Quality of transition to end-of-life care for cancer patients in the intensive care unit

Setting

This study was approved by the local Committee for Clinical Research. The study took place on the ICU in a 269-bedded specialist adult cancer hospital located across two sites in London. The hospital serves the local population and is also a tertiary referral centre for patients with cancer from across the United Kingdom and abroad. The 16-bedded mixed medical and surgical ICU admits approximately 1,400 patients per year and has capacity for both level 2 (single organ support, or extensively post-operative care) and 3 (two or more organ support, or advanced respiratory support) care [26]. All referrals and admissions to ICU are discussed with the ICU consultant. No formal EOLC protocol was in place in the ICU during the study period following the abolition of the Liverpool Care Pathway [27]. However, several of the recommendations for improving communication at the end-of-life by the American College of Critical Care Medicine had been adopted in the unit [15]. These included formal communication skills training of ICU and Palliative Care Consultants, interdisciplinary team rounds, palliative care team input at weekly ICU multidisciplinary team meetings and regular family meetings.

All medical patients with cancer admitted to the ICU over a 6-month period from 01.03.2013 to 01.08.2013 were included in this study. Patients <16 years, without a cancer diagnosis, or who had undergone elective or emergency surgery were excluded as the focus on this study was primarily on medical cancer patients. For those patients who had more than one admission during the study period, only the patient’s last admission was included in the analysis. Sepsis was defined according to the Surviving Sepsis Guidelines [28].

Study design

A retrospective review of patient records was carried out to explore patients’ demographics and co-morbidities; cancer type and stage; reason for hospital and ICU admission; documented communication between teams, patients and relatives (detailed from care records, team meetings and daily ward round annotations in the electronic patient records); documented palliative care team involvement; documented evidence of resuscitation and ceiling of care decisions; documented quality of EOLC provision, and ICU mortality.

To determine the care process and quality of EOLC transitions on the ICU, quality indicators recommended by American College of Critical Care Medicine were used in the qualitative documentary analysis (Table 1) [15]. The quantitative variables were categorised into process and outcome domains. The first five points (Domains 1–5) were used in the analysis on processes of EOL transition and provision of EOLC in our ICU. Communication not only included documented presence of patient, family and team members, but also in some cases, a qualitative analysis of the documented discussion. At the time of this study, standardised symptom assessment tools were not routinely used or integrated into the electronic documentation. Symptom control (Domain 5) was defined, for simplicity, as documented evidence of pharmacological and non-pharmacological control of symptoms such as pain, nausea, vomiting, diarrhoea, and anxiety. Assessments were made prior and after intervention to determine if patients’ had achieved adequate symptom control. Domains 6 and 7 were excluded as they relate to organisational processes that could not be determined through patient records review. Palliative or non-curative treatment intent was defined as patients with no further active treatment options available, but those that could be on phase 1 trials or be receiving treatment to control or maintain disease stability but without curative intent. In contrast, curative intent was defined as patients who were receiving active treatment for their cancer with an intention of cure [29].

Patient records were examined using the Electronic Patient Record (EPR), IntelliView Clinical Information Portfolio (ICIP), chaplaincy database, and written notes. All data was double checked for accuracy by two investigators. All data were entered onto an excel database and collated into 46 variables under the aforementioned process and outcome categories. These variables were then categorised and tabulated into the five domains (Domains 1–5). Data were pseudo-anonymised, and dealt with Good Clinical Practice guidelines and United Kingdom Data Protection Act.

Identification of cohorts: ‘EOLC’ vs ‘full active ICU care’

Following data collection, patients were divided into two cohorts: patients who received full active management with organ support throughout their ICU stay and whose main focus of care remained prolongation of survival with curative intent (‘full active ICU care’ cohort); and those patients who had transitioned from full active management to EOLC during their ICU stay or died on the ICU (‘EOLC’ cohort) (Fig. 1). The transition to EOLC was ascertained on an individual patient basis and decisions were made when the teams (parent medical team, ICU team and sometimes the palliative care team) jointly concluded that continuation of life-prolonging organ support would be futile and not in the patients’ best interests. Families (and patients, if they had capacity) were involved in these discussions. It is recognised however, that some of EOLC cohort may have received life-prolonging organ support during the early part of their ICU stay before transitioning to EOL care. Equally, symptom control was given in the full active care cohort where needed.

