Introduction

Decisions to limit life support are among the most important clinical decisions encountered by patients, families, and providers. Decisions to limit life support have been associated with high mortality rates, as well as a high degree of psychological distress in families and providers [1–7]. Understanding how life support decisions are made has been identified as a priority by European and American professional societies [8, 9].

Characteristics of the individual intensivist have been associated with varied rates and timing of life support decisions [10]. Physician-level factors influencing decisions to limit life support include geography, gender, religion, personal life support preferences, practice setting, specialty, surgical investment, experience, medical errors, and familiarity with life support [10–19]. While previous studies have focused almost exclusively on survey responses to hypothetical scenarios, we hypothesized that important clinically relevant insights would be gained by approaching the topic utilizing qualitative methods. Our objective was to explore factors that contribute to physician variability in decisions to limit life support.

Methods

Participants

We employed a random sampling strategy of 15 physicians and ten nurses. Five physicians were enrolled from each of the following: (1) medical ICU staffed by pulmonary critical care intensivists, (2) surgical ICU staffed by anesthesiologists with critical care training, and (3) surgical ICU staffed by general surgeons with critical care training. In each ICU, pulmonologists, anesthesiologists, and surgeons, respectively, take primary responsibility for patient care and life support decision making. We contacted 25 physicians with at least 10 years ICU experience via email to achieve our goal of 15. After preliminary analysis, data saturation was not achieved, so two additional pulmonary critical care physicians were recruited. We contacted 35 nurses with at least 3 years ICU experience to achieve our goal of ten (five nurses from surgical ICU and five from medical ICU settings). The Mayo Clinic Institutional Review Board approved the protocol.

Data collection

Interviews were conducted by two interviewers (M.E.W. and L.R.) in private conference rooms while the participants were off-duty. Interviews lasted approximately 1 h and were conducted using an open ended interview guide (see Appendix A). Interview questions were based on literature review, clinical observations, and expert opinion. The questions were pilot tested and modified using a small group of ICU physicians who were not study participants. Additional questions and probes were used based on participant responses. Interviews were audiotaped, transcribed verbatim, and verified for transcription accuracy.

Data analysis

Utilizing principles of ground theory, open and selective coding of each transcript was performed to identify categories and factors. Qualitative software was utilized to manage the data (NVivo, QSR International, Doncaster, Victoria, Australia). Two investigators [M.E.W. (physician) and L.R. (nurse)] performed the open coding of the first four interviews and together generated a list of factors. The subsequent transcripts were independently coded and new factors were added if passages did not fit into the existing list. At the end of the coding process, the investigators met to clarify the list of factors, review each coded statement, and select representative statements. Data saturation was achieved when no new factors were identified. Trustworthiness of the analyses was assessed by having an investigator, who was blinded to the coding assignments, independently code 30 % of passages. The average κ statistic was 0.88 (range 0.79–0.97). Excellent interrater reliability is represented by a κ > 0.8.

Results

Eleven factors within four categories were identified that influenced physician variability in decisions to limit life support (Table 1). Tables 2, 3, 4, 5 show additional quotes for each factor. Electronic supplementary Table 6 shows the demographic characteristics of participants.

Table 1 Frequency of factors that influence the decision to limit or continue life support
Full size table
Table 2 “Physician work environment” factors and illustrative quotes
Full size table
Table 3 “Physician experiences” factors and illustrative quotes
Full size table
Table 4 “Physician attitudes” factors and illustrative quotes
Full size table
Table 5 “Physician relationship with patient and family” factors and illustrative quotes
Full size table

Category 1: physician work environment

Workload and competing priorities

In a high workload environment, communicating with families about decisions to limit life support may be a lower priority than other competing tasks. Although important, addressing decisions to limit life support was not measured or rewarded. Two physicians explained: “When things are extremely busy, (it) becomes difficult to sit down and have a (family) discussion.” “(Addressing life support decisions) is not something that anyone has ever given me feedback on. (But I do) get metrics on how many operations I do, how much I bill, a wound infection rate, line infection rate, how many papers (I) published last year. It will be noticed if you don’t write the paper, but it won’t be noticed if you don’t talk to the family.”

