McEwen et al. (1) conclude that diabetes is much more likely to be reported on the death certificates of diabetic individuals who die of cardiovascular diseases (CVDs). They calculated the odds of recording diabetes anywhere on the death certificate by underlying cause of death (UCOD) and concluded that if the UCOD was CVD, the certifying physician was more likely to report diabetes on the death certificate and that if the UCOD was cancer, the certifying physician was less likely to report diabetes on the death certificate. Their analysis assumes that the certifying physician is solely responsible for determining the UCOD.
In reality, the final, coded UCOD might not be the same as that indicated by the certifying physician. In cases where certifying physicians report an incorrect sequence of causal relationships between conditions, a series of selection rules determined by the World Health Organization is used to select the UCOD (2). In example 1, diabetes was the intended UCOD of the certifying physician. He/she, however, incorrectly reported asthma above diabetes in the sequence. According to selection rule 1, asthma would be selected as the UCOD because asthma is not allowed to be due to diabetes.
Example 1:
a) acute myocardial infarction
b) asthma
c) diabetes
In contrast, in example 2, acute myocardial infarction (AMI) was the intended UCOD of the certifying physician. He/she, however, incorrectly reported diabetes above AMI in the sequence. According to selection rule 2, diabetes would be selected as the UCOD because diabetes cannot be due to AMI.
Example 2:
a) diabetes
b) AMI
We published a recent study indicating that of those deaths in which diabetes was reported in Part I of the death certificates in the U.S., 28% had incorrect causal sequences reported by the certifying physicians (3). As a result, we argue that it is not appropriate to use the final, coded UCOD to retrospectively predict the certification practices of physicians with regard to diabetes.
