Accuracy of language data in home care and long-term care assessments in Ontario, Canada

Design — Descriptive cross-sectional study using linked health administrative data.

Setting and participants — The study included adults (aged ≥20 years) in Ontario, Canada, who answered the Canadian Community Health Survey between 2011 and 2018 and had assessments in home care and long-term care databases.

Methods — We linked survey respondents to assessment data from home care databases [Resident Assessment Instrument–Home Care (RAI-HC) and International Resident Assessment Instrument (InterRAI)] or from the long-term care database [Continuing Care Reporting System (CCRS)]. The answer to “What is the language you speak most often at home?” was our reference standard, identified from the Canadian Community Health Survey. We reported the data accuracy and agreement levels for the language variables in the health care databases compared with our reference standard.

Results — Among survey respondents, we identified 8532 individuals with assessments in RAI-HC, 10,690 in InterRAI, and 4667 in CCRS. The accuracy of identifying true language preference was 88% in RAI-HC, 79% in InterRAI, and 82% in CCRS. The RAI-HC language variables had a sensitivity of 72.0% (95% CI, 68.2-75.7) for identifying a non-English language, InterRAI variables had a sensitivity of 66.9% (95% CI, 63.6-70.2), and CCRS variables had a sensitivity of 68.1% (95% CI, 63.1-73.0). The specificity of these variables for identifying a non-English language was high across all 3 databases (98.4%, 98.6%, and 98.5%) and very high for the “needs interpreter” variable [99.7% (95% CI, 99.5-99.9)].

Conclusions and implications — The accuracy of identifying true language preference was moderately sensitive and highly specific in home care and long-term care databases in Ontario, Canada. Our findings support the use of these variables to identify linguistic groups.