Volunteer Outreach and Predictive Modeling: Rapid Randomized Quality Improvement Project for New Patient Attendance in a Primary Care Safety-Net.

Nonattendance at new patient appointments leads to missed opportunities for engagement in care, lost revenue, and suboptimal resource utilization. To assess the effectiveness of outreach calls to new patients, prioritized by a no-show predictive algorithm and conducted by volunteers, on visit attendance. Rapid randomized quality improvement project. Patients with new patient appointments at an urban safety-net adult primary care clinic scheduled to occur between August 1, 2024 and September 30, 2024. Estimated probability of visit no-show for patients was calculated using a predictive algorithm embedded in the electronic health record and used to sort lists of patients with upcoming appointments. Every other patient received an outreach call from a trained volunteer within 3 business days of their appointment plus usual automated reminder messages versus usual automated reminder messages alone. New patient visit attendance compared between intervention and control groups. We conducted subgroup analyses of attendance by visit modality (in-person vs. telehealth), preferred language, and quartile of predicted no-show probability. Patients in the intervention group (n = 281) had higher visit attendance than those in the control group (n = 280): 68.0% versus 54.1% (p < 0.01). There was a significant difference in attendance for in-person (70.7% vs. 51.7%; p < 0.01) but not telehealth (60.6% vs. 61.2%; p = 0.94) visits. Patients who preferred English had the biggest increase in attendance (17.2%; p < 0.01). Patients in the second and third quartiles of predicted no-show probability (31%-38% and 39%-45% predicted probability) had the biggest increases in attendance (22.2% [p = 0.01] and 15.4% [p = 0.05]). Outreach calls for new patients, prioritized by a no-show predictive algorithm and conducted by volunteers, can be a feasible and effective approach to improving visit attendance in a targeted fashion. Further investigation is needed to understand how to better support non-English preferring patients and patients with telehealth appointments.