Effects of misattribution in assigning cause of death on prostate cancer mortality rates

Abstract

PURPOSE: Estimate the magnitude of misattribution in death certification and to determine whether misattribution could explain temporal trends in New Mexico prostate cancer mortality. METHODS: Investigators retrospectively reviewed medical records to classify cause of death for men with prostate cancer who died in New Mexico in either 1985 or 1995. The investigator-assigned cause of death, either prostate cancer or another cause, was compared with the New Mexico Bureau of Vital Statistics' (BVS) assignment. Net misattribution was the difference between BVS and investigators in the number of deaths attributed to prostate cancer relative to the number attributed by investigators. RESULTS: Concordance for death certification between BVS and investigators was higher in 1995 (95.8%) than 1985 (85.1%), p <0.01, but net misattribution was higher in 1995 (16.8% vs. 1.9%). During this time, the crude prostate cancer mortality rate in New Mexico increased 33.5%, from 18.5 to 24.7 per 100,000. However, after adjusting for misattribution, mortality increased only 15.9%, from 18.2 to 21.1 per 100,000. CONCLUSIONS: Net misattribution in death certification was higher in 1995 than 1985. The adjusted increase in crude mortality rates was substantially less than the observed increase, suggesting that misattribution explained about half of the observed increase.