Computational phenotyping of cancer-related cognitive impairment in older adults

PROJECT SUMMARY/ABSTRACT Cancer-related cognitive impairment (CRCI) involves cognitive changes and issues that cancer patients experience before, during, and after cancer treatment. CRCI is one of the most common symptoms reported by patients at different treatment stages: 30% before, 75% during, and 35% after chemotherapy. However, significant barriers persist in understanding and evaluating the prevalence, mechanisms, and management of CRCI in older patients. Even though randomized controlled trials stand as the gold standard method for evidence evaluation, their traditional design faces challenges in assessing CRCI because cognitive function is often not considered an endpoint. Additional factors, such as a limited patient sample size for older adults, variations in patient characteristics, and treatment protocols, can all lead to a lack of representativeness. The goal of this study is to improve the existing data collection and detection of CRCI for older adults in real-world electronic health records (EHR) data by developing ‘CRCIPheno’, a novel computational phenotyping approach that utilizes machine learning and natural language processing (NLP) techniques to systematically extract and standardize various CRCI-related characteristics found within clinical notes and neuroimaging reports. The validated CRCIPheno will systematically identify episodes of CRCI in a large population of older adults who receive care from multiple healthcare institutions and will be deployed to the Rochester Epidemiology Project (REP) -- a unique example of a medical records linkage system in the United States with almost half a century of activity. The project has the following specific aims: 1) develop, benchmark, and evaluate CRCIPheno to identify cancer-related cognitive impairment outcomes from EHRs, and 2) estimate the incidence of CRCI in older adults with colon cancer within the REP cohort. Accomplishing these aims will yield novel measures and techniques for identifying late-emerging effects and aging phenotypes among cancer survivors using real-world data. In addition to advancing the field of cancer and aging, this career development award will position the principal investigator as an independent researcher in the field of translational biomedical informatics who focuses on accelerating the secondary use of EHRs to support cancer and aging research.