Completeness of incidence registration

In the Nordic countries the accuracy and completeness of cancer incidence registration are known to be high, probably better than in any other part of the world.

In Sweden, however, there is a 4% underregistration of incident cases because the Swedish system does not use death certificates as a source of information. In addition to the cases which would be based on death certificate only (DCO), those cases which actually are diagnosed before death but which could only be traced starting from the death certificate information are also missed in Sweden. In other Nordic countries there is a follow-back system (inquiries sent to the treating hospitals) which finds additional information for most of the cases first known through death certificates. The final proportion of DCO cases, e.g., in Finland is less than 1%. The underregistration in Sweden varies by cancer sites, but even in cancers with the highest proportion (leukaemia 18%; Mattsson 1984) the effect to the cancer incidence maps is only about one step in the colour scale.

There is no real information about the completeness of cancer registration in the remaining areas of this Atlas. There are probably shortcomings much larger than the well-documented incomparability problem in Sweden. In some site-specific maps there are obvious classification problems (cf. result chapters), but in general is seems that a satisfactory comparability between areas was reached.

Mortality statistics are often prepared in a routine manner by statistical offices mainly for administrative purposes rather than for scientific use. Therefore less effort is put in confirming the accuracy of the diagnoses or details in codings. In the present Atlas there are many examples of mortality exceeding incidence. In some extreme situations (in a case of a fatal disease with a rapidly decreasing incidence) this might be possible but the examples in this Atlas rather indicate an error: either the incidence rate is too low (underregistration) or mortality too high (e.g., metastases coded as primary cancers, c.f., mortality/incidence ratio for liver cancer).

As far as the completeness of registration is similar for both incidence and mortality, the mortality/incidence rate ratio can be considered as a rather good proxy for survival. The main factor which may make the mortality/incidence ratios incomparable between different areas is the co-mortality due to causes other than the cancer in question.

Despite the weaknesses mentioned above it seems that the pattern of country-specific mortality/incidence ratios rather well resembles that obtained in relative survival rates in the sophisticated EUROCARE program (Berrino et al. 1999). Therefore one may believe that also the variation seen within countries really reflects differences in cancer patient survival. In practically all cancer types the mortality/incidence ratio was lowest (i.e., survival highest) in the Nordic countries except in Denmark, and highest in the post-socialistic countries.

The observation period for Russia (both incidence and mortality) and Lithuania (mortality) only included two or three of the last years of the ten-year study period of this Atlas; i.e., there is a 3.5-4 year postponement in the mid-point of observation periods in comparison to other countries. This produces too high rates for cancer sites with increasing incidence and too low rates for decreasing sites. The maximal annual changes in the site-specific rates in the 1980s in most countries were below 5% (or some 15% in 4 years). This would correspond to a maximal comparability error of one colour step in the relative scale between Russia and Lithuania in comparison with the other countries. If there is underregistration, this may compensate the error in increasing sites but add it for decreasing sites.