Data Analytics and Predictive Modeling for Appointments No-show at a Tertiary Care Hospital

Abstract

This study aims to develop an accurate machine learning model for predicting no-shows in pediatric outpatient clinics at King Faisal Specialist Hospital and Research Centre (KFSH&RC), and understand pediatric patients' characteristics who are most likely will not show to their scheduled appointments. Appointment no-show data collected from KFSH&RC data warehouse over the period (01 Jan – 31 Dec 2019). We analyzed a dataset that consists of 101,534 scheduled appointments for 35,290 pediatric patients. No-shows over the mentioned period was 11,573 for 8,105 patients. Three machine-learning algorithms, namely logistic regression, JRip, and Hoeffding tree, were compared to find the best one. The no-show rate in pediatric outpatient clinics was 11.39%. Accuracy, precision, recall, and F-score were selected to evaluate the built models performance. The precision and recall of the three models was around 90%. The F-score of the three models was similar and equal to 0.86. These models improved our capability to identify pediatric patients’ characteristics at high risk of not attending their appointments.