only 2% of the studies were open trials. Intention-to-treat analysis is essential to generalizations of intervention effectiveness to the targeted population as it takes into consideration attritional effects on overall intervention effectiveness. Those retained throughout follow-up tests may be more likely to be highly motivated individuals in behavior change and thus, inflate intervention effectiveness. However, only twenty-four percent of the studies explicitly mentioned intention-to-treat in the data analysis while 34% conducted specific intention- to-treat analysis. The definitions of intention-to-treat, in terms of inclusiveness of the participants, varied across studies. Intention-to-treat analysis in these studies tended to be less inclusive to those lost at early stages of the interventions. About one-third of the studies included people who did not complete intervention or follow-up tests in the data analyses. Only 12% included people who did not start the allocated intervention after the assignment. Few studies, 6%, included people who satisfied the criteria of enrollment, but were not assigned to any intervention. Therefore, use of intention-to-treat analysis in these studies did not compensate for their overall lack of attrition analysis on those lost at early stages of the interventions. Furthermore, the ways that missing data was handled in an intention-to-treat analysis were mostly not reported in the studies involving any use of intention-to-treat analysis. Therefore, an estimation of the effect of intention-to-treat analysis on intervention effectiveness is limited. In addition, there is a general lack of information on allocation concealment, blinding, and intention-to-treat analysis in the study reports. Mean ofconfidence rating 113 on the evaluations of allocation concealment, blinding, and intention-to-treat analysis is relatively lower, compared with that of the five domains of study reports. However, compared with allocation concealment and blinding, more information was provided on intention-to-treat analysis. Eighty-six percent of the evaluations on intention-to-treat analysis were based on strong inference or explicit statements with a mean of 4.24 (SD=0.744). In contrast, the mean of confidence rating on the evaluations of allocation concealment was 3.53 (SD=0.789), involving only 34% of the evaluations on the primary studies based on strong inference or explicit statements. The mean of confidence rating on blinding was 4 (SD=0.639), including 80% of the evaluations based on strong inference or explicit descriptions. Level of Study Quality by Subgroups Sixty-eight percent of the studies were conducted with moderate study quality, while 32% had solid study quality. Dominant cultural groups in the studies, in general, did not significantly co-vary with study quality (F =0.734, df=4, 45, p=0.574). Figure 4.2 shows that the studies targeting African Americans as a cultural group appeared to have a higher level of study quality. In a two-tailed test, these studies are only slightly associated with a higher level of study quality (r=0.109, p=0.45). On the other hand, studies targeting White Americans as a major cultural group were moderately related to a lower level of study quality (r=-0.258) close to a significant level (p=0.071) in a two- tailed test. No other studies targeting different cultural groups significantly explained variation of study quality, while their correlation with the variation of study quality was also very low (r<0. 1). Therefore, the distribution of different levels of study quality was 114 Figure 4.2 Level of Study Quality by Culturally Sensitive Interventions Cultural Category [:1 MSM Drug users African Americans E Hispanic Americans [III White Americans Count . . _ . .I. Moderate Solid Accumulative quality level Conclusion The study quality of primary studies determines collective quality of a broad spectrum of research, which can be assessed in terms of the validity of findings and generalization. In meta-analysis the causal relationships that are aggregated relate to causal inferences associated with the constructs of the primary studies under review. Therefore, biases that influence causal inferences in primary studies can be carried forward cumulatively to confound aggregate causal inferences. Similarly, the generalization ofa meta-analysis is also limited to the constructs targeted in the primary studies. Quality of construct validity in the primary studies will directly impact the 115 generalizations that a meta-analysis can make. Thus, an evaluation of the quality by study domains can be an integral part of a meta-analysis on intervention effectiveness This study systematically evaluates the quality of 50 C SHAPI studies which were obtained through exhaustive searches of various publication sources. These studies were drawn from a population of CSHAPI studies in the U. S. which were homogeneous in terms of the following features: incorporating at least one form of culturally sensitive component, targeting a cultural group, conducted with either randomized experiments or quasi-experiments involving pre-tests and post-tests, including condom use or unprotected sex behavior as a behavioral measure, and published between 1996 and 2007. However, these studies were heterogeneous with respect to demographic characteristics of the participants, intervention objectives, and culturally sensitive components. I had reviewed in Chapter 3 aggregate demographic characteristics of the participants, study designs, characteristics of intervention treatments, and intervention settings. For example, contrary to one-third of the studies targeting Afiican Americans as a cultural group, only 8% involved White Americans as a major cultural group. The participants of the CSHAPI studies were more likely to be African Americans, adults, females. single people, and those with low socioeconomic status. An advantage of meta-analysis is that it is based on multiple studies. It increases the number of its construct inferences, while enhancing statistical power. However, lumping together a heterogeneous set of studies may hamper causal inference from the meta-analysis. This kind of study which involves heterogeneous studies may be preferable to creating an explicit typology of the intervention constructs and to test the effects on causal relationships (Shadish et al., 2002). This approach also helps unravel 116 impacts on the causal relationships from various potential moderators which may not be easily evaluated in a single study. The heterogeneity of the CSHAPI studies has provided a chance to systematically evaluate the impact of demographic characteristics, types of interventions, behavioral variables, and settings on the causal relationships between the intervention treatments and intervention effectiveness, if these CSHAPI studies have been conducted properly. The heterogeneity of these studies also enhances their collective external validity. This study found that in general, the CSHAPI studies possessed moderate or solid quality with 92% using randomized experiments. In subgroup analyses, it was found that the quality of CSHAPI studies did not differ by the aggregate demographic characteristics of participants except that studies with White Americans as a cultural majority were moderately correlated with a lower level of study quality. Therefore, study quality is not a factor that may explain the variation in effectiveness of interventions targeting different subpopulations. A meta-analysis of the intervention effectiveness should examine a potential interaction effect of study quality and the studies involving White Americans as a cultural majority. This study found major deficiencies in the quality of CSHAPI studies that may hamper causal inferences and generalizations that these studies can make. Threats to internal validity among these studies included a lack of random sampling, paucity of information on those failing out before follow-up tests are completed, especially those at the early stages of interventions, and inadequate attention to attrition analysis that examines the impact of attrition on intervention effectiveness. Ninety-two percent of the CSHAPI studies randomly assigned the participants to different intervention conditions to 117 ensure baseline equivalency between different intervention conditions. Randomization in the intervention assignments may improve causal inference because samples in different intervention conditions were not systematically different from each other. However, randomized assignments did not prevent sampling bias in terms of the selection of participants. Sampling bias may exist in both of the intervention and control/comparison groups, in terms of their similarities to the targeted population, and may influence generalizability of the study findings. Moreover, randomized assignments cannot prevent impact on intervention effectiveness from systematically differential attrition in the participants or between different intervention conditions. Therefore, any causal inferences the CSHAPI studies can make were contingent on those subjects retained in the interventions. To what extent the participants retained in data analysis represented the targeted populations were not well examined in these studies. Furthermore, this study found that 76% of the CSHAPI studies had moderate or high level of quality with respect to data analysis. In general, these CSHAPI studies had sufficient sample sizes although statistical power may be low in analysis involving different types of sex partners and sexual behavior. Therefore, type II error was less likely to occur when testing about general sex behavior or sex partners Multivariate analysis was also used in about half of the studies to control for the impact of moderators on causal relationships. The major threats to statistical conclusion validity among the studies included an overall lack of a description on intervention implementation (66% of the studies), a lack of report or estimate on reliability of behavioral measures (98%), lack of intention-to- treat analysis (66%), and little detail on the ways that missing data had been handled in 118 intention-to-treat analyses (98%). Only 16% of the studies had the participants rating the interventions, while 62% involved community participation in formative research or pilot tests. The success of the intervention implementation was less likely to be examined in the participants’ perspectives. Moreover, unreliability of behavioral measures may weaken the relationships between two variables (Shadish et al., 2002). Lack of intention-to-treat analysis may not impact estimation of intervention effectiveness or causal inference if attrition is equivalent between different intervention conditions in randomized experiments. For example, if in both intervention conditions, attrition rates of younger people were the same, age would not be a factor confounding a causal inference between intervention treatment and intervention effectiveness. However, if attrition was systematically differential among the participants, such as a higher attrition rate among younger participants, a lack of intention-to-treat analysis may still hamper generalization of the intervention effectiveness to the targeted population. In addition, there are various ways to handle missing data in an intention-to-treat analysis, such as imputing good outcomes, poor outcome, or group means. If the intention-to-treat analyses in the CSHAPI studies had handled missing data in different ways, the impact of intention-to-treat analysis on intervention effectiveness may be cancelled out in a meta-analysis. On the other hand, if the intention-to-treat analyses are consistently biased in specific ways, such as imputing good outcomes in missing data, the aggregate intervention effectiveness may be biased in a specific direction. Only 17.6% of the studies involving intention-to-treat analysis reported the ways missing data had been handled. Given these threats to statistical conclusion validity, it is hard to judge whether 119 the statistical conclusions in the CSHAPI studies were biased upward or downward, while the statistical conclusions may be more contingent on those participants retained. longer in the interventions. This study found threats to construct validity in these CSHAPI studies, including mono-method bias (100%), lack of use or description of allocation concealment (76%), and a lack of practice or description of double blinding (84%). All of the studies had relied on self-reports as behavior measures, while two-thirds of the studies pointed out the limitation of a self-report method that may be influenced by social desirability. Without allocation concealment, the respondents’ perception to the interventions may confound their responses to the intervention treatments and measures that were being tested in the interventions. Double-blinding trials for the participants and the outcome assessors may work to reduce experimenter expectancies about desirable or undesirable responses to the participants receiving different intervention treatments. These threats to construct validity may contribute to criticisms leveled against the CSHAPI studies that the effectiveness of CSHAPIs was largely due to the participants’ perceptions of social expectancies, instead of the intervention treatments they had received. The social desirability of HIV/AIDS safe sex behavior may have inflated the effectiveness of CSHAPIs. Moreover, this study found moderate to poor external validity in over 70% of the CSHAPI studies by inferences derived from the evaluation of conclusions that these studies had made. In general, these studies did not discuss limitations of generalization due to non-random sampling, attrition, and lack of evaluation on intervention 120 implementation. Limitation of self-reports as behavioral measures was noted by two- thirds of the studies. Generalization of this study, or any meta-analysis, is straightforward to some extent. The primary studies under evaluation which composed the study population itself in a well-conducted meta-analysis do not require a significant test to validate correlational or causal relationships between variables in the study population. Significant tests in a meta—analysis confer different meanings from those in the primary studies in that generalization in a meta-analysis is based on an aggregate study level. The meta-study-population targeted by a meta-analysis is a hypothetical one. Primary studies in a meta-analysis could be considered derived from random sampling in the hypothetical meta-study-population. Generalization to the meta-study-population is of interest to scientists, who are concerned with the replication of relationships between variables. Policy makers may also find this useful in terms of whether the pattern of intervention effectiveness could be replicated in the future in an aggregate of interventions or in a large-scale multi-site study. One limitation of this specific study includes a lack of the estimate of reliability. Because quality evaluators must have advanced training in statistics and be substantially familiar with experiment design and research of HIV/AIDS prevention interventions, training of additional coders was not conducted before the dissertation was completed. In addition, the scoring system used in this study did not provide a direct estimate of study quality. However, the quality scores or levels obtained by this scoring system indicated a number of important quality issues being addressed in the primary studies. 121 Moreover, the collective study quality is less likely applied to studies targeting White Americans, youths, males, those with steady partners or being married, or others with high socioeconomic status. The discussion of collective study quality is also less generalizable to the studies conducted outside of community settings or involving different types of HIV/AIDS risk behavior. Furthermore, the collective study quality did not represent quality in other CSHAPI studies involving different types of intervention treatments, conducted with one-group designs or quasi-experiments without pre-tests, conducted outside of the U. S., or published before 1996. Finally, to what extent quality of the CSHAPI studies should be improved or could be enhanced is a complex issue that depends on many factors. Resources available to researchers may limit their options in terms of sampling, study designs, and intensity and complexity of intervention. Availability of the resources may be a political and social issue dependent on how important the research is to the broader population. For a highly stigmatized disease, such as AIDS, whether the US. government understands the importance of HIV/AIDS prevention will also influence funding available to the research. Study design is also a result of decision making which takes into consideration various advantages and disadvantages that have been associated with randomized experiments and observational studies. Causal inferences are more readily supported by randomized experiments. However, findings of observational community studies are more generalizable to the community settings where HIV sex risk behaviors take place. Moreover, whether the quality of interventions could be improved may also depend on history. Self-report of sex behavior as a behavioral measure, despite of its bias. was used by all of the CSHAPI studies because there was no alternative measure that 122 could be readily used by the researchers. However, when social expectancy of sex behavior is highly varied, the impact of social desirability on self-reports of sex behavior may be reduced. What academic communities may do to improve study quality is to establish standards of quality evaluation. These have been largely descriptive to date; there is no generally accepted scoring system available. A scoring system around domains in study reports or four types of validity may enhance the examination of collective quality in research. It also offers a guideline of study quality for intervention studies. References Alderson, R, Green, S., & Higgins, J. P. T. (2004, March). Cochrane reviewers' handbook. Retrieved January 3 1 st, 2004, from the World Wide Web: http://mchochrane.org/resources/handbook/hbook.htm AIS-Nielsen, B., Chen, W., Gluud, C., & Kjaergard, L. L. (2003). Association of funding and conclusion in randomized drug trials: a reflection of treatment effect or adverse events? The Journal of American Medical Association, 290(7), 921-928. Campbell, D. T., & Stanely, J. C. (1963). Experimental and quasi-experiment designsfor research. Chicago: Rand McNally. Chalmers, T. C., Smith, H. J., Blackburn, B., et a1. (1981). A method for assessing the quality of a randomized control trial. Controlled Clinical Trials, 2, 31-49. Colditz, G. A., Miller, J. N., & Mosteller, F. (1989). How study design affects outcomes in comparisons of therapy. 