“me.” .Ifi l - g A 1,4; 4 r 1.3 . .V .5... it . <4... Wuymhuhu$ : A a .1 NA 3% u .3 .3 fit. .. 1.8- . to . nub...” . .3. mi” L. k . 4 V15 ‘ ndwm-l.- a ‘ 173.31}, . . 1.144).» :32‘ , ,tw1‘ltrfiunuWW4cr1lu , . 14'112 . ‘ ‘ Wkflxmamv I 1.: an. a . 1 .,,......I.mrLuufln.. 39,34... r.‘ i 1* 895255359 5.235%... . .- Au... Amflluuwumwuuhwmmfl - .51.... 1M... .. . «unwflwmfl’ . itmmnuwMMWMWHMWAIMrftml. 55:... .53. 3.. 33:4,: . uflflcéxumg: . . I)ate 0—7639 imamintuituuumisuigxmmi 3 1293 0155 LIBRARY Michigan State University This is to certify that the thesis entitled Developing and Testing a Measurement Model for Perceived Barriers to Condom Use: A Cross-Cultural Study presented‘by Kenzie Alynn Cameron has been accepted towards fulfillment of the requirements for Master of Arts , Communication degreein a; WW Major professor 19 November 1996 MS U is an Affirmative Action/Equal Opportunity Institution PLACE N RETURN BOX to romovothio ohookout trom your rooord. TO AVOID FINES Mum on or bdoro dot. duo. DATE DUE DATE DUE DATE DUE nsu Io An Affirmative Action/Equal oppomnny trunnion ‘ W ”3-9.1 DEVELOPING AND TESTING A MEASUREMENT MODEL FOR PERCElVED BARRIERS TO CONDOM USE: A CROSS-CULTURAL STUDY By Kenzie Alynn Cameron A THESIS Submitted to Michigan State University in partial fiilfillment of the requirements for the degree of MASTER OF ARTS Department of Communication 1 996 ABSTRACT DEVELOPING AND TESTING A MEASUREMENT MODEL FOR PERCEIVED BARRIERS TO CONDOM USE: A CROSS-CULTURAL STUDY By Kenzie Alynn Cameron A model suggesting that the perceived barriers dimension of the Health Belief Model (Rosenstock, 1974) is second-order unidimensional is presented and tested using confirmatory factor analysis procedures. Subjects were college students (N = 365) from two separate cultures (United States N = 178; Kenya N = 187). The health threat used was contraction of HIV, and the perceived barriers dimension was tested by responses of subjects to items measuring perceived barriers to condom use. The proposed second- order unidimensional model was not consistent with the data. Alternative models were proposed and tested. Implications of understanding the perceived barriers dimensions are discussed in terms of applying such findings to communication campaigns aimed to increase behavior of adoption of a recommended response to a health threat. This thesis is dedicated to my family: Jule, E. Alan, and Bain Cameron, without whose assistance, guidance, listening ears, support, belief, and love I could not have completed this work. iii ACKNOWLEDGMENTS I would like to thank my adviser, Dr. Kim Witte, who, in addition to her role as adviser, also provided me the opportunity and the support to attempt a cross-cultural study such as this one. I would also like to thank my committee members, Dr. Franklin J. Boster and Dr. Charles T. Salmon, who provided me invaluable assistance and guidance along the way. In addition, I would like to thank Betty H. La France, whose assistance in demystifying the various computer programs I used to analyze my data was far past valuable; Janet K. McKeon, who listened patiently to me and provided me with many thoughts on the thesis experience as well as on surviving graduate school; Robb Sinn, whose late-night e-mails urged me to keep going, even in the face of adversity; and Julie Burgess, who now knows more than she ever wanted to about this thesis, but always took the time to listen and to care. Additional thanks go to Deb Tigner, for her help in ensuring that the questionnaires reached Kenya; to Dr. Charles B. K. Nzoika, of the University of Nairobi; and to Solomon Nzyuko, Dr. David Nyamwaya, and Peter Ormondi, of the African Medical and Research Foundation. I would also like to acknowledge an All-University iv Research Initiative Grant awarded by Michigan State University, which funded an earlier project that led to the development of this project. My greatest thanks go to my family - Mom, Dad, and Bain - they alone know all of the steps that it took to get me where I am today, and they are the ones who made such steps possible. TABLE OF CONTENTS LIST OF TABLES ............................................................................................... viii LIST OF FIGURES .............................................................................................. ix CHAPTER 1 INTRODUCTION ................................................................................................. 1 CHAPTER 2 LITERATURE REVIEW ...................................................................................... 5 CHAPTER 3 METHODS ............................................................................................................ 17 CHAPTER 4 RESULTS ............................................................................................................ 22 CHAPTER 5 DISCUSSION........... ........................................................................................... 46 CHAPTER 6 CONCLUSION .................................................................................................... 54 APPENDIX A INDIVIDUAL BARRIERS TO CONDOM USE .................................................. 57 APPENDIX B RELATIONAL BARRIERS TO CONDOM USE ................................................. 59 APPENDD( C PHYSICAL BARRIERSTO CONDOM USE ...................................................... 61 APPENDIX D KNOWLEDGE-BASED/SELF—EFFICACY BARRIERS TO CONDOM USE ..... 62 APPENDIX E SOCIAL NORMS/CULTURAL VALUES BARRIERS TO CONDOM USE ...... 63 APPENDIX F STRUCTURAL BARRIERS TO CONDOM USE ............................................... 64 APPENDIX G BARRIERS TO CONDOM USE: THE 36 ITEMS USED FOR THE PROPOSED SIX-FACTOR MODEL ................................................................... 65 APPENDD( H BARRIERS TO CONDOM USE: THE 16 ITEMS USED FOR THE FOUR-FACTOR ANALYSIS ............................................................................... 68 LIST OF REFERENCES ...................................................................................... 70 LIST OF TABLES Table 1 - The Observed Correlations and Factor Loading Matrix for the Proposed Six-Factor Model ........................................................................ 26 Table 2 - The Expected and Error Correlation Matrix for the Proposed Six-Factor Model ........................................................................ 28 Table 3 - The Observed Correlations and Factor Loading Matrix for the Four-Factor Model ..................................................................................... 34 Table 4 - The Expected and Error Correlation Matrix for the Four-Factor Model ..................................................................................... 3 5 Table 5 - The Observed Correlations and Factor Loading Matrix for the Single-Factor Model .................................................................................. 40 Table 6 - The Expected and Error Correlation Matrix for the Single-Factor Model ................................................................................... 41 Table 7 - The Observed Correlations and Factor Loading Matrix for the Single-Factor Model using Data fiom the United States Sample Only ......... 42 Table 8 - The Expected and Error Correlation Matrix for the Single-Factor Model using Data fi'om the United States Sample Only ......... 43 Table 9 - The Observed Correlations and Factor Loading Matrix for the Single-Factor Model using Data from the Kenyan Sample Only .................. 44 Table 10 - The Expected and Error Correlation Matrix for the Single-Factor Model using Data from the Kenyan Sample Only .................. 45 viii LIST OF FIGURES Figure 1 - Proposed Second-Order Unidirnensional Factor Model Of Perceived Barriers to Condom Use .................................................................... l4 Chapter 1 INTRODUCTION The prevalence and spread of Acquired Immune Deficiency Syndrome (AIDS) is a constant threat to society as the number of AIDS cases continues to increase, even in the face of many HIV/AIDS prevention campaigns. Recent epidemiological records show unceasing increases in reported AIDS cases from around the world (World Health Organization, 1995, 1996). As of June 30, 1996, the Weekly Epidemiological Record, published by the World Health Organization (WHO), documented a 19% increase in reported AIDS cases of adults and children since July 1, 1995 (WHO, 1996). Based upon available data Obtained from a country-by-country analysis, WHO estimates that there have been over 7.7 million AIDS cases in adults and children worldwide (WHO, 1996). WHO estimates that 21 million adults and 800,000 children are currently living with HIV/AIDS and that approximately 25.5 million adults, in addition to over 2.4 million children, have been infected with HIV since the pandemic had its start in the late 19708 and early 1980s (WHO, 1996). The June 30, 1996 estimate Of 21 million adults living with HIV is a 24% increase fi'om the December 15, 1995 report, in which WHO estimated that approximately 17 million adults were infected with HIV (WI-IO, 1995, 1996). 2 The majority Of HIV cases in the world are currently found in Sub-Saharan Africa, where HIV seroprevalance rates have been suggested to range from less than 1% up to 20% of the adult population (AIDS Analysis Africa, 1993). AIDS was made a “notifiable disease” by the Kenyan government under the Public Health Act in 1987, at which time 1,497 cases had been reported from all of the Kenyan provinces (Agata, Muita, Muthami, Gachihi, & Pelle, 1993; Rachier, 1993). Recent epidemiological reports rank Kenya as the fourth highest country in terms of estimated HIV infection, with current estimates being in excess of 1,000,000 cases of HIV infection within the provinces of Kenya (WHO, 1995). Current estimates also place the United States (700,000 cases) in the top ten countries of estimated numbers of HIV infections (WHO, 1995). Such reports indicate the increasing effect Of HIV/AIDS on populations throughout the world. HIV/AIDS is an issue Of global importance; a disease that is not bound by a country’s borders. Research is needed to aid the development of prevention campaigns and intervention strategies in order to assess how best to dissenrinate information about the contraction of HIV, and to alert people to efiicacious preventive measures against such contraction. Many prevention campaigns have been developed in the fight to decrease the spread Of HIV/AIDS. However, despite the existence of these campaigns, and the focus of these campaigns on prevention, individuals are continuing to engage in unsafe, and risky, sexual behavior (Manning, Balson, Barenberg, & Moore, 1989; Sereno & Dunn, 1994; Sheer & Cline, 1994). Although current HIV/AIDS prevention campaigns appear to be increasing individuals’ levels Of knowledge about HIV/AIDS, research reveals that such increased knowledge is not resulting in an apparent increase in condom use, specifically among college students in the United States (Baldwin, Whiteley, & Baldwin, 3 1990; Serene & Dunn, 1994; Sheer & Cline, 1994). In addition, as college students appear to Often engage in risky behaviors, research suggests that students have a greater probability of contracting HIV /AIDS than the average person (Baldwin & Baldwin, 1988; Sheer & Cline, 1994). Research suggests that the most rapid increase Of HIV contraction is expected to be found among young adults (Chesney, 1994). Further, it has been noted that knowledge-based HIV/AIDS programs have not been efi‘ective cross culturally in changing behavior: "One Of the mistakes that has been made in AIDS programs is the assumption that if you provide a lot Of information and improve knowledge, that will afl‘ect the epidemic. It has been quite evident that in most countries now, sometimes as high as 90 percent know about AIDS, how it is transmitted and how it can be prevented, but that has not led to people actually changing their behavior and sustaining it” (N ovicki, 1992, p. 28). Increasing individuals’ knowledge about HIV/AIDS is not leading to a change in their sexual behavior; there do not appear to be increases in the use Of condoms as a preventive measure against contraction of HIV. In the summer of 1988, a brochure, “Understanding AIDS,” was sent by the Public Health Service to every household in America (Gerbert & Maguire, 1989; Koop, 1988). This brochure contained a message from then Surgeon General C. Everett Koop. His message encouraged families to talk about AIDS, and to learn more about AIDS. In addition, the message noted the behaviors that place individuals at risk for contraction Of HIV, as well as explained how one could use condoms to serve as a protective barrier against contraction of HIV (KOOp, 1988). Although public response to this brochure was positive (Gerbert &Maguire, 1989), increased behavior of condom use is still lacking. Thus, if knowledge is not the key tO behavior change, we must look for other potential 4 variables that may be hindering individuals fi'om adopting the recommended preventive action of using condoms. Ifresearch is able to identify those factors that are acting as barriers to condom use, such research could then be incorporated into current HIV/AIDS prevention campaigns in an attempt to break down the barriers to condom use. Using the dimension of perceived barriers fi'om the Health Belief Model (Rosenstock, 1974) as a base, this study (1) will elicit barriers to condom use that individuals perceive to be in existence, (2) will test a proposed factor structure of these perceived barriers to condom use, and (3) will test the proposed factor structure Of perceived barriers to condom use cross culturally, using an United States and a Kenyan sample, through administration of a questionnaire focused on perceived barriers to condom use. Due to the fact that HIV/AIDS infection estimates continue to increase worldwide, there is an urgent need for cross-cultural analysis and research that can be used by health educators to develop culturally-appropriate HIV/AIDS prevention programs. Such a nwd has been demonstrated by a number Of researchers (Marin & Marin, 1990; Peterson & Marin, 1988; Vlfrtte & Morrison, 1995; Yep, 1992). This thesis, focusing on perceived barriers to condom use, may ofl‘er useful insight into the development of such prevention programs. Chapter 2 LITERATURE REVIEW The Hgth Belii Mme] The Health Belief Model (HBM) has served as an organizing framework for much of the work