Palliative healthcare professionals give greater weight to advice from a prognostic algorithm rather than advice from another healthcare professional when predicting 2-week survival: the ADJUST study


Year:

Session type:

Andrea Bruun1, Nicola White1, Linda Oostendorp1, Victoria Vickerstaff1, Adam Harris1, Christopher Tomlinson2, Steven Bloch1, Patrick Stone1
1University College London (UCL), 2Imperial College London

Abstract

Background

Healthcare professionals’ (HCPs) survival estimates are often inaccurate. In cases of doubt, a second opinion or a prognostic algorithm can help predict survival. There is limited research on how HCPs differentially weight advice from colleagues or prognostic algorithms when providing survival estimates. This study assessed the extent to which HCPs altered survival estimates, depending on perceived source of advice.

Method

Participants (palliative care doctors, nurses, and other HCPs) completed an online decision-making task (ClinicalTrials.gov;NCT04568629). Recruitment was via UK hospices or online seminars (October 2020 – April 2021).

Participants were randomised into one of two arms: those in arm 1 were informed that advice came from a validated prognostic algorithm; those in arm 2 were informed that advice came from another HCP. Both groups in fact received the same advice which came from a prognostic algorithm that has been shown to be as accurate (but no more accurate) as a clinical prediction of survival.

Participants reviewed five vignettes. The task involved three steps: (1) participants provided an initial probability estimate of 2-week survival (0% ‘certain death’ – 100% ‘certain survival’) based on clinical information provided; (2) participants received advice perceived to come from either another HCP or a prognostic algorithm (according to study arm); (3) participants provided a final (possibly revised) probability estimate.

Weight of Advice (WOA) was calculated for each vignette. WOA represents the difference between initial and final estimates (0 ‘no change from initial estimate’ – 1 ‘complete change to the advice received’). Means of WOA for the two arms were compared.

Results

285 out of 323 HCPs (88%) completed all five vignettes. Participants who perceived advice coming from a prognostic algorithm changed their final estimate significantly more towards the advice (WOA = 0.43) than those who perceived advice coming from another HCP (WOA = 0.31) (coef. = -.12 [95% CI -.17 – -.07], p<0.001).

Conclusion

HCPs are more likely to alter their prognosis in response to advice perceived as coming from a prognostic algorithm than from another HCP, even if the algorithmic advice is no more accurate than another HCP.

Impact statement

Understanding how HCPs respond to advice from different sources may help to improve prognostic decision-making.