Developing Prediction Model for Hospital Appointment No-Shows Using Logistic Regression

Abstract

Patient no-shows are a significant problem in health care which leads to increased cost, inefficient utilization of capacity, and discontinuity in care. With the existing available patient appointment history, the research aims to predict the appointment no-shows of patients in a public hospital using the method of logistic regression. Based on characteristics of the appointment history data, the features are divided into demographic variables, appointment characteristics, and clinical characteristics. By considering these features and its multiple combinations, a fivefold cross-validation technique is used to choose the best feature combination for the best prediction model. From the analysis, it is found that appointment characteristics give better predictions. The study tested the model with appointment characteristics, and the performances are evaluated using accuracy, specificity, precision, recall, and F1-score. The model is evaluated using receiver operating characteristic curve and precision-recall curve. Hospitals can employ the model to predict appointment no-shows and implement mitigation strategies based on the outcome of the prediction.