The comparison of the coding behaviors of death certificates for diabetic patients among physicians of different countries reported by Lu et al. (1) is interesting and provides some explanations for a lower proportion of mortality from cardiovascular disease (CVD) coded as the underlying cause of death in Taiwanese diabetic patients (2). However, the data presented should be interpreted with caution in order to avoid overinterpretation.
Based on a lower coexistence of CVD and diabetes in death certificates with mention of diabetes, Lu et al. concluded that their results partially supported the first hypothesis of a less likely experience of CVD death in Taiwanese diabetic patients than in their counterparts in Sweden and the U.S. This conclusion should at least be based on a strict assumption that the behavior of physicians among different countries to simultaneously enter CVD and diabetes in the death certificates for diabetic patients having CVD was not different. However, this presumption was not based on evidence and was actually contradictory to their later conclusion of different behaviors of filling the death certificates among physicians of various countries.
The third hypothesis suggested by Lu et al. that “coders in Taiwan were more likely to assign diabetes as the underlying cause of death than in other countries” was supported by their comparison analyses. However, the use of multiple-cause-of-death data could at most reflect the prevalence of certain diseases in the deceased patients, and it is still unknown whether the actual “cause of death” can be attributed to CVD.
The use of data from the Asia Pacific Collaboration cohort study (3,4) to contradict the first hypothesis could be misleading. First, the hazards ratios reported in that study compared diabetic versus nondiabetic subjects in the Asian and Australasian populations, respectively; Asian diabetic patients were not compared with Australasian diabetic patients (3). Second, the analyses in that study grouped all Asian populations from China, Japan, Hong Kong, Singapore, and Taiwan, not individuals specifically from Taiwan. Third, the two cohorts from Taiwan were not representative of general diabetic patients in Taiwan, because the Kinmen Neurological Disorders Survey recruited residents of Kinmen Island with an age range of 50–93 years and the Cardiovascular Disease Risk Factors Two Township Study recruited residents of two townships (one Hakka community and one Fukienese community) in Taiwan Island Proper with an age range of 20–92 years (4). Both cohorts excluded a significant proportion of younger diabetic patients, and the case numbers of diabetic patients were relatively small (∼219 and 155, respectively) (3). With a median follow-up of 2.9 and 6.0 years in the respective cohorts, there would not be sufficient mortality cases for making significant inferences for Taiwanese diabetic patients.
It should also be mentioned that the approach of Lu et al. of studying cause of death of diabetic patients using death certificates without following a cohort of diabetic patients might have neglected a high proportion of the deceased diabetic patients not having diabetes mentioned in their death certificates.
