The utility of the surprise question in identifying patients most at risk of death.

Abstract

8Background: Understanding the cancer patient’s prognosis in all illness phases is important. Evidence suggests that the “Surprise Question” (SQ) -- “Would you be surprised if this patient died within the next year?” -- may be useful in identifying those most at risk of death, but prior studies are limited by the relatively small number of patients and clinicians included. Methods: From July 2012 to October 2014, oncology clinicians at Dana-Farber Cancer Institute were invited to enroll in a randomized controlled trial on the impact of a structured intervention to improve conversations about end-of-life goals -- the Serious Illness Care Program. The SQ was asked about each patient seen by the enrolled clinician. We used a weighted propensity score approach to calculate adjusted proportions of survival at 6 months and 1 year, clustering by SQ clinician. To determine which variable was most predictive of death, we fit a multivariable Cox model, and found the variable that led to the largest increase in the Cox-regression goodness-of-fit c-statistic. Results: 81 oncology clinicians (59 oncologists, 18 nurse practitioners, and 4 physician assistants) enrolled in the study and we had complete data on 4617 patients: 3821 (83%) for whom the clinician answered “Yes” to the SQ and 796 (17%) for whom the answer was “No.” Propensity-adjusted 1-year survival for “Yes” patients was 93% (95%CI 91-96%) compared to 53% (95%CI 46-60%) for the “No” patients (p < 0.0001). The SQ was more predictive of patient death than type of cancer, age, cancer stage, or time since diagnosis. Sensitivity of the “No” response was 59% (95%CI 49-68%) and specificity was 90% (95%CI 86-93%). Positive predictive value was 49% (95%CI 45-54%) and negative predictive value was 93% (95%CI 90-95%). Conclusions: The Surprise Question identifies cancer-center patients at high risk of death within one year better than clinical variables such as cancer type, stage, patient age, or time since diagnosis.