Flow chart of the death certificate survey in Flanders, Belgium. The initial sample was 6927 of all 54881 deaths in 2007. From non-response analyses we found that response was impossible for 725 deaths eg because the physician did not have access to the patient’s medical file or the patient could not be identified. In total 3623 were returned and the response rate was 58.4% (3623/6202 eligible cases). We selected non-sudden deaths as denominator in all analyses. Considering the educational trajectory is likely not complete in minors and early adults, we elected a cut-off age of 25 years (23 cases deleted). Of the remaining non-sudden deaths, 755 cases (27.7%) had missing values for educational attainment and were discarded. This resulted in a final sample of 1951 patients.