Article Text

Download PDFPDF

P-230 Artificial intelligence in palliative care: a systematic review to identify its scope of use
  1. Osamah Ahmad1,
  2. Sarah Stanley2,
  3. Stephen Mason3 and
  4. Amara Nwosu4,5 6
  1. 1University of Liverpool, Liverpool, UK
  2. 2Marie Curie Hospice Liverpool, Liverpool, UK
  3. 3Palliative Care Unit, Liverpool, UK
  4. 4Lancaster Medical School, Lancaster, UK
  5. 5Marie Curie Hospice Liverpool, Liverpool, UK
  6. 6Liverpool University Hospitals NHS Foundation Trust, Liverpool, UK


Background New technologies, such as artificial intelligence (AI), supported by novel ways of linking and analysing data, are transforming the way that healthcare data is analysed. AI is increasingly being used to support healthcare delivery, and examples of palliative care application are emerging. AI is an umbrella term covering a variety of intertwined sub-concepts: machine learning refers to machine algorithms automatically improving themselves through experience, and neural networking refers to a form of this mimicking the way the human brain works. Deep learning is another form of machine learning, and natural language processing may refer to a variety of AI algorithms used to understand text intended for human recipients. However, the current scope of (and potential) use of AI in palliative care delivery has not been fully explored. The aim of this project was to define the scope of use of AI methodologies in palliative care studies.

Methods A systematic review of literature was conducted in accordance with the PRISMA guidelines. Four electronic databases were searched, in addition to grey literature searches. AI was used as an umbrella term to include keyword searches for the following: machine learning, deep learning, neutral networks and natural language processing.

Results Twenty-seven relevant articles were selected. The majority of studies described people with cancer (n=10, 37%), from general palliative (n=8, 30%) and intensive care populations (n=4, 15%). Studies using natural language processing were most common (n=12, 44%), with others mainly utilising machine learning (n=10, 37%), deep learning (n=3, 11%) and neural network (n=2, 8%) methodologies. A variety of outcomes were covered, with most studies predicting survival (n=8, 30%), identifying goals of care (n=6, 22%), analysing serious illness conversations (n=2, 9%) and reporting if palliative care best practice recommendations had been followed in clinical care (n=2, 9%).

Conclusion Most palliative care AI studies report cancer, use natural language processing and machine learning methods, to predict survival and analyse goals of care. Future studies need to explore how different AI methods can support palliative care, whilst carefully assessing the risks and limitations, to ensure effective use in the management of serious illness.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.