Internal Validation of a Risk Score for Prediction of Postoperative Atrial Fibrillation after Cardiac Surgery

Abstract Background: Postoperative atrial fibrillation (POAF) after cardiac surgery has great clinical and economic implications. Many attempts have been made to identify risk factors aiming at a better evaluation of prophylactic treatment strategies. Objective: To perform an internal validation of a risk score for POAF. Methods: A prospective cohort of 1,054 patients who underwent myocardial revascularization and/or valve surgery was included. The risk score model was developed in 448 patients, and its performance was tested in the remaining 606 patients. Variables with a significance level of 5% in the cohort were included and subjected to a multiple logistic regression model with backward selection. Performance statistics was performed using the c-statistic, the chi-square and the Hosmer-Lemeshow (HL) goodness-of-fit, Pearson's correlation coefficient. Results: Four variables were considered predictors of outcome: age (≥ 70 years), mitral valve disease, the non-use or discontinuation of beta-blockers and a positive water balance (> 1,500 mL). The ROC curve was 0.76 (95% confidence interval [CI]: 0.72-0.79). The risk model showed a good ability according to the performance statistics - HL test x(2) = 0.93; p = 0.983 and r = 0.99 (Pearson's coefficient). There was an increase in the frequency of POAF with the increase of the score: very low risk = 0.0%; low risk = 3.9%; intermediate risk = 10.9%; and high risk = 60.0%; p < 0.0001. Conclusion: The predictive variables of POAF allowed us to construct a simplified risk score. This scoring system showed good accuracy and can be used in routine clinical practice.

摘要
### 背景
术后心房颤动（Postoperative atrial fibrillation, POAF）作为心脏手术后的并发症，对临床诊疗与医疗经济成本均具有重要影响。既往已有诸多研究尝试识别其危险因素，以期更科学地评估预防性治疗策略。

### 研究目的
对术后心房颤动风险评分开展内部验证。

### 研究方法
本研究纳入1054例行心肌血运重建术及/或瓣膜手术的前瞻性队列患者。该风险评分模型以448例患者作为建模集，剩余606例患者用于检验模型性能。将队列中显著性水平为5%的变量纳入分析，通过向后逐步筛选法构建多因素logistic回归模型。采用c统计量、卡方检验、Hosmer-Lemeshow（HL）拟合优度检验以及Pearson相关系数进行模型性能评估。

### 研究结果
最终筛选出4个预后预测变量：年龄≥70岁、二尖瓣疾病、未使用或停用β受体阻滞剂（beta-blockers），以及体液正平衡＞1500mL。受试者工作特征（ROC）曲线下面积为0.76（95%置信区间[CI]：0.72~0.79）。性能评估结果显示该风险模型具备良好的区分度与校准度：HL检验χ²=0.93，P=0.983；Pearson相关系数r=0.99。随着风险评分升高，术后心房颤动的发生频率显著上升：极低危组0.0%、低危组3.9%、中危组10.9%、高危组60.0%（P＜0.0001）。

### 研究结论
本研究筛选出的术后心房颤动预测变量可用于构建简化风险评分系统。该评分系统具有良好的预测准确性，可应用于临床日常诊疗实践。