A comparison of the sources of cancer mortality in China.

OBJECTIVE: To compare the validity of mortality data from available sources in China. MATERIALS, METHODS: Two large-scale surveys have provided accurate national-level rates; the most recent involved deaths occurring in a random 10% sample of the population during 1990-1992. Since then, the only readily available sources are two on-going surveillance systems, which provide annual estimates of mortality--the "Disease Surveillance Points" (DSP) sample survey, and that established by the Center of Health Information and Statistics (CHIS) of the ministry of health, the results of which are published by WHO. They were compared with respect to the representativeness of the populations covered and the rates obtained. RESULTS: Neither source covers a random sample of the Chinese population, with respect to age group, sex, and urban-rural residence, although the DSP population is the more representative of the national population in this respect. Sex and region (urban/rural) specific age-standardized mortality rates from the CHIS dataset were, however, closer to those from the (1990-1992) national survey, than those calculated from DSP data. CONCLUSIONS: The CHIS data is the preferred source for estimation of national mortality, and study of time trends, but requires appropriate weighting (by age, sex, rural/urban residence). The within-stratum estimates are more stable than those of DSP, because of its larger sample size.