Integrated palliative care: triggers for referral to palliative care in ICU patients

Palliative care within intensive care units (ICU) benefits decision-making, symptom control, and end-of-life care. It has been shown to reduce the length of ICU stay and the use of non-beneficial and unwanted life-sustaining therapies. However, it is often initiated late or not at all. There is increasing evidence to support screening ICU patients using palliative care referral criteria or “triggers”. The aim of the project was to assess the need for palliative care referral during ICU admission using “trigger” tools. Electronic record review of cancer patients who died in or within 30 days of discharge from oncology ICU, between 2016 and 2018. Patients referred to palliative care before or during ICU admission were identified. Three sets of palliative care referral “triggers” were applied: one that is being tested locally and two internationally derived tools. The proportion of patients who met any of these triggers during their final ICU admission was calculated. Records of 149 patients were reviewed: median age 65 (range 20–83). Most admissions (89%) were unplanned, with the most common diagnoses being haemato-oncology (31%) and gastrointestinal (16%) cancers. Most (73%) were unknown to palliative care pre-ICU admission; 44% were referred between admission and death. The median time from referral to death was 0 day (range 0–19). On ICU admission, 97–99% warranted referral to palliative care using locally and internationally derived triggers. All “trigger” tools identified a high proportion of patients who may have warranted a palliative care referral either before or during admission to ICU. The routine use of trigger tools could help streamline referral pathways and underpin the development of an effective consultative model of palliative care within the ICU setting to enhance decision-making about appropriate treatment and patient-centred care.


Background
In recent years, studies have shown that there has been an increasing number of cancer patients benefiting from intensive care support [1]. Initiation of a "trial of intensive care unit (ICU) therapy" in patients with advanced cancer is becoming more common [1]. It is estimated that 18-30% of cancer patients use intensive care services [2,3]. Mortality for cancer patients admitted to ICU is similar to general ICU patients at approximately 27-43% [3][4][5]. However, a proportion of cancer patients who survive their ICU stay die several months later from their underlying cancer [6]. Early identification and recognition of cancer patients in ICU who are either at risk of a poor outcome and/or have needs that might benefit from specialist palliative care involvement is key to facilitating prompt palliative care referral and review [7].
Early involvement of palliative care for patients with cancer has been proven in studies to improve patient experience, reduce symptom burden, support communication, and promote patient choice [8][9][10][11][12]. Furthermore, early palliative care has been shown to reduce the use of acute care services including inpatient hospital admissions and emergency department attendances [13][14][15]. Within the ICU setting, the delivery of palliative care has shown to benefit decisionmaking and symptom control, as well as reducing the length of ICU stay and the use of non-beneficial and unwanted lifesustaining therapies [7,13,16,17].
Prominent organisations such as the World Health Organization, the European Society for Medical Oncology [18], and the American Society of Clinical Oncology [19] recommend that palliative care should be delivered early on, alongside standard oncology care, thereby ensuring that patients with or at risk of unmet palliative care needs are identified proactively [20,21].
However, there are several potential barriers to palliative care referral [16,[22][23][24]. Inaccurate prognostication, underrecognition of dying, and a historical association between palliative care and end-of-life care can result in palliative care referrals often being initiated late in the course of a critical illness [13,14,25]. The lack of standardised referral criteria can also contribute to an inequitable access to palliative care services [22][23][24]26].
Identification of patients who may benefit from palliative care review in the ICU and the best time to refer them to the speciality has been the subject of several studies and reviews [27,28]. There is increasing evidence to support screening ICU patients using referral criteria or "triggers" [13,[28][29][30][31][32], as this may help to identify those patients who would benefit from a formal assessment of palliative care needs and offers a pragmatic approach to integrating palliative care in this setting.
The aims of this study were to: 1) Examine how many patients, who died on a cancerspecific ICU or within a month of ICU discharge, were referred to palliative care before death 2) Identify the proportion of patients who may have been referred to palliative care if a palliative care referral "trigger" tool had been used, either within 6 months before ICU admission or during ICU admission

Setting and cohort
This study was approved as a service evaluation by the Royal Marsden Committee for Clinical Research. This study was carried out in a 269-bedded specialist adult cancer hospital located across two sites in London. The hospital is a tertiary referral centre for patients with cancer from across the UK and abroad. There is a 16-bedded mixed medical and surgical ICU, which admits approximately 1,400 patients per year and has capacity for both levels 2 (single organ support, or extensive post-operative care) and 3 (two or more organ support, or advanced respiratory support) care.
The study included all patients who had died either in ICU or within 30 days of an ICU admission during a 2-year period, between 01 April 2016 and 31 March 2018. There were no exclusion criteria.

