2. Materials and Methods
Institutional ethics approval was granted prior to the study. Extracted data were pseudonymised and securely stored on a password-protected system in accordance with General Data Protection regulation.
2.1. Setting
This study was conducted at a tertiary care institution in London and data were gathered from two general ICU wards, which comprised a total of 45 critical care beds, providing both level 2 (single-organ support/extensive post-op care) and level 3 (advanced respiratory or multi-organ support) care.
2.2. Cohort
All ICU deaths between 19 March and 30 October 2019 were included, with no exclusion criteria. These dates align with the introduction of the Electronic Healthcare Record (EHR) in March 2019, facilitating accurate data retrieval. October was selected as the endpoint to minimise the effects of the SARS-CoV-2 pandemic on ICU admissions.
2.3. Study Design
The hospital EHR was used to identify all ICU deaths occurring in this period. Data collected included demographics, reason for ICU admission, length of ICU stay, whether a PC referral was made and the timing of referral relative to death. For those patients not referred to PC, three “trigger tools” were retrospectively and separately applied at the point of ICU admission:
Hua et al. [
16]—a five-point tool validated in ~400,000 ICU patients (USA).
Zalenski et al. [
17]—a five-point system from ~600 ICU patients (USA).
Royal Marsden [RM] tool [
18]—a seven-point consensus-based tool designed for oncology admissions (UK).
Patients were deemed suitable for PC referral if they met at least 1 criterion of the trigger tool being applied.
Table 1 sets out the criteria of each tool for referral to PC. The Hua et al. and Zalenski et al. tools were selected for their validation in large cohorts, with Hua et al. representing the most extensive study of its kind. The RM tool was included due to our study centre’s significant oncology caseload. The RM tool was only applied to those patients with an active cancer diagnosis.
2.4. Data Analysis
Descriptive statistics were employed to summarise the dataset. For categorical variables, absolute frequencies (n) and percentages (%) were calculated. Means and medians with ranges were used for continuous data.
Each patient record underwent individual examination. For patients who were referred to PC, the number of ICU days prior to referral and the interval from PC referral to death were recorded. Additional data collected included age, gender, type of ICU admission (planned or unplanned), underlying diagnosis, reason for ICU admission, cause of death, and total ICU length of stay (see
Supplementary Materials S2).
For patients not referred to PC, clinical notes were reviewed against predefined trigger tools. Eligibility for PC referral was determined if any single criterion from each trigger tool was met at the time of ICU admission. No further statistical analyses were required for this study.
4. Discussion
Our study highlights a significant gap in palliative care provision, with up to 75% of ICU decedents meeting referral criteria but not being referred. Although the benefits of PC are well recognised, there appears to be a barrier to engage the service until the final day or days of illness, potentially limiting its effectiveness.
4.1. Barriers to Referral
One prevailing barrier to PC involvement in the ICU is the view that symptom management, often regarded as the core function of PC, is already well-addressed within the ICU setting by clinicians skilled in advanced supportive care [
19]. Despite this, almost all the patients referred to PC in our study were referred for symptom control, indicating likely complex and refractory symptoms. Thirty-six percent of referred patients did not receive a face-to-face review, with the majority of those referred the day before or the day of death. This gives PC practitioners, who are already a stretched resource, limited time to review these complex patients but also at a time when the key benefits of symptom control are likely to be limited.
As society has moved towards improved healthcare and more medicalised, institutionalised death, western societies have become remote from discussions and experiences associated around death and dying [
20]. This is reflected in the low profile of PC and EOLC, even within the medical profession. Palliative care remains a relatively new specialty and has historically received limited emphasis within medical curricula. It has often been associated primarily with end-of-life care in hospice settings. Consequently, many clinicians—particularly in intensive care units—may lack sufficient training, familiarity, and confidence in initiating PC consultations. This is amplified by the enduring belief that PC is synonymous with the withdrawal of curative efforts. As a result, PC is often only considered by ICU physicians once a poor prognosis is unequivocally established, due in part to concerns that early referral might be perceived as prematurely limiting active treatment.
