I-Corps: Software platform to predict patient no-shows using machine learning algorithms

West Virginia University Research Corporation, Morgantown WV

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is the development of an AI-based healthcare patient scheduling system. Patient no-shows to medical appointments are a major logistical and economic challenge for clinics and hospitals, with an estimated average yearly no-show rate for primary care/specialty medical appointments of 20 - 35% leading to significant revenue loss. In addition, patients who failed to keep an appointment were 70% more likely not to return within 18 months, and older patients experiencing chronic illnesses are likely not to return to their physicians’ offices after missing just one appointment. Missing an appointment has been shown to create a significant negative impact on disease management leading to increased morbidity and increased costs. Factors that cause patients to miss their appointments have been identified, and the proposed system offers a proactive solution that may result in higher appointment adherence. Reduced missed appointments may decrease economic costs in medical services such as imaging studies, and downstream services resulting such as pharmacy services and medical equipment services. The proposed technology may increase operational efficiency by minimizing empty timeslots, increasing revenue, and also may lead to a significant boost in patient and clinician satisfaction and improved patient outcomes. This I-Corps project is based on the development of software tool that employs AI-based predictive models that has the potential to more accurately predict patient no-show and cancellation rates. Initial testing of the proposed AI-based algorithm resulted in no-show prediction rates of up to 90% as compared to other software applications using least-squares predictive techniques, which realized a 68% prediction rate. In addition, the use of a patient’s historical demographic data allows the system to learn/improve, increasing prediction accuracy over time. Currently available solutions are not dynamically adaptive and are dependent on individual physicians for protocols, which increases variability in scheduling and hence difficult to implement at a system level. The proposed algorithm provides the likelihood of appointment adherence based on patient’s historic behavior, and clinics and hospitals as well as patients may benefit by scheduling appointments more efficiently. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.