regarding health behavior and compliance (Janz & Becker, 1984). The HBM was developed as an attempt to explain the behavior of people who were not taking advantage of preventive health services Ofl‘ered to them, even when the service was Ofi‘ered at little or no cost (Rosenstock, 1974). In addition, the HBM allowed researchers to study the behavior Of individuals who were not suffering fi'om a disabling disease, yet who were sufi‘ering fi'om diseases that could be attended to through preventive care (Rosenstock, 1974). Early tests of the HBM focused on diseases such as tuberculosis, cervical cancer, and influenza, among others. Currently, the HBM continues to be used in research regarding individuals' responses to preventive health behavior (Calnan & Rutter, 1986; Yep, 1993). The five dimensions identified by the HBM that influence preventive health behaviors are: (a) perceived susceptibility (e.g., one's perception Of the risk of contracting a condition), (b) perceived severity (e.g., one's beliefs about the seriousness of the 6 disease), (c) perceived benefits (e.g., one's beliefs regarding the benefits of performing a recommended response), (d) perceived barriers (e.g., one's beliefs about the negative aspects or "costs" Of a particular health action), and (e) the dimension of cues to action (e.g., a stimulus triggering the action process) (Janz & Becker, 1984; Kirscht & Joseph, 1989; Rosenstock, 1974;1V1tte, Stokols, Ituarte, & Schneider, 1993). In the HBM, perceived benefits and perceived barriers act in tandem to produce a final “cost analysis” (perceived benefits minus perceived barriers), which influences one’s likelihood of taln'ng action, along with perceived threat, which is affected by perceived susceptibility, perceived severity, and cues to action (J anz & Becker, 1984;Rosenstoc1g 1974). Although the dimension of perceived barriers is theorized to be an antecedent to one’s likelihood of taking action (Janz & Becker, 1984; Rosenstock, 1974), the conceptualization of this dimension has been vague. Most researchers do not define barriers as a construct; rather, they provide examples Of barriers (Clark, 1983). Little is known about this factor of perceived barriers, and suggestions regarding a definitive measurement of this dimension are lacking. Although many researchers have used the dimension of perceived barriers in their work, conceptualizations of perceived barriers have ranged from cultural barriers to a specific health action (Clark, 1983; Ruiz, 1985) to personal barriers hindering an individual’s action to take preventive health measures (Sereno & Dunn, 1994). These numerous and seemingly distinct conceptualizations of various researchers of the perceived barriers dimension cause one to question if there may be multiple dimensions Of barriers underlying an overall perceived barriers dimension. Specifically, this study seeks to test the possibility that the perceived barriers dimension suggested by 7 the HBM is indeed multidimensional, that is, that there may be a number of dimensions or categories of barriers, that are conceptually distinct, yet also can be encompassed under an overall barrier dimension. A multidimensional factor structure can be considered to be analogous to a hierarchy. In this study, the overall dimension of perceived barriers would be at the top Of the hierarchy, and various subdimensions (to be explained later) would be subsumed under the overall perceived barriers dimension. Although a review of the literature suggests the possibility Of a multidimensional structure, there is a lacuna in the literature testing this possibility. Thus, the goal of this study is to develop and test a multidimensional factor structure of the overall factor of perceived barriers. i ° f e Per 'ved Barriers Dimension of the Health Belief M e1 Numerous studies have indicated that the perceived barriers dimension of the HBM is often the most significant dimension Of the model (Champion, 1992; Janz & Becker, 1984; Sereno & Dunn, 1994). However, as noted, there is a lack of a well- developed conceptualiution of the perceived barriers dimension in the literature (Melnyk, 1988). Moreover, a review of recent literature regarding barriers to health care uncovered “considerable confirsion regarding the barrier variable bOth theoretically and empirically, which apparently results fi'om a lack of methodological rigor in defining and Operationalizing the concept” (Melnyk, 1988, p. 196). Barriers specific to the health field and to HIV/AIDS issues have been conceptualized in various ways across a multitude of studies. Researchers have examined the potential negative aspects of a particular health action such as pain or diflicrrlty involved in performing a health action such as receiving an immunization (Janz & Becker, 1984); medication side efl‘ects as barriers to patient compliance (Kelly, Marnon, & Scott, 8 1987); the costs Of treatment, in addition to varying cultural meanings of illness (F abrega, 1977); the costs Of money, time, and emotional energy (Jones, Jones, & Katz, 1988); the costs of taking action regarding one’s health (Calnan & Rutter, 1986; Witte & Morrison, 1995); lack of access to preventive health measures, such as condoms (Cameron, Witte, Lapinski, & Nzyuko, 1996); racism (Comely, 1976); a perceived lack of information regarding AIDS (Gerbert, Maguire, Bleecker, Coates, & McPhee, 1991); students’ stereotypes of patients as a barrier to clinical decision making (Johnson, Kurtz, Tomlinson, & Howe, 1986); cultural barriers, ranging fiom cultural heritage (e. g., belief in spiritism or witchcraft), to language barriers (Clark, 1983; Ruiz, 1985); sociocultural barriers (Quesada & Heller, 1977); poverty and social isolation as barriers to effective AIDS prevention (Bowser, 1992); a lack of skills Of how to practice safer sex (Chesney, 1994); dificulty in innovation dissemination and implementation (Orlandi, 1987); psychosocial benefits to unsafe sex acting as barriers to safer sex (Soho, 1993); condom usage as a barrier to sexual firlfillment (Allen et al., 1992); the influence of sensation seeking and the connected desire to take risks while engaging in sexual activity (Sheer & Cline, 1995); negative perceptions toward condoms and condom use, including perceptions Of embarrassment or repulsion (Cline, Freeman, & Johnson, 1990; Sheer & Cline, 1994); . personal, interpersonal, and social norms barriers acting upon an individuals desire to practice safer sex (Serene & Dunn, 1994); fears of loss of partner’s trust or feelings of embarrassment associated with requesting and using condoms (Choi, Rickman, & Catania, 1994); barriers to condom use seen as hindering or affecting pleasure, intimacy, partner’s perception, fiiends’ perceptions, communication, and perceived need of condom use in a 9 sexual relationship (W endt & Solomon, 1995); and the role Of response and self-eflicacy in decisions regarding condom use (W itte, 1992; Witte, Berkowitz, Cameron, & McKeon, 1995). Based upon such studies, it appears as though the dimension of perceived barriers is considered by researchers tO be one worthy Of study, especially as the HBM dimension of perceived barriers has been repeatedly shown to be significant (Champion, 1992; Janz & Becker, 1984; Sereno & Dunn, 1994). However, what the literature is lacking is a common structure and conceptualization Of the dimension of perceived barriers, the dimension that arose fi'om the HBM. Thus, firrther research investigating a conceptualization or fiamework of specific barriers to preventive health is warranted. Numerous studies have examined the effect of perceived barriers to condom use as a safer sex practice (Cameron et al., 1996; Sheer & Cline, 1994, 1995; Sereno & Dunn, 1994; Wendt & Solomon, 1995; \Vrtte, 1992;1Vrtte et al., 1995). Although HIV/AIDS prevention campaigns appear to be increasing knowledge Of HIV/AIDS, research indicates that individuals continue to engage in risky sexual practices (Baldwin & Baldwin, 1988; Fisher & Misovich, 1990; Sheer & Cline, 1994, 1995). In an attempt to better understand the dimension of perceived barriers, this study will examine perceived barriers to condom use when engaging in sexual activity. As this dimension Of perceived barriers appears to be significant, perhaps a categorization and specified measurement model of perceived barriers can enhance future campaigns so that greater self-protective behavior change results. Indeed, if the perceived barriers dimension is as powerful as it appears to be, perhaps firture prevention campaigns should focus more on diminishing perceived barriers to the recommended response than on individuals’ perceived susceptibility and perceived severity of the specific disease and perceived benefits of performing the recommended 10 response. Such a focus on perceived barriers may greatly enhance future campaigns directed at the prevention of health threats and risks. Pr B ' Framew rk This study proposed that the dimension of perceived barriers in the HBM is multidimensional; specifically, that there are six separate dimensions underlying the overall construct of perceived barriers. First, a description and conceptualization of each of these six proposed dimensions is Ofl‘ered. Second, a description of second-order unidimensionality, as it will be hypothesized that these six dimensions form a second-order unidimensional factor structure, is presented. Third, a description of how factor analytic procedures may be used to test such a model is ofl‘ered. Finally, hypotheses regarding the proposed model are presented. Pr B ° Dim ions Based upon a carefirl review of the literature, it was determined that the various barriers suggested in past research could be subsumed in six dimensions: perceived individual psychological barriers, perceived relational psychological baniers, physical barriers, knowledge-based/self-eficacy barriers, social norms/cultural values barriers, and structural barriers. These six dimensions are conceptualized as follows. Perceived individual psychological barriers to condom use (hereafter referred to as individual barriers) afl‘ect the individual personally at a cognitive level, either through an individual’s beliefs and attitudes (e. g., “I am not likely to use a condom because condoms spoil the mood”) or through an individual’s feelings about him/herself (e. g., “I am likely to use a condom because I am not embarrassed to use a condom”). Past research indicating a focus on such individual barriers includes that Of Sheer and Cline (1994, 1995), Choi, l l Rickman, and Catania (1994), and Sereno and Dunn (1994). Perceived relational psychological barriers to condom use (hereafter referred to as relational barriers) are either those barriers that focus on the individual’s perceptions of the relationship (e. g., “I am not likely to use a condom because my partner will not trust me if I suggest condom use”) or perceptions about one’s partner (e.g., “I am not likely to use a condom because my partner is not infected”). Choi, Rickrnarr, and Catania (1994), Sereno and Dunn ( 1994), and Wendt and Solomon (1995) have explored such relational issues as barriers to safer sex. Physical baniers to condom use are those that afl‘ect the individual in some physical sense (e.g., “I am not likely to use a condom because condoms reduce sensation”). The work of Allen et al. (1992) and Wendt and Solomon (1995) examines such physical barriers to safer sex practices. Knowledge-based/self-eficacy barriers to condom use are those that afi‘ect the individual through the individual’s knowledge (or lack of knowledge) and beliefs regarding HIV/AIDS related issues (e. g., “I am not likely to use a condom because condoms carry HIV”) and perceived self-eficacy barriers to an individual’s use Of condoms (e.g., “I am not likely to use a condom because I have never been taught how to use condoms”). Past research focusing on knowledge-based or self-eflicacy issues regarding safer sex includes that Of Chesney (1994), Gerbert et al. (1991), Witte (1992), and “frtte et al. (1995). Social norms/cultural values barriers to condom use are those that address the individual’s perceptions of his/her culture and the specific social practices of that culture (e.g., “I am not likely to use a condom because to use a condom would suggest that I am a prostitute”), as well as religious values that the individual may perceive to be social norms (e. g., “I am not likely to use a condom because my religion discourages 12 the use of condoms”). Researchers who have examined barriers similar to this dimension include Quesada and Heller (197 7) and Sereno and Dunn (1994). Finally, structural barriers to condom use are conceptualized as those barriers that exist outside of the individual, yet these barriers hinder the individual’s effort to perform the action of engaging in condom use (e.g., “I am not likely to use a condom because condoms break.” “I am not likely to use a condom because I cannot afford condoms”). The research of Bowser (1992), Calnan and Rutter (1986), Fabrega (1977), Jones, Jones, and Katz (1988), and Witte and Morrison (1995) has examined such a dimension of structural barriers. Thus, the six proposed dimensions were developed by attending to the various dimensions of perceived barriers that have been studied in past research. F r i The above six factors of individual barriers, relational barriers, physical barriers, knowledge-based/self-eficacy barriers, social norms/cultural values barriers, and structural barriers are hypothesized to fit a factor structure of second-order unidirnensionality (see Figure l). A second-order unidimensional model is a measurement model which specifies a relationship between the underlying variables and the general construct being measured (Hunter & Gerbing, 1982). In order to understand such a model, one should first be familiar with the idea of a factor structure. A factor can be measured by a number of. items. For example, in this study, individual barriers to condom use constitutes a factor. When subjects indicated their individual barriers to condom use, they were asked to respond to multigle items which were intended to measure this factor of individual barriers (specifics regarding the development Of the scale are found in Chapter 4). Similarly, multiple items were used to measure eaih of the six proposed factors Of perceived barriers 13 to condom use. Thus, six scales, each consisting Of multiple items, were used to measure these six factors. Through such a process, six separate factors, all relating to perceived barriers to condom use, were hypothesized to exist. In order to develop a second-order unidimensional model, one must apply the factor analysis procedure to a correlation matrix at a difl‘erent level (Hunter & Gerbing, 1982). At this level of analysis, one takes the six factors that have been developed and considers each of the factors to be a single item measuring an overall factor (here, the overall factor is perceived barriers to condom use). Specifically, one now uses the correlations between the six [39193 to assess if each of these factors can be considered to underlie the overall factor of perceived barriers to condom use. Here, the “scale” measuring perceived barriers to condom use consists Of six items (the six factors previously established). A second-order unidimensional model is consistent with the data when the correlations among the factors are subjected to confirmatory factor analysis and fit a single-factor model (Hunter & Gerbing, 1982). Thus, these underlying factors are here being treated as items, and are then subjected to confirmatory factor analysis procedures. 