PURPOSE: Currently, the Tampa VA hospital provides outpatient specialty services for central Florida using the primary medical facility and several satellite outpatient clinics. This practice results in huge travel burden to outpatients. In fact, many outpatient consult requests could be scheduled at closer clinics. So far, almost all effort has been devoted to centralized outpatient scheduling, whereas little consideration is given to conceptualize distributed spatial scheduling to minimize travel burden. We combined simulation and integer programming to model optimal resource distribution which minimizes patients' travel time and distance.

METHOD: We developed a simulation model that iteratively generates new consult requests and determines the patients to be seen along the decision horizon. We proposed an integer programming model to compute the optimal visit sites based on the satisfied patient demand. We compared the current practice and the proposed innovative approach of delivering care. We tested four innovative configurations with varying numbers of physicians or additional requirements on physician travel. These tests sequentially added modeling complexity and progressively approximated the real situation.

RESULTS: Using the simplest configuration, our model yielded an average 40.0% decrease over the corresponding benchmark in travel distance, and an average 32.7% decrease in travel time. With additional complexity, our model yielded at least 23.9% decrease on average in travel distance, and at least 20.9% decrease on average in travel time. In other variations in system constraints, we also obtained significant improvement in both measures. These improvements are stable to varying parameters in the consult request generation. The optimal location can vary iteratively with the demand.

CONCLUSIONS: Simulation analyses show that the current practice can be improved by physically locating physicians in a distributed fashion for patients requiring care. For further empirical analysis, our model can be populated with any existing or proposed demand and demographics. This model is useful for validating the suitability of iterative patient and site scheduling. It is also useful for later site selection of new clinical facilities.