Development and internal and external validation of a nomogram model for frailty risk among hospitalised older people using comprehensive geriatric assessment data

• Published in: BMC Geriatrics
• Main Authors: Hong Lyu, Wenhui Jiang
• Format: Article
• Language: English
• Published: BMC 2023-11-01
• Subjects: Frailty; Frailty prediction; Frailty prevention; Frailty Nomogram prediction model; Comprehensive geriatric assessment
• Online Access: https://doi.org/10.1186/s12877-023-04426-8

Abstract Background Currently, there are few such studies about establishing the frailty prediction model on the basis of the research on the factors influencing frailty in older patients, which can better predict frailty and identify its risk factors, and then guide the formulation of intervention measures precisely, especially in the hospital setting in China. Meanwhile, comprehensive geriatric assessment (CGA) can provide measurable and substantial health improvements for frail older people. The study aimed to develop a nomogram model for frailty risk among hospitalised older people using CGA data and validated its predictive performance for providing a basis for medical staff to grasp the risk and risk factors of older inpatients’ frailty conveniently and accurately, and to formulate reasonable nursing intervention plan. Methods We used CGA data of individuals over age 64. Demographic characteristics, geriatric syndrome assessment, and frailty assessment based on the FRAIL scale were included as potential predictors. Significant variables in univariate analysis were used to construct risk models by logistic regression analysis. We used the root mean square (rms) to develop the nomogram prediction model for frailty based on independent clinical factors. Nomogram performance was internally validated with Bootstrap resampling. The final model was externally validated using an independent validation data set and was assessed for discrimination and calibration. Results Data from 2226 eligible older inpatients were extracted. Five hundred sixty-two older inpatients (25.25%) suffered from frailty. The final prediction model included damaged skin, MNA-SF, GDS-15, Morse risk scores, hospital admission, ICI-Q-SF, Braden score, MMSE, BI scores, and Caprini scores. The prediction model displayed fair discrimination. The calibration curve demonstrated that the probabilities of frailty predicted by the nomogram were satisfactorily matched. Conclusions The prediction model to identify hospitalised older people at high risk for frailty using comprehensive geriatric assessment data displayed fair discrimination and good predictive calibration. Therefore, it is inexpensive, easily applied, and accessible in clinical practice, containing variables routinely collected and readily available through consultation. It will be valuable for grasp older inpatients at high risk of frailty and risk factors in hospital setting to guide the formulation of intervention measures precisely for reversing and preventing frailty.