End-of-life perceptions among physicians in intensive care units managed by anesthesiologists in Germany: a survey about structure, current implementation and deficits

Background Structural aspects and current practice about end-of-life (EOL) decisions in German intensive care units (ICUs) managed by anesthesiologists are unknown. A survey among intensive care anesthesiologists has been conducted to explore current practice, barriers and opinions on EOL decisions in ICU. Methods In November 2015, all members of the German Society of Anesthesiology and Intensive Care Medicine (DGAI) and the Association of German Anesthesiologists (BDA) were asked to participate in an online survey to rate the presence or absence and the importance of 50 items. Answers were grouped into three categories considering implementation and relevance: Category 1 reflects high implementation and high relevance, Category 2 low and low, and Category 3 low and high. Results Five-hundred and forty-one anesthesiologists responded. Only four items reached ≥90% agreement as being performed “yes, always” or “mostly”, and 29 items were rated “very” or “more important”. A profound discrepancy between current practice and attributed importance was revealed. Twenty-eight items attributed to Category 1, six to Category 2 and sixteen to Category 3. Items characterizing the most urgent need for improvement (Category 3) referred to patient outcome data, preparation of health care directives and interdisciplinary discussion, standard operating procedures, implementation of practical instructions and inclusion of nursing staff and families in the process. Conclusion The present survey affirms an urgent need for improvement in EOL practice in German ICUs focusing on advanced care planning, distinct aspects of changing goals of care, implementation of standard operating procedures, continuing education and reporting of outcome data. Electronic supplementary material The online version of this article (doi:10.1186/s12871-017-0384-5) contains supplementary material, which is available to authorized users.


Background
The question of end-of-life (EOL) care has recently reached modern intensive care medicine. Several aspects need to be considered, such as differences in ethnicity, country and medical specialty. An updated summary of published statements on EOL care in the ICU from national societies has currently been presented, highlighting commonalities and differences within and between international regions [1]. Therein, the complexity of EOL care in the ICU within different ethical and cultural environments, particularly relating to withholding and withdrawing life-sustaining treatment while ensuring the alleviation of suffering, has been recognized. Subsequently, the World Federation of Societies of Intensive and Critical Care Medicine has encouraged their member societies to lead the debate and to develop national guidelines and recommendations regarding EOL care in the ICU within each country [1].
At present, data are sparse on the factors associated with EOL decisions in Germany [2]. In Germany, the majority of surgical and interdisciplinary ICUs are managed by anesthesiologists who are also in charge of patient care. Since the status of structural aspects and the relevance of EOL in German ICUs are unknown, we aimed to explore current practice, barriers and opinions on EOL decisions. Due to the increasing importance of EOL-decisions on ICUs and paucity of data, the working group "Epidemiology and Ethics", a scientific working group with emphasis on ethics within the Scientific Working Group Intensive Care Medicine (WAKI) of the German Society of Anesthesiology and Intensive Care Medicine (DGAI), has been supported by the DGAI who addressed EOL in ICU as one of their strategic goals in 2013. The hypotheses of the present survey were: 1.) EOL standard operating procedures (SOPs) are not regularly used in ICU's and 2.) There is a discrepancy between the actual (implementation of items) and the desired status (importance of items) on EOL decisions.

Methods
After the working group had recommended conducting the survey, the methods were discussed in meetings, telephone conferences and emails by the members of the DGAI working group "Epidemiology and Ethics". The literature about EOL in ICU was researched and with help from an epidemiologist this survey was developed for those items identified as most relevant. Questions were phrased according to publications regarding EOL care and guidelines, such as the Surviving Sepsis Campaign (SSC) 2012 guidelines [3]. After endorsement by the DGAI, 97 questions were reduced to 59 items in order to limit the questionnaire time to 15 min. In the qualitative pilot test, all members of the working group considered the questions relevant, all agreed with the questions, and none of the participants found areas lacking. In November 2015, all members of the DGAI and the Association of German Anesthesiologists (BDA) were invited by email (n = 17,044) to participate in the online survey via "www.Umfrageonline.com". Only one reminder was sent out before closing on December 23rd, 2015.

