Identifying Care Home Residents in Electronic Health Records - An OpenSAFELY Short Data Report

Schultze, Anna; Bates, Chris; Cockburn, Jonathan; MacKenna, Brian; Nightingale, Emily; Curtis, Helen J; Hulme, William J; Morton, Caroline E; Croker, Richard; Bacon, Seb; McDonald, Helen I; Rentsch, Christopher T; Bhaskaran, Krishnan; Mathur, Rohini; Tomlinson, Laurie A; Williamson, Elizabeth J; Forbes, Harriet; Tazare, John; Grint, Daniel J; Walker, Alex J; Inglesby, Peter; DeVito, Nicholas J; Mehrkar, Amir; Hickman, George; Davy, Simon; Ward, Tom; Fisher, Louis; Evans, David; Wing, Kevin; Wong, Angel YS; McManus, Robert; Parry, John; Hester, Frank; Harper, Sam; Evans, Stephen JW; Douglas, Ian J; Smeeth, Liam; Eggo, Rosalind M; Goldacre, Ben

Background: Care home residents have been severely affected by the COVID-19 pandemic. Electronic Health Records (EHR) hold significant potential for studying the healthcare needs of this vulnerable population; however, identifying care home residents in EHR is not straightforward. We describe and compare three different methods for identifying care home residents in the newly created OpenSAFELY-TPP data analytics platform. Methods: Working on behalf of NHS England, we identified individuals aged 65 years or older potentially living in a care home on the 1st of February 2020 using (1) a complex address linkage, in which cleaned GP registered addresses were matched to old age care home addresses using data from the Care and Quality Commission (CQC); (2) coded events in the EHR; (3) household identifiers, age and household size to identify households with more than 3 individuals aged 65 years or older as potential care home residents. Raw addresses were not available to the investigators. Results: Of 4,437,286 individuals aged 65 years or older, 2.27% were identified as potential care home residents using the complex address linkage, 1.96% using coded events, 3.13% using household size and age and 3.74% using either of these methods. 53,210 individuals (32.0% of all potential care home residents) were classified as care home residents using all three methods. Address linkage had the largest overlap with the other methods; 93.3% of individuals identified as care home residents using the address linkage were also identified as such using either coded events or household age and size. Conclusion: We have described the partial overlap between three methods for identifying care home residents in EHR, and provide detailed instructions for how to implement these in OpenSAFELY-TPP to support research into the impact of the COVID-19 pandemic on care home residents.