HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Respond to This Article Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Cook, D. Search for Related Content PubMed PubMed Citation Articles by Cook, D.

Levels of Care in the Intensive Care Unit: A Research Program

	Abstract

Top Abstract Methods Cardiopulmonary Resuscitation... Provision of Advanced Life... Clinical Prediction of ICU... Clinicians' Discomfort With Life... Discussion References

 
A multidisciplinary research program on levels of care was conducted in 15 adult intensive care units in North America, Europe, and Australia. The program addressed advance directives for cardiopulmonary resuscitation, provision of advanced life support, and clinicians’ discomfort with evolving treatment plans. The results indicated that the factors that determined the establishment of directives for advance life support differed from the factors that informed a decision to limit or withdraw support after admission to an intensive care unit. In addition, clinicians’ prognoses were imprecise and often an underestimation of the probability of short-term survival. Finally, some degree of discomfort was common in care providers in the intensive care unit, most often because they thought interventions were excessive and not compatible with an acceptable future quality of life. The provision of advanced life support mandates explicit decision making about how life-support measures should be used.

We therefore formulated research questions to improve our understanding of the use of life-support technology in relation to targeted pathophysiological changes. By focusing on key questions about the provision of life-prolonging interventions at the time of ICU admission and during the clinical evolution of patients’ illnesses and by evaluating the changing perspectives of different stakeholders in the process, we gained insight into contemporary practice in a number of important domains. In this article, we describe the program as an integrated whole designed to understand these issues and the complex interactions between the issues.

Respect for patients’ autonomy suggests that life support interventions are most appropriate when consistent with patients’ values, rather than targeted to specific organ dysfunction.

 

	Methods

Top Abstract Methods Cardiopulmonary Resuscitation... Provision of Advanced Life... Clinical Prediction of ICU... Clinicians' Discomfort With Life... Discussion References

 
The Levels of Care Research Program involved bedside ICU nurses, residents, and attending physicians in 15 ICUs in 4 countries (Canada, United States, Australia, and Sweden) in a research program that modeled the multidisciplinary delivery of critical care. We enrolled consecutive patients 18 years or older who received mechanical ventilation within 24 hours of admission to an ICU. The Figure gives the characteristics of the patients we studied to achieve the 6 objectives in the research program. The research ethics boards of all institutions involved in the study approved the project. We were not required to obtain informed consent because we made no attempt to influence patients’ care.

Our fields of inquiry were directives about cardiopulmonary resuscitation, advanced life support, clinical prediction of ICU mortality, and clinicians’ discomfort with life support. In the following sections, we report the key objective for each field, the specific methods used to achieve each objective, the results we obtained, and our interpretation of the results.

	Cardiopulmonary Resuscitation Directives in the ICU

Top Abstract Methods Cardiopulmonary Resuscitation... Provision of Advanced Life... Clinical Prediction of ICU... Clinicians' Discomfort With Life... Discussion References

 
Advance directives, particularly, do-not-resuscitate (DNR) directives, are thought to promote self-determination and result in decisions about life support that are consistent with patients’ values. Many seriously ill patients lose their decision-making ability. Furthermore, wishes of patients’ surrogates and clinicians correlate only modestly with the patients’ wishes, and barriers to advance care planning exist in healthcare systems. For patients requiring ICU admission, decisions about resuscitation in the event of a cardiopulmonary arrest are particularly important. When these directives are explicit, they formalize a plan for resuscitation or no resuscitation. When no explicit directive is established, the default directive is to perform cardiopulmonary resuscitation (CPR) whether or not this intervention is consistent with patients’ values. We therefore sought to define the frequency with which explicit directives are available at the time of ICU admission and the factors that influence the establishment of such directives both at the time of ICU admission and during the course of critical illness.

Surrogates’ and clinicians’ life support wishes correlate only modestly with patients’ wishes.

 

Objective 1: To Determine the Prevalence, Predictors, and Procurement Pattern of CPR Directives Within 24 Hours of Admission to the ICU
  Specific Methods.   For objective 1,¹ we documented age, sex, score on the Acute Physiology and Chronic Health Evaluation (APACHE) II,² multiple organ dysfunction score (MODS),³ admitting diagnosis, and other baseline factors in 2916 patients (see Figure). During the first 24 hours after admission to the ICU, we determined whether there was an explicit directive to resuscitate or not resuscitate or no explicit directive. These directives were established per usual clinical practice, and their existence was determined by research personnel during the first day of ICU admission. For patients with explicit directives, we ascertained when the directives were established and what factors determined that establishment. Using polychotomous logistic regression, we identified factors associated with an explicit resuscitation directive versus no resuscitation directive.

View larger version (26K): [in this window] [in a new window]   Flow of patients in the Level of Care Research Program contributing to the 6 objectives. Abbreviations: CPR, cardiopulmonary resuscitation; ICU, intensive care unit; LOS, length of stay; MV, mechanical ventilation.

Within 24 hours of ICU admission, only 11% of patients had a resuscitation directive, and ICU residents had established 46% of these.

 

Increases in patients’ age were strongly predictive of DNR directives; odds ratios were 3.4 for patients 50 to 64 years old, 4.4 for patients 65 to 74 years old, and 8.8 for patients more than 75 years old. In contrast, increasing acute severity of illness as indicated by the APACHE II scores at the time of ICU admission was predictive of both resuscitate and DNR directives. Patients admitted on the weekend or at night were more likely to have an explicit directive than were patients admitted during the week or in the daytime, whether the directive was to resuscitate or not resuscitate. Inability to participate in decisions increased the likelihood of a DNR directive (odds ratio 3.7) more than the likelihood of a resuscitate directive (odds ratio 1.7; P = .001 for the difference). Patterns of obtaining and documenting CPR preferences differed between countries, between cities within countries, and between ICUs within cities; that is, we found a strong center effect.