1: Medical. Statistics in Medicine, 8, 441-454. Cook, T., & Campbell, D. (1979). Quasi—experimentation: Design and analysis issues for field settings. Boston: Houghton Mifflin. Darbes, L. A., Kennedy, G. E., Peersman, G., Zohrabyan, L., & Rutherford, G. W. (2002). Systematic review of H] V behavioral prevention research in Latinos. University of California San Francisco. Retrieved November 9, 2007, from the World Wide Web: http://hivinsite.ucsf.edu/InSite?page=kb-07-04-l 1 Flores, S. A., & Crepaz, N. f. t. H. P. R. S. T. (2004). Quality of study methods in individual- and group-level HIV intervention research: critical reporting elements. AIDS Education and Prevention, 16(4), 341-352. Greenland, S. (1994). Quality scores are useless and potentially misleading. American Journal of Epidemiology, I 40, 300-301. Herbst, J. H., Becker, C., Mathew, A., McNally, T., Passin, W. F., Kay, L. S., et al. (2007). The effectiveness of individual-, group-, and community-level HIV behavioral risk-reduction interventions for adult men who have sex with men: A systematic review. American Journal of Preventive Medicine, 32(4S), S3 8—S67. Huwiler-Mfintener, K., Jiini, P., Junker, C., & Egger, M. (2002). Quality of reporting of randomized trials as a measure of methodologic quality. Journal of A merican Medical Association, 28 7(21 ), 2801-2804. Jadad, A. R., Moore, R. A., Carroll, D., Jenkinson, C., Reynolds, D. J ., Gavaghan, D. J ., et al. (1996). Assessing the quality of reports of randomized clinical trials: is blidnin g necessary? Controlled Clinical Trials, 17, l-12. 124 mm, P., Altman, D. G., & Egger, M. (2001). Systematic reviews in health care: assessing the quality of controlled clinical trials. British Medical Journal, 323, 42-26. Jiini, P., Witschi, A., Bloch, R., & Egger, M. (1999). The hazards of scoring the quality of clinical trials for meta-analysis. The Journal of the American Medical Association, 282, 1054-1 060. Kjaergard, L. L., Villumsen, J., & Gluud, C. (2001). Reported methodologic quality and discrepancies between large and small randomized trials in meta-analyses. Annals of Internal Medicine, I 35, 982—989. Liberati, A., Himel, H. N., & Chalmers, T. C. (1986). A quality assessment of randomized control trials of primary treatment of breast cancer. Journal of Clinical Oncology, 4, 942-951. Lipsey, M. W., & Wilson, D. B. (2001). Practical Meta-analysis (Vol. 49). Thousand Oaks, California: Sage Publications Inc. Lohr, K. N. (2004). Rating the strength of scientific evidence: relevance for quality improvement programs. International Journal for Quality in Health Care, 16(1), 9- l8. Lyles, C. M., Crepaz, N., Herbst, J. H., Kay, L. S., & Team, f. t. H. A. P. R. S. (2006). Evidence-based HIV behavioral prevention from the perspective of the CDC's HIV/AIDS prevention research synthesis team. AIDS Education and Prevention, 18(Suppl. A), SZl-S31. Lyles, C. M., Kay, L. S., Crepaz, N., Herbst, J. H., Passin, W. F., Kim, A. S., Rama, S. M., Thadiparthi, S., DeI.uca, J. B., Mullins, M. M., & Team, f. t. H. A. P. R. S. (2007). Best-evidence interventions: findings from a systematic review of HIV behavioral interventions for US populations at high risk, 2000-2004. American Journal of Public Health, 97(1), 133-143. May, G. S., DeMets, D. L., Friedman, L. M., Furberg, C., & Passamani, E. (1981). The randomized clinical trial: bias in analysis. Circulation, 64, 669-673. Moher, D., Cook, D. J ., Eastwood, S., Olkin, 1., Rennie, D., & Stroup, D. F. (1999). Improving the quality of reports of meta-analyses of randomized controlled trials: the QUOROM statement. Lancet, 354, 1896-1900. Moher, D., Pham, B., Jones, A., Cook, D. J., Jadad, A. R., Moher, M., et al. (1998). Does quality of reports of randomized trials affect estimates of intervention efficacy reported in meta-analyses? Lancet, 352, 609-613. Moher, D., Schulz, K. F ., & Altman, D. G. (2001). The CONSORT statement: revised recommendations for improving the quality of reports of parallel group randomized trials. Retrieved April 15, 2008, from the World Wide Web: htLpd/wwwconsort- statementorgl. Moja, L. P., Telaro, E., D'Amico, R., Moschetti, 1., Coe, L., & Liberati, A. o. b. o. t. M. S. G. (2005). Assessment of methodological quality of primary studies by systematic reviews: results of the metaquality cross sectional study. British Medical Journal, 10,1136-1140. Montenegro, R., Needleman, I., Moles, D., & Tonetti. M. (2002). Quality of RCTs in Periodontology - a systematic review. Journal of Dental Research, 81(12), 866-870. Orwin, R. G., & Cordray, D. S. (1985). Effects of deficient reporting on meta-analysis: A conceptual framework and reanalysis. Psychological Bulletin, 97, 134-147. Patton, M. Q. (1978). Utilized-focused evaluation. Beverly Hills: Sage Publications, Inc. Reisch, J. 8., Tyson, J. E., & Mize, S. G. (1989). Aid to the evaluation of therapeutic studies. Pediatrics, 84, 815-827. Rossi, P. H., Lipsey, M. W., & Freeman, H. E. (2004). Evaluation: A systematic approach. Thousand Oaks: Sage Publications. Schrimshaw, E. W., Rosario, M., Meyer-Bahlburg, H. F. L., & ScharfMatlick, A. A. (2006). Test-retest reliability of self-reported sexual behavior, sexual orientation, and psychosexual milestones among, gay, lesbian, and bisexual youth. Archives of Sexual Behavior, 35(2), 225-234. Schroder, K. E. E., Carey, M. P., & Vanable, P. A. (2003). Methodological challenges in research on sexual behavior: II Accuracy of self-reports. Annals of Behavioral Medicine, 26, 104-123. Schulz, K. F ., Chalmers, l., Hayes, R. J ., & Altman, D. G. (1995). Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effect in controlled trial. The Journal of the American Medical Association, 273, 408-412. Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and Quasi- Experimental Designs for Generalized Causal Inference. Boston: Houghton-Mifflin. Sindhu, F ., Carpenter, L., & Seers, K. (1997). Development of a tool to rate the quality assessment of randomized controlled trials using a Delphi technique. Journal of Advanced Nursing, 25, 1262-1268. Soares, H., Daniels, S., Kumar, A., Clarke, M., Scott, C., Swarm, S., & Djulbegovic, B. (2004). Bad reporting does not mean bad methods for randmized trials: 126 observational study of randomized controlled trials performed by the Radiation Therapy Oncology Group. British Medical Journal, 328, 22-25. Stroup, D. F ., Berlin, J. A., Morton, S. C., et al. (2000). Meta-analysis of observational studies in epidemiology: a proposal for reporting meta-analysis of Observational studies in Epidemiology (MOOS) group. The Journal of the American Medical Association, 283, 2008-2012. Wingood, G. M., & DiClemente, R. J. (1996). HIV sexual risk reduction interventions for women: a review. American Journal of Preventive Medicine, 12(3), 209-217. Wortman, P. M. (1994). Judging research quality. In Cooper & Hedges (Eds), The Handbook of Research Synthesis (pp. 97-110). New York: Russell Sage Foundation. 127 CHAPTER 5 EFFECTIVENESS OF CULTURALLY SENSITIVE HIV/AIDS PREVENTION INTERVENTIONS IN THE US, 1996-2007: A META- ANALYSIS Introduction Most of the HIV/AIDS prevention interventions to date have a focus on changing people’s perception of preventive behavior. However, more recent interventions have included behavior skills training. Because culture may influence people’s perception of the cost and benefit of sexual behaviors, incorporation of cultural sensitivity into intervention treatment may fit into the perception-oriented framework of the existing interventions. For this reason, many HIV/AIDS risk behavior reduction interventions have incorporated either more or less cultural sensitivity into intervention components in the belief that people will be more responsive to intervention treatments which are contingent on their community subcultures. However, sexual behavior is more than a cultural product. It may also be social and economic. The relative importance of cultural, social, and economic factors on people’s decision making about safe sex behavior may be conditioned by people’s social positions. For example, gender roles in a specific culture may reinforce women’s submissive behavior in sex relationships, while women who are more economically independent from their sex partners may have more power in their relationship. In addition, decisions about safe sex behavior may depend on specific social contexts in which sex behavior takes place. People’s perception of the cost and benefit of specific safe sex behavior may be related to their evaluation of risk for HIV infection associated with specific types of sex partners, as well as other factors which are not necessarily 128 related to disease prevention, such as maintaining intimacy with steady partners. Due to the complexity of the factors involved in enacting safe sex behavior, investigating intervention effectiveness should take into consideration the factors associated with people’s social positions and other factors related to intervention components. This study is a meta-analysis of the effectiveness of culturally sensitive HIV/AIDS prevention interventions (CSHAPIs). It tested hypotheses developed in Chapter 2 by analyzing the long-term effectiveness of CSHAPIs in comparison with that of non-CSHAPIs. In addition, it tested the relationships between intervention effectiveness and people’s social positions (including memberships to cultural groups, racial/ethnic status, socioeconomic status, and gender status). Furthermore, it explored the relationships between intervention effectiveness, important intervention components (such as intervention objectives and culturally sensitive indicators), and study quality. It investigated intervention effectiveness on either increasing condom use or reducing unprotected sex. This study follows the suggestions from the QUOROM statement (Moher, Cook, Eastwood et al., 1999) in the report of study methodology and findings. Methodology Samples This study reviewed 50 culturally sensitive HIV/AIDS prevention intervention (CSHAPI) studies and 18 non-culturally sensitive HIV/AIDS prevention intervention (non-CSHAPI) studies that were gathered based on a set of inclusion and exclusion criteria, which have been documented elsewhere (the methodology section in Chapter 3). The difference between CSHAPls and non-CSHAPIs was that the non-CSHAPI studies 129 did not report any culturally sensitive indicator. Because non-CSHAPIs did not specifically incorporate cultural elements of the majority in the samples, the more-than- 70% rule in the criteria of identifying a cultural majority in the CSHAPI studies was not applied to the non-CSHAPI studies. Therefore, this review included the non-CSHAPI studies involving mixed racial/ethnic samples. Among the 68 studies, the meta-analysis analyzed 31 CSHAPI studies and 10 non-CSHAPI studies. Although the 68 studies provided empirical behavioral outcomes of either condom use, unprotected sex, or both, 16 studies did not provide sufficient information on which effect sizes can be calculated. In addition, 8 studies had large baseline differences between intervention and control/comparison groups, which were not taken into consideration in estimating intervention effectiveness. Because large baseline differences between intervention and comparison/control groups would confound comparison of odds at a post-test or follow-ups, effect sizes derided from odds ratio with large baseline differences (more than 5%) were excluded from the meta-analysis. In addition, 1 of the 68 studies did not specify the time of post-tests. One study did not report a pre-test for the specific behavioral outcome. One study was excluded because the estimation of intervention effectiveness highly outperformed all other studies under review (Mishra, Sanudo, and Conner, 2004). This study targeted Hispanic farmers in isolated farming camps where a choice of sexual partners was largely limited to sex workers. In terms of behavioral outcomes, 8 CSHAPI studies and 4 non-CSHAPI studies overlapped between the meta-analysis of condom use and unprotected sex. Twenty CSHAPI studies and 6 non-CSHAPI studies were included in the meta-analysis of 130 condom use. Nineteen CSHAPI studies and 8 non-CSHAPI studies were used to investigate unprotected sex. Dependent Variables Dependent variables included condom use and unprotected sex. Condom use by definition is protected sex. Unprotected sex is frequency of sex deducted by protected sex. Because these two behaviors may be highly correlated with each other, and they were not independent in a single study, any analysis involved only one of the behaviors. Studies using both behaviors as dependent variables would be used for analysis involving either dependent variable. Studies had used various measures for condom use or unprotected sex. For studies involving multiple measures for the same behavior constructs, this study used the best measure available for level of risk for HIV infection. The order of the best measures for condom use is as follows: percentage of condom use, consistent condom use (yes/no), frequency of condom use, numbers of days or weeks with protected sex, and condom use at last sex. The order of the measures for unprotected sex is the number of unprotected sex, percentage of unprotected sex, consistent unprotected sex (yes/no), and unprotected sex at last sex. The last two measures of unprotected sex, however, were not been widely used by researchers. Validity Assessment Validity assessment was part of the criteria for inclusion and for testing quality evaluation in Chapter 4. The studies in meta-analysis involved either randomized 131 controlled experiments or quasi-experiments with a pre-test and a post-test. Both designs may support a causal inference that behavioral changes of the participants were due to the intervention treatments they had received, if the interventions had been conducted properly (Cook and Campbell, 1979). In Chapter 4, I evaluated whether the 50 CSHAPI studies, from which the 31 CSHAPI studies in the meta-analysis were derived, had been conducted properly in terms of internal validity. In addition to assessments of the external validity of these studies, I also discussed internal validity of a meta-analysis based on these primary studies. The quality of study report was not an indicator of methodological quality (Huwiler-Mfintener, Jt'ini, Junker, and Egger, 2002). Estimation of effect size is not a standard practice in the research of HIV/AIDS prevention intervention, nor is it the only way to report intervention effectiveness. Therefore, a lack of any discussion on effect size is not a criterion of study quality assessed in Chapter 4. In addition, given that most of the 50 studies provided sufficient sample sizes and involved random assignments, a lack of report on sample size for each intervention condition is more related to the issue of differential attrition between intervention conditions than an issue of study report. The issue of differential attrition had been assessed in the domains of study analysis and result. Therefore, this study assumed that the pattern of collective study quality did not differ systematically between studies whether reporting the information on effect size or not. Relationships between important study quality indicators and intervention effectiveness were tested in sensitivity tests as part of the meta-analysis. 132 Data Abstraction The coding was entirely conducted by the author. As the author was also the one searching and retrieving eligible studies, blinding in the process of coding was not possible. However, the coding was less likely to be biased in favor of any funding sources in support of the primary studies, or with any intervention projects in interest of grant applications, due to a lack of the author’s affiliation with them. Multi-Follow-ups The studies chosen for conducting meta-analysis had different frequency of post- tests conducted at different time points after the interventions were completed. If a specific time interval is chosen, the number of follow-ups may vary from study to study. For this reason, when a given study started the first follow-up, yet another study may be at its second later follow-up. This study tested time effect on intervention effectiveness. In order to control for intervention effectiveness confounded by different time points of follow-ups across studies, the meta-analysis established four time points for evaluation of time effect on intervention effectiveness. It defined follow-ups conducted right after the interventions were completed as post-tests, follow-ups at 1, 2, 3, and 4 months after the interventions as Time 1, follow-ups at 5 and 6 months as Time 2, and follow-ups at 12 months or beyond as Time 3. In addition, some studies had conducted 9-month follow-ups as a second follow- up, while some other studies had included it as a third follow-up. In order to control for 133 dependency of effect sizes in a meta-analysis, 9-month follow-ups were excluded from the data analysis. Type of Effect Sizes This study represents effect sizes as an indicator of intervention effectiveness, which may quantify and standardize intervention effectiveness. It provides a statistical basis for combinations of intervention effectiveness across studies involving different behavioral measures of the same behavioral construct. Because most of the studies under review reported mean differences as a measure of intervention effectiveness, this study uses Cohen’s d” as a measure of effect sizes, which was obtained by calculating standardized gain score differences between intervention and control/comparison groups in each of the studies. Gain scores were calculated by subtracting the post-test mean from the pre-test mean. Divided by pooled standard deviations of the gain scores from both intervention and comparison/control groups, the gain scores were standardized and could be combined across studies involving different behavioral measures of the same behavioral constructs. Effect sizes, derived from standardized gain score differences between intervention and control groups, have taken into consideration baseline differences between groups. 