VALIDATION OF AUTISM SCREENING ASSESSMENTS: COMPARISON OF THE SOCIAL COMMUNICATION QUESTIONNAIRE, SOCIAL RESPONSIVENESS SCALE AND 23Q WITH DSM - 5 IN ASSESSING FOR AUTISM SPECTRUM DISORDER (ASD) IN UGANDA By Jorem Emmillian Awadu A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Rehabilitation Counselor Education Doctor of Philosophy 20 21 ABSTRACT VALIDATION OF AUTISM SCREENING ASSESSMENTS: COMPARISON OF THE SO CIAL COMMUNICATION QUESTIONNAIRE, SOCIAL RESPONSIVENESS SCALE AND 23Q WITH DSM - 5 IN ASSESSING FOR AUTISM SPECTRUM DISORDER (ASD) IN UGANDA By J orem Emmillian Awadu The lack of validated autism spectrum disorder (ASD) screening instruments hampers rehabilitation efforts for children with ASD in Africa. In the present study, the psychometric properties of two ASD screening instruments, that have been well - validated in the United States (US) context, were examined in the Ugandan context. These instruments were the Social Responsiveness Scale - Second Edition (SRS - 2; Constantino & Gruber, 2012) and the Social Communication Questionnaire (SCQ; Rutter, Bailey & Lord, 2003) . T hese two instruments were cross - validated with the 23Questions Screener (23Q; Kakooza - Mwesige et al., 2014) - a developmental disability screening measure for assessing ASD symptoms among children in Uganda. Two distinct groups of study participants were r ecruited: (a) 51 caregivers of children aged 4 - 17 years with a known diagnosis of ASD and (b) 56 caregivers of typically developing children (non - ASD) of the same age range. Caregivers of children with ASD were recruited either during routine clinical fo llow - up visits at a national hospital or through a special needs school with both sites drawing cases from the city of Kampala and the surrounding districts. Caregivers of typically developing children (i.e., children not identified or suspected of having a developmental disability) were recruited either during medical wellness visits or through schools within the same region. Results showed that the SRS - 2 and SCQ were both reasonably reliable and valid for use in distinguishing Ugandan children with ASD fr om those who are typically developing. Summation of items 19 - 23 of the 23Q (typically used to screen children below five years of age) yielded higher internal consistency reliability than a combination of all ten available 23Q ASD items (i.e., items 14 23). For discriminant validity, children with a diagnosis of ASD scored significantly higher than their non - ASD peers on all study instruments. Based on the sample results, the optimum SRS - 2 cut - score was lowered to 62, while that of the SCQ was retained at 15. A cut - score of one or more on the 23Q, despite having high sensitivity yielded unacceptably low specificity. Increasing the 23Q cut - score to 3 or more improved specificity but lowered sensitivity albeit to a moderate range. Copyright by J OREM EMMILLIAN AWADU 20 21 v To my parents, Mrs. Jane Margaret Iyobu Akope (R.I.P.) and dad, Mr. John Ejou Akope. For you, my wife, Esther Akwii, and son, Jethro Victor Awadu To you, Dr. Kevin Aanyu, and all my siblings vi ACKNOWLEDGEMENTS First, I would love to thank Dr. Gloria Lee my academic advisor. You constantly challenged me to be better. The world needs more professors like you. Dr. Martin Volker, I do not know how best to say THANK YOU in a befitting way. Ours was a chance meetin g on academic pursuits. However, you took me up wholeheartedly. Your guidance enabled me to succeed in this endeavor. You have created one of the few psychometricians in Africa. To Drs. Michael Leahy and Andrew Nay, yes, we finally made this happen. It wa s an honor having you on my dissertation committee. Your questions, whether in the lift in Erickson Hall or during our meetings made this study possible. Uniquely, I would like to thank Drs. Connie Sung, John Kosciulek, Hung Jen Kuo, Todd Lewicki, Su Pi, and Cary Roseth. Thank you for all the support over the years. To my mentors: Drs. Catherine Abbo, Angelina Kakooza - Mwesige, Amara Ezeamama, and Michael Boivin, I am grateful for your guidance and support. To Rebecca Akello, Annet Birabwa, Julius Caesar Ojuka, Wankya Ronald Kiduma, Robert Katende, and the administrators in Butabika National Mental Health Referral hospital and Finally, I would love to thank the administrators of the College of Education, Michigan State University. Your several research scholarships to me made this ambitious student research project possible. vii TABLE OF CONTENTS x CHAPTER 1: INTRODUCTION .......... 1 Autism Spectrum Disorder ................................ ................................ ................................ ......... 1 Assessing for ASD and its Challenges in Africa ................................ ................................ ........ 2 Purpose of the Study ................................ ................................ ................................ ................... 4 Potential Implications of the Study ................................ ................................ ............................. 5 Research Questions ................................ ................................ ................................ ..................... 6 CHAPTER 2: LITERATURE REVIEW ................................ ................................ ........................ 7 Autism Spectrum Disorder ................................ ................................ ................................ ......... 7 Etiology and Prognosis ................................ ................................ ................................ ............... 8 Global Burden ................................ ................................ ................................ ........................... 10 Autism Diagnosis ................................ ................................ ................................ ...................... 10 Advantages of Autism Screening Tools ................................ ................................ ................... 12 Challenges Faced in the Utilization of Autism Screening Tools ................................ .............. 13 Autism Diagnostic and Screening Instruments ................................ ................................ ......... 1 3 Diagnostic and Statistical Manu al for Mental Disorders, Fourth Edition, Text Revision (DSM - IV - TR) ................................ ................................ ................................ ....................... 1 4 Diagnostic and Statistical Manual for Mental Disorders, Fifth Edition (DSM - 5) Criteria .. 15 Autism Diagnostic Interview, Revised ................................ ................................ ................. 18 Autism Diagnostic Observation Schedule - 2 (ADOS - 2) ................................ ....................... 2 2 Social Communication Questionnaire (SCQ) ................................ ................................ ....... 25 Social Responsiveness Scale, Second Edition (SRS - 2) ................................ ........................ 29 Childhood Autism Rating Scale, Second Edition (CARS - 2) ................................ ................ 33 Gilliam Autism Rating Scale (GARS - 3) ................................ ................................ ............... 36 23 Question Questionnaire ................................ ................................ ................................ .... 38 Justification for Using the SRS and SCQ ................................ ................................ ................. 39 Summary ................................ ................................ ................................ ................................ ... 40 CHAPTER 3: METHODOLOGY ................................ ................................ ................................ 41 Purpose of the Study ................................ ................................ ................................ ................. 41 Research Questions and Hypotheses ................................ ................................ ........................ 41 Research Q uestion 1 ................................ ................................ ................................ ............. 41 Research Q uestion 2 ................................ ................................ ................................ ............. 42 Research Q uestion 3 ................................ ................................ ................................ ............. 42 Research Q uestion 4 ................................ ................................ ................................ ............. 43 Research Q uestion 5 ................................ ................................ ................................ ............. 44 Research Q uestion 6 ................................ ................................ ................................ ............. 44 Participants ................................ ................................ ................................ ................................ 4 5 Procedure ................................ ................................ ................................ ................................ .. 46 Review and M odification of the SRS - 2 and SCQ ................................ ................................ 46 viii Training of the R esearch A ssistants for I nstrument A dministration ................................ .... 47 Fidelity C heck. ................................ ................................ ................................ ...................... 48 Field - testing ................................ ................................ ................................ .......................... 4 9 Measures ................................ ................................ ................................ ................................ ... 50 Social Communication Questionnaire (SCQ) - Lifetime Form ................................ ............ 50 Social Responsiveness Scale, Second Edition (SRS - 2) - School - Age Form ........................ 5 1 The 23 Question Screener ................................ ................................ ................................ ..... 51 Data Checking and Entry ................................ ................................ ................................ .......... 5 2 Data Analysis ................................ ................................ ................................ ............................ 52 CHAPTER 4: RESULTS ................................ ................................ ................................ .............. 54 Demographic Characteristics of Parents ................................ ................................ ................... 54 Demographic Characteristics of Children of the Participating Parents ................................ .... 5 7 Research Questions and Hypotheses ................................ ................................ ........................ 58 Research Q uestion 1 ................................ ................................ ................................ ............. 59 Research Q uestion 2 ................................ ................................ ................................ ............. 6 2 Research Q uestion 3 ................................ ................................ ................................ ............. 6 4 Research Q uestion 4 ................................ ................................ ................................ ............. 6 6 Research Q uestion 5 ................................ ................................ ................................ ............. 7 3 Research Q uestion 6 ................................ ................................ ................................ ............. 8 3 CHAPTER 5: DISCUSSION ................................ ................................ ................................ ........ 8 7 Reliability ................................ ................................ ................................ ................................ .. 8 8 The SCQ. ................................ ................................ ................................ .............................. 8 8 The SRS - 2.. ................................ ................................ ................................ ........................... 8 9 The 23Q.. ................................ ................................ ................................ .............................. 9 0 Validity ................................ ................................ ................................ ................................ ..... 9 1 Mean Differ ences ................................ ................................ ................................ .................. 9 1 Sensitivity and Specificity ................................ ................................ ................................ .... 9 3 Alternative Optimal Cut - scores ................................ ................................ ............................ 9 6 Correlations Among Study Instruments ................................ ................................ ............. 100 Strengths and Limitations ................................ ................................ ................................ ....... 10 1 Clinical Implications ................................ ................................ ................................ ............... 10 5 Research Implications and Future Research Directions ................................ ......................... 10 6 Conclusions ................................ ................................ ................................ ............................. 10 9 APPENDICES ................................ ................................ ................................ ............................ 11 1 APPENDIX A : Table 21 ................................ ................................ ................................ ......... 11 2 APPENDIX B : Table 22 ................................ ................................ ................................ ......... 11 4 APPENDIX C: IRB Clearance in Uganda ................................ ................................ .............. 11 8 APPENDIX D: MSU IRB Clearance ................................ ................................ ..................... 1 20 APPENDIX E: ......... 12 1 APPENDIX F: Permission to Conduct Research with Children at Butabika National Hospital, Uganda ................................ ................................ ................................ .................... 12 2 APPENDIX G: Fidelity Checklist ................................ ................................ .......................... 12 3 APPENDIX H: Assent Form (8 - 17 years) ................................ ................................ .............. 12 4 ix APPENDIX I: Consent Form ................................ ................................ ................................ . 12 9 APPENDIX J: Demographic Form ................................ ................................ ......................... 13 3 APPENDIX K: 23 Questions Questionnaire ................................ ................................ ........ 13 5 REFERENCES ................................ ................................ ................................ ........................... 13 8 x LIST OF TABLES Table 1. Demographic Characteristics of Parents ................................ ................................ ......... 55 Table 2. Demographic Characteristics of Children ................................ ................................ ....... 58 Table 3. Sensitivity and Specificity of the SCQ Total Raw Score based on Cut - score of 15 ..... 67 Table 4. Sensitivity and Specificity of the SRS - 2 Total Score Based on Raw Cut - score of 70 ... 70 Table 5. Sensitivity and Specificity of the SRS - 2 Total Standardized Cut - score (T - scores) of 70 ................................ ................................ ................................ ................................ ....................... 70 Table 6. Sensitivity and Specificity of the SRS - 2 Total Raw Cut - score of 85 ............................. 70 Table 7. Sensitivity and Specificity of the SRS - 2 Total Standardized Cut - score (T - scores) of 85 ................................ ................................ ................................ ................................ ....................... 71 Table 8. Sensitivity and Specificity of the 23Q Based on Using a Cut - score of 1 or more on the Ten summed ASD Items ................................ ................................ ................................ ............... 72 Table 9. Receiver Operating Characteristic (ROC) Curve of the SCQ Total Cut - Scores ............ 74 Table 10. Sensitivity and Specificity of the SCQ Based on the Cut - score of 15 .......................... 75 Table 11. Sensitivity and Specificity of the SCQ Based on the Cut - score of 10 .......................... 76 Table 12. Summary of Sensitivity and Specificity of Recommended Versus Alternative SRS - 2 Raw and Standardized Cut - Scores ................................ ................................ ................................ 76 Table 13. Receiver Operating Characteristic (ROC) Curve of the SRS - 2 Alternative Total Cut - Scores Using Raw Scores ................................ ................................ ................................ ............. 77 Table 14. Detailed Breakdown of Sensitivity and Specificity Based on an SRS - 2 Raw - score of 62 Classification Analysis ................................ ................................ ................................ ............. 79 Table 15. Receiver Operating Characteristic (ROC) Curve of the SRS - 2 Standardized Alternative Total Cut - Scores (U.S. Norms) ................................ ................................ ................................ ..... 80 Table 16. Sensitivity and Specificity Based on an SRS - 2 T - score Classification of 62 in Study Children ................................ ................................ ................................ ................................ ......... 81 Total Cut - scores ................................ ................................ ................................ ............................ 82 xi Table 18. 23Q Sensitivity and Specificity of the 23Q Based on a cut - score of 3.0 ...................... 83 Table 19. Sensitivity and Specificity Based on the 23Q Cut - score of 4.0 ................................ .... 83 Table 20. Research Hypotheses Findings Summary ................................ ................................ ..... 85 Table 2 1 . Reliability of SCQ Total Scores Among Non - ................................ ................................ ................................ ......... 112 Table 2 2 . SRS - 2 Total Scores Reliability Among Non - ASD Group if Item is Deleted ............. 114 1 C HAPTER 1 : INTRODUCTION Autism Spectrum Disorder The fifth edition of the Diagnostic and Statistical Manual (DSM - 5) defines Autism Spectrum Disorder (ASD) as a condition characterized by impairment in two domains: social - communication and interaction; and r estrictive, repetitive behaviors, activities, or interactions (American Psychiatric Association [APA], 2013). In the Diagnostic and Statistical Manual, Fourth Edition, Text Revision (DSM - IV - TR), autistic disorder was defined by impairment in the three area s of social interaction, communication, and restrictive, repetitive, and stereotyped patterns of behavior (APA, 2000). Autism did not have its separate category under Pervasive Disorder, Childhood Disintegrative Disorder, and Pervasive Developmental Disorder Not Otherwise Specified (APA, 2000). In addition to the major recategorization of the five disorders, the diagnostic criteria of autism merged the social and communic ation subdomains, which were independent domains in the DSM - IV - TR. The repetitive restricted behaviors, activities or interactions, and the requirement that onset of impairments occur before the age of three years of age have held constant across the vario us iterations to the autism diagnostic criteria (APA, 2013; Lecavalier & Norris, 2010; Volkmar, 2012; Young & Rodi, 2013). In recent years, most researchers and practitioners have used the terms autism and autism spectrum disorder interchangeably to refer to the same disorder. Such interchangeable usage will occur in the context of the present study as well. As of 2015, it was estimated that one in every 145 children worldwide had autism spectrum disorder (Presmanes Hill, Zuckerman, & Fombonne, 2015). Some researchers posit a prevalence rate of one percent and 1.5% for ASD in developing and developed countries 2 respectively. One in 54 children in the United States, by age the age of eight, has an ASD diagnosis, according to the most recent estimate from the Autism and Developmental Disabilities Monitoring [ADDM] Network (Maenner, Shaw, Baio, et al., 2020). Well - developed population - based prevalence studies for ASD in Africa are lacking, but researchers and practitioners suspect that prevalence may be higher t han in developed countries. For instance, Kakooza - Mwesige and colleagues (2014) estimated that 1.2 1.3/100 children have ASD in Uganda. Although concerns about the sampling procedure put this estimate in doubt (Abubakar, Ssewanyana, & Newton, 2016), it n evertheless provides a preliminary, best available estimate of the prevalence in Africa. It is important to note that the estimates from Presmanes Hill et al. and Kakooza - Mwesige et al. occurred close in time to each other because evidence from developed c ountries suggests that the prevalence of ASD diagnosis has been increasing over time and that the most recent estimate in the U.S. from the Centers for Disease Control and Prevention (CDC) puts the U.S prevalence rate at 1 in 54 (Maenner et al., 2020). As sessing for ASD and its Challenges in Africa Researchers and practitioners agree that the lack of validated ASD screening assessments hampers the diagnosis and rehabilitation efforts of children with ASD (Abubakar et al., 2016; Franz, Chambers, von Is enburg, & de Vries, 2017). Most clinicians in Africa rely on the DSM - IV - TR (American Psychiatric Association, 2000) for ASD diagnosis, which although readily available, needs extensive clinical training and advanced education for one to administer appropri ately. In more developed countries, a clinical psychologist needs doctoral training to administer advanced diagnostic instruments such as the Autism Diagnostic Observation Schedule (ADOS) 3 (Lord et al., 2000), and the Autism Diagnostic Interview - revised ( ADI - R) (Rutter, Le Couteur, & Lord, 2003), which are considered the gold standard for ASD diagnosis. Moreover, unlike most developing countries, diagnosticians in the developed world are utilizing the DSM - 5 despite arguments for or against its use. Such ad vanced training is limited in Africa, is held by a few psychologists, or neurologists, and yet the number of people in need of their services is staggeringly high (Mullan et al., 2011). Besides, the widespread practice is that primary healthcare workers wi th limited training run rural health centers in Africa (Mullan et al., 2011). The few better - trained and equipped clinicians often live and work in the urban areas (Tekola et al. 2016). The characteristics of ASD are closely related to other neurodevelop mental disorders. Without a precise differential screening tool, ASD is likely to be misdiagnosed for other disorders such as Attention Deficit and Hyperactivity Disorder (ADHD), Intellectual Disability (ID) or other closely related neurodevelopmental ASD comorbidities. The thin line between ASD and other neurodevelopmental disorders calls for good differential diagnostic skills, which only well - trained clinicians possess, as well as instruments. Subsequently, the risk of misdiagnosis may be high in Africa because of the limited number of practitioners trained to competently carry out a precise diagnosis of ASD. Given this problem, it is not surprising that (Smith, Malcolm - Smith, & de Vries, 2017) . Furthermore, there is often a long waitlist before a confirmation of a diagnosis takes place for a child with ASD, let alone receiving services. A useful screening tool that does not require advanced level professionals will streamline the assessment proce ss and better enable children that are at risk for ASD to receive appropriate services early. This practice is likely to help focus 4 The need for early detectio n using a screening tool is critical as it will help with early identification of children at risk of ASD by practitioners with less advanced diagnosis training serving primarily in the rural areas, hence increasing the number of referrals for better diagn ostic assessment. Referrals ensure that the highly specialized efforts wielded by the few well - trained professionals are focused on those identified as being at higher risk so that they can be referred to rehabilitation services promptly (Eaves, Wingert, & Ho, 2006). Uganda is among the few African countries where there has been an attempt at validating a culturally sensitive screening instrument, which caters for children at risk of ASD and other neurodevelopmental disorders l ike epilepsy, cerebral palsy , and cognitive impairment, among others (Kakooza - Mwesige et al., 2014). There was a limited attempt at validating the Autism Diagnostic Observation Schedule, Second Edition (ADOS - 2) (Smith, Malcolm - Smith, & de Vries, 2017) in South Africa with a small sam ple size ( N = 7). It was translated into the Afrikaans version to generate guidelines to enhance cultural sensitivity and appropriateness. Kakooza - Mwesige and others (2014) stressed the need for further work in the validation of autism - related screening instruments in low resource countries to assist in the identificati on of affected individuals and boost rehabilitation efforts. Purpose of the Study Without deliberate efforts to identify ASD assessment and screening instruments that fit the local context, we run the risk of missing the benefits accruing from early di agnosis and management of ASD in African countries, including Uganda. It is likely that a substantial proportion of children with ASD will not receive rehabilitation services. This poses long - term negative impact in their adulthood, as well as overall econ omic impact because of additional 5 services needed for impacts that could have been alleviated or reduced. The primary purpose of this study, therefore, is to investigate the psychometric properties of two commonly used, well - validated ASD screening instrum ents used in the US with children in Uganda. The Social Responsiveness Scale, Second Edition (SRS - 2; Constantino & Gruber, 2012) and Social Communication Questionnaire (SCQ; Rutter, Bailey & Lord, 2003) will be assessed for utility in the screening of ASD symptoms for children aged 4 - 18 years in Uganda. The ten ASD specific items of the 23Questions screener (23Q; Kakooza - Mwesige et al., 2014), an instrument developed in the African context to screen for a broad array of neurodevelopmental disorders, includ ing ASD, will be used to examine concurrent and discriminant validity. Potential Implications of the Study The results of this study will provide an initial step in understanding the utility of the SCQ and SRS - 2 in applying to children at risk of ASD in the Ugandan culture. The study will have significant application to the identification of valid, reliable, and efficient ASD screening tools, which will aid in the early identification of those at risk of ASD. The broader community of professionals, such a s teachers, counselors, social workers who may not have advanced ASD identification training such as in medicine and psychology will also use them. Combined with ease of use by front - line clinicians, it will allow for improved early screening and identific ation of at - risk children and help free up time for pediatricians, and psychologists to conduct formal ASD diagnoses. Children diagnosed with ASD will then be more likely to be referred for appropriate services with delay improving the likelihood of better adult outcomes. Therefore, results of this study could potentially lead to the validation of the first tool specifically adapted for ASD screening in Uganda. This will improve the likelihood of early identification and access to appropriate services for c hildren with ASD and their families. 6 Research Questions 1) In the Ugandan context, do total scores for the three ASD instruments (i.e., SCQ, SRS - 2, and 23Q) yield adequate internal consistency relative to values reported in the test manual or available studie s for those with ASD? 2) In the Ugandan context, do total scores for the three ASD instruments (i.e., SCQ, SRS - 2, and 23Q) yield adequate internal consistency relative to values reported in the test manual or available studies for non - ASD cases? 