Study design
This was a two-part retrospective study with the first part relating to palliative care referral before ICU admission, the "pre-ICU admission analysis", and the second relating to palliative care referral during ICU admission, the "analysis of ICU admission". In both parts of the project, trigger tools were retrospectively applied to explore the potential impact of each individual tool on palliative care referrals.
A literature search was carried out using the keywords "trigger tools" and "intensive care" to identify palliative care referral trigger tools for inclusion in this study. As there is no evidence to date as to the most effective or useful trigger tool in this setting, a pragmatic approach was adopted to decide on the tools for inclusion in this study (Table 1). For the first part of the study, the "pre-ICU admission analysis", these included: • "The Royal Marsden (RM)" trigger tool: a palliative care referral "trigger" tool that was developed locally, which had been previously tested against acute oncology admissions [33] • The "Hui et al." trigger tool: a set of palliative care referral criteria for outpatient specialty palliative care which had been devised through a process of international Delphi consensus [34] For the second part of the study, the "analysis of ICU admission", three trigger tools were included: The trigger tools were retrospectively applied using the clinical data which would have been available at the time, as documented in the medical notes. As defined by the parent studies, patients were identified as being eligible for a palliative care referral if they were positive for any one or more of the palliative care referral "triggers". For the "pre-ICU admission analysis", patients were assessed as to whether they met the palliative care referral criteria at any time in Table 1 Trigger tools used in the study Palliative care referral "trigger" tool "Hui et al." trigger tool [34] "Royal Marsden" trigger tool (locally derived tool) [ the 6 months prior to the ICU admission. In the "analysis of ICU admission", patients who would have been eligible for a palliative care referral using a trigger tool after they had been admitted to ICU were identified. Patient records were examined using the electronic patient record (EPR), IntelliVue Clinical Information Portfolio (ICIP), and written notes. Data collection included patient demographics (age, gender, cancer diagnosis and stage), the reason for ICU admission [20], date, and cause of death. Referral to palliative care, reason for referral and the number of days between the ICU admission, earliest palliative care referral, and death were recorded.
Data were collected by two investigators. Pseudonymised data were entered into a database and were handled in accordance with Good Clinical Practice guidelines and General Data Protection Regulation.

