谵妄预测模型与早期谵妄预测模型在ICU患者谵妄预测中的应用价值

摘要

目的 研究谵妄预测模型(PRE-DELIRIC)和早期谵妄预测模型(E-PRE-DELIRIC)在预测ICU患者发生谵妄的能力,并比较2种预测模型的应用价值.方法 选择2017年1-4月入住中南大学湘雅医院重症医学科的265例ICU患者,应用PRE-DELIRIC和E-PRE-DELIRIC模型计算每例患者2种模型对应的分值及预算发生率,对2种预测模型的灵敏度、特异度、约登指数、阳性预测值、阴性预测值进行相关分析,通过评价2名ICU医护对另外20例患者评估结果的组内相关系数,完成评定者间信度评价.结果 未发生谵妄196例,实际发生谵妄69例,谵妄患者和非谵妄患者的PRE-DELIRIC分值分别为(46.41±14.05)、(20.08±9.96)分,两者比较差异有统计学意义(t=14.34,P<0.05);谵妄患者和非谵妄患者的E-PRE-DELIRIC分值分别为(36.23±13.34)、(14.45±9.56)分,两者比较差异有统计学意义(t=14.59,P<0.05);PRE-DELIRIC和E-PRE-DELIRIC对ICU总体患者谵妄发生预测的ROC曲线下面积均大于0.9,PRE-DELIRIC的95％置信区间为0.891～0.965,E-PRE-DELIRIC的95％置信区间为0.861～0.937,2种预测模型的辨别力均较好,PRE-DELIRIC和E-PRE-DELIRIC的灵敏度分别为0.841、0.812,特异度分别为0.939、0.913;PRE-DELIRIC分值取36.5分和E-PRE-DELIRIC分值取30.5分是两者预测谵妄发生率的最佳分值;PRE-DELIRIC的约登指数为0.779,优于E-PRE-DELIRIC的0.725.结论 PRE-DELIRIC和E-PRE-DELIRIC评分模型在ICU患者应用的辨别力及校验力均较好,适合用于中国文化背景下的ICU患者谵妄发生的预测.PRE-DELIRIC对总体患者的预测性能较优越,但应用时间上有一定的局限性,E-PRE-DELIRIC模型则在患者入ICU时即可使用.%Objective To assess and compare the predicted abilities of PRE- DELIRIC and E-PRE-DELIRIC two delirium prediction models in ICU patients. Methods From January 2017 to April 2017, 265 critically ill patients who met the inclusion criteria were included in this study. The clinical data of all the included patients, respectively calculated the scores with both PRE-DELIRIC and E-PRE-DELIRIC rating software were collected. Forecasting performance of two different models of discrimination and calibration were assessed. 2 medical staff assessed 20 patients with the Chinese version PRE-DELIRIC and E-PRE-DELIRIC,while the intraclass correlation coefficients were accumulated to evaluate the inter-rater reliability. Results The average scores of PRE-DELIRIC were46.41 ± 14.05in delirium group and20.08 ± 9.96 in patients without delirium, the difference between scores was statistically significant (t=14.34, P<0.05). The average scores of E-PRE-DELIRIC were36.23±13.34in delirium group and14.45±9.56 in patients without delirium, the difference between scores was statistically significant (t=14.59, P<0.05). The AUROCC and its 95％CI of the PRE-DELIRIC and E-PRE-DELIRIC for prediction of delirium were 0.928(0.891-0.965) and 0.904(0.861-0.947) respectively in all patients. Discrimination was generally good for two models. The PRE-DELIRIC and E-PRE-DELIRIC sensitivity were 0.841, 0.812, specificity was 0.939, 0.913, the threshold values of PRE-DELIRIC model was 36.5％and E-PRE-DELIRIC model was 30.5％. The Youden′s index of PRE–DELIRIC model was 0.779, which better than 0.725 of E-PRE-PREDIRIC model. Conclusions The PRE-DELIRIC and E-PRE-DELIRIC models both have high accuracy in predicting delirium of patients in intensive care unit. PRE-DELIRIC does better performance than E-PRE-DELIRIC, but has imitation in terms of time. E-PRE-DELIRIC model can be used in ICU patients who develop delirium within 24 h following admission.