Low perioperative hemoglobin as an independent detrimental predictor of functional outcome after basal ganglia hemorrhage: development and validation of a composite prognostic model

TL;DR

Low hemoglobin levels before and after surgery are linked to worse recovery in patients with basal ganglia hemorrhage, and a new model helps predict outcomes.

Contribution

A novel composite prognostic model integrating perioperative hemoglobin levels for predicting functional outcomes in basal ganglia hemorrhage patients.

Findings

Lower preoperative, postoperative, and mean hemoglobin levels were independently associated with poor outcomes.

The model showed strong predictive performance with an AUC of 0.84.

Calibration and decision curve analysis confirmed the model's clinical applicability.

Abstract

Basal ganglia hemorrhage (BGH) is a life-threatening neurosurgical emergency associated with substantial mortality and disability. Accurate postoperative prognosis assessment remains challenging due to multifactorial influences. Hemoglobin (HB), as the key determinant of oxygen delivery, may play a critical role in neurological recovery, yet the prognostic significance of perioperative HB fluctuations in BGH has not been fully elucidated.

A retrospective cohort of 213 surgically treated BGH patients from 2020 to 2023 was analyzed. Perioperative HB indices, including preoperative (Pre-HB), postoperative (Post-HB), and mean HB (Mean-HB) levels, were evaluated alongside clinical data. Functional outcome at 6 months was determined based on the modified Rankin Scale (mRS). Least absolute shrinkage and selection operator (LASSO) regression together with multivariate logistic regression models were utilized to screen for independent risk variables, followed by construction of a composite predictive model. Model discrimination, calibration, and evaluation of the model’s clinical applicability were conducted using receiver operating characteristic (ROC) analysis, calibration plots, and decision curve analysis (DCA).

Patients with poor prognosis exhibited significantly lower Pre-HB, Post-HB, and Mean-HB levels (all P < 0.05). Multivariate analysis confirmed these variables as independent predictors of adverse outcome. The proposed model provides a practical and data-driven tool that demonstrated good predictive performance (AUC = 0.84) in a single-center retrospective cohort. Calibration and DCA demonstrated good consistency and potential clinical applicability.

Perioperative declines in HB are independently associated with poor postoperative outcomes in BGH. The proposed HB-integrated model provides a reliable, dynamic tool for individualized risk prediction, facilitating precision perioperative management and optimized recovery strategies.

Figures