A novel dynamic nomogram based on clinical features and laboratory indicators for diagnosis of post-neurosurgery intracranial infection

Abstract

Objective: Intracranial infection is a serious complication after neurosurgery. However, the early diagnosis of post-neurosurgical intracranial infection (PNICI) remains challenging. The purpose of this study was to compare clinical characteristics and common laboratory indicators in patients with and without intracranial infections after neurosurgery and construct a diagnostic model of PNICI and assess its diagnostic efficacy.

Methods: A total of 623 patients who underwent neurosurgery from January 2018 to October 2021 were enrolled and divided into a training set and a validation set. SPSS 22.0 software was used to compare the differences in basic information and laboratory examination results between the two groups to screen out valuable indicators. Subsequently, a nomogram for the diagnosis of PNICI was established. Then, the receiver operating characteristic (ROC) curve, calibration diagram, and decision curve analysis (DCA) were performed to evaluate the discriminative ability, consistency, and clinical usefulness of the nomogram.

Results: The diagnostic model of PNICI consisted of seven variables: meningeal irritation, fever, postoperative drainage, cerebrospinal fluid (CSF) white blood cells, CSF chlorine, the CSF/blood glucose ratio, and blood neutrophil percentage. The model achieved an area under the ROC curve of 0.958 in the training set and 0.966 in the validation set. At the optimal cutoff of 0.397, the training set demonstrated 90.4% sensitivity and 90.8% specificity. The calibration curves and DCA curves of the nomogram demonstrated that the model exhibited good goodness of fit and showed a net benefit from its use.

Conclusions: We developed an easily applicable nomogram using routinely available indicators. This tool enables early risk stratification for PNICI, facilitating timely interventions that may reduce infection-related complications. However, multicenter prospective validation data are required to further confirm the clinical utility.

Keywords: diagnosis; intracranial infection; neurosurgery; nomogram.

Figure 1

Flow chart of the present study.

Figure 2

Nomogram for the diagnosis of PNICI. Draw vertical lines from each variable value to the “Points” axis. Subsequently, sum all points to obtain total points. Align total points with the probability axis to determine individual PNICI risk. Defining categorical variables, 0 = no, 1 = yes. (a) The nomogram of PNICI and (b) the web-based dynamic nomogram (https://nomogram-pnici.shinyapps.io/DynNomPNICI/). CSF_WBC: CSF white blood cells; GluR: the CSF/blood glucose ratio; B_NEU%: blood neutrophil proportions.

Figure 3

Verification of the diagnostic efficiency of the PNICI model. (a) and (b) are the ROC curve of the model. (a) ROC curve of the training set, the area under the ROC curve (AUC) was 0.958 (sensitivity 90.4%, specificity 90.8%). (b) ROC curve of the validation set, the AUC was 0.966 (sensitivity 94.8%, specificity 90.5%). (c) and (d) are the calibration curve of the model. The observed PNICI is plotted on the y-axis and the probability of PNICI measured by nomogram is plotted on the x-axis. The dotted line represents the apparent curve. The solid line is calculated by bootstrapping (resample: 1,000) and represents the discrimination of PNICI using a nomogram, and the diagonal dashed line represents the ideal line showing the diagnostic probability is consistent with the observed probability. (c) Training set’s calibration curve, mean absolute error = 0.009. (d) Validation set’s calibration curve, mean absolute error = 0.006.

Figure 4

Decision curve analysis of the nomogram for PNICI. The y-axis represents the net benefit and the x-axis indicates the threshold probability. The red solid line represents the net benefit of the model. The gray line represents the assumption that all patients are PNICI. The black line represents the assumption that no patients are PNICI. (a) Training set’s DCA curve and (b) validation set’s DCA curve.