This repository corresponds to manuscript as cited.
Lee AK, Diaz-Ramirez LG, Boscardin WJ, Smith AK, Lee SJ. J Am Geriatr Soc. 2022 July 6. https://doi.org/10.1111/jgs.17932
Title: A comprehensive prognosis tool for older adults: Predicting Death, ADL Disability, and walking Disability simultaneously.
Authors: Alexandra K. Lee, L. Grisell Diaz-Ramirez, W. John Boscardin, Alexander K. Smith, Sei J. Lee
Volume: 70, 2022, PMID:35792836 https://doi.org/10.1111/jgs.17932
Many clinical and financial decisions for older adults depend on the future risk of disability and mortality. Prognostic tools for long-term disability risk in a general population are lacking. We aimed to create a comprehensive prognostic tool that predicts the risk of mortality, of activities of daily living (ADL) disability, and walking disability simultaneously using the same set of variables.
We conducted a longitudinal analysis of the nationally-representative Health and Retirement Study (HRS). We included community-dwelling adults aged ≥70 years who completed a core interview in the 2000 wave of HRS, with follow-up through 2018. We evaluated 40 predictors encompassing demographics, diseases, physical functioning, and instrumental ADLs. We applied novel methods to optimize three models simultaneously while prioritizing variables that take less time to ascertain during backward stepwise elimination. The death prediction model used Cox regression and both the models for walking disability and for ADL disability used Fine and Gray competing-risk regression. We examined calibration plots and generated optimism-corrected statistics of discrimination using bootstrapping. To simulate unavailable patient data, we also evaluated models excluding one or two variables from the final model.
In 6646 HRS participants, 2662 developed walking disability, 3570 developed ADL disability, and 5689 died during a median follow-up of 9.5 years. The final prognostic tool had 16 variables. The optimism-corrected integrated area under the curve (iAUC) was 0.799 for mortality, 0.685 for walking disability, and 0.703 for ADL disability. At each percentile of predicted mortality risk, there was a substantial spread in the predicted risks of walking disability and ADL disability. Discrimination and calibration remained good even when missing one or two predictors from the model. This model is now available on ePrognosis
Given the variability in disability risk for people with similar mortality risks, using individualized risks of disabilities may inform clinical and financial decisions for older adults.
disability; mortality; prediction.
link: https://agsjournals.onlinelibrary.wiley.com/doi/10.1111/jgs.17932