3) In the Uganda n context, do total scores from the three instruments (i.e., SCQ, SRS - 2, and 23Q) yield substantial mean differences consistent with construct - related (i.e., ASD construct) differentiation between ASD and non - ASD groups? 4) In the Ugandan context, will the u se of the recommended screening cut - scores for each of the three instruments (i.e., SCQ, SRS - 2, and 23Q) result in adequate sensitivity and specificity using DSM - 5 ASD diagnosis vs. non - ASD cases as the outcome variable? 5) Beyond the recommended cut - scores f or screening reported in each test manual, are there more optimal cut - scores for each of the three instruments (i.e., SCQ, SRS - 2, and 23Q) in the Ugandan context? 6) Are there significant convergent relationships among the total scores of the three instrumen ts (i.e., SCQ, SRS - 2, and 23Q) in the Ugandan context? 7 CHAPTER 2 : LITERATURE REVIEW Autism Spectrum Disorder Autism Spectrum Disorder (ASD) is a disorder that is marked by impairment in social communication or social interaction as well as restricted, repetitive patterns of behavior, interests, and activities in individuals (APA, 2013). Leo Kanner first reported autism in 1943, where he referred to it as infantile autism. According to Kanner, children who had infantile autism tended to isola te themselves from others, had limited language abilities and enjoyed doing a specific activity repetitively. This profile mimics the description of a child with autistic disorder, according to the fourth edition of the Diagnostic and Statistical Manual, T ext Revision (DSM - IV - TR). In 1946, Hans Asperger reported another group of boys with similar characteristics who appeared to be higher functioning. Cases who demonstrated this pattern, which Asperger referred to as having autistic psychopathy of childhood, later came to be referred to as Asperger's Disorder (Miller & Ozonoff, 1997). Kanner is credited with being the first to call the condition autism. Both Kanner and mented about autism in the English - speaking world because the original descriptions of children with autism were in German. It was not until Van Krevelen (as cited in Wing, 1986) whose paper was urging s autism/syndrome was published. Lorna - speaking scientific community 8 by describing a series of cases like those of Asperger (Klin & Volkmar, 2005; Woodbury et al., 2005). There have been hea ted debates about the clinical difference between autism and - IV. Most clinicians argued that making a - IV was difficult (Woodbury et al., 2005). The signi was, therefore, challenging to make a conclusive diagnosis in situations wher e a parent or other primary caregiver was not accurate on developmental milestones of the child (Klin & Volkmar, 2005). Researchers re - Syndrome using the DSM - IV and showed that they would be re - assigned to a diagnosis of research and clinical disagreements have been the major contributors to the merging of the different autism subdomains in one, i .e., autism spectrum disorder in the current DSM - 5. Etiology and Prognosis Autism Spectrum Disorder etiology has been associated with both genetic and environmental factors, but not a single biomarker has been identified for the diagnosis of one with ASD (Bello - Mojeed, Omigbodun, Bakare, & Adewuya, 2017). Parental history of mental pregnancy, bearing children while of advanced age (males 50 years and older, and fe males 40 years and older), genetic factors, environmental toxins are considered potential risk factors for ASD (APA, 2013). However, genetic factors have the most robust empirical support for ASD 9 etiology thus far. Research in twin concordance rates for au tism indicates hereditary estimates between 37% to above 90% (Geschwind, 2011). The age of onset for ASD is between 12 to 24 months (about 2 years). Some children are, however, diagnosed later beyond three years of age (APA, 2013; Meng et al., 2013). Rese arch that enables the identification of younger children at risk of ASD has been ongoing. Children as young as six months old can now be screened for ASD, although most accurately diagnosed at two years when apparent deficits can be ascertained. ASD onset is characterized by either rapid or gradual child developmental regression in language, social interaction and restricted, repetitive patterns, behavior, or activities (APA, 2013). More boys than girls are diagnosed with ASD, with a male to female ratio o f between 4:1 and 5:1, and most females diagnosed with ASD in clinical samples having co - occurring intellectual disability (APA, 2013). Symptom masking, support, and compensation, especially as related to societal gender roles, may mask the deficits in soc ial abilities among higher - functioning females with ASD (APA, 2013; Lai, 2015). Thus, detection and diagnosis are more difficult in females (especially those who are high functioning). Autism is a lifelong disorder although those diagnosed with it currentl y have a better outlook to life when compared to more than 50 years back (APA, 2013; Lord et al., 2018; Meng et al., 2013). More individuals with ASD can work and live independently in the community depending on the level of severity. Those whose language and intellectual abilities are less affected (non - impaired or intact) can function better in adulthood. Early identification and interventions have been associated with better adult outcomes (Lord et al., 2018). 10 Global Burden Autism prevalence rates a re projected to be increasing worldwide (APA, 2013; Sharma, Gonda & Tarazi., 2018). Greater awareness, better diagnosis, and screening instruments, and change in the diagnostic criteria from the DSM - IV - TR to the DSM - 5 are some of the reasons advanced for t he increasing prevalence (Magaña & Vanegas, 2017). Having a child with ASD has been associated with poor mental and physical health outcomes for parents. Parents of children with ASD often report higher levels of stress and depression resulting from const antly worrying about their child (Lee et al., 2009) . Diagnosis, rehabilitation, and service needs of children with ASD are expensive and often lead to many negative outcomes, such as spousal conflicts (Gona et al., 2015; Schlebusch, Dada, & Samuels., 2017). This results in poor quality of life for parents. This is further worsened by negative societal attitudes and the limited number of professionals capable of providing the services needed for the functioning of children with ASD. Parents have a few resources for respite care and as a result isolate themselves, which further increases the risk for stress. The much n eeded social and professional support sought by parents could be affected, hence, negatively affecting the rehabilitation efforts for the child with ASD (Fewster & Gurayah, 2015; Franz, Chambers, von Isenburg, & de Vries, 2017; Gona et al., 2015; Kakkar & Srivastava, n.d .) . Autism Diagnosis The first clinical documentation of the criteria to diagnose autism was captured in the Diagnostic and Statistical Manual of Mental Disorders, Third Edition [DSM - III] in 1980 and the International Classification of Diseases [ICD - 10 ]. In the DSM - III, autism was recognized as a unique disorder in children with no known etiology, but with a different clinical presentation. There have been several changes in the categorization of autism right from the first time it was 11 classified for cl inical purposes in the DSM - III - TR (DSM - III, APA, 1980). Research in autism increased as a result and led to changes in the descriptions and diagnostic criteria in the subsequent versions of the DSM. For instance, in the DSM - IV - TR, autism was categorized un der Pervasive Developmental Disorders (PDDs) with other related diagnostic entities. The five PDDs (CDD) and Pervasive Developmental Disorder - Not Otherwise Spec ified (PDD - NOS), a designation for cases that did not meet full criteria for another PDD but shared significant core symptoms (APA, 2000). In the latest revision of the Diagnostic and Statistical Manual for Mental Disorders, Fifth Edition (DSM - 5), the prio r diagnostic designations of autistic disorder, - NOS under DSM - IV - TR were re - organized into the single diagnostic category of Autism Spectrum Disorder (Magana & Vanegas, 2017). The merging of three of the PDDs into one diagnos is of ASD in the current DSM - 5 has been met with mixed reactions. Some self - advocates, parents of children with ASD and the concerned community at large, argue that the DSM - 5 changes reduce the chance that a child is diagnosed with autism, hence impacting disability insurance (Magana & Vanegas, 2017). Others argue that the change will increase the likelihood of reliable and accurate diagnosis (i.e., reduce the likelihood of missed or false negative c ases). For example, Magana and Vanegas (2017) showed that the DSM - 5 increased the likelihood of children of Hispanic descent being diagnosed with ASD when compared to the DSM - IV - TR. These authors indicated that the DSM - 5 identified more children with ASD o f Spanish speaking parents, with the social communication and social interaction domains being significantly different for the ASD and developmental disability groups under the DSM - 5 12 revision. This was not the case under the DSM - IV for the same demographic group (Magaña & Vanegas, 2017; Vanegas, Magaña, Morales, & McNamara, 2016) . Despite the different views on the DSM - 5, autism diagnosis takes advanced tra ining for one to appropriately utilize the various diagnostic and screening tools. Often, these advanced diagnostic assessments rely on a combination of parental reports on child developmental history or milestones, and child observation by multidisciplina ry teams (pediatric psychiatrists, psychologists, counselors among other professionals) (Lord et al., 2018). Advantages of Autism Screening Tools For children living in developing countries like Uganda, an ASD diagnosis is often tedious and expensive wit h little guarantee of an accurate diagnosis, especially for those living in rural areas (Bakare & Munir, 2011; Bello - Mojeed et al., 2017; Mazurek, Curran, Burnette, & Sohl, 2019). Professionals trained in advanced autism assessment are scarce, with most li ving in urban areas, hence reducing the chances for early diagnosis for children and family living in the rural settings due to inaccessibility issues. Autism screening, however, does not require advanced training for one to administer. This reduces diagno sis delay for individuals from a low socioeconomic status as well as costs to parents, as screening tools are less expensive than advanced diagnostic assessment. Also, timely and appropriate referrals lessen the effect of potential societal stigma to both parents and their children through timely access to services and support by more knowledgeable professionals (Kakooza - Mwesige et al., 2014; Mazurek et al., 2019; Woolfenden, Sarkozy, Ridley, & Williams, 2012). Self - rating scales such as the Social Communication Questionnaire [SCQ] (Rutter, Bailey & Lord, 2003), Social Responsive Scale, second edition [SRS - 2] (Constantino & Gruber, 2012) and others that are administered to parents or primary caregivers of childre n often reduce the 13 time required of the clinicians to make a diagnosis. They are cost - effective, less complicated to administer and can identify those in need of further clinical assessments. Challenges Faced in the Utilization of Autism Screening Tools Finding a uniform autism screening tool catering for all cultural and demographic variance is complicated. Most are developed in North America and not validated for use in African cultures (Durkin et al., 2015). Such unvalidated instruments are nonetheless utilized in ASD screening by professionals in some African countries. The possibility of misdiagnosis stemming from these unvalidated tools is therefore high. False positives resulting from using non - validated screening tools could lead to inappropriate i ntervention referral, stigma, anxiety, and extra financial burden on the parents as well as the healthcare system (Marlow et al., 2019). Further, younger children with less visible ASD characteristics are hard to identify especially where primary healthcar e providers are not well trained. Autism screening instruments are equally expensive, and demand a better trained human resource, which is scarce in Africa. Professionals with limited ASD training often fill the human resource gap (Marlow, Servili, & Tomli nson, 2019). Autism Diagnostic and Screening Instruments Diagnosis of ASD involves a combination of procedures. In psychology, behavioral observation, and taking of developmental history from primary caregivers of the person being assessed form the mo st reliable basis for an autism diagnosis. Below are highlighted standard diagnostic and screening instruments used in the identification of autism. 14 Diagnostic and Statistical Manual for Mental Disorders, Fourth Edition, Text Revision (DSM - IV - TR) According to the DSM - IV - TR, to be diagnosed with autism, one must meet the set criteria for autistic disorder. Specific features outlined in the three core areas, i.e., A, B and C as delineated in the DSM - IV - TR must be met. One must have a total of six or more items across all the sub - categories (A, B, or C) with at least two from sub - category: A) qualitative impairment in social interaction, and one under both B) qualitative impairments in communication and C) restricted repetitive and stereotyped patterns of behavior, interests, and activities (APA, 2000). For qualitative impairment in social interaction (criteria A), one must have at least two or more features of the following: 1) marked impairment in the use of nonverbal behaviors like eye - to - eye gaze, facial expression, body postures and gestures to regulate social interaction. 2) failure spontaneous seeking to share interests or enjoyment, and 4) a lack of socia l or emotional reciprocity (APA, 2000). For qualitative impairment in communication (criteria B), one must have at least one of the following: 1) a delay or a total lack of the development of spoken language (not accompanied by alternative modes of commun ication like gestures. 2) a marked impairment in the ability to initiate and sustain a conversation with others (among those with adequate speech). 3) repetitive or stereotyped use of language (echolalia), and 4) lack of varied, spontaneous make - believe pl ay that is appropriate to their developmental level (APA, 2000). A person must meet at least one criterion for the Restricted Repetitive and Stereotyped Behavior (criteria C): 1) intense or abnormal preoccupation with one or more interest(s). 2) inflexib le adherence to routine or rituals. 3) stereotyped and repetitive motor mannerisms (hand 15 or finger flapping or twisting, or complex whole - body movements, and 4) persistent preoccupation with parts of objects. It is important to note that to be diagnosed w ith autistic disorder, evidence of delay or abnormalities manifested before three years of age is required. Also, the symptom presentation should not have been better explained by a diagnosis of Rett's Disorder or Childhood Disintegrative Disorder (APA, 20 00). their language development, but with sustained impairment in social interaction, development of restricted, repetitive patterns of behavior, interests, and activitie s. These must cause significant impairment in social, occupational, or other areas of daily functioning. While for a diagnosis of PDD - NOS, one must have pervasive impairment in social interaction and either the presence of stereotyped behavior, interests a nd activities, or substantial communication impairment(s), or due to a late age of onset, some symptoms being subclinical or insufficient to meet criteria fo r another PDD, unusual variation in symptomology, etc. (APA, 2000). Diagnostic and Statistical Manual for Mental Disorders, Fifth Edition (DSM - 5) Criteria According to the DSM - 5 (APA, 2013), to be diagnosed with an autism spectrum disorder, one must meet t he diagnostic criteria established by having deficits in all the two outlined criteria, that is: 1) social communication and social interaction domain (criteria A), and 2) at least two of the four outlined criteria under the restricted, repetitive sub - doma in (criteria B). Criterion A delineates the social communication and interaction aspects of a person with autism, which includes: 1) deficits in social - emotional reciprocity, 2) difficulty in nonverbal 16 communication behaviors, and 3) deficits in developin g, maintaining and understanding relationships (APA, 2013). Criteria B pertains to the restricted and repetitive behavior, interaction and activities subdomain of a person suspected to have autism. Under this subdomain, one must have any two of the four c riteria outlined: 1) stereotyped or repetitive motor movements or use of objects, 2) insisting on sameness, inflexibility in routines, or ritualized patterns of verbal or nonverbal behavior, 3) highly restricted, fixated interests that are abnormal in inte nsity or focus, and 4) hyperactivity to sensory input or unusual interest in sensory aspects of the environment (APA, 2013). The deficits outlined above must have been present in the early developmental period. However, such difficulties may not be fully manifest until social demands exceed the limited capacities of the individuals and may be masked by learned strategies later in life (APA, 2013). Also, the symptoms must cause severe impairment in social, occupational, or other areas of the t functioning. The outlined disturbances ought not to be better explained by intellectual disability (intellectual developmental disorder) or global development delay (APA, 2013). Specifiers . The DSM - 5 goes further in delineating different specifiers of s ymptom severity utilized in the diagnosis of individuals with autism. The various levels in the specifiers include: Level 1 : In this level, such a person will have difficulty in the social communication domain. Individuals with this specifier require supp ort. Without the necessary support in place for them, the difficulty in this domain makes their impairments noticeable. They experience difficulties in initiating and sustaining social interaction. For example, such individuals may be 17 able to speak in full sentences, engage in communication with others, but their back - and forth interaction with others fails. Their attempts to make friends are unsuccessful and odd. For the restrictive, repetitive behaviors domain, the individual is inflexible in their behav ior, and it significantly interferes with their functioning in one or more contexts. This behavioral rigidity makes it hard for them to switch between activities. They also experience difficulties with planning, which hinder their independence (APA, 2013). Level 2 : This level is characterized by increased difficulty in both verbal and nonverbal communication skills. Unlike in level 1, even with the availability of support, social impairments are evident in the person. Self - initiated social interactions are limited i n nature, and they have odd or reduced responses to social interaction from others. Such a person will, for example, speak simple sentences and their interaction with others limited to a specific narrow particular interest and a markedly odd nonverbal comm unication. The rigidity in behavior or other restricted, repetitive interests appears often enough in various settings to be evident to a casual observer. They experience distress and difficulty changing focus or action. These individuals require substanti al support to function (APA, 2013). Level 3 : For one to receive level 3 for a specifier, he or she must require very substantial support. People with this specifier have immense difficulties in verbal and nonverbal communication, which impairs their inter action with others. Their initiation, ability to sustain or respond to the interactions or conversation with others is very limited. They have less intelligible speech, few words and rarely initiate interaction. Even when they do, it is to meet their own n eeds (APA, 2013). They mostly respond to straightforward social approaches by others. Their behaviors are not flexible, and they have incredibly high difficulty coping with change and other 18 restricted or repetitive behaviors. This remarkably affects their functioning in all contexts. They also experience great difficulty in changing their focus or action. Severity specifiers (i.e., levels 1, 2 and 3) are used to describe the current symptoms of the person being diagnosed. However, in assigning individuals specifiers, professionals should be cognizant of the fact that ASD severity may vary depending on where (context) and when (time) in which the behavior unfolds. Autism Diagnostic Interview, Revised The Autism Diagnostic Interview - Revised (ADI - R) (Rutter, Le Couteur, & Lord, 1994) is an instrument administered to parents or other primary caregivers of a chi ld suspected to have autism by professionals knowledgeable in autism symptoms and presentation. It has two forms, that is, the summary and lifetime forms. The interview entails having a parent or caregiver of a child answer questions about symptoms related to social communication, interaction and restrictive, repetitive activities and behaviors (Lord et al., 2000). Lord and colleagues published the ADI - R in 1994 after modifying the Autism Diagnostic Interview in 1989 (Lord et al., 1994). The ADI - R administe red to parents of children with cognitive abilities corresponding to a developmental level of at least two years. The instrument has 93 yes or no items that gather background as well as current information in the three areas of language and communication, reciprocal and social interaction as well as restricted, repetitive, stereotyped behaviors in the child. The ADI - R scores range from 0 (behavior not - present at all), 1 (behavior mildly present, 2 (behavior significantly present), to 3 (behavior severely p resent). The severity code of 3 is treated as 2 to reduce the undue weight that could be placed on individual items (Lord et al., 1994). During its administration, the interviewer can ask follow - up questions about the child to 19 appropriately score the behav ior. The scale scores are totaled and coded by the test administrator. Administration of the interview takes between 1.5 to 2.5 hours to complete. It provides five different algorithms, i.e., two utilized for diagnostic purposes and three to assess current functioning as well as help with treatment planning. A child with recommended cut - scores in the three domains, that is, social reciprocity (10), communication (8 or 7 for verbal and non - verbal cases respectively) and restrictive repetitive behaviors (3) i s considered to have autism. Also, these behaviors should have manifested in at least one of the subdomains before three years of age. The ADI - R was initially validated on a sample of 20 children (i.e., n = 10 with an autism diagnosis and n = 10 without au tism, but with a mental handicap or language delay). They ranged in age from 30 to 39 months. The ADI - R scores yielded intra - class correlation coefficients ranging from 0.93 to 0.97 and test - retest reliability estimates of 0.93 to 0.97 among interviewers. Further ADI - R validity evidence was gathered through participation of another group of parents ( N = 50) who were seeking help for their preschool children. Among these parents, n = 25 had children with autism and a second group of n = 25 had children with ranged from 36 to 59 months with mental age ranging from 21 to 74 months. Out of the total sample, interrater reliability was assessed using a subsample of 20 ch ildren (i.e., 10 with autism and 10 without autism) with multiple raters. This inter - rater analysis yielded kappa coefficients ranging from .63 to .89 for each item. Intra - class correlations were above .92 for all subdomain and domain scores.) Only one of the 25 children with autism failed to meet the algorithm criteria for autism under the ADI - R. Two of the 25 children without autism, but with mental handicap and language, yielded ADI - R scores consistent with autism. Thus, the ADI - R criteria for autism, 20 as outlined by Lord and colleagues (1994), provided good discrimination between the two groups (Rutter, Le Couteur, & Lord, 2003). Application of the ADI - R has occurred in other linguistic and cultural contexts. For example, in a study that utilized a Spani sh version of ADI - R, Magaña and Smith (2013) compared 48 Latino adolescents and adults who had ASD with a matched sample of non - Latino Whites using the summary and lifetime scores on the ADI - R. They found no significant differe nce between the two groups on total impairment of social reciprocity or communication; however, the Latino group had lower levels of restricted and repetitive behaviors compared to their White counterparts. Eight of the Latino and three of the White childr en did not meet the autism cut - score for the repetitive behavior domain with an interrater reliability of .88 for the interviewers. According to the study authors, this might have happened because Latino mothers often place more emphasis on interpersonal s ocial development as compared to goals related to the development of autonomy among their children (Magaña & Smith, 2013) . Tsuchiya and others (2013) investigated the psychometric properties of a Japanese version of the ADI - R (i.e., the ADI - R - JV). To establish interrater reliability, they utilized a total sample of 51 individuals aged 2 - 19 years. These were divided into two groups: 1) 35 individuals who were clinically referred for ASD, and 2) 9 control group with no ASD diagn osis. For discriminant and diagnostic validity, the authors recruited 317 individuals aged 2 - 19 years and divided them into three groups: 1) n = 138 with autistic disorder (AD), 2) n = 28 PDD - NOS, and 3) n = 90 with non - ASD. Intra - class correlations of 0.80 were obtained for the three different domains for interrater reliability. For discriminant validity, the mean scores for those with AD were higher than the other two groups. The sensitivity and specifici ty of the ADI - R - JV were 21 reported at 0.92 and 0.89 respectively. However, the sensitivity of those below five years of age was 0.55 (Tsuchiya et al., 2013) . Another study with the Polish version of the ADI - R (ADI - R - PL) established psychometric validation of the assessment for use in Poland (Chojnicka & Pisula, 2019) . The study consisted of 125 individuals aged two years old and above. They were divided into two groups, 1) n = 65 individuals diagnosed with ASD and 2) n = 60 controls, among whom 18 had non - spectrum disorders and 42 we re typically developing children. The norming sample comprised a total of 178 individuals ( n = 118 with ASD and n = 60 with no ASD). Intra - class correlations for children of two years old and above ranged from .96 - 1.00 while the test - retest reliability fo r the various domains was .88 - .91 for those with two years and above. Cronbach alpha for all domains ranged from .85 to .95 except for the areas of reciprocal peer interaction (.64) and restrictive repetitive behaviors (.63). The authors also reported the Comparative Fit Index (CFI) for the ADI - R - PL at .88 and .89 (slightly lower than the recommended estimate of .90 and above) for the two and three factor model, respectively. A root means square error of approximation (RMSEA) of .08 was established for bot h models, which is above the recommended .06 to indicate a good model fit. They concluded the three - factor model following the norming study by Lord (1994) for the ADI - R - PL. In summary, evidence suggests that the ADI - R is a good diagnostic tool, but also one that requires substantial training to administer, score, and interpret appropriately. It takes a long time to administer and requires extensive training. The ADI - R should not be used as a single tool for diagnostic purposes, but with other complimentar y assessments such as the ADOS - 2 (Falkmer et al., 2013). Limited studies were published in the use of the ADI - R in other languages or cultures. 