Cost Effective Data Mining Approach for Outpatient Scheduling : Analyzing the Performance of Appointment Scheduling Systems

Poor scheduling and operations management can be very costly for clinics. There is a trade-off between the preference of patients for short waiting times and the preference of facilities for minimal idle times of medical staff. In outpatient scheduling, the common practice of assigning time slots to patients based on simple patient classification (e.g., 45 minutes for a new patient; 30 minutes for a return patient) is not efficient. By using the characteristics of each patient of a clinic, we develop a neural network prediction model for the visit time of each patient. We then develop a rounding method for estimating the duration of patients’ examinations to improve the performance of the scheduling system. Compared to other studies, we conduct a more nuanced examination of common assumptions of appointment rules in the literature. We relax some of these assumptions and conduct sensitivity analysis. We use the inputs of our model to evaluate the performance of each of the three appointment scheduling systems (the individual-block rule, the plateau-dome rule, and the OFFSET rule) combined with three sequencing policies in the literature. Empirical and simulation modeling help us develop interesting findings and insights. According to our analysis, the proposed appointment system reduces the overall cost function involving the patients’ waiting times, the doctors’ idle times, and the facility’s overtime by as much as 32.5% compared to the common classification method. Using patients’ specific characteristics improved the cost function of every appointment rule tested in this study. This study contributes to the literature and practice