AN ANALYSIS OF RISK-FACTORS FOR PREVALENT CORONARY HEART-DISEASE BY USING THE PROPORTIONAL ODDS MODEL

A proportional odds model (POM) for grades of severity of prevalent coronary heart disease (CHD) is used to explore and contrast risk factors for men and women, by using cross-sectional data from the Scottish heart health study. The grades of CHD used are no disease, angina grade I, angina grade II and myocardial infarction (MI). Separate models are deemed necessary for the two sexes, since some of the risk factors act differently for men and women. After adjustment for other risks, the significant (p < 0.05) risk factors found for men are age, parental history of CHD, housing tenure, fibrinogen, high density lipoprotein-cholesterol, total cholesterol and smoking status. For women, the first five of these are again found significant, plus body mass index, occupational social class and marital status. The proportional odds assumption appears to be satisfied reasonably well for the model applied to the data for males, but the data for females need to be split by age before the assumption is reasonable. The four-grade POM performs better than the binary (logistic regression) model, although a three-grade POM, with angina considered as a single category, appears to be better still. It is concluded that there is no need to consider angina and MI totally separately, nor any advantage in combining them into a single disease definition, in risk factor modelling. The POM is a useful tool which is easy to apply with modern software and deserves more attention in epidemiological investigations.