14 Figure 1: Proposed Second-Order Unidimensional Factor Model of Perceived Barriers to Condom Use Perceived Barriers tO Condom Use Indivrdual Physical Social orms/ Psychological Barriers Cultural Values Barri s Barriers U1 U2 U. W1 W2... wh Y1 Y2... Y,- J. ., Perceived Knowledge- Structural Relational Based/ Self- Barriers Psychological Efl'rcacy Barriers Barriers V1 V2 V8 X1 X2 X Zl Z2 1 1'1 i ll 1 l". evr 6-12.. evg 6x1 6x2... exi 621 622-.- ezk 15 Confirmatory factor analysis is a statistical technique that allows one to assess the validity of a scale through two separate tests: internal consistency and parallelism (Hunter & Gerbing, 1982). Internal consistency is a test used to determine if the Obtained correlations between items in the scale (the items measuring the particular factor) are as the factor model would predict (Hunter, 1977). In order to determine internal consistency, one compares the Observed correlations with the expected correlations and determines the amount of error (Hunter, 1977). The observed correlations are the correlations among the items that one Obtains. The expected correlations are calculated by using the internal consistency theorem and determining the product Of the factor loadings (the qualities Of the items, or the correlation between the item and its true score in the population) (Hunter, 1977). The larger the difference between the Observed and expected correlations, the larger the error. Parallelism is a test to Observe the relationships between items and other factors, i.e., factors that the item is not intended to measure. Again, expected and Observed correlations are viewed, and error is calculated. Expected correlations, when testing for parallelism, are calculated using the parallelism theorem, where the product of the factor loadings Of m of the items is Obtained, and then multiplied by the correlation between the two separate factors that each item is purported to measure (Hunter, 1977). W In this study, it is hypothesized that six factors underlie the overall dimension of perceived barriers to condom use. Therefore, the following hypothesis is advanced: 16 H1: The proposed factor structure of the perceived baniers dimension suggested by the Health Belief Model will fit a second-order unidimensional factor structure, with the factors Of individual barriers, relational barriers, physical barriers, knowledge-based/self-eflicacy barriers, social norms/cultural values barriers, and structural barriers underlying the factor Of perceived barriers to condom use (see Figure 1). In addition, it is expected that the proposed factor structure will be replicated in cross- cultural tests; that is, that the model will fit the data when tested through confirmatory factor analysis with two culturally different samples (in this study, the data were Obtained from subjects fi'om both the United States and the Kenyan cultures). Therefore, the following hypotheses are proposed: H2: The proposed factor structure will replicate cross-culturally such that: H2a: When data from the United States sample only are used, the proposed factor structure will be replicated. H2b: When data fi'om the Kenyan sample only are used, the proposed factor structure will be replicated. Chapter 3 NIETHODS This study will serve as a measurement study to test the proposed second-order unidimensional factor structure of the perceived barriers dimension of the HBM, and to determine if such a factor structure is replicable cross-culturally. The specific concerns of this study, then, are twofold: (1) to determine if the proposed second-order unidimensional factor structure is consistent with the data Obtained when these barriers are subjected to factor analysis procedures, and (2) to determine if the proposed second-order unidimensional factor structure is consistent with the data Obtained from two culturally distinct samples. P 'ci ts The total sample consisted of 365 college students. The subsamples were collected at two large universities: one large university located in the midwestern United States (N=178), and one large university located in a major city in Kenya CN=1 87). One hundred and sixty-four males and 200 females participated in this study. The ages of the participants ranged fi'om 19 to 46 (M = 22.03). Due to missing data in a number Of the questionnaires (subjects not answering all items in the questionnaire), the average number 17 18 Of subjects in the analyses used in this study was N=320 (approximately N = 160 for each culture). Students in the United States sample were enrolled in a variety of undergraduate communication classes, both upper and lower levels, and received extra credit for their participation in this study. Students in the Kenyan sample were enrolled in a sociology course and were asked to complete the questionnaire by their instructor. The questionnaires were completely anonymous and confidentiaL as no personal identification appeared at any place on the questionnaire. After completing the questionnaire, which took approximately 10—1 5 minutes to complete, the participants were thanked, debriefed, and provided with a handout about HIV/AIDS, which included information about the disease and preventive action that can be taken to avoid contraction of the disease.1 Ligament The questionnaire consisted of 119 items, 76 of which specifically related to perceived barriers to condom use. The remaining items on the questionnaire pertained to demographic information; questions regarding the subjects’ sexual experiences and sexual practices; reports of intended and actual condom usage by the participants; items intended to measure subjects’ ability to delay gratification; perceptions of susceptibility to and severity of HIV/AIDS, and perceptions of benefits to condom use (the other dimensions of the HBM); questions regarding an individual’s exposure to individuals (1) infected with HIV, (2) diagnosed as having AIDS, and (3) having died of AIDS; and an individual’s HIV status, including questions as to whether or not the participants had been tested for HIV. Developmgrt of the items for the overall questionnaire. In order to develop the items to be used to assess perceived barriers to condom use, a variety of methods were 19 used. First, a literature search was performed to become familiar with barriers that other researchers may have addressed when conducting studies that related to existing barriers to condom use. Second, participants at a Health Communication Conference that focused upon “Communicating Health with Unique Populations”2 were asked to complete a survey that asked them to list reasons that they could think Of or that they may have heard voiced by their clients (as many Of the participants were actively working in the fields of HIV/AIDS counseling and education) as to why individuals may choose not to use a condom when engaging in sexual encounters.3 Third, Sexual Health counselors at the midwestem university where a subsample of the data were collected were also asked to complete the above-mentioned survey. Fourth, during an AIDS prevention project in Kenya in the summer of 1995, focus groups were conducted with native Kenyans along the Trans-Afiica Highway, and those participants, as well as the health workers with whom we spoke, were asked to provide reasons as to why they may choose not to use condoms during sexual activity.‘ Following a process suggested by Hunter and Gerbing (1982), after the collection Of the numerous barriers items from these cross-cultural groups, the items were placed into a priori clusters by the researcher through an evaluation Of content of the items. As noted earlier, these categories were developed through careful analysis Of the literature. Wm. For purposes Of this study, the culture of the participants was determined to be that culture in which the individuals currently resided. Therefore, those individuals who completed the questionnaire at the large midwestem university were coded as “United States culture” and those who completed the questionnaire at the large university in the major Kenyan city were coded as “Kenyan culture.” The questionnaire 20 did ask for individuals to indicate their citizenship; 91% of those individuals completing the questionnaire in the United States identified “American” or “United States” as their citizenship and 97.3% Of those individuals completing the questionnaire in Kenya indicated “Kenyan” as their citizenship.’ Thus, it was believed that separating those individuals by culture based upon the culture in which they currently resided was warranted. Pilot tesnn' g. Once the barrier items had been solicited from the various sources described above, they were developed into statements to be included in the questionnaire. The questionnaire was then pilot tested to ensure that the individual items were ones that subjects could understand. Five individuals were administered the questionnaire, and were asked to complete the questionnaire as if they were a subject, but also to indicate if any items were confusing or otherwise diflicult to comprehend. After completing the questionnaire, each of the pilot subjects spoke individually with the researcher and ofl‘ered feedback on the questionnaire. The suggestions were discussed, and, when appropriate, the questionnaire was altered to reflect this input from the pilot subjects (e.g., word choice was altered on a few items to ensure that the question would be understood by subjects). In addition, the questionnaire was read by two native Kenyan individuals to ensure that the questionnaire was appropriate for Kenyan subjects. The questionnaire for both samples (United States and Kenyan) was administered in English.‘5 D is The 76 perceived barriers items were separated into the six a priori clusters suggested by this study. These clusters, and the items that were believed to be measures of each of these clusters, had been determined and assigned before the data analysis and 21 were thus imposed upon the data. Confirmatory factor analysis procedures were used to analyze the data. For this study, data analysis consisted Of four steps: (1) tests of internal consistency of each of the six proposed perceived barriers factors in order to develop an internally consistent scale for each of the six factors, (2) tests of parallelism among the six scales developed to measure each barrier factor, (3) a test Of second-order unidimensionality (using the factor correlations between the six factors as the items to measure the overall perceived barriers construct) to complete the test of Hypothesis 1, and (4) testing Hypotheses 2a and 2b to determine if the proposed factor structure is replicable across cultures. Thus, confirmatory factor analysis procedures were used to analyze these data and test the proposed hypotheses. Chapter 4 RESULTS The results Obtained from this data analysis will be presented in order to parallel the data analysis procedures. Specifically, internal consistency Of the six proposed underlying factors will be discussed, followed by a discussion of tests of parallelism among these six factors. As will be seen, firrther analyses were necessary due to the results Obtained. These analyses will be discussed and the results are presented. As noted earlier, the 76 original barrier items were separated into six distinct a priori clusters. These clusters are the six distinct factors that have been described above. Following is a description Of each cluster, along with a description of the process through which the original 76 items were reduced to 36 total items, which were then used to test the hypotheses Of a second-order unidimensional factor structure model to be replicable across cultures. him Consisteng Analysis Each Of the six proposed factors Of perceived barriers to condom use were subjected to analyses Of internal consistency in order to ascertain that the multiple items were indeed measuring the factor they were intended to measure. Following is a 22 23 description of each factor, as well as a description of the process through which the 76 original items were reduced to a 36-item scale. Ing'vidual bm’gs to condom use. Initially, the proposed scale to measure individual barriers to condom use consisted Of 23 items (see Appendix A), generated by practitioners, counselors, students, and other individuals, as noted above. Participants were asked to respond to the statements using a five-point Likert scale, where 1 was i “Strongly Disagree” and 5 was “Strongly Agree.” Due to the fact that most of the items were phrased negatively (i.e., “I am NOT likely to use a condom because...”), to circle a 5 meant that the participant felt that the item did indeed pose some sort of barrier to condom use. Seven of the original 23 items were recoded as needed so that to strongly agree with an item indicated that one perceived the item to be a barrier to condom use. When subjected to confirmatory factor analyses, the items that were intended to measure individual barriers to condom use failed in tests of internal consistency, suggesting that the 23 items did not measure only one factor. The items were then analyzed again for content validity, and some items were removed as they did not, in post hoc assessment of the items, appear to be a good fit for the factor of individual barriers to condom use. Other items had been questioned as to their meaning