论文部分内容阅读
目的对急性生理和慢性健康状况Ⅱ(Acute Physiology and Chronic Health EvaluationⅡ,APACHEⅡ)评分预测老年危重症患者住院病死率的效能进行验证。方法 421例老年危重症患者分为生存组281例和死亡组(住院死亡)140例,收集入住ICU 24h内相关指标,计算APACHEⅡ评分及其预测的住院病死率;应用ROC曲线评估APACHEⅡ评分预测住院病死率的辨别力,采用Hosmer-Lemeshow(HL)goodness-of-fit检验评估APACHEⅡ评分的校准力,采用标准化死亡比值(standardized mortality ratio,SMR)检验APACHEⅡ评分预测住院病死率的准确性,采用Brier评分(Brier score,BS)评估APACHEⅡ评分预测住院病死率的综合能力。结果生存组APACHEⅡ评分[17.0(12.0,24.0)分]低于死亡组[26.5(22.0,31.0)分],APACHEⅡ评分预测住院病死率[7.7(0.5,47.4)%]低于死亡组[71.1(26.1,90.4)%],ICU住院时间[7(4,11)d]较死亡组[11(5,19)d]短,住院时间[27(14,51)d]较死亡组[15(6,34)d]长(P<0.05);生存组患者合并慢性心功能不全(NYHA心功能Ⅲ~Ⅳ级)、痴呆、肿瘤比例(14.9%、14.2%、10.7%)低于死亡组(28.6%、23.6%、19.3%)(P<0.05);APACHEⅡ评分24分确定为最佳截断值,预测住院病死率的AUC为0.77(95%CI:0.72~0.81,P<0.05),HL检验显示校准不良(χ~2=16.690,P=0.034),预测老年重症患者群体住院病死率的准确性良好(SMR=0.90),模型综合能力一般(BS=0.20)。结论 APACHEⅡ评分在预测老年危重症患者住院病死率上有良好的辨别力和总体预测准确性,但校准不良,应专科化模型改善校准。
Objective To verify the efficacy of Acute Physiology and Chronic Health Evaluation Ⅱ (APACHE Ⅱ) in predicting in-hospital mortality in elderly critically ill patients. Methods A total of 421 elderly critically ill patients were divided into survival group (n = 281) and death group (death in hospital), and the related indicators were collected within 24 hours of ICU admission. APACHEⅡscore and in-hospital mortality were calculated. APACHEⅡscore was used to predict hospitalization To evaluate the accuracy of the APACHEⅡscore in predicting the in-hospital mortality by the standardized mortality ratio (SMR) using Hosmer-Lemeshow (HL) goodness-of-fit test, Score (Brier score, BS) Assess the combined ability of the APACHE II score to predict in-patient mortality. Results The APACHEⅡscore of the survival group was lower than that of the death group [26.0 (22.0, 31.0)], and the APACHE Ⅱ score was lower than that of the death group (71.1 (12.4) ICU length of stay [7 (4,11) d] was shorter than that of death group [11 (5,19) d] and length of stay [27 (14,51) 6,34) d] were longer than those in the death group (P <0.05); patients in the survival group had chronic heart failure (NYHA class Ⅲ ~ Ⅳ), dementia and tumor proportions (14.9%, 14.2% and 10.7% 28.6%, 23.6%, 19.3% respectively) (P <0.05). The APACHEⅡ score of 24 was the best cutoff value. The AUC of predicting in-hospital mortality was 0.77 (95% CI: 0.72-0.81, Poor calibration (χ ~ 2 = 16.690, P = 0.034) showed good accuracy of in-hospital mortality (SMR = 0.90) and general ability of model (BS = 0.20). Conclusion The APACHE Ⅱ score has good discrimination and overall predictive accuracy in predicting the in-hospital mortality of critically ill elderly patients, but the calibration is poor. Specialized models should be used to improve the calibration.