Design
The survey included nine items about the structure of the ICU and 50 items about prognostic scores, reporting of individual patient outcomes, goals of care, patient autonomy, standard operating procedures, quality management, limitations of life-sustaining therapy, nursing aspects and concepts of care for dying patients (Table 1). For each item, participants were asked to state current practice, desired status and importance. Current practice was rated on a modified four-point Likert scale as "yes, always", "often", "sometimes" or "no, never". If "sometimes" or "no, never" was marked, participants were asked to rate whether they considered this as a deficiency, and thus, the desired status. Relevance of an item for the respondents in the implementation domains "sometimes" or "no, never" was estimated as: relevance = "deficiency yes" divided by "deficiency yes and no". Importance was ranked on a four-point Likert-scale ranging from "not important"to "very important". To reduce complexity and for a clearer presentation, we focused on three subgroups regarding the implementation domain: "always/ often", "sometimes" and "never", and on two subgroups regarding the importance: "important" and "not important".
Considering implementation, relevance and resulting implication, we focused on three subgroups:

Results
Eight-hundred twenty-one anesthesiologists (4.8% of 17,044 DGAI or BDA member email adresses) responded. The mailing list classified 870 heads of anesthesiology departments. There are no data available how many of these departments run ICUs. After the first email reminder, there was only minimal increase in responses. Thus, no further reminders were sent out. Only completed surveys (n = 541) were included in the analysis. Out of the 541 responders, 417 stated the name of their department, resulting in 305 reponders from different departments. Thus, the response rate reflects more than 1/3 of anesthesiology departments in Germany. Almost all questions were answered by the participants, thus, demonstrating the ability of the questionnaire to distinguish between different respondents.

Structural items
There was a higher likelihood of physicians working in university hospitals to participate in the survey: 19% of respondents, but only 8% of the ICU's involved were from university hospitals. Thirty-two percent of respondents worked in maximal care hospitals (level 1), 29% in priority care hospitals (level 2), and 39% in basic and regular care hospitals (level 3). Predominantly, experienced physicians participated in  this inquiry: 12% were department heads, 8% ICU directors, 32% anesthesia and intensive care specialists and 16% residents. The participants had experience in intensive care medicine less than 5 years in 26%, between 5 and 10 years in 26% and more than 10 years in 48%.
Regarding the institutions the respondents worked at, 53% were public or community hospitals, 17% were privately owned, 24% pertained to churches and 6% to common welfare organizations. The characteristics of the involved ICU's are presented in Table 2. Since multiple responses were possible regarding treated patients, the sum of treated patients exceeds 100%.
Mean percentages of responses in the three categories are presented in Table 3 regarding implementation (always / often), importance (important), implementation discrepancy (difference between percentage values regarding "importance" and "status of implementation") and relevance ("deficiency yes" divided by "deficiency yes and no" in the implementation domains "sometimes" or "no, never"). The highest level of importance was found for sufficient Category 1 items which lacked implementation in ¼ of respondents. Unsatisfactory Category 3 items were ranked less important than sufficient Category 1 items but still reached high importance levels. There was a substantial implementation discrepancy characterizing those items with the greatest need for a change in practice, i.e., unsatisfactory Category 3 items. The relevance was profoundly lower than the importance in all categories.

