Mortality in Elderly Gen Surg Patients With 'DNR' Orders
Our analysis of more than 25,000 patients in the American College of Surgeons National Surgical Quality Improvement Program database shows that elderly patients who undergo emergency general surgical procedures suffer very high mortality and morbidity and that they are more likely to die within 30 days of the operation if they carried DNR orders preoperatively. Although the alternative for many of these patients should they have refused operation would have been likely death, the findings of our study nevertheless serve to underscore the ominous outcomes associated with emergency general surgical intervention in the elderly population. To our knowledge, this is the most detailed description of early postoperative outcomes among elderly DNR patients requiring emergency general surgery to be published. Our specification of diagnosis-specific postoperative mortality and morbidity rates will enable general and acute care surgeons to use objective data rather than anecdotal observation when advising elderly DNR patients about the anticipated risks of emergency operation. Although a better understanding of these outcomes may not alter a patient's decision to undergo operation, it will nevertheless provide surgeons with greater confidence in their ability to provide unbiased counsel to patients and/or their health care proxies. Alternatively, receiving a more objective presentation of potential postoperative outcomes may dissuade some patients from pursuing emergency operation, depending on their particular goals of care. Either way, studies drawn from the oncology literature clearly demonstrate that the quality of prognostic information can have a significant effect on patient treatment decisions. Therefore, any resource that adds to the ability of surgeons to predict postoperative outcomes in the elderly population must necessarily be viewed as useful.
In addition to its practical utility, the findings of our study may also help to elucidate potential causes for the independent association between preoperative DNR status and postoperative mortality that we and others have demonstrated. Specifically, we believe that the major contributing factor for the higher mortality among DNR patients in our propensity-matched cohort was their greater likelihood (compared with non-DNR patients) of succumbing to major postoperative complications. Although "failure to rescue" is the traditional term that is used to describe death in the setting of postoperative complications, such a moniker is potentially misleading when used to describe mortality among patients who undergo emergency operation. "Failure to rescue" implies that patient death due to postoperative complications has occurred despite every and all attempts to prevent such death. A close examination of the findings of our analysis suggests "failure-to-pursue rescue" as a more accurate descriptor of the excess mortality suffered by elderly DNR patients who experience major postoperative complications, as this term better reflects the possible disinclination among such patients to accept aggressive management of these complications.
The results of our analysis do not suggest that increased presence and/or severity of comorbid illnesses are responsible for the greater postoperative mortality rates experienced by DNR patients. Although DNR patients in our overall study sample did seem to be more acutely and chronically ill than non-DNR patients, our use of propensity matching seemed to adequately adjust for this baseline difference in health status. Specifically, the non-DNR patients who were included in our smaller, propensity-matched cohort were uniformly well matched to DNR patients with respect to incidence and severity (when severity is known) of all of the preoperative variables included in our analysis. In addition, preoperative DNR status did not seem to influence the incidence of major postoperative complications in our matched cohort (40.2% for non-DNR patients vs 42.1% for DNR patients, P = 0.38). We would have expected to find a higher rate of major postoperative complications in DNR patients if they were in some way "sicker" than the non-DNR patients who were included in our matched cohort. In the absence of such a finding, we cannot conclude that disparate degrees of comorbid illness explain the discrepancy in surgical mortality that we describe.
Similarly, our findings do not support the existence of an overt bias among physicians against aggressive preoperative or intraoperative management of elderly DNR patients. Our inclusion of markers of aggressiveness of preoperative/intraoperative care in our propensity-matching algorithm theoretically adjusts for such differences. Even before this adjustment, however, a comparison of the unmatched study sample suggests that DNR patients were managed just as aggressively as non-DNR patients in the preoperative period. For example, DNR patients were as likely or more likely to receive packed red blood cell transfusion and/or mechanical ventilation before operation. Similarly, a review of diagnosis-specific operative times and total work relative value units suggests that the operations performed in DNR patients were just as complex as those performed in non-DNR patients. Taken together, these findings argue against less aggressive preoperative or intraoperative care as a reason for the higher postoperative mortality experienced by elderly DNR patients.
