Ontology highlight
ABSTRACT: Background
Limitations in instrumental activities of daily living (IADL) hinder a person's ability to live independently in the community and self-manage their conditions, but its impact on hospital readmission has not been firmly established.Objective
To test the importance of IADL dependency as a predictor of 30-day readmissions and quantify its impact relative to other morbidities.Design
A retrospective cohort study of the population-based Health and Retirement Study linked to Medicare claims data. Random forest was used to rank each predictor variable in terms of its ability to predict readmission. Classification and regression tree (CART) was used to identify complex multimorbidity combinations associated with high or low risk of readmission. Generalized linear regression was used to estimate the adjusted relative risk of readmission for IADL limitations.Subjects
Hospitalizations of adults age 65 and older (n = 20,007), from 6617 unique subjects.Main measures
The main outcome was 30-day all-cause unplanned readmission. The main predictor of interest was self-reported IADL limitation. Other key predictors were self-reported complex multimorbidity including chronic diseases, geriatric syndromes, and activities of daily living (ADL) limitations, along with demographic, socioeconomic, and behavioral factors.Key results
The overall 30-day readmission rate in the study was 16.4%. Random forest analysis ranked ADLs and IADL limitations as the two most important predictors of 30-day readmission. CART identified hospitalizations of patients with IADL limitations and diabetes as a subgroup at the highest risk of readmission (26% readmitted). Multivariable regression analyses showed that ADL limitations were associated with 1.17 (1.06-1.29) times higher risk of readmission even after adjusting for other patient covariates. Risk prediction was modest though for even the best model (AUC = 0.612).Conclusions
IADL limitations are key predictors of 30-day readmission as demonstrated using several machine learning methods. Routine assessment of functional abilities in hospital settings could help identify those most at risk.
SUBMITTER: Schiltz NK
PROVIDER: S-EPMC7573020 | biostudies-literature |
REPOSITORIES: biostudies-literature