Utility of the “Surprise” Question to Identify Dialysis Patients with High Mortality

Materials and Methods

Results

Of 150 consecutive hemodialysis patients in three dialysis units, 147 (98%) agreed to participate in the study. Eighty-one (55%) patients were male; 88% were white and 12% were black. The cause of ESRD was diabetes for 40% of the patients, hypertension for 19%, glomerular disease for 6%, renal artery stenosis for 6%, polycystic kidney disease for 5%, interstitial nephritis for 3%, and unknown or the other for 21%. For 34 (23%) of the patients, the nurse practitioners said that they would not be surprised if the patient died within the next year. Compared with the “yes” patient group, the “no” patient group was older and had lower mean serum albumin and Karnofsky Performance Status scores and a higher mean CCI score (Table 1).

At 12 mo, 22 of the 147 patients had died, for an overall mortality of 15%. The mortality rate for the “no” group was 10 (29.4%) of 34 patients and for the “yes” group was 12 (10.6%) of 113 (P = 0.032).

In univariate logistic regression analysis, the “surprise” question response, the CCI score, and the Karnofsky Performance Status score were significantly associated with patient status at 12 mo (Table 2). For those for whom the “surprise” question was answered “no,” the odds of dying within 1 yr were 3.507 times the odds of dying for those for whom the “surprise” question was answered “yes.” For each 1-point increase in the CCI score, the odds of dying were 1.417 times higher than for those whose score was 1 less. In the multivariate analysis, only the CCI score was significantly associated with patient status at 12 mo (Table 3).

In the survival analyses, there was a significant difference in mean days alive at 12 mo between the “no” and “yes” response groups for the “surprise” question and the higher and lower groups for the CCI score (Figure 1). There was no difference in mean days alive at 12 mo for patients in the lower and higher Karnofsky Performance Status score groups (351.0 ± 33.6 versus 354.3 ± 32.0; P = 0.473).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Survival curves for “surprise” question response and comorbidity score in days alive at 12 mo. Data are means ± SE. (A) Curves of “yes” and “no” response groups to the “surprise” question, “Would I be surprised if this patient died in the next year?” (B) Curves of the lower (<8) and higher (≥8) Charlson Comorbidity Index (CCI) score groups.

Discussion

The dialysis patient population is among the sickest of patient populations. Dialysis patients have a high number of comorbid illnesses and a symptom burden comparable to that of cancer patients (18,19). They have a known shortened life expectancy, an annual mortality rate approaching 25%, and a high in-hospital mortality rate (2,3). Dialysis patients have been recognized as a patient population for whom palliative care is appropriate (1). There would be great utility in a simple screening assessment tool that would identify a subset of dialysis patients who are at the highest risk for early death. This study was conducted to assess whether the “surprise” question would be such a tool that would identify the sickest dialysis patients, those who unquestionably should receive priority for palliative care.

The intent of the “surprise” question is to counter the tendency of physicians to overestimate prognosis (11). The “surprise” question prompts clinicians to think about their patients in a new way. Instead of asking clinicians whether the patient will be dead in 1 yr, it asks them to consider whether they think that the patient's death in 1 yr is enough within the realm of possibility that it would not surprise them. In this respect, the “surprise” question is unlike other predictors of prognosis, such as the Acute Physiology and Chronic Health Evaluation II (APACHE II) system in which a derived score is associated with a particular likelihood of 1-yr survival (20). The “surprise” question has been found to be effective in helping physicians identify patients in a primary care population who are terminally ill and for whom palliative care referral is appropriate (11). It has not been previously tested in a rigorous manner in chronic disease populations (21), and, to our knowledge, this is the first report of the use of the “surprise” question in dialysis patients.

In this study of dialysis patients, use of the “surprise” question was effective in identifying a subgroup of dialysis patients who were significantly sicker than the majority of dialysis patients who were being treated in the three dialysis units. At baseline, compared with the “yes” group, the “no” group was older, had more comorbid conditions, had a lower functional status, and had a worse nutritional status as reflected by a lower serum albumin level (Table 1). The odds of dying within 1 yr for the patients in the “no” group were 3.5 times higher than for patients in the “yes” group. In the univariate logistic regression analysis, the “surprise” question along with functional status and comorbidity score were the only three variables that were significantly associated with prediction of patient death at 12 mo. In a previous systematic literature review of dialysis patient studies, functional status and comorbidity along with age and nutritional status were the only variables that were independently associated with poor prognosis (8).

An answer to the “surprise” question is simpler to obtain than calculation of the CCI, which requires a chart review. Asking the “surprise” question is something that nephrologists and other nephrology clinicians could institute monthly on rounds to screen patients and identify those to refer for immediate palliative care consultation. Palliative care consultants are experts in pain and symptom management and advance care planning (7). Pain is known to be undertreated in dialysis patients (22,23), and in their training, most nephrologists are not prepared to manage pain in their patients (24). Because of the debilitated health state of patients in the “no” response group, advance care planning is particularly important to help these patients articulate and prioritize their goals of care in the context of their health state and to express their preferences regarding a proxy medical decision-maker, cardiopulmonary resuscitation, intubation and mechanical ventilation, tube feeding, circumstances in which they would want to stop dialysis, and dying at home with hospice care (25).

The “surprise” question identifies patients who are at highest risk for death, but because of their shortened life expectancy and high symptom burden, most dialysis patients are good candidates for palliative care. In absolute numbers, more patients in this study in the “yes” group died than in the “no” group. Nonetheless, the “surprise” question performed quite well because our dialysis population had an annual gross mortality of only 15%, significantly less than the average gross mortality of 24% reported for the ESRD dialysis population in the US Renal Data System (2). Despite this low mortality, it classified close to one quarter of patients in the “no” group. By all parameters known to predict mortality in ESRD, these patients in the “no” group were sicker than the patients in the “yes” group.

This study has several limitations. First, it was conducted in a dialysis patient population that is not as ethnically diverse as the US ESRD population. Only 12% of the patients in this study were black compared with 37% in the December 31, 2005, point prevalent dialysis population in the United States (2). Second, the study was conducted only of hemodialysis patients in three dialysis units in the same geographic region. Third, the study was conducted on the basis of the responses to the “surprise” question of a small number of clinicians: Three nurse practitioners. Although the “surprise” question has been shown to be effective in helping primary care clinic–based physicians identify patients who are appropriate for palliative care (11), it is possible that the accuracy of the clinicians in our study might not be matched by others. It is also possible that physicians may be more accurate in classifying patients according to the “surprise” question than nurse practitioners. In favor of the approach used in our study is the finding of the efficacy and simplicity of the use of another single screening question—“Are you depressed?”—to identify patients who are appropriate for palliative care intervention in another patient population (26). This study also offers a simple way to distinguish better those patients whose death is not unexpected from those who “are not supposed to die,” a major challenge to our health care system that is struggling with providing and measuring quality care at the end of life (27).

Conclusions

The “surprise” question worked quite well to identify a subset of hemodialysis patients with a high risk for early death in the next year. It identified those with older age, more comorbid illnesses, lower functional status, and lower serum albumin levels, factors that all have been independently determined to be predictors of mortality in dialysis patients (8). This study validates the use of the “surprise” question by nephrology nurse practitioners. Further research is needed to determine whether (1) the “surprise” question will work equally well or better in the hands of nephrologists as compared with nephrology nurse practitioners, (2) the “surprise” question will work in more ethnically diverse dialysis patient populations, and (3) use of the “surprise” question will lead to interventions that improve the quality of end-of-life care for dialysis patients.

Disclosures

None.