Credit risks in long-term care: a novel assessment of private long-term care institutions for sustainable development under the silver economy in China.

As private long-term care institutions play an increasingly important role in meeting long-term care needs in an aging society, assessing their credit risk is essential for supporting institutional sustainability, reducing information asymmetry, and strengthening risk-aware governance under the silver economy. However, existing studies have largely focused on subjective service-quality evaluations, with limited attention to objective, interpretable institution-level assessment based on publicly available data. Using publicly available records from 594 private long-term care institutions in Hangzhou, China, this study develops and validates an interpretable machine-learning framework for credit risk assessment. The framework integrates class-imbalance handling techniques with multiple machine-learning algorithms to improve predictive performance and identify the factors most closely associated with institutional creditworthiness. The results show that XGBoost with SMOTE achieves the best overall performance among the benchmark models across accuracy, sensitivity, specificity, and AUC. Feature importance and SHAP analyses identify four key predictors: registered capital and the number of tax-paying employees are positively associated with predicted creditworthiness, whereas financing history and patent counts are negatively associated with it in this setting. To enhance practical applicability, the final model is deployed in a Streamlit-based web application for real-time credit rating and visualization. Overall, this study extends credit risk assessment to the long-term care sector and provides an objective, interpretable, and deployable decision-support framework for families, managers, and regulators, with implications for strengthening the sustainable governance of private long-term care institutions.