22 Initial results appeared to be mixed in terms of its psychometrics when applied to diverse cultures. Additional studies are warranted. Autism Diagnostic Observation Schedule - 2 (ADOS - 2) The Autism Diagnostic Observation Schedule - 2 (ADOS - 2; Lord et al., 2012) is a semi - r disconfirm a potential diagnosis of ASD (Lord et al., 2012). Assessment using the ADOS - 2 entails activities requiring the child or person being assessed to elicit behaviors characteristic of people with ASD. It is a standardized assessment of communicati on, social interaction, play or imaginative use of materials as well as restricted and repetitive behaviors for individuals suspected of having ASD. Administration of the ADOS - coding, and scorin g them using the scoring algorithm provided by the test providers. The choice of the module to be administered depends on one's expressive language skills, chronological age, - 2 has five modules: Module 1: children with no language skills; Module 2: children with minimal language skills (are not verbally fluent); Module 3: children and young adults with phrase speech; Module 4: older adolescents and adults and Module 5: a toddler module, which is utilized for toddlers aged 12 - 30 months (about 2 and a half years). Modules 1 and 2 can be administered to adolescents or adults with expressive language levels meeting the two modules although their interpretation must be made with caution beca use the instrument validation samples did not include adults or adolescents (Kamp - Becker et al., 2018). It takes between 40 to 60 minutes to administer each module with the possibility of scores ranging from 0 (the behavior targeted is not present) to a sc ore of 2 or 3 (indicating the presence of a targeted behavior). The ADOS - 2 was found to have 23 an inter - rater reliability ranging from .78 - .98 (for modules 1 - 3) among research trained assessors (Lord et al., 2012) and a median inter - rater reliability of . 74 - .84 (for all the four modules) among clinically trained ADOS users conducting clinical work across 13 sites in Sweden (Zander et al., 2016). Limited studies have been conducted in the use of the ADOS in other cultural contexts. For instance, the ADOS - 2 (ADOS - 2 - PL) was validated for use in Polish using a sample of 407 study participants. They were divided into two groups, 1) n = 193 with ASD, and 2) n = 208 without ASD (78 with non - autism spectrum and 130 typically developing). Its test - retest estimate s ranged from .71 - .95 for all algorithms except for the restrictive repetitive behaviors algorithm of module 2, which was .41, module 3 (.54) and module 4 (.65). The internal consistency of all modules in the ADOS - 2 - PL was above .70 for all modules except module 3, , all modules were above 0.80 were higher in comparison to those without ASD (Chojnicka & Pisula, 2019) . - - translated ADOS - 2 was used to assess 47 children. Forty were from the community while seven were clinical samples referred to Red Cross War Malcolm - Smith & de Vries., 2017). Four of the seven children referred to the clinic had prior clinical diagnoses of ADHD, intellectual disability, global development delay and ectodermal dysplasia - cleft syndrome. The inclusion criteria of all participants were: being colored, ability to speak Afrikaans, from a low SES and having a child aged 1 - 20 years. All participants were divided into 24 two groups. Group 1 consisted of 24 caregivers with children under 10 years of age, with 20 parents from the community sa mple and four from the clinical sample. Group 2 consisted of 23 caregivers whose children were 10 or more years old, with 20 parents from the community sample and three from the clinical sample. The study authors reported limited psychometric properties fo r the instrument. However, they pre - piloted the translated ADOS - 2 (Afrikaans version) with four caregivers from the clinical sample and the examiner. The community samples were utilized for the validation of the appropriateness of social interactions and activities. For the clinical sample, language, activities, and cultural appropriateness of the ADOS - 2 Afrikaans version were performed. Caregiver comments about the instruments were documented and thematically analyzed. Children and caregivers were not fa miliar with certain items, including the bathtub, remote controlled bunny, or baby dolls, among others. They were, however, familiar with plastic cups, plates, and utensils clearly indicating the inappropriateness of some test items. Such discrepancies aff ect the clinical validity as deductions of scores would have been made in such contexts. The ADOS - 2 Afrikaans version by Smith and researchers (2017) provides an initial exploration of the application of the ADOS - 2 in the African culture. In summary, the A DOS is often used by well - trained professionals to make ASD autism. Its potential application in the African context, in translation and in terms of cultural app ropriateness, is currently being explored. Both the ADOS and ADI - R are reliable instruments in distinguishing children with and without ASD. However, for proficiency in its administration, one requires considerable training. Also, it is time - consuming to administer both instruments (Falkmer et al., 2013). Their prices are 25 equally prohibitive to most professionals and yet one must continuously purchase additional copies from the test authors. Autism screening tools are suitable alternatives because they ar e cheaper, require less training to administer, and are not time - consuming and yet with acceptable to high reliability as shown below. Social Communication Questionnaire (SCQ) The Social Communication Questionnaire (SCQ) was initially referred to as the Autism Screening Questionnaire (ASQ; Berument, Rutter, Lord, Pickles, & Bailey, 1999). The SCQ cannot be used to diagnose ASD in isolation but is useful as a screener for identif ying those in need of further assessment by professionals who are appropriately trained and equipped to diagnose autism. It takes about ten minutes to fill out and five minutes to score. The SCQ has been validated for use with individuals whose chronologic al age is over four years and have a mental age of at least two years (Gau et al., 2011; Lord et al., 1994). The items for the SCQ were adapted from the diagnostic scoring algorithm of the ADI - R, which incorporate content that is also consistent with the DSM - IV diagnostic criteria for autistic disorder (APA, 1994). There are two different forms of the SCQ, that is, the Lifetime and Current forms. Both consist of 40 items although item wording is slightly different across the forms to reflect the purpose a nd timing of the assessment. The Lifetime form is intended for diagnostic screening and the Current form is intended for assessing change over time by reflecting behavior from the most recent three - month period (Rutter, Bailey, & Lord, 2003). Only the Life time form was used in the present study. The SCQ Lifetime form, referred to simply as the SCQ going forward, is a 40 - item instrument filled out by parents or caregivers most familiar with the child suspected of having ASD. All items are responded to using a dichotomous (i.e., yes/no) scale (Berument et al., 1999; 26 w ill be included in the calculation of the total score. For those with sufficient language, their SCQ total score can range from 0 to 39, based on the summation of items 2 through 40. For those without sufficient language, the total score can range from 0 t o 33, as only items 8 through 40 are included in the calculation of the total. An optimal cut - score of 15 or more is utilized to differentiate ASD from other diagnoses. A slightly lower cut off is suggested, but not specified, for those with a substantial language impairment or other risk factors. Those under this condition are referred for further evaluation by more experienced professionals for a final diagnostic determination (Rutter et al., 2003). Only the SCQ total score has been validated for screen ing purposes. However, the measure contains subsets of items reflecting each of the three domains of the ADI - R and the three core symptom domains of autistic disorder under DSM - IV. Though scores can be derived for these domains, information about domains i s considered more important for purposes of - specific screening. The first domain reflects smiling, interes t in other children and offering comfort to others. The second is about communication issues and contains 13 items. This domain includes items asking if the child uses conventional gestures, reciprocates while in conversation and whether he or she has ster eotyped utterances, etc. The third domain assesses restricted, repetitive, and stereotyped patterns of behavior. It consists of eight items which ask about such things as stereotyped body, hand, or finger mannerisms; ritualistic behavior; circumscribed int erests and unusual pre - occupations by 27 the child; etc. Also, the SCQ has an item related to self - injurious behavior, as well as another on current language functioning (Berument et al., 1999; Rutter et al., 2003). According to Rutter and researchers (2003) , the primary standardization of the SCQ was with a total of 200 individuals who were participants of previous autism - related studies. The studies entailed participants with the various subdomains of autism under the DSM - IV together with those who had non - ASD disabilities often utilized for comparison purposes. The total number of cases utilized for analysis of the SCQ total score purposes was 214 (i.e., 177 males and 37 females) with ages ranging through two to 18 years. According to the parents, a total o f 157 children had the verbal ability while 57 did not. A total of 213 children had their prior diagnostic information available. The 157 children who had verbal or language ability were divided into three groups. The first group were those with a diagnosi s of autism ( n = 71). The second group were those with ASD ( n - NOS (CDD). The third group included those with non - spectrum dis orders ( n = 37) and with mostly language impairment, mental retardation, and ADHD. Data analysis was done on those with and without language. Those with language were further sub - divided into four distinct groups (representing the different age groups in the sample). The test authors noted that the children without language had the high est SCQ scores and that those with language had moderate SCQ scores that did not differ among much for the different age groups (2 - 4 years, 5 - 6 years, 7 - 10 years or 11 and older). Children with no language, those aged 2 - 4, and 5 - 6 years all had al pha indexes of internal consistency of 0.84 while those aged 7 - 10 or 11 and older had .89 and .93, respectively. The authors further categorized the psychometric properties of the SCQ using the diagnostic classification and found 28 an alpha index of intern al consistency of .81 for autism, .86 for ASD and .92 for non - spectrum diagnoses. The SCQ has been studied in other cultures. The SCQ has been used in studies with differing levels of specificity and sensitivity found, thus the adjustment of cut - off point s was done to match local populations. For instance, in a study with the Greek version of the SCQ among a sample of 130 children aged between seven to ten years, it was found that the SCQ had a specificity and sensitivity of .99 and .96 respectively (Zarok anellou, Kolaitis, Vlassopoulos, & Papanikolaou, 2017). The cut - off points for the ASD and non - ASD group was 15 when assessed using the rater operator characteristic (ROC). To distinguish between ASD and the high - functioning ASD group, the cut - off point wa s adjusted to 11. Another cultural study on the SCQ was conducted by Gau and others (2011). The researchers administered a Chinese version of the SCQ to 736 caregivers of children aged two to 18 years. The 736 study participants all had children clinical ly diagnosed with Autistic Disorder, - NOS according to the DSM - IV diagnostic criteria. All study caregivers filled out the Chinese version of the SCQ. Among the 736 participants, 317 were also interviewed using the ADI - R to serve as a measure to evaluate the concurrent validity of the Chinese version of the SCQ. An exploratory factor analysis revealed a three - factor structure, i.e., social interaction, repetitive behaviors, and communication. The study authors found test - retest re liability of .77 - .78 of intra - class correlations, and an internal alpha estimate of consistency ranging from .73 to .91 for the three subscales. Correlation between the Chinese SCQ and ADI - R subscales ranged between high to moderate. The SCQ social inter action subscale correlated highly (i.e., .63) with the reciprocal social interaction, and communication (.54 and .65 for verbal and non - verbal communication respectively) but moderately with the stereotyped behavior 29 subscale (i.e., .21) of the ADI - R subsca les. The SCQ repetitive behavior subscale had a correlation of .49 and .36 for stereotyped behaviors and reciprocal social interactions respectively as well as .39 and .31 for communication (verbal and non - verbal communication respectively) on the ADI - R su bscales. The SCQ social communication subscale correlated at .56 and .32 (verbal and non - verbal respectively), .37 (stereotyped behaviors) and .31 (reciprocal social interaction) with the ADI - R subscales. Gau and colleagues (2011) found a sensitivity and s pecificity of .71 and .77 respectively using the authors recommended cut - score of 15. However, when they utilized a cut - score of 12, the sensitivity increased to 0.86 while specificity fell to .60. The researchers concluded that the Chinese version of the SCQ was useful in assessing for the broader autism spectrum (now collectively called ASD under DSM - 5) but did not distinguish between DSM - In summary, most researchers contend that the SCQ i s useful in identifying those at risk of autism, but those interested in using it should consider adjusting its cut - off scores depending on the age and purpose as well as using it in combination with another autism measure (Bölte, Poustka, & Constantino, 2 008; Corsello et al., 2007; Gau et al., 2011; Zarokanellou et al., 2017). It has been validated in more than 11 studies (mostly with White culture) and found to have good clinical utility. However, the need for adjustments in the differential cut - off score s in different contexts as well as the non - optimal utility with children aged 2 - 3 years old (Norris & Lecavalier, 2010) . Limited studies exist in the application of the SCQ to other cultures and none to the culture targeted for this study. Social Responsiveness Sca le, Second Edition (SRS - 2) The SRS - 2 was designed to be a rapid measure of autistic symptoms. The instrument is useful in the identification of mild changes in autistic behaviors among individuals resulting 30 from interventions targeting the reduction of ce rtain behaviors unique to autism. The SRS - 2 has three different forms, i.e., Preschool, School Age and Adult Forms. The instrument assesses social or interpersonal behaviors, communication, and repetitive behavior characteristic of individuals with ASD (Co nstantino & Gruber, 2012). The instrument has a total of 65 Likert type items with a scale ranging from 1 (not true), 2 (sometimes true), 3 (often true), and 4 (almost always true). A person familiar with the individual with autism fills out the assessmen t according to the frequency of the behavior related to autism. Of the 65 items, 35 relate to the social impairment criteria as elaborated in the DSM - IV, i.e., the extent to which a child recognizes social cues, their capacity to interpret and appropriatel y respond to them as well as their general tendency to socially engage with others. Twenty items relate to the stereotyped or restricted range of interests and six to language deficit assessment. The remaining four items relate to different symptoms associ ated with autism, but also often observed in other neurodevelopmental or psychological disorders (Constantino & Gruber, 2012). The assessment takes 15 - 20 minutes to administer. Total raw scores (ranging from 0 to 195) are generated indicating the variou s levels of severity index for social deficits in a person suspected to have ASD. The total raw scores are converted into T - scores and utilized to make clinical or educational decisions depending on whether the individual is severely or moderately affected in their daily social interactions. There are four levels of severity in the SRS - 2. A T - score of 59 or below is considered within clinically normal limits. T - scores of 60 - 65 are indicative of mild to moderate limitation in activities of daily social interactions. In the DSM - IV - TR, children in this category were mild cases of ASD, PDD - T - scores of 66 - 75 are associated with 31 childre n with moderate deficits in their social interactions, i.e., have substantial interference with their daily social interactions. Children in this category were those having moderate severity of ASD. This falls into Autistic Disorder, PDD - sorder in the DSM - IV and the Social Communication Disorder and ASD in the current DSM - 5. Finally, those with T - scores of 76 or higher displayed deficiencies in reciprocal social behavior, which cause a severe disturbance in daily social interactions. Acco rding to the authors, the SRS - 2 is not as a standalone assessment in ASD diagnosis. The SRS - 2 T - for comprehensive ASD assessment. A standard error of measurement of 2 T - score points is advisable when making clinical or educational decisions to cater to assessment error in different contextual settings. The SRS - 2 was normed on preschool children (2. 6 - 4.6 years), school - aged children (4 - 18 years) and adults (19 years and above). Parents or daycare attendants rate the preschool forms while parents and teachers rate the school - age form. The School Age form standardization caters for normative data of children aged 4 - 18 years (to be utilized for this study). The study included a total of 2,025 reports. Parents/caregivers had 1,014 reports and teachers had 1,011 ratings. The sample consisted of 493 (48.7%) males and 518 (51.2%) females recruited from 2 0 sites in 13 states representative of all the four U.S Census Bureau regions (East, West, Midwest, and South). There were 23.2%, 16.2%, 15.9%, and 44.7% children for the West, East, Midwest, and South regions respectively. Their ethnic/racial background p ercentages were 5.7% (Asians), 15.8 % (Black/African American), 16.6% (Hispanic/Latino), 0.3% (Native American), 59.5% (White) 32 percentages were 13.7% (less than high school graduate), 26.1% (high school graduate), 24.8% (some college), and 35.4% for four years of college or more. The alpha internal consistency for the SRS - 2 School Age Form was .95 and .96 for parents and teacher reports, respectively. The inter - rate r reliability between parents and teachers was = .61. Constantino and Gruber (2012) recommended the cut - scores of 70 and 85 to identify ASD conditions in unselected general and clinical populations, respectively. According to them, for example, a cut - s core of 70 yielded a sensitivity of .78 and specificity of .94. They also specified that 93% of all children above the raw score of 70 were found to have a diagnosis of ASD on follow - up clinical assessment. Several studies were identified in the use of t he SRS in other cultural contexts. In one study conducted in Tehran (Iran), the Farsi version of the SRS - 2 was administered to parents and teachers of 533 children (191 boys and 342 girls) aged 7 - 11 years old (Tehrani - Doost et al., 2018) . All study children had had no record of previous developmental or psychiatric disorder. The SRS - 2 and SCQ were administered to parents and teachers. The authors reported a concurrent validity of .43 between the SRS - at .99 and .90 respectively supporting the DSM - 5 criteria for ASD. Tehrani and colleagues concluded that the Farsi version of SRS was a valid measure in helpi ng to identify children with difficulties in communication and social responsiveness within the community. Cen and colleagues (2017) investigated the psychometric properties of the Mandarin version of the SRS in Mainland China using a sample of 749 child ren aged 4 - 14 years. The sample included parents of 411 non - ASD children and 338 children who were clinically diagnosed with ASD. An internal consistency between .87 - .92 was found for the total scale 33 while its test - retest reliability was 0.81 - 0.94. Ac cording to the authors, the ROC data for the Mandarin version of the SRS reported that it accurately classifies 69.2 97.2% of youth with ASD. The study indicated that the parent - report SRS was a reliable and valid ASD screening instrument that can be use d in China, but separate cut - scores for screening and clinical purposes had to be established. They suggest that a cut - score of 56.5 with a sensitivity of .95 specificity of .90 is effective in discriminating between children with and without ASD (Cen et al., 2017) . In another study, a German version of the SRS was administered to 1) parents (mothers and fathers) of 838 non - ASD children, and 2) 527 with ASD and other psychiatric diagnoses, i.e., ADHD, conduct disorders, personality disorders, mental retardation, enuresis, and others. An internal consistency of .91 - .97, a test - retest reliability of .84 - .97 and interrater r eliability of .76 and .95 was found (B ölte, Poustka, & Constantino, 2008) . A ROC analysis of ASD versus other clinical diagnoses yielded an area under curve [AUC (Area Under the ROC Curve)] of .83 with a total score of 85 having a sensitivity of .73 and specificity of .81. The SRS is easy to administer and has been validated for use in many countries with good psychometric properties. However, cutoff points are not uniform in other populations hence the need for more multicultural validation studies in places like Africa where the SRS has not been validated. Childhood Autism Rating Scale, Second Edition (CARS - 2) The Childhood Autism Rating Scale (CARS) was initially called the Childhood autism (Schopler, Reichler, DeVellis, & Daly, 1980). The scoring of the instrument i s based on observations by experienced raters together with a combination of clinical examination, reports 34 by parents or other primary caregivers of a child being assessed and psychological tests (Chlebowski, Green, Barton, & Fein, 2010). The CARS or CARS - 2 is used for clinical or educational decisions making as well as assessing change in autism symptomology resulting from rehabilitation efforts. It assesses the presence and severity of autism symptoms and is scored on a Likert scale ranging from 1 (withi n abnormal) (Schopler et al., 1980). The CARS - 2 had good internal consistency in the ranges of .94 and inter - rater reliability of .71 (Schopler et al., 1980). The to tal composite scores possible range from 15 to 60. A person with a total score of less than 30 is not considered to have autism while those with total scores above 30 are considered to have autism. The original CARS was criticized for producing many false positives and over - diagnosing children of two years of age with autism (Chlebowski et al., 2010). Chlebowski and others (2010) investigated the use of CARS in a large sample of toddlers and preschoolers aged two and four years for evaluation of ASD. They found a Cronbach alpha of .91 and proposed increasing the cut - off point for two - year - old from 30 to 32 while retaining the cut - off point of 30 The instrumen t has been modified to include the standard version (CARS2 ST) as well as catering for those with high functioning autism (CARS2 - HF; Vaughan, 2011). The standard version (CARS2 ST) is administered to children under the age six or those above age six, bu t with significantly impaired communication skills or an IQ of 79 and below. It has a total of 15 items that assess for behaviors characteristic of individuals with autism. The completion of ratings for the CARS2 - ST can be done based on information from a single interview or direct observation. CARS2 - ST was normed on a total of 1,034 people (78% males 35 - 36 and were from all the four geographical regions of the U.S Census Bureau. The CARS2 - ST to tal score for internal consistency was .93. The High Functioning version (CARS2 - HF) was designed to assess individuals who are high functioning and suspected to be at - risk for ASD. It has a total of 15 items related to autism symptomology. To assign ratin gs on the CARS2 - HF, the rater must have information from more than one source. This can, for example, be information from direct observation and interviews. Also, important to note is that in the assignment of ratings, the assessed individual is compared t o non - ASD peers of the same age with values ranging from 1 - 4 for all the items. To increase sensitivity across sections, the authors provide for midpoint scores across each rating from 1 - 4 (Vaughan, 2011). It is, therefore, permissible to give midpoin t scores like 1.5, 2.5 or 3.5 if raters believe it depicts the best estimate of the behavior being assessed. The CARS2 - HF was normed on 994 people with 78% of them being male and 22% females. Their age ranged between 6 and 57 years and represented all the four regions of the U.S Census Bureau (West, Midwest, Northeast, and South). The internal consistency estimate for the CARS - 2 - HF was 0.96 while inter - rater reliability derived from 239 samples was .95. The CARS - 2 - HF total score and relationship to the pres ence of ASD (sensitivity) was .80 and a specificity of .87. Studies have been conducted to investigate the application of the CARS - 2 in other cultures. In one study conducted in Tanzania, 41 children aged 2 - 14 years were assessed for autism using the CA RS - 2 (Harrison, Zimak, Sheinkopf, Manji, & Morrow, 2014) . Thirty of the children were found to have autism while 11 received a diagnosis of global delay. Children suspected to have Fragile X - syndrome ( n = 5), cerebral palsy ( n = 1), downs syndrome ( n = 1), intellectual disability ( n = 5) and language delay ( n = 1) but could not be confirmed because of 36 the lack of appropriate assessment instruments constituted the group with global delay. A clinical psychologist synthesized information from the various as sessments (parent report, observation during play using the CARS - 2 and adaptive function) before providing a final diagnostic decision following the DSM - IV or DSM 5 criteria. A univariate ANOVA using CARS - 2 showed that children with ASD had significantly h igher scores then their peers with general delay. The average rating for children with ASD on the CARS - 2 was 37.75 ( SD = 5.75) while that for those with general delay was 27.15 ( SD = 6.79). Although not a validation study, Harrison, colleagues (2017) showe d that it is possible to validate autism screening and diagnostic instruments in Africa. In a validation study in Lebanon, the CARS2 - ST Lebanese version was found to have an internal consistency of .99, interrater reliability of .99 and test - retest reliability (conducted after a month from baseline assessment) of .89. This was based on a total study sample of 109 children ( n = 90 with ASD and n = 19 with intellectual disability) aged 4 - 9 years (Akoury - Dirani, Alameddine, & Salamoun, 2013) . The researchers in this study excluded item 14 (Level and Consistenc y of Intellectual Response) because it loaded on a factor of its own and never correlated with other items in the Lebanese version. This Lebanese version of the CARS2 - ST therefore consists of 14 items and not 15 as in the original assessment. Gilliam Autis m Rating Scale (GARS - 3) Gilliam developed the Gilliam Autism Rating Scale (GARS - 3) to assess children and young adults aged three to 22 years for autism (Karren, 2017). The measure is administered to parents, teachers or other people who spend considerabl e time with the person being assessed. The GARS - 3 was developed following the DSM - IV as well as criteria set out by the Autism Society of America (Lecavalier, 2005). The GARS was criticized for limited sensitivity and 37 inter - rater reliability when compared to what was reported by Gilliam (Lecavalier, 2005; South et al., 2002). The GARS - 3 was normed based on ratings from parents, teachers, educational diagnosticians, psychologists, speech clinicians, and teacher assistants. A total of 1,859 children and youn g adults diagnosed with autism, aged three to 22 years, and living in the U.S. formed the study sample for the GARS - 3 (Karren, 2017). The instrument can be administered as a questionnaire or as a structured interview. It is composed of up six subscales wit h a total of 58 items following the DSM 5 definition of ASD (APA, 2013). The subscales include restricted, repetitive behaviors (13 items), emotional responses (8 items), maladaptive speech (7 items), social interaction (14 items), social communication (9 items), and cognitive style (7 items). Rating of autism behaviors is on a Likert type scale ranging from 0 (not at all like the individual), 1 (not much like the individual), 2 (somewhat like the individual) to 3 (very much like the individual; Karren, 201 7). All items of the measure contribute to the final composite score for the scales called the Autism Index. The Autism Index is utilized to determine the presence or absence of autism. The test developers recommend that children who have selective mutism not be assessed using the maladaptive speech and cognitive style subscales, but with the other four remaining sub - scales (Karren, 2017). The scores in the four or six subscales are totaled (depending on whether a child has been rated on four scales because the y have selective mutism or all the six subscales for those without selective mutism), then converted into percentiles indicating those equal to or below the person being rated in the normative sample. The percentiles are then scaled to give the Autism Inde x with a mean of 100 and a standard deviation of 15. Higher Autism Index scores are 38 indicative of severe autism symptoms. The overall internal consistency across all the age groups was .94 and .93 for the four and six subscales respectively (Gilliam, 2014) . Studies have been conducted to investigate the use of the GARS in other cultures. Caregivers of children aged 3 - 16 years in South Texas were utilized for the validation of the Spanish version of the GARS - 2 (Jacks on, Little, & Akin - Little, 2013) . Parents of a total of 100 children composed of 77 with ASD and 23 without ASD (ADHD n = 9, mental retardation n = 2, other diagnosis n = 3 or no diagnosis n = 9). Out of the entire study sample, 92% of participants repo rted their ethnicity to be Mexican or Mexican American with the remaining eight percent of the GARS - 2, Spanish Version. Correlations among the three scales rang ed from .81 to .83. It was also found that participants with ASD had significantly higher scores than their peers in the other diagnostic groups. Test - = .98 although this entailed analysis o f data from only 10 of the 100 participants who responded to the call to participate in the re - test study. 