Machine Learning Reveals How Depression Influences Chest Pain Localization and Its Predictive Value for Coronary Artery Disease

Depression is independently associated with chest pain and increased cardiovascular risk, regardless of coronary artery disease (CAD) status. However, limited research has examined how chest pain characteristics differ in individuals with and without depression. This study evaluates the relationship between depression and chest pain localization to inform CAD risk assessment. Data from the National Health and Nutrition Examination Survey (NHANES) between 2005 and 2020 were analyzed for adults completing the chest pain questionnaire. Survey-weighted propensity score matching created depression-stratified cohorts. Chest pain localization and inter-regional correlations were compared using survey design-adjusted methods. Random Forest models with survey-weighted bootstrap replicates were used to estimate the relative importance of pain locations in predicting CAD, stratified by depression status. A total of 2,208 individuals were matched (1,104 per cohort). Depressed individuals more frequently reported chest pain in the lower sternum (p = 0.045), left chest (p = 0.002), and epigastrium (p = 0.039). In depressed individuals, atypical pain regions (epigastrium, lower sternum, neck, arms) were more predictive of CAD, whereas in non-depressed individuals, typical regions (chest, upper sternum) were stronger predictors. These findings were robust to temporal validation and stricter definitions of depression. Depression modifies the predictive value of chest pain localization for CAD. A "depressed chest pain profile" involving non-traditional locations was more strongly associated with CAD in those with depression, while a more central "non-depressed pain profile" was more predictive in those without. These findings underscore the importance of integrating mental health context into chest pain evaluation.