Predicting the risk of nursing home placement of elderly persons using a population-based stratification score

Objective: To develop and validate a novel score predictive of nursing home placement in elderly. Study design: Population-based case-control study based on healthcare utilization databases of Lombardy, a region of Northern Italy. Methods: The 2.4 million citizens aged ≥65 years who on January 1, 2018 lived outside nursing home formed the target population. Cases were citizens who experienced nursing home admission (the outcome of interest) until December 31, 2019. Cases were matched 1:1 by gender, age, and municipality of residence to one control. Conditional logistic regression was fitted to select candidate predictors (the exposure to 69 clinical conditions and 11 social and healthcare services) independently associated with the outcome. The model was built from the 26,156 cases, and as many controls (training set), and applied to a validation set (15,807 case-control couples). Predictive performance was assessed by discrimination and calibration. Results: Twenty-one factors were identified as predictive of nursing home admission and were included in the “Elderly Nursing Home Placement” (ENHP) score. Mental health disorders and chronic neurological illnesses contributed most to prediction of nursing home admission. ENHP performance showed an area under the receiver operating characteristic curve of 0.77 and a remarkable calibration of observed and predicted outcome risk. Conclusions: A simple score derived from data used for public health management may reliably predict the risk of nursing home placement in elderly. Its use by healthcare decision makers allows to accurately identify high-risk individuals who need home services, thereby avoiding admission to nursing homes.