Prognostic value of the “surprise question” among UPMC Hillman Cancer Center patients with select stage IV cancer diagnoses.

Abstract

25 Background: The “Surprise Question” — Would I be surprised if this patient died in the next 12 months? — was developed to help clinicians predict when patients are nearing the end of life. Limited data has shown that the “Surprise Question” is modestly predictive of mortality (CMAJ 2017 Apr 3;189(13):E484-E493), though its performance seems to be superior among cancer patients (Palliat Med 2014 Jul;28(7):959-964). Via Oncology Pathways, a platform used by UPMC Hillman Cancer Center and other institutions nationwide to guide treatment decisions, asks physicians the “Surprise Question” when a new treatment plan is implemented for patients with metastatic cancer. We assessed the “Surprise Question’s” ability to predict survival among Hillman Cancer Center patients with select stage IV diagnoses. Methods: We queried the UPMC Hillman Cancer Center Registry Information and Reporting Services for cases of colorectal, non-small cell lung, prostate, pancreatic, and breast cancer with clinic visits between 1/1/2016 and 12/31/2017 and residence in Allegheny County, the primary referral base for the UPMC Hillman Cancer Center network’s flagship facility. Results: The “Surprise Question” was completed for 1,584 patients with metastatic disease of the 5,330 patients that were screened. “No” was the response for 891 patients (56.3%). Mortality at 12 months for patients for whom the answer to the “Surprise Question” was “no” was 63.1%, compared to 32.5% for patients for whom the answer was “yes” (P < 0.0001). The sensitivity of the “Surprise Question” was 71.4% (95% CI 69.0 – 73.8%), and the specificity was 58.7% (95% CI 56.3 – 61.0%). The positive predictive value was 63.1% (95% CI 60.9 – 65.2%) and negative predictive value 67.5% (64.8% – 70.2%). Finally, the positive likelihood ratio was 1.73 (95% CI 1.58 – 1.89) and negative likelihood ratio 0.49 (0.43 – 0.55). Conclusions: While a “no” response to the “Surprise Question” for UPMC oncology patients with select stage IV diagnoses was more likely to predict 12-month mortality than a “yes” response, the “Surprise Question” was only modestly predictive of 12-month mortality. Future work will focus on determining if there are patient populations for whom the “Surprise Question” is more predictive and assessing the ability of the “Surprise Question” to predict other clinical outcomes, such as ED visits and hospitalizations.