Predictive Modeling of No-Shows in Orthopaedic Oncology Clinic

Introduction: Missed clinic appointments (“no-shows”) have far-reaching consequences on healthcare delivery. Despite extensive research on predictors of no-show, few studies address orthopaedic oncology clinics, which face unique challenges because of time-sensitive diagnoses, resource-intensive care, and patient-specific barriers. This study aims to identify predictors of no-show appointments and develop machine learning models to forecast missed visits in an orthopaedic oncology clinic at a high-volume sarcoma referral center. Methods: We retrospectively analyzed 11,667 scheduled appointments from March 2013 to September 2024, including demographic, socioeconomic, clinical, and behavioral variables. Predictors were identified through univariate and multivariate statistical and effect size analyses, and machine learning models were applied to predict no shows. Model performance was assessed with multiple scoring metrics, with feature importance quantified using SHapley Additive exPlanations values. Results: The overall no-show rate was 11%. Key predictors included historical no-show behavior, marital status, diagnosis type, and modifiable factors. Multivariate analysis revealed electronic medical record reminders (adjusted odds ratio [aOR] = 0.733, P = 0.002), appointment confirmations (aOR = 0.193, P < 0.001), and scheduling lead time (aOR = 1.005, P < 0.001) as the strongest modifiable interventions to decrease no-show rates. Machine learning models demonstrated robust sensitivity and specificity values, quantified as an area under the receiver operating characteristic curve of 0.83 and F1 score of 0.43. SHapley Additive exPlanations analysis highlighted that scheduling lead time and prior no-show behavior strongly influenced predictions. Conclusion: Effective electronic reminders, appointment confirmations, and tailored scheduling approaches with shorter lead times are key practice management strategies to mitigate no shows. Although past patient behavior, unmarried status, and several comorbidities may increase no-show risk, other demographic and socioeconomic factors were not associated with no-shows in our analysis. The multifactorial nature of no shows and limited predictive model performance necessitates nuanced, equitable approaches to enable targeted resource allocation for patients with elevated no-show risk, ensuring access and improving outcomes in orthopaedic oncologic care.