Background Research suggests that given the choice, most people would prefer to die at home. However, patients with haematological malignancies (leukaemia, lymphoma and myeloma) are much more likely to die in hospital than people with other conditions. The reasons underlying this are the subject of much speculation, but there is little evidence to support the various conjectures made. This study investigated end-of-life issues in blood cancer patients, focussing on preferred and actual place of death.

Methods Data are from the Haematological Malignancy Research Network, a specialist population-based cohort that follows all patients from diagnosis to death and abstracts detailed clinical information from hospital records (www.HMRN.org). The present study, which is based on adult patients diagnosed since September 2004 who died between September 2011 and August 2012, collected additional data about all place of death discussions, the people involved in these and decisions made. All analyses were conducted using Stata 14.

Results Overall, 886 patients were included in the study. Hospital death was the most common (56.9%), followed by home (20%), care home (12.1%) and hospice (10.2%). Evidence that preferred place of death had been discussed was found for 52.0% of patients, of whom 61.6% had multiple (up to eight) discussions. Preferred place of death changed as patients moved across the disease trajectory, with the proportions at first and final discussion respectively being: home 51.4% vs 41.9%; hospital 11.1% vs 17.1%; hospice 12.4% vs 16.9%; care home 10.9% vs 14.1%; and other 0.7% vs 1.7%. Hospital was significantly more likely to be preferred when the discussion closest to death did not involve the patient (OR 3.6, 95% Confidence Intervals (CI) 1.99–5.57). The likelihood of hospital being expressed as a preference was also greater in patients/families in the 7 days before death than it was several months before (<7 days vs >8 weeks, OR 13.9, 95% CI 3.31–58.49). There was also a suggestion, albeit non-significant, that home circumstances might impact on the decision, with hospital more likely to be preferred when patients lived with others than alone (OR 1.7, 95% CI 0.97–2.99).

Conclusion Preferred place of death is a complex concept, with many interrelated factors. We found variations depending how often it was discussed (i.e. proximity to death); who was asked (patients or relatives); and who the patient lived with. Ignoring the dynamic nature of these decision processes risks over-simplifying preferences, and failure to develop flexible systems that can respond to changes.