[Reliability of cancer as the underlying cause of death according to the Mortality Information System and Population-Based Cancer Registry in Goiânia, Goiás State, Brazil]

Cad Saude Publica. 2014 Feb;30(2):296-304. doi: 10.1590/0102-311X00024813.
[Article in Portuguese]

Abstract

This study assessed the reliability of cancer as the underlying cause of death using probabilistic linkage between the Mortality Information System and Population-Based Cancer Registry (PBCR) in Goiânia, Goiás State, Brazil, from 2000 to 2005. RecLink III was used for probabilistic linkage, and reliability was assessed by Cohen's kappa and prevalence-adjusted and bias-adjusted kappa (PABAK). In the probabilistic linkage, 2,874 individuals were identified for the reliability analysis. Cohen's kappa ranged from 0.336 to 0.846 and PABAK from 0.810 to 0.990 for 14 neoplasm groups defined in the study. For reliability of the 35 leading cancers, 12(34.3%) presented kappa values under 0.600 and PABAK over 0.981. Among the neoplasms common to both sexes, crude agreement ranged from 0.672 to 0.790 and adjusted agreement from 0.894 to 0.961. Sixty-seven percent of cases classified by the Mortality Information System as "cancer of ill-defined sites" were reclassified according to the PBCR. This study was useful for the classification of cancer mortality estimates in areas covered by the PBCR.

Publication types

  • English Abstract

MeSH terms

  • Brazil / epidemiology
  • Cause of Death
  • Female
  • Health Information Systems*
  • Humans
  • Male
  • Neoplasms / classification
  • Neoplasms / mortality*
  • Prevalence
  • Registries*
  • Reproducibility of Results