Multi-objective machine learning framework for welfare-optimized health insurance design in infectious disease management (Gastroenteritis)

用于传染病管理(肠胃炎)中福利优化健康保险设计的多目标机器学习框架

阅读:1

Abstract

OBJECTIVE: To develop and validate a multi-objective machine learning framework for welfare-optimized health insurance design in infectious disease management, explicitly modeling trade-offs among patient financial protection, provider efficiency, and insurer sustainability through empirically-derived stakeholder utilities and social welfare functions. SETTING: The national health insurance system in Iran, utilizing a total of 378,403 claims of gastroenteritis hospitalizations across public, private, and charitable hospitals in 2023. DESIGN: Observational study employing 5 supervised machine learning models to predict stakeholder-specific utilities—patient (health outcomes, out-of-pocket burden, affordability), provider (length-of-stay efficiency, performance quality), and insurer (coverage adequacy, subsidy management, cost containment). Utilities were aggregated using six social welfare functions (Utilitarian, Nash, Atkinson, Rawlsian, Geometric Mean, Convex Combination) with empirically-derived stakeholder weights. Pareto frontier analysis identified welfare-dominant policy configurations across insurance arrangements. RESULTS: Only 2.46% of observed policies achieved Pareto efficiency, indicating substantial allocative inefficiency. “Win-Win-Win” configurations (16.9% of efficient policies) dominated 54.1% of all alternatives, demonstrating simultaneous welfare gains across stakeholders without requiring zero-sum trade-offs. The equity-fiscal sustainability correlation (r=-0.509) was nearly twice the magnitude of the equity-efficiency correlation (r=-0.267), identifying fiscal capacity—not operational inefficiency—as the binding constraint on patient-centered insurance design. Nash and Geometric Mean social welfare functions achieved superior aggregate welfare (mean utilities 4.01 and 8.84 respectively) with exceptional stability (SD = 0.15 and 1.02), while provider-insurer aligned policies imposed catastrophic patient burdens (mean utility − 22.39). Patient utility was driven predominantly by out-of-pocket burden (weight = 0.80), with health gains contributing modestly (0.12). CONCLUSION: Multi-stakeholder welfare optimization in health insurance is empirically feasible and does not require inherent efficiency-equity trade-offs. Fiscal constraints, rather than operational limitations, constitute the primary barrier to equitable insurance expansion, necessitating complementary revenue mobilization strategies alongside benefit design reforms to achieve sustainable universal financial protection. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12889-026-26285-9.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。