Discussion
The present survey revealed a discrepancy between current practice of EOL and perceived importance and lack of particular feedback, education, and tools. This was especially important for outcome data and advanced care planning. For the first time, in greater than 1/3 of anesthesiology departments running ICUs in Germany, a survey revealed a valuable insight in current practice, First of all, we wanted to know whether prognostic scores play a role in EOL decisions in German ICUs. SAPS II is regularly recorded on a daily basis in Germany for calculation of diagnosis related groups (DRG's) and reimbursement. Therefore, we expected that SAPS II might be used for estimation of prognosis, also, which might have influence on EOL decision Fig. 1 EOL items Q8-11 of high implementation and high relevance (sufficient Category 1). Data are presented as "blob-o-grams" were the number of respondents in each category is represented by a circle whose area is proportional to the number. Importance (x-axis) and status of implementation (y-axis) are rated on modified Likert scales. Q = Question. C1 = sufficient Category 1 Percentage of responses in regards to: status of implementation (always / often), importance (important), implementation discrepancy (difference between percentage values regarding "importance" and "status of implementation always / often") and relevance ("deficiency yes" divided by "deficiency yes and no" in the implementation domains "sometimes" or "no, never") making. The results of the present survey do not corroborate this hypothesis. Although SAPS II [4] (Simplified Acute Physiology Score) or SOFA [5] (Sequential Organ Failure Assessment) scores were originally developed for estimation of prognosis, they were hardly used for this purpose in the present survey and categorized as inessential Category 2 (Additional file 7: Figure S7). The prognostic performance of SAPS II is poor [6], better with SAPS 3 [7] and SOFA [5], all predict outcome for a given subgroup of patients, but fail for a single patient. Due to limited ICU resources, physicians seek prediction tools to facilitate allocation of ICU beds to patients which might benefit best. In this context, an initial refusal and final triage score provided objective data for rejecting patients that will die even if admitted to the ICU and survive if refused [8]. The mortality benefit regarding ICU admission appeared greater for the elderly [9].
In absence of adequate prediction models of outcome, the majority of respondents judged feedback on outcome data as very important for individual decision making (Q 4-7, attributed to unsatisfactory Category 3, Fig. 3). A considerable variability between hospitals and physicians in terms of EOL care in ICUs is due, in large part, to the lack of compelling evidence or professional consensus for specific approaches that ensure patients of receiving the care they would want if fully informed about their prognosis and likely outcomes [10]. EOL decisions might be improved by establishing interdisciplinary rounds, advanced health care planning, and structured feedback on outcome data.
Questions Q8-18 regarding curative vs. palliative goals of care were answered to be in the sufficient Category 1 (Fig. 1, Additional files 1 and 2: Figures S1 and S2). However, the answers to these goals of care bear a high Fig. 2 EOL items Q44-46 of low implementation and low relevance (inessential Category 2). Data are presented as "blob-o-grams" were the number of respondents in each category is represented by a circle whose area is proportional to the number. Importance (x-axis) and status of implementation (y-axis) are rated on modified Likert scales. Q = Question. C2 = inessential Category 2 risk of societally acceptable responses. The issue of "setting goals of care"(Q9 and Q10) has been addressed by the Surviving Sepsis Campaign (SSC) 2012 guidelines [3] and reiterated in the 2016 guidelines [11]. Initiatives to enhance care in the ICU highlight the importance of incorporating goals of care together with the prognosis into treatment plans [12]. It has been reported that less than half of the ICU physicians felt comfortable in having EOL discussions with patients' families [13]. Therefore, even well-intended clinicians may miss valuable opportunities to address and clarify families' misunderstandings and concerns regarding goals of care at EOL [14]. Moreover, while the estimations of risks and prognosis may change during hospital stay, patients and families often are unable to move beyond the very first prognostic statements. We emphasize that these skills regarding curative vs. palliative goals of care should be regarded as essential and trained accordingly.
The participants identified SOPs dealing with psychosocial (Q27) and spiritual (Q28) problems as an issue of unsatisfactory Category 3 (Additional file 9: Figure S9). In Germany, an approach to these aspects has been published [15], specifying ethical principles, legal basics, patient autonomy, decision making and implementation regarding limitation of care and change in goals of care, unity of patient and family, cultural and religious influences as well as conflicts and burnout in the caregiving team. In Canada, a series of guidelines address withdrawal, distress and discontinuation [16]. There is a critical need to reframe EOL care planning, not prioritizing life extension over good death [17]. Thus, tools are already available to enable a shift from unsatisfactory Category 3 to sufficient Category 1.