23 Question Questionnaire The 23 Question Questionnaire (23Q) was developed in the context of the African culture (Kakooza - Mwesige et al., 2014). It is a 23 - item scale that is used to assess neurodevelopmental disorders for children aged 2 - 9 years. The 23Q includes all ten items of the original Ten Questions (TQ) - an instrument utilized to assess 22,125 children aged 2 - 9 years for childhood dis ability in Jamaica, Pakistan, and Bangladesh (Durkin, 1995). Kakooza - Mwesige and colleagues (2014) in collaboration with a Technical Advisory Group (TAG), which included Ugandan and American clinicians as well as epidemiologists, modified the original TQ developed by Durkin (1995) into 23 (yes/no) response items. They 39 utilized it to assess 1,169 children in Uganda aged 2 - 9 years for neurodevelopmental disabilities. They specifically included an additional five items for all children aged 2 - 9 years, fiv e items for children less than five years of age and an additional three items assessing for visual, hearing or seizure impairments. They translated and back - translated the entire tool into Luganda - a local language used by most people in the study area. If a child failed one or more items on the instrument, he or she was considered positive for potential neurodevelopmental disorders and referred for further clinical evaluation by specialists. Out of a total of 320 children who screened positive on the 23Q , only eight children had ASD after further clinical evaluation by experts. Kakooza - Mwesige and colleagues reported a sensitivity ranging from .55 to .80 for the 23Q. Justification for Using the SRS and SCQ There is not one good screening or diagnostic tool that has been psychometrically validated for use in Uganda and Africa at large. The few validation studies have demonstrated promising results in identifying children with autism (Harrison et al., 2014; Kakooza - Mwesige et al., 20 14; Smith et al., 2017) . The ADOS and ADI that are considered the gold standard in ASD diagnosis take a lot of time to administer and require extensive training for one to utilize. Professionals trained to this level are few and stretched by the high nu mber of patients in Uganda. The SRS and SCQ were chosen for use in this study because they are administered directly to the parents or other primary caregivers who spend considerable time with the child. Both do not require trained raters and yet have stro ng reliability and validity with the DSM - IV and ICD - 10 algorithms for autism criteria. Kakooza - Mwesige and colleagues (2014) were able to show that screening positive on both the additional ASD items in the 23Q and original TQ items increased the sensitivi ty for receiving a diagnosis of ASD from .49 (with TQ items only) to .69 40 (with both TQ and ASD items). Further evaluation of the five autism specific items on the 23Q is needed given that it improved the sensitivity of the TQ. Summary Assessing for autism takes extensive training and experience working with an autism population to be competent to make a diagnosis (Volker & Lopata, 2008). Autism assessment requires collaborative effort among professionals (Witwer & Lecavalier, 2007). Even with such training and experience, one is still expected to consider information from other professionals before making a diagnosis. There are few practitioners trained to such levels in Uganda. Children with ASD must wait long before a formal diagnosis, or at the worst, go undiagnosed. The situation is worse for children in rural settings because most professionals who are knowledgeable in ASD live and work in urban areas (Bakare & Munir, 2011; Gona et al., 2015). Such extensive assessments like the ADOS - 2 and ADI - R conside red the gold standard in ASD are a preserve for Ugandan children who can make it to the health facilities in urban settings. Even then their utility in settings where professionals meant to make diagnostic decisions are overwhelmed by the large patient to professional ratio is little. Having instruments that do not require advanced level training can enable the few well - trained professionals to focus their efforts on individuals with a higher likelihood of having ASD. Furthermore, screening tools such as th e SCQ, SRS or the ten ASD specific questions on the 23Q can be utilized by less trained professionals often running rural health centers as well. This reduces the long waits parents, especially in low - and medium - income countries, who are often subjected t o the waitlist of as long as six months or more (Bello - Mojeed et al., 2017). 41 CHAPTER 3 : METHOD OLOGY Purpose of the Study The purpose of this study was to investigate the psychometric properties (i.e., internal consistency reliability, concurrent validity, and clinical discriminant validity) of the Social Responsiveness Scale, Second Edition (SRS - 2), Social Communication Questionnaire (SCQ), and select 23Q items when used to assess for ASD among children ages four to18 years old in Ugand a. This chapter lays out the six research questions, their associated hypotheses, study participants, procedures, and instruments. Prior to field testing, a systematic procedure describes how items of the SRS - 2 and SCQ were reviewed and modified based on a small group of of research assistants, data management, and data analyses are presented. Research Questions and Hypotheses Research Q uestion 1 In the Uga ndan context, do total scores for the three ASD instruments (i.e., SCQ, SRS - 2, and 23Q) yield adequate internal consistency relative to values reported in the test manual or available studies for those with ASD? Hypothesis 1a . Within an ASD sample in the Ugandan context (as determined by the DSM - the SCQ was anticipated to be > .80. Hypothesis 1b . Within an ASD sample in the Ugandan context (as determined by the DSM - the SRS - 2 was anticipated to be > .90. 42 Hypothesis 1c . Since no existing data were available for the internal consistency of a composite made up of the ASD - related items from the 23Q, no specific reliability value or range of values was/were predicted for this hypothesis. However, it was anticipated that the internal consistency for a 23Q composite for cases with ASD would be low er than that of the SRS - 2 and SCQ due to: a) small number of ASD - related items from the 23Q for ASD screening and b) ASD screening had typically been done at the item level for the 23Q. Research Q uestion 2 In the Ugandan context, do total scores for the th ree ASD instruments (i.e., SCQ, SRS - 2, and 23Q) yield adequate internal consistency relative to values reported in the test manual or available studies for non - ASD cases? Hypothesis 2a . For a non - ASD sample in the Ugandan context, the internal consistency for the total score of the SCQ was hypothesized to be > Hypothesis 2b . Within a non - ASD sample in the Ugandan context, the internal consistency for the tot al score of the SRS - 2 was predicted to be > Hypothesis 2c . Since no existing data was available for the internal consistency of a composite made up of the ASD - related items from the 23Q, no specific reliability estimate was ma de for this hypothesis. It was, however, anticipated that the internal consistency for the 23Q composite for non - ASD cases would be lower than that of the SRS - 2 and SCQ due to: a) small number of ASD - related items from the 23Q for ASD screening, and b) ASD screening has typically been done at the item level for the 23Q. Research Q uestion 3 In the Ugandan context, do total scores from the three instruments (i.e., SCQ, SRS - 2, and 43 23Q) yield substantial mean differences consistent with construct - related (i.e., ASD construct) differentiation between ASD and non - ASD groups? Hypothesis 3a . For the SCQ total score in the Ugandan context, it was anticipated that the mean of the ASD grou p would be significantly and substantially higher than the mean of the non - ASD group. Hypothesis 3b . For the SRS - 2 total score in the Uganda context, it was anticipated that the mean of the ASD group would be significantly and substantially higher than the mean of the non - ASD group. Hypothesis 3c . For the 23Q total score of the ASD - related items, it was anticipated that the mean of the ASD group would be significantly and substantially higher than the mean of the non - ASD group. Research Q uestion 4 In the Ugandan context, will use of the recommended screening cut - scores for each of the three instruments (i.e., SCQ, SRS - 2, and 23Q) result in adequate sensitivity and specificity using DSM - 5 ASD diagnosis vs. non - ASD cases as the outcome variable? Hypo thesis 4a . Within the Ugandan context, a classification analysis comparing ASD vs. non - ASD based on the SCQ cut - score reported in the manual and the DSM - 5 ASD vs. non - ASD diagnostic outcome would result in a significant Chi - square value, as well as sensiti vity of > .80 and specificity of > .80. Hypothesis 4b . Within the Ugandan context, a classification analysis comparing ASD vs. non - ASD based on the SRS - 2 cut - score reported in the manual and the DSM - 5 ASD vs. non - ASD diagnostic outcome would result in a si gnificant Chi - square value, as well as sensitivity of > .80 and specificity of > .80. 44 Hypothesis 4c . Using the screening criterion for the ASD items on the 23Q (i.e., one or more ASD - related items affirmatively endorsed) compared to the DSM - 5 ASD vs. non - ASD diagnostic criterion, no specific screening sensitivity or specificity target values were hypothes ized because of a lack of clear prior findings in the context of a full classification analyses. However, lower sensitivity and specificity for the 23Q ASD - related items was hypothesized relative to the SCQ and the SRS - 2, as these other instruments consist of many more items and likely better represent the ASD construct. Research Q uestion 5 Beyond the recommended cut - scores for screening reported in each test manual, are there more optimal cut - scores for each of the three instruments (i.e., SCQ, SR S - 2, and 23Q) in the Ugandan context? Research Q uestion 6 Are there significant convergent relationships among the total scores of the three instruments (i.e., SCQ, SRS - 2, and 23Q) in the Ugandan context? Hypothesis 6a . Within the Ugandan context, the correlation across the instruments on the total scores of SRS - 2 and SCQ was hypothesized to be significant, positive, and substantive (i.e., > .50). This was based on the reported correlations between the SRS - 2 total score and SCQ total score ranging from .68 in non - ASD group to .50 for the ASD - specific group in the SRS - 2 manual. Hypothesis 6b . Within the Ugandan context, the correlation between the SRS - 2 total score and 23Q summed ASD - related items score was hypothesized to be significant and positive but lower than the correlation between the SRS - 2 and the SCQ tota l scores. 45 Hypothesis 6c . Within the Ugandan context, the correlation between the SCQ total score and the 23Q summed ASD - related items score would be significant and positive but lower than the correlation between the SRS - 2 and the SCQ total scores. Participants The researcher collected data from a total of 107 participants. All participants were caregivers, who were defined as a parent or person who spends considerable time with the child providing primary care. Study participants were recruited duri ng routine clinical follow - up visits, from regular and special needs schools as well as autism assessment or rehabilitation centers within Kampala and the surrounding districts. Inclusion criteria of caregivers were as follows: 1) live within Kampala and t he surrounding districts; 2) adult caregivers of 20 years and older; 3) proficiency in speaking and comprehension of the English language; 4) spent at least 20 hours per week having regular physical interaction with the child; 5) no cognitive disabilities that would impede their judgment. The study was composed of two groups. Group 1 included 51 caregivers of children aged between 4 and 18 years with a confirmed diagnosis of ASD. With the help of the school, clinic and hospital administrators, caregivers w hose children had a diagnosis of autism were invited to be part of the study. Upon agreement to participate, caregivers were requested to provide formal medical or psychological documentation confirming an autism diagnosis for their child. The research ass istant reviewed and consulted with the primary researcher before a decision for inclusion in the study based on the availability of clinical documentation. Only parents with clear autism diagnostic documentation for their child were included in the study. Caregivers of children meeting the inclusion criteria were consented to participate in the study. The children were assented (depending on whether they were above or below seven years of age as per the 46 Ugandan IRB protocol) before the beginning of the stud y. Those without proper documentation were referred for further assessment with pediatric psychiatrists or clinical psychologists in Butabika and Mulago national referral hospitals. Group 2 involved 56 caregivers of typically developing children (without A SD, not identified with or suspected of a developmental disability) within the same age range (i.e., 4 to 18 years). A similar caregiver recruitment procedure as the first group was followed, with the difference being that caregivers were asked about the d evelopmental history of their child to rule out the possibility of including undiagnosed autism or a closely related disorder. Caregivers of children meeting the inclusion criteria were consented to participate in the study. The children were assented (dep ending on whether they were above or below seven years of age as per the Ugandan IRB protocol) before the beginning of the study. Those identified, as being at risk for undiagnosed autism were excluded and referred for further clinical workups with pediatr ic psychiatrists or clinical psychologists in Butabika and Mulago national referral hospitals. Procedure Review and M odification of the SRS - 2 and SCQ Both the SRS - 2 and SCQ were developed based on the U.S. culture. To ensure that the items were applicabl e to the Ugandan culture in terms of linguistic use, context, and content, a request was made of the test patent holders (i.e., Western Psychological Services [WPS]) to purchase the two instruments for pilot use in Uganda. When approved, both instruments w ere reviewed by five Ugandan parents of varying education levels (i.e., high school, college, and graduate) and three Ugandan trained psychologists (i.e., one community and two clinical psychologists). They provided feedback on the items of the SRS - 2 and S CQ after reviewing the items by communicating via e - mail with the researchers in the U.S. As suggested by the parents 47 and psychologists, item review and adjustment were conducted with the primary researcher, and two PhD level researchers familiar with both the clinical and research aspects of the SRS - 2 and SCQ. One researcher also had extensive research experience in psychological testing and measurement. All unclear items were modified for better comprehension and cultural fit. A total of seven SRS - 2 items (i.e., items 5, 25, 29, 30, 39, 50 and 64) were adjusted. For ben e parents in the pilot phase Six SCQ items (i.e., items 2, 3, 5, 6, 11, and 34) were iterated to fit the Ugandan research her/him that involves taking decision was made basing on the fact that Most reviewers and parents in the pilot study were concerned over the meaning of the word fied version of both the SRS - 2 and SCQ were then submitted to WPS for final approval to be administered to the parent participants. 48 Both the SRS - 2 and SCQ are filled out by caregivers most familiar with the child or person being evaluated. However, for this study, the administration of both instruments was changed. Research assistants administered the items directly to the study caregi vers by reading them out and filling in the caregiver responses. The 23Q administration did not change given that it was set up to be administered to the person most familiar with the child by a trained professional as opposed to having the parent fill it out directly. Training of the R esearch A ssistants for I nstrument A dministration Two research assistants working in Butabika and Mulago National Referral Hospitals and known to the lead researcher through prior research engagement were emailed about the part - time research job opportunity to which they agreed to participate. Study details were shared with them through a meeting organized by the lead researcher while in Uganda. Both research Makerere University in Uganda and administered psychological tests as part of their routine work. The lead researcher provided two half - day trainings to the research assistants on the administration of study instruments, consent and conducting of outreach . Four hours of training were dedicated to proper administration of the study instruments and methods of successfully initiating, conducting, and closing interviews. Fidelity C heck On the third day of training, each research assistant administered study instruments to a parent of a non - study child with the primary researcher observing the entire procedure (i.e., how they initiated, conducted, and closed interviews) and assessing them for fidelity. A fidelity checklist was created with input from the prima ry researcher as well as the major academic advisor to the lead researcher (Appendix G). 49 Field - testing When granted, the Institutional Review Board (IRB) clearances for each of the Ugandan sites and study host institution (Michigan State University), the researcher reached out to the in person for potential study collaboration. Study details together with a request for permission to utilize the school or hospital , as a study site were communicated with the management. When permitted, the researcher provided the administration with letters inviting parents to be part of the study. Principals or managers helped identify potential caregivers of children aged four to 18 Mental Health hospital. Only information of children whose parents were interested in participating in the study was abstracted from available hospital a nd school records. The school or hospital staff then made telephone calls to interested parents reminding them of the invitation to be part of the study. Once identified, potential study participants were informed about the ongoing study and asked if they were interested in participating. If they were, a parent together with his/her child (if above seven years) were requested to meet with the researcher for further details about the study. The child was requested to meet with the researcher as per the info rmed consent guidelines by the Makerere University School of Health Sciences IRB. The research assistants explained the study in its entirety to the potential study participants using simple English and provided an opportunity for them to ask questions bef ore questioning if they were willing to be part of the study. If they agreed to participate in the study, they were consented and/or assented, screened and included in the study. 50 Data collection took 11 months. The primary researcher together with the tw o research assistants administered the study instruments to caregivers during the first month of data adherence to acceptable interview guidelines for the first month. Both research assistants reached excellent fidelity levels (i.e., 95%) in the administration of instruments and principles of good ethical practice when interacting with research participants. The two research assistants continued with the data col lection as the primary researcher returned to the U.S. to continue with his studies. All study caregivers were given an anonymous identification number generated for them and a packet of instruments. A trained research assistant administered all the assess ments and demographic information to the caregiver on a one - on - one basis in a secured room within the school, hospital, or other private premises. Instrument administration for each parent took between 35 - 40 minutes. For all completed protocols, the resear ch assistants double - checked each measure for completeness and accuracy in the presence of the caregiver. If errors were identified, caregivers were requested to clarify. Measures Social Communication Questionnaire (SCQ) - Lifetime Form The SCQ Lifetime Form assesses traits of ASD throughout the lifetime of an individual suspected of having ASD (Lord et al., 1994; Rutter, Bailey, & Lord, 2003). It is a 40 - item instrument, with each item scored dichotomously (i.e., Yes/No), and administere d to the parent or another caregiver most familiar with the child suspected of having ASD. Self - administration of the screening measure takes approximately ten minutes, and the scoring can be completed in five minutes. Thirty - nine of the 40 items are summe d to calculate the total score, which can 51 range from 0 to 39. Please refer to chapter two for detailed information about the psychometrics of the SCQ. Social Responsiveness Scale, Second Edition (SRS - 2) - School - Age Form There are three SRS - 2 forms, the School - Age (4 18 years), Preschool (2.5 and 4.5 years), and the Adult (Self/Relative/Other Report) forms. The SRS - 2 School - Age form is a parent or caregiver completed questionnaire, which is used to assess for presence and severity of social impairment o f autism spectrum disorder (Constantino & Gruber, 2012). The original SRS (Constantino & Gruber, 2005) and current SRS - 2 School - Age form consist of 65 items rated on a Likert scale of 1 to 4 (1 = not true, 2 = sometimes true, 3 = often true, 4 = almost alw ays true). Please refer to chapter two for detailed information about the psychometrics of the SRS - 2. Self - administration of the SRS - 2 takes between 15 to 20 minutes. For the current study, a research assistant as opposed to them filling it out by themselv es administered the SRS - 2 to the parents. Its administration took between 25 to 45 minutes. Please refer to chapter two for detailed information about the psychometrics of the SRS - 2. The 23 Question Screener The 23Q has 23 items used to assess Ugandan chil dren aged 2 - 9 years of age for different neurodevelopmental disabilities (Kakooza - Mwesige et al., 2014). The 23 items have a neurodevelopmental disability if they fail one or more items. However, it is not used for diagnostic purposes, but is a screener. Children identified as positive are referred for further clinical evaluation. Not much was reported about the psychometrics of the instrument. Kakooza - Mwesi ge and colleagues reported a sensitivity ranging from 0.55 to 0.80 for the 23Q. However, 52 for the current study, only ten of the 23Q items specific to ASD screening were used. Please refer to chapter two for detailed information about the psychometrics of t he 23Q. Data Checking and Entry By the end of the data collection, the research assistants transported all completed instruments to a secure office in Mulago hospital. All hard copies of the instruments were scanned and uploaded to OneDrive (a secure password - protected online platform) f or review by the lead researcher. The hard copies were then locked in a secure cabinet at the Global Health Uganda offices in Mulago National Referral hospital. The research assistants then created password protected database files and sent by e - mail to t he lead researcher at Michigan State University and stored through OneDrive, a secured cloud - based tool designed for secure data storage and sharing among approved research collaborators across sites. The lead researcher then entered all the data into the SPSS database system in the study office at Michigan State University, with double data entry and error checklist verification for quality control. In Uganda, the research assistants ensured that all hard copies of study data were securely transported to a nd stored in locked file cabinets within the Child Health Uganda (CHU) offices in Tororo Main hospital. Data Analysis This study provides critical data concerning the reliability, validity, and screening utility of the SCQ, SRS - 2, and 23Q ASD - specific it ems in the Ugandan cultural context. Descriptive statistics for all measure scores were reported, as well as correlations among scores within and between instruments. Internal consistency estimates (i.e., coefficient alpha) for the SCQ and SRS - 2 composite scores are reported. Item analyses (e.g., item - total correlations, alpha if item deleted, item discrimination, etc.) were conducted for the SCQ and SRS - 2, while only item level 53 analyses were conducted for the 23Q ASD items, as these items were not typicall y summed into behavior in a child consistent with ASD and was coded as 1, while the other three responses (i.e., 0 indicating absence of behavior. Discriminant validity of the SCQ and SRS - 2 composite scores was examined through: 1) the point - biserial correlation between test score and the DSM - 5 dichotomous diagnostic outcome (i.e., ASD vs. non - ASD); 2) predicted me an differences between children with a known ASD diagnosis and those with no known - ASD diagnosis (Non - ASD children); and 3) Receiver Operating Curve (ROC) analyses to determine optimal screening cut - scores for the SCQ and SRS - 2 when attempting to different iate between known ASD and non - ASD children. Screening sensitivity and specificity estimates were reported for various cut - scores for the SCQ and SRS - 2, as well as for the number of item endorsements on the 23Q. Optimum screening cut - scores maximized true positive and true negative screening classifications, while minimizing false negative and false positive screening classifications. False negatives were deemed the most problematic screening error, as true positives and false positives go on for further di agnostic evaluation for more accurate differentiation in the real world while true negatives and false negatives both end up getting screened out. All data analyses were performed using SPSS version 25. 