Data analysis

Quantitative data were analysed using SPSS statistics software (version 20, Chicago, IL, USA). Comparison was made between the two cohorts using standard descriptive statistics. For categorical variables such as organ support required in ICU and performance status, Fisher’s exact test and Chi squared tests were performed to compare data between the two groups. Continuous data was analysed using the Mann–Whitney U (for non-parametric data including age) and student t test for parametric data. A statistical significance p < 0.05 was used when comparing the two cohorts.

Multiple logistic regression was used to investigate the joint effect of clinical variables on the whether or not patients transitioned to the EOLC cohort or not. Factors with p < 0.1 on univariate analysis were included in the multivariate modelling. Variables with p > 0.1 were excluded in order to reduce the number of predictor variables. Variable selection was carried out using a stepwise method, with p < 0.05 and p > 0.1 as criteria for entry and removal, respectively. Only factors with p < 0.05 were retained in the final model.

Data from the EOLC cohort were further analysed using descriptive statistics to explore the quality of EOLC being delivered on the ICU.

Text data from all medical records (ICIP and hospital EPR) was analysed using qualitative thematic analysis. Undertaking textual analysis enriched the quantitative data yielded [30, 31]. A mixed analysis allowed examination and exploration of unusual quantitative findings [32]. Thematic analysis provided a structure for identifying patterns in data and involved steps of data codification; testing and connecting codes, searching and establishing themes; reviewing and corroborating themes.

Quantitative analysis comparing ‘Full active ICU care’ with ‘EOLC’ cohorts

Overall ICU mortality was 34% for all medical patients admitted to the ICU and 3.7% (118/3,203 patients) for all hospital in-patients during the study period.

Quantitative analysis of the quality of transition to EOL

Of the 38 patients who were included in the EOLC cohort, only one patient was admitted for symptom control alone. All other patients were initially admitted for full active ICU management.

Most patients in the EOLC cohort (37/38 patients; 97.4%) had clear documented evidence of transition from full active care to EOLC.

ICU and in-hospital mortality

Most patients (29/38 patients; 76.3%) in the EOLC cohort died on the ICU. Of the remaining 9 patients, 7 died on the ward; one died in hospice and one was transferred to their home country whilst remaining intubated. In-hospital mortality for the EOLC cohort was (36/38 patients; 94.7%).

Resuscitation decisions

On admission to ICU most patients in the EOLC cohort (34/38 patients; 89.5%) were for full resuscitation. In the EOLC cohort ‘do not attempt cardiopulmonary resuscitation’ (DNACPR) orders were completed for 47% patients within 48 h of ICU admission. Most patients (97.4%) in the EOLC cohort had documented evidence of DNACPR at some point during their ICU stay. Sixty-nine percent of patients died within 3 days of DNACPR decision. The time from DNACPR decision to death was a median of 1.5 days (range 0–93 days). In contrast, amongst the ‘full active ICU care’ cohort, just 2/45 patients (4.4%), had a DNACPR order documented during their ICU stay.

Documented quality markers for transition to EOLC

Qualitative thematic analysis

Data were analysed qualitatively for all patients under a priori framework analysis categories of prognostication, decision-making and EOLC based on the quantitative findings. These were modified according to the 19 sub-themes (Fig. 2) and resulted in five main themes (Table 5).

1. Achieving a consensus to initiate end of life transitions

1.1 ‘I have discussed his management with [oncologist] this morning. He agrees that it would not be appropriate to start [mechanical ventilation] [. . .] It is unlikely he will survive beyond the next few hours and even with full invasive ventilation it is unlikely he will survive beyond the next few days. We have therefore moved to palliation. [Wife] is in agreement that it would not be appropriate to intubate and ventilate him. (patient 79)

1.2 ‘Prof [oncologist] reiterated our conversation we had with [patient] and [husband] this morning about the change from curative to palliative intent. Then we stressed the importance of symptom control, how we were going to involve the palliative care team, and how we would ensure reducing anxiety, distress and any discomfort…’ (patient 63)

1.3 ‘[patient] agrees to intubation [. . .] but does not want CPR or any heroic measures to prolong her life.’ (patient 59)