  Interpretation.   Only 11% of critically ill patients had CPR directives established within 24 hours of ICU admission. In Canada, for patients without explicit CPR directives, CPR is the current default decision if a cardiopulmonary arrest occurs. However, half of the patients in this study for whom directives were known expressed a wish not to be resuscitated, suggesting that many patients who undergo CPR by default are being subjected to an intervention that they would not explicitly choose.

The finding that CPR choices were influenced by the time of admission suggests that factors such as availability of beds in the ICU and the need to ration beds also determine the outcome of discussions about resuscitation. During evenings and weekends, more family members of patients and more ICU residents, but relatively fewer ICU physicians, may be in the hospital; these factors may influence the dialogue with residents about resuscitation. The significant center effect also suggests variable interest among centers in establishing directives soon after ICU admission. When addressing a patient’s preference for CPR, clinicians are also influenced by their level of training and experience, the characteristics of the patient, and the setting in which the clinicians work. To meet the challenge of eliciting valid CPR preferences, ICU teams should work toward wider adoption of culturally appropriate, locally adapted, and effectively implemented guidelines for discussions with patients about CPR.

Admission timing, that is, weekend or weekday ICU admission, influenced CPR choices.

 

Objective 2: To Determine the Effect of Patients’ Previous Functional and Employment Status on CPR Directives Established Within 24 Hours of Admission to the ICU
  Specific Methods.   For objective 2,⁴ we focused on 1008 patients (see Figure) in the first 24 hours after admission to the ICU. Each patient’s attending ICU physician rated the patient’s global functional status 1 month before admission as severely limited, somewhat limited, totally independent, or unknown on the basis of the patient’s history and physical examination and discussion with the patient’s family, hospital colleagues, family physicians, or nursing home personnel. These data were collected independently by clinicians blinded to each other’s responses. Research nurses also documented each patient’s employment status 1 month before admission; patients were categorized as unemployed, not working outside the home, retired, employed, or of unknown employment status. Using polychotomous logistic regression, we examined the relation between CPR plans and functional and employment status as perceived by the ICU team.

  Results.   A total of 10% of the patients were judged to be severely functionally limited 1 month before ICU admission, 22% were judged somewhat limited, 62% were judged totally independent, and 6% had unknown functional status. Severe impairment in functional status was moderately associated with an explicit plan to resuscitate (odds ratio 2.2 relative to no explicit directive) and strongly associated with an explicit DNR plan (odds ratio 6.2 relative to no explicit directive; P = .01 for the difference). This relationship was not influenced by age, sex, APACHE II score, surgical status, employment status, or city. Severely limited functional status was strongly associated with an explicit DNR directive for patients who could not participate in decisions (odds ratio 8.2) and more weakly associated for patients who could participate (odds ratio 1.7). We also found that patients who were unemployed were more likely to have an explicit resuscitation directive rather than no explicit directive (odds ratio 5.5).

Patients who were unemployed were more likely to have an explicit resuscitation directive rather than no explicit directive.

 

  Interpretation.   As perceived by the ICU team, impaired functional status before admission to the ICU is strongly associated with DNR directives in patients unable to participate in decision making; for patients able to participate in decision making, the association is weaker. Possibly, clinicians and substitute decision-makers (who might underestimate patients’ psychological adjustment to functional impairment) are more inclined than a patient to decide that the patient would want to forgo CPR. However, our observations also raise the concern that patients might have made different decisions if they had been able to participate in discussions. Moreover, our results also suggest that clinicians may be more inclined to clarify CPR directives for patients who are unemployed or work only at home than for patients who are employed outside the home.

These observations emphasize the challenges of ensuring that CPR decisions are not adversely affected by patients’ inability to participate or by physicians’ conscious or subconscious bias about patients’ functional and employment status. The influences of employment status might be reduced by improved training of clinicians. In the absence of valid advance healthcare directives completed by well-informed patients, clinicians should emphasize in discussions with a patient’s family that the patient’s wishes should determine the goals of treatment. The decisions should be those that the patient would make for himself or herself if able to do so.

Objective 3: To Determine the Rate and Determinants of DNR Directives Established During an ICU Stay for Patients Receiving Mechanical Ventilation
  Specific Methods.   For objective 3,⁵ we focused on 765 patients (see Figure) who had an explicit DNR directive established more than 24 hours after admission to the ICU. We determined the rate with which DNR directives were established and the factors strongly associated with these directives. We collected baseline data, including sex, age, baseline APACHE II score, and admission diagnosis. Each day, we calculated the MODS and recorded whether life-support treatment (mechanical ventilation, use of inotropic agents, dialysis) was administered or withheld or withdrawn, the patient’s ability to participate in decision making, and whether or not a DNR directive was established. In addition, we asked each patient’s attending physician and bedside nurse to independently estimate the patient’s previous function, the probability of the patient’s survival during the stay in the ICU and the stay in the hospital, and probable function and cognitive status 1 month after discharge from the hospital.

  Results.   DNR directives were established for 231 patients (30%); 62% of the directives were established between days 2 and 10 after admission to the ICU. Of these 231 patients, 90% were unable to participate in decisions. Compared with patients without DNR directives, patients with DNR directives were more likely to have withdrawal of mechanical ventilation (48% vs 5%, P < .001) and to die in the ICU (74% vs 15%, P < .001).

