Related Experiment Video
Updated: Jan 10, 2026
Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
External Validation of the Saga Fall-Related Injury Risk Model and Exploration of Common Factors in Multiple
Shizuka Yaita1,2, Naoko E Katsuki1, Risa Hirata1
1Department of General Medicine, Saga University Hospital, Saga, Japan.
Purpose:
The Saga Fall-related Injury Risk Model (SFIRM) was developed in an acute care hospital to predict fall-related injuries based on six factors upon admission: age, sex, emergency transport, medical referral letters, history of falls, and bedriddenness ranks. This study aims to validate the applicability of the model across various hospitals through external validation using data from multiple hospitals. Additionally, the common predictors of fall-related injuries across these hospitals were explored.
Patients And Methods:
This multicenter, retrospective, observational study included patients aged 20 years and older who were admitted to 8 hospitals (chronic-care, acute-care, and tertiary acute-care) between April 2018 and March 2021. A calculated sample size of patients was selected and the area under the curve (AUC) of the SFIRM was determined for fall-related injuries during hospitalization. Multivariate analyses were conducted for each hospital using the surveyed factors as covariates and fall-related injuries as outcomes. The significant factors associated with fall-related injuries were compared across hospitals.
Results:
From 144,777 patients, 2376 were randomly sampled and analyzed. Among them, 51 patients (2.1%) experienced falls during hospitalization and 35 (1.5%) sustained fall-related injuries. The AUC of SFIRM was 0.617 (95% confidence interval 0.534-0.701). In multivariate analyses by hospital, age and bedriddenness ranks were significantly associated with fall-related injuries in five hospitals, whereas male sex, history of falls, and diabetes were significantly associated with fall-related injuries in four hospitals.
Conclusion:
The SFIRM demonstrated low discrimination in a population from various hospitals. The predictive models for fall-related injuries require redevelopment and validation to suit various hospitals. In the multivariate analyses across hospitals, age, bedriddenness ranks, male sex, history of falls, and diabetes mellitus were common and significant factors associated with fall-related injuries. These factors are most favorable for developing a predictive model for fall-related injuries.