1.4 I talked with [oncologist] yesterday who is keen that we pursue every therapeutic option for [patient] at present and I discussed my reluctance to consider intubation for this gentleman because none of our intensive care treatments were improving his respiratory function [. . .] [Palliative care consultant] talked with the family and indeed I talked to [patient] yesterday who was adamant that he didn’t want further escalation. (patient 16)

2. Concomitant prognostication and managing uncertainty

2.1 ‘Explained to [husband] that [patient] is gravely unwell at present, and that she may not survive this episode. We have explained to [husband] that we will do everything we can in CCU, and will have more of an idea within the next 24 to 48 hours of how her condition will progress. If she deteriorates further, then we will do everything we can to keep her comfortable… (patient 81)’

2.2 ‘Unable to wean off FiO2 [oxygen via oscillator mode on ventilator] and respiratory demands continue to increase. [patients]’ condition continued to deteriorate and discussions were had between the family, oncology and CCU team.’ (patient 83)

2.3 Even were his GVHD to resolve imminently he would still be in multi-organ failure[. . .] (patient 47)

3. Parent medical and critical care team decision-making

3.1 ‘in view of limited cancer prognosis and, in event of requiring emergency sedation, intubation and ventilation that recovery to the point of acceptable quality of life would be unlikely’ (patient 40)

3.2 ‘After discussion with [oncologist] and the family we decided not for further active management but continued high quality supportive care’ (patient 84)

4. Integrative palliative care

4.1 ‘we stressed the importance of symptom control, how we were going to involve the palliative care team, and how we would ensure reducing anxiety, distress and any discomfort. (patient 17)

4.2 ‘His continuing deterioration is indicative of end of life events and we discussed his management with the palliative care team (patient 15).

5. Family-centred versus patient-centred care

5.1 ‘Consider discontinuation of NIV at an interval after son’s arrival on Monday. In the meantime—if [patient] wishes us to discontinue the NIV earlier, this would be appropriate—with sensitive communication to the family’ (patient 27)

5.2 ‘They [family] understand that it is highly unlikely that Mrs D would be able to be weaned from invasive ventilation and agree as to the previous set ceiling of care of NIV [non-invasive ventilation]. They understand she has developed multi-organ failure and that our priority of care now would be solely her comfort.’ (patient 30)

5.3 ‘Family did not want active treatment if no hope of recovery’ (patient 85)

5.4 ‘Wife does not want us to try to communicate this to him now but would rather she did that herself if he became more alert later on.’ (patient 80)

5.5 ‘We will leave her treatment as it as at present instead of withdrawal as her son is in transit from USA.’ (patient 30)

When initiating end-of-life transitions, where patients had capacity, decisions were made collaboratively with families and teams. When supportive care was deemed appropriate, rather than focusing on active treatment, palliative care input was quickly sought. The excerpt (1.1) exemplifies consensus building as well as highlighting the consensus between teams in the next theme.

Parent medical and critical care team decision-making referred to how critical care teams sought early joint decision-making with parent medical teams when there was doubt over the possibility of critical illness recovery (theme 3). This planning meant that some patients were transferred out of the unit and received EOLC on the ward.

Poor critical illness related outcomes and cancer prognosis made it straightforward to reach consensus and rationalise decisions. Prognosis was communicated where possible (2.1) and where there was uncertainty regarding outcome from the critical illness perspective, it was more likely that patients died with full intensive care treatments in situ (2.2, 2.3).

Furthermore, teams negotiated to try critical care treatment for a period of time, allowing opportunity for anti-cancer treatments to work, and allowing for a trial of ICU to reverse acute problems. If families were reluctant for EOLC transitions and wished for full active treatment to continue, the intensivists would often continue with the caveat that if there was no improvement in a specified time, they would initiate a transition to end-of-life (2.1).

If families (and patients, when able) were receptive to EOLC transition, then the critical care team acted swiftly to focus on EOLC, either placing a ceiling of care (limitation of treatment) (5.2), or withdrawing treatment (5.1, 5.3).

Integrative palliative care referred to how palliative care input was sought on the majority of patients who actually died in critical care, despite the short time to death from withdrawal or limitation of treatment (theme 4). This facilitated the focus on comfort measures, and away from high-intensity treatment at end-of-life, but depended on timely referral. Some intensivists advocated an integrative approach where palliative care became involved early, to assist in advanced care planning.