54 CHAPTER 4 : RESULTS This study assessed the psychometric properties of the SRS - 2, SCQ and 23Q among children aged 4 - 18 years in Uganda. The present chapter reports on the demographics and findings for all six study research questions and associated hypotheses. Research questions 1 and 2 relate to f - ASD groups respectively. Research questions 3 and 4 show findings for validity and classification analyses for all the study instruments respectively. Research questions 5 and 6 yield results fr om the exploration of alternative cut - scores and convergent validity of the total scores for the study instruments. Demographic Characteristics of Parents ASD and non - A SD groups. When considering both groups together, a total of 102 parents (out of 107) provided demographic information. Five parents had incomplete demographic forms involving several unanswered items. All parents in both groups were of African ethnicity ( 100%, n = 107). For age, 28 (52.8%) and 27 (55.1%) of non - ASD and ASD parents were between 31 and 40 years of age. For geographic residence, 49 (87.5%) from the non - ASD group and 30 (58.8%) from the ASD group lived in Kampala district. For marital status, 62.5% ( n = 35) and 54.9% ( n = 28) of the non - ASD and ASD groups were married. Most mothers, (i.e., 89.3% ( n = 50) of the non - ASD group and 74.0% ( n = 37) of the ASD group) were employed. For fathers, 96.3% ( n = 52) of non - ASD and 90.7% ( n = 39) of ASD gro ups were employed. Most parents in the non - ASD (73.9%, n = 34) and ASD (78.6%, n = 22) groups earned a monthly income between 0 (no income) and 3,000,000 Ugandan Shillings (U.S $830). Parents from the non - ASD group mostly had a secondary or polytechnic edu cation (33.9%, n 55 (25.0%, n = 14). In contrast, most parents of children with ASD had completed ordinary level or 11 years of school (34.0%, n = 17) followed by a bachelors, that is, 16 years of school (24.0%, n = 12). Table 1 . Demographic Characteristics of Parents Non - ASD ( n = 56) n (%) ASD Group ( n = 51) n (%) Total Sample ( N = 107) n (%) Parent Age (Years) < 20 -- 1 (2.0%) 1 (1.0%) 21 - 30 8 (15.1%) 11 (22.4%) 19 (18.6%) 31 - 40 28 (52.8%) 27 (55.1%) 55 (53.9%) 41 - 50 15 (28.3%) 8 (16.3%) 23 (22.5%) 51 - 60 2 (3.8%) 2 (4.1%) 4 (3.9%) District Kampala 49 (87.5%) 30 (58.8%) 79 (73.1%) Luwero -- 1 (2.0%) 1 (0.9%) Mbarara -- 1 (2.0%) 1 (0.9%) Tororo -- 1 (2.0%) 1 (0.9%) Wakiso 7 (12.5%) 18 (35.3%) 25 (23.1%) Ethnicity African 56 (100.0%) 51 (100.0%) 107 (100.0%) Marital Status Single 8 (14.3%) 8 (15.7%) 16 (15.0%) Married 35 (62.5%) 28 (54.9%) 63 (58.9%) Living with domestic partner 11 (19.6%) 9 (17.6%) 20 (18.7%) Divorced -- 3 (5.9%) 3 (2.8%) Widowed 2 (3.6%) 3 (5.9%) 5 (4.7%) Mother Employed No 6 (10.7%) 13 (26.0%) 19 (17.9%) Yes 50 (89.3%) 37 (74.0%) 87 (82.1%) Mothers' Occupation Private Business 10 (17.9%) 22 (43.1%) 32 (29.9%) Housewife 5 (8.9%) 5 (9.8%) 10 (9.3%) Professional 37 (66.1%) 8 (15.7%) 45 (42.1%) Domestic -- 3 (5.9%) 3 (2.8%) None 1 (1.8%) 10 (19.6%) 11 (10.3%) Others 3 (5.4%) 3 (5.9%) 6 (5.6%) Father Employed No 2 (3.7%) 4 (9.3%) 6 (6.2%) Yes 52 (96.3%) 39 (90.7%) 91 (93.8%) 56 Table 1 ( ) Father's Occupation None 5 (8.9%) 12 (11.2%) 17 (15.9%) Professional 25 (44.6%) 14 (13.1%) 39 (36.4%) Private Business 15 (26.8%) 17 (15.9%) 32 (29.9%) Others 11 (19.6%) 8 (7.5%) 19 (17.8%) Income 0 - 3,000,000 34 (73.9%) 22 (78.6%) 56 (75.5%) 3,001,000 - 6,000,000 10 (21.7%) 4 (14.3%%) 14 (18.9%) 6,001,000 - 9,000,000 1 (2.2%) -- 1 (1.4%) 9,001,000 - More 1 (2.2%) 2 (7.1%) 3 (4.1%) Parent Education ( Years) Primary Completed (7 Years) 1 (1.8%) 2 (4.0%) 3 (2.8%) Some O' Level (9 Years) 1 (1.8%) 7 (14.0%) 8 (7.5%) O' Levels Completed (11 Years) 10 (17.9%) 17 (34.0%) 27 (25.5%) A' Level Completed (13 Years) 3 (5.4%) 6 (12.0%) 9 (8.5%) Some Post - Secondary or Polytechnic (15 years) 19 (33.9%) 3 (6.0%) 22 (20.8%) Bachelors (16 Years) 14 (25.0%) 12 (24.0%) 26 (24.5%) Masters or Post - Graduate (18 Years) 8 (14.3%) 3 (6.0%) 11 (10.4%) Note. Calculations are based on different sample sizes due to missing data for some demographic variables: a total of 107 parents. b Parental age was based on a to tal of 102 parents, Mother employed ( N = 106), father employed ( N = 97), estimated income ( N = 74), and education level ( N = 106). 57 Demographic Characteristics of Children of the Participating Parents Parents provided the demographic information for all 107 study children ( n = 51 with ASD, n = 56 non - ASD). A detailed breakdown of their demographic characteristics as reported by parents, are summarized in Table 2. All study children were of African ethnicity with their ages ranging betw een 4 17 years. The children in the non - ASD group were slightly older than those in the ASD group. The age range for the children was 4 - 15 years ( = 7.22, D = 2.46) for the ASD group and 4 - 17 years ( = 9.34, D = 3.52) for the non - ASD group. The average age for ASD first diagnosis was 4.1 years ( D = 1.8) with 80.4% ( n = 41) of the ASD group having a comorbidity while 19.6% ( n =10) had no known comorbidity. Males diagnosed with ASD were 66.7% ( n = 34) with 33.3% ( n = 17) of the ASD group being f emale. Children in the non - ASD group had no known ASD diagnosis and were typically developing. Therefore, demographic information related to age at first diagnosis, comorbidity of ASD or professional who first diagnosed the child was not applicable to the non - ASD group. Available data showed that medical doctors diagnosed 38.8 % ( n = 19), psychiatric clinical officers 24.5% ( n = 12), clinical psychologists 18.4% ( n = 9), psychiatrists 12.2% ( n = 6), other and occupational or speech therapists 6.1% ( n = 3) o f the children with ASD (Table 2). 58 Table 2 . Demographic Characteristics of Children Demographic Characteristics ASD Non - ASD Total ( n = 51) ( n = 56) ( N = 107) Child Age in Years = 7.22 ( D = 2.46) = 9.34 ( SD = 3.52) = 8.34 ( SD = 3.24) Range (4 - 15) Range (4 - 17) Range (4 - 17) Gender Males 59.6% ( n = 34) Males 40.4% ( n = 23) 100.0% ( n = 57) Females 34.0% ( n = 17) Female 66.0% ( n = 33) 100.0% ( n = 50) African 100.0 % ( n = 51) 100.0% ( n = 56) 100.0% ( N = 107) Age at First Diagnosis = 4.1 ( SD = 1.8) N/A N/A Comorbidity of ASD No 19.6% ( n = 10) N/A N/A Yes 80.4% ( n = 41) Professional who first diagnosed child Medical Doctor 38.8% ( n =19) N/A N/A Clinical Psychologist 18.4% ( n = 9) N/A N/A Psychiatrist 12.2% ( n = 6) Psychiatric Clinical Officer 24.5% ( n = 12) N/A N/A Other & Occupational or Speech Therapist 6.1% ( n = 3) N/A N/A Note. Age was missing for 2 cases in the ASD group. Research Questions and Hypotheses This section reports the study findings in relation to each of the stated research questions and associated hypotheses. Research questions 1 and 2 related to findings about the reliability of the study instruments. All instruments were assessed for overall reliability among all study participants (i.e., ASD and non - ASD combined) before analyses were conducted within individual groups (i.e., ASD and non - ASD separately). Research question 3 pertained to the 59 validity of the three study instruments for differen tiating between groups known to differ on the intended construct (i.e., ASD and non - ASD). For this hypothesis, differentiation was assessed via mean differences between groups. Continuing with the assessment instrument validity for differentiating between groups, research question 4 examined the screening utility of prescribed cut - core article[s]). Screening utility was examined in terms of the sensitivity and specifici ty of each instrument with its recommended cut - score. Research question 5 explored the screening utility of alternative cut - scores for the three instruments in the Ugandan context. While there were no specific directional hypotheses, tentative hypotheses w ere set based on the range of screening cut - scores using ROC curves. Research question 6 involved examining the convergent validity of the total or composite scores for each instrument via their inter - correlations with each other (i.e., correlations betwee n the instruments). The results of each research question and hypothesis are detailed below. Research Q uestion 1 In the Ugandan context, do total scores for the three ASD instruments (i.e., SCQ, SRS - 2, and 23Q) yield adequate internal consistency relative to values reported in the test manual or available studies for those with ASD? Hypothesis 1a . Within an ASD sample in the Ugandan context (as determined by the DSM - 5 of the SCQ was anticipated to be > .80. The Cronbach alpha value obtained from the total scores of the SCQ among the ASD sample was .61. A closer inspection of the reliability statistics showed that if item 4 (inappropriate question) was deleted, the Cronbach alpha would increase to .64, which is not a 60 significant change in the reliability. Given that the reliability is below the hypothesized value of .80, hypothesi s 1a was not supported. Hypothesis 1b . Within an ASD sample in the Ugandan context (as determined by the DSM - of the SRS - 2 was anticipated to be > .90. The Cronbach alpha value obtained from the total score of the SRS - 2 in the ASD group ( n = 51) was .84. Item analyses revealed that any other item deletion would not substantively improve the reliability of the SRS - 2 in this sample. Despite an improvement in in ternal consistency, the estimate was still below the hypothesized value of .90. Therefore, h ypothesis 1b was not supported. Hypothesis 1c . Since no existing data were available for the internal consistency of a composite made up of the ASD - related items from the 23Q, there were no guiding estimates for this hypothesis. However, it was anticipated that the internal consistency for a 23Q composite for cases with ASD would be lower than that of the SRS - 2 and SCQ due to a) small number of AS D - related items from the 23Q for ASD screening and b) ASD screening had typically been done at the item level for the 23Q. The 23Q has ten items intended to screen for ASD (i.e., items 14 to 23 in the order they appear on the instrument form). Items 14 to 18 are used to screen for ASD among children of all ages, while 19 to 23 are additional items administered to children below five years of age for the same purpose (i.e., assess for ASD). Therefore, in typical practice, parents whose children are below fiv e years of age respond to a total of 10 items while those with children five years and above respond to only five items. In the present study, all ten ASD items were administered to all study participants, regardless of age . An analysis of internal consist ency when combined into a 61 composite (i.e., total score consisting of the sum of all ten ASD items of the 23Q) yielded a reliability of this subset of 23Q items improved. Without these items, reliability increased from .60 to .73. Item analyses conducted after items 14 and 15 were removed, indicated that removal of thr ee additional items further improved internal consistency of the 23Q. These three items eleted, the reliability increased from .73 to .86. Although the retained items were intended for the screening of ASD in children less than five years of age (i.e., 5.7% of the entire study sample with ASD), the retained items resulted in a reliability of .86 for the 23Q among the whole ASD group. Therefore, combining items 19 through 23 into a total score, for cases with ASD in the Ugandan context, yielded an internal consistency reliability within a range considered adequate for screening. When the total score of the five ASD items recommended for administration to children of all ages (i.e., items 14 18, the ones suggested for deletion by the above analysis) Given that the hypothesized reliability was made based on a total score of all the ten ASD - specific items (i.e., .60), hypothesis 1c was supported because the 23Q reliability was below that of the SRS - 2 and SCQ at .84 and .61, respectively. However, it is also noteworthy that there was evidence that the internal consistency could be substantively improved through item deletions. 62 Research Q uestion 2 In the Ugandan context, do total scores for the three ASD instruments (i.e., SCQ, SRS - 2, and 23Q) yield adequate internal consistency relative to values reported in the test manual or available studies for non - ASD cases? Hypothesis 2a . For a non - ASD sample in the Ugandan context, the internal consistency for the total score of the SCQ was hypothesized to be > .80 (i. - item SCQ for the non - ASD children was .72. Given that it is below the hypothesized value of .80, hypothesis 2a was not supported . The SPSS output following calculati item analyses that indicate how internal consistency may improve if each item is deleted while inspection of that SPSS out put which showed that the overall SCQ reliability would increase (though not substantively). When item 19 was deleted, reliability slightly increased to .73 for the item set, as shown in Appendix A. Further inspection, following the deletion of item 19, sh owed that deletion of any other items would not yield substantive improvements in reliability. Note that item 19 was deleted only for purposes of this analysis. Thus, it was retained for other analyses used to examine other research questions and hypothese s. Hypothesis 2b . Within a non - ASD sample in the Ugandan context, the internal consistency for the total score of the SRS - 2 was predicted to be > - 2 was .89, a value that is slightly b elow the hypothesized score. The reliability did not change even with weaker items deleted, as shown in Appendix B, because all individual items were highly correlated and 63 - item consis tency. Therefore, hypothesis 2b was not supported. Hypothesis 2c . Since no existing data was available for the internal consistency of a composite made up of the ASD - related items from the 23Q, no specific reliability estimate was made for this hypothesis. It was, however, anticipated that the internal consistency for the 23Q composite for non - ASD cases would be lower than that of the SRS - 2 and SCQ due to: a) small number of ASD - related items from the 23Q for ASD screening, and b) ASD screening has typicall y been done at the item level for the 23Q. The Cronbach alpha value obtained from the total score of the ten ASD items (i.e., items 14 - 23) for the 23Q when administered to all parents of children regardless of age was .83. Five items were specifically fo r children aged five years and below (items 19 - 23). The reliability of these five ASD items administered to parents of all children regardless of age (i.e., items 14 - 18) without the additional items of children less than five years of age (i.e., items 19 - 23) was low at .21. Therefore, when used without the additional items for children who are less than five years of age, the composite reliability made up of the ASD - related items moved from relatively high to unacceptably low (see Nunnally & Bernstein, 1 994 for standards). A substantial number of children were screened out as false negatives if the additional items were not administered. Also, utilizing a tota l of nine items, the 23Q reliability increased to .90. Further, when just items 19 - 23 (the five items intended to screen children who are less than five years of age) were utilized, the 23Q yielded a reliability of .99. However, since the reliability of the 23Q summed scores of the ten ASD items (.83), the strict target of this hypothesis, was higher than that of the SCQ at .72 and lower than that of SRS - 2 at .89 for the non - ASD children, hypothesis 2c was not 64 supported . (Yet, the analyses regarding diffe rent item subsets were very informative for how the reliability might be improved). Research Q uestion 3 In the Ugandan context, do total scores from the three instruments (i.e., SCQ, SRS - 2, and 23Q) yield substantial mean differences consistent with construct - related (i.e., ASD construct) differentiation between ASD and non - ASD groups? Hypothesis 3a . For the SCQ total score in the Ugandan context, it was anticipated that the mean of the ASD group would be significantly and substantially higher than th e mean of the non - ASD group. Independent samples t - tests were conducted to examine the mean differences between the two groups. Assumptions of equal variance were tested using the Levene's test. Results indicated that there was a significant difference in the variability of the SCQ scores across the two g roups. Thus, an independent samples t - test was conducted with equal variances not assumed. A significant mean difference was found ( t [95.252] = 21.765, p < 0.001, d = 4.278; for ASD group, M = 22.61, SD = 4.315, n = 49; for non - ASD group M = 5.41, SD = 3. 701, n = 56), which was consistent with the ASD group scoring substantially higher, on average, than the non - ASD group on the SCQ. The effect size d value of 4.278 was considered large according to Cohen's (1988) effect size standards. Overall, hypothesis 3a was supported. Hypothesis 3b . For the SRS - 2 total score in the Ugandan context, it was anticipated that the mean of the ASD group would be significantly and substantially higher than the mean of the non - ASD group. The SRS - 2 was standardized on U.S chil dren and therefore application of the U.S. T - scores to Ugandan children might have been inappropriate. There was a need to find out if the 65 unstandardized and standardized scores (based on U.S norms) equally discriminated between ASD and non - ASD children in this study. Regardless of using raw (unstandardized scores) or U.S standardized score norms in the SRS - 2 manual, significant group mean differences existed between the ASD and non - ASD groups on the total SRS - 2 scores. For both score types, the ASD group h ad higher mean scores than those in the non - ASD group. Using unstandardized (raw) t - test for the SRS - 2, with equal variances not assumed, resulted in t ( 93.949) = 24.057, < 0.001, d = 4.687 ( ASD M = 111.16, SD = 20.233, n = 51; non - ASD M = 26.24, SD = 15.616, n = 55) . Using standardized T - scores, based on US norms -- as provided for in the ma nual, resulted in a significant and large -- standardized effect size ( t [97.993] = 16.468, < 0.001, d = 3.214; ASD M = 77.88 SD = 8.118, n = 50; non - ASD M = 53.24, SD = 7.118, n = 55 ). Despite the 1.47 reduction in effect size d when using US standardized scores compared to raw scores, the difference between the groups in d , the estimate for the SCQ - 2 total scores was considered large regardl ess of whether raw (based on the mean of non - ASD study children) or U.S standardized scores (derived from the manual) were used. Therefore, hypothesis 3b was supported. Hypothesis 3c . For the 23Q total score of the ASD - related items, it was anticipated th at the mean of the ASD group would be significantly and substantially higher than the mean of the non - ASD group. Significant group mean differences existed between the two groups on the total scores for the ten ASD items of the 23Q with the ASD group havi ng a higher mean score compared to the non - groups. Thus, the t - test was adjusted for this issue. A t - test for the 23Q with equal variances not 66 assumed yielded t (10 3.216) = 3.009, = .003, d = 0.578 (for ASD group M = 4.06, SD = 1.943, n = 51; for non - ASD M = 2.95, SD = 1.872, n = 56 ). The mean difference was significant, and the effect size was moderate in magnitude. Also, when analyzed using only the five ASD items intended for children less than five years of age (i.e., items 19 - 23), the results yielded a significant and large mean difference ( t [75.529] = 5.287, < 0.001, d = 1.049, equal variances not assumed; ASD M = 3.08, SD = 1.036, n = 51; non - ASD M = 2.21, SD = .563, n = 56 ) in the expected direction. Overall, h ypothesis 3c was specifically aimed at the 10 - item set of 23Q ASD items and was supported. Research Q uestion 4 In the Ugandan context, will use of the recommended screening cut - scores for each of the three instruments (i.e., SCQ, SRS - 2, and 23Q) result in adequate sensitivity and specificity using DSM - 5 ASD diagnosis vs. non - ASD cases as the outcome variable? For sensitivity, the SCQ identified almost all of children with ASD (98%) using the recommended cut - score (i.e., 15) and screened out 94.6% of the non - ASD children (specificity). The SRS - 2, with the raw cut - scores of 85 and 70 (recommended for clinical and research samples respectively) identified 84.0% a nd 98.0% of cases with ASD (sensitivity) respectively, and 100.0% and 96.4% of cases without ASD (specificity) respectively. SRS - 2 screening results differed when the cut - scores were applied in the standardized T - score metric (see results for both raw and standardized scores under hypothesis 4b). Endorsing one or more items of the 10 ASD specific items on the 23Q identified all children with ASD (i.e., sensitivity of 100.0%), but had a low specificity, i.e., correctly identified none of study children witho ut ASD, respectively. Below, amidst addressing the specific hypotheses, is a detailed breakdown for classification analyses and screening utility of the SCQ, SRS - 2, and 23Q. 67 Hypothesis 4a . Within the Ugandan context, a classification analysis comparing ASD vs. non - ASD based on the SCQ cut - score reported in the manual and the DSM - 5 ASD vs. non - ASD diagnostic outcome would result in a significant Chi - square value, as well as sensitivity of > .80 and specificity of > .80. According to Lord and others (1994), an SCQ cut - off score of 15 or more is indicative of a strong possibility for a clinical diagnosis of ASD. To investigate the applicability of this cut - score among Ugandan children, a classification analysis utilizing a cut - score of 15 or more was conducted on a total of 105 participants ( n = 49 with ASD, n = 56 non - ASD). The Chi - square test was used to statistically assess the significance of the relationship between SCQ classifications and actual diagnostic status of the cases. The Chi - square ( 2 [1, = 105] = 85.926, < .001) was statistically significant. The cut - score of 15 yielded a sensitivity of 95.9% and specificity of 94.6% (see Table 3) for children in this study. This implies that using the SCQ cut - score of 15 correctly identified 96% of th e sample who had ASD while accurately excluding 95% without ASD. Given that the Chi - square result was significant and that both sensitivity and specificity were higher than predicted (i.e., > .80 ), hypothesis 4a was supported. Table 3 . Sensitivity and Specificity of the SCQ Total Raw Score based on Cut - score of 15 Research Group Total Non - ASD ASD SCQ Classification (15 Cut - score) Non - ASD Count 53 2 55 % within Research Group 94.6%** 2.0%* ASD Count 3 47 50 % within Research Group 6.0%* 95.9%** Total Count 56 49 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the total row and total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false 68 negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Hypothesis 4b . Within the Ugandan context, a classification analysis comparing ASD vs. non - ASD based on the SRS - 2 cut - score reported in the manual and the DSM - 5 ASD vs. non - ASD diagnostic outcome would result in a significant Chi - square value, as well as sensitivity of > .80 and specificity of > .80. Constantino and Gruber (2012) recommended 70 and 85 as cut - scores for general research and clinical decision - making purposes respectively. The results of the classification analysis for the raw and standardized cut - score of 70 (based on U.S norms in the manual) are in Tables 4 a nd 5 respectively. Those for the raw and standardized cut - score of 85 (based on U.S. norms in the manual) are in Tables 6 and 7, respectively. For the SRS - 2 raw cut - score of 70, results indicated that the Chi - square was significant ( 2 [1, = 105] = 93.3 61, < .001, with a continuity - corrected Chi - square value of 89.622 and associated with a - value < .001, sensitivity of .98 (i.e., 98%), and specificity of .964 (i.e., 96.4%). When applied in the context of norm - referenced (standardized) T - scores, the SRS - 2 standardized cut - score of 70 resulted in a significant Chi - square ( 2 [1, = 105] = 80.105, < .001, with a continuity - corrected Chi - square value of 76.588 and associated with a - value < .001, sensitivity of .86 (i.e., 86%), and specificity of 1.00 (i.e., 100%). These results indicated that the raw SRS - 2 cut - score of 70 accurately identified 98.0% (compared to 86.0% identified using the standardized cut - score of 70) of children with ASD, while only 2% of ASD cases were missed and inaccurately sc reened out (i.e., false negative cases) in comparison to 14.0% falsely screened out as non - ASD using the same standardized score (i.e., norm - referenced T - score of 70). Also, 70 as a raw score was able to accurately screen out 96.4% of the non - ASD cases. Th us, only 3.6% of non - ASD cases were identified as having ASD (i.e., false positive cases). The standardized (norm - referenced T - score) 69 cut - score of 70, however, accurately screened out all non - ASD children (i.e., had a specificity of 100%). For the SRS - 2 ra w (unstandardized) cut - score of 85, results indicated that the Chi - square was significant ( 2 [1, = 105] = 77.000, < .001), with a continuity - corrected Chi - square value of 73.540 and associated - value < .001, sensitivity of .84 (i.e., 84.0%), and specificity of 1.0 (i.e., 100.0%; Table 6). However, using a norm - referenced standardized T scale cut - score of 85, a sign ificant Chi - square resulted ( 2 [1, = 105] = 17.769, < .001, with a continuity - corrected Chi - square value of 15.429, associated - value of < .001), sensitivity of .28 and specificity of 1.0 (see Table 7). These results indicated that using a raw (un standardized) cut - score of 85 accurately identified 84.0% cases with ASD and accurately screened out 100% of non - ASD cases. Thus, the SRS - 2 mis - classified 16.0% of children with ASD as non - ASD cases. Despite having a specificity of 1.0, the standardized cu t - score of 85 had a very low sensitivity compared to its raw (unstandardized) cut - score (i.e., a false negative rate of .72). Both the raw (unstandardized) and standardized cut - score of 70 yielded high sensitivity and specificity. However, only the raw (un standardized) cut - score of 85 yielded a good sensitivity (i.e., .84) in this study. The standardized cut - score of 85 had low sensitivity (i.e., .28). Also, both SRS - 2 cut - scores (i.e., 70 and 85) had excellent specificities, classifying close to all non - AS D children as true negatives. Since the SRS - 2 is usually scored and interpreted using standardized T - scores, results for this score type were used to evaluate hypothesis 4b. For standardized T - scores, results for the cut - score of 70 (i.e., sensitivity of 9 8.0 and specificity of 96.4) were consistent with the hypothesis, but results for the cut - score of 85 (i.e., sensitivity of .28 and specificity of 1.0) fell well short of the .80 benchmark for sensitivity. Thus, overall, hypothesis 4b was supported for a c ut - score of 70, but not for a cut - score of 85. 70 Table 4 . Sensitivity and Specificity of the SRS - 2 Total Score Based on Raw Cut - score of 70 Research Group Total Non - ASD ASD SRS - 2 Non - ASD Count 53 1 54 % within Research Group 96.4%** 2.0%* ASD Count 2 49 51 % within Research Group 3.6%* 98.0%** Total Count 55 50 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the Total row and Total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Table 5 . Sensitivity and Specificity of the SRS - 2 Total Standardized Cut - score (T - scores) of 70 Research Group Total Non - ASD ASD SRS - 2 Non - ASD Count 53 1 54 % within Research Group 96.4%** 2.0%* ASD Count 2 49 51 % within Research Group 3.6%* 98.0%** 48.5% Total Count 55 50 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the Total row and Total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Table 6 . Sensitivity and Specificity of the SRS - 2 Total Raw Cut - score of 85 Research Group Total Non - ASD ASD SRS - 2 Non - ASD Count 55 8 63 % within Research Group 100.0%** 16.0%* ASD Count 0 42 42 % within Research Group 0.0%* 84.0%** Total Count 55 50 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the total row and total column, the upper left 71 quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Table 7 . Sensitivity and Specificity of the SRS - 2 Tota l Standardized Cut - score (T - scores) of 85 Research Group Total Non - ASD ASD Group SRS - 2 Non - ASD Count 55 36 91 % within Research Group 100.0%** 72.0%* ASD Count 0 14 14 % within Research Group 0.0%* 28.0%** Total Count 55 50 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the Total row and Total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Hypothesis 4c . Using the screening criterion for the ASD items on the 23Q (i.e., one or more ASD - related items affirmatively endorsed) compared to the DSM - 5 ASD vs. non - ASD diagnostic criterion, no specific indices for screening sensitivity or specificity target values were used because of a lack of clear prior findings. However , lower sensitivity and specificity for the 23Q ASD - related items was hypothesized relative to the SCQ and the SRS - 2, as these other instruments consist of many more items and likely better represent the ASD construct. The Chi - square test was used to asses s the relationship for statistical significance between 10 - item 23Q dichotomous classifications and known DSM - 5 ASD vs. non - ASD classifications. The 2 x 2 classification analysis (i.e., cross - tabulation of ASD vs. non - ASD according to 23Q and ASD vs. non - A SD using DSM - 5 diagnosis) is presented in Table 8. The endorsement of one or more of the 10 ASD items of the 23Q, as recommended by the authors, for one to screen positive for a potential ASD diagnosis yielded a sensitivity of 1.0 and specificity of 0.0 am ong children in the study. The recommended endorsement score of one or more despite identifying 72 all ASD cases equally had a very high false positive rate. The sensitivity of the 10 summed ASD items for the 23Q was in the same range as that of the SRS - 2 and SCQ cut - scores of 70 and 15, respectively (i.e., .98 for both cut - scores) but better than that of the SRS - 2 clinical cut - score of 85 (i.e., .82). For specificity, the 23Q ten summed ASD items performed much poorer, identifying none (i.e., 0.0%) of true ne gative children, when compared to both SRS - 2 cut - scores of 70 and 85 (i.e., 96.3% and 100% respectively) and SCQ cut - score of 15 at 94.6% in this study sample. Therefore, given that the 23Q sensitivity was slightly higher than that of the SRS - 2 (for both g eneral and clinical cut - scores) and that of the SCQ cut - score of 15, but substantially lower in specificity when compared to both the SRS - 2 and SCQ, hypothesis 4c was not supported . However, it should be clear that the 10 ASD 23Q items ended up screening a ll cases positive for ASD and none of the cases negative for ASD (regardless of actual classification). This suggests very poor overall performance relative to the SCQ and SRS - 2 when the endorsement of one or more items is used as the cut - off for the 10 23 Q ASD items. Table 8 . Sensitivity and Specificity of the 23Q Based on Using a Cut - score of 1 or more on the Ten summed ASD Items Research Group Total Non - ASD ASD 23Q ( > 1 Cut - score) ASD Count 56 51 107 % within Research Group 100.0%* 100.0%** Total Count 56 51 107 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the total row and total column, the upper left quadrant is false positive cases and upper right quadrant is true posi tive cases (i.e., sensitivity as a percent). Though not explicit in the table, specificity was 0.00% (i.e., all non - ASD misclassified as ASD by the 23Q). 