Shift changes and handoffs

Physician shift changes delayed addressing life support decisions by disrupting continuity of care. Because it took time for each physician to personally assess the patient and family prior to reaching a life support decision, in some instances there was not enough physician continuity of care to reach a decision prior to rotating off service. One physician noted, “We change staff positions once a week so that sometimes nobody ever gets enough longitudinal involvement to be able to make a decision on the patient.”

Additionally, physician approaches to handoffs regarding life support decisions varied. One physician noted that he routinely conducted handoffs “at the bedside,” while another physician noted that handoffs occurred via “telephone” or “email.” Sometimes key information about patient wishes was not effectively communicated during handoffs. In one example, an intensivist discussed the plan of care to continue life support during a handoff to a nighttime intensivist, but did not discuss the specific patient wishes to continue life support nor the family dynamics. That evening, multiple family members unexpectedly desired to withdraw life support, and without contrary information about the patient’s wishes, the evening intensivist withdrew life support. The daytime intensivist felt this could have been avoided with a more complete handoff.

Incorporation of nursing input

Both nurses and physicians felt that nurses offered a unique perspective on the assessment of patient and surrogate wishes and wellbeing. The degree to which physicians sought out and incorporated the nursing perspective was variable. One nurse noted, “(Some intensivists) deal with the family, and we (nurses do) our thing and there is no cross communication between us. And then, we (nurses) don’t offer anything. Even if we do, it is not acknowledged. You learn just kind of by knowing who the (intensivist) is and where they kind of see (your role in decision making).” One physician offered a differing viewpoint: “Nurses have much more insight (into the patient and family). (When) the nurses have had a conversation with the patient or the family that offers valuable insight and information, I’d be wise to think about that.”

Category 2: physician experiences

Experiences of unexpected patient survival

Some patients survived and recovered despite physician recommendation to limit life support. These outcomes of surprising survival influenced future decisions such that most intensivists perceived more uncertainty in communicating prognosis and were more open to families who wished to continue life support. Seven out of the 17 physicians noted that a single surprising outcome was a significant turning point in their decision making. One physician recounted, “This patient surprised all of us by having a perfectly acceptable quality of life when we had all given up on him. Now I think I’m much more circumspect and I don’t deal in absolutes. I was embarrassed that I had been so certain when the patient’s mother, who I considered an ignorant, unrealistic person—she was absolutely right!”

Many intensivists also noted that a lack of exposure to the long term expected or unexpected outcomes of their ICU patients was a barrier to counseling patients. One physician noted, “I don’t see these patients post ICU stay. So my world view is skewed by this. The only patients (I) see are doing poorly and come back to the ICU.” Other intensivists noted that their “work in the chronic ventilator unit” provided them exposure to long term outcomes of critically ill patients. This exposure enabled them to better counsel patients regarding chronic critical illness.

Experiences of limiting life support in physician’s family

Intensivists who experienced decisions to withdraw life support for their own family members could better understand, empathize with, or provide support to patients and families during future life support decisions. One intensivist noted, “4 days after (my daughter) graduated from college someone ran a red light and broadsided her. I sat in the ICU and watched her die from her brain injuries for the next 30 h. That has had a very, very profound impact on me and how I make these decisions and how I talk to people, from having been on the other side of the bed. I have the ability and an insight and I talk to families in ways that others generally cannot because they haven’t been on the other side of the bed.”

