Development and Validation of a Nomogram for Predicting Frailty Risk Among Older Patients With Ischaemic Stroke

ABSTRACT

Aim

To investigate the risk factors associated with frailty in older patients with ischaemic stroke, develop a nomogram and apply it clinically.

Design

A cross-sectional study.

Methods

Altogether, 567 patients who experienced ischaemic strokes between March and December 2023 were temporally divided into training (n = 452) and validation (n = 115) sets and dichotomised into frail and non-frail groups using the Tilburg Frailty Indicator scale. In the training set, feature selection was performed using least absolute shrinkage and selection operator regression and random forest recursive feature elimination, followed by nomogram construction using binary logistic regression. Internal validation was performed through bootstrap re-sampling and the validation set was used to assess model generalisability. The receiver operating characteristic curve, Hosmer–Lemeshow test, Brier score, calibration curve, decision curve analysis and clinical impact curve were used to evaluate nomogram performance.

Results

The prevalence of frailty was 58.6%. Marital status, smoking, history of falls (in the preceding year), physical exercise, polypharmacy, albumin levels, activities of daily living, dysphagia and cognitive impairment were predictors in the nomogram. Receiver operating characteristic curve analysis indicated outstanding discrimination of the nomogram. The Hosmer–Lemeshow test, calibration curve and Brier score results confirmed good model consistency and predictive accuracy. The clinical decision and impact curve demonstrated notable clinical utility. This free, dynamic nomogram, created for interactive use and promotion, is available at: https://dongdongshen.shinyapps.io/DynNomapp/.

Conclusion

This nomogram may serve as an effective tool for assessing frailty risk in older patients with ischaemic stroke.

Relevance to Clinical Practice

The nomogram in this study may assist healthcare professionals in identifying high-risk patients with frailty and understanding related factors, thereby providing more personalised risk management.

Reporting Method

TRIPOD checklist.

Patient or Public Contribution

No patient or public contribution.