Diagnostic Accuracy of ASRS and SRS in Screening ASD in Chinese Community Children

Background:  National prevalence of Autism Spectrum Disorder (ASD) is unknown in China; research on ASD has started recently. Up to now, very few screening and diagnostic instruments were available for the Chinese population, instruments that require population norms are still very scarce. An important initiative funded by the Ministry of Health was launched in 2013 in China to boost autism awareness and clinical and research expertise. As part of this recent program, a national epidemiological survey of ASD among the school-aged population of China was designed. In order to conduct the screening phase of this survey involving 120,000 children from 8 provinces, we needed an instrument, properly validated and adapted for the Chinese population. Based on a representative community-based reference sample and aged matched clinical ASD cases from 4 cities of China, we re-examined the factor structure that was proposed in the initial US study. We evaluated the modified Chinese version of Autism Spectrum Rating Scale (ASRS) and the existing Social Responsiveness Scale (SRS) to select a better screening tool for the large scale epidemiological survey.

Objectives:   The study aimed at comparing the diagnostic accuracy of ASRS and SRS in screening ASD in community children aged from 6 to 12 years old.

Methods:   The study sample consist of 1588 normal children from community (ASRS T-score<70) and 190 children with diagnosed ASD by DSM-IV from clinics from 4 cites of China. Parents of selected children were invited to complete the ASRS and SRS scales with 1-2 weeks interval. Two of the sites were randomly assigned to administer ASRS first; the rest of two sites did the SRS first and followed by ASRS with 1-2 weeks interval. Receiver Operating Characteristic (ROC) analysis was performed by using ASD disease status as gold standard (1 for cases, 0 for community controls) and ASRS as well as SRS scores as test variables, respectively. Area under curve (AUC) and 95%CI, sensitivity, specificity, positive and negative predicts value and positive rate were evaluated and compared between ASRS and SRS performances. The optimal cutoff of ASRS was proposed for screening ASD to obtain best sensitivity and specificity.

Results:  The study sample averagely aged 8.8 years old (range 6-12 years), with 51.1% and 87.9% males in normal controls and clinical cases (P<0.001). 31.2% and 65.2% questionnaires were administered by father and mother respectively. Both ASRS and SRS showed excellent performance in ROC analysis with AUC of 0.95(95%CI: 0.93-0.96) and 0.97 (95%CI: 0.96-0.98), respectively, with some overlap in the 95%CIs. Using the cutoff of 57 for ASRS and 60 for SRS, ASRS and SRS showed similar sensitivity (96.3% vs 94.2%), specificity (82.8% vs 82.0%) and positive predict value (39.0% vs 38.4%) and negative predict value (99.0% for both). The positive rate of ASRS and SRS scales were 17.23% and 17.9% respectively. Father administering the questionnaire or mother doing did not bias the comparison between the two screening scales.

Conclusions:  ASRS shows excellent performance in screening ASD from Chinese population aged 6-12 years old, which is comparable with SRS.