by the participants (e.g., item 12, “I am likely to use a condom because using condoms is masculine”) and were thus deleted from scale. Finally, using confirmatory factor analysis procedures and Observing tests of internal consistency, other items were deleted so that the scale used to measure perceived individual barriers to condom use loaded on one factor (the factor Of individual barriers) within sampling error (further discussion of sampling error below). The original 23 items, along with their means and standard deviations are reported in 24 Appendix A Following the elimination of items based on the analysis described above, the final scale of perceived individual barriers to condom use consisted of six items (the items are in bold in Appendix A). These six items provided a scale with moderately low reliability (or = .67). The scale fit a single-factor model. The errors obtained (by calculating the difl‘erence between the expected and Observed correlations, as is done for tests Of internal consistency) all fell within the range Of sampling error. Specifically, the percent of items for which the error was greater than sampling error was 0.00% (see items 1-6 in Table l for Observed correlations and factor loadings; see items 1-6 in Table 2 for expected correlations and residuals). These six items were then used as the scale in the proposed six-factor model to measure individual barriers to condom use. W. Using the same procedures as described above regarding individual barriers to condom use, a scale to measure relational barriers to condom use was developed and analyzed. The original scale (using items placed in the a priori category of relational barriers tO condom use) consisted of 13 items (see Appendix B for a listing of the items, their means, and standard deviations). Three of the original 13 items were recoded as needed. The original 13 items failed to fit a single factor. After further analysis based on content validity and factor analyses, as described above, this scale was reduced to four items (those items in bold in Appendix B). The reliability of this four-item scale measuring relational barriers to condom use was fair at or = .75. In addition, these four items also loaded on a single factor with the percent of items for which the error was greater than sampling error again being 0.00% (see items 7-10 in Tables 1 and 2). These four items, then, were used as a scale to measure relational barriers to condom use. 25 Phyg'gg barriers tO condom use. The scale to measure physical barriers to condom use was developed using the same procedures. Originally, the scale intended to measure physical barriers to condom use consisted Of nine items (see Appendix C for a listing of the items, their means, and standard deviations). Four of the original nine items were recoded as needed. The original nine items failed to fit a single factor. Upon further analysis, using content validity and factor analysis, this scale was reduced to 7 items (those items in bold in Appendix C). The reliability of this seven-item scale measuring physical barriers to condom use was low at or = .62. These seven items loaded on a single factor, establishing internal consistency with the percent of items for which the Obtained error was greater than sample error being 4.76% (see items 11-17 in Tables 1 and 2). These seven items were then taken to be a scale measuring physical barriers to condom use. Knowlfige—baseflself-eflicagy barriers to condom um. Originally, the proposed scale to measure knowledge-based/self-eflicacy barriers to condom use consisted of nine items (see Appendix D for a listing of the items, their means, and standard deviations). Four Of the nine items were recoded. As the nine items failed to fit a single factor, content validity and internal consistency were evaluated, which served to reduce the scale to seven items (those items in bold in Appendix D). The reliability Of this seven-item scale measuring knowledge-based/self-eflicacy barriers to condom use was low at a = .65. These seven items fit a single factor with the percent of items for which the error was greater than sampling error being 0.00% (see items 18-24 in Tables 1 and 2). Thus, the scale to measure knowledge-based/self-eflicacy barriers consisted of seven items. 26 Table l The Observed Correlations and Factor Loading Matrix for the Proposed Six-Factor Model“ 12 345 67 8 9101112131415161718 1 3_0 2 31 33 3 35 38 fl 4 20 25 25 15 5 21 20 31 08 Q 6 29 28 28 26 18 26 7 36 39 50 26 23 36% 8 36 29 39 25 34 32 524_5_ 9 36 38 42 22 21 37 57 5261 10 17 32 29 12 24 24 33 29 37 _2_0 ll 39 26 27 12 26 20 22 22 29 14 09 12 -03241219040009-011211020§ 13 21 55 35 19 27 244434 35 30 27 2764; 14 ll 10 10 27 03 17 12 08 10 08 -0212 09 Q 15 34 40 37 26 20 30 46 33 43 37 15 24 48 14 32 l6 14 45 290926 25 34 33 30 37 24 19 56 06 34g l7 18 20 27 18 15 27 32 30 29 17 18 -05 20 08 21 26 19 18 47 35 29 18 24 19 25 24 18 07 27 07 22 07 27 19 16 24 19 34 25 36 l6 16 27 34 29 26 20 14 ~03 19 10 25 13 14 22 20 24 l9 17 34 19 21 12 19 08 16 07 09 09 22 18 17 08 22 21 37 18 33 12 40 31 37 35 32 l9 19 01 27 -0125 14 22 21 22 17 29 20 35 27 29 14 22 24 28 13 24 22 16 29 20 22 24 23 17 20 15 14 12 25 l8 17 27 05 09 08 07 05 33 19 22 18 24 29 29 34 20 34 27 33 31 33 26 17 00 23 -02 34 20 18 26 25 54 28 46 16 29 25 46 41 35 21 45 -0428 06 30 23 26 33 26 18 23 19 43 21 25 13 13 15 19 15 18 13 25 24 15 10 19 27 31 31 32 34 13 28 35 38 47 21 l7 16 27 30 42 31 34 18 28 40 19 23 21 37 40 25 24 35 21 27 06 25 08 29 22 28 29 29 30 20 33 24 29 21 36 31 37 20 18 02 21 09 28 20 16 17 30 34 31 38 25 25 30 47 50 60 33 23 02 34 06 35 23 34 10 31 39 25 37 19 29 30 46 50 45 28 28 01 22 12 27 24 34 25 32 31 26 34 12 21 22 31 32 26 16 21 -02 22 -02 23 25 21 31 33 34 18 22 12 30 25 26 33 38 29 24 -0421 08 18 23 28 16 34 04071115061200-0108071120041114100208 35 22 27 30 18 29 28 36 40 39 19 26 ~13 29 04 21 20 42 27 36 22 25 32 15 15 28 40 42 41 24 18 06 39 13 19 26 27 14 I 545866393751696464454918592661494157 R 4752 60 32 384975677845 33 12 54 146051 41 28 P 43705641'3946635160493028801662673140 KB 6455 5746545654 55 52 38 33 1440 18 6038 3849 SN 62 45 58 46 46 50 63 62 69 41 44 10 43 24 54 40 46 38 S 56 5164365057 66 72 75 47 50 03 57 174752 6048 27 Table 1 (continued) 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 19 g 20 17 E 21 29 14 33 22 07 28 20 Q 23 10 12 26 13 E 24 24 24 42 27 12 35 25 30 24 35 20 14 33 Q 26 18 42 08 35 13 24 17 l_l 27 27 20 23 23 21 20 31 23 2.2 28 19 30 38 24 23 31 30 22 31 Q 29 21 16 49 20 28 51 33 25 23 38 3; 30 33 21 43 21 14 33 43 19 40 37 28 g 31 28 25 54 21 19 51 40 15 36 36 45 55 48 32 26 134015 20 29 43 12 22 18 30 27 30H 33 210938 142625 28 08 24 31 21 35 3021& 34 10 20 02 23 02 03 01 23 13 09 07 03 -08 -02 02 35 24 10 46 23 16 34 30 08 26 33 35 45 52 21 31 36 2010 2415 10 2125 14 33 20 22 37 29 26 23 -05 30 E I 514456 52 34 57 65 49 56 59 52 60 59 48 46 18 51 45 R 41 21 46 33 25 46 54 23 53 40 47 726440 48 05 51 55 P 29 29 34 47 33 35 49 38 63 46 36 5047 34 38 23 41 47 KB394258433159595047606355705446215636 SN 444563413361573354575768704645125845 S 503174453756633259555773664151006148 I R P KB SN S 100 92 97 109 101 103 92 100 84 75 89 99 97 84 100 77 83 91 109 75 77 100 102 106 101 89 83 102 100 102 103 99 91 106 102 100 raga-owi- ‘The underlined numbers in the diagonal show the reliabilities for each item (without decimals). Thefactorloadings completethematrirebolded loadings indicateanitemispartofthefactor where I = individual barriers, R = relational barriers, P = physical barriers, KB = knowledge- based/sclf-emcacy barriers, SN = social nouns/cultural values barriers, and S = structural barriers. 28 Table 2 Th E Err r orrelation Matrix for the Pro osed Six-Factor Model" 12 3 4 5 6 7 8 9101112131415161718 00 01 01 01 01 01 03 03 05 23 18 21 02 02 21 02 19 31 0002 01 02 01 0703 08 1008 1001 05 070304 36 38 01 07 06 05 01 05 01 08 05 16 01 03 14 08 06 21 23 26 06 06 01 01 06 05 00 08 11 21 03 16 06 03 20212414 01021106081506020302020404 28 30 34 20 19 01 01 00 03 05 14 16 O9 01 08 11 08 374045272535 0201000409060207081302 33 36 40 24 23 31 50 00 01 05 17 ll 01 02 05 12 01 3941 4728273758 52 021006170003140911 10 2224 28 17 16 21 33 30 35 0200000213120510 11 16161912111518171912 0603 0704040915 12 15 16 17 ll 10 14 18 16 18 ll 08 05 08 07 00 14 04 13 42 45 51 30 29 40 50 45 52 30 24 22 04 02 02 05 08 14 09091106060810091006050413 04050301 15 32 35 40 23 22 31 39 35 40 24 l9 17 50 10 08 02 04 16 353843252433423844252019541142 0506 17 16 17 19 12 ll l6 l9 18 20 12 09 09 25 05 19 21 04 18 28 31 35 21 20 27 27 25 29 17 12 ll 30 06 23 25 12 19 23 25 28 16 15 22 22 20 23 14 09 08 24 05 18 20 09 19 20 25 26 31 17 17 23 23 21 25 14 10 09 26 05 20 22 10 21 21 34 37 41 25 23 33 32 29 34 20 13 12 35 07 28 30 14 28 22 25 27 31 19 17 24 24 22 26 14 10 09 26 05 21 22 10 21 23 19 20 22 13 12 17 17 l6 18 ll 07 07 19 04 15 16 08 15 24 35 37 43 25 24 33 33 30 35 20 14 13 36 07 28 31 14 29 25 31 33 38 22 21 29 37 34 39 23 14 13 39 08 29 32 15 29 26 18 19 22 13 12 17 21 20 23 13 08 08 22 04 17 18 08 16 27 29 31 36 21 20 28 36 32 37 21 l3 13 36 08 28 30 14 27 28 31 33 38 22 21 29 37 34 39 23 14 13 39 08 29 32 15 29 29 31 33 38 22 21 29 37 34 39 23 l4 13 39 08 29 32 15 29 30 37 39 45 27 25 35 45 41 47 28 17 16 45 O9 35 39 18 34 31 38 41 46 27 25 36 46 42 49 28 l8 17 47 O9 36 39 18 35 32 23 25 28 16 15 22 30 27 32 18 ll 10 30 06 23 25 12 21 33 29 31 34 21 20 27 38 34 4O 23 14 13 37 07 29 31 15 27 34 000000000000000000000000000000000000 35 34 36 41 25 24 32 45 41 48 27 16 15 45 09 35 37 17 32 36 27 29 33 20 19 26 36 33 38 22 l4 13 35 07 27 30 14 25 \OMQQM-bUNI—I 29 Table 2 (continued) 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 110103080206230002090103010805041205 000719020008050400141308160113070904 081408110709080304150507090613111101 001713160105063013010202080409150705 010217100010080907160800030610060504 050202050806040800110805060002120402 121105100100090801120102000112000904 0902060001 01 070706 1003 090805 01 01 01 09 031702020902040810040213040602080903 10 060201140606020600020305000206070802 11050306030203310704130406101010111004 12 110011150113171003071114161217202807 13 051708041213110909141811250816041604 14 051708110109022122000103030801110506 15 070203081806010714000100090011141408 16 070516020311090301101216150008101704 17 050208121404110220130116160913022513 18 0301070303030403090012241010110805ll 19 0106100201080506030105000900100100 20 16 1010 010100 28 03 06 08 09 05 0513 201812 212324 050808011109041503121506020906 06 04 02 \OOOQONM-bWNI-i 22 17 18 25 00 02 21 00 02 06 09 10 04 09 23 05 07 2312131813 06 0304051007030609020406 24 2325342518 05 130416 08 090407 030410 25 222434261835 0200020104002002010604 26 13 14 19 14 10 19 19 05 03 06 03 08 02 09 23 12 02 27 21 23 32 23 17 33 31 18 00 08 03 02 00 05 13 08 06 28 22 24 34 26 18 35 32 19 31 06 02 04 05 01 09 03 09 29 22 24 34 26 18 35 32 19 31 32 11 05 07 09 07 01 07 30 28 30 40 30 21 41 39 22 37 39 39 07 02 01 03 03 03 31 28 30 42 31 22 42 40 23 38 40 40 48 00 07 08 08 06 32 17 18 25 19 14 25 23 14 22 23 23 29 30 00 02 04 06 33 21 22 32 23 17 32 30 17 29 30 30 36 37 21 02 00 02 34 000000000000000000000000000000 0405 35 25 28 3728 2038 3620 34 36 36424425 3100 00 36 20 22 30 22 16 31 29 16 27 29 29 34 35 20 25 00 30 I"In this matrix, the lower diagonal presents the expected/predicted correlations (without decimals) for items of all six factors; the upper diagonal presents the residuals. 30 Social norms/cultural values barriers to con_dom use. On the questionnaire, nine items were intended to measure social norms/cultural values barriers to condom use (see Appendix E for a listing of the items, their means, and standard deviations). Three Of the original nine items were recoded as needed. These nine items failed to fit a single factor. After analyzing content validity and performing a factor analysis on these nine items, this scale was reduced to seven items (those items in bold in Appendix E). The reliability of this seven-item scale measuring social norms/cultural values barriers to condom use was fair at or = .77. These seven items loaded on a single factor, where the percent of those items for which the error was greater than sampling error was 4.76% (see items 25-31 in Tables 1 and 2). Social norms/cultural values barriers to condom use were then measured by this seven-item scale. SEEM barriers to condom use. The scale to measure structural baniers to condom use was developed with the same procedures as outlined above. The original scale consisted of 13 items (see Appendix F for a listing of the items, their means, and standard deviations). Four of the 13 items were recoded as needed. The proposed scale, consisting of the 13 items, failed to fit a single factor. Following analyses of content validity and internal consistency, this scale was reduced to five items (those items in bold in Appendix F). These five items provided a scale with low reliability at or = .47. The five-item scale fit a single-factor model with the percent Of items for which the error was greater than sampling error being 0.00% (see items 32-36 in Tables 1 and 2). These five items were then used as the scale in the proposed six-factor model to measure structural barriers to condom use. 31 si for Parallelism As Hypothesis 1 proposed a second-order unidimensional model, in addition to the tests of internal consistency Of the six factors performed above, tests for parallelism of the six-factor model (all 36 barrier items) were performed in order to begin to test Hypothesis 1 (see Appendix G for a listing of the 36 items used for further analyses). Using the six specified factors, which fit the model in terms of internal consistency, tests Of parallelism indicated that the percent of items for which error was greater than sampling error was 21.64%, indicating that the proposed six factor barrier scale failed to fit the model. As these six dimensions failed to fit a first-order factor structure, it was impossible to complete the analysis for Hypothesis 1, to determine if these items fit a second-order unidimensional factor structure with the six barrier factors underlying the factor of perceived barriers. In order to test that hypothesis, one would take the correlations among the six factors (the factor correlation matrix) and subject that matrix to a factor analysis. In so doing, the individual factors would have replaced the scale items as measures of the overall factor, the procedure that is followed to test a