Development and internal validation of a frailty risk stratification model for older patients with spinal...

Frailty is a common geriatric syndrome in older adults and is closely associated with adverse outcomes such as prolonged hospitalization, functional decline, and mortality. Older patients with spinal tuberculosis often experience chronic inflammatory consumption, malnutrition, and activity limitation, which may contribute to a higher frailty burden. However, evidence on early identification of Fried frailty in this specific population remains limited.

To develop and bootstrap internally validate a single-center model for identifying Fried frailty in older patients with spinal tuberculosis using routinely available early in-hospital clinical information.

This was a single-center retrospective observational study including older patients with spinal tuberculosis who were hospitalized between January 2024 and January 2026. Variables available at admission or during the early stage of hospitalization, including demographic characteristics, nutritional and inflammatory laboratory indicators, activities of daily living, and imaging burden, were collected. Frailty status was assessed using the Fried frailty phenotype. Candidate predictors were prespecified based on the literature, clinical judgment, and early availability at admission. A multivariable logistic regression model was then used to develop the stratification model and construct a nomogram. Model performance was assessed using discrimination, calibration, and decision curve analysis, and bootstrap resampling was used for internal validation.

Among the 323 patients included, 260 were identified as frail, yielding a frailty prevalence of 80.50%. Multivariable analysis showed that age (OR = 1.124, 95% CI: 1.057-1.196), non-married status (OR = 6.078, 95% CI: 2.902-12.730), and the number of involved vertebrae (OR = 1.278, 95% CI: 1.014-1.610) were associated with higher odds of frailty, whereas body mass index (BMI) (OR = 0.835, 95% CI: 0.746-0.934) and Barthel Index score (OR = 0.956, 95% CI: 0.934-0.978) were associated with lower odds of frailty. The area under the receiver operating characteristic curve of the model was 0.826 (95% CI: 0.779-0.873). At the Youden-derived cutoff value of 0.835, the sensitivity was 0.665 and the specificity was 0.905. Bootstrap internal validation showed a C-index of 0.826, with an optimism-corrected C-index of 0.812. The Hosmer-Lemeshow test yielded a p-value of 0.215, and the Brier score was 0.125, indicating good calibration. Decision curve analysis suggested possible net benefit across a broad range of threshold probabilities; however, this exploratory finding should not be interpreted as evidence that use of the model improves clinical outcomes.

The nomogram developed in this single-center retrospective study may serve as an auxiliary screening and early in-hospital risk-stratification tool for identifying Fried frailty in older patients with spinal tuberculosis. However, it should not be interpreted as external validation, validation of the Fried phenotype or individual clinical instruments, or evidence that use of the nomogram improves clinical outcomes. External validation and prospective impact studies are needed before broader clinical application.