Machine Learning Accurately Predicts Costs and Identifies Disparities for Common Outpatient Hand Surgeries.

Purpose: Costs of outpatient hand procedures vary widely, and the influence of social determinants of health is not well characterized. Machine learning (ML) offers a way to quantify cost drivers and identify disparities. We sought to (1) assess whether ML models can accurately predict the costs of common outpatient hand procedures (2) identify cost determinants and (3) assess whether socioeconomic or racial disparities exist in cost patterns. Methods: A retrospective cohort study was conducted using the New York State Ambulatory Surgery and Services Database (2015-2018) including patients undergoing open carpal tunnel release, endoscopic carpal tunnel release, trigger finger release, and distal radius fracture open reduction and internal fixation. ML models were trained to predict total procedure costs, and feature importance analyses identified key cost drivers. Associations between costs, Social Deprivation Index quintiles, and racial/ethnic composition were evaluated using Kruskal-Wallis and Dunn's tests and Pearson/Spearman correlations. Results: ML models trained on 61,460 patients demonstrated strong predictive performance. Patient zip code and hospital identifier were consistently top cost predictors. Significant disparities were observed: higher costs were associated with both high and low extremes of Social Deprivation Index, as well as higher proportions of Asian, Black, and Hispanic patients. In contrast, White patients were overrepresented in lower-cost quartiles. Conclusions: ML accurately predicted procedure costs and revealed consistent socioeconomic and racial disparities in cost patterns for common outpatient hand surgeries. These findings highlight the need for value-based policies designed to reduce inequities and improve access to high-quality, affordable surgical care. Type of study/level of evidence: Therapeutic III.