Integrating preoperative frailty into super-early and early perioperative assessment enhances prediction of delirium in patients undergoing elective craniotomy: a secondary analysis of a prospective cohort study.

BACKGROUND

Postoperative delirium (POD) is a common complication in neurosurgical patients, leading to worse outcomes. However, effective super-early prediction models incorporating preoperative frailty remain unclear.

METHODS

A secondary analysis of 800 adults undergoing elective craniotomy at Beijing Tiantan Hospital (2017-2018) was conducted. POD was assessed using CAM-ICU on postoperative days 1-3. Preoperative frailty was measured by the 5-factor modified Frailty Index (mFI-5).Four models: super-early (only preoperative variables), early (additionally adding intraoperative variables) and respectively further integrating mFI-5, were developed by logistic regression. Model performance was evaluated by the area under receiver operating characteristic curve (AUC), calibration curves and decision curve analysis (DCA).The incremental value of adding frailty was evaluated by net reclassification improvement (NRI) and integrated discrimination improvement (IDI). SHAP analysis assessed feature importance.

RESULTS

POD occurred in 157 patients (19.6%). Multivariate analysis identified four preoperative variables (age, educational level, nature of the intracranial lesions, and supratentorial lesions) and three intraoperative variables(operation duration, blood loss and hypotension) as independent risk factors. The early model with frailty had the highest AUC(0.843, 95% CI: 0.810-0.875), but the super-early model with mFI-5 also achieved an AUC of 0.813 (95% CI: 0.777-0.849), comparable to the early model without frailty (AUC: 0.804, 95% CI: 0.768-0.840; P = 0.582).Calibration curves demonstrated good agreement between predicted and observed POD risks, and DCA confirmed the clinical utility of all four models. Adding mFI-5 significantly improved both models and SHAP confirmed frailty as the top predictor.

CONCLUSIONS

Preoperative frailty substantially enhances POD prediction. An super-early model integrating mFI-5 enables timely risk stratification before surgery, supporting proactive delirium prevention.