Improving One-year Fall Risk Screening Methods Using Longitudinal Changes In Modifiable Predictors

PURPOSE: Improving methods of identifying older adults at risk for falls is an important public priority due to the major consequences of falls for the health and well-being of older adults. The established STEADI toolkit is widely used for clinical fall risk screening, but more robust models are needed to improve the precision of risk estimation. The purpose of the study was to develop methods appropriate for evaluating changes in one-year fall risk or in response to interventions. METHODS: The study using longitudinal data from the National Health and Aging Trends Study. The study build on a previously developed continuous fall risk model, but employs a transition matrix model to predict likelihood of a fall in each year of follow-up. A total of 5691 unique individuals and 23402 year-over-year risk predictions were used to develop the model. Indicators explored in the model included falls in the past year, physical function, subjective fall risk factors and participant demographics. RESULTS: The developed model had a positive predictive value of 0.576 and a negative predictive value of 0.838. The low risk threshold demonstrated few false negatives (sensitivity = 0.895) and the high risk threshold demonstrated few false positives (specificity = 0.877) when applied to a holdout sample. Model coefficients showed limited interaction between fall history and function, suggesting that changes in function have a similar impact on changes in fall risk, with fall history impacting absolute risk. An interaction between fall risk and gender was noted, suggesting that fall history may increase a female’s risk for falls in the following year more than it does for male’s. CONCLUSIONS: The newly developed model outperforms the previous model and provides more utility for screening and research. The creation of high and low-risk cutoff scores improves sensitivity and specificity, offering advantages for clinical and community-based screening efforts. Individual predictor variable coefficients suggest that changes in function impacts changes in risk independent of fall history, though fall history is important for determining absolute risk. Notably, age dropped out of the final model suggesting that increases in fall risk with age are due to modifiable functional factors, rather than as an unavoidable consequence of aging. Supported by: U01CE003490- 01: Process and Outcome Evaluation of the Walk with Ease program for Fall Prevention