Using longitudinal electronic health records to measure multilevel social disadvantage: Colorectal cancer screening among urban safety-net patients
Category: Cancer Health Disparities, Early Detection & Risk Prediction
Conference Year: 2018
Introduction. Social disadvantage predicts colorectal cancer (CRC) outcomes across the cancer care continuum for many populations and places. Social disadvantage is part of meaningful use electronic health records (EHR) requirements, but frequently is inadequately assessed with cross-sectional, sector-specific measures (e.g., income at cohort entry) that are strikingly homogenous for vulnerable populations. EHRs can be linked to external data sources to create more informative measures of social disadvantage at multiple levels. Purpose. By linking patient residential address in the EHR with external geospatial datasets, we investigate associations between novel measures of social disadvantage and CRC screening. Methods. We identified urban safety-net patients eligible and due for CRC screening from the Parkland-UT Southwestern PROSPR cohort. We used the EHR to assess one-time receipt of colonoscopy or fecal immunochemical test (FIT) screening in the 18 months following cohort entry (defined as a primary care visit). We geocoded and linked EHR data to housing and Census data to generate measures of social disadvantage at the parcel and block group level. We fit multilevel logistic regression models to control for patient sociodemographics, comorbidity, and healthcare utilization. Results. Among 32,965 safety-net patients, 45.1% received screening; there was limited block-group-level variation (e.g., 0.005). In adjusted models, measures of patient-level social disadvantage (e.g., sex ) and healthcare utilization were associated with CRC screening receipt. Of all nine measures in our patient-level housing disadvantage (e.g., value of housing) and neighborhood-level physical (e.g., vacancy) and social disadvantage (e.g., poverty) groups, only 3 were significant: residential mobility, zoning, and majority race. Conclusions. Address-based linkage of EHRs to external datasets can expand measurement of multilevel social disadvantage. Investigating social disadvantage in safety-net settings may be constrained by homogeneity (e.g., floor/ceiling effects). Therefore, more applications of these linkage methods and measures of patient housing, neighborhood physical, and neighborhood social disadvantage to more heterogeneous patient populations are needed.
Keywords: Neighborhoods; social disadvantage; electronic health record; geographic information systems (GIS); colorectal cancer screening