Early Detection of Mild Cognitive Impairment, Alzheimer's Disease and Other Dementias using EHR

PROJECT SUMMARY The aging population has led to an increase in cognitive impairment (CI), including mild cognitive impairment and dementia. Alzheimer’s disease is the most common cause of dementia, with more than 6 million people currently affected and an estimated increase to 15 million by 2060, causing significant public health concerns. However, a recent study indicated that clinicians are not aware of CI in more than 40% of their patients, which results in missed opportunities for appropriate care plans, leading to adverse clinical outcomes. The clinical diagnosis of CI requires an extensive evaluation with a battery of standardized tests and questions to patients and caregivers. However, these assessments are not routinely performed in the majority of healthcare institutions, resulting in a significant delay in diagnosis. Electronic health records (EHRs) contain significant amounts of relevant information that is routinely recorded as part of clinical care. Indeed, our preliminary study shows that early signals of CI exist in EHRs, several years before clinical diagnoses. However, little is known about systematically analyzing patient health data in EHRs and how their temporal trends are associated with the development of CI. To address this gap, we will utilize unique resources to identify EHR patterns that rapidly detect the development and risk of CI: (a) Mayo Clinic Study of Aging (MCSA) cohort with longitudinal cognitive assessments and extensive clinical characterization. This will provide an ideal gold standard to assess the validity of EHR-derived CI; and (b) Rochester Epidemiology Project (REP), which provides access to longitudinal EHRs from multiple healthcare institutions. The primary goal of this study is to develop an informatics tool to extract patient health conditions related to CI from EHR data from multiple healthcare institutions (Aim 1, informatics). We will then characterize temporal health trends of CI patients by mining routinely-collected longitudinal EHR data (Aim 2, population health); and will develop a predictive model to early identify patients at high risk of CI using temporal trends of patient health (Aim 3, clinical practice). The tools developed will be deployed in public to facilitate further clinical research. In summary, the proposed research opens up new avenues for utilizing the routine EHRs to facilitate early detection of CI by characterizing patient’s temporal health trends (a potential surrogate of assessment-based clinical diagnosis). Widespread adoption of informatics tools can potentially lead to early detection of CI in healthcare settings and the ability to improve treatment plans and health outcomes for these patients.