Factors independently associated with establishment of a DNR directive after ICU admission were age (hazard ratios were 1.4 for patients 50–64 years old, 1.8 for those 65–74 years old, and 2.3 for those older than 75 years), medical versus surgical diagnosis (hazard ratio 1.8), MODS (hazard ratio 1.7 for each 5-point increment), physician’s prediction of ICU survival (hazard ratios were 15.0 for <10% chance of survival, 5.0 for 10%–40% chance, and 4.0 for 41%–60% relative to >90% chance), and physician’s perception that the patient preferred to limit life support (hazard ratios were 5.8 for no advanced life support and 3.2 for partial advanced life support compared with full life-support measures). Compared with patients who did not have any withdrawal of life support, patients who also had at least one life-support treatment withdrawn were more likely to die receiving mechanical ventilation and to have shorter ICU and hospital stays.

  Interpretation.   In contrast to a decade ago, today DNR directives are established earlier, rather than just before withholding or withdrawal of life support.⁶ Age was significantly associated with DNR decisions established during the ICU stay for patients more than 50 years old. Although in previous studies¹ age was more strongly associated with DNR decisions made within 24 hours of ICU admission, age was not an independent determinant of the decision to withdraw mechanical ventilation.⁷ Thus, patients’ age appears to influence resuscitation decisions in the event of a cardiopulmonary arrest but has less influence on the more complex and consequential decisions to limit life support. We also found that physicians’ impressions of patients’ wishes influenced the likelihood of a DNR order.

	Provision of Advanced Life Support in the ICU

Top Abstract Methods Cardiopulmonary Resuscitation... Provision of Advanced Life... Clinical Prediction of ICU... Clinicians' Discomfort With Life... Discussion References

 
Mechanical ventilation is the main form of advanced life support in the ICU. Although most critically ill patients are successfully weaned from mechanical ventilation, other patients die despite ongoing mechanical ventilation or after the withdrawal of ventilatory support in anticipation of death. Of the life-support treatments, mechanical ventilation is stopped most often.

Other studies^8,9 have suggested that factors such as a patient’s age, illness severity, wishes, and past and future quality of life influence decisions to withhold and withdraw treatment. Because nurses and physicians often have different perspectives on end-of-life care,^10–12 we elicited perspectives from several ICU team members.

Objective 4: To Determine the Influence of Baseline and Time-Dependent Factors on Withdrawal of and Death During Mechanical Ventilation
  Specific Methods.   For objective 4,⁷ we focused on 851 patients (see Figure). We recorded baseline characteristics, MODS, decision-making ability, life support, DNR orders, clinicians’ predictions of survival and future functional and cognitive status, and clinicians’ perceptions of patients’ preferences. Daily, we determined whether patients were weaned from mechanical ventilation, died while receiving mechanical ventilation, or had mechanical ventilation withdrawn. To compare the determinants of withdrawal of ventilatory support with the determinants of death during mechanical ventilation, we used Cox proportional hazards regression analysis.

  Results.   A total of 63% of the patients were successfully weaned, 17% died while receiving mechanical ventilation, and 20% died after ventilatory support was withdrawn. Of the 166 patients who had ventilatory support withdrawn, 87% died in the ICU and an additional 9% died in the hospital, for an overall hospital mortality of 96%. Among the 851 patients in the sample, 66% of the patients who died in the ICU died after withdrawal of one or more life-support treatments (mechanical ventilation, use of inotropic agents, or dialysis).

Patients who had ventilatory support withdrawn were less likely than those who died while receiving mechanical ventilation to receive inotropic agents (69% vs 90%, P < .001), and once receiving inotropic agents, were more likely to have these drugs withdrawn (63% vs 41%, P < .001). In the multivariate analysis, 4 factors were significant independent predictors of withdrawal of ventilatory support: use of inotropic agents (hazard ratio 1.8), physician’s prediction of ICU survival <10% (hazards ratio 3.5), physician’s prediction that the patient would have impairment of cognitive function so severe that the patient would not leave the hospital (hazards ratio 2.5), and physician’s perception that the patient preferred to limit life support (hazards ratio 4.2).

  Interpretation.   Patients who had withdrawal of ventilatory support or who died while receiving mechanical ventilation had a shorter ICU stay than did patients who were weaned from mechanical ventilation. These findings contrast with those of studies of 2 decades ago in which patients who eventually died in the ICU had a longer ICU stay and greater resource consumption than did patients who lived.¹³

The factors of age, previous functional status, illness severity, and organ dysfunction were not independently associated with withdrawal of ventilatory support. Rather the strongest determinants of withdrawal reflected physicians’ perceptions of a patient’s preferences to limit life support and the probability of ICU survival and future cognitive function and dependency on inotropic agents.

	Clinical Prediction of ICU Mortality

Top Abstract Methods Cardiopulmonary Resuscitation... Provision of Advanced Life... Clinical Prediction of ICU... Clinicians' Discomfort With Life... Discussion References

 
Because the clinical decision to limit life support is influenced by prognostication of potential survival and future quality of life, we evaluated the reliability of such estimates as they evolve over time. Scoring systems used to measure severity of illness (APACHE and organ dysfunction scores) are of limited value in the prognostic assessment of individual patients. Nonetheless, clinicians and patients’ families would all benefit from accurate prognostic information when making decisions such as whether and when patients might benefit from an ICU admission. Moreover, patients’ families often ask about prognosis during meetings with clinicians, and completeness of these family discussions is a determinant of families’ satisfaction.¹⁴ Clinically useful prediction models remain elusive, and the ability of ICU nurses and physicians to recognize which patients have the highest risk for death during the course of critical illness has not been analyzed.

