Using novel natural language processing approaches to examine age-friendly communication about nursing Nome residents with dementia

Abstract Background and Objectives The Age-Friendly Health Systems model has emerged as a geriatric care model focusing on the 4Ms: what matters, mobility, mentation, and medications. Representation of the 4Ms in interdisciplinary communication could have implications for outcomes including avoidable nursing home-to-hospital transfers. The objective of this article was to explore the association between the 4Ms found in text messages and avoidable transfer of residents with and without Alzheimer’s Disease and Related Dementias (ADRD). Research Design and Methods The researchers merged data from two primary datasets: (a) text messages (n = 30,066) between nursing home (NH) healthcare workers from an HIPAA-compliant messaging platform and (b) resident transfer data (n = 3,687) from NHs (n = 16) that participated in a 5-year (2016–2020) Centers for Medicare and Medicaid Services demonstration project. They used natural language processing to extract 4M terms from text messages and fit generalized linear mixed models to estimate avoidable NH-to-hospital transfer, controlling for resident characteristics, NH characteristics, and the 4Ms. Results Merged data contained 1,031 observations grouped by temporal proximity to the transfer date. Cardiopulmonary resuscitation status, late-stage ADRD, NH bed size, and location were associated with avoidable NH-to-hospital transfer of residents. Text messages containing terms representing mentation and mobility were also associated with avoidable NH-to-hospital transfers. Discussion and Implications These results suggest that natural language processing can be used to identify components of age-friendly care in unstructured data. Association between resident level and nursing home factors and avoidable transfers should be considered as nursing homes implement strategies to reduce avoidable transfer of residents with ADRD.