Department of Neurosurgery, the First Affiliated Hospital of Jinan University, Guangzhou, 510630, China.
Department of Neurosurgery, Shanxi Bethune Hospital, Shanxi Academy of Medical Sciences, Tongji Shanxi Hospital, Third Hospital of Shanxi Medical University, Taiyuan, 030032, China.
Neurosurg Rev. 2023 Nov 21;46(1):308. doi: 10.1007/s10143-023-02208-9.
The incidence of pneumonia in ICU patients with TBI is very high, seriously affecting the prognosis. This study aims to construct a predictive model for pneumonia in ICU patients with TBI and provide help for the prevention of TBI-related pneumonia.Clinical data of ICU patients with TBI were collected from the Medical Information Mart for Intensive Care (MIMIC)-IV database and hospital data. Variables were screened by lasso and multivariate logistic regression to construct a predictive nomogram model, verified in internal validation cohort and external validation cohort by receiver operator characteristic (ROC) curve, calibration curve and decision curve analysis (DCA).A total of 1850 ICU patients with TBI were enrolled in the study from the MIMIC-IV database, including 1298 in the training cohort and 552 in internal validation cohort. The external validation cohort included 240 ICU patients with TBI from hospital data. Nine variables were selected from the training cohort by lasso regression and multivariate logistic regression, and a pneumonia prediction nomogram was constructed. This nomogram has a high discrimination in training, internal validation and external validation cohorts (AUC = 0.857, 0.877, 0.836). The calibration curve and DCA showed that this nomogram had a high calibration and better clinical decision-making efficiency.The nomogram showed excellent discrimination and clinical utility to predict pneumonia, and could identify pneumonia high-risk patients early, thus providing personalised treatment strategies for ICU patients with TBI.
颅脑损伤（TBI）患者 ICU 中肺炎的发病率非常高，严重影响预后。本研究旨在构建 TBI 患者 ICU 肺炎预测模型，为 TBI 相关肺炎的预防提供帮助。
从医疗信息存储库-重症监护（MIMIC-IV）数据库和医院数据中收集 TBI 患者 ICU 的临床数据。通过lasso 和多变量逻辑回归筛选变量，构建预测列线图模型，通过接收者操作特征（ROC）曲线、校准曲线和决策曲线分析（DCA）在内部验证队列和外部验证队列中进行验证。
共纳入 1850 例来自 MIMIC-IV 数据库的 TBI 患者 ICU，其中 1298 例为训练队列，552 例为内部验证队列。外部验证队列包括 240 例来自医院数据的 TBI 患者 ICU。通过 lasso 回归和多变量逻辑回归从训练队列中筛选出 9 个变量，构建了肺炎预测列线图。该列线图在训练、内部验证和外部验证队列中具有较高的区分度（AUC=0.857、0.877、0.836）。校准曲线和 DCA 表明，该列线图具有较高的校准度和更好的临床决策效率。
该列线图对预测肺炎具有优异的判别力和临床实用性，可早期识别肺炎高危患者，从而为 TBI 患者 ICU 提供个性化治疗策略。