Furthermore, decisions about life support and treatment plans can be complex, dynamic, and emotionally challenging, and these plans may cause clinicians discomfort. Discomfort with treatment plans among clinicians, in turn, may have a profound effect on patients’ care and may result in conflict among clinicians and between clinicians and patients’ families when commencing, continuing, withholding, or withdrawing advanced life-support treatment.

Objective 5: To Determine the Predictive Ability of and Outcomes Associated With Daily Clinicians’ Estimates of Low Probability of ICU Survival for Patients Receiving Mechanical Ventilation
  Specific Methods.   For objective 5,¹⁵ we undertook a daily evaluation of organ dysfunction; use of and plan for mechanical ventilation, inotropic agents, or hemodialysis; DNR directives; and patients’ preferences to limit life support for 851 patients (see Figure). Each day after morning rounds, we asked the ICU physician and each patient’s nurse to predict the probability of each patient’s survival in the ICU (<10%, 10%–40%, 41%–60%, 61%–90%, >90%). Each respondent was blinded to the others’ ratings and to the patient’s APACHE II score and MODS. We used Cox regression analysis to examine baseline and time-dependent determinants of ICU mortality, including physicians’ predictions of ICU survival. In the regression analysis, we included only predictions made at least 48 hours before a patient’s death in the ICU or discharge from the unit to avoid inflating the accuracy of clinicians’ predictions that simply reflected an imminent decision to withdraw life support.

  Results.   For 341 patients (40%), on at least one occasion, a physician’s predicted ICU survival rate was less than 10%. ICU mortality was 71% and 12%, respectively, for patients with low (<10%) versus higher (>10%) predicted probability of survival (P < .001). Actual ICU survival rates exceeded physicians’ estimated survival for all ranges of survival prediction.

We found that physicians’ estimated probability of ICU survival of less than 10% influenced the patterns of life-support provision and limitation. Mechanical ventilation, use of inotropic agents, and dialysis (all P < .001) were withdrawn more often when physicians predicted that the probability of a patient’s ICU survival was less than 10%. Independent predictors of ICU mortality were APACHE II scores (hazards ratio 1.2 for 5-point increase), daily MODS (hazards ratio 2.5 for 5-point increase), use of inotropic agents (hazards ratio 2.1), use of dialysis (hazards ratio 0.5), patients’ preference to limit life support (hazards ratio 10.2), and physicians’, but not nurses’, prediction that the probability of survival was less than 10%. These associations held when we repeated the analysis and excluded all patients who had ventilatory support withdrawn. In other words, physicians’ predictions that a patient’s probability of ICU survival was less than 10% were significantly associated with death in the ICU, even after adjustment for patients’ preferences to limit life support, dependency on inotropic agents, need for dialysis, and baseline and evolving severity of illness (APACHE II and organ dysfunction scores).

  Interpretation.   Physicians tend to overestimate risk of ICU mortality and cannot judge with sufficient accuracy which patients will survive. We found that physicians’ prediction of a low probability of ICU survival was more strongly associated with actual ICU mortality than were clinical factors such as illness severity, evolving or resolving organ dysfunction, and use of inotropic agents and that an estimated low probability of ICU survival strongly influenced provision and limitation of life-support treatments.

These results emphasize a further challenge to valid end-of-life discussions, namely, the provision of an accurate prognosis. Therefore, while we continue to develop and evaluate objective predictions of outcomes in the ICU, the prognosis understood by ICU physicians remains strongly influential (although not necessarily as accurate as we would like) in the framing of end-of-life discussions.

Physicians tend to overestimate risk of ICU mortality and cannot judge with sufficient accuracy which patients will survive.

 

	Clinicians’ Discomfort With Life-Support Treatments

Top Abstract Methods Cardiopulmonary Resuscitation... Provision of Advanced Life... Clinical Prediction of ICU... Clinicians' Discomfort With Life... Discussion References

 
Finally, recognizing that uncertainty exists in understanding the specific wishes of patients and their families, and in providing a reliable estimate of ultimate prognosis, and that differences exist between clinicians in evaluating these parameters, we focused on clinicians’ comfort with the provision or withdrawal of life-support treatments.

Objective 6: To Determine the Incidence and Predictors of Healthcare Providers’ Discomfort With Advanced Life Support Plans for Patients Receiving Mechanical Ventilation
  Specific Methods.   For objective 6,¹⁶ we focused on 657 patients (see Figure) and recorded the daily treatment plan for mechanical ventilation, use of inotropic agents, and dialysis. We used 5 categories to describe the life-support plans; for mechanical ventilation they were (1) continue mechanical ventilation, (2) withdraw ventilatory support in anticipation of death, (3) wean the patient from mechanical ventilation in anticipation of improvement, (4) use mechanical ventilation if necessary, and (5) withhold mechanical ventilation. Each bedside nurse, resident. and physician recorded how comfortable he or she was with the life-support plan, and if not comfortable why not with respect to whether the plan was too technologically intense or not intense enough. We used hierarchical logistic modeling to determine predictors of the life-support plan.

  Results.   We detected discomfort with the life-support plan during 8% of 16 354 daily observations. Despite high levels of overall comfort (92% of patient-day observations), at least one ICU clinician expressed discomfort with the plan for 43% of patients on at least one occasion. Discomfort, when it occurred, was experienced most often by nurses (42%); the percentages were lower for physicians (35%) and residents (24%). Clinicians more often experienced discomfort because plans were too technologically intense rather than because the plans were insufficiently intense (94% vs 6%, P < .001). The primary reasons for discomfort were perceptions that a patient’s family had overestimated the patient’s chances of survival (52%) or future quality of life (46%), that life support was prolonging the dying process (36%), and that resources were being used inappropriately (30%).