35 . . . Cohen’s d= Gain score (1) — Gain score (2)/ Pooled SD of gain scores Gain score (1) = Group 1 mean difference Gain score (2) = Group 2 mean difference Pooled SD of gain scores = Sqr((((n(l) - l) * SD of gain score (1)2 + (n(2) - 1) * SD of gain score (2)2)/ (n(l)+ n(2) - 2») SD of gain score (I) = Sqr(SD at pre-test(1)2 + SD at post-test(l)2 - 2 * CorrPrePost * SD at pre-test( 1) * SD at post-test( l )) SD of gain score (2) = Sqr(SD at pre-test(2)2 + SD at post-test(2)2 - 2 * CorrPrePost * SD at pre-test(2) * SD at post-test(2)) SE ofd= Sqr(l / n(l) + 1 / n(2) + d2 / (2 * (n(l) + n(2)))) 134 In addition, for studies reporting odds ratios as a measure of intervention effectiveness, the odds ratios were transformed to Cohen’s din order to combine with effect sizes derived from standardized gain score differences. Odds ratios involving large baseline differences (more than 5%) were excluded from the meta-analysis as indicated earlier in the “Sample” section. Effect sizes were retrieved for condom use, unprotected sex, or both when information of effect sizes was available. This study uses Comprehensive Meta-analysis software36 for estimating effect sizes, pooling effect sizes, and conducting meta-analysis. Missing Data Frequently missing information in the reviewed studies includes a lack of reports on sample sizes for each intervention and control/comparison group and a lack of reports on reliability of behavioral outcomes. This study estimated reliability of behavioral outcomes based on Kalichman and Stevenson’s report (1997) on reliability of self- reported sexual behavior. They reported reliability of unprotected sex (r=0.78) and condom use (r=0.65). Data Analysis Combining Eflect Sizes." Types of Effect Sizes, Study Quality. and Sample Sizes Sensitivity tests were conducted to test variation between different types of effect sizes, such as effect sizes derived from odds ratios or standardized gain score differences, at post-test, Time 1. Time 2, and Time 3. This study found that effect sizes at each time 36 . http://www.meta-analysnscom/ 135 point did not significantly differ by types of effect sizes. Therefore, this study pooled together effect sizes derived from either standardized gain score differences or odds ratios, controlling for each time point. Furthermore, a sensitivity test ofCSHAPI effectiveness by accumulative study quality found that effect sizes at each time point did not vary significantly due to differences in accumulative study quality. Therefore, effect sizes were included in data analysis, regardless of their accumulative study quality. When pooling effect sizes, this study weighted each effect size by its inverse of variance, which is partially impacted by its sample size. Studies with a large sample size tend to have smaller variance and therefore, are more precise in estimating effect sizes. Therefore, studies with large sample sizes contributed more in the pooled effect sizes than other studies with small sample sizes. Variation ofIntervention Effectiveness Homogeneity tests were conducted to assess the extent of variation of intervention effectiveness by culturally sensitive indicators and other essential intervention components. A comparison of intervention effectiveness was made between CSHAPIs (yes) and non-CSHAPIs (no) in terms of whether the existence of culturally sensitive indicators may count for variation of intervention effectiveness. In addition, this study explored whether essential components of the CSHAPIs may have contributed to the heterogeneity in the CSHAPI effectiveness. The variables of intervention components included types of culturally sensitive indicators, intensity of cultural sensitivity. use of theory (yes/no), types of intervention objectives, types of 136 intervention activities, same attention (same level of intervention exposures to each intervention and control/comparison group, yes/no), and levels of intervention (individual/gmup/community). Types of culturally sensitive indicators included community participation, pride, gender roles, social responsibility, social experience, matching message, and matching facilitators. Refer to Chapter 3 for the definition of each culturally sensitive indicator. Intensity of cultural sensitivity was measured as the number of culturally sensitive indicators reported in a single intervention. Five types of intervention objectives were included: increasing awareness of susceptibility, motivation, intrapersonal competency (such as self-esteem and self- regulation), interpersonal skills (such as communication), interpersonal growth (such as community empowerment), promoting behavioral norms, and behavioral skills training (such as condom use). Intervention activities included skills demonstration and role demonstration. The former had a focus on practical skills of condom use, while the latter had more emphasis on the social relationships involved in safe sex behavior. In addition, levels of intervention exposures were calculated by frequency of the intervention sessions multiplied by the average duration (minutes) of each session. Levels of intervention included individual, group, and community. Variables of study quality included accumulative study quality and important quality indicators. Accumulative study quality was measured as total quality score combining quality scores from five domains of study reports. The important quality components included use of randomized assignment, allocation concealment, double blinding, monitoring mechanisms, and intention-to-treat analysis. Refer to the 137 methodology section of Chapter 4 for details on the definitions and measures of the accumulative study quality and the individual quality indicators. Subgroup Analysis This study conducted the Chi-square test and Fisher’s exact test to compare sample differences between CSHAPI and non-CSHAPI studies in terms of demographic characteristics of the participants. It also conducted subgroup analysis, using homogeneity tests, in order to examine whether demographic characteristics of the intervention participants may explain variation of effect sizes across the studies. Categorical demographic variables under assessment included cultural majorities in the samples and the participants’ aggregate age, sex, race/ethnic status, and socioeconomic status (SES). Point estimates of effect sizes for each subgroup were provided. This study also tested strength of the relationships between intervention effectiveness and various continuing demographic variables, such as age and years of education, in meta-regression analysis. This study used both fixed-effect and mixed-effect models in meta-analysis. It usually started with the fixed-effect model. When heterogeneity, in terms of effect sizes, was found within sub-category studies, the mixed-effect model may be used to supplement the analysis. Publication Bias This study involved only published studies which may be subject to publication bias. A publication bias is present when studies with significant findings were more 138 likely to be accepted for publication. Because studies with small sample sizes are more likely to have more variance, intervention effectiveness must be at a higher level in order to gain a significant result. Therefore, the study population of this meta—analysis may represent those having a higher level of intervention effectiveness. In order to assess the extent of publication bias, this study used Rosenthal’s approach (1979), which involves calculating a fail-safe number of missing studies which would bring a p value to more than 0.05. For example, if a fail-safe number is 100, it means that 100 studies are needed in order to reject the hypothesis that the magnitude of effect sizes is not equal to zero. This study also displayed precision funnel plots of effect sizes at different time points in order to show the distribution of effect sizes as a function of sample sizes. Findings Demographic Characteristics: Sample differences between CSHAPI and Non-CSHAPI studies Table 5.1 shows that CSHAPI studies were significantly different from non- CSHAPI studies in terms of cultural categories of the samples (X2=11.185, df=4, p=0.025). Two-by-two Fisher’s exact test shows that compared with non-CSHAPI studies, CSHAPI studies were significantly more likely to target African Americans as a cultural group (X2=6.093, df= l, p=0.014). 139 Table 5.] Demographic Characteristics of Participants by CSHAPI and non-CSHAPI studies All Studies CSHAPIs Non-CSHAPls . 3 Studies by cultural groups Drug user studies 18 14 4 77.8% 22.2% MSM studies 1 l 5 6 45.5% 54.5% African American studies* 19 18 l 94. 7% 5.3% Hispanic American studies 10 9 1 90% 10% White American studies 5 4 1 80% 20% General studies 5 0 5 27.8% . . . . 38 goal/ethnic majority White Americans* 14 7 7 50% 50% African Americans“ 32 29 3 90.6% 9.4% Hispanic Americans 10 9 I 90% 10% Asian and Pacific Islanders 1 1 0 I 0000 Mixed racial/ethnic groups* 1 I 4 7 36.4% 63.6% . . 39 Sex majority Male 20 '13 7 65% 35% Female 23 20 3 87% 13% Mixed gender group 23 15 8 65.2% 34.8% * indicates that the difference between the CSHAPI and non-CSHAPI studies was significant (p<0.05). 37 . . . . . - . Number and percentage of studies With their samples consrsting of more than 70% of a Single cultural roup. Refer to Chapter 2 for the definition of cultural groups. 8 . . . . . . Number and percentage of studies With their samples consrsting of more than 70% of a Single racial/ethnic group. 9 . . . . Number and percentage of studies With a gender group consrsting of more than 70% of the samples. Percentage was estimated based on available studies reporting gender composition in the samples. 140 Table 5.2 Aggregate Estimate of Age, Socioeconomic Status, and Marital Status of Participants by CSHAPI and Non-CSHAPI Studies 40 Age ( mean 1 Proportion of males Employment . 41 part-time (%) Family income< $10,000/year (%) Education Years (mean)* Marital Status Married or steady relationships (%) Separated Retention Unemployed/ employed High school or less (%) Single, never married (%) /divorced/widowed (%) CSHAPIs Non- T-test (If p CSHAPIs 28.52 28.17 -.137 56 .892 k=41 k=17 SD=8.744 SD=9.496 44.6% 64.3% 1.793 64 0.052 k=48 k=18 SD=0.419 SD=0.327 80.8% 68.3% 16 k43=i7 k=1 SD=0.122 63.9% 54% 10 k=ll k=l SD=0.22I 10.9 15.4 3.301 19 .0004 k=19 k=2 SD=I .87 SD=0.134 60.6% 37.8% -1 .792 18 0.09 k=15, k=5 SD=0.244 SD=0.255 55.7% 58.5 0.186 15 .955 k=13 k=4 SD=0.224 SD=0.379 25.5% 36.3 0.776 17 .449 k=17 k=2 SD=0.192 SD=0.023 21.1% 37.3 1.841 8 .103 k=8 k=2 SD=0.] 17 SD=0.061 76% 76% 0.005 56 .996 k=4] k=17 SD=0.152 SD=0.133 * indicates that the difference between the CSHAPI and non-CSHAPI studies was significant (p<0.05). 40 . . . . . Median of age was estimated based on aggregate data of available studies reporting mean age of the samples. 41 . . . Mean of the proportion reported by available studies. ’2 Number of the studies provides the information of unemployment or part-time employment. 141 In terms of racial/ethnic status, Table 5.2 shows that racial/ethnic status of the samples varied between CSHAPIs and non-CSHAPIs (X2=18.347, df=4, p=0.001). Two- by-two Fisher’s exact test indicates that compared with non-CSHAPI studies, CSHAPI studies were more likely to target African Americans as a major racial/ethnic group in the samples (X2=9.076, dfil , p=0.003). White Americans and mixed-racial/ethnic samples were also significantly related to CSHAPIs. There was no significant difference between CSHAPI and non-CSHAPI studies in terms of the number of Hispanic American studies. In addition, Table 5.2 shows that in a two-tailed t-test, samples in the CSHAPIs tended to have a higher level of education by years (t= 3.301, df=l9, mean difference=4.47, C [=1 .63-7.3). There were no other significant differences in aggregate demographic characteristics between CSHAPI and non-CSHAPI studies. Further analysis shows that there was no difference in terms of type of experimental design (RCT/Quasi-experiment) and attrition rate between CSHAPI and non-CSHAPI studies. However, the CSHAPI studies were more likely to be conducted at a group level and were more likely to make same-attention comparisons between intervention and control/comparison groups than non-CSHAPI studies. Comparison of Effectiveness of C SHAPIs with Non-CSHAPIs Unprotected Sex Table 5.3 shows that in general, the magnitude of intervention effectiveness on reducing unprotected sex behavior ranged from a moderate to a small level at different time points The effect sizes tended to decrease over time. There was a moderate reduction of effectiveness from posttest (d=-0.27, p<0.005, k=6) to Time 1 (d=-0. l 7, 142 p<0.005, k=17) and a slight decrease to Time 2 (d=-0.13, p<0.005, k=11). However, there was a slight increase of intervention effectiveness one year after the interventions were completed (d=-0.16, p<0.005, k=10, Time 3). In a longitudinal perspective, the pattern of effectiveness of non-CSHAPIs was similar to that of all interventions. The intervention effectiveness was strongest at post- test (d=-0.3 7, p<0.005, k=3), moderately decreased at Time 1 (d=-0.22, p<0.005, k=4) and at Time 2 (d=0.l4, p=0.16, k=2), yet slightly improved at Time 3 (d=-0.13,p=0.21, k=2). However, the intervention effectiveness at Time 3 was still much smaller than that at post-test. In contrast, changes of the effectiveness of CSHAPIs over time showed an opposite pattern to that of all interventions and of non-CSHAPIS. In general, the effectiveness of CSHAPIs tended to be steady but increased gradually in the long term. It increased over time from post-test (d=-0.14, p=0.14, k=3), to Time 1 (d=-0.16, p<0.005, k=1 3), and to Time 2 (d=-0.l 8, p<0.005, k=9). There was a slight reduction of effectiveness one year after the interventions were completed (d=-0.17, p<0.005, k=8, Time 3). However, the effectiveness at Time 3 exceeded that at post-test. Furthermore, meta-regression analysis showed that the regression coefficients of the relationships between cultural sensitivity and intervention effectiveness were positive at post-tests, but became negative at later follow-ups. It indicated an increasing rate of the effectiveness of CSHAPIs over that of non-CSHAPIS in the long term, although the regression coefficients were not significant at post-test, Time 1, and Time 3. However, since this meta-analysis gathered most of the published studies in the study population, the differences between CSHAPIs and non-CSHAPIs were evident in the study population 143 even if the findings may not be generalizable to a larger study population. Besides, the meta-analysis at post-test and Time 3 was limited by the small number of studies available for data analysis. Meta-analysis with a small number of studies is less likely to gain a significant result. The advantage of CSHAPIs over non-CSHAPIs was most evident and generalizable at Time 2. In a comparison of the effectiveness of CSHAPIs to that of non- CSHAPIs at Time 2, a large Qw value (1 7.85, p=0.058, k=11) in a homogeneity test indicated heterogeneity among the effect sizes. Furthermore, a large between-group Qb value (9.21, p=0.002) and small within-group Qw values (6.21, p= 0.62 for CSHAPIs; 2.43, p=0.12 for non-CSHAPIS) indicated dissimilarity of effect sizes between CSHAPIs and non-CSHAPIs A meta-regression analysis showed that at Time 2, CSHAPIs were more effective than non-CSHAPIs in terms of reducing unprotected sex (B=-0.31, p<0.005, 02:0.00, k=10), while Tau-squares showed that there was no study-level random effect on intervention effectiveness (0'2 =0.005) at post-test, Time 1, and Time 2. 144 Table 5.3 Comparison of the Effectiveness of CSHAPIs with non-CSHAPIs on Reducing Unprotected Sex Behavior Post-test All CS Non-CS CSvs Non-CS -_All C8 Non-CS CSvs Non-CS —All CS Non-CS CSvs Non-CS _AII cs Non-CS CSvs Non-CS Summary Homogeneity Homogeneity ES within group betweeLgroups k d C I Q,» % of Qb Regression B pvalue pvalue variance pvalue value c variance 6 -0.27 (-0.39)-(-0. 14) 4.67 0.00 p<0.005 p=0.46 3 —0. 14 (j-0.33)-(0.05) 1.27 ' 0.00 p=0.14 p=0.53 3 -0.37 (-0.54)-(-O.2) 0.23 0.00 p<0.005 p=0.89 6 3.2 0.23 p=0.08 p=0.08 02:0.00 17 -0.17 (-0.24)-(-0.ll) 7.84 0.00 p<0.005 p=0.95 13 -0. 16 (-0.23)-(-0.09) 4.59 0.00 p<0.005 p=0.97 4 -0.22 (-0.38)-(-0.06) 2.87 0.00 p<0.005 p=0.4l 17 0.39 0.06 p=0.53 p=0.53 02:0.00 I l -0.13 (-0.2)-(-0.05) 17.85 43.98 p<0.005 p=0.06 9 -0.18 (g-0.26)-(-0.1) 6.21 0.00 p<0.005 p=0.62 2 0.14 (-0.05)-0.3l 2.43 58.85 p=0.16 p=0.12 11 2.67 -0.31 p<0.00 p<0.005 02=0.00 10 -0.16 (-0.25)-(-0.08) 11.22 19.78 p<0.005 p=0.26 8 -0. l 7 (-0.26)-(-0.08) 11.1 36.92 ' p<0.005 p=0.13 2 -0.13 (-0.34)-(0.07) 0.03 0.00 [7:021 p=0.87 10 0.09 -0.04 p<0.76 p=0.76 02:0.01 145 Table 5.4 Comparison of the Effectiveness of CSHAPIs with Non-CSHAPIS on Enhancing Condom Use Behavior Post-test All CS Non-C S CSvs Non-CS 7n cs Non-CS CS vs. Non-CS 711 CS Non-CS CSvs Non-C S T_3 All CS N on-C S CSvs Non-C S Summary Homogeneity Homogeneity ES within group between groups k d CI Qw % of Qb Regression B pvalue pvalue variance pvalue value 0 variance 8 0.38 0.26-0.50 15.06 53.51 p<0.005 p=0.04 5 0.3 0.17-0.46 4.34 7.78 p<0.005 p=0.36 3 0.52 0.31-0.72 8.18 75.56 p<0.005 p=0.02 8 2.54 -0.2 p=0.11 p=0.11 02:0.04 16 0.26 0.18-0.34 24.67 39.2 p<0.005 p=0.06 12 0.27 0.18-0.35 18.28 39.82 p<0.005 p=0.08 4 0.22 0.02-0.43 6.26 52.08 p=0.04 p=0.l 16 0.13 0.04 p=0.72 p=0.7l 02:0.02 10 0.22 0.13-0.3 13.33 32.49 p<0.005 p=0.15 10 0.22 0.13-0.3 13.33 32.49 p<0.005 p=0.15 0 10 6 0.32 0.21-0.43 14.75 66.11 p<0.005 p=0.01 5 0.34 0.22-0.45 13.95 71.32 p<0.005 p=0.01 1 0.16 (-0.l9)- p=0.37 0.52 6 0.8 0.17 [2740.37 p=0.37 02:0.05 146 Condom use In general, the long-term changes of intervention effectiveness on condom use were similar to that of unprotected sex, while the former was more effective than the latter by about 60% (Cohen’s d for condom use deducted by d for unprotected sex and then divided by d of unprotected sex). The intervention effectiveness on enhancing condom use tended to decrease over time. The magnitude of the effect sizes on enhancing condom use decreased from post-test (d=0.3 8, p<0.005) to Time] (d=0.26, p<0.005) by 40% and to Time 2 (d=0.22, p<0.005) by 15%, and increased at Time 3 (d=0.32, p<0.005) by 45%. However, the effect size at Time 3 was still smaller than that at post-test. In general, both of the non-CSHAPI and CSHAPI effectiveness tended to decrease over time until Time 2 with non-CSHAPIS having a higher decreasing rate. While CSHAPI effectiveness decreased by 10% at Time I, non-CSHAPI effectiveness decreased by 58%. At Time 2, the CSHAPI effectiveness decreased from Time 1 by 19%. There was non-CSHAPI study under review at Time 2. Furthermore, the CSHAPI effectiveness increased one year afier the interventions were completed (Time 3) by 55%, while no non-CSHAPI effectiveness decreased again from Time 1 to Time 3 by 27%. Meta-regress analyses of the relationships between cultural sensitivity and intervention effectiveness at different time points also indicated an increase of the effectiveness of CSHAPIS over that of non-CSHAPIs. The regression coefficient changed its negative direction to a positive one with an increasing magnitude (B=-0.2, p=0.11 at post-test; B=0.04, p=0.71 at Time 1; [#017, p=0.37 at Time 3). 