Category 3: physician attitudes

Investment in a good surgical outcome

Surgeon and non-surgeon intensivists, as well as nurses, observed that when a physician was personally invested in a good surgical outcome, there was less inclination for that surgeon to address life support decisions, limit life support, or recognize a poor prognosis. For example, one surgeon intensivist noted, “(When a surgeon) has performed an elective surgery (and) there have been complications, I find tremendous reluctance for the surgeon to approach the family about end-of-life issues. People feel guilty that their surgery went bad. It is extremely uncomfortable to have contributed to a patient’s poor outcome.”

Specialty perspective

Compared to specialists, intensivists were described as more likely to have a comprehensive, big-picture view of the patient’s clinical situation. Many participants perceived that specialists focused on prolonging life by curing disease at the expense of other aspects such as prognosis, long term functional status, and quality of life. One intensivist noted, “One of the things that is brought to the table by an intensivist is a broader view of a patient’s condition—an incorporation of various subspecialty assessments. With oncology, it may be that the cancer is entirely curable; however, when bringing in the other aspects, the overall prognosis is different than that of the specific organ system.”

Values and beliefs

A physician’s ethical and religious beliefs regarding life support were observed to be reasons for variability. One intensivist noted, “Some physicians just simply do not believe that withdrawal of life support is appropriate in any circumstance—that it is a disrespect of life. (A while ago, I remember one cardiologist who) believed it was killing someone to deprogram their pacemaker. (These) kinds of beliefs are not easily modifiable.”

Category 4: physician relationship with patient and family

Hearing the patient’s wishes firsthand

Intensivists who reported having heard the patient’s wishes firsthand, rather than summarized by another provider, surrogate decision maker, or advance directive, more strongly advocated for those wishes than other providers who did not have a firsthand account. In describing one difficult decision, an experienced nurse commented, “It is a little easier to accept when you have somebody face-to-face, (the patient) telling you this is what I want, please honor my wishes (rather than looking at the advance directive where) the wishes are spelled out.” An intensivist commented that without a firsthand knowledge of patient wishes, “You cannot really assert against 15 family members who have made up their mind (to withdraw support). It is very difficult to explain (the rationale for continuing life support).”

Engagement in family communication

When faced with life support decisions, some intensivists readily engaged (prioritized or devoted time to) decision making, and other intensivists disengaged (avoided or delayed) decision making. While factors (such as workload) influenced physician engagement for a particular encounter, participants also noted that some physicians, regardless of such circumstances, were routinely and predictably more engaged than their colleagues in approaching life support decisions. Reasons cited for this engaged approach included interest, training, comfort, and general experience with providing end-of-life care. One nurse noted, “I know one physician who is very, very good with (life support decisions) and he tends to talk about them right away with the family. He is extremely interested in end-of-life care (and) really (approaches) it differently than other (physicians).” An intensivist described, “I know some (physicians) who don’t have that much interest in (end-of-life care), so they just go with the flow. (They) want to avoid complex decisions with a difficult family or patient.”

Family negotiation

When disagreement arose between physicians and families regarding life support, some intensivists willingly accepted the family wishes (deferred to their substituted judgment) and other intensivists pushed back against the family wishes, attempting to negotiate with families to make a different decision. Some physicians were more likely to offer a specific recommendation to limit or continue life support. The following experience highlights both approaches, first a physician who pushed back against family wishes to withdraw support and an incoming physician who did not push back, but accepted the family wishes. An intensivist recounted, “(The patient’s) wife said (the patient) didn’t want life support. (But) I said, ‘No, this is imminently reversible. We can get past this.’ I resisted (the wife) strongly because (the patient) was getting better. A new (intensivist) came in on Monday, (the wife) gave the same drill all over again and they withdrew life support and (the patient) died that day. And he was going to survive with minimal to no impairment. But, (the wife) said stop, they stopped, and he died.” Another physician noted that when he recommended life support withdrawal, but the family wished to continue, he would readily try to persuade the family by “keep engaging (the family) in discussion for a long time—as long as it takes.” Another intensivist offered an alternative approach, believing that he was “sophisticated enough not to demand that the patient’s family withdraw.”