In addition to significant associations of discomfort with ICU admission factors (age, a medical diagnosis, increasing organ dysfunction, and poor functional status), several daily patient factors were also associated with clinicians’ discomfort: patients with worse organ dysfunction (odds ratio 1.6 for each 5-point increase on the MODS), patients who required dialysis (odds ratio 2.5) or for whom there was a plan to withhold dialysis (odds ratio 2.0), and patients during the first week of their ICU stay (odds ratio 1.8). In contrast, for patients treated with mechanical ventilation who were subsequently extubated, the plan to withhold future mechanical ventilation was predictive of comfort with the life-support plan (odds ratio 0.2; the odds ratio <1 indicates that clinicians were less uncomfortable if a plan existed to forgo reintubation and mechanical ventilation of extubated patients). We also found a center effect.

  Interpretation.   Clinicians often experience discomfort about life-support plans for patients receiving mechanical ventilation. Discomfort occurs more often among nurses than among physicians and is more likely for older, more severely ill medical patients and for extubated patients for whom there are no plans to withhold mechanical ventilation. Because they spend more time with patients and patients’ families at the bedside than residents and physicians do, nurses may have a better understanding of patients’ values and preferences than the other clinicians do.

These results emphasize the importance of effective communication and collaborative planning among ICU team members. Discomfort among members of the ICU team should prompt further discussions among team members, and with patients and the patients’ families. Reconsidering treatment plans and examining the roots of clinicians’ discomfort is necessary to improve the quality of decision making, which involves optimal processes as well as outcomes. Our goal should be more appropriate, compassionate, and harmonious care for critically ill patients.

	Discussion

Top Abstract Methods Cardiopulmonary Resuscitation... Provision of Advanced Life... Clinical Prediction of ICU... Clinicians' Discomfort With Life... Discussion References

 
In the Level of Care Research Program, we addressed several interdependent research questions by doing a series of detailed analyses of data from a multinational observational study. Other research program models exist, such as the US Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatment,¹⁷ which consisted of a cohort study and a randomized trial that generated multiple additional observational studies. Other models include observational studies to design and interpret the results of randomized clinical trials such as the models we used for a research program on stress ulcer prophylaxis¹⁸ or the single-center, multistudy management of several projects conducted in a single ICU.¹⁹ The Table gives outcomes from the Level of Care Research Program from several perspectives.

Among the more striking findings of these studies was the difference between factors associated with establishing a DNR directive and with making the decision to withdraw life-support treatment.^1,4,7 Patients’ age, disease severity, and functional impairment before ICU admission all increased the frequency with which clinicians established a DNR order.^1,4 We were not surprised that factors specific to the ICU influenced these decisions. However, the extent to which nonclinical factors influenced decisions was unanticipated. For example, the frequency with which patients had a DNR established differed among countries and between cities within countries, between ICUs within cities, and between patients admitted on weekends and weekdays and during the day or at night. We were similarly surprised by the observation that the frequency of DNR orders starts increasing when patients are 50 years old and that previous impairment in functional status appears to increase the likelihood of a DNR order in patients who cannot participate in decision making but not in patients who can participate.

Frequency of DNR orders starts increasing when patients are age 50.

 

Prior impairment in functional status increases the likelihood of a DNR order in patients unable to participate in decision making, but not in patients able to participate.

 

Factors influencing the decision to withdraw life-support treatment versus continue full treatment until death differ from those that guide the initial provision of advance directives. We studied mechanical ventilation as the prototypical life-support treatment, the one most commonly used and withdrawn. We found that patients’ age, disease severity, and functional status did not influence the decision to withdraw mechanical ventilation.⁷ Instead, physicians’ prediction of poor probability of survival became an important factor, as it does in the provision and limitation of other life-support therapy. The justifications for these predictions are difficult to untangle. To what extent do physicians have insights beyond physiological variables (our analysis suggests that to some extent, they do), and to what extent is this insight a self-fulfilling prophecy (eg, a physician thinks that a patient will not survive, life support is withdrawn, and the patient dies)? The impact of physicians’ predictions holds true even for patients who do not have mechanical ventilation withdrawn.¹⁵ Physicians’ perceptions that patients would have severely impaired cognitive function long-term and physicians’ perceptions that patients would prefer limitation in life-support treatment were also predictors of the decision to withdraw mechanical ventilation.

The influences of a patient’s age, disease severity, and previous functional status seem to disappear as a critical illness evolves and the response (or lack thereof) to ICU interventions becomes more influential. Our analysis of clinicians’ discomfort provides some additional insight. First, discomfort was almost always a result of the perception that life-support plans were too intense rather than not intense enough.¹⁶ Second, the predictors of discomfort included patients’ age, APACHE score, and previous functional status. Earlier, we found that physicians obtain DNR orders through dialogue with older, sicker, functionally impaired patients and the patients’ families.¹ Maintaining life support for such patients who do not respond to treatment is likely to generate the feelings of discomfort.¹⁶

The most powerful predictor of withdrawal of life-support treatment was physicians’ perception that patients preferred to limit life support. However, how congruent this perception was with the patients’ true wishes is unclear. Patients perceived to prefer limited life support accounted for 30% of those who had withdrawal of ventilatory support. That finding means that some of the 70% of patients who had ventilatory support withdrawn may have preferred intensive life support. Of patients who died while receiving mechanical ventilation, physicians thought that 11% preferred limited life support. In these 11%, why did intense advanced life support continue until death?