147 Compared with unprotected sex, effect sizes of enhancing condom use were more heterogeneous as indicated by significant Qw values at post-test (Qw=15.06, p=0.04), Time 1 (Q1 £24.67, p= 0.055), and Time 3 (Qw=14.75, p= 0.01). Moreover, there were small Qb values at post-test, Time 1, and Time 3 while no non-CSHAPI studies were available at Time 2 for a homogeneity test. The small Qb values at different time points indicated that cultural sensitivity was not a sufficient factor to explain the variance of effect sizes at each time point. In addition, tau—square in meta-regression analyses pointed out an existence of study-level random effect on intervention effectiveness at three over four time points Therefore, this study explored other study factors that may explain variation in CSHAPI effectiveness especially on increasing condom use. Table 5.5 consists of the findings of mixed-effect and fixed-effect analyses on the effectiveness of C SHAPIs at different time points. When effect sizes were heterogeneous as indicated by percentage of variance, mixed-effect analyses were more precise in estimating pooled effect sizes. Mixed-effect analysis resulted in an increase in the magnitude of the pooled effect sizes. Table 5.6 shows the effect size of each study included in the meta-analysis at different time points. 148 Table 5.5 Effectiveness of CSHAPIs by Time Points and Dependent Variables Unprotected sex Condom use Summary Homogeneity Summary Homogeneity of ES within amp of ES within group k (1 Cl Qw % of k d CI Qw % of p value vari. p value vari. Fixed Posttest 3 -0.14 (-0.33)- 1.27 0.00 5 0.3 0.17- 4.34 7.77 (0.05) p=0.53 0.46 p=0.36 T1 13 -0.16 (-0.23)- 4.59 0.00 12 0.27 0.18- 18.28 39.82 (-0.09) p=0.97 0.35 p=0.08 T2 9 -0.18 (-0.26)- 6.21 0.00 10 0.22 0.13- 13.33 32.49 (-0.1) p=0.62 0.3 p=0.15 T3 8 -0.17 (-0.26)- 11.1 36.92 5 0.34 0.22- 13.95 71.32 (-0.08) p=0.13 0.45 p=0.01 Mixed Posttest 3 -0.14 (-0.33)- 5 0.33 0.17- (0.05) 0.48 T1 13 -0.16 (-0.23)- 12 0.29 0.17- (-0.09) 0.4 T2 9 -0.18 (-0.26)- 10 0.23 0.13- (-0.1) 0.34 T3 8 -0.19 (-0.3)- 5 0.38 0.15- (-0.07) 0.61 149 Table 5.6 Effect Sizes (Cohen’s d) of Culturally Sensitive Interventions by Time Point Condom Unprotected Sex Study Post- T1 T2 T3 Post- T1 T2 T3 test test Sterk (2003) 0.42 Harris (1998) 0.48 0.57 0.36 Wechsberg (2004) -0.24 -0.23 McCoy (1998) 0.23 Latkin (2003) 0.79 Cottler (1998) <0.005 Nyamathi (1998) <-0.005 -0. 16 McMahon (2001) -0.45 -0.49 Hershberger (2003) 0.20 Avants (2004) <-.005 Choi (1996) -0.14 Rosser (2002) -0.09 -0.36 Kelly (1997) 0.52 -0.28 Kegeles (1996) -0.24 Koniak-Griffin (2003) -0.24 -0.13 -0.01 Peragallo (2005) 0.33 0.17 Raj (2001) 0.51 0.80 Harvey (2004) 0.39 -0.23 Mishra (1996) 0.65 Kalichman (1996) 0.41 -0.01 Kalichman (1999) 0.37 0.01 -0.13 -0.14 Carey (1997) 0.29 <0.005 -0.26 -0.1 1 Carey (2000) 0.33 Jemmott (1999) -0.01 -0.20 Jemmott Ill (1998) <0.005 0.01 0.01 -0.01 -0.01 <-0.005 Robinson (2002) -0.15 Nyamathi (1997) -0.25 Stanton (1996) 0.70 0.21 DiClemente (2004) 0.36 Jemmott (2007) 0.35 <0.005 0.71 -0.33 -0.19 -0.24 Kalichman (2005) 0.34 0.37 -0.13 -0.29 Finally, both of the pooled effect sizes of condom use and unprotected sex pointed out the advantage of C SHAPls in the long term. Figure 5.1 and Figure 5.2 depicted the patterns of longitudinal changes of CSHAPI effectiveness on different dependent 150 variables in comparison with non-CSHAPIs. Although non-CSHAPIS were more effective than CSHAPIS at posttest, the advantages tended to attenuate over time. In contrast, the effectiveness of CSHAPI increased (for reducing unprotected sex) or at least remained at similar levels (for enhancing condom use). The effectiveness of CSHAPIs was evident one year after the interventions were completed when the magnitude of the pooled effect sizes at Time 3 became larger than those of post-tests for dependent variables. Figure 5.1 Intervention Effectiveness by Cultural Sensitivity and Time Points: Unprotected Sex 0.20 0.10 ' 0.00 -0.10 Effect Size -I— CSHAPI .‘1 . -—-I-— - PI -0.2o Nflm -0.30 ‘ .. -o.40 7‘ 11m Point 151 Figure 5.2 Intervention Effectiveness by Cultural Sensitivity and Time Points: Condom Use 0.60 0.50 0.40 . , .‘s’ 9 0 30 -I-CSHAP1 E +Non-CSHAPI H 0.20 ‘ 0.10 :I' l O-m .Lw. 1.., V I” Posttest T1 T2 T3 'fime Point The Effectiveness of CSHAPIs by Subgroup Differences Due to the long-term changes in CSHAPI effectiveness as discussed above, this section focuses on Time 1 and Time 3, which involve somewhat different sets of studies. Table 5.7 and Table 5.8 show that either cultural category or race/ethnicity of the intervention participants was not significantly related to intervention effectiveness for both dependent variables at Time 1. At Time 3, race/ethnicity was a predictor of intervention effectiveness on reducing unprotected sex (Qb =8.1, p=0.04, k=5). Studies involving White Americans (d=-0.3, p<0.005, k=2) or mixed racial/ethnic samples (d=- 152 0.49, p<0.005, k=1) were more effective than African American studies (d=-0.l4, p=0.02, k=l) or Hispanic American studies (d=-0.01, p=0.94, k=l). The White American studies were MSM studies. The mixed racial/ethnic samples were drug user studies. It appears that in general, racial/ethnic status had more impact on distribution of intervention effectiveness than memberships to subcultural groups. Table 5.7 CSHAPI Effectiveness on Enhancing Condom Use Behavior: a Subgroup Analysis * indicates significant within-group heterogeneity (p<0.05) Time 1 Time 3 Homogeneity Summary Homogeneity Summary between Subgroup ES between Subgroup ES groups groups Qb k d C] k Qb k d CI k p value (df) p value I value (d/) p value Cultural 3.25 12 1.36 5 Categom p=0.2 (2) p=0.24 (1) AA 0.21 0.09- 7 0.31 0.19- 4* p<0.005 0.33 p<0.005 0.43 DU 0.22 0.01- 2* 0 p=0.04 0.43 HA 0.38 0.23- 3 0 p<0.005 0.53 MSM 0.52 0.19- 1 p<0.005 0.85 Ra_ce/ 18.28 12 1.36 5 Ethnicity p=0.07 (2) p=0.24 (1) AA 0.21 0.11- 9 0.31 0.19- 4* p<0.005 0.31 p<0.005 0.43 HA 0.38 0.23- 3 0 p<0.005 0.53 WA 0.52 0.19- 1 p<0.005 0.85 g 8.1 1 12 9.87 5 p=0.02 (2) p=0.01 (2) Female 0.37 0.25- 8 0.46 0.30- 2* p<0.005 0.48 p<0.005 0.62 Male 0.37 (-0.03)- 1 0.52 0.19- 1 p=0.07 0.77 p<0.005 0.85 Mixed 0.12 (-0.02)- 3 0.1 (-0.08)- 2 p=0.09 0.25 p=0.29 0.29 153 Table 5.8 CSHAPI Effectiveness on Reducing Unprotected Sex Behavior: a Subgroup Analysis Time 1 Time 3 Homogeneity Summary Homogeneity Summary between groups Subgroup ES between Subgroup ES groups Qb k (1 CI k Qb k (1 CI p value (dj) p value I value (df) p value Cultural 1.93 13 6.23 8 Category p=0.59 (3) p=0.1 (3) AA -0.13 (-0.22)- 8 -0.14 (-0.27)- p=0.01 (-0.34) p=0.04 (-0.01) DU -0.24 (-0.47)- 1 -0.33 (-0.57)- p=0.05 (-0.00) p=0.01 (-0.09) HA -0.24 (-0.40)- 2 -0.01 (-0.20)- p<0.005 (-0.08) p=0.94 (0. 1 d8) MSM -0. 12 (-0.37)- 2 -0.30 (-0.49)- p=0.34 (-0.13) p<0.005 (-0.10) m 1.25 13 8.1 8 Ethnicity p=0.74 (3) p=0.04 (3) AA -0.14 (-0.23)- 9 -0.14 (-0.27)- p<0.005 (-0.06) p=0.02 (-0.02) HA -0.24 (-0.40)- 2 -0.01 (-0.20)- p<0.005 (-0.08) p=0.94 (0.18) WA -0.09 (-0.53)- l -0.30 (-0.49)- p=0.35 (-0.35) p<0.005 (-0.01) Asian A -0.14 (-0.44)- I p=0.37 (-0.16) Mixed -0.49 (-0.26)- p<0.005 (-0.08) SQ 2.97 13 6.87 8 p=0.23 (2) p=0.03 (2) Female -0.23 (-0.33)- 7 -0.14 (-0.26)- p<0.005 (-0. 12) p=0.03 (-0.02) Male -0. 12 (-0.34)- 3 -0.35 (-0.52)- p=0.25 (-0.09) p<0.005 (-0. l 8) Mixed -0.09 (-0.21)- 3 -0.02 (-0.21)- p=0.13 (-0.03) p=0.88 (0.18) * indicates significant within-group heterogeneity (p<0.05) In addition, at Time 1, the studies involving a sex majority were more effective than others with mixed-sex samples for both dependent variables. The relationship was significant only for condom use. However, it was significant at Time 3 for both dependent variables. This contrast is most evident for condom use (d=0.l for mixed-sex samples, d=0.46 for female samples, d=0.52 for male samples). 154 Table 5.9 Meta-regression on Relationships between Age, SES, and CSHAPI Effectiveness on Enhancing Condom Use at Time 1 and Time 3 Time 1 Time 3 Fixed Mixed Fixed Mixed k B 0'2 B (If k B 0'2 B 02 pvalue 1 (If pvalue 1 df pvalue 1 df pvalue 1 df Age 1 1 0.00 0.02 0.00 0.02 5 0.03 0.0 0.03 0.0 p=0. 16 p=0.45 p<0.005 0 p<0.005 0 Income 5 -0.03 0.02 0.05 0.02 0 <= p=0.94 p=0.93 10,000 Not full- 3 -2.34 0.03 -1.83 0.03 0 time p=0.13 p=0.47 employed Education/ 5 0.08 0.00 0.08 0.00 0 years p=0.7 p=0.7 Education/ 3 0.23 0.05 0.18 0.05 0 less than p=0.52 p=0.82 coHege Table 5.10 Meta-regression on Relationships between Age, SES, and Effectiveness of Culturally Sensitive Interventions on Reducing Unprotected Sex Behavior at Time 1 and Time 3 Time 1 Time 3 Fixed Mixed Fixed Mixed k B 02 B 0’2 k B 0'2 I3 02 p value 1 df p value 1 df p value 1 df p value 1 df Age 1 1 -0.00 0.00 -0.00 0.00 7 -0.01 0.00 -0.01 0.00 p=0.47 p=0.47 p<0.005 p<0.005 1ncome<= 4 0.13 0.00 0.13 0.00 0 10,000 p=0.78 p=0.78 Not full- 3 -0.33 0.00 -0.33 0.00 0 time p=0.68 p=0.68 employed Education/ 6 -0.03 0 00 -0.03 0.00 3 -0.18 0.00 -0. 18 0.00 years p:0.32 p=0.32 p=0.02 p=0.02 Education/ 4 -0.37 0.00 -0.37 0.00 3 0.14 0.00 0.14 0.00 less than p=0.49 p=0.49 p=0.68 p=0.68 coHege 155 Table 5.9 and Table 5.10 contain information of meta-regression of CSHAPI effectiveness on age and SES. Age had very minimal influence on intervention effectiveness on both dependent variables (B=0.03, p<0.005 for condom use, B=-0.00, p=0.00 for unprotected sex). Education by years was positively related to intervention effectiveness on both dependent variables at Time 1, however, at a weak and insignificant level. At Time 3, this relationship was still weak although it was significant (B=-0.18, p=0.02). All other SES indicators did not have consistent relationships with the two dependent variables, and the relationships were not significant. For example, family income less than $10,000 per year was positively related to intervention effectiveness on increasing condom use in mixed-effect analysis, while it was negatively related to intervention effectiveness for reducing unprotected sex. In contrast, less-than-full-time employment was positively related to intervention effectiveness on reducing unprotected sex, but was negatively associated with intervention effectiveness on increasing condom use. Table 5.11 shows that all of the culturally sensitive indicators were positively related to CSHAPI effectiveness for increasing condom use, except for self-pride at Time 1 and Time 3, matching facilitator at Time 1, and community participation at Time 3. Self-pride was the only culturally sensitive indicator significantly associated with less CSHAPI effectiveness for increasing condom use at both time points. Furthermore, the relationships between intervention effectiveness and community participation or matching facilitators were different between Time 1 and Time 3. Community participation had a positive relationship with intervention effectiveness at Time 1, but 156 was negatively related to increasing condom use at Time 3. In contrast, matching facilitators was negatively associated with an increase of condom use at Time 1, but was positively related to intervention effectiveness for increasing condom use at Time 3. In addition, cultural intensity (the number of culturally sensitive indicators) had minimal impact on CSHAPI effectiveness on increasing condom use. At Time 1, cultural intensity was slightly related to more intervention effectiveness (0:0.05, p=0.28 in mixed-effect model). At Time 3, cultural intensity was slightly related to decreasing condom use(B=-0.12, p=0.36 in mixed-effect model). However, neither relationships were significant. The relationships between culturally sensitive indicators and intervention effectiveness were similar for both dependent variables. For example, this study found a persistently negative impact of (ethnic) pride on both dependent variables (significant only for condom use) at both time points. The relationships between several culturally sensitive indicators and intervention effectiveness on reducing unprotected sex varied over time. For example, gender roles had a positive relationship with intervention effectiveness on reducing unprotected sex in the long term. Table 5.12 shows that the relationship between gender roles and intervention effectiveness on reducing unprotected sex was negative and insignificant at Time 1, but was positive and significant at Time 3. Similarly, matching facilitators had a negative relationship with intervention effectiveness at Time 1, but the relationship became positive by Time 3. In contrast, community participation was negatively related to intervention effectiveness at Time 3 for both behaviors while it had a positive relationship with intervention effectiveness at Time 1. 157 Table 5.11 Culturally Sensitive Indicators and CSHAPI Effectiveness on Increasing Condom Use Time] Time3 Qb Summary Qb Summary p value Subgroup ES p value Subgroup ES K=12 (1 Cl % of k =5 d CI % of dfil p value vari. dfil p value vari. Community 0.02 1.36 participation p=0.88 p=0.24 Yes 0.27 0.08- 38.58 8 0.31 0.19- 76.17 p<0.005 0.42 p<0.005 0.43 * No 0.25 0.17- 56.24 4 0.52 0.19- 0.00 p<0.005 0.37 p<0.005 0.85 Pride 3.71 7.53 p=0.05 p=0.01 Yes 0.13 (-0.04)- 34.05 3 0.23 0.1- 74.81 p=0.14 0.29 p<0.005 0.37 * No 0.31 0.22- 30.65 9 0.57 0.37- 18.43 p<0.005 0.41 p<0.005 0.78 Gender role 2.31 p=0.12 Yes 0.48 0.19- 0.00 3 p<0.005 0.77 No 0.25 0.16- 44.89 9 0.34 0.22- 71.32 p<0.005 0.33 p<0.005 0.45 * Social 3.04 responsibilit p=0.08 Y. Yes 0.53 0.22- 0.00 2 p<0.005 0.83 No 0.24 0.16- 40.14 1 0.34 0.22- 71.32 p<0.005 0.33 0 p<0.005 0.45 * Social 5.35 Experience p=0.02 Yes 0.53 0.29- 0.00 3 p<0.005 0.78 No 0.23 0.14- 30.4 9 0.34 0.22- 71.32 p<0.005 0.32 p<0.005 0.45 * Matching 2.67 0.18 facilitators p=0.1 p=0.67 Yes 0.22 0.13- 37.05 1 0.34 0.23- 78.21 p<0.005 0.32 0 p<0.005 0.45 * No 0.38 0.22- 23.51 2 0.21 (- 0.00 p<0.005 0.55 p=0.47 0.36)- 0.79 Matching 3.87 12 message p=0.05 Yes 0.38 0.24- 0.00 p<0.005 0.52 No 0.2 0.1-0.31 42.87 0.34 0.22- 71.32 p<0.005 p<0.005 0.45 * *indicates significant within-group heterogeneity (p<0.05) 158 Table 5.12 Culturally Sensitive Indicator and CSHAPI Effectiveness on Reducing Unprotected Sex Timel Time3 Qb Summary Qb Summary p value Subgoup ES p value Subgroup ES K=l3 d CI % k K=8 (1 CI % k df=1 p value var df-=1 p value var Communig 0.13 0.01 participation p=0.7] p=0.93 Yes -0.15 -(0.24)- 0.0 9 -0.‘16 (-0.31)- 67.97 3 p<0.005 (-0.07) 0 p=0.03 (-0.02) * No -0.18 -(0.33)- 0.0 4 -0.17 (-0.28)- 17.46 5 p=0.01 (-0.04) 0 p<0.005 (-0.06) Pride 0.69 3.02 p=0.4l p=0.08 Yes -0. 12 (-0.24)- 0.0 5 -0.02 (-0.21)- 0.00 1 p=0.07 (-0.01) 0 p=0.88 (0.18) No -0.18 (-0.27)- 0.0 8 -0.21 (-0.31)- 25.72 7 p<0.005 (-0. 1) 0 p<0.005 (-0.1 1) Gender role 0.06 3.87 p=0.8 p=0.05 Yes -0.14 (-0.34)- 0.0 3 -0.49 (-0.82)- 0.00 1 p=0.2 (-0.07) 0 p<0.005 (-0.16) No -0.17 (-0.24)- 0.0 1 -0.15 (-0.24)- 17.02 7 p<0.005 (-o.09) 0 0 p<0.005 (-0.05) Social 0.59 3.46 responsib- p=0.44 p=0.06 j|_it_y Yes -0.23 (-0.41)- 0.0 2 -0.01 (-0.20)- 0.00 1 p=0.02 (-0.04) 0 p=0.94 (-0.18) No -0.15 (-0.23)- 0.0 I -0.21 (-0.31)- 21.44 7 p<0.005 (-0.07) 0 1 p<0.005 (-0.1 1) Social 1.18_ 0.33 Experience p=0.28 p=0.57 Yes -0.23 (-O.37)- 0.0 3 -0.13 (-0.29)- 83.56 2 p<0.005 (-0.09) 0 p=0.13 (0.04) * No -0.14 (-0.22)- 0.0 1 -0.19 (-0.29)- 0.00 6 p<0.005 (-0.05) 0 0 p<0.005 (-0.08) Matching 1.45 0.33 facilitators p=0.23 p=0.57 Yes -0.14 (-0.22)- 0.0 l -0.19 (-0.29)- 0.00 6 p<0.005 (-0.05) 0 1 p<0.005 (-0.08) No -0.24 (-0.39)- 0.0 2 -0.13 (-0.29)- 83.56 2 p<0.005 (-0.09) 0 p=0.13 (-0.04) Matching 0.39 message p=0.53 Yes -0.2 (-0.34)- 0.0 4 0 p<0.005 (-0.06) 0 No -0.15 (-0.23)- 0.0 9 -0.17 (-0.26)- 36.92 8 p<0.005 (-0.06) 0 p<0.005 (-0.08) *indicates significant within-group heterogeneity (p<0.05) 159 Furthermore, the long-term impact of social experience and social responsibility was not conclusive in this study. Social experience and social responsibility were positively related to intervention effectiveness for both dependent variables at Time 1, but was negatively associated with intervention effectiveness on reducing unprotected sex at Time 3. Matching message was positively related to intervention effectiveness at Time 1 for both behaviors. Its impact at Time 3 was, however, not tested. Similar to CSHAPI effectiveness on condom use, cultural intensity in intervention treatment had little relationship with intervention effectiveness on reducing unprotected sex at Time 1(B=0.00, p=0.93, tau-square=0.00 in mixed-effect model). However, at Time 3, more cultural intensity was slightly related to less reduction in unprotected sex (B=-0.12, p=0.36, tau-square=0.04 in mixed-effect model). The negative impact of cultural intensity on reducing unprotected sex at Time 3 was similar to that of condom use. Table 5.13 shows that the relationships between many intervention objectives and activities and intervention effectiveness on condom use were in the same direction at both time points. However, at Time 3, these relationships became significant. Among them, interpersonal skills, norm, and behavioral skills training were positively related to more condom use, while intrapersonal competency was negatively related to more condom use. Furthermore, interpersonal growth was the only intervention component having a significantly negative relationship with condom use at both time points, while skills demonstration was the only one having a significantly positive relationship at both time points. 