International cancer survival comparisons indicate the first three months after diagnosis as a key period to explain much of the relative survival variations and this suggests this time is where greatest survival gains in the UK are to be found. However, little is understood about factors associated with early death in cancer. This PhD aims to investigate factors associated with early death with a view to identifying targets for improving cancer outcomes. The PhD study comprised five main parts starting with a literature review followed by a study of routine cancer registry data in an area with high quality data to establish current evidence and trends in early death in cancer. A secondary analysis of the International Cancer Benchmarking Partnership (ICBP) ‘Awareness and Beliefs about Cancer’ survey was undertaken to investigate the role of barriers in anticipated time to help seeking for possible cancer symptoms. This was followed by a nested case control study of factors associated with dying with three months of diagnosis of colon cancer in Northern Ireland and lastly a cohort study investigating the role of General Practice referral patterns in explaining variation in cancer mortality. Five year relative survival for colon, rectal and breast cancer improved significantly since the early 1990s with little evidence of improvement in lung cancer. However, outcomes at 3 months from diagnosis continue to contribute significantly to excess deaths with the highest rates attributed to the oldest patients. The case control study found early death associated with those unmarried, living alone and the most deprived as were constitutional symptoms, more frequent GP use and A&E attendance before diagnosis. Secondary analysis of the International ABC survey highlighted the strong association between perceived barriers to seeing a GP and anticipated time to help seeking for possible cancer symptoms though provided limited evidence for such barriers explaining international cancer survival differences. There was limited evidence of an association between higher rates of urgent suspected cancer referral in General Practice and cancer mortality, though an association with earlier stage disease suggests this deserves further investigation. This thesis contributes additional research to the evidence base on early cancer death and identifies insights into population groups that should be targeted to improve outcomes in the early stages after diagnosis. It is clear that earlier diagnosis should be a urgent priority for this group of patients, in particular pre-hospital delays should be addressed.