Srilata Moningi, Padhy Narmada, Padmaja Durga, Gopinath Ramachandran
Department of Anaesthesiology and Critical Care, Nizam's Institute of Medical Sciences, Hyderabad, Andhra Pradesh, India.
Ann Card Anaesth. 2015 Apr-Jun;18(2):161-9. doi: 10.4103/0971-9784.154468.
To validate the Parsonnet scoring model to predict mortality following adult cardiac surgery in Indian scenario.
A total of 889 consecutive patients undergoing adult cardiac surgery between January 2010 and April 2011 were included in the study. The Parsonnet score was determined for each patient and its predictive ability for in-hospital mortality was evaluated. The validation of Parsonnet score was performed for the total data and separately for the sub-groups coronary artery bypass grafting (CABG), valve surgery and combined procedures (CABG with valve surgery). The model calibration was performed using Hosmer-Lemeshow goodness of fit test and receiver operating characteristics (ROC) analysis for discrimination. Independent predictors of mortality were assessed from the variables used in the Parsonnet score by multivariate regression analysis.
The overall mortality was 6.3% (56 patients), 7.1% (34 patients) for CABG, 4.3% (16 patients) for valve surgery and 16.2% (6 patients) for combined procedures. The Hosmer-Lemeshow statistic was <0.05 for the total data and also within the sub-groups suggesting that the predicted outcome using Parsonnet score did not match the observed outcome. The area under the ROC curve for the total data was 0.699 (95% confidence interval 0.62-0.77) and when tested separately, it was 0.73 (0.64-0.81) for CABG, 0.79 (0.63-0.92) for valve surgery (good discriminatory ability) and only 0.55 (0.26-0.83) for combined procedures. The independent predictors of mortality determined for the total data were low ejection fraction (odds ratio [OR] - 1.7), preoperative intra-aortic balloon pump (OR - 10.7), combined procedures (OR - 5.1), dialysis dependency (OR - 23.4), and re-operation (OR - 9.4).
The Parsonnet score yielded a good predictive value for valve surgeries, moderate predictive value for the total data and for CABG and poor predictive value for combined procedures.
验证Parsonnet评分模型在印度背景下预测成人心脏手术后死亡率的能力。
本研究纳入了2010年1月至2011年4月期间连续接受成人心脏手术的889例患者。为每位患者确定Parsonnet评分，并评估其对院内死亡率的预测能力。对全部数据以及冠状动脉旁路移植术（CABG）、瓣膜手术和联合手术（CABG合并瓣膜手术）亚组分别进行Parsonnet评分的验证。使用Hosmer-Lemeshow拟合优度检验和受试者工作特征（ROC）分析进行模型校准以评估区分度。通过多因素回归分析从Parsonnet评分中使用的变量评估死亡率的独立预测因素。
总体死亡率为6.3%（56例患者），CABG为7.1%（34例患者），瓣膜手术为4.3%（16例患者），联合手术为16.2%（6例患者）。全部数据以及各亚组的Hosmer-Lemeshow统计量均<0.05，表明使用Parsonnet评分预测的结果与观察到的结果不匹配。全部数据的ROC曲线下面积为0.699（95%置信区间0.62 - 0.77），单独测试时，CABG为0.73（0.64 - 0.81），瓣膜手术为0.79（0.63 - 0.92）（具有良好的区分能力），联合手术仅为0.55（0.26 - 0.83）。为全部数据确定的死亡率独立预测因素为低射血分数（比值比[OR] - 1.7）、术前主动脉内球囊泵（OR - 10.7）、联合手术（OR - 5.1）、透析依赖（OR - 23.4）和再次手术（OR - 9.4）。
Parsonnet评分对瓣膜手术具有良好的预测价值，对全部数据和CABG具有中等预测价值，对联合手术具有较差的预测价值。
