Background: In mid 1998, a question "Was the deceased a smoker five years ago?" was introduced on the newly revised South African death notification form.
Design: A total of 16 230 new death notification forms from 1998 have been coded, and comparison of the prevalence of smoking among those who died of different causes was used to estimate, by case–control comparisons, tobacco attributed mortality in South Africa. Cases comprised deaths from causes known (from other studies) to be causally associated with smoking, and controls comprised deaths from medical conditions expected to be unrelated to smoking. Those who died from external causes, and from diseases strongly related to alcohol consumption, were excluded.
Subjects: Reports were available from 5340 deceased adults (age 25+), whose smoking status was given by a family member.
Results: Significantly increased risks were found for deaths from tuberculosis (odds ratio (OR) 1.61, 95% confidence interval (CI) 1.23 to 2.11), chronic obstructive pulmonary disease (COPD) (OR 2.5, 95% CI 1.9 to 3.4), lung cancer (OR 4.8, 95% CI 2.9 to 8.0), other upper aerodigestive cancer (OR 3.0, 95% CI 1.9 to 4.9) and ischaemic heart disease (OR 1.7, 95% CI 1.2 to 2.3).
Conclusion: If smokers had the same death rate as non-smokers, 58% of lung cancer deaths, 37% of COPD deaths, 20% of tuberculosis deaths, and 23% of vascular deaths would have been avoided. About 8% of all adult deaths in South Africa (more than 20 000 deaths a year) were caused by smoking.