80 THE ACCURACY OF CAUSE-OF-DEATH CODING IN THE NETHERLANDS: A STUDY INTO RELIABILITY

P. Harteloh, K. de Bruin, J. Kardaun Statistics Netherlands, THE HAGUE, The Netherlands

Background: Cause-of-death statistics are an important source of information for epidemiological research or policy decisions. The reliability or consistency of these statistics is important for interpreting trends in time or differences between populations. Attribution of observed differences to determinants of health should not be biased by variations in the coding or selection of the underlying cause of death. Objectives: We measured the reliability of cause-of-death statistics in order to improve internal consistency and external interpretations. Methods: A sample of 8,215 death certificates from the month of May 2005 was recoded (manually) by the four coders of Statistics Netherlands. Reliability was measured by calculating agreement between (intercoder) or within (intracoder) coders and by calculating the chance of reproducing the original code for the underlying cause of death. Results: The agreement between coders in coding and selecting the underlying cause-of-death was 78% (intercoder variation: 22%). The (mean) intracoder agreement was 89% (intracoder variation: 11%). The chance of reproducing the original code was 87%. Agreement among coders was associated with the specificity of the ICD-10 code (chapter, three digits, four digits), the nature of the disease (ICD-10 chapter), the age of the deceased and the number of diseases reported on the death certificate. For important causes of death such as cancers and acute myocardial infarction the reliability of cause-of-death statistics appeared to be high ([90%). For chronic diseases such as diabetes and renal insufficiency the reliability was low (\60%). Conclusions: When interpreting cause-of-death statistics reliability should be taken into account. The consistency of mortality statistics could be improved by (automatic application of) coding or selection rules, the interpretation by analysis based not only on underlying causes of death, but on comorbidity as well (multiple cause coding).