Estrategia diagnóstica temprana para infecciones bacterianas del sistema nervioso central después de neurocirugía: un...

Accurate diagnosis of post-neurosurgical bacterial infection of central nervous system is challenging due to the nonspecific nature of clinical signs and cerebrospinal fluid (CSF) parameters, which often overlap with sterile postoperative inflammation. This study aimed to develop and validate a stepwise diagnostic strategy integrating readily available clinical and basic laboratory indicators to improve the early identification of post-neurosurgical bacterial infection of the central nervous system. A retrospective cohort study was conducted at Tianjin Huanhu Hospital, a tertiary neurosurgical center, from October 2018 to June 2025. We enrolled 176 patients suspected of post-neurosurgical bacterial infection of the central nervous system who underwent CSF metagenomic next-generation sequencing (mNGS). Six diagnostic prediction models, combining clinical features (fever, altered mental status) and CSF parameters (white blood cell count, glucose levels), were constructed. Their diagnostic performance was evaluated against a composite reference standard (mNGS, culture, and clinical treatment response) using receiver operating characteristic analysis. The area under the curve (AUC), sensitivity, and specificity were calculated. Among the 6 models, two demonstrated superior performance. Model 5 (T > 38.0°C + [CSF white blood cell ≥ 2000 × 106/L OR CSF glucose < 2.2 mmol/L OR CSF/Blood glucose ratio < 0.4]) achieved an AUC of 0.768. Notably, Model 6 (T > 39.0°C + Altered Mental Status + Intermittent Fever), relying solely on clinical indicators, achieved a comparable AUC of 0.769. For individual parameters, a high fever threshold (T > 39.8°C) and profoundly low CSF glucose (<1.01 mmol/L) showed high specificities of 99%and 97%, respectively, for ruling in infection. A diagnostic strategy combining severe clinical manifestations (high fever and altered mental status) with high-threshold CSF parameters enables effective risk stratification for post-neurosurgical bacterial infection of the central nervous system. The high performance of a purely clinical model (Model 6) offers a valuable tool for rapid bedside assessment, especially in resource-limited settings. Future prospective, multicenter studies are recommended to validate these algorithms and further refine variable definitions for broader clinical application.