Background: There is some previous evidence that diagnosis of cancer at death, recorded as registry death certificate only records, is associated with problems of access to care. Methods Records from the Northern and Yorkshire Cancer Registry for patients registered with breast, colorectal, lung, ovarian or prostate cancer between 1994 and 2002 were supplemented with measures of travel time to general practitioner and hospital services, and social deprivation. Logistic regression was used to identify predictors of records where diagnosis was at death. Results There was no association between the odds diagnosis at death and access to primary care. For all sites except breast, the highest odds of being a cancer diagnosed at death fell among those living in the highest quartile of hospital travel time, although it was only statistically significant for colorectal and ovary tumours. Those in the most deprived and furthest travel time to hospital quartile were 2.6 times more likely to be a diagnosis at death case compared with those in the most affluent and proximal areas. Conclusions There is some evidence that poorer geographical access to tertiary care, in particular when coupled with social disadvantages, may be associated with increased odds of diagnosis at death.