Predicting Patient Satisfaction in Indian Healthcare Using Artificial Intelligence: A Data-Driven Approach to Patient Relationship Management

Background: Patient satisfaction serves as a vital measure of healthcare quality, especially in regions with limited resources, such as Chhattisgarh, India—a state characterized by its tribal populations and underdeveloped medical infrastructure. This research employs Artificial Intelligence (AI) to forecast patient satisfaction levels, specifically the overall satisfaction, with the goal of improving patient relationship management (PRM) in a public hospital setting in Chhattisgarh. Methods: Data from 107 patient surveys were examined, encompassing demographic factors (e.g., age group, gender, income level, and frequency of visits), service quality aspects (e.g., timeliness, accessibility, communication, system efficiency), and views on technology (e.g., technology quality and usability). An XGBoost regression model was developed to predict the overall satisfaction, complemented by SHapley Additive exPlanations (SHAP) for model interpretability. Additional analyses involved Pearson correlations, multiple linear regression, and t-tests. Missing values (under 5%) were handled through k-Nearest Neighbors (k-NN) imputation. The study did not involve preregistration or animal testing. Results: The XGBoost model yielded a root mean squared error (RMSE) of 0.39 and a coefficient of determination, R² of 0.90. SHAP highlighted communication (mean SHAP value = 0.72, p < 0.001), system efficiency (0.48, p < 0.01), and technology usability (0.35, p < 0.05) as primary influencers. Correlations revealed strong links, such as between communication and the overall satisfaction (correlation coefficient, r = 0.82, p < 0.001). Regression analysis supported the significance of communication (β = 0.70, p < 0.001) and system efficiency (β = 0.45, p < 0.01). Patients with very frequent visits showed reduced satisfaction (mean = 3.5 vs. 4.0 for occasional visitors, p < 0.001). Conclusions: Artificial Intelligence demonstrates strong potential for predicting patient satisfaction, emphasizing the roles of communication and operational efficiency. These insights could guide targeted PRM interventions in Chhattisgarh to better serve tribal and low-income groups. However, given the modest sample size from a single site, results should be viewed as preliminary, warranting larger-scale validation.