H-RL-MUSYA is a multi-objective safe-reinforcement-learning framework exploring budgets across nutrition, mental health, behavioral risk and accident prevention in Thailand’s Health Region 10. It generated 127 Pareto-efficient policies; a representative modeled compromise reported 34.1% more DALYs averted, 31.3% better cost-effectiveness and a health-equity Gini reduction from .243 to .187 versus historical allocation.
Key findings
- The model identified 127 Pareto-efficient policies. Its representative compromise reported 847,293 DALYs averted, a 34.1% improvement, 31.3% better cost-effectiveness and a Gini change from .243 to .187. A 12-month pilot reported +23.1% composite health improvement and 91% acceptance; these are study-defined results, not population randomized-trial effects.
Why this matters globally
Health budgets involve conflicting values globally. Pareto frontiers can make efficiency-equity trade-offs visible and deliberation more transparent when data, assumptions, constraints and public rights are disclosed.
Thai researcher contribution
A multi-university Thai team connected optimization engineering, data science and public health around a regional allocation problem grounded in Thailand.
Limitations to consider
The abstract does not detail data provenance, composite definitions, confounder control or pilot design. Historical comparisons invite confounding and secular trends; large DALY estimates are model-sensitive and may overlap across interventions. There is no randomized or external regional validation, and equity cannot be reduced to one Gini metric.