老年重症肺炎患者死亡危险因素分析及预测模型构建

Risk factors of death and construction of prediction model in elderly patients with severe pneumonia

目的探讨老年重症肺炎患者死亡的危险因素,构建其死亡风险预测模型,为早期防治提供依据。方法对300例老年重症肺炎患者的临床资料进行回顾性分析,并将其按照2:1的比例随机分为建模组200例(生存149例,死亡51例)和验证组100例(生存78例,死亡22例)。老年重症肺炎患者死亡的影响因素进行单因素分析及多因素分析,并建立风险模型,分别应用建模组(内部验证)和验证组(外部验证)的数据绘制受试者工作特征曲线评估模型的预测效能。结果重症肺炎患者死亡率在血清降钙素原表达、器官受累数、急性生理学及慢性健康状况评分系统Ⅱ评分及是否存在感染性休克、呼吸衰竭方面存在显著性差异(P<0.05)。Logistic回归分析显示,器官受累数≥3个、急性生理学及慢性健康状况评分系统Ⅱ评分≥29分、存在呼吸衰竭、存在感染性休克、血清降钙素原表达≥4 ng·ml^(-1)是老年重症肺炎患者死亡的独立危险因素。重症肺炎患者死亡风险预测模型表达式为:P=1/[1+e^(-(-4.340+0.593×血清降钙素原+1.281×器官受累数+0.873×急性生理学及慢性健康状况评分系统Ⅱ评分+0.885×呼吸衰竭+0.872×感染性休克))],Hosmer-Lemeshow检验模型的拟合优度(P>0.05);利用建模组数据对模型进行内部验证,曲线下面积为0.884(95%CI为0.826~0.943),灵敏度为85.91%,特异度为84.31%;利用验证组数据对模型进行外部验证,预测重症肺炎患者死亡的曲线下面积为0.888(95%CI为0.817~0.959),灵敏度为81.94%,特异度为82.14%。结论老年重症肺炎患者死亡影响因素较多,风险预测模型能较好地筛选危险因素,临床可据此甄别高危人群,以期早期预防与治疗,降低死亡率。

Objective To explore the risk factors of death in elderly patients with severe pneumonia and construct the prediction model of death risk so as to provide reference for early prophylaxis and treatment.Methods Retrospective analyses were performed on clinical data of 300 elderly patients with severe pneumonia.They were randomly divided into modeling group(n=200,survival 149,death 51)and validation group(n=100,survival 78,death 22)according to the ratio of 2 to 1.Single and multiple factor analyses were performed for death influencing factors of elderly patients with severe pneumonia,constructed risk model,and forecasting efficacies assessed with receiver operator characteristic(ROC)curve charted with the data of modeling(internal validation)and validation(external validation)group.Results Death rates of patients with severe pneumonia had significant differences in expression of serum procalcitonin,number of involvement organs,score on acute physiology and chronic health evaluation scoring system(APACHE II)as well as whether existing infectious shock and respiratory failure(P<0.05).Logisitic regression analysis showed that number of involvement organs≥3,APACHE II score≥29 points,respiratory failure,septic shock and serum PCT≥4ng/mL were independent risk factors of death in patients with severe pneumonia.The expression of mortality risk prediction model of patients with severe pneumonia was as follow:P=1/[1+e^(-(-4.340+0.593×serum PCT+1.281×number of involvement organs+0.873×APACHE II score+0.885×respiratory failure+0.872×septic shock))].The fit goodness of Hosmer-Lemeshow test model was(P>0.05).Internal verification was performed using the data of modeling group,the area under the ROC curve(AUC)was 0.884(95%CI 0.826~0.943),sensitivity 85.91%and specificity 84.3%;external verification was performed using the data of verification group,the area under the ROC curve(AUC)was 0.888(95%CI 0.817~0.959),sensitivity 81.94%and specificity 82.14%.Conclusions There are more death influencing factors of elderly patients with severe pneumonia,risk prediction model could preferably screen risk factors and on this basis high-risk death groups are identified in order to conduct early prevention and treatment and lower death rate.