Abstract

Objective: To increase the cross-state comparability of diabetes mortality statistics related in the US and Mexico.

Methods: We used multinomial logistic regression to estimate the effects of individual and community factors on a death for which diabetes was recorded as one of the multiple contributing causes of death (MCD) being assigned to diabetes as the underlying cause of death (UCD), versus to cardiovascular, other non-communicable, or communicable diseases. We used the model to estimate state-level diabetes death rates that are standardized in the individual and community factors.

Results: Deaths with diabetes as one of the multiple causes of death were more likely to be assigned to cardiovascular causes if they occurred in hospitals or had a necropsy, and if the decedents were from states with higher BMI and SBP, more educated, or insured. Adjusting for individual- and community-level factors substantially increased the cross-state correlation of diabetes UCD and diabetes MCD mortality rates. The adjustment also reduced the number of direct diabetes deaths by 10% in the USA and by 24% in Mexico. In the US, UCD diabetes deaths declined most in Utah, New Mexico, New Jersey and Louisiana, and increased in California and Hawaii. In Mexico, adjusted diabetes deaths were smaller than observed ones in all states by 3-34%. An additional 126,300 IHD and stroke deaths in the US and 19,497 in Mexico were attributable to high blood glucose.

Conclusions: There is a need to improve the comparability of diabetes cause of death assignment, especially in relation to cardiovascular diseases.