Individual heterogeneity, educational attainment and cardiovascular mortality: a pooled analysis of Norwegian health surveys.

Background: Smoking, physical inactivity, cholesterol level and systolic blood pressure are well-established risk factors for cardiovascular diseases (CVD). However, even among individuals with similar levels of these factors, a substantial degree of variation in risk may still remain. We investigated the variation in this unexplained risk across educational levels. Methods: The study population (N=451 800) was from Norwegian health surveys and linked to the National Education Database and the Norwegian Cause of Death Registry. We used survival analysis with frailty models to measure unobserved heterogeneity (frailty variation). Models were stratified by three educational levels. We highlight the degree of heterogeneity by presenting Gini coefficients and indicate how much the unobserved heterogeneity differed across levels of education from the estimated parameters of the frailty distributions. Lorenz curves were plotted for a graphical representation of inequalities in individual risk. Main results: The estimated frailty variances were 24.96, 34.12 and 42.37 in the low, middle and high education groups, respectively, before adjusting for risk factors and 3.76, 7.12 and 7.82 after adjusting for risk factors. The corresponding Gini coefficients were 0.75, 0.84 and 0.86 for low, middle and high education groups, respectively. Conclusion: A large share of the variation in CVD mortality was explained by the observed risk factors. However, a substantial individual unobserved variation in CVD mortality remained after adjusting for these risk factors. Our findings indicate that the unobserved variation in CVD mortality could vary somewhat between educational groups, but that education does not explain the major share of the remaining substantial heterogeneity.