160 In addition, this study found that intervention intensity was not associated with intervention effectiveness on increasing condom use at both time points. The regression coefficients ranged from 0.01 at Time 1 (p=0.49, 02=0.03, k=10) to smaller than 0.005 at Time 3 (p=0.58, 02:0.11, k=5). Similarly, total time span of intervention was not related to intervention effectiveness. Its regression coefficients ranged from smaller than 0.005 at Time 1 (p=0.34, 02:003. #10) to smaller than -0005 at Time 3 (p=0.85, 02:007. k=5). Therefore, in general, there were more intervention objectives that had a significant relationship with intervention effectiveness on increasing condom use than culturally sensitive indicators. Intervention objectives appear to play an important role in determining C SHAPI effectiveness, not less than culturally sensitive indicators. Table 5.14 shows that important study quality indicators appear to be predictors of intervention effectiveness as evidenced by significant Qb and insignificant Qw in homogeneity tests. Use of monitoring mechanisms, double blinding, and intention-to- treat analysis were significantly related to less intervention effectiveness on increasing condom use at Time 1, while monitoring mechanisms and blinding remained the significant relationships at Time 3. Higher retention rate was significantly related to less intervention effectiveness at Time 1. When individuals who would otherwise drop out were retained in the interventions, they were also the ones less likely to change their behavior and thus dilute the intervention effectiveness. However, a higher retention rate made the intervention findings more generalizable to the targeted population. 161 Table 5.13 Relationships between Intervention Components and CSHAPI Effectiveness on Increasing Condom Use Time 1 Time 3 Homogeneity Summary Homogeneity Summary between groups Subgroup ES between Subgroup ES groups Qb** p value (I p k Qb” p d p k k=12 value k=5 value value Theogy-based 3.59 0.06 Yes 0.29 0.00 1 1 0.34 0.00 5* No 0.02 0.87 1 0 Objectives Susceptibility 1.12 0.29 Yes 0.39 0.00 2 0 No 0.25 0.00 10* 0.34 0.00 5* Motivation 0.97 0.32 0.18 0.67 Yes 0.34 0.00 6 0.21 0.47 1 No 0.24 0.00 6* 0.34 0.00 4* Intra-personal 2.79 0.1 7.53 0.01 competency Yes 0.21 0.00 7* 0.23 0.00 2* No 0.36 0.00 5 0.57 0.00 3 Inter-personal 2.25 0.13 4.47 0.03 skill Yes 0.33 0.00 7 0.44 0.00 3 No 0.20 0.00 5* 0.20 0.02 2* Inter-personal 6.59 0.01 9.20 0.00 growth Yes 0.04 0.68 1 0.09 0.38 1 No 0.32 0.00 1 I 0.46 0.00 4 Norm 2.96 0.09 0.84 0.36 Yes 0.57 0.00 1 0.39 0.00 3 No 0.25 0.00 1 1 0.28 0.00 2* Behavior skill 0.68 0.41 5.14 0.02 training Yes 0.29 0.00 9 0.46 0.00 2* No 0.21 0.01 3* 0.20 0.01 3 Activity Skills 5.05 0.03 5.14 0.02 demonstration Yes 0.35 0.02 6 0.46 0.00 2* No 0.16 0.00 6* 0.20 0.01 3 Role 0.4 0.53 demonstration Yes 0.32 0.00 5 0 No 0.25 0.00 7* 0.34 0.00 5* * indicates significant heterogeneity within group. ** was byfixed-effect analysis 162 Table 5.14 Relationships between Study Quality and CSHAPI Effectiveness on Condom Use Time 1 Time 3 Homogeneity Summary Homogeneity Summary between groups Subgroup ES between groups Subgroup ES k=12 K=5 Q1,” ’6‘" d pvalue k Q1," '3’” d pvalue k p value p value P value p value 1 df c2 1 4f (62) Accumula 0.02 -0.02 t_iv_e p=0.48 p=0.72 guality (0.02) (0.07) Monitor 6.74 9.25 p=0.01 p<0.005 Yes 0.14 0.03 4 0.23 <0.005 3 No 0.37 < 0.005 8 0.63 <0.005 2 Randomiz 3.15 £11in p=0.08 Yes 0.25 <0.005 1 0.34 <0.005 5 I III No 0.8 0.01 1 0 Blinding 6.67 9.20 p=0.04 p<0.005 df=2 df=1 Double 0.04 0.68 1 0.09 0.38 1 Single 0.32 <0.005 1 0.46 <0.005 4 0 Open 0.41 0.18 1 0 Intention" 8.57 1.36 to treat p<0.005 p=0.24 Yes 0.15 0.01 5 0.31 <0.005 4 * No 0.40 <0.005 7 0.52 <0.005 1 m 0.01 7.53 attention p=0.93 p=0.01 Yes 0.27 <0.005 7 0.23 <0.005 2 No 0.26 <0.005 5 0.57 <0.005 3 Retention -0.64 -0.07 p=0.04 p=0.97 (0.00) (0.06) k=12 k=4 *indicates significant heterogeneity within group. ** was by fixed-effect analysis *** was by mixed-effect analysis In comparison with condom use, most of the relationships between various intervention objectives and CSHAPI effectiveness on reducing unprotected sex were in 163 the same direction as condom use at both time points, while the relationships with unprotected sex tended to be insignificant (p>0.05). The exceptions were interpersonal skills at Time 1 and role demonstration at Time 3, both of which were associated with less effectiveness on reducing unprotected sex although they were related to more effectiveness on increasing condom use. On the other hand, enhancing susceptibility and norms were positively associated with intervention effectiveness on both dependent variables at two time points at an insignificant level. In addition, time effect on the relationships between various intervention objectives and intervention effectiveness (the changes of relationships over time) were similar for both dependent variables. For example, both dependent variables had a negative relationship with intrapersonal competency and a positive relationship with interpersonal skills at Time 1, and the relationships became significant at Time 3. In addition, both dependent variables had a negative relationship with interpersonal growth and a positive relationship with behavioral skills training and skills demonstration at Time 3. Intervention effectiveness on unprotected sex had a positive relationship with motivation/intention at Time 1, but the relationship became negative at Time 3. Both relationships were, however, not significant. Furthermore, this study found that either intervention intensity or total time span of intervention was not related to intervention effectiveness of both dependent variables (0:0.00 or -0.00, p>0.05 for unprotected sex). Similarly, the direction of relationships between study quality and intervention effectiveness on unprotected sex were similar to those with condom use at both time points, except for intention to treat and same attention at Time 1. In contrast to condom 164 use, intention-to-treat analysis at Time 1 was related to more intervention effectiveness on reducing unprotected sex. Same-attention designs between intervention and control groups were related to less intervention effectiveness. However, none of the relationships between important study quality indicators and intervention effectiveness on reducing unprotected sex at Time 1 were significant. At Time 3, the direction of relationships between important study quality indicators and intervention effectiveness was the same for both dependent variables. However, significant results differed by the dependent variables. For example, the relationship between monitoring mechanisms or double blinding and intervention effectiveness was significant for condom use, but not for unprotected sex. However, similar to condom use, a higher retention rate was strongly related to less intervention effectiveness on reducing unprotected sex at Time 3 (B=1.01,p=0.01). Publication Bias Table 5.15 shows that in order to change the hypothesis on CSHAPI effectiveness, the number of missing studies needed varied by time points and dependent variables. At Time 1, the fail-safe N is 116 for condom use and 45 for unprotected sex, while at Time 3, the number of required missing studies is 42 for condom use and 29 for unprotected sex. Table 5.15 shows that the required missing studies outnumber the observed studies. Figure 5.3 and 5.4 are funnel plots of intervention effectiveness on condom use at Time 1 and Time 3. At Time 1, the fact that more studies were centered on the right hand side of the vertical line indicated that the studies in the meta-analysis had a bias towards a higher level of intervention effectiveness on increasing condom use. Although this 165 pattern is less apparent at Time 3, there was some study bias existing at Time 3. The funnel plot shows that studies with larger samples had more extreme effect sizes than those with small samples. However, this is probably due to the small number of studies available at Time 3. Table 5.15 Fail-Safe N43 of CSHAPI Effectiveness by Dependent Variables and Time Points Post-test Time 1 Time 2 Time 3 Observed Missing Observed Missing Observed Missing Observed Missi N N N N N N N ng N Condom use 5 24 12 1 16 10 71 5 42 Unprotected 3 0 1 3 45 9 35 8 29 sex Figure 5.3 Funnel Plot of CSHAPI Effectiveness on Increasing Condom Use at Time 1 f -2 _ __ _____, I Funnel Plot of Precision by Std diff in means 1 i 1 . . 1 I 1 141 1 1 I 1 . 1 1 , 1 I 1 1 I '21 1 11 I I 101 Precision (1/Std Err) ‘3' I I1 1 1 1 1 I s I1 1’ j i I 4 I // (‘1 \ I 1 2 1 -/ \ I 0 l .1; _ - - ._ __-_,,-._...-____i 1 I -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 1 Std diff in means I 43 N is equal to k, number of studies, used in previous chapters. 166 Figure 5.4 Funnel Plot of CSHAPI Effectiveness on Increasing Condom Use at Time 3 I Funnel Plot of Precision by Std diff in means I I 1 1 I 1 I l 14 I 1 1 I I 1 1 1 1 '21 ’ 1 '1 I . 1 g) l 10 I Precision (l/Std Err) J.‘ 1 I ' 1 I I 8 I 1 . 1 1 / ; 1 4 I "I “1 \\ I I I /.v / 1 \“\ I 1 2 1. / 1 1 1 -~~~— r” 5 “we - 1 0 “' “fl _ __._ * t _.....L t —_ a I -2.0 -1.5 -l.0 -0.5 0.0 0.5 1.0 2.0 I I Std diff in means I Figure 5.5 Funnel Plot of CSHAPI Effectiveness on Reducing Unprotected Sex at Time 1 I“ ___fl_ __ “d _______V I Funnel Plot of Precision by Std diff in means 1 ' ’ 1 I 141 I I I I 12 I I I I '10 I Precision (l/Std Err) I” I i . if) I I 8 i .1105" I I ' 1 *1 ’ -1 1 1 2;. \ I 4 ’ ,// I I). I 1 ,1 1 I . l ‘ 1 1 1e 1 -- 1 1 1 ° ‘ '5 ” i r i i ’ T 1 I -2.0 -I .5 -I .0 -0.5 0.0 0.5 1.0 2.0 1 I Std diff in means I 167 Figure 5.6 Funnel Plot of CSHAPI Effectiveness on Reducing Unprotected Sex at Time 3 -2.0 -l .5 -l .0 -0.5 0.0 0.5 1.0 1.5 2.0 Std diff in means ; if. i __ __ ___ __ _____ ___ . ,_‘.. I_—e——-—._. ___-ow 1 Funnel Plot of Precision by Std diff in means 1 '41 III 1 I , 1 1 '21 1 I I I lOI Precision (l/Std Err) I 1 ,5: I I (”I I1 1 8I 11 I , 1. Q .1 II ‘1 I 61 1 I \ I 1 I IIIIIII I \II I 41 1 I 2 1 ., - f, I x ,\ ‘‘‘‘‘ - a“ I 1 0 C— —— + . g - - __ - _ 1:“ J Figure 5.5 and 5.6 are funnel plots of intervention effectiveness on reducing unprotected sex at Time 1 and Time 3. Study bias existed in an opposite direction between Time 1 and Time 3. At Time 1, intervention effectiveness were biased toward a lower level of effectiveness, while at Time 3, more studies had a higher level of effectiveness. Compared with condom use, the distribution of intervention effectiveness on reducing unprotected sex was more centered around the mean value. It showed that the effect sizes of reducing unprotected sex were more homogeneous than those for condom use. In addition, the fact that the effect sizes of condom use and unprotected scattered in the neck section of funnel plots suggests that the primary studies have large sample sizes that are more precise in estimating the population mean of intervention effectiveness. 168 Conclusion This meta-analysis centered on testing the research hypotheses developed in Chapter 2 and exploring variables that may explain variation of the CSHAPI effectiveness. One hypothesis is that CSHAPIs were more effective in increasing condom use or reducing unprotected sex than non-CSHAPIs. This study found that CSHAPIs were more effective than non-CSHAPI in the long term for both dependent outcomes. At post-test, non-CSHAPIS appeared to be more effective. At Time 2 (5 or 6 months after the interventions were completed), CSHAPIs became significantly more effective than non-CSHAPI in reducing unprotected sex. At Time 3, one year after the interventions, the advantages of CSHAPIs over non-CSHAPIs remained for both dependent variables at an insignificant level. However, the small number of studies available at Time 3 for a meta-analysis may have led to the insignificant result in the meta-analysis. As discussed in Chapter 3, 25% of the 50 eligible CSHAPI studies had 3-month follow-ups as the last test. The mean number of months of the last follow-ups was 8.6. Therefore, testing generalizability of the long-term effectiveness of CSHAPI effectiveness is limited by availability of empirical studies. However, as one of the primary purposes of meta-analysis is to generate a synthesis of the existing studies, the finding of the long-term advantages of CSHAPIs over non-CSHAPIS was still applicable to the existing studies. The second hypothesis is that CSHAPI participants with lower SES were less likely to increase condom use or to reduce unprotected sex than those with higher SES. This study found mixed evidence for both dependent variables in support that CSHAPI 169 was less effective in terms of changing the behavior of lower-SES participants. For example, less-than-full-time employment was positively related to intervention effectiveness on reducing unprotected sex (B=-2.34, p=0.13, k=3), but was negatively associated with intervention effectiveness on increasing condom use (B=-O.33, p=0.68, k=3). There was also a time difference in terms of the relationships between SES and C SHAPI effectiveness. Less-than-college education was moderately associated with a reduction of unprotected sex (B=-O.37, p=0.49, k=4), but was negatively related to reducing unprotected sex at Time 3 (B=O.l4, p=0.68, k=3). This hypothesis testing, however, is limited by a small number of studies reporting comparable SES data that may be used for a comparison in a research synthesis. In addition, there were no CSHAPI studies on condom use available at Time 3 for testing this hypothesis. On the other hand, the zero value in tau-square obtained through the meta- regression on reducing unprotected sex by less-than-college education indicates that there was no study random effect existing in the regression. Therefore, other intervention components (such as types of intervention objectives) or study quality (such as a retention rate) may explain the variance of CSHAPI effectiveness. For example, while lower SES may be negatively related to intervention effectiveness, it may also be related to a higher attrition rate. People with lower SES were more likely to drop out before follow-ups were completed. Therefore, the negative impact of lower SES on intervention effectiveness may be offset by a lovVer retention rate in a single study. The finding that retention rate had a strong relationship with intervention effectiveness ([3:- 0.64, p=0.04, k=12) indicates that people retained in the interventions were more likely to change risk behavior. 170 In addition, the quality assessment in Chapter 4 indicated that only 42% of the 50 eligible CSHAPI studies provided information on attrition in each intervention condition or tested whether there was differential attrition between intervention conditions. Therefore, the potential bias caused by differential attrition between intervention conditions among the participants with different SES was unlikely to be adjusted in the estimation of intervention effectiveness. Therefore, testing the second hypothesis on the relationship between SES and CSHAPI effectiveness is limited by the collective study quality and a lack of consistency of SES measurements among the studies. Furthermore, this meta-analysis provided strong evidence supporting that CSHAPIs were more effective when targeting a single sex as a majority in the samples. It found that the studies involving a sex majority were more effective than others with mixed-sex samples for both dependent variables at Time 1, and the relationships were significant at Time 3. CSHAPIs involving single sex as a majority and conducted at a group level may provide a social environment in which intervention messages were more readily accepted as a gender norm. This study also found that CSHAPI effectiveness did not differ by cultural category. It is not surprising as CSHAPIs by definition were tailored to each subculture. However, this study found that studies with samples involving different racial/ethnic composition significantly explained variance of intervention effectiveness on reducing unprotected sex between racial/ethnic groups. As the studies within racial/ethnic subgroups (for reducing unprotected sex) were homogeneous (small Qw), the racial/ethnic composition in the intervention samples is a factor that may explain variation of intervention effectiveness. Studies involving White Americans (MSM) as a 171 majority in the samples or involving mixed racial/ethnic samples were more effective at reducing unprotected sex than the other studies. As White American studies by default incorporated mainstream culture while adding additional culturally sensitive indicators that were included in other subcultures, White American studies may have more cultural intensity than all other studies. However, my work study found that cultural intensity is not significantly . associated with intervention effectiveness. It is possible that the culturally sensitive indicators were not applicable for the MSM studies or White American studies. It is also possible that certain types of culturally sensitive indicators were more important than the others. Therefore. including certain culturally sensitive indicators had more impact on intervention effectiveness than the number of culturally sensitive indicators, which might not include the most important culturally sensitive indicator. Moreover, this study tested whether specific culturally sensitive indicators or intervention objectives were related to a higher level of intervention effectiveness. The findings suggested a negative impact of an intervention focus on individual social psychological strength. For example, this study found a persistently negative relationship between (ethnic) pride as a culturally sensitive indicator and intervention effectiveness for both dependent variables at Time 