Suicide risk is highest among older adults. Since 70% of aging adults will require some form of long-term care (LTC), residential LTC potentially constitutes an important setting for preventing suicide. We applied data science methodologies to analyze textual narratives from suicide decedents in the National Violent Death Reporting System (NVDRS) (N= 50,438 deaths aged 50+, 2003–2015). Narratives were analyzed using supervised machine learning (ML) algorithms. Algorithms were trained with confirmed cases related (true positive) and not related (true negative) to LTC verified from the textual narratives via keywords related to transitioning/residing in LTC (e.g., “nursing,” “assisted,” “care”, “moving,” etc.). ML results were compared to NVDRS-identified data on location and circumstances at the time of death. We find evidence of misclassification of suicide location, which results in an underestimate of the frequency of suicide related to LTC. Results have implications for improving surveillance of suicide risk in later life.
展开▼
