Integrating eSAGE with EHR data using machine learning for the early detection and monitoring of cognitive impairment in individuals including those underserved - Alzheimer’s disease (AD) is a major public health crisis. It is estimated that in 2022, 6.5 million Americans ages 65 or older lived with AD, and 5 million among the same population had mild cognitive impairment (MCI). By 2025, the number of AD cases will reach 7.2 million. The healthcare costs for individuals with Alzheimer’s or other dementias are substantial. The total payments in 2022 for all individuals with AD or other dementias were estimated to be $321 billion; by 2050, the annual payments for AD healthcare will total almost $1 trillion. However, resources are limited, particularly for underserved populations, and identification of cognitive impairment is often delayed so long that more effective treatments are underutilized. Therefore, expert panels have continued to stress the need for validated, brief, case-finding cognitive assessment tools, especially self-administered tests that allow for completely unsupervised administration and can accurately identify those with MCI. However, adding to the difficulties of AD diagnosis during the MCI stage are the significant disparities in the prevalence of AD and access to healthcare in racial/ethnic minority groups and other under-resourced and/or underserved populations. Black and Hispanic individuals are disproportionately more likely than White individuals to have AD, yet their socioeconomic disadvantages impede the early detection of MCI or AD. Thus, having easy access to low/no-cost validated and accurate self-administered cognitive assessments that can be taken in any clinic or non-clinic setting is highly consequential for the early detection of cognitive impairment. The goal of this project is to leverage the electronic Self-Administered Gerocognitive Examination (eSAGE), the metadata collected during eSAGE test-taking, the rich EHR data, and advanced ML techniques to develop such tools, which are particularly accessible to all individuals, including those socioeconomic disadvantaged, AD-vulnerable populations. To achieve the goal, we have three Aims. Aim 1 is to develop behavioral data and metadata, and ML methods to enhance the scoring and predictive ability of eSAGE. Aim 2 is to combine eSAGE test data and metadata/behavioral data with EHR data and develop new ML approaches to increase predictive capacity for cognitive status detection. Aim 3 is to test and validate the new ML- and EHR-enhanced eSAGE with the OSU Wexner Medical Center Memory Disorders Clinic, and with the underserved rural and minority populations in Ohio through outreach and tele-health (tele-cog) to evaluate their cognitive complaints. Successful completion of this project will produce a novel, translational, ML-enhanced eSAGE smart app and its integration within EHR systems, to improve our understanding of behavioral and clinical characteristics of cognition impairment, facilitate the identification of cognition impairment, and ultimately have a translational impact on AD identification and management. The project will enable a widely accessible, easy-to-use, and highly accurate cognitive assessment tool to be available for underserved, AD-vulnerable populations.