hypothesis Of second-order unidimensionality (Hunter & Gerbing, 1982). However, as the 36 items used to test the notion of a six-factor first-order structure failed to fit that model, the test for second-order unidimensionality was rendered irrelevant in this case. Thus, the data were not consistent with the hypothesized model. Hypotheses 2a and 2b were contingent upon the data being consistent with Hypothesis 1, as Hypotheses 2a and 2b hypothesized that the proposed factor structure 32 was replicable cross-culturally. Due to the lack Of consistency of the data with Hypothesis 1, Hypotheses 2a and 2b were, in effect, rendered irrelevant and could not be tested. Further Analysis As none of the hypotheses could be adequately tested based upon the result that a confirmatory factor analysis failed to find the expected first-order factor structure with the proposed six factors, a post hoc hypothesis was Offered. Although the six-factor model failed in tests of parallelism, the success of the six factors in tests of internal consistency indicate that there may indeed be multiple dimensions underlying the overall perceived barriers dimension. Thus, the post hoc hypothesis suggested is a revision of Hypothesis 1: Post Hoc Hypothesis 1a: The perceived barriers dimension suggested by the Health Belief Model will fit a second-order unidimensional factor structure, with multiple factors underlying the overall factor of perceived barriers to condom use. In order to test this post hoc hypothesis, confirmatory factor analysis procedures were again employed. Upon careful analysis of the six factors originally proposed, it appeared that two of the six proposed factors (physical barriers and knowledge-based/self— efl'rcacy barriers) were causing the model to fail in tests of parallelism. Therefore, afier repeated attempts to reduce the number of items fi'om the scales measuring these two factors, it was decided that these two factors would be removed entirely from further tests Of the factor model. Following the removal of these two factors, slight alterations were made to the remaining four factors (removal Of individual items fiom these remaining four factors) and a l6-item scale that was consistent with a four-factor model was found to fit the data. 33 The four remaining factors were those Of individual barriers, relational barriers, social norms/cultural values barriers, and structural barriers to condom use. Each of these four factors was measured by four items, thus allowing one to perform tests of both internal consistency and parallelism. The 16 items used in this four-factor analysis, along with their means and standard deviations, are reported in Appendix H. Internal consisteng analysis of the four-factor model. The four items used to measure the dimension of individual barriers to condom use (see Appendix H) provided a scale with low reliability (or = .60). The scale fit a single-factor model with the percent Of items for which error was greater than sampling error being 0.00% (see items 2, 3, 5, and 6 in Table 3 for Observed correlations and factor loadings, and in Table 4 for expected correlations and residuals). These four items were then used as the scale in the four-factor model to measure the dimension of individual barriers to condom use. The four items used to measure the dimension of relational barriers to condom use (see Appendix H) provided a scale with fair reliability (or = .75). The scale fit a single- factor model with the percent of items for which error was greater than sampling error being 0.00% (see items 7, 8, 9, and 10 in Table 3 for observed correlations and factor loadings, and in Table 4 for expected correlations and residuals). These four items were then used as the scale in the four-factor model to measure the dimension of relational barriers to condom use. The four items used tO measure the dimension of social norms/cultural values barriers to condom use (see Appendix H) provided a scale with low reliability (or = .62). 34 Table 3 Tho ngrved Correlations and Factor Loading Matrix for the Four-Factor Model“ 2356789102526273032333536 I R SNS 2 3_1 3 38_Z 5 2031_'Z 6 28281829 7 39502336; 8 2939343252_5 9 38 42 21 37 57 5291 10 32 29 24 24 33 29 37__9 25284629254641352199 2623192125131315191719 27 31 32 13 28 35 28 47 21 31 2339 30313825304750603343194041 32 26 34 21 22 31 32 26 1643 12 22 271_ 33 18 22 30 25 26 33 38 29 28 98 24 35 21_4 35 27 30 29 28 3640 39 19 30 08 26 45 21 313_ 36 25 32 15 28 40 42 41 24 25 14 33 37 26 23 30_8 I 566941457 06466526142505949455448 10095 97 97 R 5260384975 5 6 67 78 45 54 23 53 72 40 48 51 55 95 100 93 96 65 72 43 57 32 60 69 48 44 50 50 97 93 100 95 73 714462 21 52 7143495653 97 96 95 100 SN52 62 4150 6 S 48 59 47 516 " The numbers used to identify the items are from Appendix H. The underlined numbers in the diagonal show the reliabilities for each item (without decimals). The factor loadings complete the matrix, bolded loadings indicate an item is part of a factor where I = individual barriers, R = relational baniers, SN = social norms/cultural values barriers, and S = structural barriers. 35 Table 4 The Expectod Ed Error Correlation Matrix for the Four-Factor Model“ 2 3 5 6 7 8 9 10 25 26 27 30 32 33 35 36 2 0 3 3 0 7 4 8 3 6 2 6 3 8 3 4 3 38 3 3 2 5 7 0 8 2 8 7 6 10 7 3 5 23 28 0 6 8 9 7 7 8 11 2 4 11 7 6 6 25 31 18 5 3 4 5 0 11 2 0 4 4 4 5 7 39 48 29 31 2 1 0 7 9 6 0 0 9 3 3 8 36 44 26 29 50 0 l 6 7 1 7 4 1 4 7 9 42 50 30 33 58 52 2 6 8 3 11 7 2 3 2 10 24 29 17 19 33 30 35 3 6 4 4 2 8 5 1 25 31 38 22 25 39 35 41 24 1 3 4 19 1 0 4 26 17 21 13 14 22 20 23 13 18 4 3 l 7 9 2 27 33 40 24 26 41 37 44 25 34 19 l 3 4 6 3 30 37 45 27 30 47 43 49 29 39 22 41 1 4 9 3 32 23 28 17 18 31 28 3 18 24 13 25 28 0 3 3 33 26 32 19 21 35 32 36 21 27 15 28 31 21 4 3 35 30 37 22 24 39 36 42 24 30 17 32 36 24 27 0 36 29 35 21 23 37 35 39 23 29 16 30 34 23 26 30 * The numbers used to identify the items are fiom Appendix H. In this matrix, the lower diagonal presents the expected/predicted correlations (without decimals) for items Of the four factors; the upper diagonal presents the residuals. 36 The scale fit a single-factor model with the percent Of items for which error was greater than sampling error being 0.00% (see items 25, 26, 27, and 30 in Table 3 for Observed correlations and factor loadings, and in Table 4 for expected correlations and residuals). These four items were then used as the scale in the four-factor model to measure the dimension Of social norms/cultural values barriers tO condom use. The four items used to measure the dimension of structural barriers to condom use (see Appendix H) provided a scale with low reliability (or = .58). The scale fit a single- factor model with the percent of items for which error was greater than sampling error being 0.00% (see items 32, 33, 35, and 36 in Table 3 for observed correlations and factor loadings, and in Table 4 for expected correlations and residuals). These four items were then used as the scale in the four-factor model to measure the dimension of structural baniers to condom use. Maia for ogflolism for the four-factor model. An analysis for parallelism revealed that the percent of items for which error was greater than sampling error was 4.17%, indicating that the four-factor model fit the data. Using tests for parallelism allowing for variation in the quality of the items, the factor Of individual barriers revealed x2 (9, N = 320) = 4.12, p > .05, the factor of relational barriers revealed it” (9, N = 320) = 4.89, p > .05, the factor of social norms/cultural values barriers revealed x2 (9, N = 320) = 6.91, p > .05, and the factor of structural barriers revealed x2 (9, N = 320) = 1.89, p > .05. Thus, confirmatory factor analysis procedures indicated that such a four-factor model fit the data.7 Analysis for second-order unidimensi_orm. In order to test the hypothesis that such a four-factor model was second-order unidimensional, the factor correlation matrix 37 Obtained from confirmatory factor analysis of the four-factor model was entered as the matrix to be use in the subsequent factor analysis. In such a way, by using the correlations of the factors with each other, one is able to test for evidence of second-order unidimensionality (Hunter & Gerbing, 1982). Although each of the factors was highly correlated (see Table 3), they failed to fit a single-factor model, which would be consistent with the hypothesis of second-order unidimensionality. A possibility as to why such strongly correlated factors do not fit a single factor is that the items underlying each factor may constitute a Guttman simplex. Thus, an analysis was performed to determine if there was evidence for a Guttman simplex. In order to test if the data are consistent with the existence Of a Guttman simplex, the mean difl‘erence between the variables (the individual items) is correlated with the correlation of the variables (Guttman, 1955; Hunter & Boster, 1987). Ifa Guttman simplex exists, analysis should show a high negative correlation. NO evidence of a Guttman simplex was found in these data, the obtained correlation was -.22. Thus, although the l6-items used to measure perceived barriers to condom use (four items for each of the four proposed dimensions) fit a four-factor model, they failed to fit a second-order unidimensional model, as suggested by post-hoe Hypothesis la. These data were therefore not consistent with post hoc Hypothesis 1a. As noted, the correlations among the factors in the successful four-factor model (see Table 3), were quite strong. Such strong correlations indicated that the items underlying these four factors are not conceptually distinct; rather, they appear to be alternate indicators Of the same factor. In addition, the high correlations between the factors suggest problems of multicollinearity were the factors considered to be separate 38 factors. Although the four-factor model does fit the data, it appears as though a simpler model may fit the data as well. Thus, a one-factor model was tested. Andysis Of o ono-factor barrier model. Using the 16 items used in the successfirl four-factor analysis, confirmatory factor analysis procedures were used to test whether or not a simpler model (one factor) would fit the data. After the deletion of four of the 16 items, the remaining 12 items (those items in bold in Appendix H) formed a “perceived barrier scale” with good reliability ((1. = .85). The scale fit a single factor model with the percent of items for which error was greater than sampling error being 1.52% (see Table 5 for observed correlations and factor loadings, and Table 6 for expected correlations and residuals). groom gaysis of the factor model. As the proposed second-order unidimensional factor models (both the six-factor model and the four-factor model) failed to fit the data, tests of Hypotheses 2a and 2b were not performed on the six— and four- factor models. However, the sample was split and separate factor analysis were performed for the United States data and the Kenyan data to determine if the one-factor model (consisting of 12 items) was replicable in cross-cultural analysis. When data fi'om the United States sample were used, the one-factor analysis (with the 12-item scale providing good reliability, or = .87) failed to fit a one-factor model (see Table 7 for observed correlations and factor loadings, and Table 8 for expected correlations and residuals). The percent of items for which error was greater than sampling error was 12.12%. When the subsample consisting of the Kenyan data was analyzed, the one-factor model (with the 12-item scale providing good reliability, or = .80) fit a one-factor model of the data (see Table 9 for Observed correlations and factor loadings, and Table 10 for 39 expected correlations and residuals). The percent Of items for which error was greater than sampling error was 3.03%. Therefore, the data were not consistent with the initial hypotheses. Due to this fact, Hypotheses 2a and 2b could not be tested. A post hoc hypothesis was Ofl‘ered, as there appeared as though the data may be consistent with the notion of multidimensionality of the perceived barriers dimension. Although a multidimensional solution with four factors was found to fit the data, the same factor model did not fit a second-order unidimensional factor model, as hypothesized. As the correlations among the four factors were quite strong, it appeared that even though the four-factor model fit the data, the items may actually be alternate indicators of the same factor. A simpler, one- factor model was then tested and found to fit the data. This one-factor model was then tested cross-culturally (as suggested by Hypotheses 2a and 2b) and it was found that the one-factor model was not replicated in the United States subsample, but was replicated in the Kenyan subsample. 40 Table 5 Th rr la '0 d Factor Loadin Matrix for the Sin e-Factor Model“ 2 3 6 7 8 10 25 27 30 33 35 36 13 2 21 51 3 38 99 62 6 28 28 25 49 7 39 50 36 92 72 8 29 39 32 52 59 69 10 32 29 24 33 29 21 46 25 28 46 25 46 41 21 99 58 27 31 32 28 35 38 21 31 29 54 30 31 38 30 47 50 33 43 40 £9 70 33 18 22 25 26 33 29 28 24 35 21 46 35 27 30 28 36 40 19 30 26 45 31 29 54 36 25 32 28 40 42 24 25 33 37 23 30 22 54 B 51 62 49 72 69 46 58 54 70 46 54 54 100 * The numbm‘s used to identify the items are fi’om Appendix H. The underlined numbers in the diagonal show the reliabilities for each item (without decimals). The factor loadings complete the matrix; here, the items load on one overall perceived barriers factor (B). 4 1 Table 6 The Ema god Error Correlation Matrix for the Single-Factor Model“ 2 3 6 7 8 10 25 27 30 33 35 36 6 3 2 6 32 2 5 4 25 30 1 2 37 45 35 2 35 43 34 50 10 23 29 23 33 32 25 30 36 28 42 40 27 27 28 33 26 39 37 25 31 30 36 43 34 50 48 32 41 38 33 23 29 23 33 32 21 27 25 32 35 28 33 26 39 37 25 31 29 38 25 36 28 33 26 39 37 25 31 29 38 25 29 OOQGNN 9 0 1 0 3 5 5 5 7 4 2 3 7 2 1 1 8 2 1 2 1 3 QQNHOWWNWI—I 3 1 2 1 5 1 6 4 1 2 1 * The numbers used to identify the items are fi'om Appendix H. In this matrix, the lower diagonal presents the expected/predicted correlations (without decimals) for items of the overall perceived barriers factor; the upper diagonal presents the residuals. 