【作者】 李倩；

【导师】 王书会；

【作者基本信息】 山东大学 ， 护理（专业学位）， 2022， 硕士

【摘要】 研究目的调查口腔癌手术患者医院感染现状,探讨口腔癌手术患者医院感染的危险因素,构建Logistic回归和决策树风险预测模型,以便于早期筛查口腔癌术后医院感染的高风险人群。通过评估医院感染所致的经济负担,包括患者的直接和间接经济负担,为合理利用医疗资源提供相关参考依据。研究方法本研究收集山东省某三级甲等综合医院2018年1月1日～2021年12月31日期间826例口腔癌手术患者作为研究对象,采用自行编制的《口腔癌手术患者医院感染目标监测调查表》、《口腔癌手术患者医院感染经济负担调查表》收集患者的相关临床资料,按照8:2的比例随机分为建模组(660例)和检验组(166例)。在建模组通过单因素分析筛选口腔癌手术患者医院感染的危险因素,将P<0.05具有统计学意义的因素纳入多因素Logistic回归,建立Logistic回归预测模型,通过Hosmer-lemeshow检验拟合效果,利用受试者工作特征曲线(Receiver Operating Characteristic,ROC)曲线下面积(Area under curre,AUC)检验预测效果;利用卡方自动交互检测(Chi-squared automatic interaction detector,CHAID)算法将P<0.05的相关因素构建决策树预测模型(设置父节点最小样本量为20,子节点为10,最大树深度3),应用Risk统计量及AUC进行模型评价。利用检验组数据对模型进行内部验证,采用Z检验比较Logistic模型与决策树模型的AUC。采用倾向评分匹配法,按照1:1倾向条件进行配对,探讨口腔癌手术患者医院感染导致的经济负担。研究结果1.共纳入口腔癌手术患者826例,其中男性494例(59.81%),女性332例(40.19%);年龄平均为58.90±12.970岁,建模组660例口腔癌手术患者中,男性400例(60.61%),女性260例(39.39%),年龄平均为58.77±12.984岁;检验组166例口腔癌手术患者中,男性94例(56.63%),女性72例(43.37%),年龄平均为59.43±12.938岁。2.826例口腔癌手术患者中,医院感染129例,共计148例次,医院感染发生率为15.62%,医院感染例次发生率为17.92%。建模组的医院感染发生率为16.36%,医院感染例次发生率为19.24%;检验组的医院感染发生率为12.65%,医院感染例次发生率为13.25%。3.129例口腔癌术后感染中,以手术切口感染最为常见,74例(57.36%),其中深部手术切口感染63例(48.84%)、浅表手术切口感染11例(8.53%),其次是肺部感染32例(24.81%),多部位感染19例(14.73%);在建模组,手术切口感染63例(58.33%),其中深部手术切口感染55例(50.92%),浅表手术切口感染8例(7.41%),肺部感染24例(22.22%),多部位感染18例(16.67%);在检验组,手术切口感染11例(52.38%),其中浅表手术切口感染3例(14.29%),深部手术切口感染8例(38.09%);肺部感染8例(38.10%),多部位感染1例(4.76%)。4.共检出病原菌101株,革兰氏阴性菌65株(64.36%),其中鲍曼不动杆菌(15.84%)检出最多,其次是铜绿假单胞菌(11.88%)、肺炎克雷伯菌(9.91%);革兰氏阳性菌33株(32.67%),其中表皮葡萄球菌(5.94%)检出最多,其次是金黄色葡萄球菌(4.95%)、咽峡炎链球菌(4.95%)和草绿色链球菌(3.96%);真菌2株(1.98%)。5.Logistic回归分析显示口腔癌术后医院感染的5个独立危险因素包括:糖尿病、静脉血栓、气管切开、总住院时间、抗菌药物联合使用;决策树CHAID算法筛选出的6个解释变量:抗菌药物联合使用、静脉血栓、一级护理天数、糖尿病、引流管数量及总住院时间;根据危险因素构建的Logistic回归和决策树风险预测模型,在建模组的AUC分别为0.879和0.870,在检验组的AUC分别为0.817和0.826,说明两模型具有良好的区分度;两预测模型的AUC进行比较(P=0.3795,Z=0.879),差异无统计学意义。6.采用1:1倾向评分匹配后,感染组患者的住院总费用为84055.31元/例,非感染组患者住院总费用为72354.87元/例,医院感染所致患者直接经济负担为11700.44元/例,其中,西药费所致的负担最大(4267.68元),其次是卫生材料费(3301.12元)和治疗费(945.75元);医院感染使口腔癌患者住院时间延长4.5天/例,间接经济负担为2163.69元。结论与意义口腔癌术后医院感染率较高,应加强对口腔癌手术患者的目标性监测。糖尿病、静脉血栓、气管切开、总住院时间、抗生素联合使用、一级护理天数、引流管数量(>2)是口腔癌术后感染的危险因素,根据以上危险因素实施有效的、具有针对性的医院感染防控措施。同时,医院感染给口腔癌患者带来了巨大的经济负担,并延长了住院时间;本研究构建了 Logistic回归和决策树风险预测模型,两模型具有较好的区分度和准确度,有助于早期筛查和判断口腔癌术后医院感染高风险人群,及早制定合理有效的干预策略,降低医院感染发生概率。更多还原