73 Research Q uestion 5 Beyond the recommended cut - scores for screening reported in each test manual, are there more optimal cut - scores for each of the three instruments (i.e., SCQ, SRS - 2, and 23Q) in the Ugandan context? For this study, the SCQ cut - score of 10 (as compared to 1 5 recommended by SCQ authors) was found to more optimally distinguish between ASD and non - ASD children. (However, it must be noted that the sensitivity and specificity of the cut - score of 15 were both excellent according to standards from Nunnally and Bern stein [1994].) The cut - score of 10 identified all children at risk of ASD, which was not the case with 15. For the SRS - 2, a cut - score of 62 performed better than the author - recommended cut - scores of 70 and 85 (for research and clinical decision making, res pectively). Using the 23Q, a total cut - score of 3 performed better than 1 (i.e., the author - recommended cut - score) in simultaneously identifying those at - risk of ASD, while screening out those who do not have ASD. Below is a detailed breakdown of the explo ration of alternative cut - scores showing analyses that led to these new cut - scores for the three study instruments. Exploration of alternative cut - scores for the SCQ . A Receiver Operating Characteristic (ROC) Curve was utilized in assessing for an alternat ive cut - score yielding a more optimal balance between sensitivity and specificity for the SCQ total score in the study sample (see Table 9). The area under the ROC curve (AUC) quantifies the proportion of accurately classified cases or how well an instrume - score differentiates between groups -- such as those with and without ASD for this study. The ROC sensitivity and specificity proportions each range from 0 (accurately identifying none) to 1 (accurately identifying all). 74 Table 9 . Receiver Operating Characteristic (ROC) Curve of the SCQ Total Cut - Scores Possible SCQ Cut - Scores Sensitivity Specificity - 1.00 1.000 0.000 .50 1.000 0.071 1.50 1.000 0.161 2.50 1.000 0.232 3.50 1.000 0.304 4.50 1.000 0.411 5.50 1.000 0.518 6.50 1.000 0.679 7.50 1.000 0.768 8.50 1.000 0.839 10.00** 1.000 ++ 0.929 + 12.00 0.980 0.946 14.00 0.959 0.946 + 15.00* 0.959 ++ 0.946 + 17.50 .857 1.000 18.50 .837 1.000 19.50 .735 1.000 20.50 .673 1.000 21.50 .633 1.000 22.50 .592 1.000 23.50 .531 1.000 24.50 .367 1.000 25.50 .245 1.000 26.50 .204 1.000 27.50 .082 1.000 28.50 .061 1.000 30.00 .020 1.000 32.00 .000 1.000 Notes. Bolded black numbers show approximations of the sensitivity and specificity of different SCQ cut - scores. * Shows an approximation of the author recommended cut - score of 15 (which is the same as the values for a cut - score of 14 depicted in the table) while ** shows the performance of 10 as an alternative cut - score. + and ++ their respective sensitivity and specificity. The cut - score of 15 as proposed by the manual had a sensitivity of .96 and specificity of .95 (see Table 10), thus, close to all the children in the study sample were screened out as true negatives or in as true positives. Further analysis of the ROC curve showed that lowering the cut - 75 score to 10 appropriately distinguished non - ASD children from those with ASD by identifying all ASD chi ldren as ASD (i.e., sensitivity of 1.0) although specificity dropped to .93 (see Table 11). With the 49 children with ASD in the study sample, the cut - score of 10 correctly identified all 49 children (a positive predictive value of 1.0). For the non - ASD gr oup, the cut - score of 10 displayed a negative predictive value of .93 (i.e., identified 52 out of 56 children without ASD). The false positive percentage using 10 as a cut - score was low at 7.1% (4 of 56 without ASD) with a false negative rate of 0% (0 out of 49 children with ASD) for this study sample (see Table 11). Table 10 . Sensitivity and Specificity of the SCQ Based on the Cut - score of 15 Research Group Total Non - ASD ASD SCQ Classification (15 Cut - score) Non - ASD Count 53 2 55 % within Research Group 94.6%** 2.0%* ASD Count 3 47 50 % within Research Group 6.0%* 95.9%** Total Count 56 49 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the total row and total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). 76 Table 11 . Sensitivity and Specificity of the SCQ Based on the Cut - score of 10 Research Group Total Non - ASD ASD SCQ Classification (10 Cut - score) Non - ASD Count 52 0 52 % within Research Group 92.9%** 0.0%* ASD Count 4 49 53 % within Research Group 7.1%* 100.0%** Total Count 56 49 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the total row and total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Exploration of alternative cut - scores for the SRS - 2 . A summary of the performance of the cut - scores 70, 85 and 62 when used both as unstandardized (raw) and standardized cut - scores is provided in Table 12. Standardized scores were based on T - scores provided in the SRS - 2 manual while raw scores were unstandardized scores of p arental rating of study children on the SRS - 2, respectively. Table 12 . Summary of Sensitivity and Specificity of Recommended Versus Alternative SRS - 2 Raw and Standardized Cut - Scores Cut - Score Sensitivity Specificity False Positive False Negative 70 (Raw) 70 (Standardized) 98.0% 86.0% 96.4% 100% 3.6% 0.0% 2.0% 14.0% 85 (Raw) 85 (Standardized) 84.0% 28.0% 100.0% 100.0% 0.0% 0.0% 16.0% 72.0% 62 (Raw) 62 (Standardized) 100% 100% 96.4% 96.4% 3.6% 3.6% 0.0% 0.0% 77 Notes. All standardized scores are based on U.S. T - scores provided in the SRS - 2 Manual. Raw scores are based on unstandardized raw scores of study children. Following a review of the ROC curve coordinates and associated sensitivity and specificity values using raw scores (see Table 13), the cut - score of 62 performed slightly better than the author recommended cut - scores of 70 and 85. Table 13 . Receiver Operating Characteristic (ROC) Curve of the SRS - 2 Alternative Total Cut - Scores Using Raw Scores Positive if Greater Than or Equal To a Sensitivity Specificity 2 1.000 0.000 4 1.000 0.018 7 1.000 0.036 9.5 1.000 0.073 10.5 1.000 0.109 11.5 1.000 0.145 12.5 1.000 0.164 14 1.000 0.218 15.5 1.000 0.236 16.5 1.000 0.255 17.5 1.000 0.273 18.5 1.000 0.327 19.5 1.000 0.364 20.5 1.000 0.400 21.5 1.000 0.436 22.5 1.000 0.491 23.5 1.000 0.509 24.5 1.000 0.564 25.5 1.000 0.600 26.5 1.000 0.636 28 1.000 0.655 29.5 1.000 0.673 30.5 1.000 0.691 32 1.000 0.745 34 1.000 0.782 35.5 1.000 0.800 36.5 1.000 0.818 38.5 1.000 0.836 42.5 1.000 0.873 45.5 1.000 0.891 47 1.000 0.909 78 Table 13 ( ) 52.5 1.000 0.927 57.5 1.000 0.945 62*** 1.000 0.964 68.5 0.980 0.964 70* 0.980 0.964 71.5 0.980 0.982 72.5 0.961 0.982 76 0.941 0.982 79.5 0.922 0.982 80.5 0.902 0.982 82 0.902 1.000 83.5 0.882 1.000 84.5** 0.843 1.000 86 0.824 1.000 89 0.804 1.000 93.5 0.784 1.000 97 0.745 1.000 99 0.725 1.000 101.5 0.706 1.000 103.5 0.686 1.000 104.5 0.667 1.000 106 0.647 1.000 107.5 0.627 1.000 109 0.608 1.000 111 0.569 1.000 112.5 0.549 1.000 114.5 0.490 1.000 117.5 0.471 1.000 120 0.412 1.000 121.5 0.392 1.000 122.5 0.373 1.000 124 0.353 1.000 125.5 0.275 1.000 127 0.235 1.000 128.5 0.196 1.000 129.5 0.176 1.000 130.5 0.137 1.000 131.5 0.098 1.000 134 0.078 1.000 136.5 0.059 1.000 138.5 0.039 1.000 147 0.020 1.000 79 Table 13 ( ) 155 0.000 1.000 Notes. *** shows the performance (sensitivity and specificity) of an alternative cut - score of 62. * and ** depict the sensitivity and specificity of the author recommended research and clinical cut - scores (i.e., 70 [same sensitivity and specificity as 68.5 in th e table] and 85 respectively) among study children. When used to classify children in the study, 62 as a cut - score (raw) classified all children with ASD correctly, that is 1.0 for ASD and .96 for correct classification of non - ASD (see Table 14). This was slightly better than the sensitivity of .98 but the same for the specificity (i.e., .96) obtained when using the raw cut - score of 70. The alternative cut - score of 62 also performed - score for clinical decision makin g (i.e., 85 raw cut - score) which had a sensitivity of .84. However, for specificity, the unstandardized or raw cut - score of 85 identified all non - ASD correctly (i.e., specificity of 1.0) hence performing better than 62 in that respect. Table 14 . Detailed Breakdown of Sensitivity and Specificity Based on an SRS - 2 Raw - score of 62 Classification Analysis Research Group Total Non - ASD ASD Classification with 62 as Alternative Cut - score Non - ASD Count 53 0 52 % within Research Group 96.4% 0.0% ASD Count 2 50 38 % within Research Group 3.6% 100.0% Total Count 55 50 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the total row and total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Among standardized T - scores based on U.S. norms provided in the manual, the RO C curve coordinates showed that a T - score of 62 was a better cut - score than standard scores of 70 80 or 85 (see Table 15). When applied to classify the sample in this study, the cut - score of 62 maintained a high sensitivity (i.e., 1.00) with a slight drop in specificity to .964 (See Table 16). Table 15 . Receiver Operating Characteristic (ROC) Curve of the SRS - 2 Standardized Alternative Total Cut - Scores (U.S. Norms) Positive if Greater Than or Equal To a Sensitivity Specificity 39.00 1.000 0.000 40.50 1.000 0.036 41.50 1.000 0.091 42.50 1.000 0.109 43.50 1.000 0.127 44.50 1.000 0.164 45.50 1.000 0.236 46.50 1.000 0.255 48.00 1.000 0.291 49.50 1.000 0.327 52.00 1.000 0.345 54.50 1.000 0.382 55.50 1.000 0.473 56.50 1.000 0.564 57.50 1.000 0.673 58.50 1.000 0.782 59.50 1.000 0.873 60.50 1.000 0.927 62.00 1.000 0.964 64.00 0.980 0.964 65.50 0.960 0.982 66.50 0.920 0.982 67.50 0.900 0.982 68.50 0.860 0.982 69.50 0.860 1.000 70.00 0.860 1.000 71.00 0.780 1.000 72.50 0.720 1.000 73.50 0.680 1.000 74.50 0.600 1.000 75.50 0.540 1.000 76.50 0.500 1.000 77.50 0.480 1.000 78.50 0.420 1.000 79.50 0.400 1.000 81 Table 15 ( ) 80.50 0.360 1.000 81.50 0.340 1.000 83.00 0.320 1.000 85.00 0.280 1.000 85.50 0.240 1.000 86.50 0.180 1.000 87.50 0.140 1.000 88.50 0.100 1.000 89.50 0.060 1.000 90.50 0.040 1.000 94.50 0.020 1.000 99.00 0.000 1.000 Note. Scores are in the norm - referenced T - score metric (U.S. norms). Cut - scores were standardized for sex (males and females combined) as rated by a parent. Table 16 . Sensitivity and Specificity Based on an SRS - 2 T - score Classification of 62 in Study Children Research Group Total Non - ASD ASD Classification Using 62 as T Cut - score Non - ASD Count 53 2 55 % within Research Group 96.4%** 3.6%* ASD Count 0 50 50 % within Research Group 0.0%* 100.0%** Total Count 55 50 105 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the Total row and Total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Exploration of alternative cut - scores for the 23Q . In the 23Q screening, a cut - score of 1, as used by Kakooza - Mwesige and colleagues (2014) for identification of children with ASD, failed to adequately discriminate between the children in the different groups (Table 17). Despite identifying all children wi th ASD, the cut - score of 1 equally screened all non - ASD individuals in 82 as false positives. A further classification analysis was performed using the cut - scores of 3 (Table 18) and 4 (Table 19). A sensitivity and specificity of .77 and .68 for the cut - score of 3 and .55 and .80 for the cut - score of 4 were attained. Although 4 as a cut - score had a higher specificity than 3, it had a much lower sensitivity. Therefore, a cut - score of 3 appeared to balance sensitivity and specificity more appropriately (compared to cut - scores of both 1 and 4) for the Ugandan children in this sample. Thus, for research question five, each of the three instruments (i.e., SCQ, SRS - 2, and 23Q) yielded different, more optimal alternative cut - scores of 10, 62 and 3, respectively. Table 17 . Total Cut - scores Positive if Greater Than or Equal To a Sensitivity Specificity Table 17 ( ) 1.0 1.000 0.000 2.0 .961 0.036 3.0 + .765 + 0.679 + 4.0 .549 0.804 5.0 .353 0.857 6.0 .196 0.857 7.0 .137 0.857 8.0 .078 0.964 9.0 .020 1.000 10.0 .000 1.000 Note . Bolded number represents the sensitivity and specificity of the cut - score consistent with + represents the performance of alternative cut - scores. 83 Table 18 . 23Q Sensitivity and S pecificity of the 23Q Based on a cut - score of 3.0 Research Group Total Non - ASD ASD 23Q Classification Using 3.0 Cut - score for ASD Non - ASD Count 38 12 50 % within Research Group 67.9%** 23.5%* ASD Count 18 39 57 % within Research Group 32.1%* 76.5%** Total Count 56 51 107 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the total row and total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Table 19 . Sensitivity and Specificity Based on the 23Q Cut - score of 4.0 Research Group Total Non - ASD ASD 23Q Classification based 4.0 Cut - score Non - ASD Count 45 23 68 % within Research Group 80.4%** 45.1%* ASD Count 11 28 39 % within Research Group 19.6%* 54.9%** Total Count 56 51 107 % within Research Group 100.0% 100.0% Note. ** indicates accurate or true screening results and * indicates inaccurate or false screening results. Each column sums to 100%. Ignoring the Total row and Total column, the upper left quadrant is true negative cases (i.e., specificity as a percent), uppe r right quadrant reflects false negative cases, lower left quadrant is false positive cases, and lower right quadrant is true positive cases (i.e., sensitivity as a percent). Research Q uestion 6 Are there significant convergent relationships among the t otal scores of the three instruments (i.e., SCQ, SRS - 2, and 23Q) in the Ugandan context? Research question 6 investigated the correlations between the total scores of the three instruments (i.e., SCQ, SRS - 2, and 23Q). Below is a breakdown of the convergent validity 84 hypotheses and results for each. The total sample (i.e., ASD and non - ASD cased combined) was used to calculate all correlations. Hypothesis 6a . Within the Ugandan context, the correlation across the instruments on the total scores of SRS - 2 and SCQ was hypothesized to be significant, positive, and substantive ( > .50). This is based on the reported correlations between the SRS - 2 total score and SCQ total score ranging from .68 in non - ASD group to .50 for ASD - specific group in the SRS - 2 manual. The SRS - 2 and SCQ total scores had a significant, positive and substantive correlation: (1 05) = .93, < .001. The significance, direction, and general magnitude of the relationship were as predicted. Thus, hypothesis 6a was supported. Hypothesis 6b . Within the Ugandan context, the correlation between the SRS - 2 total score and 23Q summed ASD - related items score will be significant and positive but lower than the correlation between the SRS - 2 and the SCQ total scores . There was a non - significant positive correlation (105) = .19, = .053 between the SRS - 2 total score and the 23Q summed ASD - related items (10 ASD 23Q items, i.e., 23Q items 14 23). When only the five 23Q items (i.e., 19 - 23) typically used in screening for ASD in children below five years of age were utilized, the correlation between the five - item 23Q score and the SRS - 2 was higher and significant ( [105] = .35, < .001). When compared to the magnitude of the correlation between the SCQ and SRS - 2 total scores (i.e., r = .93), the correlations involving variations of the summed 23Q ten ASD - related items with the SRS - 2 total score were both lower (i.e., r = .19 and r = .35, respectively) and only significant when the five ASD items meant to assess children below five years (items 19 23) were used. Hypothesis 6b was not supported for the 10 - item 23Q summative scale (i.e., items 14 23), because the correlation was not significant. 85 Hypothesis 6c . Within the Ugandan context, correlation between the SCQ total score and the 23Q summed ASD - related items score will be significant and positive but lower than the corr elation between the SRS - 2 and the SCQ total scores. The correlation between the SCQ total score and 23Q summed 10 ASD - related items was positive but not significant (105) = .17, = .082 . The correlation magnitude was lower than that between the SRS - 2 and SCQ total scores (i.e., r = .93). However, the SCQ total score correlation with the five ASD items used to screen for ASD among children younger than five years (items 19 - 23) was moderate and significant (i.e., r [105] = .36, < .001) but still ha ving a lower correlation than that between the SRS - 2 and SCQ total scores. Hypothesis 6c was not supported when using all the 10 ASD 23Q items (items 14 23), because the correlation was not significant. Findings from all research questions and related hypotheses are summarized below in Table 20. This table is intended to put all of the important results in one place for the reader. Table 20 . Research Hypotheses Findings Summary Hy potheses Hypotheses Outcome 1a) SCQ reliability for ASD group to be > .80. Not supported 1b) SRS - 2 reliability for ASD group to be > .90. Not supported 1c) 23Q composite reliability lower than that of the SRS - 2 and SCQ for ASD group. Supported 2a) SCQ reliability for non - ASD group to be > .80. Not supported 2b) SRS - 2 reliability for non - ASD group to be > .90. Not supported 86 Table 20 ( ) 2c) 23Q composite reliability for non - ASD group to be lower than that of the SRS - 2 and SCQ. Not supported 3a) SCQ mean of the ASD group to be higher than that of the typically developing group. Supported 3b) SRS - 2 mean of the ASD group to be higher than that of the typically developing group. Supported 3c) 23Q mean of the ASD group to be higher than that of the typically developing group. Supported 4a) SCQ to have a sensitivity and specificity of > .80. Supported 4b) SRS - 2 to yield a sensitivity and specificity of > .80. Supported for a cut - score of 70, but not a cut - score of 85 4c) The 23Q ASD - related items to have a lower sensitivity and specificity relative to the SCQ and the SRS - 2. Not supported, but much worse at classification overall 5) No specific hypotheses were set for alternative cut - scores. All three instruments yielded alternative optimal cut - scores (i.e., 10, 62 and 3 for the SCQ, SRS - 2 and 23Q, respectively). 6a) Correlations of the SRS - 2 and SCQ to be > .50. Supported 6b) Correlations of the SRS - 2 and 23Q ASD - related item s hypothesized to be lower than the correlations between the SRS - 2 and the SCQ. Not supported when using 10 - item 23Q ASD summative scale items (i.e., 14 23), because correlation was not significant. 6c) Correlations between the SCQ and the 23Q hypothesized to be lower than the correlation between the SRS - 2 and the SCQ. Not supported when using the 10 item 23Q ASD summative scale (items 14 23), because correlation was not significant. 87 CHAPTER 5 : DISCUSSION The primary purpose of this study was to investigate the psychometric properties of caregiver - rated SCQ (Rutter, Bailey & Lord, 2003), SRS - 2 (Constantino & Gruber, 2012), and 23Q (Kakooza - Mwesige et al., 2014) in screening 4 - to - 18 - year - old children for ASD in Uganda. In this chapter, psychometric findings from the SCQ, SRS - 2, and 23Q in the present study are compared to results from similar validation studies and put into context within the literature. Study strengths and limitations, clinical implications, and research implications and future directions will also be discussed. As revealed in the prior chapter, internal consistency reliability findings in the present study sample varied considerably for each of the three instruments, depending on whether the estimate came from the ASD group, non - ASD group, or combined sample. Of the three instruments, the SRS - 2 yielded consistently higher reliability across the three conditions. In terms of mean differences between ASD and non - ASD groups, the SCQ and SRS - 2 yielded effect size d estimates that were > 4.00 in the raw score metric, while the set of 10 ASD items from the 23Q yielded a more modest d = .58 effect size. Sensitivity and specificity findings for both the SCQ and SRS - 2 were very good (all > .90 for b oth sensitivity and specificity) based on recommended cut - scores of 15 (SCQ) and 70 (SRS - 2 raw score), as well for more optimal sample - specific cut - scores of 10 (SCQ) and 62 (SRS - 2 raw score). However, sensitivity and specificity results for the set of 10 ASD items from the 23Q ranged from very poor (for the recommended cut - score of one or more) to modest (.77 sensitivity and .68 specificity based on an optimized sample - specific cut - core of three or more). Finally, the correlation between the SCQ and SRS - 2 total scores was high (> .90), while the correlations between the summative 88 score for the 10 23Q ASD items and both the SCQ and the SRS - 2 total scores were low and non - significant (< .20). Reliability The SCQ An SCQ internal consistency reliability of .61 was found in the ASD sample and an estimate of .72 was obtained among those without ASD. For research purposes, a reliability of .70 or higher is considered adequate, while .90 and above is recommended for assessments from which critical clinical decisions are to be made (Nunnally & Bernstein, 1994). Internal consistency estimates of .80 or higher are considered appropriate for screening purposes (Salvia, Ysseldyke, & Bolt, 2010). Thus, the SCQ reliability reported in this study for the A SD sample was below that found by Lord and colleagues (2003), that is, .81 and .86 (autism - specific and broader ASD groups respectively), and .92 among the non - spectrum disorders. This is also below that reported by Bolte and others (2008) for the German S CQ version (i.e., .83 among those with ASD), although they did not report the reliability in the non - autism group. All other SCQ validation studies reviewed for this purpose reported overall internal consistency reliability based on data from both ASD and non - ASD cases pooled together as one sample -- and not separate estimates for ASD and non - ASD. Therefore, further comparisons were made based on this type of overall internal consistency estimate. In the present study, an overall SCQ internal consistency of .93 resulted when ASD and non - ASD cases were combined and treated as one sample. This estimate was similar to those found in studies with the Arabic (.92) and Greek (.91) versions of the SCQ (Aldosari et al., 2020; Zarokanellou et al., 2017), but generall y higher than values reported for the Turkish (.89) and German (.83) versions of the SCQ (Avcil et al., 2014; Bolte et al., 2008). 89 In summary, the internal consistency reliability for the SCQ for children with ASD in the Ugandan context (i.e., based on A SD cases only) was lower than expected and fell short of standards for both research and clinical use. In the non - ASD sample from Uganda, the internal consistency estimate met research standards, but fell short of standards for screening and clinical use. However, when the internal consistency reliability was estimated based on the pooled ASD and non - ASD cases, the SCQ internal consistency estimate was adequate; met standards for research, screening, and clinical use; and was generally comparable to SCQ est imates from other cultural studies. The SRS - 2 The SRS - 2 internal consistency reliability estimates among the ASD and non - ASD groups were .84 and .89 respectively , which is adequate for both screening (Salvia, Ysseldyke, & Bolt, 2010) and research purposes - - but below the .90 minimum recommended for use in clinical diagnosis or other important decisions about an individual (Nunnally & Bernstein, 1994). It is noteworthy that the internal consistency estimate for the combined sample [i.e., both ASD and non - ASD cases pooled together] was .97, which meets standards for research, screening, and use in individual decision making though, in practice, one should not use a single measure or source for important clinical decisions. The group - specific internal consistenc y reliability estimates were also lower than those reported for a larger clinical sample in the SRS - 2 manual (i.e., .95 for those with ASD and .97 for unaffected siblings, based on parent ratings; Constantino & Gruber., 2012). However, they are similar to estimates found for the Mandarin version in Taiwan (i.e., .87 and .85 for ASD and control groups, respectively) as well as .89 for non - ASD groups in both the Vietnam and China (Mandarin) studies (Cen et al., 2017; Nguyen et al., 2019; Wang et al., 2012). A lso, the reliability found among those with ASD for this study (.84), was 90 lower than the .96 found in the Vietnamese and German SRS studies, and the .92 result in both the Mandarin and Spanish SRS studies in China and Mexico. The internal consistency among the non - ASD group in the present study (.89) was similar to, though slightly lower than, estimates reported for the German and Spanish SRS versions (i.e., .91 and .92, respectively; Bolte et al., 2011; Fombonne et al., 2012). Though differences in variabi lity play a role in these differences across studies, it is not clear if these differences in sample variability and reliability estimates truly reflect culture - specific or language - specific issues. Yet, these studies provide what are currently our best es timates of SRS reliability in these contexts. In summary, the SRS - 2 internal consistency estimates were lower than expected when calculated separately for ASD and non - ASD groups in the Ugandan context though standards were met for research and screening use in both groups, and the .97 internal consistency value for the two groups combined met standards for research, screening, and individual decision making. However, the group - specific estimates were consistent with a number of other estimates r eported in other cultural contexts, and these were consistently higher than the group - specific estimates obtained for the SCQ in the present study. The 23Q This was the first study to examine the internal consistency reliability of a summed subscale made u p of the 10 ASD items from the 23Q in this case, screening children with a known diagnostic status (i.e., either ASD or non - ASD). Internal consistency estimates of .60 and .83 were found for the 10 ASD items of the 23Q among the ASD and non - ASD groups, res consistency for this 10 - item scale (23Q items 14 - 23) when evaluated with all ASD and non - ASD cases combined (i.e., with ASD and non - ASD groups combined) was .65. Howeve r, the 91 estimate increased to .92 when only items 19 23 (five items more typically administered only to parents of children less than five years of age) were included. In addition, a summed composite consisting of just items 19 23 with an optimal cut - sc ore identified more children at risk of ASD than when all 10 ASD items available on the 23Q (i.e., items 14 23) were used. It is of critical importance to mention that items like these administered to children less than five years of age on the 23Q (i.e. , 19 23) are regularly utilized in other instruments as part of the assessment of older children (e.g., SRS - 2). Taken together, these findings suggest that, in the context of 23Q usage, items 19 - 23 on the 23Q should be considered as standard for rating a ll children regardless of age -- and not used exclusively to rate children younger than five years old. Overall, all three instruments met internal consistency standards for research, screening, and clinical use under at least some conditions (e.g., SCQ in the combined sample, SRS - 2 in the combined sample, and 23Q in the combined sample when a subgroup of five items were used). However, the SRS - 2 performed relatively better than the other two instruments under all conditions (i.e., ASD only, non - ASD only, a nd combined