Data analysis
Data were described using descriptive methods. Median and range were used to describe continuous non-parametric clinical data, with counts and percentages used for discrete variables.
A patient is considered to meet the requirements for a palliative care referral based on the "trigger" tool if he/she is positive for any one of the criteria within the individual "trigger" tool.
Sixty-two percent of patients (n = 92) were referred to palliative care before death. Of these 25% (n = 37) were referred to palliative care before ICU admission. The median time between the first palliative care referral and death for those referred before ICU admission was 38 days (range 0 to 1145). Reasons for referral to palliative care were categorised into six different subgroups-end-of-life care, symptom control, psychosocial support, team support, advance care planning, and other. Referrals to palliative care before ICU admission were mainly done for help with symptom control (81%, n = 30), psychosocial support (62%, n = 23), or in a minority of patients (11%, n = 4) for advance care planning purposes.
During ICU admission, 56% (n = 84) of patients were referred to palliative care. Out of these, 35% (n = 29) were already known to palliative care before admission to ICU. There were 8 patients who had previously been referred to and were known to palliative care but were not re-referred during their ICU admission. The main reason for referral to palliative care during ICU admission was for symptom control, with 75% (n = 63) of referrals stating this as one of the reasons for referral. The other two most common reasons for referral were psychosocial support, 70% (n = 59), and end-of-life care, 59% (n = 50). One-third (n = 28) of the referrals to palliative care included advance care planning as a reason for review.
The median number of days between ICU admission and referral was 7 (range 2-24). The median number of days between palliative care referral in ICU and death was 0 (range 0-19).
Out of the 92 patients who were referred to the palliative care team, 14% (n = 13) did not have a face-to-face review, but advice was given over telephone consultations. Nine of these patients (69%) had been referred on the day of death for symptom control and end-of-life care. The remaining four patients had also been referred for help with symptom control, and all but one of them had been referred the day before death.
Thirty-eight percent of patients (n = 57) were not referred to palliative care at all before death.
In the "pre-ICU admission analysis", 71% (n = 106) of the patients met at least one of the criteria for palliative care referral in the 6 months before ICU admission using the locally derived "Royal Marsden trigger tool". A smaller proportion of 59% (n = 88) of the patients would have warranted a palliative care referral using the "Hui et al." trigger tool.
The data from the "analysis of ICU admission" part of the study showed that a high percentage of patients would have met the criteria for a palliative care referral on admission to ICU. Using the "Royal Marsden trigger tool", 96% of the patients (N = 143) met at least one of the triggers, whereas 99% (N = 149) and 97% (N = 146) of the patients were positive using the "Zalenski et al." trigger tool and the "Hua et al." trigger tool, respectively. The proportions of patients meeting each of the criteria in the individual trigger tools are presented in Table 2.
The use of the ICU-specific "Zalenski et al." and "Hua et al." trigger tools would have also resulted in a palliative care referral being made much earlier before death. All patients who were positive for "Zalenski et al." and "Hua et al." trigger tools met at least one of the criteria on the day of their ICU admission. The median time between becoming positive for either trigger tool and death was 8 days (range 0-70).

Table 2
Retrospective application of trigger tools to the data, including full breakdown of the individual triggers