42 Table 7 Th rr ' an F orLoadin MatrixfrheSin e-F rMO l in Dog fi'om tho [1th Statos Sample Only“ 2367810252730333536 B 2 29 54 3 42 92 72 6 26 31 25 49 7 33 63 30 99 70 8 29 57 27 73 92 72 10 42 37 28 33 26 29 50 25 35 50 24 56 67 21 99, 63 27 31 33 31 33 37 26 32 92 57 30 39 49 29 44 46 30 38 40 go 68 33 25 40 48 38 39 38 26 40 45 99 62 35 34 38 30 35 46 21 40 53 64 42 A9 65 36 25 32 27 24 24 36 27 25 23 29 29 19 44 B 54724970725063 5768626544 100 * The numbers used to identify the items are fi'om Appendix H. The underlined numbers in the diagonal show the reliabilities for each item (without decimals). The factor loadings , complete the matrix; here, the items load on one overall perceived barriers factor (B). 43 Table 8 Tho Exoogfl mo Error Corrolation Matrix for the Single-FMor Model using Doro fi'om the United States Sample Only" 2 3 6 7 8 10 25 27 30 33 35 U.) 0\ 2 3 0 5 10 15 1 0 2 8 1 1 3 39 4 l3 5 1 5 8 0 5 9 0 6 26 35 4 8 3 7 3 4 18 2 5 7 38 50 34 23 2 12 7 4 5 ll 7 8 39 52 35 50 10 22 4 3 6 l 8 10 27 36 25 35 36 11 3 4 7 12 14 25 34 45 31 44 45 32 4 5 13 1 1 27 31 41 28 40 41 29 36 l 5 16 0 30 37 49 33 48 49 34 43 39 3 20 7 33 33 45 30 43 45 31 39 35 42 2 2 35 35 47 32 46 47 33 41 37 44 40 0 36 24 32 22 31 32 22 28 25 30 27 29 * The numbers used to identify the items are from Appendix H. In this matrix, the lower diagonal presents the expected/predicted correlations (without decimals) for items of the overall perceived barriers factor, the upper diagonal presents the residuals. Table 9 Tho 99m Corrolofiorg and Factor Loading Matrix for the Single-Factor Modol using D fi In th K Sam 1e Onl * 2367810252730333536 B 2 19 43 3 31 29 50 6 24 20 11 42 7 36 38 29 99 68 8 24 24 28 37 51 64 10 21 22 18 34 30 19 44 25 22 44 23 44 27 21 99 60 27 25 23 20 22 30 17 29 19 43 30 22 27 25 42 46 35 45 34 59 70 33 09 08 08 17 26 21 28 13 27 11 33 35 14 19 19 28 30 14 21 02 30 20 19 39 36 14 22 19 35 44 10 21 24 37 13 20 22 47 B 43 50426864446043 7033 3947 100 * The numbers used to identify the items are from Appendix H. The underlined numbers in the diagonal show the reliabilities for each item (without decimals). The factor loadings complete the matrix; here, the items load on one overall perceived barriers factor (B). 45 Table 10 Tho floootfi Ed Error Correlation Matrix for the Single-Factor Model osing Dato fi'om me Konyao figmolo Only" 2 3 6 7 8 10 25 27 30 33 35 36 2 9 6 3 22 1 6 18 21 7 29 34 29 4 7 4 4 1 0 2 3 1 5 \rt-aoots MACON 8 28322 1019221 744 8 25 26 30 25 8 9 1 urbane \rhat— \r 8 8 4 5 2 30 28 5 41 38 26 4 2718221 29281926 4 3030352 4744304130 33 14 17 14 22 35 17 20 16 27 32 36 20 24 20 5 9 6 5 5 6 8 1 5 4 3 1 3 1 5 3 2 1 3 7 21 15 20 14 23 25 17 23 17 27 13 30 21 28 20 32 16 18 6 2 1 3 1 l 7 4 5 3 2 * The numbers used to identify the items are from Appendix H. In this matrix, the lower diagonal presents the expected/predicted correlations (without decimals) for items Of the overall perceived barriers factor; the upper diagonal presents the residuals. Chapter 5 DISCUSSION Past research indicates the importance of the dimension of perceived barriers to a . recommended response in the face Of a health threat (Champion, 1992; Janz & Becker, 1984; Sereno & Dunn, 1994). This study attempted to discern a multi-factor model to explain this important dimension Of the Health Belief Model. Were a multi-factor solution to fit the data, such information could prove useful to the development of future communication campaigns that focus on promoting preventive measures (recommended responses) to health threats. This study focused on the recommended response of the use Of condoms to prevent contraction Of HIV/AIDS. Although the hypothesized six-factor second-order unidimensional model failed to fit the data, the results Obtained in this study provide an interesting contribution to the notion of perceived barriers to a response recommended to prevent a health threat. As can be seen by the extensive amount Of research regarding barriers to preventive health actions presented in the literature review, the dimension of perceived barriers is one that has been used in many research projects. Although this study did not succeed in finding the second-order unidimensional factor structure proposed, the data 46 47 were consistent with a four-factor model. Perhaps the four factors used in the four-factor model are not as conceptually distinct as hypothesized. A single-factor model did fit the data, yet it did not continue to fit the data when tested cross-culturally. An explanation for the lack of success Of any of the models tested in this paper could be predicated on the idea that individuals have various understandings of each Of the barriers suggested. As noted in Chapter 4, some individuals questioned the Mug of some of the barrier items (e. g., “I am likely to use a condom because using condoms is masculine”). Across sexes, this item may have been invalid; females would most likely perceive no need to respond to this item. In addition to the meaning of a particular item for an individual, one could also look to psychological literature regarding the functions of attitudes as an explanation for the varied responses Obtained (Katz, 1960). Although Katz focused on studying individuals’ attitudes, such a functional approach could be applied to this study, in terms of the individuals’ attitudes toward the barrier items to which they responded, and how their attitudes may have influenced their understanding of the various items. Katz (1960) suggested that there are four firnctions of attitudes, each of which is aroused by different stimuli. The four firnctions are instrumental, ego-defensive, value- expressive, and knowledge functions. Instrumental (or adjustive or utilitarian) attitudes are those that are based upon the notion that an individual attempts to maximize rewards and minimize punishments, thus attaining the maximum utility possible fi'om the attitude. An ego-defensive attitude is one that derives fi'om an individual’s attempt to protect one’s ego from forces that may threaten the ego. The value-expressive function of attitudes addresses the desire Of the individual to present a positive self-image and to remain true to 48 the type of person the individual perceives him/herself to be. The final function suggested by Katz is the knowledge function, which -is fueled by an individual’s need and want of knowledge that will provide order and meaning so that the individual can “understand” the world in which s/he lives (Katz, 1960). It is possible that the numerous items developed with the intention of measuring difl‘erent aspects Of perceived barriers to condom use may have been processed by the subjects in various ways, as suggested by the firnctional approach of attitudes. In addition, each of the barrier items may have served as varying stimuli, producing diverse functional attitudes. Whereas an item such as “I am not likely to use a condom because condoms carry HIV” may have been intended to measure an individual’s knowledge about HIV- related issues (i.e., understanding that condoms are not contaminated or infected), an individual may have perceived this item as a threat to one’s self and one’s health if the individual focused on the aspect of the contamination of condoms. Perhaps such an ego- defensive function is more related to what this researcher conceptualized as being an individual barrier (something that affects the self), and the subject responding to this item may have responded thinking Of self-preservation and protecting the self; as Opposed to responding from a more rational view of knowledge (i.e., understanding how HIV is and is not transmitted). Thus, perhaps the six factors proposed by the researcher were not consistent with the data due to the individual perceptions of the subjects regarding what firnction the various barrier items may have stimulated. As a multidimensional factor structure was consistent with the data regarding a four-factor model, the idea that there may be multiple dimensions underlying the perceived 49 barriers dimension should not be dismissed. An altered focus, potentially using a fi'amework such as Katz’s attitude functions, may be more appropriate in discerning what the multiple dimensions may be. Further, as this model was tested cross-culturally, it is important to note that a possible explanation Of the lack of the success of the model could be due to cultural constraints that were not addressed in this study. The barrier items were developed using input from researchers and other individuals fiom both the United States and the Kenyan cultures. Attempts were made to address barriers that exist in both cultures. Yet, as the questionnaires were identical across cultures in terms of the barrier items, it is possible that some items included on the questionnaires may have seemed to be confusing or irrelevant to members of one of the two cultures. For example, item number 61, “I am not likely to use a condom because tO use a condom would suggest that I am a prostitute” (intended as a social norms/cultural values barrier) may have raised questions in the United States culture, as condom use is promoted by many health organizations, without any connection made to prostitution. It is also possible that the same item may not have been as appropriate as thought, even to the subjects in the Kenyan sample, where prostitutes (commercial sex workers) may actually have great dificulty in Obtaining condoms (Cameron et al., 1996). Another aspect of the cross-cultural nature Of this study is worthy of attention. Although HIV/AIDS is a threat across cultures, the impact Of the epidemic may be perceived in various ways across cultures. The spread of HIV/AIDS in Kenya is more extreme than the spread in the United States, as can be seen by the number of estimated 50 cases Of HIV infection (see Chapter 1). In addition to the barrier items holding difl’erent meanings for members Of difl‘erent cultures, as suggested above, a factor structure itself may take on a difl‘erent form when applied cross-culturally. Perhaps individuals Of one culture will perceive a higher degree of barriers to condom use than individuals in another culture. These perceptions could cause varying potential frameworks of the dimension Of perceived barriers to condom use to emerge in different cultures. A preliminary work using a portion of the data collected for this study analyzed only the two dimensions of individual and relational barriers to condom use (Cameron, 1996). A research question in this preliminary study focused on determining whether or not the variable of culture (United States versus Kenyan) appeared to influence perceived barriers to condom use. Results of two-tailed t-tests indicated that subjects in the Kenyan sample perceived that the barriers to condom use were greater than did those subjects in the United States sample, whether the barriers be individual or relational (Cameron, 1996). These results suggest that culture does play a role in an individual’s perceptions of barriers to condom use. Thus, it is possible that a framework Of perceived barriers to condom use would not remain constant when tested cross-culturally. Indeed, when the one-factor model presented in' this study was tested, the model was not consistent across cultures. This study, then, has provided a framework of perceived barriers to condom use, albeit a fiamework that was not consistent with the data. Further research could focus on developing a fi'amework that is applicable within a single culture, and then attempt to test such a model cross-culturally. This study suggests potential dimensions of a perceived 51 barriers framework, and is consistent with the notion that there may be multiple dimensions underlying the perceived baniers dimensions suggested by the HBM. Limitoo'ons There are at least four limitations to this study. First, the imposition of a factor structure on the dimension of perceived barriers was not consistent with the data in this study. The six dimensions (individual barriers, relational barriers, physical barriers, knowledge-based/self-eficacy barriers, social norms/cultural values barriers, and structural barriers) were developed through a careful review of the literature, as well as through the solicitation of items from various health professionals. However, in the end, the researcher is the one who suggested these six dimensions. Perhaps there is multidimensionality in the perceived barriers dimension suggested by the Health Belief Model, but the factors suggested and conceptualized in this study may not be appropriate divisions of the perceived barriers dimension. Second, the hypotheses suggesting second-order unidimensionality may be premature, as no existing factor structure of the dimension of perceived barriers was used to test this hypothesis. Thus, by hypothesizing that there was a second-order unidimensional factor structure, the requirements placed on this study may have been too stringent as a first attempt to further investigate the perceived barriers dimension. A more structured approach, which would include but not be limited to the development of scales measuring various dimensions of the perceived barriers dimension, if the barrier dimension is indeed multidimensional, should first be attempted. After such scales are developed and tested, then research could proceed, in an attempt to determine whether or not those 52 scales might fit a second-order unidimensional model. In this study, the second-order unidimensionality hypothesis was tested and found to fail. Third, the reliabilities of the scales that were used to measure the proposed factors were at times quite low. It is possible that the low reliabilities of the scale could provide information about the perceived barriers dimension, for example, it is possible that items used to measure the specific barrier dimension proposed would have been more appropriately placed as a measure of one of the other proposed barrier dimensions. Fourth, the length of the questionnaire may have contributed to the occurrence of response bias by the participants. Although the 76 perceived barriers items included items to be recoded, it appeared, when looking carefully at the original data, that some individuals simply chose a response and used that response throughout the majority of the questionnaire. The limitations all suggest that the development of a multi-stage study may be appropriate to measure this phenomenon of perceived barriers to condom use. Ideas as to how such a study may be developed are presented in the following section. Fogge Directions As suggested by the limitations, research regarding the perceived barriers dimension suggested by the Health Belief Model could be enhanced by a carefully