Background: Linking cancer registry and administrative data can reveal health care use patterns among cancer patients. The Canadian Cancer Registry (CCR) contains personal health insurance numbers (HINs) that facilitate linkage to hospitalization information in the Discharge Abstract Database (DAD).
Data and methods: Valid HINs, captured in the CCR or obtained through probabilistic linkages to provincial health insurance registries, were used to deterministically link prostate, female breast, colorectal and lung cancers diagnosed from 2005 through 2008 with the DAD for fiscal years 2004/2005 to 2010/2011.
Results: At least 98% of tumours diagnosed from 2005 through 2008 had valid HINs in the CCR or obtained through probabilistic linkages. For provinces submitting day surgeries to the DAD, linkage rates to at least one DAD record were higher for female breast (95.6% to 98.1%), colorectal (96.9% to 98.7%) and lung cancers (92.8% to 96.3%) than for prostate cancers (77.2% to 91.6%). Among linked records, agreement was high for sex (99% or more) and complete date of birth (97% or more); the likelihood of a consistent diagnosis in the CCR and on at least one linked DAD record was higher for female breast (86.8% to 97.2%), colorectal (94.6% to 97.7%) and lung cancers (90.3% to 95.5%) than for prostate cancers (77.4% to 87.8%).
Interpretation: Deterministically linking the CCR and DAD using personal HINs is a feasible and valid approach to obtaining hospitalization information about cancer patients.
Keywords: Administrative databases; data linkage; health care; health insurance number.