An adaptive and interpretable case grouping framework for healthcare payment systems

The diagnosis-intervention packet is an innovative medical insurance reimbursement system based on case-mix, developed in China to facilitate more precise medical service management. Despite the intended role of the auxiliary catalog in subdividing case groups, its formulation remains exploratory and lacks a systematic methodological foundation.

To bridge this gap, this study proposes a case grouping framework that synthesizes optimization modeling, intelligent algorithms, and structured representation techniques. Specifically, a mathematical optimization model is developed to minimize intragroup variation, and case severity is quantified using nonnegative LASSO regression and the natural breaks method. To improve the self-adaptive and interpretable capacity of the algorithm for solving the model, an adaptive learning differential evolution algorithm based on Q-learning and a hypercube learning strategy is designed, and the case grouping pathways are represented using tree structures.

Empirical analysis demonstrates that the proposed framework effectively enhances intragroup homogeneity and intergroup heterogeneity of the case groups under established constraints, thereby providing a robust foundation for the development of the auxiliary catalog of diagnosis-intervention packet. Furthermore, its generalizability is validated by extending the framework to subdivide diagnosis-related groups.

The proposed framework offers a theoretically sound and practically feasible method for case grouping in the medical insurance payment systems. Its flexibility and scalability support the development of more accurate and equitable reimbursement schemes.