developed multi-stage study. As past research has indicated that the perceived barriers dimension of the Health Belief Model is an important one, research regarding this dimension is warranted. Initially, an applicable framework of the perceived barriers dimension needs to be developed. This fiamework may be developed through various 53 processes, including but not limited to statistical analyses such as exploratory factor analysis. As noted above in the limitations section, the underlying dimensions suggested here were the product of the researcher, and perhaps there are multiple dimensions to perceived barriers, yet these dimensions may be of a far different ilk than, e.g., individual and relational barriers. Once an appropriate fi'amework is established, if one does appear to fit the dimension of perceived baniers to a recommended response, then the step of administering the scale to determine if the fiamework fits a second-order unidimensional model would be appropriate. However, the necessary first step is to determine reliable scales to measure the suggested dimensions of the perceived barriers dimension. If specific clusters of barriers to a recommended response, here, condom use, can be determined, such information could be incorporated into future prevention campaigns in an attempt to better address barriers to condom use. Such campaigns would then have the potential to not only increase awareness of knowledge of HIV /AIDS, as current campaigns have achieved, but also to promote an increased use of condoms as a preventive measure against contraction of HIV. Current HIV/AIDS prevention campaigns appear to be increasing knowledge about the disease, yet there remain unbroken barriers that are hindering individuals fi'om using condoms to protect themselves against contraction of HIV. If research can determine what these barriers to condom use rrright be, then firrther study can determine if such barriers can be addressed in a mass media format, such as communication campaigns, or if other measures need to be taken in order to break down these barriers to condom use. Chapter 6 CONCLUSION The results of this study suggest that the perceived barriers dimension of the Health Belief Model does not fit a second-order unidimensional factor model. Vlfrth the final analysis suggesting a one-factor model of barriers to condom use, this study indicated that, although barriers to condom use may appear to be of sharply different foci (e.g., falling into various categories such as individual baniers, physical barriers, etc.), these barriers may not be discernible to individuals as barriers of different ilk. It is also important to recognize that the six barrier dimensions suggested here were imposed upon these data. The six barrier dimensions were developed from a review of the research in an attempt to categorize the numerous barriers found in multiple tests of the HBM. However, the barrier dimensions suggested in this paper were merely one researcher’s conception of appropriate categories of barriers, as gleaned fi'om the literature, past studies regarding the HBM, as well as self-reports of individuals who are involved first- hand with HIV/AIDS prevention campaigns and the promotion of safer sex. The data obtained in this study were not consistent with the hypothesis that the perceived barriers dimension of the HBM forms a second-order unidimensional factor 54 55 structure with six underlying dimensions. The lack of consistency of the data with this hypothesis does not preclude the possibility that multiple factors may underlie the perceived barriers dimension. Rather, the results of this study indicate that the particular factors suggested and measured here are not those factors that underlie the perceived barriers dimension. Perhaps with alternate factor structures, a multidimensional model could be developed and tested. This study does indicate that there are barriers to condom use in existence, and that people perceive these barriers as hindering them in practicing safer sex. Understanding what categories of barriers exist may be crucial in developing future communication campaigns to promote condom use. In addition to focusing on one’s susceptibility to contraction of HIV, and the severity of the disease, focusing on promoting condom use as an efficacious and possible alternative to unprotected sex may strengthen firture HIV/AIDS prevention campaigns. 56 ENDNOTES 1. The handout provided to the participants in the United States also included phone numbers of various organizations and hotlines through which participants could receive firrther information about HIV/AIDS. 2. “Communicating Health with Unique Populations” conference held at Michigan State University, April 8-9, 1995. 3. Conference participants were asked to respond to the following: “Please list any barriers that you are aware of, or that you may have heard individuals voice, regarding reasons an individual would not use condoms to prevent contraction of HIV/AIDS.” 4. See Mtte, Nzyuko, & Cameron (1996) and Cameron, Witte, Lapinski, & Nzyuko (1996) for further information. 5. In the United States sample there were no respondents indicating Kenyan citizenship. In the Kenyan sample, 1.1% of the respondents indicated United States citizenship. 6. English is one of the national languages of Kenya. The class in which this questionnaire was administered in Kenya was conducted in English. 7. Note that, in confirmatory factor analysis, the null hypothesis is that the model fits the data. Therefore, the desired significance levels are p > .05 so that one fails to reject the null hypothesis that the model fits the data. APPENDICES APPENDIX A APPENDIX A Individ1g1 Bgo'ers to Condom Use“ *Note: Individuals were provided with the following directions to answer the 76 barrier items: “Please answer the following questions on a scale of 1 to 5, with 1 meaning “Strongly Disagree,” 2 meaning “Disagree,” 3 meaning “Neither Agree nor Disagree,” 4 meaning “Agree,” and 5 meaning “Strongly Agree.” Please note that the questions refer to the use of male condoms during sexual intercourse. Please read each question thoroughly and carefully.” Item M SD 1. I am NOT likely to use a condom because condoms reduce 1.92 1.14 spontaneity in sexual interactions. 2. I am NQT likely to use a condom because to use a condom suggests 1.76 1.13 that I do not trust my partner. 3. I am NOT likely to use a condom because I do not know how to 1.61 0.87 talk about condom use. 4. I am NOT likely to use a condom because if I use a condom, people 1.31 0.70 willthinkthatlaminfectedwithI-IIV. 5. I am NOT likely to use a condom because I do not want to buy them 1.52 0.83 because someone I know might see me and then they will know that I am having sex. 6. I am likely to use a condom because I know that I am NOT 2.56 1.53 invincible. (R) 7. I am NQT likely to use a condom because condoms spoil the mood.2.00 1.10 8. I am NOT likely to use a condom because condoms are inconvenient 2.32 1.27 to use when having repeated sexual intercourse. 9. I am NOT likely to use a condom because I am going to die anyway. 1.30 1.22 57 58 Item M SD 10. I am NOT likely to use a condom because I get a thrill when 1.63 0.89 I take chances. 11. I am likely to use a condom because I am NOT embarrassed 2.16 1.29 to use condoms. (R) 12. I am likely to use a condom because using condoms is masculine.(R) 3.75 1.06 13. I am likely to use a condom because using condoms is an expression 3.37 1.27 of love. (R) 14. I am NQT likely to use a condom because I (or my partner) use 2.61 1.41 another form of birth control. 15. I am likely to use a condom because my friends use condoms. (R) 3.79 1.10 16. I am NQT likely to use a condom because people might think that I 1.61 0.88 sleep around. 17. I am likely to use a condom because I am vulnerable to contracting 2.28 1.46 a sexually transmitted disease. (R) 18. I am NOT likely to use a condom because I just don't care. 1.48 0.90 19. I am NOT likely to use a condom because people might think that 1.37 0.71 I am homosexual. 20. I am likely to use a condom because I am NOT embarrassed to ask 2.23 1.30 my partner to use condoms. (R) 21. I am NOT likely to use a condom because only alternative health 1.88 1.12 practices will protect me against sexually transmitted diseases. 22. I am likely to use a condom because I feel that I have the power to 2.50 1.34 enforce condom use in my relationship. (R) 23. I am NQT likely to use a condom because condom use reduces 2.12 1.14 intimacy. (R) Item was recoded for analysis. Items in bold are those used in the 36-item scale (Appendix G). APPENDIX B APPENDIX B Relogiofl Bfloro Lo Condom Use Item M SD 24. I am NOT likely to use a condom because if I suggest condom use 1.81 1.12 my pytper will think that I have been unfaithful. 25. I am NQT likely to use a condom because condom use will 1.83 1.12 encourage my partner to have other partners. 26. I am NOT likely to use a condom because my partner is too 1.77 1.03 embarrassed to use condoms. 27. I am NQT likely to use a condom because my partner will not 1.75 1.01 trust me if I suggest condom use. 28. I am NOT likely to use a condom because if I suggest condom use, 1.77 1.04 M will think that I sleep around. 29. I am NOT likely to use a condom because my partner behaves that 1.72 0.92 condoms aren’t masculine. 30. I am NOT likely to use a condom because my partner is 2.22 1.32 not infected. 31. I am likely to use a condom because my partner wants to use 2.66 1.38 condoms. (R) 32. I am likely to use a condom because my partner is willing to use 2.30 1.28 condoms. (R) 33. I am NOT likely to use a condom because I know my partner 2.44 1.41 well enough. 34. I am NOT likely to use a condom because I can trust W. 2.44 1.38 59 60 Item M SD 35. I am likely to use a condom because using a condom turns 3.70 1.04 mum on. (R) 36. I am NQT likely to use a condom because my partner will reject me 1.66 0.84 if I suggest condom use. (R) Item was recoded for analysis. Items in bold are those used in the 36-item scale (Appendix G). APPENDIX C APPENDIX C Physig Barriers to Condom Use Item M SD 37. I am N91 likely to use a condom because condoms are too small. 1.58 0.85 38. I am likely to use a condom because condoms are comfortable. (R) 3.55 1.07 39. I am N91 likely to use a condom because condoms reduce 2.24 1.20 sensation. 40. I am likely to use a condom if I am using drugs. (R) 3.30 1.47 41. I am N91 likely to use a condom because condoms are not 1.99 1.18 natural. 42. I am Nfl likely to use a condom because I don't like the feel 2.33 1.18 of condoms. 43. I am likely to use a condom ifI am drunk. (R) 3.05 1.37 44. I am M likely to use a condom because I am allergic to latex. 1.91 0.99 45. I am likely to use a condom because condoms enhance pleasure.(R) 3.78 0.94 (R) Item was recoded for analysis. Items in bold are those used in the 36-item scale (Appendix G). 61 APPENDIX D APPENDIX D Knowledge-BagflSelf-Eficacy Barriers to Condom Use Item M SD 46. I am NOT likely to use a condom because I have never been 1.50 0.82 taught how to use condoms. 47. I am N9T likely to use a condom because condoms carry HIV. 1.35 0.74 48. I am likely to use a condom in order to prevent contraction 1.88 1.34 of potentially fatal diseases. (R) 49. I am likely to use a condom because I know how to use a condom 2.39 1.25 correctly. (R) 50. I am NOT likely to use a condom because I have never used 1.81 1.09 condoms before. 51. I am N9T likely to use a condom because I do not know where 1.44 0.74 I can obtain condoms. 52. I am likely to use a condom because sex is risky. (R) 2.11 1.25 53. I am NOT likely to use a condom because I have never practiced 1.93 1.95 using condoms. 54. I am NOT likely to use a condom because whether or not I 1.46 0.84 contract a sexually transmitted disease is out of my control. (R) Item was recoded for analysis. Items in bold are those used in the 36-item scale (Appendix G). 62 APPENDDK E APPENDIX E Sogal Norms/991mm] Values Barriers to Cdeom Us_e Item M SD 55. I am NOT likely to use a condom because I would 1.55 0.83 be stigmatized if I did. 56. I am likely to use a condom because I am responsible for 1.84 1.24 what happens to my health. (R) 57. I am N9T likely to use a condom because my religion 1.91 1.18 discourages the use of condoms. 58. I am N9T likely to use a condom because no one has ever 1.53 0.81 told me that I should use a condom. 59. I am likely to use a condom because I believe that people whom 3.08 1.29 I respect use condoms. (R) 60. I am NOT likely to use a condom because it is inappropriate 1.53 0.78 for women to suggest condom use. 61. I am NOT likely to use a condom because to use a condom would 1.49 0.82 suggest that I am a prostitute. 62. I am N9T likely to use a condom because people may avoid me 1.51 0.73 flldm 63. I am likely to use a condom because using condoms is okay 3.07 1.13 with my fiiends. (R) (R) Item was recoded for analysis. Items in bold are those used in the 36-item scale (Appendix G). 63 APPENDIX F APPENDIX F 51mm Bg'ers to andom Use Item M SD 64. I am NOT likely to use a condom because, in past use, when 1.78 0.99 I used a condom it broke. 65. I am NQT likely to use a condom because, in past use, when 1.59 0.84 I used a condom I (or my partner) got pregnant anyway. 66. I am likely to use a condom because condoms are easily accessible.(R) 2.49 1.26 67. I am NOT likely to use a condom because I cannot afford 1.49 0.79 condoms. 68. I am likely to use a condom because using a condom will protect 1.89 1.27 me against HIV/AIDS. (R) 69. I am NOT likely to use a condom because condoms are 1.53 0.82 too expensive to buy. 70. I am NOT likely to use a condom because, in past use, when I 1.53 0.79 used a condom, I contracted a sexually transmitted disease anyway. 71. I am NOT likely to use a condom because providers refuse 1.57 0.77 to provide me with condoms. 72. I am NOT likely to use a condom because I cannot obtain condoms 1.75 0.93 in an anonymous way. 73. I am likely to use a condom because condoms are reliable. (R) 2.57 1.14 74. I am likely to use a condom because I can get condoms 3.04 1.24 for free. (R) 75. I am NOT likely to use a condom because providers are unable 1.64 0.80 to provide me with condoms. 76 I am T lik to e a condom because con oms break. 