Evaluating the performance of administrative data elements used to document homelessness among psychiatric patients in Ontario, Canada: a validation study

Abstract

Housing information about patients admitted to hospitals is routinely documented in health administrative records, which are increasingly used to inform planning and research. This study aimed to evaluate the accuracy of these records in identifying people experiencing homelessness (PEH) at discharge from psychiatric hospitalizations. We linked a reference standard of individuals admitted to a designated psychiatric unit or provincial psychiatric hospital who consented to participate in the research arm of a program for PEH to healthcare administrative data from Ontario, Canada. We evaluated the performance of five variables within the Ontario Mental Health Reporting System (OMHRS) that directly capture homelessness status: an ICD-10-CA diagnosis of Z59.0 (homelessness); a postal code recorded as unknown or not applicable (indicating absence of a fixed address at the time of admission); living situation at admission recorded as homeless; usual residential status recorded as homeless; and planned or actual discharge destination recorded as homeless. The evaluation included 94 participants. Participants were a median age of 35 years (IQR, 28–46), 62.8% were male, and 33.0% reported being homeless for two or more years in the past decade. The best-performing individual variable accurately identified 30.9% (95% CI, 22.4–40.8%) of participants as experiencing homelessness. Considering any of the five variables improved identification to 43.6% (34.0-53.7%). Participants identified as homeless in OMHRS were slightly older (median age 36 vs. 33), more likely to be male (65.9% vs. 60.4%), and more often reported prolonged homelessness (≥ 2 years in the past decade: 41.4% vs. 26.4%). Despite multiple opportunities to document homelessness in psychiatric hospital records, more than half of individuals known to be experiencing homelessness at discharge were not identified as such in OMHRS administrative data, indicating systematic under-identification. Improving the documentation of housing status within psychiatric settings is necessary to ensure administrative data can reliably support care planning and health system decision-making.