高龄脓毒症患者28d死亡预测模型的构建及验证

Development and validation of predictive model for 28-day mortality in very older patients with sepsis

目的构建预测高龄脓毒症患者入院28 d内死亡的列线图并进行验证,以期早期识别高危人群,改善预后。方法连续收集2022年1月1日至2022年11月30日首都医科大学附属北京朝阳医院急诊科收治的、年龄≥80岁的脓毒症患者,记录患者的临床资料及28 d预后,随机分为建模组(70%)与验证组(30%),对建模组患者采用最小绝对值收敛和选择算子算法(LASSO)及多因素Cox回归分析确定患者死亡的独立危险因素,并以此建立列线图。并在验证组中进行验证。结果507例高龄脓毒症患者的病死率为31.2%,在建模组中将LASSO回归分析筛选出的变量纳入多因素Cox回归分析中显示,增龄[风险比(HR)=1.059,95%置信区间(95%CI)=1.017~1.103,P=0.005]、认知功能障碍(HR=2.100,95%CI=1.322~3.336,P=0.002)、衰弱(HR=2.561,95%CI=1.183~5.545,P=0.017)、低平均动脉压(HR=0.987,95%CI=0.976~0.998,P=0.017)、低前白蛋白(HR=0.997,95%CI=0.994~1.000,P=0.040)、高尿素氮(HR=1.028,95%CI=1.010~1.045,P=0.001)及高降钙素原(HR=1.008,95%CI=1.001~1.016,P=0.019)为高龄脓毒症患者死亡的独立危险因素。基于上述独立危险因素建立列线图,校准曲线、时间依赖性曲线下面积、临床决策曲线分别显示该模型在建模组与验证组中具有较好的校准度、区分度及临床实用性。结论增龄、认知功能障碍、衰弱、低平均动脉压、低前白蛋白、高尿素氮、高降钙素原是高龄脓毒症患者28 d死亡的独立危险因素,基于上述危险因素构建的列线图预测效能较好,可为预后评估提供帮助。

Objective To develop and validate a predictive nomogram for 28-day mortality among very older patients with sepsis,to identify high-risk patients early and improve prognosis.Methods This study was conducted from January 1,2022,to November 30,2022.Very older patients aged≥80 years with sepsis admitted to the emergency department of Beijing Chao-Yang Hospital,Capital Medical University were consecutively recruited.Their clinical data within 24 h of admission and 28-day mortality was recorded.The participants were divided into training(70%)and validation cohort(30%)(random number).In the training cohort,the risk factors of 28-day mortality were selected via least absolute shrinkage and selection operator(LASSO)regression analysis and multivariable Cox proportional hazard model,and a nomogram was developed.The prediction model was verifi ed in validation cohort.Results In total,507 very older patients with sepsis were included,among which the mortality rate was 31.2%.In training cohort,the independent risk factors for 28-day mortality were identifi ed:increased age[hazard ratio(HR)=1.059,95%confi dence interval(95%CI)=1.017-1.103,P=0.005],cognitive impairment(HR=2.100,95%CI=1.322-3.336,P=0.002),frailty(HR=2.561,95%CI=1.183-5.545,P=0.017),decreased mean arterial pressure(HR=0.987,95%CI=0.976-0.998,P=0.017),decreased prealbumin(HR=0.997,95%CI=0.994-1.000,P=0.040),increased blood urea nitrogen(HR=1.028,95%CI=1.010-1.045,P=0.001),increased procalcitonin(HR=1.008,95%CI=1.001-1.016,P=0.019)via LASSO regression analysis and multivariable Cox regression analysis.The nomogram was developed using these seven predictors.In the training and validation cohorts,the calibration curves,time-dependent AUC curves,and decision curve analysis showed that the nomogram had good calibration degree,discrimination and clinical net benefi ts.Conclusions Increased age,cognitive impairment,frailty,decreased mean arterial pressure,decreased prealbumin,increased blood urea nitrogen,and increased procalcitonin are independent risk factors for 28-day mortality in very older patients with sepsis.The nomogram,which included the seven predictors,have good predictive performance,and might be helpful for prognosis assessment.