Can the ‘surprise question‘ be used to correctly identify people nearing the end of life?

The surprise question (SQ) (“Would you be surprised if this patient died within the next χ months?”) offers an alternative to standard prognostic estimates | BMJ Supportive & Palliative Care


Methods: We searched numerous databases, including: Medline, Embase, CINAHL, AMED. Studies were included if they reported the SQ and were written in English.

Results: Out of the 357 studies identified, 22 were included in the review. In these studies, 25 718 estimates were reported. The results showed a wide variation in the reported accuracy of the SQ, with sensitivity ranging from 11.6% to 96.6% and specificity ranging from 13.9% to 78.6%. The AUROC score across the studies ranged from 0.512 to 0.822. Doctors appeared to be more accurate than nurses at recognising people in the last year of life (c-statistic=0.735 vs. 0.688).

Conclusions: The performance of the SQ varied greatly across the studies. Further work is required to understand the processes by which clinicians arrive at their prognostic estimates, to refine the accuracy of the SQ and to compare its performance against other more sophisticated prognostic tools, particularly in populations where a higher proportions of deaths occur.

Full reference: Vickerstaff, V. et al. (2017) 60 Can the ‘surprise question‘ be used to correctly identify people nearing the end of life?: a review. BMJ Supportive & Palliative Care. Vol. 07 (Issue 03) A371.


How are non-numerical prognostic statements interpreted and are they subject to positive bias?

Frank, clear communication with family members of terminally ill or incapacitated patients has important implications for well-being, satisfaction with care and sound decision-making | BMJ Supportive & Palliative Care


Objectives: Numerical prognostic statements, particularly more negative ones, have been found to be interpreted in a positively biased manner. Less precise non-numerical statements, preferred by physicians, and particularly statements using threatening terms (dying vs surviving) may be even more subject to such biases.

Methods: Participants (N=200) read non-numerical prognostic statements framed in terms of dying or surviving and indicated their interpretation of likelihood of survival.

Results: Even the most extreme statements were not interpreted to indicate 100% likelihood of surviving or dying, (eg, they will definitely survive, 92.77%). The poorness of prognoses was associated with more optimistically biased interpretations but this was not, however, affected by the wording of the prognoses in terms of dying versus surviving.

Conclusions: The findings illuminate the ways in which commonly used non-numeric language may be understood in numeric terms during prognostic discussions and provide further evidence of recipients’ propensity for positive bias.

Full reference: Moyer, A. et al. (2017) How are non-numerical prognostic statements interpreted and are they subject to positive bias?