# Laboratory frailty index improves prediction of in-hospital falls among older adults

## TL;DR

A laboratory-based frailty index improves the accuracy of predicting in-hospital falls among older adults when added to a traditional tool.

## Contribution

The study introduces a laboratory-based frailty index that enhances fall risk prediction in older hospitalized patients.

## Key findings

- FI-lab was independently associated with in-hospital falls after adjusting for STRATIFY.
- Adding FI-lab to STRATIFY increased the area under the ROC curve from 0.674 to 0.715.
- The net reclassification improvement was 0.413, indicating better risk classification.

## Abstract

To explore the association between Frailty Index based on laboratory tests (FI-lab) and in-hospital fall risk in older adults, and to explore whether incorporating FI-lab improves the predictive accuracy of a traditional fall risk prediction tool.

We conducted a retrospective cohort study using electronic medical records from patients aged ≥ 60 years who were admitted to Nagoya University Hospital in 2020. We assessed fall risk using the St. Thomas’s Risk Assessment Tool in Falling Elderly Inpatients (STRATIFY). We calculated FI-lab based on 35 common laboratory parameters tested on admission. Each fall was reported prospectively by nurses through computer-based incident report forms. The relationship between FI-lab and in-hospital falls was analyzed using multivariate binomial logistic regression. Predictive performance was compared using the area under the receiver operating characteristic curve (AUROC) and net reclassification improvement (NRI). Missing data were not imputed and internal validation used 1000-bootstrap optimism-correction.

Data for 5984 patients were included (mean age 73 years, 63.5% male). The mean FI-lab score was 0.31 ± 0.16. Falls occurred in 175 patients (2.9%) during a median hospital stay of 9 days. FI-lab was associated with falls independently of STRATIFY. Adding FI-lab to STRATIFY significantly improved its predictive accuracy, increasing AUROC from 0.674 to 0.715 (p = 0.018), with NRI of 0.413 (p < 0.001). Calibration slope after internal validation was 0.97.

FI-lab on admission was independently associated with in-hospital fall risk and improved the predictive ability of STRATIFY. FI-lab could be a valuable component in more accurate fall prediction.

The online version contains supplementary material available at 10.1007/s40520-025-03090-9.