Exploring autism spectrum profiles via network analysis of parent-reported ASSQ patterns

The behavioral profile of autism spectrum disorder (ASD) is highly heterogeneous. Analyses relying on total scores often fail to capture the associative patterns among individual behavioral items. This study applied network analysis to parent-reported Autism Spectrum Screening Questionnaire (ASSQ) data to explore the internal associative structure among core ASD behavioral domains and to identify key items within the network.

This retrospective study included 995 children and adolescents aged 7-16 years diagnosed with ASD according to ICD-10 criteria at Xiamen Xianyue Hospital between 2019 and 2025, whose parents completed the ASSQ. Given the 3-point Likert scale of the ASSQ, two parallel network modeling strategies were employed to enhance robustness: a dichotomized (Ising model) and an ordinal (Gaussian Graphical Model, GGM) approach to explore associations among ASSQ items. Comorbidities and psychiatric medication history were included as covariates. Central and bridge nodes were identified using expected influence and bridge expected influence, respectively, and the optimal model was selected. Network Comparison Tests were conducted to examine differences in global strength and structure between sex (male/female) and age groups (7-11/12-16 years).

(1) The ordinal network model (ASSQ-GGM network) demonstrated better centrality stability and was therefore selected for reporting. (2) The ASSQ-GGM network identified the following central nodes: Q11 (inability to adjust, EI = 1.238), Q27 (unusual posture, EI = 1.154), and Q15 (fails to make relationships with peers, EI =  1.047). Bridge nodes were: Q26 (unusual facial expression, bEI = 0.806), Q11 (inability to adjust, bEI = 0.719) and Q12 (lacks empathy, bEI=0.710). (3) Network Comparison Tests found no significant differences in network structure by sex or age, which may be due to insufficient or imbalanced subsample sizes. The possibility of sex- or age-specific network patterns cannot be ruled out.

This study presents a network of interrelated parent-reported ASD-related behavioral features based on ASSQ items. No significant sex or age differences in network structure were found. These findings provide insights into the associative patterns among ASSQ items. As this study is based on cross-sectional, guardian-reported data, the findings are hypothesis-generating and do not support causal inferences. Any implications for intervention require validation through longitudinal or experimental studies.

Not applicable.