The fall and rise of cancer registration in Estonia: The dangers of overzealous application of data protection.

Background: The population-based Estonian Cancer Registry (ECR) has maintained a database of cancer cases since 1968. Between 2001 and 2007 the ECR was prohibited from linking cancer records to death certificates. In January 2008, the prohibition was lifted, and two years later the ECR was able to begin tracing back unmatched deaths. This paper estimates the effect of the linkage ban on reported cancer incidence and survival.

Methods: Incident cancers in 2001-2007 were extracted from the ECR database in May 2018 to allow for late registrations. Two datasets were created: one with all incident cases and another without death-certificate-initiated (DCI) cases. Using both datasets, age-standardised incidence rates (ASIR) and their ratios; age-standardised five-year relative survival ratios (ARSR) and excess mortality rate ratios were calculated.

Results: In 2001-2007, 46,535 incident cancers were registered in the ECR. Of them, 2299 (4.9 %) were DCI cases. The inclusion of DCI cases increased the ASIR for overall cancer by 6 % in men and 3 % in women. An increase ≥10 % in ASIR for lung, liver and pancreatic cancer was observed. The effect of accrued DCI cases to the ARSR was minor. Excess mortality in the dataset without DCI cases was 4 % underestimated in men and 3 % in women.

Conclusion: Biases in cancer incidence and survival measures generated by the temporary record linkage ban were largely correctable by using trace-back procedures when this became possible. Nevertheless, this type of ban and the arguments put forward to justify it, harm disease registration and register-based research.