The possibility that a significant divergence occurs between a patient’s wishes and his or her physician’s perception of those wishes suggests a productive line of future research. Although we collected data on patients’ demographic characteristics, physiological status, life-support treatments administered, and health providers’ perceptions of the intensity of the life-support treatment plan, we did not attempt an extended follow-up, and we did not have any way to measure alignment of patients’ perceptions and wishes with the perceptions and wishes of their families. Enormous ethical and logistical challenges are associated with measurement of perceptions of critically ill patients and patients’ families in intense distress faced with participation in discussions about agonizing decisions. An understanding of the process of end-of-life decision making that goes beyond the insights of our studies will require both qualitative and quantitative studies.

What do our results suggest for the present? Clinicians should have a consistent, compassionate approach to explaining CPR and the outcomes expected with advanced life support. Clinicians should adapt the shared decision-making model,³⁰ as appropriate, to the needs of different patients and the patients’ families, sensitively counseling the patients and families through the decision-making process. Educational programs are needed for clinicians and clinicians-in-training to address the challenges of this process, with formal evaluation to ensure that training programs are meeting this need. Patients do not want their fate determined by the serendipity of the ICU in which they land, or by the ICU team to which their care is entrusted. Our next challenge is to develop and consistently deliver quality multidisciplinary end-of-life care for all dying patients and their loved ones in the ICU.³¹ Research programs are currently being conducted to address this goal.

	ACKNOWLEDGMENTS

 
This study was funded by the following national and provincial agencies: Medical Research Council of Canada; Department of Health, Province of Nova Scotia; Physicians Services Incorporated of Ontario; British Columbia Medical Services Foundation; and the Research Committee of the Orebro County Council, Sweden.

Additional funding was provided by the following regional peer-review agencies: Father Sean O’Sullivan Research Center, St. Joseph’s Hospital; Health Services Research Fund, London Health Sciences Center; University Internal Medicine Research Fund, Dalhousie University; Faculty of Medicine Intramural Grant, Dal-housie University; Camphill Medical Center Research Fund, Halifax; and the Queen Elizabeth II Health Sciences Research Fund, Halifax. D. Cook is a research chair of the Canadian Institutes for Health Research.

We thank the Canadian Critical Care Trials Group for their support of this research. We express our appreciation to the research nurses, bedside nurses, residents, and attending physicians who participated in this study.

Site investigators and research nurses were as follows: Peter Dodek, MD, and Carol Honeyman, RN, St. Paul’s Hospital, Vancouver, British Columbia; John Marshall, MD, Debra Foster, RN, and Chanel McKenna, General Division, Toronto Hospital, Toronto, Ontario; Neil Lazar, MD, and Marilyn Steinberg, RN, Western Division, Toronto Hospital; David Leasa, MD, and Sue Langdon, RN, University Hospital, London Health Sciences Center, London, Ontario; Ann Kirby, MD, Mary Katherine Scott, RN, and Mary van Soeren, RN, St. Joseph’s Hospital, London Health Sciences Center; Deborah Cook, MD, and Ellen McDonald, RN, St. Joseph’s Hospital, Hamilton, Ontario; Allan McLellan, Serge Puksa, MD, and Andrea Tkaczyk, RN, Henderson Division, Hamilton Health Sciences Corporation, Hamilton, Ontario; Christine Bradley, MD, Nicole Krolicki, RN, and Susan Caldwell, RN, General Division, Hamilton Health Sciences Corporation; Cindy Hamielec, MD, and Nancy Merrill, RN, McMaster Division, Hamilton Health Sciences Corporation; Graeme Rocker, MD, and Mary Gordon MacKenzie, PhD, Victoria General Hospital and Halifax Infirmary, Halifax, Nova Scotia; Joseph Varon, MD, and Cheryl Keenan, RN, M.D. Anderson Cancer Center, Houston, Tex; Mitchell Levy, MD, and Mary Beth Fucci, RN, Rhode Island Hospital, Providence, RI; Peter Sjokvist, MD, and Mia Svantesson, RN, Orebro Hospital, Orebro, Sweden; and Simon Finfer, MD, Malcolm Fisher, MD, Joey Penfold, RN, and Anne O’Connor, RN, Royal North Shore Hospital, Sydney, Australia.

At the Methods Center, Lisa Buckingham and Nicole Zytaruk served as study coordinators, Lisa Buckingham provided database management, and Suzanne Duchesne, Sandi Reeve, Barbara Jedrzejowski, and Laurel Raftery provided data entry.

To purchase electronic or print reprints, contact The InnoVision Group, 101 Columbia, Aliso Viejo, CA 92656. Phone, (800) 809-2273 or (949) 362-2050 (ext 532); fax, (949) 362-2049; e-mail, reprints{at}aacn.org.

	REFERENCES

Top Abstract Methods Cardiopulmonary Resuscitation... Provision of Advanced Life... Clinical Prediction of ICU... Clinicians' Discomfort With Life... Discussion References

 

