Validation of an Integrated Autism Screening Model Using DSM-5, CARS-2, and CASD Among Children

Technical Summary: Validation of an Integrated Autism Screening Model Using DSM-5, CARS-2, and CASD Among Children in Sabah, Malaysia

Problem Statement
Autism Spectrum Disorder (ASD) identification in Sabah, Malaysia, faces significant challenges due to geographic fragmentation, limited access to developmental specialists in rural districts, and a shortage of intervention centers relative to the growing prevalence of ASD. While international literature suggests that DSM-5 criteria may underestimate mild ASD cases—particularly in children with higher cognitive function or less obvious communication difficulties—there is a lack of empirical evidence regarding the diagnostic agreement of DSM-5 with established screening tools (CARS-2 and CASD) within the specific socio-cultural and healthcare context of Sabah. The disparity between urban and rural access to services exacerbates the risk of delayed diagnosis and missed early intervention opportunities.

Methodology
The study employed a quantitative cross-sectional design to evaluate the validity and diagnostic agreement of an integrated screening model.

• Participants: A purposive sample of 50 children aged 4 to 6 years was recruited from four autism intervention centers in Sabah (NASOM Kota Kinabalu, Seri Mengasih Centre, EF Borneo Intervention Program, and Persatuan C.H.I.L.D. Sabah). All participants had been referred for developmental assessment due to suspected ASD.
• Instruments: The study utilized four assessment measures:
  1. DSM-5: Applied to assess social communication deficits and restricted/repetitive behaviors.
  2. Childhood Autism Rating Scale, Second Edition (CARS-2): Used to evaluate the severity of autism symptoms across multiple domains.
  3. Autism Spectrum Disorder Checklist (CASD): A structured screening tool designed for practicality in low-resource settings to identify behavioral characteristics.
  4. Clinical Report Summary: A multidisciplinary clinical impression serving as the reference standard.
• Data Analysis: Data were analyzed using SPSS version 30. The study utilized descriptive statistics, Spearman's rank correlation to examine relationships between scores, and weighted Cohen's kappa to determine the level of agreement in severity classifications (mild, moderate, severe).

Key Results

• Convergent Validity: All assessment variables showed significant positive correlations ($p < .001$).
  • DSM-5 scores demonstrated very strong correlations with CARS-2 ($r = .923$), CASD ($r = .885$), and Clinical Reports ($r = .887$).
  • The strongest correlation was observed between the CASD Checklist and Clinical Report Summary ($r = .996$).
  • CARS-2 also showed very strong correlations with CASD ($r = .936$) and Clinical Reports ($r = .938$).
• Diagnostic Agreement: Weighted Cohen's kappa values indicated moderate to substantial agreement across severity classifications ($\kappa = .593$ to $.768$).
  • The highest agreement was found between DSM-5 and CARS-2 ($\kappa = .768$).
  • Agreement between DSM-5 and CASD was moderate-to-substantial ($\kappa = .698$).
  • Agreement between DSM-5 and the Clinical Report Summary was moderate ($\kappa = .593$), suggesting that while scores correlate highly, DSM-5 severity classifications may occasionally differ from broader clinical impressions, particularly for children with mild symptoms.
• Descriptive Statistics: Mean scores across DSM-5, CARS-2, and CASD were closely comparable (approx. 2.25–2.26), indicating that the instruments identified similar overall levels of autism-related characteristics, though clinical reports tended to indicate slightly higher severity (Mean = 2.41).

Significance and Claims
The study claims to provide local empirical evidence supporting the use of an integrated autism screening model in Sabah. The authors assert that combining DSM-5 criteria, CARS-2, and CASD offers a reliable and consistent approach for identifying ASD characteristics and severity.

• Complementary Strengths: The integrated model leverages the diagnostic criteria of DSM-5, the structured behavioral severity assessment of CARS-2, and the sensitive early symptom screening of CASD.
• Addressing Local Constraints: The authors posit that this multidimensional approach is particularly valuable in Sabah's context, where rural distance, limited specialist availability, and incomplete developmental histories can hinder early diagnosis.
• Clinical Utility: The findings suggest that the integrated model can improve diagnostic accuracy, reduce missed cases of mild ASD, enhance referral decisions, and support earlier intervention planning by facilitating collaboration between healthcare and education professionals.
• Limitations: The authors acknowledge the modest sample size (50 children) from selected intervention centers and note that future research should involve larger, more diverse samples and longitudinal designs to predict long-term developmental outcomes.

The paper concludes that while no single tool is sufficient for independent use in this resource-constrained environment, the convergence of these three instruments with clinical reports validates the proposed integrated framework for improving early ASD detection in Sabah.