Cognitive Computing Models Now Available in Epic
Cognitive Computing Models, also known as Predictive Analytics, allow OCHIN members to use statistical modeling to calculate the probabilities for future events. The models can help clinicians identify patients at higher risk for negative outcomes. Risk scores, which are derived from the models, can be used in Best Practice Advisories (BPAs), Reporting Workbench Reports, Patient List, SnapShot Reports, Dashboards, and other areas of Epic to help identify patients for early intervention. As part of our ongoing Healthy Planet implementation, OCHIN implemented three new clinical models with the August 2019 EP, including:
Risk of Hypertension
Risk of Initial Myocardial Infarction
Risk of Negative Outcomes of Type 2 Diabetes.
Risk of Hypertension predicts the patient’s probability of developing hypertension over the next two years. This score can be used by clinicians and care managers to review a patient’s risk, drill down into the contributing factors, and help determine what interventions could lower the hypertension risk.
The Risk of Initial Myocardial Infarction score can be used by care managers and other clinicians to see an adult patient's risk of an initial myocardial infarction in the next two years so they can take preventive measures. The model is applicable to patients with overt ASCVD and can be used to complement existing cardiovascular disease (CVD) risk scores and better inform shared decision-making between patients and providers.
Using the Risk of Negative Outcomes of Type 2 Diabetes, care managers and other clinicians can review diabetes-related complications a patient is expected to develop over the next two years. Then they can drill down into the contributing factors, and determine what can be done to lower that risk.
OCHIN is currently planning the implementation of two additional Cognitive Computing Models in upcoming EPs. The Risk of Opioid Abuse of Overdose model will be available for use by clinicians and care managers to review a patient’s risk of opioid dependency or overdose in the next year. Additionally, a custom model is being developed to predict a patient’s HIV risk, with the expectation of using this score to identify patients for Pre-Exposure Prophylaxis (PrEP) intervention.
When predictive models are integrated into users' regular workflows, they can identify and reach out to at-risk patients, book appointments for those patients, and document patient care. The goal is better outcomes for patients, more integrated workflows for clinicians and care managers, and more money saved by members and payers.