1. Cook DJ, Guyatt GH, Rocker G, et al. Cardiopulmonary resuscitation directives on admission to the intensive care unit: an international observational study. Lancet. 2001;358:1941–1945.[Medline]
2. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med. 1985;13:818–829.[Medline]
3. Marshall JC, Cook DJ, Christou NV, Bernard GR, Sprung CL, Sibbald WJ. Multiple organ dysfunction score: a reliable descriptor of a complex clinical outcome. Crit Care Med. 1995;23:1638–1652.[Medline]
4. Guyatt GH, Cook DJ, Weaver B, et al. Influence of perceived functional and employment status on cardiopulmonary resuscitation directives. J Crit Care. 2003;18:133–141.[Medline]
5. Sinuff T, Cook DJ, Rocker G, et al. DNR directives are established early in mechanically ventilated intensive care unit patients. Can J Anesth. 2004; 51:1034–1041.[Abstract/Free Full Text]
6. Rapoport J, Teres D, Lemeshow S. Resource use implications of do not resuscitate orders for intensive care unit patients. Am J Respir Crit Care Med. 1996;153:185–190.[Abstract]
7. Cook DJ, Rocker G, Marshall J, et al. Withdrawal of mechanical ventilation in anticipation of death in the intensive care unit. N Engl J Med. 2003;349:1123–1132.[Abstract/Free Full Text]
8. Faber-Langendoen K. The clinical management of dying patients receiving mechanical ventilation: a survey of physician practice. Chest. 1994; 106:880–888.[Abstract/Free Full Text]
9. Vincent JL. Forgoing life support in Western European intensive care units: the results of an ethical questionnaire. Crit Care Med. 1999;27:1626–1633.[Medline]
10. Cook DJ, Guyatt GH, Jaeschke R, et al.. Determinants in Canadian health care workers of the decision to withdraw life support from the critically ill. Canadian Critical Care Trials Group. JAMA. 1995;273:703–708.[Abstract]
11. Sjokvist P, Nilstun T, Svantesson M, Berggren L. Withdrawal of life support: who should decide? Differences in attitudes among the general public, nurses and physicians. Intensive Care Med. 1999;25:949–954.[Medline]
12. Ferrand E, Lemaire F, Regnier B, et al. Discrepancies between perceptions by physicians and nursing staff of intensive care unit end-of-life decisions. Am J Respir Crit Care Med. 2003;167:1310–1315.[Abstract/Free Full Text]
13. Detsky AS, Stricker SC, Mulley AG, Thibault GE. Prognosis, survival, and the expenditure of hospital resources for patients in an intensive care unit. N Engl J Med. 1981;305:667–672.[Abstract]
14. Heyland DK, Rocker GM, Dodek PM, et al. Family satisfaction with care in the intensive care unit: the results of a multicenter study. Crit Care Med. 2002;30:1413–1418.[Medline]
15. Rocker G, Cook DJ, Sjokvist P, et al. Clinician predictions of intensive care unit mortality. Crit Care Med. 2004;32:1149–1154.[Medline]
16. Griffith L, Cook DJ, Hanna S, et al. Clinician discomfort with life support plans for mechanically ventilated patients. Intensive Care Med. 2004; 30:1783–1790.[Medline]
17. The SUPPORT Principal Investigators. A controlled trial to improve care for seriously ill hospitalized patients: the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatment (SUPPORT) [published correction appears in JAMA. 1995;275:1232]. JAMA. 1995;274:1591–1598.[Abstract]
18. Cook D, Heyland D, Marshall J, the Canadian Critical Care Trials Group. On the need for observational studies to design and interpret randomized trials in ICU patients: a case study in stress ulcer prophylaxis. Intensive Care Med. 2001;27:347–354.[Medline]
19. Foster D, Cook DJ, Granton J, Steinberg M, Marshall J. Use of a screen log to audit patient recruitment into multiple randomized trials in the intensive care unit. Canadian Critical Care Trials Group. Crit Care Med. 2000;28:867–871.[Medline]
20. Rocker G, Cook DJ, Weaver B, et al. Withdrawal of mechanical ventilation in the intensive care unit [abstract]. Crit Care Med. 2003;31:A84.
21. Cook DJ, Weaver B, McDonald E, et al. Predicting mortality in the ICU [abstract]. Crit Care Med. 2003;31:A554.
22. Griffith L, Cook DJ, Hanna S, et al. Predictors of clinician discomfort with daily life support plans for mechanically ventilated patients [abstract]. Crit Care. 2003;7(suppl 2):P252.
23. Rocker G, Cook DJ, Griffith L, et al. Physician predictions of ICU survival: a multicentre study [abstract]. Am J Resp Crit Care Med. 2004; 169:A624.
24. Sinuff T, Cook DJ, Rocker G, et al. Resuscitation directives in the intensive care unit [abstract]. Am J Resp Crit Care Med. 2004;169:A623.
25. Adhikari N, Sinuff T, Cook DJ, Schunemann H, Griffith L, Rocker G, Walter S. Mortality predictions in the intensive care unit [abstract]. Proc Am Thorac Soc. 2005;2:A424.
26. Cook DJ, Todd TRJ. The Canadian Critical Care Trials Group: a collaborative educational organization for the advancement of adult clinical ICU research. Intensive Care World. 1997;14:68–70.
27. Rocker G, Heyland D, New research initiatives in Canada for end-of-life and palliative care. CMAJ. 2003;169:300–301.[Free Full Text]
28. Schroeder SA. The legacy of SUPPORT. Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments. Ann Intern Med. 1999;131:780–782.[Free Full Text]
29. Cook DJ, Sibbald WJ. The progress, the promise and the paradox of technology assessment in the intensive care unit. CMAJ. 1999;161:1118–1119.[Free Full Text]
30. Carlet J, Thijs LG, Antonelli M, et al. Challenges in end-of-life care in the ICU. Statement of the 5th International Consensus Conference in Critical Care: Brussels, Belgium, April 2003. Intensive Care Med. 2004;30:770–784.[Medline]
31. Clarke EB, Curtis JR, Luce JM, et al. Quality indicators for end-of-life care in the intensive care unit. Crit Care Med. 2003;31:2255–2262.[Medline]

Journal Club Article Discussion Points

When critically appraising this issue’s AJCC journal club article, "Levels of Care in the Intensive Care Unit: A Research Program," consider the questions and discussion points listed below.

