National adaptations of the ICD rules for classification--a problem in the evaluation of cause-of-death trends

J Clin Epidemiol. 1997 Apr;50(4):367-75. doi: 10.1016/s0895-4356(96)00426-x.


In 1981, the registration routines of the Swedish cause-of-death register were adjusted. The aim of this study was to assess what influence these changes in registration practice might have had on the cause-of-death trends after 1981. The Eighth Revision of the International Classification of Diseases (ICD-8) was used throughout the study period (1976-1985). Significant changes in the registered number of cases were found in 13 of the 18 diagnostic groups scrutinized. Four main types of outcomes were observed: (a) the number of underlying causes increased while the number of contributing causes decreased or vice versa; (b) the number of both underlying and contributory causes changed in the same direction, due to the transfer of a diagnostic group from one ICD category to another; (c) the number of both underlying and contributory causes changed in the same direction, but not due to the transfer of a diagnostic group; or (d) the number of either underlying or contributory causes changed, but not both. In general, the altered registration practice led to more conditions that are often considered as terminal complications to other diseases being registered as the underlying cause of death. While most of the 1981 instructions meant a more literal application of the ICD-8, those concerning cardiac valvular diseases deviated substantially from it. We conclude that (a) important changes in registration practice may occur at any point in time, and not only in connection with the implementation of a new version of the ICD; and (b) national adaptations of the ICD coding instructions may amount to a reversal of the instructions included in the ICD manuals. These findings must be considered when comparing cause-of-death statistics from different countries, and both underlying and contributing cause-of-death statistics should be considered in such analyses of cause-of-death trends.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Cause of Death / trends*
  • Data Collection
  • Disease / classification*
  • Humans
  • Registries*
  • Sweden / epidemiology