Another deficiency identified was the lack of support regarding "changes in goals of care"(Q34, 35, 37), RID Fig. 3 EOL items Q4 -Q7 of high importance that are rarely implemented but are considered to be highly relevant (unsatisfactory Category 3). Data are presented as "blob-o-grams" were the number of respondents in each category is represented by a circle whose area is proportional to the number. Importance (x-axis) and status of implementation (y-axis) are rated on modified Likert scales. Q = Question. C3 = unsatisfactory Category 3 (Re-evaluate Indication and De-escalate) and CTC (Comfort Terminal Care) (Additional files 9 and 10: Figures S9 and S10). Checklists may be helpful, and guidelines for changes in goals of care [18] utilizing DNR (Do Not Resuscitate), DNE (Do Not Escalate), RID and CTC have recently been published, reevaluating, documenting and changing on demand goals of care on a daily basis. However, although acuity of illness and organ dysfunction consistently predicted mortality in critically ill patient populations, only elements of the past medical history were positively associated with a DNR order [19]. The WELPICUS study has achieved world-wide consensus on key EOL issues and terminology [20]. However, EOL decisions are very variable between regions, countries, individual ICU' s and individual clinicians in the same ICU [21,22], often driven by the views of individual physicians and hospital norms [10]. Instead of "no escalation of treatment", a "time-limited trial" of life support has been advocated [23]. In practice, withholding preceded or accompanied withdrawal in >90% of patients [22], and was more likely to occur during on-call hours [24]. It is noteworthy that withholding and withdrawal reflect the limitation of life sustaining treatments, but it is essential in the communication with the team, the patient and the relatives that a change in goals of care does not mean the cessation of medical care [25].
Another area of need in the present survey was EOL education (Q40, Additional file 10: Figure S10). The World Health Organization defines palliative care as 'an approach that improves the quality of life of patients and their families' [26]. In everyday practice, adoption of the ' ABCDs' of EOL critical care is applicable: Attitudes, Behaviours, Compassion and Dialogue [27]. Presently, between 10% and 20% of the population at large now die in the ICU underlining the importance of EOL care to everyday practice and training [25], being extended to EOL orders (Q41) for continuing care after death for relatives (Q43) (Additional file 10: Figure S10). Many clinicians and families equate withholding or withdrawing as giving up [17]. Communication and intervention withdrawal practice guidelines that highlight EOL care as part of, rather than separate from, critical care and education [28] are available and may be crucial in supporting ICU teams to help make good death more accessible [17].
Unfortunately, only a minority of DGAI and BDA members participated in the survey. Thus, the results are not representative. Moreover, due to the voluntary participation, it must be appropriately considered that the most unsatisfied colleagues may have responded willingly, or just the most engaged in EOL. Many anesthesiologists do not work in ICUs and hospitals. Thus, we cannot state how many respondents would be eligible and representative for an EOL survey. It remains unclear, whether EOL is regarded not important or not as a problem. Also, due to the voluntary participation, the distribution of health care providers of respondents was not a representative selection. In 2015, in Germany, 38% of hospitals running ICUs were allocated to public or community hospitals (vs. 53% in present survey), 21% (vs. 17%) to privately owned, and 41% (vs. 30%) to churches and to common welfare organizations [29]. 1177 out of 1956 hospitals stated ICU beds, and 416 intensive care medicine departments [29]. No representative data were available regarding the percentage of hospitals and ICUs in Germany treating ≤ 500 up to >2000 patients / year as given in Table 2. To our surprise, some colleagues reported that they were not permitted to participate by their hospital chief executive officers (CEO's) for the fear of disclosing sensitive information. Physicians informed their CEO's and data security engineers, and some were not allowed to participate due to safety concerns with the online survey via "www.Umfrageonline.com", because this evaluation might not be anonymised as confirmed. This barrier to scientific investigations driven by economic competition is worrisome and could likely increase in the future.
The present survey was only addressed to anesthesiologists, and therefore, the results cannot be generalized to other specialties, professions, or persons affected by EOL decisions, such as nurses, palliative care experts, or families. It is appreciated that surveys regarding EOL care might show different results for different participating subpopulations [10,21,22].

Conclusions
The present survey reveals an urgent need for improvement in EOL practice in German ICUs. Improvement might be achieved by focusing on desirable quality of life, advanced care planning, continuing EOL education and feedback on outcome data. A shift from unsatisfactory Category 3 to sufficient Category 1 may be enabled by generating awareness regarding deficits in EOL care and deliver already available tools via specialist societies, such as the German Society of Anesthesiology and Intensive Care Medicine (DGAI). To improve EOL care in ICUs run by anesthesiologists in Germany, therapeutic indications have to be clean-cut, followed by decision making and implementation by the main players, the physicians and nurses, the patients, their legal representatives and families.

Additional files
Additional file 1: Figure S1. EOL items Q12-15 of high implementation and high relevance (sufficient Category 1). Data are presented as "blob-ograms" were the number of respondents in each category is represented by a circle whose area is proportional to the number. Importance (x-axis) and status of implementation (y-axis) are rated on modified Likert scales. Q = Question. C1 = sufficient Category 1. (JPEG 1380 kb) Additional file 2: Figure S2. EOL items Q16-19 of sufficient Category 1. Data are presented as "blob-o-grams" were the number of respondents in each category is represented by a circle whose area is proportional to the number. Importance (x-axis) and status of implementation (y-axis) are