Association between stomach pain and depression in middle-aged and older chinese adults: Evidence from CHARLS with machine learning and SHAP analysis.

BACKGROUND

Emerging evidence indicates a bidirectional relationship between gastrointestinal symptoms and psychological distress, yet population-based evidence concerning stomach pain and depression remains scarce. This study investigates the association between stomach pain and depression within a nationally representative cohort of middle-aged and older adults in China.

METHODS

Data were obtained from the China Health and Retirement Longitudinal Study (CHARLS). Multivariable logistic regression was employed to evaluate the association between stomach pain and depression, adjusting for sociodemographic, behavioral, and clinical covariates. Subgroup analyses were performed based on age, sex, residence, and comorbidities. Six machine learning models were trained to predict depression using stomach pain and related features. The extreme gradient boosting (XGBoost) model was further interpreted via Shapley Additive Explanations (SHAP).

RESULTS

Stomach pain was significantly associated with higher odds of depression (adjusted odds ratio [OR]: 2.36; 95% confidence interval [CI]: 2.15-2.59). Subgroup analyses revealed consistent associations among females, rural residents, and individuals with hypertension or dyslipidemia. XGBoost demonstrated the highest predictive performance among the machine learning models (area under the receiver operating characteristic curve [AUROC]: 0.988 in training; 0.884 in testing), followed by light gradient boosting machine (LGBM) and CatBoost. SHAP analysis indicated that rural residence, education level, age, and stomach pain were the most influential predictors of depression risk.

CONCLUSIONS

Stomach pain is independently associated with depression among middle-aged and older adults in China. Routine assessment of gastrointestinal symptoms could aid in the early identification and stratification of depression risk in aging populations.