The Centers for Medicare and Medicaid Services (CMS) have issued guidance to health systems regarding intensive care risk segmentation for patients assigned to a provider team. Multiple approaches to identify those who may benefit from intensive care management strategies have been proposed. One approach is to use a "two-step" risk stratification (a tool for identifying and predicting which patients are at risk) that employs an algorithm based on electronic data allowing population-level stratification, while also adding the care team's perception of risk to adjust the risk stratification of patients, as needed. Clinicians' contributions to model performance in this setting remain untested, and there are numerous studies in clinical settings in which clinicians fail to provide highly accurate predictions around mortality, length of stay, and care utilization. This study will explore the predictive performance of a model implemented within the University of Pennsylvania Health System in the electronic health record (EHR) of multiple outpatient practices, among all patients who were signed a utilization risk score, along with a clinician annotation of the score in the EHR. The goal of the study is to describe the predictive performance of the model and compare it to the model with augmented with input from clinicians.