Wong Jonathan M, Sin Nancy L, Whooley Mary A
From the School of Medicine (J.M.W.), University of California, Irvine, California; Doris Duke Clinical Research Fellowship Program (J.M.W.) and Departments of Medicine (N.L.S., M.A.W.) and Epidemiology & Biostatistics (M.A.W.), University of California, San Francisco, California; and Veterans Affairs Medical Center (M.A.W.), San Francisco, California.
Psychosom Med. 2014 May;76(4):311-7. doi: 10.1097/PSY.0000000000000059.
Hostility is associated with adverse outcomes in patients with coronary heart disease (CHD). However, assessment tools used to evaluate hostility in epidemiological studies vary widely.
We administered nine subscales of the Cook-Medley Hostility Scale (CMHS) to 656 outpatients with stable CHD between 2005 and 2007. We used Cox proportional hazards models to determine the association between each hostility subscales and all-cause mortality. We also performed an item analysis using logistic regression to determine the association between each CMHS item and all-cause mortality.
There were 136 deaths during 1364 person-years of follow-up. Four of nine CMHS subscales were predictive of mortality in age-adjusted analyses, but only one subscale (the seven-item Williams subscale) was predictive of mortality in multivariable analyses. After adjustment for age, sex, education, smoking, history of heart failure, diabetes, and high-density lipoprotein, each standard deviation increase in the Williams subscale was associated with a 20% increased mortality rate (hazard ratio = 1.20, 95% confidence interval = 1.00-1.43, p = .046), and participants with hostility scores in the highest quartile were twice as likely to die as those in the lowest quartile (hazard ratio = 2.00, 95% confidence interval = 1.10-3.65, p = .023).
Among nine variations of the CMHS that we evaluated, a seven-item version of the Williams subscale was the most strongly associated with mortality. Standardizing the assessment of hostility in future epidemiological studies may improve our understanding of the relationship between hostility and mortality in patients with CHD.
敌意与冠心病（CHD）患者的不良预后相关。然而，流行病学研究中用于评估敌意的工具差异很大。
2005年至2007年期间，我们对656例稳定型CHD门诊患者进行了库克-梅德利敌意量表（CMHS）的九个分量表测试。我们使用Cox比例风险模型来确定每个敌意分量表与全因死亡率之间的关联。我们还使用逻辑回归进行了项目分析，以确定每个CMHS项目与全因死亡率之间的关联。
在1364人年的随访期间有136例死亡。在年龄调整分析中，九个CMHS分量表中有四个可预测死亡率，但在多变量分析中只有一个分量表（七项威廉姆斯分量表）可预测死亡率。在调整年龄、性别、教育程度、吸烟、心力衰竭病史、糖尿病和高密度脂蛋白后，威廉姆斯分量表每增加一个标准差，死亡率增加20%（风险比=1.20，95%置信区间=1.00-1.43，p=0.046），敌意得分处于最高四分位数者死亡的可能性是最低四分位数者的两倍（风险比=2.00，95%置信区间=1.10-3.65，p=0.023）。
在我们评估的CMHS的九种变体中，七项版本的威廉姆斯分量表与死亡率的关联最为密切。在未来的流行病学研究中标准化敌意评估可能会增进我们对CHD患者中敌意与死亡率之间关系的理解。
