Sex-Based Differences in Prenatal and Perinatal Predictors of Autism Spectrum Disorder Using Machine Learning With National Health Data.
Ju Sun Heo, Seung-Woo Yang, Sohee Lee and 2 others
PMID 40384623WHAT IT FOUND
Machine learning on national data identified sex-based differences in autism predictors.
Low maternal BMI raised risk for both sexes, but high BMI raised risk only for females. Low socioeconomic status raised risk for males, while high status raised risk for females. These are statistical associations, not causes.
Key findings
01Low maternal pregestational BMI was linked to higher autism risk in both sexes, while high BMI was associated with higher risk only in female offspring.
02Socioeconomic status showed a U-shaped association with autism risk: lower status increased risk for males, while higher status increased risk for females.
03Advanced maternal age was a significant risk factor for autism in both sexes, but the risk increased at different age thresholds: under 35 for males and over 35 for females.
STILL TO COME
How it was doneWhat they found
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
ASD diagnoses were based on insurance claims codes rather than standardized diagnostic assessments, which may affect accuracy and severity grading. The study excluded adult diagnoses and those made after age 14, potentially missing later-identified cases. Data was limited to a specific Korean population, restricting generalizability to other healthcare systems or ethnic groups. The model did not account for intellectual disability comorbidity or paternal age. The female subgroup had a small sample size in the training set, raising concerns about model robustness despite reported high accuracy. The study identifies statistical associations from retrospective data and cannot establish causation or clinical utility for prediction.
Declared interests
The authors declared no competing interests. The study used public national health insurance data.
The easy way to misread this
Do not interpret the high machine learning accuracy as evidence that these factors can reliably predict autism in individual patients. The model was trained on retrospective claims data where diagnosis codes may reflect healthcare access patterns (such as delivery institution level) rather than true clinical risk, and the study does not provide a validated screening tool for clinical use.