Trends in mortality rates for death-certificate-coded diabetes mellitus in an English population 1979-99

Aims Mortality statistics have customarily been coded and analysed using only one underlying cause of death. Rules for selecting the underlying cause, when more than one cause is certified on a death certificate, have changed twice in England over the past 20 years. We used data from death certificates for 1979-99 to compare mortality rates for diabetes mellitus certified anywhere on death certificates with those certified as the underlying cause. Methods Analysis of data from 18 917 death certificates that included diabetes mellitus in the former Oxford health region. Results Based on the underlying cause of death, mortality rates for diabetes varied substantially between the periods defined by rule changes. Based on mentions of diabetes anywhere on the death record, mortality rates were almost unchanged over time: they showed a non-significant rise of 0.1% per year (95% confidence interval -0.3, 0.6). Circulatory diseases were certified causes of death in 71% of all deaths in people with diabetes. Although mortality rates from circulatory diseases in the general population fell by 2.5% per year, rates for circulatory diseases in combination with diabetes did not fall. Conclusions Two explanations are possible for the lack of change in mortality rates for diabetes based on all certified mentions between 1979 and 1999. Increasing prevalence and improved survival may have resulted in no net change; and/or there may have been no improvement in survival for people whose diabetes is associated with life-threatening pathology and in particular with circulatory diseases.