Development and Validation of a Predictive Model for Postoperative Intracranial Infections in Neurosurgery with Risk...

Currently, the diagnosis of postneurosurgical intracranial infection is mainly dependent on cerebrospinal fluid (CSF) bacterial culture, which has the disadvantages of being time-consuming, having a low detection rate, and being easily affected by other factors. These disadvantages bring some difficulties to early diagnosis. Therefore, it is very important to construct a nomogram model to predict the risk of infection and provide a basis for early diagnosis and treatment.

This retrospective study analyzed postneurosurgical patient data from the Fourth Affiliated Hospital of Harbin Medical University between January 2019 and September 2023. The patients were randomly assigned in an 8:2 ratio into the training cohort and the internal validation cohort. In the training cohort, initial screening of relevant indices was conducted via univariate analysis. Subsequently, the least absolute shrinkage and selection operator logistic regression identified significant potential risk factors for inclusion in the nomogram model. The model's discriminative ability was assessed using the area under the receiver operating characteristic curve, and its calibration was evaluated through calibration plots. The clinical utility of the model was determined using decision curve analysis and further validated by the internal validation cohort.

Multivariate logistic regression analysis of the training cohort identified 7 independent risk factors for postoperative intracranial infection: duration of postoperative external drainage (odds ratio [OR] 1.19, P = 0.005), continued fever (OR 2.11, P = 0.036), CSF turbidity (OR 2.73, P = 0.014), CSF pressure (OR 1.01, P = 0.018), CSF total protein level (OR 1.26, P = 0.026), CSF glucose concentration (OR 0.74, P = 0.029), and postoperative serum albumin level (OR 0.84, P < 0.001). Using these variables to construct the final model. The area under the receiver operating characteristic curve value of the model was 0.868 in the training cohort and 0.900 in the internal validation cohort. Calibration and the decision curve analysis indicated high accuracy and clinical benefit of the nomogram, findings that were corroborated in the validation cohort.

This study successfully developed a novel nomogram for predicting postoperative intracranial infection, demonstrating excellent predictive performance. It offers a pragmatic tool for the early diagnosis of intracranial infection.