Project Summary The prevalence of cognitive impairment (CI) is expected to triple by 2050, contributing to decreased quality of life, increased medical care utilization, and additional burden on an already stressed primary care system. Many clinicians lack confidence to assess, diagnose and manage CI, and more than 50% of patients with CI are undiagnosed. To address these important problems, in phase 1 (R61) of this project, we developed and validated a machine learning model called MC-PLUS using results from brief Mini-Cog screens completed routinely at Annual Medicare Wellness exams and electronic health record (EHR) data to identify patients at elevated risk of a future CI diagnosis. We also developed, validated, and piloted a CI clinical decision support (CI-CDS) system to engage patients and clinicians in conversation about elevated CI risk, and to give clinicians the confidence and tools they need to diagnose and manage CI. Both MC-PLUS and the CI-CDS system were added into an existing web-based CDS platform that has high use rates and high primary care clinician satisfaction and is already seamlessly integrated with the Epic EHR. We are currently beginning phase 2 (R33), a large pragmatic trial with 30 primary care clinics randomized to receive CI-CDS or usual care (UC). We will evaluate the change in CI diagnosis and clinician confidence in diagnosing and managing CI among providers in CI-CDS clinics compared to those in UC clinics. If successful, the CI-CDS system will improve rates of new CI diagnosis and narrow existing sociodemographic disparities for adults with elevated CI risk identified by MC-PLUS at index visits in CI-CDS compared to UC clinics. The CI-CDS system will be available to 2 million patients annually at the study sites with the potential to disseminate more broadly through the existing non-commercialized CDS platform built on Epic EHR. However, the CI-CDS design needs to be updated and modernized from our established legacy Epic EHR pipeline to ensure its robustness, sustainability, interoperability, and scalability for dissemination to the larger community. The proposed grant supplement aims to engage our IT (Information Technology), software engineering and internal Epic EHR IT teams to modernize the CI-CDS architecture to enhance its portability, scalability and impact through the following steps: a) migrating CI-CDS to the OpenShift platform; b) converting its Epic EHR- specific integration to Fast Healthcare Interoperability Resources (FHIR)-based application programming interfaces (APIs); and c) re-architecting its patient data extraction and artificial intelligence (AI) inference pipeline for our MC-PLUS model from batch-based to a real-time model. These activities will facilitate broader impact of the tool by allowing integration into many different EHRs.

Project Narrative (From Parent Award, R33AG069770) Most experts advocate for early detection of cognitive impairment (CI) so that patients and caregivers can be prepared for making difficult decisions and to improve quality of life, but studies show that screening alone isn't sufficient to change clinician actions related to early detection. Using predictive modelling developed with machine learning methods and sophisticated clinical decision support (CDS) tools, it is possible to identify patients at elevated risk for CI and make it much easier for primary care to engage and support patients and caregivers in meaningful care planning. In this project, we implement and evaluate a low-cost, highly scalable CI-CDS system integrated within the electronic health record that has high potential to improve early CI detection and care and translate massive public and private sector investments in health informatics into tangible health benefits for large numbers of people.