【Abstract】 ObjectivesThis study aims to investigate and explore the current status and risk factors of hospital infection in patients with oral cancer surgery,and to construct Logistic regression and decision tree risk prediction model,so as to facilitate early screening of high-risk groups of hospital infection after oral cancer surgery.By evaluating the economic burden caused by hospital infection,including the direct and indirect economic burden of patients,we aim to provide relevant reference for the rational use of medical resources.MethodTotally 826 patients with oral cancer who underwent surgery in a certain grade A tertiary hospital in Shandong Province from January 1,2018 to December 31,2021 were collected in this study by cluster sampling method.The clinical data of patients were collected according to self-made ’Hospital Infection Target Monitoring Questionnaire for Patients with Oral Cancer Surgery’ and ’ Hospital Infection Economic Burden Questionnaire for Patients with Oral Cancer Surgery ’.Patients were randomly divided into model group(660 cases)and test group(166 cases)according to the ratio of 8:2.In the model group,the risk factors of hospital infection were screened by single factor analysis,and the factors with statistical significance P<0.05 were included in the multivariate logistic regression to establish the logistic regression prediction model.The fitting effect was tested by Hosmer-lemeshow,and the prediction effect was tested by area under curre(AUC)under the receiver operating characteristic(ROC)curve.Chi-squared automatic interaction detector(CHAID)was used to construct a decision tree prediction model(the minimum sample size of the parent node was 20,the child node was 10,and the maximum tree depth was 3).Risk statistics and AUC were used to evaluate the model.The external validation of the model was performed using the test group data,and AUC of Logistic model and decision tree model was compared by Z test.The propensity score weighting method was used to match according to the 1:1 propensity condition to explore the economic burden caused by nosocomial infection in patients with oral cancer surgery.Results1.A total of 826 patients with oral cancer surgery were included in this study,including 494 males(59.81%)and 332 females(40.19%).The average age was 58.90 ± 12.970 years old.All patients were divided randonmly into two groups:model group(n=660)and test group(n=166).Among model group,400 cases were males(60.61%)and 260 cases were females(39.39%),with an average age of 58.77 ± 12.984 years old.Among test group,94 cases were males(56.63%)and 72 cases were females(43.37%),with an average age of 59.43±12.938 years old.2.Among the 826 patients included in the study,129 cases of hospital infection occurred,with a total of 148 case-times,and the incidence of hospital infection was 15.62%and the incidence of hospital infection case-times was 17.92%.The incidence of hospital infection was 16.36%and the incidence of hospital infection case-times was 19.24%in model group.On the other hand the incidence of hospital infection was 12.65%and the incidence of hospital infection case-times was 13.25%in test group.3.Among 129 cases of postoperative infection of oral cancer,surgical incision infection was the most common in 74 cases(57.36%),including 63 cases(48.84%)of deep surgical incision infection,11 cases(8.53%)of superficial surgical incision infection,followed by 32 cases(24.81%)of pulmonary infection and 19 cases(14.73%)of multisite infection.In model group,there were 63 cases(58.33%)of surgical incision infection,including 55 cases(50.92%)of deep surgical incision infection and 8 cases of superficial surgical incision infection(7.41%),24 cases(22.22%)of pulmonary infection,and 18 cases(16.67%)of multi-site infection.In test group,there were 11 cases(52.38%)of surgical incision infection,including 3 cases(14.29%)of superficial surgical incision infection and 8 cases(38.09%)of deep surgical incision infection 3,8 cases(38.10%)of pulmonary infection and 1 case(4.76%)of multi-site infection.4.Among the 129 cases with hospital infection,101 strains of pathogens were detected,65 strains(64.36%)of Gram-negative bacteria,of which Acinetobacter baumannii(15.84%)was the most detected,followed by Pseudomonas aeruginosa(11.88%)and Klebsiella pneumoniae(9.91%).33 strains(32.67%)of Gram-positive bacteria were detected,of which Staphylococcus epidermidis(5.94%)detected the most,followed by Staphylococcus aureus(4.95%),Streptococcus pharyngitis(4.95%)and Streptococcus viridans(3.96%).Besides,2 strains(1.98%)of fungi were also detected.5.Logistic regression analysis showed that the five independent risk factors of hospital infection after oral cancer surgery included diabetes mellitus,venous thrombosis,tracheotomy,total hospitalization time and combined use of antibiotics and six explanatory variables screened by decision tree CHAID algorithm included combined use of antibiotics,venous thrombosis,first-class nursing days,diabetes,number of drainage tubes and total hospitalization time.Logistic regression and decision tree risk prediction models were constructed according to risk factors.The AUC of model group was 0.879 and 0.870 respectively,and AUC of test group was 0.817 and 0.826 respectively,indicating good discrimination between the two models.The AUC was compared between the two prediction models(P=0.3795,Z=0.879),and the difference was not statistically significant.6.After 1:1 propensity score matching,the total cost of hospitalization in the infection group was 84055.31 yuan per case,the total cost in the non-infection group was 72354.87 yuan per case,and the direct economic burden of patients caused by hospital infection was 11700.44 yuan per case.Among them,the burden caused by western medicine was the largest(4267.68 yuan),followed by health materials(3301.12 yuan)and treatment costs(945.75 yuan).Hospital infection prolonged the hospitalization time of oral cancer patients by 4.5 days per case,and the indirect economic burden was 2163.69 yuan.ConclusionsOral cancer patients have a high proportion of hospital-acquired infections after surgery,so the targeted monitoring of patients with oral cancer should be enhanced.Diabetes mellitus,venous thrombosis,tracheotomy,total hospital stay,antibiotic combination,number of primary care days,and number of drains(>2)are risk factors for hospital infection after oral cancer surgery,and effective and targeted hospital infection prevention and control measures should be implemented for the above risk factors.At the same time,hospital infection brings huge economic burden to patients,and prolongs the hospitalization time of patients with oral cancer surgery.In this study,Logistic regression and decision tree risk prediction models were constructed and the two models have good discrimination and accuracy,which are helpful for early screening and judging the high risk population of hospital infection after oral cancer surgery,and formulate reasonable and effective intervention strategies as soon as possible to reduce the incidence of nosocomial infection.更多还原