Association of Shanghai air pollution with postoperative infection in adolescent orthopedic patients: a study using a deep learning-based evolutionary model

Background Surgical site infections (SSI) represent severe complications in adolescent orthopedic surgery. Shanghai’s complex air pollution profile creates a critical context to investigate multi-pollutant impacts on SSI risk in this vulnerable population. Methods We analyzed 32,261 adolescent SSI cases from Shanghai (2019–2024) alongside high-resolution pollution/meteorological data. An evolutionary deep learning model (CNN-BiGRU-Attention optimized by Improved StarFish Algorithm) and generalized additive models (GAMs) assessed lagged effects, age/gender stratification, and concentration-response relationships. Results NO 2 and SO 2 showed significantly positive associations with SSI risk at lag0 (concurrent day); O 3 exhibited protective effects (strongest at lag05: −2.396% [95% CI: −3.349% to −1.443%] per 10 μg/m 3 increase); Age stratification: 7–14 ages groups demonstrated heightened sensitivity to NO 2 /SO 2 . O 3 effects varied across age groups; Gender differences: O 3 ’s negative association was stronger in males; Dose–response: NO 2 /SO 2 showed monotonic increases with no safety thresholds; O 3 displayed a straight line curve. Conclusion Multi-pollutant exposure modulates SSI risk in adolescents, with NO 2 /SO 2 as risk factors and O 3 showing context-dependent protection. Deep learning identified SO 2 /NO 2 /O 3 as dominant predictors, supporting perioperative air-quality interventions.