脑卒中手术患者医院感染风险预测列线图模型的构建

目的构建脑卒中手术患者医院感染风险预测列线图模型,为早期筛查医院感染高风险人群和制定预防策略提供一定的参考和依据。方法回顾性收集2016-2018年山东大学齐鲁医院脑卒中手术患者的临床相关资料,将患者按照7∶3的比例随机分为建模组(571例)和验证组(245例)。采用单因素和多因素Logistic回归探讨医院感染的独立危险因素,基于危险因素的回归系数构建脑卒中手术患者医院感染风险预测列线图模型。分别在建模组(内部验证)和验证组(外部验证)中采用受试者工作特征(ROC)曲线下面积(AUC)和校准曲线评估预测模型的区分度和校准度。结果共纳入816例脑卒中手术患者,医院感染213例,医院感染发生率为26.10%。Logistic回归分析显示,脑卒中类型、留置胃管、静脉血栓、手术风险分级(NNIS)、美国国立卫生研究院卒中量表(NIHSS)评分以及住院时间是脑卒中手术患者医院感染的独立危险因素(P<0.05)。依此构建的列线图模型在建模组和验证组中的ROC曲线下面积分别为0.849和0.858,具有良好的区分度;两组校准曲线显示列线图模型的预测值和实际观察值结果一致性良好(P=0.731、P=0.224)。结论本研究构建的个体化风险预测列线图模型有助于提高对脑卒中术后医院感染高危人群的筛查和早期诊断,尽早制定干预策略,以降低感染发生率。

Development of nomogram model for prediction of risk of nosocomial infection in stroke patients undergoing surgery

OBJECTIVE To construct the nomogram model for prediction of risk of nosocomial infection in stroke patients undergoing surgery so as to provide guidance for early screening of population at high risk of nosocomial infection and development of prevention strategies. METHODS The relevant clinical data were retrospectively collected from the stroke patients who underwent surgeries in Qilu Hospital of Shandong University from 2016 to 2018, the enrolled patients were randomly divided into the derivation group with 571 cases and the validation group with 245 cases in a 7:3 ratio. Univariate analysis and multivariate logistic regression analysis were performed for independent risk factors for the nosocomial infection, the nomogram model for prediction of nosocomial infection was constructed based on the regression coefficients of the risk factors. The discrimination and calibration of the prediction model were estimated by respectively using the area under curve(AUC) of receiver-operating-characteristic(ROC) and calibration belts in the derivation group(internal validation) and the validation group(external validation). RESULTS Of totally 816 stroke patients who were included in the study, 213 had nosocomial infection, with the incidence rate of nosocomial infection 26.10%. Logistic regression analysis showed the type of stroke, gastric tube indwelling, phlebothrombosis, NNIS, National Institute of Health Stroke Scale(NIHSS) score and length of hospital stay were the independent risk factors for the nosocomial infection in the stroke patients undergoing surgery(P<0.05). The area under ROC curve of the nomogram was 0.849 in the derivation group, 0.858 in the validation group, showing good discrimination. The calibration curve of two groups showed that the predicted probability of the nomogram model was in high consistency with the actual probability(P=0.731, P=0.224). CONCLUSION The individualized nomogram model for prediction of risk may facilitate the screening of the stroke patients at high risk of postoperative nosocomial infection and the early diagnosis, and it is necessary to formulate interventions as early as possible so as to reduce the incidence of infection.