Clinicians' intuitions about when terminally ill patients will die are often inaccurate

A simple method, routinely used by clinicians to help identify patients who may be approaching their last year of life, is frequently inaccurate, according to a new study led by UCL researchers and funded by Marie Curie.

The "Surprise Question" was developed as a way to recognise those patients who might benefit from palliative care. It typically asks clinicians to consider the question: "Would you be surprised if this patient died within the next 12 months?"

The study, conducted by researchers at the Marie Curie Palliative Care Research Department at UCL and published in BMC Medicine, looked at 26 previously published studies comprising 25,718 predictions made by clinicians using the "Surprise Question" over a 10-year period.

The findings reveal that the accuracy of predictions varied considerably, with clinicians tending to over-predict the number of people whom they thought would die. Over half (54 percent) of those predicted to die within a specified time period, lived longer than expected.

Clinicians made inaccurate predictions about one third of the patients who did die. Overall, they were incorrect in a quarter (25 percent) of cases.

However, clinicians were considerably better at identifying which patients would survive rather than which patients would die. They were accurate in 84 percent of these cases.

There was some evidence to suggest that the Surprise Question was slightly more accurate when used with cancer patients rather than patients with other illnesses (79 percent accurate compared to 72 percent), supporting the idea that patients with a cancer diagnosis have a more predictable disease trajectory than other illnesses.

Senior author Professor Paddy Stone (Marie Curie Palliative Care Research Department at UCL) said: "Clinicians are often, understandably, reluctant to make specific prognostic predictions and so the Surprise Question plays a role in encouraging them to identify those of their patients who may benefit from palliative or end of life care, without the need for a precise survival prediction."

"But while it does have some utility in this respect, it falls short as an accurate predictor of survival – in terms of over-identifying the number of people whom it predicts will die and missing a proportion of those who do go on to die – and therefore the search is on for more accurate ways to recognise when patients are dying, that are both more sensitive and more specific."

The Surprise Question forms part of the Gold Standards Framework proactive indicator guidance which is widely used in hospitals, hospices and General Practices to help identify patients who may be approaching the last year of life.

Incorrect identification of patients could mean that some miss out on palliative care which should be provided when a patient has been identified as terminally ill. If patients are being over-identified as terminally ill, they may be unnecessarily referred to services for which there are limited resources.

The researchers emphasise that the Surprise Question is not specifically designed or promoted as a way of predicting survival. The purpose of the Surprise Question is to raise clinicians' awareness about which of their patients may benefit from palliative care. However, the findings of the study highlight the need to develop more accurate ways of identifying such patients.

The team are now carrying out further work to develop more accurate and objective ways of predicting survival. The team are also trying to understand the decision making processes underlying clinicians' predictions, they wish to find out why some clinicians' make more accurate predictions and whether it's possible to learn from these experts and teach others to be more precise.

Professor Bill Noble, Medical Director for Marie Curie, said: "An accurate prognosis concerning the length of a terminal illness can be a really difficult thing for clinicians to get right, even for people in the advanced stages. While we may be able to improve the accuracy of predictions, these will ultimately always be expressed in terms of risk of death within a particular time frame. No two people are the same – every illness carries a variety of different possible outcomes depending on the individual and the treatment they are receiving. What is important is that clinicians provide the best possible palliative care based on the individual's need, regardless of how long they expect them to survive."