Introduction We aimed to show that utilising information from a population-based, retrospective mortality survey for the main determinants of disease is feasible.

Methods During 1989–1991, a nationwide retrospective survey of mortality was conducted. To assess the effects of smoking on all cause mortality in Chinese adults, we used a new design within this survey. For all the deceased persons (cases) we sex-matched surviving spouses as controls. Smoking information for cases was provided by spouse or other informants, at the same time as providing their own smoking history.

Results For all cause mortality, the results revealed that tobacco use was associated with 11.2% of total deaths in 1987, and more than two-thirds of these excess deaths occurred between the ages of 50 and 74 years. Although life expectancies varied with region or sex, the years of life lost attributable to smoking was almost the same in each age group. Our study also confirmed that more than 50% of the sex difference in life expectancy was accounted for by smoking. With respect to the novel design, the results revealed consistency in the results using different control groups.

Conclusion This new case-spouse control design as an alternative for control selection in case-control studies is valid and feasible.