Discussion

We identified eleven factors that ICU providers perceived are important sources of physician variability in decisions to limit life support. While some of these factors have previously been described [10–13, 18–20], other factors such as workload and competing priorities, shift changes and handoffs, experiences of unexpected patient recovery, limiting life support in one’s own family, and firsthand knowledge of patient’s wishes have not previously been well described.

Physician level variability has both advantages and disadvantages in life support decision making. One advantage is the ability to adapt decision making to a diversity of countries, cultures, religions, resource utilization, and patient and family involvement [11, 14, 21–23]. Presenting a variety of provider viewpoints (such as a long term specialist who knows the patient well versus a short term intensivist) may improve decision making. But, if the conflicting messages are not coordinated, confusion and mistrust may ensue among patients and family members. Physician variability can also lead to differences in timing, frequency, prioritization, and quality of decisions to limit life support [11], which were perceived to be possible negative consequences of many factors identified in our study. Nurse participants provided a unique perspective regarding the impact of physician variability on patient and family-centered factors. Furthermore, both medical and surgical intensivists confirmed previously described findings that surgical investment may lead to an over emphasis on continuing life support [15–17, 24].

The environment in which a physician approaches life support decisions is an important source of physician variability. While some aspects of the decision environment are not easily modifiable (such as academic versus community practice setting, hospital culture, and ICU size), [10, 11, 21, 25–27] other factors such as handoffs, staffing models, and workload may represent opportunities for improvement [28–30]. Workload has been associated with numerous poor ICU outcomes [31–36], and we identified that increased workload may be associated with delayed decisions to limit life support because of competing priorities. Offloading the time consuming task of in-depth family communication may explain why the addition of palliative care consultation improves some end-of-life outcomes [37]. As has been documented with other clinical information [38, 39], key information regarding life support decisions may not be adequately communicated during handoffs. Including information regarding patient and family preferences during handoffs is the focus of quality improvement efforts in our institution. Physician experiences and training also contribute to physician variability [11, 13, 26]. Providing end-of-life care, especially family communication, is a skill that intensivists should master [40]. The variability in how and to what level physicians master this skill highlights the importance of teaching end-of-life care and life support decision making. We identified that exposure to long term patient outcomes after an ICU stay, as well as surprising outcomes influenced decisions. Just as physicians are unlikely to forget instances of missed diagnoses [41], experiences with unexpected patient survival seemed to be associated with high emotional attachment and vivid remembrance and were perceived to change providers’ attitudes and approaches.

Our study has several limitations. It was conducted in a single tertiary care academic institution in the Midwest United States, and thus without further exploration, the results may not generalize to other settings. Additionally, we explored providers’ perceptions and there may be a discrepancy between the providers’ perceptions, surrogates’ perceptions, and what actually happens. It is possible that participants misinterpreted their own experiences and the experiences and intentions of their colleagues. Moderate physician and nurse response rates of 60 and 29 % may have limited the number, variety, and depth of factors that were identified, as well as biased the results towards the viewpoint of providers who have an interest in end-of-life decision making. Additionally, most physician participants were male and all participants were Caucasian. While the overall number of participants was not large, we reached data saturation.

Our findings have several implications. First, providers should be aware that physician-level variability exists and which factors contribute to such variability. Second, further investigation is needed to examine ways to address the potentially adverse effects of factors such as workload, competing priorities, shift changes, and handoffs. Formally incorporating nursing input into decision making and handoffs should be prioritized. Third, efforts should continue to develop standards for approaching life support decisions so that intensivist effort can be measured and recognized [42]. Fourth, training should focus on providing the skill set and team leadership necessary to provide high quality decisions [43]. Incorporating exposure to long term outcomes of critically ill patients should be explored. Fifth, means to objectively assess, document, and perhaps display patient wishes should be further investigated, especially because firsthand knowledge in the ICU is so valuable.