Ciampi A, Courteau J, Niyonsenga T, Xhignesse M, Lussier-Cacan S, Roy M
Department of Epidemiology and Biostatistics, McGill University, Montréal, Québec, Canada.
Eur J Epidemiol. 2001;17(7):609-20. doi: 10.1023/a:1015587428172.
Family history is commonly used when evaluating coronary heart disease (CHD) risk yet it is usually treated as a simple binary variable according to the occurrence or non-occurrence of disease. This definition however fails to consider the potential components of a family history which may in fact exert different degrees of influence on the overall risk profile. The purpose of this paper is to compare different predictive models for CHD which incorporate family history as either a binary variable or different types of family risk indices in terms of their predictive ability. Models for estimating CHD risk were constructed based on usual risk factors and different family history variables. This construction was accomplished using logistic regression and RECursive Partition and AMalgamation (RECPAM) trees. Our analyses demonstrate the importance of using more sophisticated definitions of family history variables compared to a simple binary approach since this leads to a significant improvement in the predictive ability of CHD risk models.
在评估冠心病（CHD）风险时，家族病史通常被采用，但它通常根据疾病的发生与否被视为一个简单的二元变量。然而，这种定义未能考虑家族病史的潜在组成部分，而这些组成部分实际上可能对总体风险状况产生不同程度的影响。本文的目的是比较不同的冠心病预测模型，这些模型将家族病史作为二元变量或不同类型的家族风险指数纳入其中，并比较它们的预测能力。基于常见风险因素和不同的家族病史变量构建了冠心病风险估计模型。这一构建过程是使用逻辑回归和递归划分与合并（RECPAM）树完成的。我们的分析表明，与简单的二元方法相比，使用更复杂的家族病史变量定义非常重要，因为这会显著提高冠心病风险模型的预测能力。
