How robust is the 'surprise question' in predicting short-term mortality risk in haemodialysis patients?

The 'surprise question' (SQ) may aid timely identification of patients with end-of-life care needs. We assessed its prognostic value and variability among clinicians caring for a cohort of haemodialysis (HD) patients.

Clinicians (29 nurses and 6 nephrologists) in each of our 3 HD units were asked to pose the SQ concerning all patients dialysing in their unit. There were 344 patients, 116 in Unit 1, 132 in Unit 2 and 96 in Unit 3.

An adverse SQ response: 'I would not be surprised if this patient were to die in the next 12 months' was reported by individual clinicians for between 6 and 43% of patients (mean 24 ± 9%). Nephrologists responded adversely for more patients than nurses did. Fifty-two patients died during the 12 months of follow-up. There were wide variations between clinicians in the predictive power of SQ responses. Mean odds ratios were significantly higher for nephrologists than for nurses. SQ responses of 49% of clinicians improved baseline models of 12-month mortality, more so for nephrologists (67%) than for senior nurses (50%) and nurses of lesser seniority (36%). Unit performance differed significantly. Agreements between clinicians on SQ responses improved the positive predictive value, i.e. the more clinicians agreed on an adverse response, the greater its predictive power.

SQ provides a unique contribution to the prediction of short-term prognosis in HD patients, though predictive power varies with clinical discipline, seniority and clinical setting. Agreements between clinicians on adverse responses may have clinical utility.