Optimal Planning of Health Services through Genetic Algorithm and Discrete Event Simulation: A Proposed Model and Its Application to Stroke Rehabilitation Care

Background. Increasing demand for provision of care to stroke survivors creates challenges for health care planners. A key concern is the optimal alignment of health care resources between provision of acute care, rehabilitation, and among different segments of rehabilitation, including inpatient rehabilitation, early supported discharge (ESD), and outpatient rehabilitation (OPR). We propose a novel application of discrete event simulation (DES) combined with a genetic algorithm (GA) to identify the optimal configuration of rehabilitation that maximizes patient benefits subject to finite health care resources. Design. Our stroke rehabilitation optimal model (sROM) combines DES and GA to identify an optimal solution that minimizes wait time for each segment of rehabilitation by changing care capacity across different segments. sROM is initiated by generating parameters for DES. GA is used to evaluate wait time from DES. If wait time meets specified stopping criteria, the search process stops at a point at which optimal capacity is reached. If not, capacity estimates are updated, and an additional iteration of the DES is run. To parameterize the model, we standardized real-world data from medical records by fitting them into probability distributions. A meta-analysis was conducted to determine the likelihood of stroke survivors flowing across rehabilitation segments. Results. We predict that rehabilitation planners in Alberta, Canada, have the potential to improve services by increasing capacity from 75 to 113 patients per day for ESD and from 101 to 143 patients per day for OPR. Compared with the status quo, optimal capacity would provide ESD to 138 (s = 29.5) more survivors and OPR to 262 (s = 45.5) more annually while having an estimated net annual cost savings of $25.45 (s = 15.02) million. Conclusions. The combination of DES and GA can be used to estimate optimal service capacity. HIGHLIGHTS: We created a hybrid model combining a genetic algorithm and discrete event simulation to search for the optimal configuration of health care service capacity that maximizes patient outcomes subject to finite health system resources. We applied a probability distribution fitting process to standardize real-world data to probability distributions. The process consists of choosing the distribution type and estimating the parameters of that distribution that best reflects the data. Standardizing real-word data to a best-fitted distribution can increase model generalizability. In an illustrative study of stroke rehabilitation care, resource allocation to stroke rehabilitation services under an optimal configuration allows provision of care to more stroke survivors who need services while reducing wait time. Resources needed to expand rehabilitation services could be reallocated from the savings due to reduced wait time in acute care units. In general, the predicted optimal configuration of stroke rehabilitation services is associated with a net cost savings to the health care system.