Simulation–Optimization of Hospital Capacity and Chronic Care Pathways under Demographic Shift: A Multi-Objective, Equity-Constrained and Spatially-Aware Framework

Abstract

In an era marked by aging populations, increasing chronic disease rates, and ongoing disparities in health services access, healthcare systems across the globe are facing a series of challenges. Healthcare systems around the world are increasingly grappling with a combination of factors, including an aging population, a growing burden of chronic diseases, and persistent inequities in health services access. Current hospital capacity planning models consider demand as constant, only indirectly include equity, and separate simulation from optimization, resulting in sub-optimal and inequitable recommendations. This paper proposes a new integrated simulation-optimization approach that combines stratified discrete event simulation (DES) with a multi-objective evolutionary algorithm (NSGA-II) with explicit equity constraints, which are based on needs, and with spatial access variations. The model framework represents the dynamic demographic demand, Markov chain chronic disease progression (with type 2 diabetes mellitus as the example disease), and travel times by zone in eight geographic zones. The need-weighted absolute wait time deviation is a formal equity metric that is directly used in the optimization loop and is thus treated as a Pareto objective for healthcare capacity planning for the first time, to the best of our knowledge. By applying a semi-synthetic analysis of 500,000 people in a region, Pareto frontier analysis shows that equity constraints reduce the total cost of the system by 9-14% when incorporating accelerated ageing, and targeted telehealth expansion can lower inequity by more than 23% without increasing cost by more than 8%. Equity-neutral optimization is found to be inequitable and to cause up to a 35% increase in wait times for low-socioeconomic status (SES) populations compared with high-SES populations, supporting the need for explicit fairness constraints. The framework is scalable: It takes about 4.9 hours of wall-clock time for 1,000 Pareto evaluations across 20 parallel cores, and generates actionable insights for resilient, fair, evidence-based design of health systems.