The surprise question: predictive accuracy in an unselected emergency department population - a prospective study in nurses and physicians

BackgroundThe 'surprise question' (SQ) asks care professionals to assess the patient's mortality risk. In this prospective study, we investigated 1) the prognostic accuracy of the SQ regarding 3- and 12-month mortality (SQ3/SQ12) answered by nurses and physicians in unselected adult emergency department (ED) patients, and in high risk subgroups (age >= 50 y; medical patients), 2) the agreement between these care professionals.Patients and methodsIn this prospective study, the SQ3 and SQ12 were scored by nurses and physicians. AU-ROC, sensitivity, specificity, and negative predictive value (NPV) were calculated.Kappa values and absolute agreement were calculated.ResultsIn total, 1958 patients were assessed. Mortality within 12 months was 13.9% in all patients (8.3% within 3 months), 22.9% in older and 20.1% in medical patients. The AU-ROC of the SQ3 was 0.639 and 0.698 for nurses and physicians, resp., and that of the SQ12 was 0.722 and 0.847, resp. For SQ3, sensitivity was 46.8-48.0%, and specificity 93.9-95.1%, with high NPV (95.6-97.0%). For SQ12, sensitivity was higher (54.1-60.8%), with specificity of 83.4-96.4%, and high NPV. AU-ROCs for the two high risk subgroups were comparable. Agreement was fair (kappa 0.255) for SQ3, and moderate (kappa 0.461) for SQ12, while absolute agreement was 91.4% and 80.5%, resp.ConclusionsThe study supports the SQ as a simple prognostic tool in the ED, with 12-month prognostic accuracy being especially reliable in high-risk patients. The agreement between nurses and physicians was fair-moderate. The SQ could play an important role in guiding ED care, especially in high-risk patients.