Development of the Patient Scheduling System to Handle Patient No-Show Problem Using Data Mining and Simulation-Based Optimization Techniques

The number of patients who do not show up for their appointments significantly impacts the delivery of care, cost, and resource planning in most healthcare systems. We studied the effects of the no-show problem on Veterans Healthcare Administration (VHA) in zone 16 and specifically Michael E. DeBakey VA Medical Center (MEDVAMC) in Houston with more than 750,000 visits per year. One of the promising methodologies to resolve the no-show problem is overbooking, i.e., booking more patients than available appointment slots. The objective of this research is to develop a methodology and decision support system to assist the appointment scheduler in patient overbooking and scheduling. The appointment scheduler needs to find the appropriate patient appointment slots that minimize the patients' waiting times and the overtime working of physicians and staff, as well as maximizing the resource utilization. Based on the relevant historical data for scheduled and walk-in patients collected for two recent years in MEDVAMC, three specific tasks have been conducted:;(1) We conducted statistical and economic analysis of no-show rates and costs in VA hospitals in zone 16 and specifically in MEDVAMC.;(2) We developed the prediction models for the patient show-up probability and the number of walk-in patients, respectively. We identified the significant factors on no-show by analyzing maximum likelihood estimates of the logistic regression model and the stepwise selection technique. Furthermore, for estimating the probability of patient's no-show, we developed two data mining models based on logistic regression and support vector machines (SVMs). We also developed the prediction model for the number of walk-ins by using the density function technique.;(3) We developed the dynamic outpatient appointment overbooking and scheduling model that builds the patient schedule sequentially through a call-in process. A simulation-based optimization model is developed that can involve the realistic characteristics of the appointment scheduling system. The proposed model relaxes unrealistic assumptions in the existing analytical models. We designed sensitivity analyses for the coefficients of the objective function and simulation iteration using numerical experiments. Finally, the efficiency of our model is tested by numerical experiments.