1 and Time 3. A focus on interpersonal growth was also negatively related to intervention effectiveness for both dependent variables, and the relationships were significant at Time 3. Similarly, interventions involving enhancing intrapersonal competency as an intervention objective were negatively related to intervention effectiveness for both dependent variables at both time points. Therefore, the findings suggest that an intervention focusing on strengthening individual 172 psychological strength through their social identity or behavioral self-regulation might not lead to positive intervention effectiveness. In addition, this study found that the impact of culturally sensitive indicators may depend on time effect. For example, facilitators had a negative impact on intervention effectiveness for both dependent variables at Time 1, but the impact became positive at Time 3. In contrast, the relationship between community participation and intervention effectiveness had the opposite pattern. As the long-term positive impact of social experience and social responsibility as culturally sensitive indicators was not conclusive or were not tested (due to lack of studies) in the meta-analysis, the long-term effectiveness of CSHAPIs in comparison with non-CSHAPI might be a function of matching facilitators. In addition, this meta-analysis found that interventions involving the objectives of interpersonal skills and behavioral skills training were positively related to more intervention effectiveness for both dependent variables at Time 3. Therefore, this meta- analysis suggested that CSHAPIs were more effectiveness in the long term when specific culturally sensitive indicators (such as matching facilitators) and intervention objectives (such as interpersonal skills and behavioral skills training) came to positively impact the intervention effectiveness. This meta-analysis is limited by a small set of studies or a lack of certain studies available for subgroup analysis or for evaluation at post-test, Time 2 (condom use), and Time 3. In addition, the findings of the relationships between intervention effectiveness and various culturally sensitive indicators or intervention objectives may be conditioned by the collective study quality of CSHAPI studies. This meta-analysis found that 173 monitoring mechanisms, double blinding, and higher retention rates were related to less intervention effectiveness for both dependent variables. The quality assessment in Chapter 4 showed that only 16% of the studies involving double-blinding assignment and about one-third of studies reporting monitoring mechanisms. Therefore, the CSHAPI effectiveness was more likely to be inflated by a lack of these study quality indicators in the primary studies. However, the inflation of the intervention effectiveness might not differ between CSHAPIs and non-CSHAPIS as this study found that the average retention rate of CSHAPIs was the same as that of non-CSHAPIS. Therefore, the findings about the long-term impact of CSHAPI in comparison with non-CSHAPI were less likely to be confounded by different attrition rates between C SHAPIs and non-CSHAPIs. 174 References * indicates non-CSHAPI studies under review. For a list of CSHAPI eligible studies, refer to Chapter 3. *Baker, S. A., Beadnell, B., & Stoner, S., et al. (2003). Skills training versus health education to prevent STDs/HIV in heterosexual women: a randomized controlled trial utilizing biological outcomes. AIDS Education and Prevention, 15, 1-14. *Boekeloo, B. 0., Schamus, L. A., Simmens, S. J., Cheng, T. L., O'Connor, K., & D'Angelo, L. J. (1999). A STD/HIV prevention trial among adolescents in managed care. Pediatrics, 103, 107-115. *Boyer, C. B., Barrett, D. C., Peterman, T. A., & Bolan, G. (1997). Sexually transmitted disease (STD) and HIV risk in heterosexual adults attending a public STD clinic: evaluation of a randomized controlled behavioral risk-reduction intervention trial. AIDS, 11, 359-367. *Castro, F. G., & Tafoya-Barraza, H. M. (1997). Treatment issues with Latinos addicted to cocaine and heroin. In J. G. Garcia & M. C. Zea (Eds), Psychological interventions and research with Latino populations (pp. 191-216). MA: Boston: Allyn and Bacon. Cook, T., & Campbell, D. (1979). Quasi-experimentation: Design and analysis issuesfor field settings. Boston: Houghton Mifflin. *Dilley, J. W., Woods. W. J ., & Sabatino, J., et al. (2002). Changing sexual behavior among gay male repeat testers for HIV: A randomized, controlled trial of a single- session intervention. Journal of Acquired Immune Deficiency Syndromes, 30, 177- 186. *Gibson, D., Lovelle-Drache, J ., Young, Y., Hudes, E., & Sorensen, J. (1999). Effectiveness of brief counseling in reducing HIV risk behavior in injecting drug users: final results of randomized trials of counseling with and without HIV testing. AIDS and Behavior, 3, 3-12. Huwiler-Mtintener, K., Jfini, P., Junker, C., & Egger, M. (2002). Quality of reporting of randomized trials as a measure of methodological quality. The Journal of the American Medical Association, 28 7(21), 2801-2804. Kalichman, S. C., & Stevenson, L. Y. (1997). Lack of positive outcomes from a cognitive-behavioral HIV-AIDS prevention intervention for inner-city men: Lessons from a controlled pilot study. AIDS Education and Prevention, 9, 299- 313. 175 *Kiene, S. M., & Barta, W. D. (2006). A brief individualized computer-delivered sexual risk reduction intervention increases HIV/AIDS preventive behavior. Journal of Adolescent Health, 39, 404-410. ' *Koblin, B., Chesney, M., Coates, T., et al. for the EXPLORE Study Team. (2004). Effects of a behavioral intervention to reduce acquisition of HIV infection among men who have sex with men: the EXPLORE randomised controlled study. Lancet, 3 64, 41-50. *Kotranski, L., Semaan, S., Collier, K., Lauby, J ., Halbert, J ., & Feighan, K. (1998). Effectiveness of an HIV risk reduction counseling intervention for out-of- treatment drug users. AIDS Education and Prevention, 10(1), 19-33. *McCusker, J ., Stoddard, A. M., Hindin, R. N., Garfield, F. B., & Frost, R. (1996). Changes in HIV risk behavior following alternative residential programs of drug abuse treatment and AIDS education. Annals of Epidemiology, 6, 119-125. *Metzler, C. W., Biglan, A., Noell, J ., Ary, D. V., & Ochs, L. (2000). A randomized controlled trial of a behavioral intervention to reduce hi gh-risk sexual behavior among adolescents in STD clinics. Behavior Therapy, 31, 27-54. Mishra, S. I., Sanudo, F., & Conner, R. F. (2004). Collaborative research toward HIV prevention among migrant farmworkers. In B. P. Bowser & S. I. Mishra & C. J. Reback & G. F. Lemp (Eds.), Preventing AIDS: community-science collaborations (pp. 69-95). NY: New York: Haworth Press. Moher, D., Cook, D. J ., Eastwood, S., Olkin, I., Rennie, D., & Stroup, D. F. (1999). Improving the quality of reports of meta-analyses of randomized controlled trials: the QUOROM statement. Lancet, 354, 1896-1900. *O'Donnell, L., Stueve, A., & Doval, A. S. e. a. (1999). The effectiveness of the Reach for Health Community by Youth Service Learning Program in reducing early and unprotected sex among urban middle school students. American Journal of Public Health, 89, 176-181. *Picciano, J ., Roffman, R., & kalichman, S., et al. (2001). A telephone based brief intervention using motivational enhancement to facilitate HIV risk reduction among MSM: a pilot study. AIDS Behavior, 5, 251-262. *Roffman, R. A., Picciano, J‘. F., & Ryan, R., et al. (1997). HIV-prevention group counseling delivered by telephone: an efficacy trial with gay and bisexual men. AIDS and Behavior. 1, 137-154. *Roffman, R. A., Stephens, R. S., & Curtin, L., et al. (1998). Relapse prevention as an intervention model for HIV risk reduction in gay and bisexual men. AIDS Education and Prevention, 10, 1-18. 176 Rosenthal, R. ( 1979). The "file-drawer problem" and tolerance for null results. Psychological Bulletin, 86, 638-641. *Stall, R. D., Paul, J. P., Barrett, D. C., Crosby, G. M., & Bein, E. (1999). An outcome evaluation to measure changes in sexual risk-taking among gay men undergoing substance use disorder treatment. Journal of Studies on Alcohol, 60(6), 837-845. *Suarez-Al-Adam, M., Raffaelli, M., & O'Leary, A. (2000). Influence of abuse and partner hypermasculinity on the sexual behavior of Latinas. AIDS Education and Prevention, 12, 263-274. *Weeks, K., Levy, S. R., Gordon, A. K., Handler, A., Perhats, C., & Flay, B. R. (1997). Does parental involvement make a difference? The impact of parent interactive activities on students in a school-based AIDS prevention program. AIDS Education and Prevention, 9, 90-106. 177 CHAPTER 6 CONCLUSION This dissertation is a systematic evaluation of culturally sensitive HIV/AIDS prevention interventions (CSHAPIs) conducted in the US and published between 1996 and 2007. The year of 1995 is the threshold of eligibility because this was the time when the public was becoming more aware of the HIV/AIDS pandemic. After 1993, federal funding was more available for promoting the idea of condom use as an effective HIV/AIDS preventive behavior. A systematic evaluation of primary studies could be highly informative in terms of assessing the best practices now available to promote condom use and reduce unprotected sex. I conceptualized C SHAPIs as an organizational product shaped by various institutional forces, which may converge at a societal, institutional, or community level and construct the social context in which CSHAPIs and individuals were situated. Therefore, understanding the importance of social context and its underlying institutional forces is crucial for a critical evaluation of CSHAPI effectiveness. In order to gain an understanding of the various institutional forces, I reviewed literature in Chapter 2 from a sociological point of view in terms of various political, social, structural, and cultural factors in history that may have facilitated the AIDS spread and have continuingly shaped individual HIV/AIDS preventive behavior. I also reviewed in Chapter 2 Sociological concepts of social oppression, subculture and gender roles in order to explain the unequal demographic distribution of the HIV/AIDS epidemic. Based on these concepts, 1 developed two research hypotheses. 178 One hypothesis tested the assumption that CSHAPIS were more effective in promoting HIV/AIDS safe sex behavior than non-CSHAPI. The other hypothesis was based on a sociological point of view that was lacking in the assumption underlying C SHAPIs. I hypothesized that C SHAPI participants with lower SES were less likely to increase condom use or to reduce unprotected sex than others with higher SES. In Chapter 3, I documented a detailed profile of the existing CSHAPIs which were heterogeneous in terms of subcultural samples, demographic characteristics of the participants, intervention objectives, and culturally sensitive indicators. It was found that CSHAPIs were more likely to target African Americans, females, adults, single people, and those with lower SES. There was limited research attention on men (particular male drug users) as a sex majority in the samples. CSHAPI studies were largely lacking in the subpopulation of adolescents. In addition, it was found that matching a facilitator was the most widely used culturally sensitive indicator. In Chapter 4, I systematically assessed collective study quality of the CSHAPls by using a quality scale. It was found that the CSHAPI studies possessed moderate or solid quality with 92% using randomized experiments. However, this quality evaluation found threats to internal validity among the primary studies. They included a lack of random sampling, paucity of information on those failing out before follow-ups were completed, and inadequate attention to attrition analysis. Threats to statistical conclusion were also identified, including a lack of description on intervention implementation, a lack of report or estimate on reliability of behavioral measures, and a lack of intention- to-treat analysis. Threats to construct validity consisted of mono-method bias, a lack of 179 use or description of allocation treatment, and a lack of practice or description of double blinding. Chapter 5 is a meta-analysis that tested the hypotheses I developed and explored factors that may explain variation in CSHAPI effectiveness. It also investigated to what extent important study quality indicators may influence the estimate of intervention effectiveness. It was found that C SHAPIs were more effective than non-CSHAPIs in the long term. However, hypothesis testing on the relationship between socioeconomic status and CSHAPI effectiveness was limited by a small number of studies reporting comparable SES data that could be used for a comparison in a research synthesis. In addition, intervention objectives appear to play a far more important role in determining CSHAPI effectiveness than culturally sensitive indicators. It was found that (ethnic) pride had a persistently negative relationship with CSHAPI at different follow-ups. Matching facilitators had a positive relationship with CSHAPI effectiveness one year after the interventions were completed. However, study quality indicators, such as double blinding, monitoring mechanisms, and a higher retention rate, were associated with less CSHAPI effectiveness. Therefore, collective study quality may condition the relationships between various intervention components and intervention effectiveness. The advantage of this dissertation is that it may reduce bias in reviewing the literature by a systematic approach using a coding sheet and quality scale. Based on multiple studies. it increases the number of its construct inferences and statistical power. However, this dissertation has several limitations. 1 included in the “Conclusion” section of Chapter 4 a detailed discussion of the study limitations. especially involving the collective study quality. One of the study limitations is that it lacks an estimation of 180 reliability. The coding was entirely conducted by the author. The second limitation is that the study findings were less generalizable to studies targeting White Americans, youths, males, those with steady partners or being married, or others with high SES. The study findings were also less generalizable to the studies conducted outside of community settings. In addition, the study findings did not represent those studies involving different types of interventions or other types of HIV/AIDS risk behavior, conducted outside of the U .S., or published before 1996. Significant testing was limited by the small number of primary studies. However, because one of the primary purposes of a meta-analysis is to generate a research synthesis of the existing studies, 9a correlation at an insignificant level is still applicable to the existing studies. However, because the number of primary studies available for subgroups was in general small, some subgroup analyses were not tested. In addition, because this dissertation involved only published CSHAPIs, a publication bias test was conducted to investigate the level of potential bias. It was found that the number of missing studies required to change the direction of the study findings outnumbers the observed studies. Further research is needed to cover the subpopulations of male drug users and adolescents. The existing CSHAPI studies had limited coverage toward these subpopulations, while these groups represent a high risk for HIV infection. In addition, to what extent the structure of the drug user networks conditions drug users” safe sex behavior has not been fully examined in the drug user studies. However, it may be an area of research which may facilitate an understanding of the relationships between drug users” social status within their drug user networks and their safe sex behavior. The 181 understanding may aid in designing an effective HIV/AIDS prevention intervention for changing unsafe sex behavior which is normative in this subpopulation. In addition, further research on the subpopulation of socially oppressed White Americans may help develop culturally sensitive indicators that may be used for all subcultures. The culturally sensitive indicators used in this dissertation were more applicable to the subpopulations of drug users, men having sex with men, African Americans, Hispanic Americans, or other racial/ethnic minorities due to a lack of research and empirical studies on socially oppressed White Americans other than White American men having sex with men. In addition, in order to develop a program theory of CSHAPI, conducting a multivariate meta-regression will help understand relative importance of various intervention objectives. activities, and culturally sensitive indicators in terms of their contribution to CSHAPI effectiveness. The meta-analysis conducted in this dissertation found that (ethnic) pride, a matching facilitator, interpersonal skills, behavioral skills training, and a sex majority in a sample were related to C SHAPI effectiveness one year after the interventions were completed. Therefore. a multivariate meta-analysis involving these variables may aid in developing a program theory of CSHAPI. Furthermore, a multivariate meta—regression including different follow-ups will help understand dynamic changes of CSHAPI effectiveness in the long term. 182 APPENDIX A CODING SHEET 183 Coding Sheet Study ID __ _ _ ID of Eligible Study Coder ID __ Eligibility Criteria El __ The study was reported in English E2 _ The study was conducted in the US If NO check to either of above, STOP E3 The study uses condom use behavior as outcome variable E4 The study uses needle exchange as outcome variable If NO check in any of above, STOP E5 __ Randomized experiment with both pre-test and post-test E6 __ Randomized experiment with post-test only E7 _ Quasi-experiment with both pre-test and post-test E8 __ One-group with pretest and posttest method E9 __ Other(p|ease specify) If NO check in either E5, E6, or E7, STOP Please specify reasons other than above for which the study will be excluded Reference for study SI Author (s): S l a S 1 b S 1 c S2 Name/Title of Study 53 Date of Publication S4 Intervention used (name) SS Source of funding Location of source Ll Location of study 1) database 2) research register 