A predictive model for decreasing clinical no-show rates in a primary care setting

Abstract

A challenging obstacle to primary health care in the United States (US) is patient no-shows or missed appointments. The no-show rate can vary from 5.5% to 50%. A no-show may contribute to increased health risks, poor continuity of care, and loss of revenue. In this study, we develop and test a predictive model of patient visits. Retrospective regression analyzed patient visits in 2014 and 2015. Dependent variables were month, day, age, gender, race, ethnicity, insurance type, visit type, and the number of previous no-shows. A threshold for classifying no-shows was determined. The model was prospectively tested on patient visits in 2016. Significant variables included age, visit type, insurance, and number of previous no-shows. The model performed at 47% sensitivity and 79% specificity. The receiver operating characteristic (ROC) area under curve (AUC) was 0.72 (95% CI, 0.69–0.76) for the model and 0.70 (95% CI, 0.65–0.74) for prospective analysis. Simulated overbooking with the model resulted in 3.67 vs. 6.87 unused appointments, P < 0.000 (mean diff 3.2, 95% CI, 2.9–3.5). It is feasible to develop and implement a predictive model for single physician practices and implementation may improve practice efficiency.