Discussion
In this multicenter study of patients 80 years old or older who were admitted to 22 ICUs across Canada, the most striking finding was the prolonged stay in the ICU, and the life support modalities employed before death and on the day of death. Overall, 35% of these very elderly ICU patients died in hospital, and mortality rates were much higher in frail compared with nonfrail patients. For those who died in hospital, this occurred a median of 10 days after ICU admission for the entire cohort. In the nested cohort, the average time from admission to death was 16 days with the dying experience being significantly prolonged if family members were "unsure" of their preferences for care. In nonsurvivors, previously documented advance directives and prior frailty had minimal to no impact on limiting the use of life-sustaining treatment or shortening the time from ICU admission to death. One quarter of family members of these very elderly patients preferred comfort measures, yet almost all of them received life-sustaining treatments in the ICU and time from ICU admission to death was 12 days. We did not identify predictors for prolonged period of life support before death in this population.
Our findings contrast starkly with data from other countries where the average time from ICU admission to death ranges from 1 to 2 days, and is shorter in patients 80 years old or older compared with younger patients. Our findings raise questions about the process of EOL care for very elderly patients admitted to the ICU in Canada. In a prior Canadian study, elderly patients reported that avoiding unnecessary prolongation of life through the use of technology was among the most important aspects of EOL care. Our findings challenge whether this "right to quality EOL care" is being realized for many very elderly patients and their families. Furthermore, this kind of high-intensity care provided at the EOL is associated with reduced quality of life in the patients' remaining days, and increased risk of poor health outcomes for surviving family members. The fact that a proportion of families were expressing a preference for comfort measures only for a proportion of these patients experiencing an "intensified death" further illustrates our concerns with quality care at the EOL.
There are economic implications of our findings. The cost of providing prolonged and nonbeneficial care to ICU patients is considerable. Furthermore, there is an opportunity cost in that ICU beds occupied by patients who receive nonbeneficial treatment are unavailable for other patients more likely to benefit from ICU admission. Delayed access to critical care for seriously ill patients has been associated with increased patient morbidity and mortality.
Given the interest in EOL care among the public, professionals, and politicians, it is imperative that care of the very elderly who have life-threatening illnesses be improved. First, we need to be sure that admission to the ICU and life-sustaining treatment is congruent with patient preferences; this should be determined in advance of critical illness. In another Canadian study of patient preferences among very elderly patients who were admitted to non-ICU hospital wards, fewer than 12% preferred full medical care including mechanical ventilation and cardiopulmonary resuscitation.
Furthermore, we found that advanced directives, as currently defined and implemented, did not appear to have an impact on limiting the overexposure to life-sustaining technologies at the EOL. The documents may contain expressed wishes for the use of life-sustaining treatments but this would be inconsistent with prior published studies of cohorts of very elderly patients expressing a predominant desire for comfort measures only. Or, this may reflect the inaccessibility, lack of awareness, or lack of clarity of such documents, or their ineffectiveness at influencing treatments. We posit that advance care planning, which includes reflections, values clarifications and conversations with others that prepare the patient and family for "in the moment decision-making," are more likely to be clinically useful than instructional directives. Evidence for this assertion comes from a randomized trial of very elderly hospitalized patients indicating that advance care planning discussions with adequate documentation of their wishes result in enhanced quality EOL care, greater family satisfaction, and fewer unwanted ICU admissions. Given that patients of families that were "unsure" of their treatment preferences had the longest dying experience (median of 24 d), a process that helps families clarify values early in the course of stay has potential for both improving quality EOL care and significantly reduce wasted healthcare resources.
Ideally, communication interventions, conversations, and decisions should occur in advance of a life-threatening illness and before admission to the ICU. However, very elderly hospitalized patients report that doctors rarely ask them about their prior wishes or treatment preferences. As a consequence, we found that the expressed patient preference agreed with the medical order on the chart only a third of the time. One approach to this problem would be for healthcare providers to elicit treatment preferences and support a shared decision-making process before ICU admission. Accordingly, we have initiated the "Just Ask" campaign, to encourage healthcare providers to probe all "at-risk" patients admitted to hospital about their prior wishes, named decision-makers, and current treatment preferences; this is followed by provision of tools to enable and guide such conversations.
At the same time as we encourage healthcare professionals to engage with elderly patients in these conversations, we need better decision-making tools to help clinicians identify nonbeneficial treatment earlier in the ICU stay. Frailty is associated with increased morbidity and mortality in the short and long term. Systematic measurement of patient frailty (and other key determinants to long-term outcomes), as well as the development and dissemination of validated clinical prediction rules may help to better identify elderly patients who are very unlikely to benefit from ICU admission, or when admitted, will be unlikely to benefit from prolonged critical care (identified earlier in the ICU stay).
Strengths of our study include the multicenter design, national engagement, and large sample size which increase the representativeness and utility of our findings. However, most of the patients in our sample were Caucasian, and had to have a family member who spoke English or French, which may limit the generalizability of these findings. The presence of an unselected hospital cohort in conjunction with a nested cohort where we were able to obtain better characterization of patients at baseline is another strength.
There are several limitations. Our data describe practices in Canada and consequently, our findings may not be generalizable to other healthcare systems. We have no control group of either younger patients or elderly patients who were not admitted to ICU, for comparative analyses. When comparing treatment limitations between those with and without advance directives, we do not have the detailed information on the content of the directive to know whether treatments were concordant or discordant with the requests of patients. Paradoxically, there were fewer treatment limitations made by physicians and longer dying periods in those patients who had preexisting advance directives compared with those who did not. We further note that we enrolled a family member who was not necessarily the legally appointed substitute decision-maker, to obtain an understanding of the patient's baseline characteristics and preferences for care. In soliciting these preferences, we provided standard definitions or statements describing different goals of care. However, we acknowledge that lay people may not understand the true meaning of life supports or use of mechanical ventilation. This may explain why so many family members preferred "comfort care" and yet their loved one had a prolonged stay in the ICU. Finally, we acknowledge that our definition of "prolonged dying" (> 7 d in ICU) is somewhat arbitrary.
In summary, we report the use of ICU treatments and outcomes of care of patients 80 years old or older who were admitted to 24 participating ICUs in Canada. We have observed that it is common for elderly patients to die in hospital, often after a prolonged ICU stay and while still receiving life sustaining-technologies. Our findings question whether hospitalized very elderly patients are achieving a "quality finish." We are not advocating that we triage potential ICU patients based on age; but rather, these results serve as a call to action to improve communication and decision-making in this high-risk population and thereby improve EOL care for our very elderly patients.