糖尿病合并脑卒中患者医院感染风险预测模型构建及经济负担评价

Construction of risk prediction model for nosocomial infection and economic burden of diabetes mellitus patients complicated with stroke

目的探讨糖尿病(DM)合并脑卒中患者医院感染风险预测模型的构建及经济负担评价。方法回顾性分析2017年2月-2020年6月淮北市中医医院收治的DM合并脑卒中患者93例,根据患者有无医院感染分为感染组(n=31)和未感染组(n=62)。收集感染各部位分泌物进行病原菌分离鉴定,记录患者年龄、糖尿病病程、血糖、白蛋白(ALB)、是否存在意识障碍、吞咽障碍和侵入性操作等。建立多因素Logistic回归风险预测模型,评估模型拟合度和预测效能。采取倾向得分匹配法进行1∶1病例对照研究,计算医院感染的直接经济负担。结果93例DM合并脑卒中患者医院感染率为33.33%,感染部位主要为呼吸道(48.39%),分离病原菌26株,主要为肺炎克雷伯菌(34.62%);Logistic回归分析显示,有意识障碍(OR=4.256,P=0.021)、糖尿病病程≥5年(OR=4.037,P=0.038)、空腹血糖(FPG)≥7 mmol/L(OR=4.419,P=0.020)、有侵入性操作(OR=5.108,P=0.008)、住院时间≥10 d(OR=4.057,P=0.044)均为影响医院感染的独立危险因素,相应的风险预测模型为:P=1/[1+e^(-4.808+1.448×(意识障碍)+1.396×(糖尿病病程)+1.486×(空腹血糖)+1.631×(侵入性操作)+1.192×(住院时间))],Hosmer-Lemeshowχ^(2)=10.396,P=0.238,模型的拟合度较好,模型预测医院感染的曲线下面积(AUC)为0.834。DM合并脑卒中医院感染患者的各项住院费用均高于未感染患者(P<0.05),医院感染的直接经济负担为15615元。结论医院感染会增加DM合并脑卒中患者的直接经济负担,临床可采用多因素Logistic回归模型预测医院感染发生。

OBJECTIVE To construct the risk prediction model for nosocomial infection in diabetes mellitus(DM)patients complicated with stroke and analyze the economic burden.METHODS A total of 93 DM patients complicated with stroke who were treated in Huaibei Hospital of Traditional Chinese Medicine from Feb 2017 to Jun 2020 were enrolled in the study and divided into the infection group with 31 cases and the non-infection group with 62 cases according to the status of nosocomial infection.The secretion specimens were collected from the infection sites.The age,course of diabetes mellitus,blood glucose,albumin,disturbance of consciousness,dysphagia and invasive procedures were recorded.Multivariate logistic regression risk prediction model was established,the degree of fitting and predictive efficiency of the model were evaluated,a 1∶1 case-controlled study was conducted by means of propensity score matching method,and the direct economic burden due to nosocomial infection was calculated.RESULTS Among the 93 DM patients complicated with stroke,the incidence of nosocomial infection was 33.33%,and the patients with respiratory tract infection accounted for 48.39%.Totally 26 strains of pathogens were isolated,34.62%of which were Klebsiella pneumoniae.Logistic regression analysis showed that the disturbance of consciousness(OR=4.256,P=0.021),course of diabetes mellitus no less than 5 years(OR=4.037,P=0.038),fasting blood glucose(FPG)no less than 7 mmol/L(OR=4.419,P=0.020),invasive procedures(OR=5.108,P=0.008)and length of hospital stay no less than 10 days(OR=4.057,P=0.044)were independent risk factors for the nosocomial infection.The probability of risk prediction model was as follows:P=1/[1+e(-4.808+1.448×(disturbance of consciousness)+1.396×(DM course)+1.486×(FPG)+1.631×(invasive operation))+1.192×(hospitalization time)],Hosmer-Lemeshowχ^(2)=10.396,P=0.238,the degree of fitting of the model was good,and the area under curve(AUC)of the model was 0.834 in prediction of nosocomial infection.The hospitalization costs of the stroke complicated with DM patients with nosocomial infection were significantly more than those of the stroke complicated with DM patients without nosocomial infection(P<0.05),and the direct economic burden due to nosocomial infection was 15615 yuan.CONCLUSION The nosocomial infection may increase the direct economic burden of the DM patients complicated with stroke.The multivariate logistic regression model can be used for prediction of nosocomial infection.