Registry‐based cancer incidence and mortality data are widely used for etiologic research, cancer control and health care monitoring and planning. The complete coverage of all cases is the key criteria of data quality but it is difficult to assess because the alternative sources of data may be flawed. The aim of our study was to examine, at a nationwide level, the completeness of the Swedish Cancer Registry (CR) regarding persons who died of cancer, based on the Cause of Death Registry (DR), and using the Hospital Discharge Registry (HDR) as an additional source of data. Individuals who died of cancer from years 1999 through 2008 recorded in DR were linked to CR and HDR. A total of 190,692 individuals were identified from DR with cancer as the underlying cause of death; the mean identification rate of concordant cancer in CR was 79.8%, depending on tumor site and age at death. Breast, bladder and prostate cancers showed the highest rate of identification, whereas bone, liver and pancreatic cancers showed the lowest rate of identification. CR had no records on 10.6% of cancer cases recorded in DR. Similarly, the identification rate in HDR was 84.5% for concordant cancer and with 9.6% of cases missing. Neither source reported cancers for 3.4% of cancer cases recorded in DR. In conclusion, some 10% of cancer deaths had no cancer records in CR or in HDR, and 3.4% were missing in both sources. The identification rate depended on tumor site, age at death and, to some extent, death outside hospital.