Association between the postoperative glycemic variability and mortality after craniotomy: a retrospective cohort study and development of a mortality prediction model.

Background: Glycemic variability (GV), typically quantified by the coefficient of variation (CV) and the root mean square of successive differences (rMSSD), has been recognized as a potential predictor of poor outcomes in critically ill patients. However, its prognostic value in neurosurgical populations remains unclear. This study investigated the association between postoperative GV and mortality following craniotomy. Methods: We retrospectively analyzed 1,969 adult ICU patients who underwent cranial surgery. GV was measured using both CV and rMSSD calculated from blood glucose values during the ICU stay. The primary outcome was 28-day all-cause mortality; the secondary outcome was 90-day mortality. Multivariable Cox regression, restricted cubic splines, threshold effect analysis, and mediation analysis via blood urea nitrogen (BUN) were conducted. A Random Survival Forest (RSF) model was developed using machine learning and interpreted with SHAP values. Results: Higher GV, as reflected by both elevated CV and rMSSD, was independently associated with increased 28-day and 90-day mortality (CV per 10-unit HR: 1.20; rMSSD per 10-unit HR: 1.02; all P < 0.01). BUN partially mediated the association between GV and mortality. GV outperformed traditional clinical scores (SOFA, GCS, CCI) in ROC analysis (CV AUC = 0.72). The RSF model achieved an AUC of 0.841 and identified GV metrics as top predictors. Conclusions: Postoperative glycemic variability, assessed by CV and rMSSD, is an independent and modifiable predictor of short- and mid-term mortality following craniotomy. These findings highlight the clinical importance of GV in postoperative risk stratification and support its integration into neurosurgical critical care.