3) reference ofeligible study 4) reference of systematic reviews on HIV/AIDS prevention intervention 5) index of conference proceedings 6) index of dissertation abstracts 7) direct contact with researcher 8) www 9) other (please specify) 184 Publication source of study I) 2) 3) 4) 5) 6) 7) 8) referred journal non-referred journal government report book dissertation thesis conference proceeding other (please specify) General Characteristics of Study SCI SC4 Social categories [select all that apply] The participants have memberships in the following social groups [check all that apply] prisoners drug users men having sex with men homeless runaways STD patients Location of intervention (city, state) sex workers students other (please specify) Time period in which the study was conducted (start) (end) Scope of sample __ local __ regional _ national Recruitment Strategies and Intervention Setting RSI R82 The place where the respondents were recruited [select all that apply] 1) Class 2) Community 3) Community center 4) Refugee center 5) Clinical center or hospital 6) Bar 7) Street 8) Prison 9) Drug treatment center l0) Other (please specify) Recruitment strategies [select all that apply] 1) Recruitment from class 2) Posters in community 3) Newspaper ads 4) Intervention outreach 5) Snowball 6) Radio 7) TV 8) Others 185 RS3 RS4 Recruitment incentives [select all that apply] I) Credit for class 2) Money: total amount 3) Free condom 4) HIV testing 5) Other (please specify) The setting where the intervention was conducted [select all that apply] I) Class 2) Community 3) Community center 4) Refugee center 5) Clinical center or hospital 6) Prison 7) Drug treatment center 8) Other (please specify) Intervention group The study had an intervention group NO. If NO, STOP here. YES. If YES, did it [select one of the following] IGIO __ explicitly claim cultural sensitivity WITH a description of culturally sensitive content/strategies [G] I ___ explicitly claim intervention culturally sensitive WITHOUT a description of culturally sensitive content/strategies IG I 2 __ report intervention design including culturally sensitive content/strategies WITHOUT claiming cultural sensitivity IG l3 __ provide non-culturally sensitive intervention If 1G I 3 is checked, go to next page about intervention contents of IG 1. Culturally sensitive content/strategies of IO] Did it [check all IG I s that apply] IGIa Participants help __ Design pilot program focus group other (please specify) IG 1 b Use pride of Ethnic Women sexual orientation (men having sex with men, lesbian, bisexual) Other (please specify) ' IG I c Gender role 1G I d Responsibility Family Community Other (please specify) IG I e Life experiences/settings __ Urban Poverty _ drug use 186 IGIf lGlg lGlh selling sex sexual orientation (men having sex with men, lesbian, bisexual) __ other (please specify) facilitator(s) matching target population film actor(s) matching target population facilitator(s) trained in cultural competency other (please specify)___ Cultural competency of facilitator(s) language compatibility language activity music compatibility language activity Cultural competency of message Other culturally sensitive content/strategies (please specify) Intervention content of [GI ICI IGI The purpose of IGI [check all that apply] lC2lGl IC3IGI no specific purpose described in the intervention to educate; HIV/AIDS knowledge gain to enhance perceived susceptibility, fear of HIV-AIDS to develop interpersonal skills (communications and social skills) to develop intrapersonal competence (ex. self-awareness, self-esteem) to develop interpersonal grth (such as group pride) to build social support, community norm other (please specify) Use of HIV/AIDS education in intervention 1) No 2) Yes 3) Not known Use of behavioral model in intervention no specific theoretical model described in the intervention Health Belief Model Theory of Reasoned Action Theory of Planned Behavior Social Cognitive Theory AIDS Risk Reduction Model Protection Motivation Model Fear Appeal Model other (please specify) 187 IG2 Intervention strategies of I61 ISIIGI the activities of IGI [check all that apply] Video Lecture role play demonstration role model Game consulting with facilitator(s) other (please specify) ISZIGI characteristics of facilitator(s) [check all that apply] not specified in report no facilitator(s) peer(s) facilitator(s) trained in HIV/AIDS education facilitator(s) trained in culturally competent skills nurse(s) facilitator(s) with same ethnic status as the participants facilitator(s) with same gender status as the participants community (group) leader(s) other (please specify) other (please specify) Intervention intensity of [C] II 1 1G] Number of session ll2IGl Length of each session (hours) II3IGI Interval between sessions (Day) ___(Hours) The study had a second intervention NO. If NO, go to p. 7. YES. If YES, did it [select one of the following] T070 I02 1 1022 IG23 explicitly claim cultural sensitivity WITH a description of culturally sensitive content/strategies explicitly claim intervention culturally sensitive WITHOUT a description of culturally sensitive content/strategies report intervention design including culturally sensitive content/strategies WITHOUT claiming cultural sensitivity provide non-culturally sensitive intervention - If IG23 is checked, go to next page about intervention content of IG2. Culturally sensitive content/strategies of IG2 Did it [check all IG25 that apply] IG2a Participants help _ Design pilot program focus group other (please specify)__ IG2b Use pride of Ethnic Women sexual orientation (men having sex with lGZc IGZd IGZe lG2f lG2g IG2h men, lesbian, bisexual) ___“ other (please specify) _ Gender role Responsibility Family Community other (please specify) Life experiences/settings Urban Poverty drug use selling sex sexual orientation (men having sex with men, lesbian, bisexual) Other (please specify) Cultural competency of facilitator(s) F acilitator(s) matching target population film actor(s) matching target population F acilitator(s) trained in cultural competency other (please specify) language compatibility language activity music compatibility language activity Cultural competency of message Other culturally sensitive content/strategy (please specify) Intervention content of IG2 ICI IGZ The purpose of IG2 [check all that apply] IC2IG2 lC3lG2 no specific purpose described in the intervention to educate; HIV/AIDS knowledge gain to enhance perceived susceptibility, fear of HIV/AIDS to develop interpersonal skills (communications and social skills) to develop intrapersonal competence (ex. self-awareness, self-esteem) to develop interpersonal grth (such as group pride) to build social support, community norm Other (please specify) ___ Use of HIV/AIDS education in intervention I) No 2) Yes 3) Not known Use of behavioral model in intervention no specific theoretical model described in the intervention Health Belief Model 189 Control Group CGI Theory of Reasoned Action Theory of Planned Behavior Social Cognitive Theory AIDS Risk Reduction Model Protection Motivation Model Fear Appeal Model Other (please specify) Intervention Strategies of IG2 IS l 102 the activities of IGZ [check all that apply] Video Lecture role play demonstration role model Game consulting with facilitator(s) other (please specify) ISZIG2 characteristics of facilitator(s) [check all that apply] not specified in report no facilitator(s) _ Peer(S) facilitator(s) trained in HIV/AIDS education facilitator(s) trained in culturally competent skills nurse(s) facilitators with same ethnic status as the participants facilitators with same gender status as the participants community (group) leader(s) other (please specify) Intervention Intensity of IG2 II I IG2 Number of session 112lG2 Length of each session (hours) II3IG2 Interval between sessions (Day) __(Hours) Study had a control group NO. IfNO, go to p. 9. YES. If YES, was it [select one of the following] __No treatment. If check, go to p. 9. _Waiting list ( number of days delayed for same intervention). Go to p. 9. _Receiving a different type of intervention. If check, answer questions below did it [select one of the following] CGIO explicitly claim cultural sensitivity WITH a description of culturally sensitive content/strategy CG I I explicitly claim intervention culturally relevant WITHOUT a description of culturally sensitive content/strategy 190 CG12 CGI3 report intervention design including culturally sensitive content/strategy WITHOUT claiming cultural sensitivity Provide non-culturally sensitive intervention If CGI3 is checked, go to next page about intervention contents of CG 1. Culturally sensitive content/strategies of CG] Did it [check all IG I s that apply] CGIa CGIb CGIc CGId CGIe CGlf CGlg CGIh Participants help Use pride of Gender role Responsibility Life experiences/settings Cultural competency of facilitator(s) Cultural competency of message Design Pilot program focus group other (please specify) Ethnic Women sexual orientation (men having sex with men, lesbian, bisexual) other (please specify) Family Community other (please specify) Urban Poverty drug use Prostitution sexual orientation (men having sex with men, lesbian, bisexual) other (please specify) Facilitator(s) matching target Population film actor(s) matching target population F acilitator(s) trained in cultural Competency other (please specify) language compatibility language activity music compatibility language activity other culturally sensitive content/strategy (please specify) I91 Intervention contents of CGI ICICGI IC2CGI IC3CGI The purpose of [GI [check all that apply] no specific purpose described in the intervention to educate; HIV/AIDS knowledge gain to enhance perceived susceptibility, fear of HIV/AIDS to develop interpersonal skills (communications and social skills) to develop intrapersonal competence (ex. self-awareness, self-esteem) to develop interpersonal growth (such as group pride) to build social support, community norm other (please specify) Use of HIV/AIDS education in intervention I) No 2) Yes 3) Not known Use of behavioral model in intervention no specific theoretical model described in the intervention Health Belief Model Theory of Reasoned Action Theory of Planned Behavior Social Cognitive Theory AIDS Risk Reduction Model Protection Motivation Model Fear Appeal Model other (please specify) Intervention Strategies of CGI ISICGI IS2CGI the activities of CGI [check all that apply] Video Lecture role play demonstration role model Game consulting with facilitator(s) other (please specify) characteristics of facilitator(s) [check all that apply] not specified in report no facilitator(s) peer(s) facilitator(s) trained in HIV/AIDS education facilitator(s) trained in culturally competent skills nurse(s) facilitator(s) with same ethnic status as the participants facilitator(s) with same gender status as the participants community (group) leader(s) other (please specify) Intervention Intensity of CG] II ICGI Number of session IIZCGI Length of each session (hours) II3CGI Interval between sessions (Day) (Hours) 192 CGZ Study had a second control group NO. IfNO, go to p.11. YES. If Yes, was it [select one of the following] __No treatment. If check, go to p. I 1. ___Waiting list (_ number of days delayed for same intervention). did it [select one of the following] CG20 CG2I CG22 CG23 Gotop.II. _Receiving a different type of intervention. If check, answer questions below explicitly claim cultural sensitivity WITH a description of culturally sensitive content/strategy explicitly claim intervention culturally relevant WITHOUT a description of culturally sensitive content/strategy report intervention design including culturally sensitive content/strategy WITHOUT claiming cultural sensitivity provide non-culturally sensitive intervention IfCG23 is checked, go to next page about intervention contents of CG2. Culturally sensitive contents/strategies of CGZ Did it [check all IG 1 s that apply] CG2a CG2b CG2c CGZd C(i2e CG2f Participants help Use pride of Gender role Responsibility Life experiences/settings Cultural competency of facilitator(s) 193 Design pilot program focus group other (please specify) Ethnic Women sexual orientation (men having sex with men, lesbian, bisexual) other (please specify) Family Community other (please specify) Urban Poverty drug use Prostitution sexual orientation (men having sex with men, lesbian, bisexual) other (please specify) Facilitator(s) matching target Population film actor(s) matching target population Facilitator(s) trained in cultural competency other (please specify) CGZg CG2b (‘ultural competency of message Other culturally sensitive content/strategy (please specify) language compatibility language activity _ music compatibility language activity Intervention contents of C62 IC I CG2 IC2CG2 IC3CG2 The purpose of IGI [check all that apply] no specific purpose described in the intervention to educate; HIV/AIDS knowledge gain to enhance perceived susceptibility, fear of HIV/AIDS to develop interpersonal skills (communications and social skills) to develop intrapersonal competence (ex. self-awareness, self-esteem) to develop interpersonal growth (such as group pride) to build social support, community norm other (please specify) Use of HIV/AIDS education in intervention I)No 2) Yes 3) Not known Use of behavioral model in intervention no specific theoretical model described in the intervention Health Belief Model Theory of Reasoned Action Theory of Planned Behavior Social Cognitive Theory AIDS Risk Reduction Model Protection Motivation Model Fear Appeal Model other (please specify) Intervention strategies of CGZ ISICG2 ISZCG2 the activities of CGI [check all that apply] Video Lecture role play demonstration role model Game Consulting with facilitator(s) other (please specify) characteristics of facilitator(s) [check all that apply] not specified in report no facilitator(s) peer(s) facilitator(s) trained in HIV/AIDS education facilitator(s) trained in culturally competent skills nurse(s) 194 F acilitator( s) with same ethnic status as the participants Facilitator(s) with same gender status as the participants community (group) leader Other (please specify) Intervention Intensity of CGZ II I CG2 Number of session II2CG2 Length of each session (hours) I13CG2 Interval between sessions (Day) (Hours) The names of intervention and control groups as [G], IGZ, CGI, and CG2 will be used in all the following questions. Sample Characteristics SCI Sample size Check if no information is available at all Check if no complete information at group level SCI SCIIGI SCIIGZ SCICGI SCICG2 Total sample size IGI IGZ CGI CGZ N 0/0 SC9 Ethnic status Check if no information is available at all Check if no complete information at group level SC9a SC9b SC9c SC9d SC9c Total I61 I62 CGI CG2 sample N % N % % N % N % 1) European American 2) African American 3) Latino American 4) Asian American 5) Native American specify) 6) Other (please 195 SC I 0 Gender _ Check if no information is available at all Check if no complete information at group level Female Male N % N % SC I 0a Total Sample SC I Ob IGI SC I CC IGZ SC 1 0d CGI SC I 0e CG2 SC6 Mean age Check if no information is available at all Check if no complete information at group level SC6a SC6b SC6c SC6d SC6c Total Sample IGI IGZ CGI CGZ SC 14 Marital Status Check if no information is available at all Check if no complete information at group level SCl4a SCl4b SCl4c SCl4d SCl4e Marital Status Total IGI IG2 CGI CG2 sample N % N "/0 N % N % N % 1) Single, never married 2) Married 3) Separated 4) Divorced 5) Widow/widower 6) Other (please specify) SC 7 Mean of education in years Check if no information is available at all __ C heck if no complete information at group level SC7a SC7b SC7c SC 7d SC7c Total Sample IGI IGZ CGI CGZ 196 SC8 Level of education Check if no information is available at all Check if no complete information at group level SC8a SC8b SC8c SC8d SC8c Education Level (%) Total 16] IG2 CGI CG2 sample I) Elementary school 2) Junior High School Student 3) Graduate of junior high school 4) Senior high school student 5) Graduate of senior hgigh school 6) College student Years in college (%) (%) "(TAT "(T/«T W 7) Vocational degree 8) Bachelor degree 9) Master student IO) Master I 1) PhD. student 12) PhD. I3) Other (please specify) SCI I Employment status Check if no information is available at all Check if no complete information at group level U N SClla SCllb Total Sam I01 162 CGI CG2 SCIlc SClld SClle SC l2 Family income % Check if no information is available at all __ Check if no complete information at group level SCIZa SC12b SC12c SC12d SC12e Family Income Total sample IGI IGZ CGI CGZ N % N % N % N % N % I) below 10,000 2) 10,000-15,000 3) l5,000-20,000 4) 20.000-25.000 5) 25,000-30,000 6) 30,000-35,000 7) above 35,000 8) other (specify) 197 SC I 3 Poverty/Low Income Check if no information is available at all Check if no complete information at group level Poverty/lower income Not povery/higher income N % N % SCl3a Total Sample SCI3b IGI SCI3e IGZ SCI3d CGI SCI3e CG2 SC 14 Attrition rate (%) Check if no information is available at all Check if no complete information at group level ARI AR2 AR3 AR4 ARS AR6 AR7 AR8 Measurement Reliability Posttest administered immediately after intervention Follow-up test: 1 month after intervention F ollow-up test: 2 months after intervention Follow-up test: 3 months after intervention Follow-up test: 6 months after intervention Follow-up test: 12 months afier intervention Follow-up test: 18 months afier intervention Other (please specify) Did the study test reliability using: [check all that apply] MRa MRb MRc MRd MRe MRf MRg Alpha Internal consistency (other) Kappa Percentage Split half Test-retest Other (please specify) Behavioral Model Variables BMa BMb BMe BMd BMe variable measure variable measure_ variable measure , reliability_ , reliability , reliability variable measure , reliability variable measure , reliability 198 BMf variable , reliability , reliability measure BMg variable measure BMh variable measure Test intervals , reliability Did the study use the following test intervals: [check all that apply] Tla TIb TIc Tld TIe TIf Tlg TIh Tli Intervention Outcome Variables Pretest Posttest administered immediately after