Discussion
In this retrospective study, most patients were not known to the palliative care team in the lead up to their ICU admission (13,17). Just over half of them were referred to palliative care at any stage, either in the months prior to their final ICU admission or during it. When patients were referred, it was generally when they were close to death. The use of palliative care referral "trigger" tools would have identified a high proportion of patients who may have benefitted from palliative care referral, consistent with findings from other studies [7,[29][30][31][32][33].
The results from this retrospective application of trigger tools suggest that there is a large proportion of patients who could have been identified for, and potentially benefitted from, a palliative care referral, both before ICU admission and during it. Admission to ICU is generally only considered for patients with potentially reversible clinical issues. With advances in treatments and improvements in outcomes, even cancer patients with advanced disease may be considered for a "trial of ICU therapy" [34]. ICU admission is, however, for most cancer patients a period of clinical and prognostic uncertainty. The aim of integrating palliative care at the point of ICU admission is to facilitate parallel planning, considering all outcomes and ensuring that clinical decision-making is informed by patient priorities.
Involvement of palliative care at an earlier stage, before patients even get to ICU, may add even greater benefit. In this study, depending on the trigger tool used, between 59 and 71% of patients may have been eligible to be seen by palliative care before their ICU admission. These patients may have benefitted from advance care planning and shared decision-making which may have impacted the decision for ICU [13,17]. There is good quality evidence from well-designed studies that palliative care should be available to patients throughout their cancer journey and that it is no longer just for patients at the end of their lives. Health care organisations, patients, and staff are challenged to embrace the culture shift that enables palliative care to be provided from as early as diagnosis, alongside active medical treatment of their cancer and intercurrent illnesses [18,19].
Not every dying patient needs specialist palliative care, and likewise palliative care is no longer reserved just for patients in the last hours and days of life. The use of palliative care "trigger tools" to identify patients who may benefit from a comprehensive palliative care needs assessment, and input represents a move away from a traditional referral model based mainly on subjective assessment of prognosis to one which is more standardised and centred around the individual needs of patients [35]. The "trigger tool" approach can underpin a "consultative model" of palliative care service provision in ICU whereby there is increased involvement and effectiveness of the palliative care team in ICU [27]. Used routinely, it can also support a triage system whereby a generalist palliative care is provided to all relevant patients by the ICU team with specialist involvement for those likely to have the most complex needs [7,16].
Additionally, the data from this study suggest that a large proportion of cancer patients may warrant palliative care reviews many months before becoming acutely unwell and requiring admission to ICU. A similar retrospective study using the RM trigger tool demonstrated that a high percentage of cancer patients admitted acutely would have been eligible for a palliative care assessment earlier on in their disease trajectory [33]. The findings from this study show that the use of trigger tools during ICU admission could result in more than double the number of patients that were referred to palliative care.
Whilst all the trigger tools picked up higher numbers of patients to refer to palliative care, they also identified patients for a referral at an earlier stage. All the patients that were identified by the three trigger tools during ICU admission were positive from the day of ICU admission. Thus, patients would have been seen earlier, compared to actual practice where most patients were referred to palliative care on the day of death. This would have allowed more time for patients and their families to benefit from a palliative care review and could have had a positive effect on patient care [13,14,17].
The selected trigger tools for this study were found through extensive literature searches, as described above. The tools selected not only appeared to be most relevant to the study cohort but were also supported by the most robust evidence [7,16,32,33,36]. The "Hua et al." and "Zalenski et al." trigger tools were studied in general ICU populations and hence have an evidence base in the group. However, as this study was carried out in a specialist adult cancer hospital, the findings are not necessarily generalisable to non-cancer settings. The trigger of "advanced or metastatic cancer"-and some of the other triggers toomay apply differently in this specific cancer population. However, even if the trigger of advanced cancer is disregarded, a significant proportion of patients were positive for a palliative care referral before ICU admission based on other triggers, e.g. physical symptoms. Similarly in the ICU, a high proportion of patients met the referral criteria based on symptom needs; organ failure; having a marker of advancing illness such as anorexia, hypercalcaemia, or any effusion; or because the oncology team wanted palliative care to be involved.
There are other triggers that appeared less relevant in this study, due to the nature of the study population. For example, no patients were positive for "admitted from a nursing facility".
Future research is needed to refine palliative care referral criteria that are specific for a cancer population. The ICUspecific tools ("Hua et al." tool and "Zalenski et al." tool) identified a similar proportion of patients for referral as the non-ICU specific Royal Marsden tool which has been previously tested in the acute oncology setting. This may suggest that it is the use of a targeted approach to the identification of patients for referral that matters most, rather than differences between the actual items included in the tools.
In this study it was not possible to identify the needs of patients who were not referred to the palliative care team. It was also not possible to robustly assess the effectiveness of involvement of the palliative care team in addressing the needs of patients. Incorporating palliative care need assessments and outcome measurements in future research in this area would increase the clinical applicability and impact of this research and provide objective evidence of the severity, breadth, and complexity of patients' palliative care needs and the impact of interventions in the ICU [37].
One of the main limitations of this study was that the data were collected retrospectively; therefore, the results are very reliant on the accuracy, completeness, and quality of documentation. It was not feasible to carry out a comparison of all palliative care referral or trigger tools in this study, and there may be other tools that exist that have not been included here. Also, regarding the trigger tools themselves, although most of the "triggers" were objective, there were some subjective "triggers". Therefore, the results may include some bias on part of the data collector and may have varied if purely objective "triggers" had been used, such as "team perceived need for palliative care". For example, one of the triggers included in both the "Hua et al." and "Zalenski et al." tools used in the "analysis of ICU admission" was around having advanced/metastatic cancer. Although this is clear cut in solid organ tumours, it is more difficult to define in haematological malignancies so was down to the authors' discretion on a case-by-case basis. It could be also argued that the findings are limited to the results of a single, tertiary centre. However, the authors believe that the findings have wider applicability and add to the growing evidence which supports the use of palliative care referral trigger tools.

Conclusion
This study has demonstrated that the use of specific sets of "trigger" tools may help to highlight patients with cancer who might benefit from a referral to palliative care. The use of a trigger tool would identify most patients during their ICU admission and many patients in the preceding 6 months prior. This supports the shift in the perception of palliative care; so it is not only considered at the end of life but is being considered much earlier in patients' disease course. The use of these trigger tools and early palliative care referral may streamline referral pathways and help with decision-making about appropriate treatment and patient-centred care.
These findings lend support to the plausibility of using trigger tools to deliver palliative care to critically ill cancer patients in clinical practice. Although the results are from a small sample size in a single tertiary centre, there is a clear need for the validation of such trigger tools in general and cancer-specific populations.