The reason for increased mortality among DNR patients that is best supported by the findings of our study is that such patients are less likely than non-DNR patients to receive aggressive therapy for major postoperative complications. Approximately 57% of DNR patients from our matched cohort died after developing a major postoperative complication compared with 41% of non-DNR patients, despite the fact that the 2 groups had no detectable difference in their physiological ability to withstand such complications. Further evidence of failure-to-pursue rescue as the primary reason for the association between preoperative DNR status and postoperative mortality is the finding that DNR patients from the matched cohort were significantly less likely than non-DNR patients to undergo reoperation in the postoperative period. Although the American College of Surgeons National Surgical Quality Improvement Program does not provide information on the indication for reoperation, we have no reason to expect this indication to differ between the 2 groups of our matched cohort. Therefore, we believe that the lower reoperation rate among DNR patients from our matched cohort reflects the fact that they are less likely to consent to such intervention when it is indicated.
An elderly patient's decision to undergo emergency operation is time sensitive and often made in the setting of severe physical discomfort. Our findings suggest that although many such patients will consent to emergency surgery, they will be more likely to decline aggressive medical intervention in the postoperative period if they had established DNR directives in place before the procedure. Although such behavior may seem paradoxical (accepting maximally invasive treatment in the form of emergency operation but declining less invasive treatment in the form of management of major complications), it seems more logical when viewed within the framework of "patient's goals of care." Patients may be willing to undergo emergency operation for a life-threatening disease process knowing that they will be provided general anesthesia during the operation and that there is a reasonable chance that surgery will immediately improve their pain and definitively treat the cause of that pain. Upon further reflection in the postoperative period (especially in the setting of a major complication), they may discover that the procedure has left them more debilitated or that the postoperative discomfort is worse than they had hoped. As a result, their willingness to undergo continued aggressive management becomes more closely aligned with the reality of their postoperative course. Support for the impact of complications of medical care on patient preference for such care comes from a multicenter study by Nathens et al of predictors of patient DNR status after severe traumatic injury. These authors found that the development of postinjury complications that resulted in end-organ dysfunction was independently associated with a patient being designated as DNR during his or her postinjury hospitalization.
Our analysis does indicate certain specific postoperative complications for which the discrepancy in subsequent mortality between non-DNR and DNR patients is relatively large. For example, we found DNR patients to be significantly more likely than non-DNR to die in the setting of postoperative renal insufficiency, myocardial infarction, organ/space surgical site infection, and pneumonia. If these discrepancies in complication-specific mortality are in fact due to a higher rate of failure-to-pursue rescue among DNR patients, then such patients may benefit from knowing that there is a possibility of improved survival should they accept aggressive management of these specific complications. Whether this improved understanding will alter the decision to forego such management will ultimately depend on a variety of factors, including individual goals of care and the perceived invasiveness of the management needed to treat the complication.
Our analysis has several important limitations. First, we lack the necessary patient-level information to conclusively determine context in which patient deaths occurred Prospective survey analysis will ultimately be required to confirm the extent to which failure-to-pursue rescue explains the excess mortality that elderly DNR patients experience after emergency general surgery. Second, the contribution that attending surgeon input has on a patient's decision to reject aggressive management of postoperative complications also cannot be assessed using information from data sources such as the American College of Surgeons National Surgical Quality Improvement Program. Third, the results of our analysis do not necessarily extend to nonelderly patients or to patients who undergo elective surgical intervention. Fourth, because the American College of Surgeons National Surgical Quality Improvement Program includes only patients who underwent an operation, we do not know the outcomes of elderly DNR patients who present with acute surgical disease but who do not receive an operation. It may be that less invasive interventions (such as percutaneous cholecystostomy drain placement or intravenous antibiotics) may adequately alleviate the symptoms of such patients while enabling their short-term survival. However, a comparison of outcomes for the full spectrum of treatment options (operation intervention vs nonoperative intervention vs comfort care) is not possible using the American College of Surgeons National Surgical Quality Improvement Program.
Despite these limitations, our study demonstrates that emergency general surgery is associated with significant morbidity and mortality in elderly patients. Furthermore, our findings suggest that one reason for the excess mortality experienced by that subgroup of elderly patients who have preoperative DNR orders is their failure-to-pursue rescue when major postoperative complications occur. Although confirmation of this finding will require prospective survey analysis, it is nevertheless reasonable to expect that the results of our study will enable general surgeons to provide more accurate and therefore more useful prognostic information to elderly patients who develop emergency general surgical conditions.