Real-world autism care in preschool children: family context, service composition, and 1-year follow-up

Background Families of preschool children with autism often use several services concurrently. We described real-world service configurations, family characteristics, and clinical follow-up over approximately 1 year. Methods This retrospective cohort included 163 of 198 caregiver questionnaire respondents at a pediatric center in Shanghai. Included children were aged 25–54 months at baseline, had received care in a classifiable service configuration for at least 10 intervention months, and had paired assessments 10–14 months apart. Configurations were parent-mediated care informed by naturalistic developmental behavioral interventions (PM; n = 58), occupational therapy/sensory integration-predominant institutional care (OT/SI; n = 49), and applied behavior analysis (ABA)-predominant institutional care ( n = 56); both institutional groups included speech-language therapy. We described family-resource characteristics and retrospectively reported selection reasons, component-specific professional and caregiver participation, and paired Childhood Autism Rating Scale (CARS) scores and five Griffiths developmental quotients (DQs), summarized by a study-defined five-domain mean (GQ-5). Secondary regressions examined baseline-adjusted follow-up associations; exploratory adjusted analyses examined domain-specific associations. Results Education, income, funding, and reported clinician recommendations were associated with configuration (all Holm p < 0.001). PM combined fewer professional hours with more parent practice. The institutional configurations shared speech-language therapy but differed in predominant components and reported hours. Mean CARS reductions were 1.52, 0.22, and 1.98 points in PM, OT/SI, and ABA, respectively; mean GQ-5 increases were 0.83, 2.27, and 4.53 points. Among 127 children with increased GQ-5, 69 also had a numerical decrease in at least one DQ. CARS contrasts were sensitive to how baseline development was modeled, whereas GQ-5 contrasts remained detectable across full-sample specifications. After Holm correction, adjusted omnibus DQ tests were significant for all five domains with clinical adjustment and four with additional family-resource/access adjustment. Some domain associations were sensitive to baseline specification. Conclusion Family context, reported service composition, and multidimensional follow-up provided complementary information about routine care. Reviewing service components and domain-level development alongside symptom and composite scores may inform individualized follow-up; these observed between-configuration differences should not be interpreted as comparative treatment effects.