Peng Wang1, Weizheng Song1, Shuwen Cheng1, Yongxiao Shuai1, Jiao Yang2, Shuang Luo3. 1. Department of Neurosurgery, Chengdu Fifth People's Hospital/Affiliated Chengdu No.5 People's Hospital of Chengdu University of TCM, Chengdu, 611130, China. 2. Department of Infection Control, Chengdu Fifth People's Hospital/Affiliated Chengdu No.5 People's Hospital of Chengdu University of TCM, Chengdu, 611130, China. 3. Department of Neurosurgery, Chengdu Fifth People's Hospital/Affiliated Chengdu No.5 People's Hospital of Chengdu University of TCM, Chengdu, 611130, China. luoshuang2000@126.com.
Abstract
BACKGROUND: Lumbar drainage (LD) is one of the common treatment techniques in neurosurgery. There is a risk of secondary meningitis when using this modality. We aim to predict the probability of the complication by designing a nomogram. METHODS: A retrospective study was conducted in a teaching hospital. Data were collected and LD-related meningitis (LDRM) was identified, mainly based on clinical manifestations and cerebrospinal fluid analysis. Univariate analysis was used to screen the risk factors, and binary logistic analysis was performed to build the prediction model, which was furtherly transferred into a nomogram. The prediction performance was evaluated by receiver operating characteristic (ROC) curve, Hosmer-Lemeshow test, and nomogram calibration plot. Internal validation was processed by using ordinary bootstrapping. RESULTS: A total of 273 patients who match the research criteria were enrolled, in which 37 cases (13.6%) were confirmed to have LDRM. Univariate analysis showed the risk factors included diabetes (p = 0.003), admission on surgical intensive care unit (p = 0.012), duration time (p < 0.001), site leakage (p < 0.001), and craniotomy (p < 0.001). In multivariate analysis, four of the variables were identified as independent risk factors to establish a prediction model, and a graphical nomogram was designed. The area under the ROC curve was 0.837, and the p value in the Hosmer-Lemeshow test was 0.610, with a mean absolute error in the calibration plot calculated as 0.022. The indices in the testing set were in good accordance with the original set when internal validation was performed. CONCLUSIONS: This is the first study to transform the prediction model of LDRM into a nomogram, which can be considered as a tool for clinicians to assess infection risk.
BACKGROUND: Lumbar drainage (LD) is one of the common treatment techniques in neurosurgery. There is a risk of secondary meningitis when using this modality. We aim to predict the probability of the complication by designing a nomogram. METHODS: A retrospective study was conducted in a teaching hospital. Data were collected and LD-related meningitis (LDRM) was identified, mainly based on clinical manifestations and cerebrospinal fluid analysis. Univariate analysis was used to screen the risk factors, and binary logistic analysis was performed to build the prediction model, which was furtherly transferred into a nomogram. The prediction performance was evaluated by receiver operating characteristic (ROC) curve, Hosmer-Lemeshow test, and nomogram calibration plot. Internal validation was processed by using ordinary bootstrapping. RESULTS: A total of 273 patients who match the research criteria were enrolled, in which 37 cases (13.6%) were confirmed to have LDRM. Univariate analysis showed the risk factors included diabetes (p = 0.003), admission on surgical intensive care unit (p = 0.012), duration time (p < 0.001), site leakage (p < 0.001), and craniotomy (p < 0.001). In multivariate analysis, four of the variables were identified as independent risk factors to establish a prediction model, and a graphical nomogram was designed. The area under the ROC curve was 0.837, and the p value in the Hosmer-Lemeshow test was 0.610, with a mean absolute error in the calibration plot calculated as 0.022. The indices in the testing set were in good accordance with the original set when internal validation was performed. CONCLUSIONS: This is the first study to transform the prediction model of LDRM into a nomogram, which can be considered as a tool for clinicians to assess infection risk.