• Author
    Rashed Alananzeh
  • Discovery PI

    Dr. Lauren E. Wessel

  • Project Co-Author

    Nicole J. Newman-Hung, MD; Kameel Khabaz, BA; Edward C. Cheung, MD; Frank A. Petrigliano, MD; Lauren E. Wessel, MD

  • Abstract Title

    Predictors of Missed Appointments in Sports Medicine Clinic

  • Discovery AOC Petal or Dual Degree Program

    Healthcare Improvement & Health Equity Research

  • Abstract

    Keywords: Health Equity, Access Barriers, Predictive Analytics

    Background: Missed appointments ("no-shows") impose a substantial operational and clinical burden on healthcare systems. Although predictors of nonattendance have been studied in primary care and general medicine settings, less is known about drivers of no-shows in sports medicine, particularly using interpretable machine-learning approaches.

    Objective: The primary purpose of this study was to identify factors independently associated with appointment no-show behavior in a sports medicine clinic. Secondarily, we sought to compare the relative influence of behavioral, scheduling, clinical, and social predictors and to compare predictive performance between machine-learning and logistic regression models.

    Methods: We performed a retrospective study at a single academic medical center between March 2013 and May 2025. Appointments were classified as completed or no-shows based on EHR disposition; cancelled, telemedicine, and rescheduled encounters were excluded. Candidate predictors included demographics, insurance/financial variables, social determinants of health (SDOH), clinical diagnoses, scheduling characteristics, reminders, and prior attendance behavior. Independent predictors were identified using multivariable logistic regression. Supervised machine-learning models were developed to evaluate nonlinear effects, with explainability assessed using SHAP values.

    Results: Among 122,710 scheduled appointments, the overall no-show rate was 5.8%. In multivariable regression, appointment confirmation was most protective (aOR 0.19, 95% CI 0.15–0.24), while prior no-show history in other clinics showed the largest risk association (aOR 10.76, 95% CI 7.63–15.15). Higher copay reduced no-show odds (aOR 0.95 per dollar, 95% CI 0.94–0.95), whereas longer scheduling lead time increased risk (aOR 1.014 per day, 95% CI 1.012–1.016; all p<0.001).

    CatBoost and XGBoost outperformed logistic regression (AUROC 0.84 vs 0.82). SHAP analysis identified copay amount as the most influential predictor (mean |SHAP|=1.14), demonstrating nonlinear protective effects. A parsimonious 4-feature model retained 93% of full model discrimination (AUROC 0.76).

    Conclusion: Appointment nonattendance in sports medicine was most strongly associated with modifiable behavioral, scheduling, and financial factors. Higher copay amounts were associated with lower odds of no-show, whereas longer scheduling lead time and greater prior no-show burden were associated with higher no-show risk.

Back