Study Synopsis: This multisite study examined advance directives, life-support decisions, and clinical prediction of mortality in adult patients receiving mechanical ventilation (MV) in the intensive care unit (ICU). The Level of Care Research Program was conducted in 15 adult ICUs in North America, Europe, and Australia. The study had 6 objectives: to assess the prevalence of cardiopulmonary resuscitation (CPR) directives within 24 hours of ICU admission, to determine the effects of functional and employment status on CPR directions, to assess the rate and determinants of CPR directives, to determine the influence of factors on withdrawal of MV, to assess clinicians’ predictions of mortality, and to determine clinicians’ discomfort with life-support plans. A total of 3099 patients were included in the study, and each objective focused on a different number of patients, ranging from 657 to 2916.

The results indicated that only 11% of patients had resuscitation directives, with half indicating resuscitation and half having do-not-resuscitate (DNR) status. A significant percentage (44%) of patients could not participate in decision making during the first 24 hours of ICU admission. Most directives (65%) were established in the ICU, and increasing age strongly predicted DNR directives, whereas severity of illness predicted both DNR and CPR directives. Severe functional limitations and unemployment were associated with DNR plans. DNR directives were made for 30% of patients, with most (62%) occurring between days 2 and 10. Factors associated with DNR directives after ICU admission were age, medical diagnosis, physician’s prediction of ICU survival, and physician’s perception that the patient preferred to limit life support. Most patients (63%) were successfully weaned from MV, and the strongest determinants of MV withdrawal were based on physicians’ perceptions of the patient’s preferences to limit life support and probability of ICU survival. Actual ICU survival rates exceeded physicians’ estimates. Clinicians’ overall comfort with life-support decisions was high (92%), yet nurses most frequently experienced discomfort due to perceptions that life-support plans were too technologically intense.

1. Description of the Study
   • What were the specific areas of interest for the study objectives?
     
   • Why were the specific areas of focus on decision making about life-support measures chosen?
     
   • Why are these areas significant/important?
     
   
   
2. Literature Evaluation
   • What has previous research on advance directives in the ICU indicated?
     
   
   
3. Methods and Design
   • Describe the study methods.
     
   • Who were the subjects and what were the inclusion criteria for participation in the study?
     
   • What data were collected?
     
   
   
4. Results
   • What were the findings of the research?
     
   • What were the differences in the study findings between factors associated with establishing a DNR directive and with making the decision to withdraw life support?
     
   • What was the most powerful predictor of withdrawal of life support?
     
   
   
5. Clinical Significance
   • What are the implications of the study for clinical nursing?
     
   
   

Information From the Authors: Deborah Cook, MSc(Epid), MD, and Ellen McDonald, RN, lead author and a co-author of this journal club article, provided additional information about the study. They explain that the research team chose to conduct the study because up to 20% of patients die in the ICU. "Early qualitative work on the moral distress of clinicians and families around life-support withdrawal led us to want to study this area further. We developed several research questions to improve our understanding of the use of life-support technology, in the context of the growing recognition that patients’ values were beginning to strongly influence life-support decisions. The Canadian Critical Care Trials Group was very keen to conduct the Levels of Care Research Program since this series of studies was a true collaboration involving bedside ICU nurses, residents, and attending physicians. Taking a program of this scope and conducting a series of projects like these had not been done before in an integrated fashion in the ICU."

Several of the study findings were surprising to the authors. Cook and McDonald share, "We were surprised that ICU patients today who die, do so sooner than those patients who ultimately survive and are discharged from the ICU. The reverse was the case 2 decades ago, such that ICU patients who ultimately died, did so after a much longer length of stay than those patients who were discharged alive."

"Another surprising finding was that 3 main determinants of ventilator withdrawal were the physicians’ prediction of ICU survival as less than 10%, perceptions of long-term severely impaired cognitive function, and perceptions that patients would prefer life-support limitation. Notably, these factors were more important determinants of ventilator withdrawal than the traditional APACHE [Acute Physiology and Chronic Health Evaluation] score. These findings reflect a process of life-support withdrawal today that is more patient-centered than in the past. They also underscore the clinical judgments that influence this decision and the importance of accurately representing patients’ wishes."

Implications for Practice: Addressing patients’ preferences for life-sustaining treatment during critical illness is an important aspect of ICU care. This study has several implications for clinical practice, including that DNR status and patients’ preferences should be clarified early during ICU hospitalization. The authors provide additional insight: "We recommend a consistent, compassionate approach to explaining the outcomes associated with CPR and advanced life support in general, for patients about to enter the ICU, as well as those already in the ICU. The shared decision-making model seems to be the dominant one in North America, but the approach must be adapted to different patients and families to guide them sensitively. Educational programs are needed for clinicians and clinicians-in-training to address the challenges of this process, with formal evaluation to ensure that training programs are meeting this need. End-of-life care should be on the agenda of all major critical care conferences, as it has been for the past few years. Finally, we also recommend development of a portfolio of quality indicators to try to raise the bar with respect to end-of-life care for all dying patients and their loved ones."

The role of the nurse in assisting family members and patients in decision making about life support is a key one. The authors agree: "Bedside nurses are the main caregivers of ICU patients on a day-to-day basis, and often learn first about the values and preferences of the patients they care for through intimate discussions with family members. We believe that nurses have crucial input into the decision-making process."

Journal Club feature commentary is provided by Ruth Kleinpell.

This Article Abstract Full Text (PDF) Respond to This Article Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Cook, D. Search for Related Content PubMed PubMed Citation Articles by Cook, D.