intervention Follow-up test: I month after intervention Follow-up test: 2 months afler intervention Follow-up test: 3 months after intervention Follow-up test: 6 months after intervention Follow-up test: 12 months after intervention Follow-up test: 18 months after intervention Other (please specify) IO Did the study test behavior towards different types of partners? NO YES If YES, specify the types of partners Did the study specify type of sexual behavior, such as vaginal and. anal sex? NO YES If YES, specify the types of sex 100 Intervention outcome variables included in study [select all that apply] IOOa Condom use lOOb __ Unprotected sex lOOc Intention to use condom IOOd Needle sharing IOOe Injection drug use IOOf HIV/AIDS related knowledge lOOg Self-efficacy lOOh Sexual communication IOOi Number of sexual panners IOOj Other (please specify) IOOk Other (please specify) IOOI __ Other (please specify) IOOm Other (please specify) I99 IOI Condom use Definition Measurement Reliability I02 Unprotected sex Definition Measurement Reliability 103 Number of sexual partners Definition Measurement Reliability IO4 Needle Sharing Definition Measurement Reliability SFO All statistics used in study [check all that apply] Proportion difference Mean difference Odds ratio X square ANOVA ANCOVA t-teSt Correlation coefficient Confidence interval Regression Other (please specify) Other (please specify) Studying Findings: Condom Use, Unprotected Sex, and Needle Sharing SFI Condom use in general SF Ia Summary statistics from which the effect size was derived: [select one] Mean, SD F-ratio fi'om ANOVA t-value p-value and df Proportion Chi-square Correlation (r) Other (please specify) 200 SFlb SFlc Outcome: Check if no information is available at all Check if no complete information at group level IGI I62 Pretest Immediate l-month fol 2-month foll 3-month fol 6-month fol IZ-month fol l8-month fol Other lease Effect size: Summary statistics from which the effect size was derived: [select one] Mean, SD F-ratio from ANOVA t-value p-value and (If Proportion Chi-square Correlation (r) Other (please specify) Outcome: Check if no information is available at all __ Check if no complete information at group level IGI IG2 CGI CG2 Pretest Immediate posttest I-month followup 2-month followup 3-month followup 6-month followup lZ-month followup l8-month followup Other (please specify) Effect size: Summary statistics from which the effect size was derived: [select one] Mean, SD ' F-ratio fi'om ANOVA t-value p-value and df Proportion Chi-square Correlation (r) Other (please specify) 201 Outcome: Check if no information is available at all _ Check if no complete information at group level IGI IGZ CGI CG2 Pretest Immediate posttest I-month followup 2-month followup 3-month followy 6-month followup l2-month followup l8-month followup Other (please specify) Effect size: SF l d Summary statistics from which the effect size was derived: [select one] Mean, SD F-ratio from ANOVA t-value p-value and df Proportion Chi-square Correlation (r) Other (please specify) Outcome: Check if no information is available at all Check if no complete information at group level IGI IG2 CGI CG2 Pretest Immediate posttest I-month followup 2-month followup 3-month followup 6-month followup 12-month followup lS-month followup Other (please specify) Effect size 202 SF2 Condom use with (type of partner) SF2a Summary statistics fi'om which the effect size was derived: [select one] Mean, SD F-ratio from ANOVA t-value p-value and (If Proportion Chi-square Correlation (r) Other (please specify) Outcome: Check if no information is available at all Check if no complete information at group level IGI IG2 CGI CGZ Pretest Immediate posttest I-month followup 2-month followup 3-month followup 6-month followup IZ-month followuL IS-month followup Other (please specify) Effect size: SF2b Summary statistics from which the effect size was derived: [select one] Mean, SD F-ratio from ANOVA t-value p-value and df Proportion Chi-square Correlation (r) Other (please specify) Outcome: Check if no information is available at all _ Check if no complete information at group level IGI IGZ Pretest Immediate l-month fol 2-month fol 3-month fol 6-month fol lZ-month fol l8-month fol Other Effect size: 203 SF2c SF2d Summary statistics from which the effect size was derived: [select one] Mean. SD F-ratio from ANOVA t-value p-value and (If Proportion Chi-square Correlation (r) Other (please specify) Outcome: Check if no information is available at all _ Check if no complete information at group level IGI IG2 CGI CG2 Pretest Immediate posttest l-month followup 2-month followup 3-month followup 6-month followup l2-month followup I8-month followup Other (please specify) Effect size: Summary statistics from which the effect size was derived: [select one] Mean, SD F -ratio from ANOVA t-value p-value and (If Proportion Chi-square Correlation (r) Other (please specify) Outcome: Check if no information is available at all Check if no complete information at group level IGI IGZ CGI CGZ Pretest Immediate posttest l-month followup 2-month followup 3—month followup 6-month followup 12-month followup l8-month followg) Other (please specify) Effect size 204 APPENDIX B QUALITY SCALE 205 Quality Scale Study ID _ __ __ __ ID of Eligible Study _ _ Coder ID___ Quality of sampling and study design 1.1 If a sample was used, were the participants randomly selected from the population of interest? ___I) Yes _0) No Mark the major methodts) of sampling __ I. screening _ _ 2. disproportionate stratification _ 3. multistage sampling __ 4. multiple frames __ 5. multiplicity sampling __ 6. two-phase screening _ 7. location sampling _ 8. snowballing __ 9. other __9) Cannot tell If Yes, go to 1.4 1.2 If a sample was used with a nonrandom sampling method, did the study explicitly mention whether the participants of the intervention were similar to the target population (fi'om which they were chosen)? ___1) Yes __0) No ___9) Cannot tell l.3 If a sample was used with a nonrandom sampling method, did the study test whether the participants of intervention group were similar to those in control groups? __I) Yes __0) No __9) Cannot tell 1.4 If the study involved a design of comparison between intervention and control groups, were the participants randomly assigned into an intervention or control group? __I) Yes __0) No __9) Cannot tell 1.5 Was information given on the number and characteristics of subjects who were eligible but refused to participate? __I) Yes Information about those who had refused to participate was provided, or a comparison was made between people who had participated and others who had been eligible but had refused to participate and who had participated. __0) No __9) Cannot tell 206 1.6 1.7 l.9 Was information given on the number and characteristics of participants who had dropped out intervention? __I) Yes Information about those who had dropped out was provided or a comparison was made between people who had completed intervention and others who had dropped out. _0) No ___9) Cannot tell Was information given on the number and characteristics of participants who were lost to follow- up before completing all elements of data collection? __ I) Yes Information about those who did not complete follow-up test was provided, or a comparison was made between people who had completed all elements of data collection and others who had not. 0) No _9) Cannot tell Was information provided on whether or not the intervention and control programs had been implemented as planned? __I) Yes. __ 1. external monitor(s) was present __ 2. the intervention participants rated the intervention, __ 3. other __0) No __9) Cannot tell Was the definition of condom use behavior adequate?-- ___I) Yes Condom use was defined as the behavior of condom use ___0) No Condom use was defined as different behavior from condom use, broadly defined as unprotected sex or narrowly defined as a subset of condom use behavior __9) Cannot tell I.IO Was the measure of condom use behavior adequate?-- ___I) Yes Condom use was measured as the behavior of condom use during a reasonable time/frame, between no earlier than the past three months or no later than the past week apart fi'om the test and the time of the test. _0) No If condom use behavior was broadly measured as unsafe sex, other methods of protective sexual behavior. or the behavior of condom use outside the above time range. __9) Cannot tell 1.] I Overall quality of sampling and study design based answers to the questions l.I-l . I 0. The following guideline can be used to assist with deriving a summary score. _ 1) Poor Most questions above were checked “Partial, " “No. or “Cannot tell " __p 2) Marginal Many of “Partial, " “Cannot tell, " and "No" checked above was between 5 and 7 times __ 3) Moderate Some of “Partial, " “Cannot tell, " and “No" were checked. _ 4) Solid Very few of “Partial, ” “Cannot tell, " and ”No" were checked 5) Outstanding All of the above were checked “Yes" 1.12 Confidence rating on quality of sampling and study design I) Very low (little basis) 2) Low (best estimate) 3) Moderate (weak inference) 4) High (strong inference) 5) Very high (explicitly stated) 207 The quality of program descriptions 2.] Were specific program objectives provided for the intervention group? the control group? __3) Yes _3) Yes __2) Partial __2) Partial __I ) No __I) No 2.2 Was the intervention content clearly described for the intervention group? the control group? __3) Yes _3) Yes __2) Partial ___2) Partial _I) No _ _ I) No The objectives were explicitly stated. The objectives were somewhat implied in the intervention content although not explicitly stated The objectives were not explicitly stated. nor was it easy to estimate by reading the content of the intervention The description of content was sufficient. The content was somewhat easy to understand but there were some issues The content was difficult to understand. There were a lot of issues. 2.3 Overall quality of program descriptions based on answers to the questions 2. I-2.2. The following guideline can be used to assist with deriving a summary score. __ 1) Poor Two “No " were checked. __ 2) Marginal One “No ” was checked. __ 3) Moderate Two “Partial " were checked. 4) Solid Only one “Partial ” was checked 5) Outstanding 2.3 Confidence rating on quality of program description 1) Very low (little basis) 2) Low (best estimate) 3) Moderate (weak inference) 4) High (strong inference) 5) Very high (explicitly stated) The quality of the study’s data analysis All were checked “Yes. ” 3.1 Was the sample sizejustified (such as, with a power calculation)? Did the study have a small sample size? _I) Yes __0) No ___9) Cannot tell 3.2 Use of statistics on major intended intervention outcomes _1) Test statistic and observed probability value are stated __2) Observed probability level was given, but test statistic value was not stated _“3) Test statistic was given, but no observed probability level given __4) If neither test statistic nor observed difference for negative trials 208 3.3 Use of posterior ,3 estimates of observed difference between intervention and control groups (I might not use this question) _I) Yes _0) No _9) Cannot tell 3.4 If intervention and controlled groups had not been equivalent at baseline. was this problem addressed in analysis? If they had been equivalent at baseline, skip this question. 3) Yes All or most major differences at baseline were addressed in analysis 2) Partial Some major diflerences at baseline were not addressed in analysis I) No Most or all of the major differences at baseline were not addressed in analysis Statistical inference 3.5 Confidence Limits (were confidence intervals provided for the proportions, rates, or means that was used as the measure of intervention outcomes?) ___1) Yes _0) No __9) Cannot tell 3.6 If follow-up tests were used in a study, were repeated measures used in analysis? If no follow-up, skip this question. _I) Yes __0) No _9) Cannot tell 3.7 Use of regression analysis correlation _I) Yes __0) No _9) Cannot tell 3.8 Were potential confounders adequately controlled for in the analysis? The potential confounders may include racial/ethnic, class, gender, sexual orientation status, or other demographic characteristics of the participants that may influence their sexual behavior. _3) Yes All or most confounders were controlled. __2) Partial Some major confounders were not controlled. __ I) No Most or all confounders were not controlled. 3.9 Were analytic specifications of condom use variables consistent with the evaluation questions or hypotheses under study? ___I) Yes _0) No _9) Cannot tell 209 3.10 Overall quality of data analysis based on answers to the questions 3.1-3.9, and questions on intention to treat. The following guideline can be used to assist with deriving a summary score. _ 1) Poor Serious issues in the analysis, or most questions were marked “No " or “Partial" __ 2) Marginal Some major issues in the analysis, or many questions were checked “No” or “Partial." __ 3) Moderate 3.] was checked “Yes; some questions were checked “No” or "Partial 4) Solid 3. I, 3. 4, 3.8 were checked “Yes 3.2 was checked the first; very few other questions were not checked “Y es ” 5) Outstanding 3.1, 3.4. 3.5. 3.6, 3.8, and 3.9 were checked “Yes"; 3.2 was checked the first. Intention to treat analysis was used. 3.1 1 Confidence rating on quality of data analysis 1) Very low (little basis) 2 Low (best estimate) 3) Moderate (weak inference) 4) High (strong inference) 5) Very high (explicitly stated) Evaluating the study’s results 4.] Were all this study’s questions answered? 3) Yes All of this study 's questions were answered. 2) Partial Some major questions were not answered. I) No None or very few of the components of the questions were answered. 4.2 Were attrition rates given for each group? 3) Yes The attrition rates were explicated stated. 2) Partial The attrition rates could be estimated. 1) No Neither could the attrition rates be explicitly stated, nor could they be estimated. 4.3 Were data given that demonstrated that refusers and people completing follow-ups were alike _I) Yes _0) No ___9) Cannot tell 4.4 Were data given that demonstrated that people who had dropped out during interventions and those who had completed follow-ups were alike? ___I) Yes ___0) No _9) Cannot tell 4.5 Were data given that demonstrated that people who did not complete follow-ups and those who had were alike? __I) Yes __0) No __I) Cannot tell 4.6 If refusers differed, were the limitations to the results discussed? __fll) Yes ___0) No __9) Cannot tell 210 4.7 If people who had dropped out differed, were the limitations to the results discussed? _1) Yes ___0) No __9) Cannot tell 4.8 If people failed to complete follow-ups differed, were the limitations to the results discussed? __1) Yes _0) No _9) Cannot tell 4.9 If refusers, people who had dropped out, or failed to complete follow-ups differ, were the limitations to the results discussed? _3) Yes ___2) No __I) Cannot tell 4.10 Overall quality of the study’s results based on answers to the questions 4.1-4.9. The following guideline can be used to assist with deriving a summary score. __ 1) Poor Most questions were not checked “Yes ” when the questions were applicable. _ 2) Marginal Few questions were checked "Y es ’ when the questions were applicable. ” __ 3) Moderate Some questions were checked “Y es ” when the questions were applicable. _ 4) Solid Many questions were checked “Y es ” when the questions were applicable. 5) Outstanding All or most questions were checked “Y es " when the questions were applicable. 4.1 1 Confidence rating on the study’s results 1) Very low (little basis) 2) Low (best estimate) 3) Moderate (weak inference) 4) High (strong inference) 5) Very high (explicitly stated) Evaluating the study’s conclusions 5.1 Were the conclusions based on the study’s data in that findings were applied only to the population represented by the sample that was included in the research? "_3) Yes _2) Partial __1) No 5.2 Were the conclusions based on the study’s data in that findings were applied only to the program(s) that were included in the research? __3) Yes ____2) Partial ___1) No 5.3 Were limitations of study design described? “_3) Yes __2) Partial -___I) No 211 5.4 Were limitations of sampling addressed? __3) Yes ___2) Partial __1) No 5.5 Were limitations of the implementation of the intervention discussed? _3) Yes _2) Partial __I) No 5.6 Were limitations of data collection discussed? _3) Yes __2) Partial ___I) No 5.7 Were limitations of data analysis discussed? __3) Yes __2) Partial ___-I) No 5.8 Overall quality of the study’s conclusions based on answers to the questions 5.1-5.7. The following guideline can be used to assist with deriving a summary score. 1) Poor All or most questions were “No " or “Partial " when the questions were applicable. ___ 2) Marginal Few questions were checked “Yes" when the questions were applicable. ” _ 3) Moderate Some questions were checked “Y es ” when the questions were applicable. __ 4) Solid Many questions were checked “Yes " when the questions were applicable. 5) Outstanding All or most questions were checked “Yes " when the questions . were applicable. 5.9 Confidence rating on quality of sampling and study design 1) Very low (little basis) 2) Low (best estimate) 3) Moderate (weak inference) 4) High (strong inference) 5) Very high (explicitly stated) This study in general : 6.1 General quality 1) Poor Sampling, study designs, and analysis were strongly biased; and/or the content of the reporting was difficult to understand __ 2) Marginal Many major issues in sampling, study designs, and analysis; and/or the content of the reporting was somewhat diflicult to understand __ 3) Moderate There were some issues in sampling, study designs, and analysis; but reporting of the studying was still sufficient. 4) Solid There were minor issues in sampling. study designs, and analvsis; and reporting of the study was sufficient. 5) Outstanding Sampling, study designs, and data analysis were adequate; and reporting of the study was sufficient. 212 6.2 Confidence rating on general quality 1) 2) 3) 4) 5) Very low (little basis) Low (best estimate) Moderate (weak inference) High (strong inference) Very high (explicitly stated) I\) b)