Recognising patients who will die in the near future: a nationwide study via the Dutch Sentinel Network of GPs

Abstract

Background: Recognising patients who will die in the near future is important for adequate planning and provision of end-of-life care. GPs can play a key role in this.

Aim: To explore the following questions: How long before death do GPs recognise patients likely to die in the near future? Which patient, illness, and care-related characteristics are related to such recognition? How does recognising death in the near future, before the last week of life, relate to care in during this period?

Design and setting: One-year follow-back study via a surveillance GP network in the Netherlands.

Method: Registration of demographic and care-related characteristics.

Results: Of 252 non-sudden deaths, 70% occurred in the home or care home and 30% in hospital. GP recognition of death in the near future was absent in 30%, and occurred prior to the last month in 15%, within the last month in 19%, and in the last week in 34%. Logistic regression analyses showed cancer and low functional status were positively associated with death in the near future; cancer and discussing palliative care options were positively associated with recognising death in the near future before the last week of life. Recognising death in the near future before patients' last week of life was associated with fewer hospital deaths, more GP-patient contacts in the last week, more deaths in a preferred place, and more-frequent GP-patient discussions about specific topics in the last 7 days of life.

Conclusion: Recognising death in the near future precedes several aspects of end-of-life care. The proportion in whom death in the near future is never recognised is large, suggesting GPs could be assisted in this process through training and implementation of care protocols that promote timely recognition of the dying phase.

Figure 1

Recognition of death in the near future in patients who died non-suddenly, by place and cause of death, n = 252 (includes 5 missing values; percentages of missing observations variables ranged between 0.4% and 5.6%).