Antecedents and Outcomes of Subjective Cognitive Decline: An Electronic Health Records and Artificial Intelligence Approach - PROJECT SUMMARY Early detection of Alzheimer’s disease and related dementias (ADRD) from electronic health records (EHRs) can facilitate participant enrollment in clinical trials and early intervention once clinically available. Subjective cognitive decline (SCD) can be an early manifestation of ADRD. Previous research in early detection of ADRD has focused on observational study cohorts, generally small in size and often with stringent medical exclusion criteria. Investigation of larger and more representative samples is needed to develop a full understanding of the underlying conditions, procedures, and/or interventions that can contribute to cognitive decline or accelerate progression to dementia in the population at large. The overall goal of this proposed research is to leverage large-scale EHR data and advanced informatics technology to develop case-finding methods for SCD and to advance the understanding of its risk factors and dementia outcomes in older adults. Preliminary data suggest that clinical notes and machine learning (ML) algorithms can be helpful to capture patients with early cognitive decline. However, identifying which patients with SCD are more likely to develop dementia is extremely challenging. During the K99 phase, the first aim will be to develop an informatics approach to identify a pre-dementia cohort (patients with evidence of a cognitive concern but no dementia). The second aim will identify the social and clinical characteristics of this cohort in the EHR, along with antecedent risk factors, and predictors for a dementia outcome. The two hypotheses are that 1) clinical conditions (eg, neuropsychiatric disorders, cardiovascular diseases, renal disease, respiratory infections, sleep disorders) and medications that deleteriously affect cognition will contribute to the initial appearance of cognitive decline; and 2) longitudinal, multimodal EHR data can be leveraged in ML models to stratify patients with high risk of dementia. To accomplish these goals, the applicant will leverage existing strengths in case identification, risk factor analyses, and prognostic modeling and gain additional knowledge and skills in three critical areas of training: (1) cognitive decline and ADRD, (2) clinical epidemiology, and (3) statistical methods. With the development of these skills, the applicant will be well positioned in the R00 phase to conduct the final aim: to study the antecedent risk factors and outcomes of SCD in a presumed SCD cohort (patients with both a subjective cognitive concern and normal performance on objective cognitive measures). Similar approaches to those used in the second aim will be employed to study the presumed SCD cohort. A highly innovative component of this project is the use of advanced artificial intelligence and large-scale EHR data for presumed SCD cohort identification, risk factor analyses, and early detection of dementia. The proposed study will provide some of the first insights into the characteristics and risk factors of SCD in the EHR, and predictors for dementia outcomes in SCD. For the applicant, this program will support a rapid transition to independence through a short period of intensive training and mentorship, which will seamlessly intertwine with the aims of the proposed project.
