# Analysis of influencing factors and prediction of falls among rural older adults in China based on a nomogram model

## TL;DR

This study identifies key factors that increase fall risk in rural older adults in China and builds a predictive model to help identify high-risk individuals.

## Contribution

A novel nomogram model is developed for predicting falls in rural older adults using factors like age, anxiety, and lifestyle.

## Key findings

- The fall prevalence among rural older adults was 24.34%.
- The nomogram model showed good discriminative performance with an AUC of 0.7215 in the training set.
- Internal validation confirmed the model's robustness with an AUC of 0.703 in the validation set.

## Abstract

To explore rural fall determinants and build a nomogram for high-risk identification. This study aims to preliminarily identify high-risk individuals and provide reference for subsequent intervention research.

Stratified multistage cluster sampling was used: one city from each of the northern, central, and southern regions of Anhui Province, one county per city, and six villages per county (18 villages total). Township hospital staff and village committee staff helped recruit participants who completed a face-to-face questionnaire and physical tests. 1,546 rural adults aged ≥60 years were enrolled. Inclusion: ≥60 years, local residence ≥6 months, communicable. Exclusion: severe illness or bedridden. A fall was explicitly defined as “an event in which the participant unintentionally came to rest on the ground, floor, or a lower level,” excluding incidents caused by sudden paralysis, stroke, or violent impact. Participants were randomly split 8:2 into training (n = 1,208) and validation (n = 338) sets. Univariate tests (Mann–Whitney/Kruskal–Wallis) screened variables; those with p < 0.05 entered multivariate logistic regression with backward stepwise selection to build the nomogram.

The observed fall prevalence was 24.34% in the study population. From the univariate and multivariate analyses of the training set, five variables were identified: age, anxiety, frailty, living arrangement, and frequency of coarse grain consumption. These variables were incorporated into the nomogram model, which exhibited an area under the ROC curve (AUC) of 0.7215 (95% CI:0.690–0.753), indicating good discriminative performance. The calibration curve demonstrated high calibration accuracy. Internal validation of the nomogram model using the validation set yielded an AUC of 0.703 (95% CI:0.644–0.762), reflecting robust discriminative ability. The Hosmer-Lemeshow test indicated good calibration in both the training set (p = 0.38) and the validation set (p = 0.08) suggesting that there is no statistically significant difference between the predicted and observed probabilities, and confirming good calibration.

The nomogram we built–incorporating age, anxiety, frailty, living arrangement, and frequency of coarse-grain intake–offers a visual risk estimation tool for estimating fall risk in rural older adults. It can readily flag high-risk individuals and provide clues for the follow-up targeted health management or intervention research.