2.18 1.07 (R) Item was recoded for analysis. Items in bold are those used in the 36-item scale (Appendix G). APPENDIX G APPENDIX C Bm'grs to Qondpm Use: The 36 Items Used for the Propcfied Six-Factor Model Item M SD 1. I am NOT likely to use a condom because I do not know how to 1.61 0.87 talk about condom use. (I #3)* 2. I am NOT likely to use a condom because condoms spoil the mood.2.00 1.10 (I #7) 3. I am NOT likely to use a condom because I get a thrill when 1.63 0.89 I take chances. (I #10) 4. I am likely to use a condom because I am NOT embarrassed 2.16 1.29 to use condoms. (R) (I #11) 5. I am NOT likely to use a condom because I just don't care. (I #18) 1.48 0.90 6. I am NOT likely to use a condom because only alternative health 1.88 1.12 practices will protect me against sexually transmitted diseases. (1 #21) 7. I am NOT likely to use a condom because condom use will 1.83 1.12 encourage my partner to have other partners. (R #25) 8. I am NOT likely to use a condom because my pprtner is too 1.77 1.03 embarrassed to use condoms. (R #26) 9. I am NOT likely to use a condom because my partner will not 1.75 1.01 trust me if Isuggest condom use. (R #27) 10. I am NOT likely to use a condom because my partner is 2.22 1.32 not infected. (R #30) 11. I am NOT likely to use a condom because condoms are too small. 1.58 0.85 (P #37) 12. I am likely to use a condom because condoms are comfortable. (R) 3.55 1.07 (P #3 8) 65 66 Item M SD 13. I am NOT likely to use a condom because condoms reduce 2.24 1.20 sensation. (P #39) 14. I am likely to use a condom if I am using drugs. (R) (P #40) 3.30 1.47 15. I am NOT likely to use a condom because condoms are not 1.99 1.18 natural. (P #41) 16. I am NQT likely to use a condom because I don't like the feel 2.33 1.18 of condoms. (P #42) 17. I am NOT likely to use a condom because I am allergic to latex. 1.91 0.99 (P #44) 18. I am NOT likely to use a condom because I have never been 1.50 0.82 taught how to use condoms. (KB #46) 19. I am NOT likely to use a condom because condoms carry HIV. 1.35 0.74 (KB #47) 20. I am likely to use a condom in order to prevent contraction 1.88 1.34 of potentially fatal diseases. (R) (KB #48) 21. I am NOT likely to use a condom because I do not know where 1.44 0.74 I can obtain condoms. (KB #51) 22. I am likely to use a condom because sex is risky. (R) (KB #52) 2.11 1.25 23. I am NOT likely to use a condom because I have never practiced 1.93 1.95 using condoms. (KB #53) 24. I am NOT likely to use a condom because whether or not I 1.46 0.84 contract a sexually transmitted disease is out of my control. (KB #54) 25. I am NOT likely to use a condom because I would 1.55 0.83 be stigmatized if I did. (SN #55) 26. I am likely to use a condom because I am responsible for 1.84 1.24 what happens to my health. (R) (SN #56) 67 Item M SD 27. I am NQT likely to use a condom because my religion 1.91 1.18 discourages the use of condoms. (SN #57) 28. I am NOT likely to use a condom because no one has ever 1.53 0.81 told me that I should use a condom. (SN #58) 29. I am NOT likely to use a condom because it is inappropriate 1.53 0.78 for women to suggest condom use. (SN #60) 30. I am NOT likely to use a condom because to use a condom would 1.49 0.82 suggest that I am a prostitute. (SN #61) 31. I am NOT likely to use a condom because people may avoid me 1.51 0.73 if I do. (SN #62) 32. I am NOT likely to use a condom because, in past use, when 1.59 0.84 I used a condom I (or my partner) got pregnant anyway. (S #65) 33. I am NOT likely to use a condom because I cannot afford 1.49 0.79 condoms. (S #67) 34. I am likely to use a condom because I can get condoms 3.04 1.24 for free. (R) (S #74) 35. I am NOT likely to use a condom because providers are unable 1.64 0.80 to provide me with condoms. (S #75) 36. I am NOT likely to use a condom because condoms break. (S #76) 2.18 1.07 *Note: The code in parentheses refers to the factor that the item is purported to measure and the number of the item on the original 76-item scale. I = individual barriers, R = relational barriers, P = physical barriers, KB = knowledge-based/self-eficacy barriers, SN = social norms/cultural values barriers, S = structural baniers. (R) Item was recoded for analysis. Items in bold were used for 16-item four-factor analysis. APPENDIX B APPENDIX B Barriers to Condom Use: The 16 Items Used for the Four-Factor Analysis Item M SD 2. I am NOT likely to use a condom because condoms spoil the mood. 2.00 1.10 (I #7) 3. I am NQT likely to use a condom because I get a thrill when 1.63 0.89 . I take chances. (I #10) 5. I am NQT likely to use a condom because I just don't care. (I #18) 1.48 0.90 6. I am NOT likely to use a condom because only alternative health 1.88 1.12 practices will protect me against sexually transmitted diseases. (1 #21) 7. I am NOT likely to use a condom because condom use will 1.83 1.12 encourage my partner to have other partners. (R #25) 8. I am NOT likely to use a condom because my partner is too 1.77 1.03 embarrassed to use condoms. (R #26) 9. I am NOT likely to use a condom because my partner will not 1.75 1.01 trust me if I suggest condom use. (R #27) 10. I am NOT likely to use a condom because my partner is 2.22 1.32 not infected. (R #30) 25. I am NOT likely to use a condom because I would 1.55 0.83 be stigmatized if I did. (SN #55) 26. I am likely to use a condom because I am responsible for 1.84 1.24 what happens to my health. (R) (SN #56) 27. I am NOT likely to use a condom because my religion 1.91 1.18 discourages the use of condoms. (SN #57) 30. I am NOT likely to use a condom because to use a condom would 1.49 0.82 suggest that I am a prostitute. (SN #61) 68 69 Item M SD 32. I am NOT likely to use a condom because, in past use, when 1.59 0.84 I used a condom I (or my partner) got pregnant anyway. (S #65) 33. I am NOT likely to use a condom because I cannot afford 1.49 0.79 condoms. (S #67) 35. I am NOT likely to use a condom because providers are unable 1.64 0.80 to provide me with condoms. (S #75) 36. I am NOT likely to use a condom because condoms break. (S #76) 2.18 1.07 I"Note: Numbers used in this Appendix and Tables 3-10 refer to the numbers of the items in the 36-item scale (see Appendix G). Items in bold were those used for the 12-item single-factor analysis. (R) Item was recoded for analysis. LIST OF REFERENCES LIST OF REFERENCES Agata, N., Muita, M, Muthami, Gachihi, G.S., & Pelle, H. (1993). Epidemiology of HIV/AIDS in Kenya. 1% Kenya HIV/AIDS/STD Conference: Prom and Abstmpts, 18. AIDS A_nflys_1§' Afi'iga. (1993). East and central Africa worst hit. May/June, p.5. Allen, S., Tice, J., Van de Perre, P., Serufilira, A, Hudes, E., Nsengumermyi, F., Bogaerts, J., Lindan, C., & Hulley, S. (1992). Effect of serotesting with counselling on condom use and seroconversion among HIV discordant couples in Africa. Britifl Mfl'pgl 1pm 2&4, 1605-1609. Baldwin, J.D., & Baldwin, II. (1988). AIDS information and sexual behavior on a university campus. Journal of SQ Education and Therapy, E, 24-28. Baldwin, J.I., Whiteley, S., & Baldwin, ID. (1990). Changing AIDS- and fertility- related behavior: The efl‘ectiveness of sexual education. Journal of Sex Research, g, 245- 262. Bowser, BR (1992). African-American culture and AIDS prevention: From barrier to ally. mm of Medicine. 157. 286-289. Calnan, M., & Rutter, DR (1986). Do health beliefs predict health behaviour? An analysis of breast self-examination. Soci_al Science and Medicine 2_2_, 673-678. Cameron, K.A (1996, August). Individual r lational barn' W. Paper presented at the biannual meeting of the International Society of Social and Personal Relationships, Banfl‘, Alberta, Canada. Cameron, K.A, Witte, K., Lapinski, M.K., & Nzyuko, S. (1996, November). Prmpting HIV Transmission along the Trans-Africa Highway in Kenya; Using my; message than: 19 99nduct a fprmative evaluation. Paper presented at the annual meeting of the Speech Communication Association, San Diego, CA Champion, V. (1992). Breast self-examination in women 65 and older. The Jpgpals of flontology, fl, 75-79. 70 71 Chesney, MA (1994). Prevention of HIV and STD infections. Preventive Mgficine, Q, 655-660. Choi, K.H., Rickman, R, & Catania, IA (1994). What heterosexual adults believe about condoms. The New England JournJal of Medicine. 3_3_l_, 406-407. Clark, M.M. (1983). Cultural context of medical practice. The Western 19pm; pf Ms. E. 2-6. Cline, RJ.W., Freemen, K.E., & Johnson, 8.]. (1990). Talk among sexual partners about AIDS: Factors differentiating those who talk from those who do not. memunicap'pn Rem 11, 792-808. Cornely, PB. (1976). Racism: The ever-present hidden barrier to health in our society. Am ' f Pu lic H 66, 246-247. Fabrega, H. Jr. (1977). Group differences in the structure of illness. 91m, meme apt! chhiatgg, 1, 379-3 94. Fisher, J.D., & Misovich, DJ. (1990). Evolution of college students’ AIDS related behavioral response, attitudes, knowledge, and fear. AIDS Education and Preventiop, 2, 322-337. Gerbert, B., & Maguire, B. (1989). Public acceptance of the Surgeon General’s brochure on AIDS. Public Health Reports, Mi. 130-133. Gerbert, B., Maguire, B.T., Bleecker, T., Coates, T.J., & McPhee, SJ. (1991). Primary care physicians and AIDS: Attitudinal and structural barriers to care. M the American Mfl'w Associatiop, 266, 2837-2842. Guttman, L. (1955). A generalized simplex for factor analysis. Mum 20, 173-192. Hunter, I. E. (1977). Cluster Anal sis: Reliabili Multiple Indicators Approach to Measurement. Paper presented at a workshop titled “Advanced Statistics” given at the U. S. Civil Service Commission. Hunter, J.E., & Boster, FJ. (1987). A model of compliance-gaining message selection. memunicatipn Monpgraphs, _5_4, 63-84. . Hunter, J.E., & Gerbing, D.W. (1982) Unidimensional measurement, second order factor analysis, and causal models. Reseagh in Organizational Behavior, 4, 267-300. Janz, N., & Becker, M.H. (1984). The health belief model: A decade later. Halt}: Education marterly, 1_1, 1-47. 72 Johnson, S.M., Kurtz, M.E., Tomlinson, T., & Howe, KR (1986). Students' stereotypes of patients as barriers to clinical decision-making. W Question. 51.. 727-734. Jones, S.L., Jones, P.K., & Katz, J. (1988). Health belief model intervention to increase compliance with emergency department patients. Medical Care 2_6_, 1172-1184. Katz, D. (1960). The fimctional approach to the study of attitudes. Public Qpinipn My. A. 163-204- Kelly, GR, Mamon, A, & Scott, J.E. (1987). Utility of the health belief model in examining medication compliance among psychiatric outpatients. Social Scieng app Mgicine, 2_5, 1205-1211. Kirscht, J.P., & Joseph, J .G. (1989). The health belief model: Some implications for behavior change, with reference to homosexual males. In V.M. Mays, G.W. Albee, & S.F. Schneider (Eds), PM prevention of AIDS: Pachglogig apprpaghg (pp. 111- 127). Newbury Park, CA: Sage. Koop, CE. (1988). Understanding AIDS. DHHS Publication No. (CDC) HHS-' 88-8404. Rockville, MD: Public Health Service. Manning, D., Balson, P.M., Barenberg, N., & Moore, TM. (1989). Susceptibility to AIDS: What college students do and don't believe. Journal of Amerim Qpllege Hflm. 3, 67-73. Marin. 3.. & Mafia, G. (1990) Introduction. WW Sciences. .12. 107-109. Melnyk, KAM. (1988). Barriers: A critical review of recent literature. Mg m 31, 196-201. Novicld, MA (1992). Fighting the AIDS Epidemic: Dr. Peter Lamptey. Afiig Repgrt, July/August, 17 - l9. Orlandi, MA (1987). Promoting health and preventing disease in health-care settings: An analysis of barriers. Preventive Medicine E. 119-130. Peterson, J .L., & Marin, B. (1988). Issues in the prevention of AIDS among Black and Hispanic men. Americap Psychologist, fl, 871-877. Quesada, G.M., & Heller, PL. (1977). Sociocultural baniers to medical care among Mexican-Americans in Texas: A summary report of research conducted by the Southwest Medical Sociology Ad Hoc Committee. Medical Care g, (supplement), 93- 101. 73 Rachier, AD. 0. (1993). HIV/AIDS and the law] lst Kenya HIV/AIDS/STD MW 55 Rosenstock, I.M. (1974). Historical origins of the health belief model. PM Edug'ep'en Meneggaphs, _2_, 328-335. Ruiz, P. (1985). Cultural barriers to efi‘ective medical care among Hispanic- American patients. m Review of Medicine, &, 63-71. Sereno, K.K., & Dunn, D. (1994). PredictmLAHDS risk reduction behavior in Ellege studehte: Communieetion and the Health Belief Model. Paper presented at the Annual Meeting of the Speech Communication Association, New Orleans, LA Sheer, V.C., & Cline, RJ. (1994). The development and validation of a model explaining sexual behavior among college students: Implications for AIDS communication campaigns. Hm Communication Research, _2_1, 280-304. Sheer, V.C., & Cline, RJ.W. (1995). Individual differences in sensation seeking and sexual behavior: Implications for communication intervention for HIV/AIDS prevention among college students. Health Communicatioh, 1, 205-223. Sobo, E.J. (1993). Inner-city women and AIDS: The psycho-social benefits of unsafe sex. Culture, Mfl'em’ e and chhiahy, 11, 455-485. Wendt, S.J., & Solomon, L.J. (1995). Barriers to condom use among heterosexual male and female college students. Joumel of American College Health, 55, 105-110. Witte, K. (1992). The role of threat and eficacy in AIDS prevention. 13111311931 Qpegegly ef Community Hfih Education, L2, 255-249. VVrtte, K., Berkowitz, J ., Cameron, K., & McKeon, J. (1995). Preventing the eprm ef gm'tal warts: Using fear appeals to promote self-protective ghaviphs. Paper presented at the annual meeting of the Speech Communication Association, San Antonio, TX. “frtte, K., & Morrison, K. (1995). Intercultural and cross-cultural health communication: Understanding people and motivating healthy behaviors. Interfltural Ed Ingmtiohd Communieetien Annual, 15, 216-246. Witte, K., Nzyuko, S., & Cameron, K.A. (1996). HIV/AIDS along the Tme- Afiig Highway in Kenye: Examining risk perceptions- recommended resmnses, and eampaigh materials. Report submitted to the All-University Research Initiative Grant program, Michigan State University. 74 Witte, K, Stokols, D., Ituarte, P., & Schneider, M. (1993). Testing the health belief model in a field study to promote bicycle safety helmets. Communicatien Remh, Q, 564-586. World Health Organization. (1995, December 15). Weeldy Epidemielegighl Ligand. E (50). 353-360- World Health Organization. (1996, July 5). Weekly Epidemiological Re_cerd, 11 (27), 205-212. Yep, G. (1992). Communicating the HIV/AIDS risk to Hispanic populations: A review and integration. Hispanie Jeurnal of Behavioral Science, fl, 403-420. Yep, GA (1993). HIV prevention among Asian-American college students: Does the Health Belief Model work? American Journal of College thh, fl, 199-205.