Leong Kui Toh Gerard, Wong Lai Yin, Aung Khin Chaw Yu, Macdonald Michael, Cao Yan, Lee Sheldon, Chow Wai Leng, Doddamani Sanjay, Richards Arthur Mark
Department of Cardiology, Changi General Hospital, Singapore, Singapore.
Health Services Research Department, Eastern Health Alliance, Singapore, Singapore.
Am J Cardiol. 2017 May 1;119(9):1428-1432. doi: 10.1016/j.amjcard.2017.01.026. Epub 2017 Feb 10.
There are limited accurate 30-day heart failure (HF) readmission risk scores using readily available clinical patient information on a well-defined HF cohort. We analyzed 1,475 admissions discharged from our hospital with a primary diagnosis of HF between 2010 and 2012. HF diagnostic criteria included satisfying clinical Framingham criteria, elevated serum N-terminal pro-natriuretic peptide, and evidence of cardiac dysfunction on transthoracic echocardiography. The patients were randomly divided into 2 groups; 60% were used as the derivation cohort and 40% as the validation cohort. Bivariate analysis and logistic regression were used to develop the model. Weighted risk scores were derived from the odds ratio of the logistic regression model. Total risk scores were computed by simple summation for each patient. The 7 significant independent predictors of 30-day HF readmission used to derive the risk scoring tool were the number of previous HF-related admission in the preceding 1 year, index admission length of stay, serum creatinine level, electrocardiograph QRS duration, serum N-terminal pro-natriuretic peptide level, number of Medical Social Service needs, and β blocker prescription on discharge. The area under the curve was 0.76. Sensitivity and specificity were 78.3% and 60.7%, respectively. The positive predictive value and negative predictive value were 18.9% and 96%, respectively. The actual observed and predicted 30-day heart failure readmission rates matched. In conclusion, we have developed the first 30-day HF readmission risk score, with good discriminatory ability, for an urban multiethnic Asian heart failure cohort with stringent diagnostic criteria. It consists of 7 easily obtained variables.
利用明确界定的心力衰竭（HF）队列中易于获取的临床患者信息，准确的30天HF再入院风险评分有限。我们分析了2010年至2012年间我院出院的1475例主要诊断为HF的患者。HF诊断标准包括满足临床弗明翰标准、血清N末端前脑钠肽升高以及经胸超声心动图显示心脏功能障碍的证据。患者被随机分为两组；60%用作推导队列，40%用作验证队列。采用双变量分析和逻辑回归建立模型。加权风险评分来自逻辑回归模型的比值比。通过对每位患者进行简单求和计算总风险评分。用于推导风险评分工具的30天HF再入院的7个重要独立预测因素是前1年与HF相关的入院次数、本次入院住院时间、血清肌酐水平、心电图QRS时限、血清N末端前脑钠肽水平、医疗社会服务需求数量以及出院时β受体阻滞剂处方情况。曲线下面积为0.76。敏感性和特异性分别为78.3%和60.7%。阳性预测值和阴性预测值分别为18.9%和96%。实际观察到的和预测的30天心力衰竭再入院率相符。总之，我们为诊断标准严格的城市多民族亚洲心力衰竭队列开发了首个具有良好鉴别能力的30天HF再入院风险评分。它由7个易于获得的变量组成。
