Death certificate only proportions should be age adjusted in studies comparing cancer survival across populations and over time

Background The proportion of cases notified by death certificate only (DCO) is a commonly used data quality indicator in studies comparing cancer survival across regions and over time. We aimed to assess dependence of DCO proportions on the age structure of cancer patients. Methods Using data from a national cancer survival study in Germany, we determined age specific and overall (crude) DCO proportions for 24 common forms of cancer. We then derived overall (crude) DCO proportions expected in case of shifts of the age distribution of the cancer populations by 5 and 10 years, respectively, assuming age specific DCO proportions to remain constant. Results Median DCO proportions across the 24 cancers were 2.4, 3.7, 5.5, 8.5 and 23.9% in age groups 15-44, 45-54, 55-64, 65-74, and 75+, respectively. A decrease of ages by 5 and 10 years resulted in decreases of cancer specific crude DCO proportions ranging from 0.4 to 4.8 and from 0.7 to 8.6 percent units, respectively. Conversely, an increase of ages by 5 and 10 years led to increases of cancer specific crude DCO proportions ranging from 0.8 to 4.8 and from 1.8 to 9.6 percent units, respectively. These changes were of similar magnitude (but in opposite direction) as changes in crude 5-year relative survival resulting from the same shifts in age distribution. Conclusions The age structure of cancer patient populations has a substantial impact on DCO proportions. DCO proportions should therefore be age adjusted in comparative studies on cancer survival across regions and over time.