PURPOSE: To explore agreement on cancer occurrence and site among Medicare
Part A, Massachusetts Cancer Registry, and death certificates.
METHODS: We linked these data sources with the cohort of the population bas
ed East Boston Senior Health Project, a component of the National Institute
on Aging's Established Populations for Epidemiologic Studies of the Elderl
y. The cohort consists of 905 subjects dying between January 1986 and Decem
ber 1990.
RESULTS: We detected the following agreements on cancer occurrence: hospita
lization data and death certificates (kappa = 0.70), hospitalization and ca
ncer registry data (kappa = 0.59), and cancer registry and death certificat
e data (kappa = 0.50). Measures of agreement changed little when the analys
es were stratified by age, sex, calendar year and place of death, autopsy p
erformance, cigarette smoking or alcohol consumption. Site-specific agreeme
nts were higher for colorectal and respiratory tract cancer compared to bre
ast and prostate across ail three comparisons.
CONCLUSIONS: The results should assist epidemiologists to better understand
the strengths and limitations of these data sources. (C) 1998 Elsevier Sci
ence Inc.