Context: There is currently no way of observing what primary care end of life care (EoLC) is like, therefore no way of providing systematic EoLC improvement. We have developed a method to gather primary care data when a death is registered. This retrospective study tests the feasibility of the approach. Objective: To develop a minimum dataset from data automatically collected from general practice software, and GP observational data. Design: Cross-sectional retrospective data collection. Setting: Australian general practices in Queensland, Victoria and Western Australia. Participants: Fifty-seven deceased patients reported by eleven GPs from the last three years. Findings: GPs provided extensive communication in the pre-death period. This included phone consultations (50%), home visits (60%), family meetings (49%), hospice or hospital consultations (22%), GP management planning (63%) and patient or family counselling (62%). GPs considered themselves as the principal care coordinator in 53% of reported cases, and part of the wider management team in 29% of reported cases. Most (68%) were satisfied at their level of involvement at the end of life. Palliative care units (87%) and hospitals (60%) provided communication within the last week of the patient’s life, mainly by fax or phone. RACFs communicated mainly by phone. Implication(s) for practice: A minimum dataset is urgently required guide practice and policy changes to meet escalating numbers of deaths. Communication and care coordination was a major part of these GP’s roles, who were satisfied with the role they played. Facilities seem aware of the need to liaise with the patient’s GP.