RESUMEN
BACKGROUND: Collaborations between clinical/operational leaders and researchers are advocated to develop "learning health systems," but few practical examples are reported. OBJECTIVES: To describe collaborative efforts to reduce missed appointments through an interactive voice response and text message (IVR-T) intervention, and to develop and validate a prediction model to identify individuals at high risk of missing appointments. RESEARCH SUBJECTS AND DESIGN: Random assignment of 8804 adults with primary care appointments to a single IVR-T reminder or no reminder at an index clinic (IC) and 7497 at a replication clinic (RC) in an integrated health system in Denver, CO. MEASURES: Proportion of missed appointments; demographic, clinical, and appointment-specific predictors of missed appointments. RESULTS: Patients receiving IVR-T had a lower rate of missed appointments than those receiving no reminder at the IC (6.5% vs. 7.5%, relative risk=0.85, 95% confidence interval, 0.72-1.00) and RC (8.2% vs. 10.5%, relative risk=0.76, 95% confidence interval, 0.65-0.89). A 10-variable prediction model for missed appointments demonstrated excellent discrimination (C-statistic 0.90 at IC, 0.89 at RC) and calibration (P=0.99 for Osius and McCullagh tests). Patients in the 3 lowest-risk quartiles missed 0.4% and 0.4% of appointments at the IC and RC, respectively, whereas patients in the highest-risk quartile missed 24.1% and 28.9% of appointments, respectively. CONCLUSIONS: A single IVR-T call reduced missed appointments, whereas a locally validated prediction model accurately identified patients at high risk of missing appointments. These rigorous studies promoted dissemination of the intervention and prompted additional research questions from operational leaders.