Answer to common clinical question may predict risk for death

Prognostication is one of the most difficult services we provide in health care. Even as a clinician with training and experience in palliative medicine, I find myself often having an intense debate within my own mind when prognosticating for a patient. Sometimes the discrete data points that help differentiate a patient with months vs. years of time are either not known, fluid in their value and evolving in their meaning. Take, for example, brain metastases in advanced breast cancer. A one-size-fits-most approach would generally tell us that the presence of cancer spread to the brain portends a prognosis measured in months. That’s true, unless, of course, you consider the exceptions. Off the top of my head — pun very much intended — I can think of HER-2/neu–overexpressed breast cancers and oligometastatic disease amenable to radiosurgery. Even 5 to 10 years ago, often when many of us trained, we did not recognize as well the different tumor subtypes or new treatment technologies to address complications that were once thought of as drivers of a short life expectancy.
But aside from seeking these discrete data points, can a Gestalt approach provide us the information we need to have patient-centered and truthful conversations with our patients? Many of us in the palliative care community are familiar with “The Surprise Question,” which is both elegant in its simplicity and practical in its execution. It involves looking at the whole person — their life-limiting primary and secondary illness, chronic comorbidities, functional status, behavior history, compliance, and other factors both in the moment and trended along the longitudinal nature of the clinical relationship — and asking ourselves, “Would we be surprised if this patient was not alive in 1 year?” Can this approach be used as an effective screening tool, both to direct patients to the type of care they want when it is recognized that time is short, and to evaluate for enrollment in clinical studies that are focused on those with short life expectancies?
Vick and colleagues recently reported results of a project to evaluate the sensitivity — the proportion of “yes” responses to the surprise question who did not die in the next 12 months or the proportion of a “no” response who did die — and specificity, or proportion who answered “no” to the surprise question who did die in the next 12 months. The setting was a large, tertiary/quaternary care cancer institute in a major metropolitan area. Among 81 oncology clinicians across 4,617 patients, the investigators found high utility for when the answer to the surprise question was “yes.” Researchers concluded that a positive response to the surprise question better predicted death than cancer type, patient age, cancer stage or time since diagnosis. Conversely, with the sensitivity of a “no” response only at 59%, the researchers concluded that about 40% of patients who would die within the next year were not captured by the surprise question.
These findings have several implications. First, they tell us that if your gut says that time is short, it probably is. The results also highlight the limitations of this approach as the end-all-be-all to prognostication in cancer. If your guts says that there is a lot of time left, then you might need a gut check, maybe about 40% of the time. Second, the utility of the surprise question as a screening question for enrollment into clinical studies seems to be reaffirmed. Because of the high specificity (90%), we have some reassurance that those who undergo interventions within studies focused on those near the end of life are likely to truly be in that time where “time is precious.” Third, we have more work to do. The generalizability of these findings remains unknown and warrants further research. Do these results apply to community-based settings during more primary or secondary cancer care? Is the Gestalt approach of clinicians across other institutions as finely tuned as the participants in this study?
These findings are an important step toward a greater understanding of how to target the right supportive care services. Ultimately, time is a discrete data point that has no value unless a patient-centered action is taken from the information. My hope is that evidence like this and other future studies will provide clinicians the tools and confidence to provide palliative and supportive care when it is needed the most, and certainly before it is too late.