Improving an outpatient clinic utilization using decision analysis-based patient scheduling

This study presents a predictive model to be used in scheduling patients in an urban outpatient clinic. The model is based upon actual patient characteristics from a physical therapy clinic within an urban health and wellness center situated in a public university. A number of reported patients' characteristics such as age, education level, distance from the clinic, historical attendance records, etc. were examined to determine if they significantly impacted the patients' missing scheduled appointments (no-shows.) Decision tree analysis was used to develop a model that assessed the likelihood of a patient's no-show, using key patient characteristics and attendance records. Such a model can be used to assist with scheduling patients in an outpatient clinic, while attempting to increase the clinic's overall utilization. Four tree growing criteria were examined to develop the model with the strongest predictive power. Predictive power of each method was assessed by using the entire dataset as well as using split sampling. The results were then compared with those of a Bayesian networks model and a neural networks model. In addition, the trade-off between the selected decision tree model's predictive power versus simplicity of the associated classification rules was examined. We also assessed the impact of various levels of overbooking on the clinic's utilization when using patients' schedules based on the predictive model.