SNAP HOME: Smart No-show Analytics Paired With Home-based Outpatient Medical Engagement

Unused clinic visits due to patient no-shows continue to plague American healthcare as a large source of waste and avoidable constraint on access. The average no-show rate across 105 studies was 23% though with wide variation (4% to 79%). No-show behavior has adverse effects on patients, providers, and healthcare organizations' operational and financial outcomes. Patients that miss clinic visits are more likely to need acute care and suffer poor health outcomes. There have been increasingly sophisticated efforts focused on predicting which patients are likely to no-show. This can allow for tactful over-booking and/or patient outreach. At Hopkins, investigators have implemented a novel machine learning based approach for identifying those patients at high-risk for no-show. Offering home visits for patients who are most likely to no-show is an appealing strategy to connect medical providers with patients who need care but are otherwise unlikely to receive it. Yet, it is unclear if this would be helpful to engage patients in their care, and encourage subsequent attendance, or if it would encourage future missed appointments, fostering a reliance on possible ongoing home visits. This study would link existing efforts with no-show prediction to home visits by internal medicine residents and evaluate its clinical impact. Patients at high-risk for no-show will be randomized into the control arm where patients will be called to remind patients of their visits. Those randomized into the intervention arm will be offered a one time home visit in lieu of their in-person visit to help understand barriers to in-person care and build rapport. Outcomes evaluated include future in-person show rates and healthcare cost/utilization