samples). Critically, the 23Q 10 - ASD - item composite performed poorly overall in terms of reliability with estimates below .70 in both the ASD - only and combined (ASD and non - ASD combined) conditions. Nevertheless, its performance improved consid erably (> .90) in the combined context when a subgroup of five more internal consistent items (23Q items 19 - 23) was used without the other five items. Validity Mean Differ ences For all three instruments, and consistent with the ASD construct assessed by each measure, the ASD group scored significantly higher in comparison to the non - ASD group -- providing initial evidence of differentiation (clinical discrimination) between clinical and non - 92 clinical samples. Both the SRS - 2 and SCQ total raw scor es yielded very large standardized mean differences (i.e., d > 4.00 for both instruments) between the ASD and non - ASD group means. The difference between groups remained very large ( d = 3.20) for the SRS - 2 even when norm - referenced T - scores (based on U.S. norms) were used. However, the SRS - 2 effect size d in the raw score metric ( d = 4.69) was relatively, and considerably, larger than in the T - score metric ( d = 3.21) -- a difference of almost 1.5 standard deviations. This difference due to metric should be co nsidered when deciding which type of score to use. Consistent with the SCQ mean difference found in the present study, in the Ugandan context, large mean differences were reported in the SCQ manual, as well as in Greek, Chinese, and Arabic validation studies (Aldosari et al., 2019; Gau et al., 2011; Lord e t al., 2003; Zarokanellou et al., 2017). Also, generally consistent with the present study results for the SRS - 2, the large SRS - 2 mean differences between ASD and non - ASD groups were reported in the SRS - 2 manual (i.e., means of 106.6 and 24.6, for ASD and unaffected/control group, respectively; d = 2.70; Constantino and Gruber., 2012) -- though the exact effect size d value was relatively smaller in the SRS - 2 manual. Further, large mean differences were reported for SRS - 2 studies in Mexico, Germany, and China (Cen et al., 2017; Cholemkery et al., 2014; Fombonne et al., 2012). For the 23Q, the standardized mean difference between ASD and non - ASD groups in the present study was in the medium range ( d = .58) for the sum of the 10 ASD items. Though significant, t his d value was considerably smaller than those observed for the SCQ and SRS - 2 and suggests relatively less differentiation between ASD and non - ASD groups for the 23Q items in the Ugandan context. However, it is noteworthy, that when only five items (23Q i tems 19 - 23) were used as a scale, the effect size d increased to 1.05. This value is still smaller than those for 93 the SCQ and SRS - 2, but it is a considerable improvement over the 10 - item version. As noted previously, these five 23Q items are typically admi nistered only when children are under five years old. However, they were completed for children of all ages (both younger and older) in the present study and appeared to function very well within this sample. Thus, consideration should be given to administ In summary, all three instruments yielded significant mean differences between ASD and non - ASD groups. However, standardized mean differences were considerably larger for the SCQ and SRS - 2 than the - ASD improved significantly when only five of the ASD - related items (i.e., items 19 - 23) were used to construct the scale. Nevertheless, even in this context, the SCQ and SRS - 2 betwee n group differences were considerably larger. Sensitivity and Specificity The SCQ. The SCQ cut - score of 15 (recommended in the SCQ manual) had excellent sensitivity and specificity (i.e., .96 and .95, respectively) in the present Ugandan sample. These values are higher than those reported for the same cut - score in the SCQ manual (i.e., sensitivity of .85 and specificity of .75; Lord et al., 2003). In terms of the SCQ in other cultural contexts, the Greek version of the SCQ had similarly high sensitivity and specificity estimates (i.e., .96 and .99, respectively). However, this was not th e case in the Arabic, Turkish, Chinese, and German SCQ validation studies. The sensitivity and specificity were .80 and .97 for the Arabic version, .71 and .77 for the Chinese version, 1.0 and .33 for the Turkish version, and .89 and .91 for the German ver sion (Aldosari et al., 2019; Avcil et al., 2015; Bolte et al., 2008; Gau et al., 2011; Zarokanellou et al., 2017). Thus, sensitivity and specificity values vary considerably among SCQ validation studies across various cultural contexts. However, the values reported for 94 the Ugandan (present study) and Greek validation studies stand out high relative to those reported from other contexts. The SRS - 2. According to the SRS - 2 manual, raw scores are typically used for large scale screening and reported in researc h studies involving clinical discriminant work, while norm - referenced T - scores are more typically used for communication in clinical and educational settings (see Constantino & Gruber, 2012). The manual indicates that a variety of different possible cut - sc ores could be used for screening -- depending on the screening context, groups being compared, base rate of ASD in that setting, purpose of a research study, etc. However, a cut - score of 70 (raw score) is recommended for screening minimal risk samples and cu t - score of 85 (raw score) for clinical samples where higher confidence in positive screens is desirable (Constantino & Gruber, 2012). In the present study, these cut - scores were applied in both the raw score and norm - referenced T - score metric. In the raw score metric context, sensitivity was .98 and specificity was .96 for a cut - score of 70, and .84 and 1.00 for a cut - score of 85. In the norm - referenced T - score context (U.S. norms), sensitivity was .86 and specificity was 1.00 for a cut - score of 70, and .2 8 and 1.00, respectively, for a cut - score of 85. Constantino and Gruber (2012) reported sensitivity of .78 and specificity of .94 for a raw cut - score of 70 in the U.S. context. In the present study sample from Uganda, sensitivity, and specificity values fo r raw cut - scores of 70 and 85, as well as for the norm - reference T - cut - score of 70, exceeded values reported by Constantino and Gruber for a raw score of 70 based on U.S. data. Constantino and Gruber (2012) did not report sensitivity and specificity estim ates for a cut - score of 85, but they did suggest that its use would generally be in a different clinical context where the base rate of ASD is higher and one is looking for a higher degree of confidence in the 95 positive screening decision (see p. 21) -- thoug h the risk false positives would be reduced, the risk of false negatives may be higher in this situation. When compared to other validation studies that utilized a raw cut - score of 85, the sensitivity (.84) and specificity (1.00) in the present study were both higher than values reported in Taiwanese (sensitivity .66, specificity .89) and German (sensitivity .74, specificity .81) validation studies (Bolte et al., 2008; Wang et al., 2012). No other study reported use of a norm - referenced T - score cut - score of 85 and the low sensitivity of .28 for that cut - score in the present study clearly recommends against it use for general screening. Overall, raw scale cut - scores of 70 and 85, as well as a norm - referenced T - score cut - score of 85, performed well in terms of general screening sensitivity and specificity in the Ugandan context. However, of the recommended cut - scores, the raw scale cut - score of 70 appeared to balance sensitivity (.98) and specificity (.96) best in this screening context suggesting good genera lizability between the U.S. and Uganda for this recommended cut off. The 23Q. Results for the 10 ASD items from the 23Q were surprising. The recommended cut - score of one performed very poorly, as it effectively screened in or identified all children in th e sample as ASD regardless of actual group membership (i.e., sensitivity = 1.00, specificity = 0.00). Using a cut - score of one was consistent with the use of these 23Q ASD items in practice, as guidelines indicate an endorsement of any one of these items w ould lead to a positive screen (Kakooza - Mwesige et al., 2014). Clearly, the SCQ and SRS - 2 performed considerably better with their respective recommended cut - scores in the present Ugandan sample. Given that this was the first study to use the ASD items fro m the 23Q as a summation scale, there were no prior studies available for direct comparison. However, Kakooza - Mwesige et al. (2014) reported sensitivities of .52 - .57 and specificity of .92 in general screening for ASD 96 in Uganda using the criterion of one o r more ASD items being endorsed taken as a positive screen. It is not clear why these ASD items appeared to perform better in this prior study. In that study, there was a very large number of non - ASD cases and only a small number of ASD cases, the items we re not aggregated, and it is not clear exactly how the positive screens for non - ASD and neurodevelopmental disabilities were dealt with in the sensitivity and specificity calculations. In the next section, concerning alternative optimal cut - scores, the sc reening effectiveness of other possible cut - scores were explored for the 23Q ASD items -- with better results in the present study sample. Overall, recommended cut - scores for the SCQ (raw score of 15) and SRS - 2 (raw score of 70) performed well in the prese - score of one for the summation scale consisting of the 10 ASD 23Q items performed very poorly, as it screened in all cases in the sample regardless of actual ASD or non - ASD group status. Other, potentially mo re optimal, cut - scores were explored in the next section. However, the recommended cut - scores of 15 for the SCQ and 70 for the SRS - 2 appear to work well in the Ugandan context at least for purposes of differentiating ASD from more typically - developing non - ASD cases. Alternative Optimal Cut - scores Given that the recommended cut - scores were established in another cultural context, it makes good sense to examine the screening efficiency of other possible cut - scores in the Ugandan context. Thus, ROC analyses were conducted for each of the three instruments. These analyses suggested one or more potentially more optimal cut - score(s) for each instrument in the Ugandan sample and the findings are discussed below. 97 The SCQ. As indicated above, the SCQ cut - score of 15, recommended in the manual, demonstrated excellent sensitivity and specificity (i.e., .96 and .95, respectively) in the present study sample . However, a review of the SCQ ROC coordinates indicated that, in this Ugandan sample, a revised cut - score of 10 identified all ASD children and had a specificity of .93. Although 10 as a cut - score identified all children with ASD, the specificity dropped slightly -- but to an acceptable degree relative to the sensitivity gain (virtually eliminating false negatives for ASD). When compared to other studies in which cut - scores lower than 15 were proposed (e.g., cut - score of 12 in the Chinese and Arabic validation studies [Aldosari et al., 2013; Gau et al., 2011] and 14.5 for a Turkish study [Avcil et al., 201 5]), the sensitivity and specificity found among Ugandan children for the alternative cut - score of 10 were higher. The finding of a slightly improved screening performance by a cut - score that is five points lower than the standard recommendation is very su ggestive. However, the relatively small sample size of the present study should be considered and makes these findings more tentative. Because of this, SCQ cut - scores between 10 and 15 should be examined in further Ugandan studies with larger samples. The SRS - 2. For the SRS - 2, lowering the cut - score from 70 to 62 (raw score) enabled all children with ASD in the present study sample to be identified (i.e., sensitivity of 1.0) with the same specificity as attained when utilizing both the raw and standardized cut - score of 70 (i.e., .96). It is noteworthy that this cut - score of 62 also performed well in a clinical study (ASD vs. unaffected siblings) reported in the manual for the SRS - 2 (sensitivity and specificity of .92; Constantino & Gruber, 2012, p. 50). Fur thermore, a raw cut - score of 62 was recommended based on results (sensitivity of .93, specificity of .98) in a Vietnamese study (Nguyen et al., 2019). Compared to both other studies (from the test manual and from Vietnam), the sensitivity of the 98 cut - score of 62 in the present Ugandan sample was higher. However, specificity in the Ugandan sample (.96) fell between the values reported in the clinical study (.93) from the SRS - 2 manual and the study in Vietnam (.98) -- suggesting a generally similar range of spec ificity across the three studies that used this cut - score. These findings suggest that the more optimal cut - score of 62 found in the present Ugandan study sample may have some generalizability across cultural contexts. However, these findings should not be over interpreted, as most SRS - 2 validation studies from other countries that were reviewed for this purpose, proposed different cut - scores than those recommended in the SRS - 2 manual. Cut - score recommendations ranged from 43 for children in Germany (Cholem kery et al., 2014) to 87 among children in Taiwan (Wang et al., 2012). Only Bolte and colleagues (2008) found a cut - score of 100 for clinical groups in a German validation study. Despite their recommendation, the cut - score of 100 had a low sensitivity of ( i.e., .56) when compared to that of the present study (i.e., 1.0) with a cut - score of 62. However, specificities from both studies were in excellent ranges, that is, .90 for the German validation and .96 in the present Ugandan study though resulting from v ery different cut - scores. The 23Q. The 23Q cut - score of 3 or more yielded a sensitivity and specificity of .77 and .68 respectively. When compared to the recommended cut - score of 1 with .96 and 0.0 for sensitivity and specificity respectively, the cut - sco re of 3 provided a much better balance of these screening indices. (Using a cut - score of 4 would improve specificity to .80 but would reduce sensitivity to .55 -- meaning 80% of those without ASD would be accurately screened out while 45% [almost half of par ticipants with ASD] would be missed by the screener.) Kakooza - Mwesige et al. (2014) reported sensitivities of .52 - .57 and specificity of .92 for the 10 ASD items ( > 1 item endorsed = positive screen) in a general sample from Uganda with only a small 99 number of ASD cases. In the present study, using the cut - score of one was not useful at all for screening, while a cut - score of 3 yielded greater sensitivity and poorer specificity than Kakooza - Mwesige et al. (2014) who counted any of the 10 ASD items being endo rsed as a positive screen. It is possibly noteworthy that the age range of participants differed across the two studies (two to nine years old in Kakooza - Mwesige et al., and four to 18 years old in the present study). This age difference may have impacted findings. However, it is not fully satisfying as an explanation, because the cut - score of one performed so poorly in the present study despite the inclusion of many children from within the same age range as Kakooza - Mwesige et al. (e.g., the 0.0 specificit y in the present study vs. the .92 specificity in the Kakooza - Mwesige et al. study cannot be explained purely on the basis of the age range difference when the same cut - score of one or more was used). Regardless of why the results were so different across the two studies, in a large screening situation (i.e., screening hundreds or thousands of cases), the result in the Kakooza - Mwesige et al. study suggests that an unacceptable percentage of those with ASD would be missed by the screener, while the results of the present study suggest that the optimum cut - score would likely result in too many false positives in the general population. These ASD items should be considered for revision and re - examined in further studies. Overall, ROC analyses revealed that c ut - scores of 62 on the SRS - 2 and 10 on the SCQ may improve their screening effectiveness in the Ugandan context. These cut - scores performed slightly better, in the present Ugandan sample, than the recommended cut - scores of 70 (SRS - 2) and 15 (SCQ). However, it is noteworthy that the recommended cut - scores also performed well (at levels close to the more optimal cut - scores) and the new cut - scores derived from the present sample data may require replication in larger and broader samples, to assure stability an d generalizability, before being recommended for more general use in Uganda. Results from ROC 100 analyses of the 10 ASD items from the 23Q suggested that, in the present sample, even best compromise cut - score of 3 (sensitivity .77, specificity .68) would stil l fail to screen out sufficient cases without ASD to be considered an efficient screener -- resulting in too many false positive cases that would require follow - up evaluations. Therefore, the SRS - 2 and SCQ, modified for the Ugandan context, appear appropriat e for use. However, the ASD items from the 23Q are more problematic and likely require revision to improve sufficiently for regular use. Correlations Among Study Instruments Convergent validity among the three instruments was examin ed via Pearson correlations between their overall composite scores. The positive correlation between the SCQ and SRS - 2 was very high (i.e., r = .93, p < 0.001), while the correlations between the sum of the 10 ASD items from the 23Q and the other two instruments were low and non - significant ( r = .19, p = .053 with SCQ; r = .17, p = .082 with SRS - 2). However, when the correlations were run again using ju st items 19 - 23 from the 23Q, the correlations involving the 23Q improved to moderate and significant (i.e., r = .35, p = 0.001 with the SRS - 2 and r = .36, p < .001 with the SCQ). The convergent validity issues found here for the sum of the 10 ASD items fro m the 23Q are further evidence of the need to closely re - examine and revise these items. Bolte and colleagues (2008), in a German validation study, were among the few authors who reported a correlation between the SCQ and SRS total scores (i.e., r = .58). These authors validation studies reviewed for this purpose administered only one of the instruments (i.e., the SRS or SCQ), in addition to other instruments, bu t not both. The correlation between the SRS - 2 and SCQ for this study was considerably stronger than that reported by Bolte and colleagues, which could be due to sample differences (e.g., inclusion of both ASD and non - ASD cases in the 101 present sample, proban d - only sample in the Bolte et al. study). Additional estimates of this relationship would be helpful in future samples to provide more context and suggest possible moderating conditions that may impact the magnitude of the correlation. This study is t he first to report correlations between 23Q summed ASD item sets and composites from the SRS - 2 and SCQ. As indicated above, the low convergent correlations involving the 23Q ASD items suggest the need to re - examine these items with an eye toward revision a nd further research. Strengths and Limitations The present study involved a number of strengths, which highlight its importance and quality of its findings. First and foremost, this is the first study to examine the validity of standardized western ins truments, such as the SCQ and SRS - 2, for ASD screening in Uganda. The need for high quality screening instruments in Uganda is high, as available local options are limited. In addition, ASD - related items from the 23Q were also evaluated and compared to the SCQ and SRS - 2 in the present study. This was important, because this survey screening instrument was previously used in Uganda. However, its performance in the present study was much poorer and more problematic when compared to the SCQ and SRS - 2 especiall y in terms of screening sensitivity and specificity, and convergent validity. Second, the simultaneous psychometric evaluation of three different screening instruments within the sample was a clear strength. This allowed for assessment of all instruments in terms of internal consistency, discrimination between ASD and non - ASD cases according to both recommended and sample - specific cut - scores, and convergent validity (i.e., correlations among the three instruments). Evaluating them simultaneously in the same sample 102 allowed for direct comparison of the instruments in the same context, with the same cases, and using the same criteria. Third, permission from the copyright holder of the SCQ and SRS - 2 (Western identify items with potentially confusing cultural - laden content and, when necessary, to adapt the content to fit the cultural context in order to maximize appropriateness, relevance, and rater comprehension. Items were subject to review by consultants in Uganda (in consultation with the author a nd feedback from the test company experts), modified for enhanced cultural relevance with suggestions from the consultants, and piloted with feedback from a group of potential participants. This feedback was used to identify potentially confusing items, co nsider further modification and/or to craft standardized, brief clarifications that the interviewer could use when requested. This was all done prior to recruitment of actual participants and execution of the study. Fourth, children across a relativ ely wide age range were included in the study (i.e., ranging in age from four years to 18 years old). This encompasses the full age range for the SRS - 2 school - age form and falls within ages appropriate for the SCQ. The 23Q was previously examined in Uganda with children ages two to nine years old (Kakooza - Mwesige et al., 2014). This study partially overlaps with and expands this age range. Fifth, trained clinical psychologists with experience working with children who have ASD were recruited to adminis ter the study instruments to parent participants. Such experience and training made the interviewers well suited for developing rapport, providing standard item explanations when needed, and for providing expert review of diagnostic records to evaluate ASD cases for study inclusion. 103 Sixth, the use of DSM - 5 diagnosis of ASD as the criterion for distinguishing between ASD and non - the ADI - R and ADOS - 2, are not readily available in Uganda. However, the DSM - 5 is used consistently in major hospitals and other diagnostic settings within the country and assuredly within the recruitment area for the study. Records were subject to review by the clinical psychologists employed by the study in Uganda to assure that the DSM - 5 was used to diagnose cases clear use of DSM - 5 criteria were not included in the study in either group.) This provided a consistent diagn ostic standard, which was the best available within the country. Despite the strengths of this study discussed above, there were also number of limitations. First, the sample size for the study ( N = 107) was relatively small when compared to many other va lidation studies. This can limit the precision of estimates derived from the sample and estimate stability across samples. This is, in part, why the lower - cut - scores derived from the ROC analyses require cross - validation to improve confidence. Second, lac k of greater geographic diversity in the sample from across Uganda is a limitation. All study participants lived within the Kampala or Wakiso districts. Thus, generalization of results may be restricted to these regions. Third, language diversity is restr icted. English and a variety of Ugandan dialects are spoken across Uganda. The majority of people around the major cities speak English and the diversity of Ugandan dialects would limit the applicability of a single Ugandan dialect translation and adaptati on of the instruments. Thus, English was chosen as the primary language for the cultural adaptation of the instruments. 104 Fourth, it was necessary to deviate from the typical expectation of having parents read and complete the rating scale instruments on their own. The relatively high rate of illiteracy worked against this procedure. To adapt, the instructions and items from the rating scales were administered to all parent participants via interview. This kept the procedure uniform, accommodated those who se reading skills were not sufficient to validly complete the scales, and readily allowed for standardized item clarifications from the interviewer, if needed. This procedure was very appropriate for this cultural context. However, the standardized adminis tration does deviate significantly from having participants read the items themselves with the typical expectation of valid comprehension. This is a generalization issue, but one that appears necessary to allow for such instruments to be broadly and meanin gfully used within this current cultural context. Fifth, though DSM - 5 diagnostic criteria were applied across the sample to establish the - R or ADOS - 2) was used to confirm diagnosis. These instrument s are not readily available or used within Uganda. In fact, there is a need to adapt such instruments for use in Uganda, as no validated adaptations of these instruments currently exist in there. Sixth, the clinical discriminant validity of these instrume nts was examined in the present study only for differentiating between ASD and generally typically - developing non - ASD groups. The differentiation between these two groups is a good start for this type of clinical discriminant validity and is generally appl icable for screening purposes. However, it is very important to understand how well these instruments discriminate between ASD and clinical groups, and especially between ASD and other types of non - ASD developmental disabilities. These types of comparisons more closely match the types of differentiations that clinicians are regularly asked 105 to make in the context of diagnosis. Other clinical groups or other types of developmental disability groups were not included in the present study. Clinical Implications The current study is a pioneering investigation of screening instruments that could be used to detect symptoms of ASD in Ugandan children. Results of this study support the contention that both the SCQ and SRS - 2 are potentially viable instrume nts needed to screen for ASD in the Ugandan context. However, further research studies are needed to confirm and expand upon these psychometric findings before full confidence in the use of these instruments is warranted. With these preliminary data, the combined sample of ASD and non - ASD showed reasonable reliability for both the SCQ and the SRS - 2. However, the SRS - 2 was relatively more reliable when compared to the other instruments when used separately on ASD and non - ASD. Thus, professionals must be mi ndful of how the reliability may be affected by the population of children to which an instrument is applied. In addition, although the 10 ASD items of the 23Q yielded low reliability as a scale, the five items originally intended only for younger children yielded considerably higher internal consistency reliability as a scale when applied across a broader range of ages. This suggests the possibility of improving the instrument for applied screening by including more developmental history items. However, th e current 23Q ASD items are not recommended for use in clinical contexts. When used in differentiating ASD from non - ASD samples among the Ugandan children in this study, the SCQ and the SRS - 2 showed considerably larger effect sizes compared to the 23Q, i ndicating that both the SRS - 2 and the SCQ have stronger discriminant validity for differentiating between cases with ASD and typically - developing non - ASD cases a distinction 106 similar to broad screening in the general population. Results of the ROC analyses showed that a lower cut - score of 62 enhanced the sensitivity and specificity of the SRS - 2 in this sample, but that the recommended screening cut - score of 70 was also acceptable. In addition, a lower cut - score of 10 of the SCQ performed better in this study -- although the recommended cut - score of 15 was also acceptable. In general, the two recommended cut - scores appear to be safer to apply in practice until the lower cut - scores are either supported or refuted by replication studies and studies involving other types of groups. It is very likely that a different cut - score may be more useful for different purposes (i.e., different differentiation needs). Further research in the Ugandan context is needed to clarify these possible situational variations. Resear ch Implications and Future Research Directions Up until now, no formal validation studies