Related Experiment Videos
Equity-Preserving Public Health Resource Allocation Using Multi-Objective Safe Reinforcement Learning: Evidence from
Nopparat Songserm1, Rapeepan Pitakaso2, Thanatkij Srichok2
1Department of Health Sciences, Faculty of Public Health, Ubon Ratchathani Rajabhat University, Ubon Ratchathani 34000, Thailand.
Summary
This study introduces H-RL-MUSYA, an AI framework for equitable public health budget allocation. It improves health outcomes and cost-effectiveness by generating adaptive strategies, enhancing regional health governance.
Area of Science:
- Public Health
- Health Economics
- Artificial Intelligence in Healthcare
Background:
- Equitable public health budget allocation across diverse intervention domains is a significant challenge in regional health governance.
- Current planning in Thailand's Health Region 10 relies on expert opinion and historical data, lacking systematic exploration of alternative allocation strategies.
- Resource allocation decisions are multi-criteria, involving health impact, cost-effectiveness, equity, and policy alignment, which are not captured by single metrics.
Purpose of the Study:
- To introduce H-RL-MUSYA (Hierarchical Reinforcement Learning for Multi-Domain Unified System of Yielding Adaptive allocations), a decision-support framework.
- To systematically generate and evaluate Pareto-efficient allocation strategies for public health budgets across nutrition, mental health, behavioral risk, and accident prevention domains.
- To assist public health practitioners in making informed, multi-criteria resource allocation decisions.
Main Methods:
- Development of the H-RL-MUSYA framework utilizing Hierarchical Reinforcement Learning.
- Application of the framework to Thailand's Health Region 10, covering 4.6 million inhabitants.
- Evaluation of generated allocation strategies based on DALYs averted, cost-effectiveness, and health equity metrics.
Main Results:
- H-RL-MUSYA identified 127 Pareto-efficient policies, yielding a compromise allocation that averted 847,293 DALYs (34.1% improvement).
- Cost-effectiveness improved by 31.3%, and the health equity Gini coefficient decreased from 0.243 to 0.187.
- A 12-month pilot study confirmed a 23.1% composite health improvement with 91% stakeholder acceptance.
Conclusions:
- AI-assisted policy exploration, like H-RL-MUSYA, can significantly enhance public health decision-making by revealing non-intuitive strategies.
- The framework quantifies equity-efficiency trade-offs in resource allocation.
- Human expertise, policy context, and deliberation remain crucial for final allocation decisions.