A risk prediction model for autism spectrum disorder integrating biopsychosocial factors: a systematic review and meta-analysis with multicenter validation

Background: Early identification of individuals at high risk for autism spectrum disorder (ASD) is crucial for optimizing intervention strategies and improving outcomes. This study aims to develop a risk prediction model integrating biopsychosocial factors through a systematic review with multicenter validation. Methods: A comprehensive search was conducted across PubMed, Cochrane Library, and Embase for articles on biopsychosocial ASD risk factors during 2010-2023. Two reviewers independently extracted data. Meta-regression analysis of 37 systematic reviews/meta-analyses identified 18 potential risk factors by Stata 16.0. Four core variables were included in the prediction model, while 14 were excluded due to low-quality evidence or insufficient data after screening. Multivariate logistic regression with least absolute shrinkage and selection operator (LASSO) variable selection derived model weights. External validation was performed in a Chinese cohort (n=1,175) from two tertiary hospitals. Model discrimination was assessed via receiver operating characteristic (ROC) curves and clinical utility by decision curve analysis (DCA). Results: Analysis of 37 systematic reviews identified four independent predictors of ASD risk: adverse childhood experiences (ACEs) [odds ratio (OR) =2.11; 95% confidence interval (CI): 1.61-2.77], preterm birth (OR =3.3; 95% CI: 1.24-7.60), antidepressant exposure during pregnancy (OR =1.17; 95% CI: 1.08-1.21), and perinatal antibiotic exposure (OR =1.52; 95% CI: 1.09-2.12). The risk model formula was: 0.82 × (ACEs) + 1.19 × (preterm birth) + 0.42 × (antidepressant exposure) + 0.21 × (perinatal antibiotic exposure). External validation showed excellent discrimination [area under the curve (AUC) =0.78; 95% CI: 0.75-0.81]. DCA confirmed significantly higher net clinical benefit compared to universal intervention strategies. Conclusions: This study developed a risk prediction model integrating biopsychosocial factors, providing an evidence-based tool for early identification of individuals at high risk for ASD.