have been done to examine the psychometric properties of established instruments from the U.S. and Europe to screen for children for ASD symptoms in Uganda. Thus, fin dings from this study provide a foundation to explore viable instruments, i.e., the SRS - 2 and the SCQ, that are well validated in the U.S. and that may be used in the Ugandan culture with some modifications. The screening utility of the 23Q, an instrument that was originally developed for the African culture to screen in the general population for children with developmental disabilities, including 10 items related to ASD, was further explored. As mentioned above, instruments imported from another culture, like the SRS - 2 and SCQ, may require modifications of items and/or utilization in order to reliably and accurately screen children with ASD in the Ugandan culture. However, as addressed in the limitations section, a combination of larger sample size, broade ning the cross - group comparisons to include other clinical groups, and perhaps expanding investigations of important culture - specific 107 elements can provide further information to improve these screening instruments for Ugandan children with ASD. Though thi s study showed that the SRS - 2 and the SCQ are potentially viable instruments for use in the Ugandan context, it is the first study to be conducted. Cross - validation of ROC findings to better understand the generality of revised cut - scores in the Ugandan co ntext is particularly important and this will be further enhanced by the incorporation of other relevant clinical groups in these studies for comparisons that reflect the kinds of differential diagnostic distinctions that clinicians most frequently face. S uch differential diagnoses may involve such distinctions as ASD vs ADHD, intellectual disability, speech/language disorders, anxiety disorders, etc. The robustness of an instrument for making such distinctions is important to understand and can suggest ave nues for instrument improvement across successive revisions. Furthermore, the 10 ASD items from the 23Q provided a foundation for ASD screening in Africa, the present study suggested significant weaknesses in the present item set that imply a need for revi sion and further evaluation of items to improve the reliability and screening effectiveness of the instrument. The issue of sample size is important for obtaining more stable estimates involving less sampling error. All other things being equal, larger s amples in psychometric studies instill more confidence in the stability of the findings across similar samples and contexts. Thus, future validation studies for the SRS - 2 and SCQ in Uganda should strive for significantly larger samples for cross - validation purposes. Another important issue is the broadening of sample representation to reflect a broader range of population characteristics (e.g., geographic regions, socioeconomic range, other settings [e.g., non - hospital], etc.) in order to better understan d the generality of the findings. In addition, 108 better sample representation can improve the accuracy of population estimates. Ultimately, larger and more representative sample reflect appropriate movement toward the longer - term goal of instrument norming. It will ultimately be helpful to establish Ugandan norms for instruments like the SRS - 2 in order to allow for standardized reference scores to be used to communicate results in clinical and educational settings. However, this type of activity should follow the proper establishment of the relevance and validity of the instrument in Uganda for screening, clinical, and educational purposes. Though the sample in the present study is generally well characterized demographically, to the extent it is possible, obt aining cognitive and language scores, as well as implementing a stronger autism diagnostic approach would enhance the quality of studies in this context. Potentially viable cognitive and language instruments exist with some validity for this context, but f - R or ADOS - 2) may not be reasonable until similar studies are performed to establish the generalization of ADI - R and ADOS - 2 diagnostic validity in the Ugandan context. In other wo rds, establishing the validity of rating scale instruments, such as the SRS - 2 and SCQ, for ASD screening, may be first steps toward validating a more comprehensive ASD diagnostic instrument for use in Uganda. Further explorations of the correlations across different ASD screening instruments, between screening instrument results and DSM - 5 diagnosis, and between screening instruments and established diagnostic instruments (e.g., ADI - R and ADOS - 2, which may also require validation work in Uganda before their wider use becomes an option) are important for understanding the larger scope of construct validity for an instrument or the multiple instruments and techniques involved. In addition, it may be fruitful to explore how multiple screening instruments m 109 strengths and weaknesses, assess differential weighting options with the goal of improving the accuracy of screening decisions beyond what is possible for a single measure, etc.). The SRS - 2 and SCQ were designed to be completed by a primary caregiver or teacher familiar with the child being assessed. In this study, a trained researcher read the items to caregivers of the child as opposed to self - administration by the careg iver or teacher. Therefore, the potential differences in the administration method should be further examined. This may be possible only within samples of literate adults, which may not generalize as well to adults with lower literacy skills, as such skill s may be correlated with other important variables. However, the findings from literate samples wherein the different approaches to administration are compared can be very informative. Finally, only a primary caregiver of a child with ASD completed the instruments in the current study. Having multiple informants across different settings that can rate the children, e.g., teachers at school can improve the generalizability of the results across different settings. In addition, having multiple raters who know the child well rate the child high on an instrument like the SRS - 2 statistically increases the likelihood of a confirmed ASD diagnosis later (Constantino & Gruber, 2012). Thus, validity data for multiple rater types can contribute meaningfully to the screening utility of the measure. Conclusions Results of the present study lend initial support for the viability of SRS - 2 and SCQ versions culturally adapted for use in Uganda. These preliminary results support the screening effectiveness of the SRS - 2 and the SCQ for distinguishing between cases with ASD and those more typically - developing non - ASD cases. This situation is similar to that which occurs in broader general population screening. However, both replication in larger, more representative 110 samples and discriminative results for ASD vs other cli nical groups are essential for understanding the robustness of these initial findings and the broader applicability of these instruments. Based on the results for the 10 ASD items from the 23Q, this measure is not recommended for critical screening situati ons. Though the instrument was initially designed for broad screening use in Uganda, the sensitivity and specificity results, in particular, from the present study indicate that its recommended cut - score performed very poorly. Furthermore, though a more op timal cut - score found through ROC analyses indicated a composite consisting of these 10 23Q ASD items can accurately screen in approximately 3/4s of those cases with ASD, even when using this improved cut - score it will also miss - classify almost 1/3 of non - ASD cases as ASD. The screening effectiveness of the SRS - 2 and SCQ was considerably better within the same sample. It is recommended that the 23Q ASD items be re - examined and revised going forward to improve its screening effectiveness. 111 APPENDICES 112 APPENDIX A : Table 21 Table 21. Reliability of SCQ Total Scores Among Non - Item Item - Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item - Total Correlation Cronbach's Alpha if Item Deleted 2 To and fro 5.29 13.371 .303 .720 3 Said same 5.02 13.036 .170 .726 4 Inappropriate 4.91 13.174 .112 .732 5 She / he I 5.23 13.491 .118 .726 6 Metaphorical 5.16 13.519 .061 .731 7 Same exactly 5.13 12.948 .247 .720 8 Ritually 5.18 13.495 .078 .729 9 Facial expressions 5.2 13.397 .128 .726 10 Hand a tool 5.05 13.288 .103 .731 11 Preoccupied 5.16 13.483 .074 .730 12 Interest parts 5.21 12.862 .385 .713 13 Interests age 5.2 12.961 .312 .716 14 Sight, feel, sound, 5.23 13.636 .049 .729 15 Hand and fingers 5.25 13.427 .172 .724 16 Body movements 5.29 13.626 .116 .726 17 Attachment 5.25 13.755 .000 .731 18 Talking friendly 5.18 12.804 .354 .713 20 Spontaneous copying 5.23 13.6 .066 .729 21 Spontaneous point 4.89 12.679 .251 .720 22 Point / pull 4.73 12.345 .353 .712 23 Head yes 5.18 12.44 .505 .704 24 Shake head 5.07 12.431 .389 .709 25 Directly face 5.3 13.815 - .011 .728 26 Show 5.25 13.355 .211 .722 27 Share other than 5.16 13.446 .088 .729 28 You to join 5.3 13.706 .099 .726 29 To comfort 5.18 13.058 .250 .719 30 Gesture 5.05 12.524 .347 .712 31 Range expressions 5.25 13.245 .269 .719 32 Spontaneously 5.29 13.081 .522 .713 33 Make - believe 5.21 12.644 .487 .707 34 Interest children 5.27 13.363 .245 .721 113 Table 2 1 ( ) 35 Responds positively 5.29 13.444 .250 .721 36 Attends to voice 5.2 13.033 .281 .718 37 Imaginatively 5.2 12.561 .486 .706 38 Cooperatively 5.27 12.963 .493 .711 Note 114 APPENDIX B : Table 22 Table 22. SRS - 2 Total Scores Reliability Among Non - ASD Group if Item is Deleted Item - Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item - Total Correlation Cronbach's Alpha if Item Deleted 1 Fidget 25.85 243.015 .03 .891 2 Facial match 26 240.667 .12 .890 3 Self - confident 26 236.778 .474 .887 4 Rigid stress 25.78 238.174 .215 .889 5 Doesn't tricks 25.44 242.621 .016 .893 6 Alone 26.02 242.722 .051 .890 7 Others think 25.04 236.517 .209 .890 8 Strange 26.16 241.176 .32 .889 9 Clings 25.84 232.88 .483 .886 10 Literal 25.93 232.958 .542 .886 11 Confidence 25.64 233.384 .386 .887 12 Feelings 25.62 231.833 .424 .887 13 Awkward turn - taking 25.89 230.543 .492 .886 14 Coordinated 26.04 233.184 .608 .885 15 Tone voice 25.84 234.176 .461 .886 16 Eye contact 26.05 237.978 .385 .888 17 Unfairness 25.55 230.141 .513 .885 115 Table 2 2 18 Difficulty friends 25.96 231.332 .610 .885 19 Gets frustrated 25.93 230.884 .595 .885 20 Unusual interests 26.11 238.877 .324 .888 21 Imitate others 25.6 231.985 .409 .887 22 Plays appropriately 25.96 233.406 .481 .886 23 Does not join 25.76 245.11 - .082 .892 24 Difficult routine 26.07 237.809 .409 .888 25 Out step 26.11 241.284 .201 .889 26 Comfort sad 25.53 229.772 .531 .885 27 Avoid social 25.8 235.385 .379 .887 28 Same over and over 26 237.296 .374 .888 29 Odd 26.07 235.328 .539 .886 30 Busy place 26.07 240.772 .254 .889 31 Mind off 25.78 241.137 .134 .890 32 Hygiene 25.64 237.051 .288 .888 33 Socially awkward 26.18 240.559 .456 .888 34 Avoids close 25.95 244.423 - 0.053 .891 35 Following conversation 25.98 230.092 0.652 .884 36 Difficult adults 25.85 239.83 0.2 .889 37 Difficulty relating 25.91 227.899 0.708 .883 38 Appropriate mood 25.49 230.44 0.469 .886 116 39 Limited range of interests 26.04 236.554 0.436 .887 40 Is imaginative 25 227 0.496 .885 41 Wanders aim 25.76 235.369 0.348 .888 42 Sounds 25.95 242.978 0.028 .891 43 Separates easily 25.2 235.163 0.193 .892 44 Understand events 25.64 232.828 0.365 .888 45 Attention away 25.69 239.81 0.176 .890 46 Serious expressions 25.64 243.088 0.003 .892 47 Too silly 26.16 237.917 0.581 .887 48 Jokes 25.42 231.803 0.388 .887 49 Well at few 25.45 236.29 0.252 .889 50 Flap 26.22 242.211 0.386 .889 51 Around subject 25.96 234.962 0.446 .887 52 Loud or noisy 25.4 228.393 0.461 .886 53 Tone voice 26.15 238.978 0.342 .888 54 People objects 26.18 241.263 0.356 .889 55 Knows close 25.09 229.677 0.414 .887 56 Between people 25.78 242.618 0.03 .891 57 Gets teased a lot 25.98 235.685 0.436 .887 58 Much on parts 26.02 239.981 0.217 .889 59 Suspicious 25.98 239.87 0.232 .889 117 60 Emotionally distant 26.00 244.185 - 0.037 .891 61 Inflexible 25.82 235.04 0.38 .887 62 Illogical for doing 25.91 239.899 0.169 .890 63 Others unusually 26.15 241.423 0.26 .889 64 Relax 26 239.148 0.245 .889 65 Space 26.11 241.284 0.235 .889 Note 118 APPENDIX C : I RB Clearance i n Uganda 119 120 APPENDIX D : MSU IRB C learance 121 APPENDIX E : Permission t o Conduct Research a 122 APPENDIX F : Permission t o Conduct Research w ith Children a t Butabika National Hospital, Uganda 123 APPENDIX G : Fidelity Checklist 124 APPENDIX H: A ssent Form (8 - 17 years ) 125 126 127 128 129 APPENDIX I : C onsent Form 130 131 132 133 APPENDIX J : D emographic Form 134 135 136 APPENDIX K : 23 Questions Questionnaire 137 138 R EFERENCES 139 REFERENCES Abubakar, A., Ssewanyana, D., & Newton, C. R. (2016). A Systematic Review of Research on Autism Spectrum Disorders in Sub - Saharan Africa. Behavioural Neurology , 2016 , 1 14. https://doi.org/10.1155/2016/3501910 Akoury - Dirani, L., Alameddine, M. , & Salamoun, M. (2013). Validation of the Lebanese Childhood Autism Rating Scale Second Edition High Functioning Version. Research in Autism Spectrum Disorders , 7 (11), 1332 1338. https://doi.org/10.1016/j.rasd.2013.08.001 Aldosari, M., Fombonne, E., Aldhalaan, H., Ouda, M., Elhag, S., Alshammari, H., Ghazal, I., Alsaleh, A., Alqadoumi, T., Thomson, R., Al Khasawneh, M., Tolefat, M., & Alshaban, F. (2019). Validation of the Arabic version of the Social Communication Questionnaire. Autism , 23 (7), 1655 1 662. https://doi.org/10.1177/1362361318816065 American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Arlington, VA: American Psychiatric Publishing. Avcil, S., Baykara, B., Baydur, H., Munir, K. M., & Inal Emiroglu, N. (2014). The Validity and Reliability of the Turkish Version of the Social Communication Questionnaire. Turkish Journal of Psychiatry . https://doi.org/10.5080/u7298 Bakare, M., & Munir, K. (2011). Autism spectrum disorders (ASD) in Africa: a perspective. African Journal of Psychiatry , 14 (3). https://doi.org/10.4314/ajpsy.v14i3.3 Baron - Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (n.d.). The Autism - Spectrum Quotient (AQ): Evidence from Asperger Syndrome/High - Functioning Autism, Malesand Females, Scientists and Mathematicians . 14. Bello - Mojeed, M. A., Omigbodun, O. O., Bakare, M. O., & Adewuya, A. O. (2017). Pattern of impairments and late diagnosis of autism spectrum disorder among a sub - Saharan African clinical population of children in Nigeria. Global Mental Health , 4 . https://doi.org/10.1017/gmh.2016.30 Bernal, G. E., & Domenech Rodríguez, M. M. (2012). Cultural adaptations: Tools for evidence - based practice with diverse populations (pp. xix - 307). American Psychological Association. Berument, S. K., Rutter, M., Lord, C., Pickles, A., & Bailey, A. (1999). Autism screening questionnaire: diagnostic validity. The British Journal of Psychiatry , 175 (5), 444 451. 140 Bölte, S., Poustka, F., & Constantino, J. N. (2008). Assessing autistic trait s: cross - cultural validation of the social responsiveness scale (SRS). Autism Research , 1 (6), 354 363. https://doi.org/10.1002/aur.49 Bölte, S., Westerwald, E., Holtmann, M., Freitag, C., & Poustka, F. (2011). Autistic Traits and Autism Spectrum Disorders : The Clinical Validity of Two Measures Presuming a Continuum of Social Communication Skills. Journal of Autism and Developmental Disorders , 41 (1), 66 72. https://doi.org/10.1007/s10803 - 010 - 1024 - 9 Bryson, S. E., Rogers, S. J., & Fombonne, E. (2003). Autis m Spectrum Disorders: Early Detection, Intervention, Education, and Psychopharmacological Management. The Canadian Journal of Psychiatry , 48 (8), 506 516. https://doi.org/10.1177/070674370304800802 Cen, C. - Q., Liang, Y. - Y., Chen, Q. - R., Chen, K. - Y., Deng, H. - Z., Chen, B. - Y., & Zou, X. - B. (2017). Investigating the validation of the Chinese Mandarin version of the Social Responsiveness Scale in a Mainland China child population. BMC Psychiatry , 17 (1), 51. https://doi.org/10.1186/s12888 - 016 - 1185 - y Chlebowski, C., Green, J. A., Barton, M. L., & Fein, D. (2010). Using the Childhood Autism Rating Scale to Diagnose Autism Spectrum Disorders. Journal of Autism and Developmental Disorders , 40 (7), 787 799. https:/ /doi.org/10.1007/s10803 - 009 - 0926 - x Chojnicka, I., & Pisula, E. (2019). Cross - Cultural Validation of the Polish Version of the ADI - R, Including New Algorithms for Toddlers and Young Preschoolers. Child Psychiatry & Human Development . https://doi.org/10.100 7/s10578 - 018 - 00865 - 2 Cholemkery, H., Kitzerow, J., Rohrmann, S., & Freitag, C. M. (2014). Validity of the social responsiveness scale to differentiate between autism spectrum disorders and disruptive behaviour disorders. European Child & Adolescent Psychiatry , 23 (2), 81 93. ht tps://doi.org/10.1007/s00787 - 013 - 0427 - 5 Corsello, C., Hus, V., Pickles, A., Risi, S., Cook, E. H., Leventhal, B. L., & Lord, C. (2007). Between a ROC and a hard place: decision making and making decisions about using the SCQ. Journal of Child Psychology a nd Psychiatry , 48 (9), 932 940. https://doi.org/10.1111/j.1469 - 7610.2007.01762.x Contantino, J. N., & Gruber, C. P. (2005). Social Responsiveness Scale (SRS) [Manual]. Los Angeles, CA: Western Psychological Services. Constantino, J. N., & Gruber. C. P. (2012). Social Responsiveness Scale, Second Edition (SRS - 2) [Manual]. Torrance, CA: Western Psychological Services. Randomized, Controlled Trial of an Interventi on for Toddlers With Autism: The Early Start Denver Model. Pediatrics , 125 (1), e17 e23. https://doi.org/10.1542/peds.2009 - 0958 141 Durkin, M. (1995). Evaluating a ten questions screen for childhood disability: Reliability and internal structure in different cu ltures. Journal of Clinical Epidemiology , 48 (5), 657 666. https://doi.org/10.1016/0895 - 4356(94)00163 - K A. (2015). Autism screening and diagnosis in low resource settings: Challenges and opportunities to enhance research and services worldwide: Enhancing Autism Research and Services Worldwide. Autism Research , 8 (5), 473 476. https://doi.org/10.1002/aur.1575 Fombonne, E., Marcin, C., Bruno, R., Tinoco, C. M., & Mar quez, C. D. (2012). Screening for Autism in Mexico: Screening for autism in Mexico. Autism Research , 5 (3), 180 189. https://doi.org/10.1002/aur.1235 Franz, L., Chambers, N., von Isenburg, M., & de Vries, P. J. (2017). Autism spectrum disorder in sub - sahar an africa: A comprehensive scoping review: Autism spectrum disorder in sub - Saharan Africa. Autism Research , 10 (5), 723 749. https://doi.org/10.1002/aur.1766 Gau, S. S. - F., Lee, C. - M., Lai, M. - C., Chiu, Y. - N., Huang, Y. - F., Kao, J. - D., & Wu, Y. - Y. (2011). Psychometric properties of the Chinese version of the Social Communication Questionnaire. Research in Autism Spectrum Disorders , 5 (2), 809 818. https://doi.org/10.1016/j.rasd.2010.09.010 Gilliam, J. E. (2014). Gilliam Autism Rating Scale (3rd ed.). Austin, TX: Pro - Ed. Gona, J. K., Newton, C. R., Rimba, K., Mapenzi, R., Kihara, M., Van de Vijver, F. J. R., & Abubakar, A. (2015). Options for Autism Spectrum Disorders (ASD) in a Multicultu ral Context on the Kenyan Coast. PLOS ONE , 10 (8), e0132729. https://doi.org/10.1371/journal.pone.0132729 Harrison, A. J., Zimak, E. H., Sheinkopf, S. J., Manji, K. P., & Morrow, E. M. (2014). Observation - centered Approach to ASD Assessment in Tanzania. In tellectual and Developmental Disabilities , 52 (5), 330 347. https://doi.org/10.1352/1934 - 9556 - 52.5.330 Jackson, L. S., Little, S. G., & Akin - Little, A. (2013). The Spanish adaptation of the Gilliam Autism Rating Scale - 2: Translation and psychometric analys is. Research in Autism Spectrum Disorders , 7 (9), 1160 1167. https://doi.org/10.1016/j.rasd.2013.06.005 Kakkar, J., & Srivastava, P. (n.d.). Challenges and Coping among Parents having Children with Autism Spectrum Disorder . 10 Kakooza - neurodevelopmental disorders in Uganda. Autism , 18 (4), 447 457. https://doi.o rg/10.1177/1362361313475848 142 Kamp - Stroth, S. (2018). Diagnostic accuracy of the ADOS and ADOS - 2 in clinical practice. European Child & Adolescent Psychiatry , 27 (9), 1193 1207. https://doi.org/10.1007/s00787 - 018 - 1143 - y Karren, B. C. (2017). A Test Review: Gilliam, J. E. (2014). Gilliam Autism Rating Scale Third Edition (GARS - 3) . Journal of Psychoeducational Assessment , 35 (3), 342 346. https://doi.org/10.1177/07342829166354 65 Klin, A., & Volkmar, F. R. (2005). Editorial Preface. Journal of Autism and Developmental Disorders , 35 (2), 141 143. https://doi.org/10.1007/s10803 - 004 - 1991 - 9 Lai, M. - C. (2015). Sex/Gender Differences and Autism: Setting the Scene for Future Research. ADOLESCENT PSYCHIATRY , 54 (1), 14. Lecavalier, L. (2005). An Evaluation of the Gilliam Autism Rating Scale. Journal of Autism and Developmental Disorders , 35 (6), 795 805. https://doi.org/10.1007/s10803 - 005 - 0025 - 6 Lee, G. K., Lopata, C., Volker, M. A., Th Smerbeck, A. M. (2009). Health - Related Quality of Life of Parents of Children With High - Functioning Autism Spectrum Disorders. Focus on Autism and Other Developmental Disabilities , 24 (4), 227 239. https://doi.org/10.1177/1088357609347371 (2000). The Autism Diagnostic Observation Schedule Generic: A Standard Measure of Social and Communication Deficits As sociated with the Spectrum of Autism. Journal of Autism and Developmental Disorders , 30 (3), 205 223. https://doi.org/10.1023/A:1005592401947 Lord, C., Rutter, M., & Le Couteur, A. (1994). Autism Diagnostic Interview - Revised: a revised version of a diagnos tic interview for caregivers of individuals with possible pervasive developmental disorders. Journal of Autism and Developmental Disorders , 24 (5), 659 685. Lord C, Rutter M, Dilavore PC, Risi S, Gotham K, Bishop SL (2012). Autism Diagnostic Observation Sc hedule, second edition (ADOS - 2) manual (part 1) modules 1 - 4. Western Psychological Services, Torrance, CA Maenner, M. J., Shaw, K. A., Baio, J., EdS1, Washington, A., Patrick, M., DiRienzo, M., Christensen, D. L., Wiggins, L. D., Pettygrove, S., Andrews, J. G., Lopez, M., Hudson, A., Baroud, T., Schwenk, Y., White, T., Rosenberg, C. R., Lee, L. C., Harrington, R. A., Children Aged 8 Years - Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2016. Morbidity and mortality weekly report. Surveillance summaries (Washington, D.C. : 2002) , 69 (4), 1 12. https://doi.org/10.15585/mmwr.ss6904a1 143 Magaña, S., & Smith, L. E. (2013). The Use of the Autism Diagno stic Interview - Revised with a Latino Population of Adolescents and Adults with Autism. Journal of Autism and Developmental Disorders , 43 (5), 1098 1105. https://doi.org/10.1007/s10803 - 012 - 1652 - 3 Magaña, S., & Vanegas, S. B. (2017). Diagnostic Utility of th e ADI - R and DSM - 5 in the Assessment of Latino Children and Adolescents. Journal of Autism and Developmental Disorders , 47 (5), 1278 1287. https://doi.org/10.1007/s10803 - 017 - 3043 - 2 Marlow, M., Servili, C., & Tomlinson, M. (2019). A review of screening tools for the identification of autism spectrum disorders and developmental delay in infants and young children: recommendations for use in low - and middle - income countries: Marlow et a l./A review of screening tools for autism and developmental delay. Autism Research , 12 (2), 176 199. https://doi.org/10.1002/aur.2033 Mazurek, M. O., Curran, A., Burnette, C., & Sohl, K. (2019). ECHO Autism STAT: Accelerating Early Access to Autism Diagnos is. Journal of Autism and Developmental Disorders , 49 (1), 127 137. https://doi.org/10.1007/s10803 - 018 - 3696 - 5 Note. Journal of Child Psychology and Psychiatry , 38 ( 2), 247 251. https://doi.org/10.1111/j.1469 - 7610.1997.tb02354.x - J. (2011). Medical schools in sub - Saharan Africa. The Lancet , 377 (9771), 1113 1121. https://doi.org/10.10 16/S0140 - 6736(10)61961 - 7 Nguyen, P. H., Ocansey, M. E., Miller, M., Le, D. T. K., Schmidt, R. J., & Prado, E. L. (2019). The reliability and validity of the social responsiveness scale to measure autism symptomology in Vietnamese children. Autism Research , 12(11), 1706 1718. https://doi.org/10.1002/aur.2179 Nordin, V., Gillberg, C., & Nyden, A. (n.d.). The Swedish Version of the Childhood Autism Rating Scale in a Clinical Setting . 7. Norris, M., & Lecavalier, L. (2010). Screening Accuracy of Level 2 Auti sm Spectrum Disorder Rating Scales: A Review of Selected Instruments. Autism , 14 (4), 263 284. https://doi.org/10.1177/1362361309348071 Presmanes Hill, A., Zuckerman, K., & Fombonne, E. (2015). Epidemiology of Autism Spectrum Disorders. In M. de los A. Rob inson - Agramonte (Ed.), Translational Approaches to Autism Spectrum Disorder (pp. 13 38). Retrieved from http://link.springer.com/10.1007/978 - 3 - 319 - 16321 - 5_2 Rutter, M., Bailey, A., & Lord, C, (2003). The Social Communication Questionnaire (SCQ) [Manual]. Torrance, CA: Western Psychological Services. 144 Schopler, E., Reichler, R. J., DeVellis, R. F., & Daly, K. (1980). Toward objective classification of childhood autism: Childhood Autism Rating Scale (CARS). Journal of Autism and Developmental Disorders , 10 (1 ), 91 103. https://doi.org/10.1007/BF02408436 Smith, L., Malcolm - Smith, S., & de Vries, P. J. (2017). Translation and cultural appropriateness of the Autism Diagnostic Observation Schedule - 2 in Afrikaans. Autism , 21 (5), 552 563. https://doi.org/10.1177/13 62361316648469 Ozonoff, S. (2002). Utility of the Gilliam Autism Rating Scale in Research and Clinical Populations . 7. Sturm, A., Kuhfeld, M., Kasari, C., & McCracken, J. T. (2017). Development and validation of an item response theory - based Social Responsiveness Scale short form. Journal of Child Psychology and Psychiatry , 58 (9), 1053 1061. https://doi.org/10.1111/jcpp.12731 Tehrani - Doost, M., Shahrivar, Z., Torabi, N ., Ansari, S., Haji - Esmaeelzadeh, M., & Saeed - Ahmadi, S. (2018). Cross - Cultural Validation and Normative Data of the Social Responsiveness Scale in a Group of Iranian General Child Population. Journal of Autism and Developmental Disorders . https://doi.org/ 10.1007/s10803 - 018 - 3773 - 9 (2013). Reliability and Validity of Autism Diagnostic Interview - Revised, Japanese Version. Journal of Autism and Developmental Disorders , 43 (3), 643 662. https://doi.org/10.1007/s10803 - 012 - 1606 - 9 Vanegas, S. B., Magaña, S., Morales, M., & McNamara, E. (2016). Clinical Validity of the ADI - R in a US - Based Latino Population. Journal of Autism and Developmental Disorders , 46 (5), 1623 1635. https://doi.org/10.1 007/s10803 - 015 - 2690 - 4 Vaughan, C. A. (2011). Test Review: E. Schopler, M. E. Van Bourgondien, G. J. Wellman, & S. R. Love Childhood Autism Rating Scale (2nd ed.). Los Angeles, CA: Western Psychological Services, 2010. Journal of Psychoeducational Assessme nt , 29 (5), 489 493. https://doi.org/10.1177/0734282911400873 Volker, M. A., & Lopata, C. (2008). Autism: A review of biological bases, assessment, and intervention. School Psychology Quarterly , 23 (2), 258 270. https://doi.org/10.1037/1045 - 3830.23.2.258 Wang, J., Lee, L. - C., Chen, Y. - S., & Hsu, J. - W. (2012). Assessing Autistic Traits in a Taiwan Preschool Population: Cross - Cultural Validation of the Social Responsiveness Scale (SRS ). Journal of Autism and Developmental Disorders , 42 (11), 2450 2459. https://doi.org/10.1007/s10803 - 012 - 1499 - 7 Journal of Autism and Developmental Disorders , 16 (4), 513 515. https://doi.org/10.1007 /BF01531716 145 Witwer, A. N., & Lecavalier, L. (2007). Autism screening tools: An evaluation of the Social Communication Questionnaire and the Developmental Behaviour Checklist Autism Screening Algorithm. Journal of Intellectual & Developmental Disability , 32 (3), 179 187. https://doi.org/10.1080/13668250701604776 Woodbury - Smith, M., Klin, A., & Volkmar, F. (2005). Asperger?s Syndrome: A Comparison of Clinical Diagnoses and Those Made According to the ICD - 10 and DSM - IV. Journal of Autism and Developmental Diso rders , 35 (2), 235 240. https://doi.org/10.1007/s10803 - 004 - 2002 - x Woolfenden, S., Sarkozy, V., Ridley, G., & Williams, K. (2012). A systematic review of the diagnostic stability of Autism Spectrum Disorder. Research in Autism Spectrum Disorders , 6 (1), 345 354. https://doi.org/10.1016/j.rasd.2011.06.008 Zander, E., Willfors, C., Berggren, S., Choque - (2016). The objectivity of the Autism Diagnostic Observation Schedule (ADOS) in naturalistic clinical settings. Eu ropean Child & Adolescent Psychiatry , 25 (7), 769 780. https://doi.org/10.1007/s00787 - 015 - 0793 - 2 Zarokanellou, V., Kolaitis, G., Vlassopoulos, M., & Papanikolaou, K. (2017). Brief report: A pilot study of the validity and reliability of the Greek version o f the Social Communication Questionnaire. Research in Autism Spectrum Disorders , 38 , 1 5. https://doi.org/10.1016/j.rasd.2017.03.001