The impact of coding process on observed cancer mortality trends in Switzerland

Official cancer mortality in Switzerland decreased by about 16% over the 9-year period 1990-1998 and this trend has often been used to suggest that secondary prevention by screening for breast cancer could be useless. However, the clear downshift observed between 1994 and 1995 for some cancers, such as female breast and prostate, and the simultaneous change in ICD classification used by the Federal Office for Statistics in 1995 (ICD-8 to lCD-10) could be related, suggesting an impact of coding process on the observed trend. For every death occurred between 1980 and 1999, the death certificates have been retrieved, the cause of death has been recoded and site-specific mortality rates have been calculated again for each year during this period. As suggested, the trend appears to be overestimated: in order to be comparable with current rates, the mortality observed before 1995 should be lowered by about 7% for men and 5% for women. The error may be partially due to attributing the cause of death to comorbidity factors not normally (and nowadays) defined as the underlying cause. Logically, the impact of such a miscoding is more important among older people and for cancer sites with long survival. For instance, the correction should be around 15% for female breast, 12% for prostate and up to 40% for testicular cancer. (C) 2004 Lippincott Williams Wilkins.