Referral does not inevitably equate to limiting treatment options, but rather the early identification of those likely to benefit ensures patients and families have the time and support to make informed decisions. This can rarely be achieved on the day of or the day before death, yet the average ICU stay before referral in this study was 7.5 days.
Integrating PC necessitates not only addressing the patient’s medical needs but also actively involving their family. Families often experience significant emotional distress, confusion, and a sense of isolation when a loved one is admitted to the ICU. They may grapple with feelings of guilt or fear that consenting to palliative interventions equates to abandoning hope or failing their relative.
However, early and empathetic communication can alleviate these burdens. By engaging families, as well as patients, in discussions about the goals of care, clinicians can clarify that PC aims to enhance the quality of life, manage symptoms effectively, and support both the patient and their loved ones. This approach fosters shared decision-making, ensuring that care aligns with the patient’s values and the family’s expectations.
PC not only provides support for patients who are in the terminal stages of their illness but can also be employed to support those living with chronic and severe illnesses. Such patients often require specialised input for symptom control, support in making long-term decisions and achieving a good quality of life. This means that PC can in fact be employed in tandem with receiving care that is intended to cure underlying disease. When life-prolonging treatments become less effective, PC shifts toward intensive symptom management and psychosocial support [
21]. Delivered by a multidisciplinary team, PC provides medical, emotional, and practical assistance but despite its broad relevance, PC remains underutilised due to misconceptions that it is reserved for imminent death.
4.2. Underlying Malignancy
For the deteriorating oncological patient, ICU admission is recognised as a valuable intervention but is typically reserved for patients with an anticipated good prognosis [
22]. Despite this such patients often represent some of the sickest patients with a high risk of in-patient death. In this study, over half of PC referrals involved oncology patients, and retrospective application of the RM tool showed most malignancy-related deaths met at least one PC referral criterion. This reflects the hospital’s status as a specialised cancer centre where PC is well integrated into care pathways and yet remains challenging in ICU.
While traditionally associated with oncological services, PC is increasingly relevant across medical specialties as people live longer with chronic conditions [
23]. Better integration of PC into hospital settings is crucial, particularly in ICUs, where mortality rates are comparatively high and predictable.
4.3. Trigger Tools
Most patients were not known to PC prior to admission to ICU, and 77% were not seen by PC before death, suggesting they were not expected to die either on the ward or in the ICU. This aligns with evidence that clinicians struggle to recognise dying patients [
24], prioritising treatment trials before accepting death, or indeed, that these patients were not expected to die. While appropriate for reversible conditions, this highlights the need for trigger tools to aid in identifying patients who may benefit from early PC involvement to facilitate parallel planning which opens avenues for complementary care.
Although multiple trigger tools exist, they provide an objective framework to recognise dying patients. This study demonstrates that the use of a trigger tool would result in earlier PC referral. Early PC integration can improve quality of care, reduce unnecessary escalation of medical interventions, reduce hospital deaths, and enhance patient and family satisfaction.
The implementation of referral trigger tools can aid in systematically identifying patients who may benefit from palliative input, ensuring timely and appropriate involvement. However, the effectiveness of these tools is limited by the current capacity of palliative care teams. Despite the growing recognition of its value and widespread need, access to PC remains limited, with many services being underfunded and with a limited workforce capacity unable to meet the growing demand. As the need for palliative support continues to expand within critical care settings, greater investment for the integration of these services is essential to ensure equitable and high-quality care for all patients.
4.4. Limitations
This study’s findings are based on a relatively small cohort of 121 patients from two ICU wards within a single tertiary care institution in London. While this setting allowed for in-depth data collection and consistency in clinical practices, the limited sample size reduces the statistical power of the study and may affect the generalisability of the results to broader ICU populations. Consequently, the findings should be interpreted as preliminary and hypothesis-generating. Further research involving larger, more diverse patient cohorts across multiple centres is necessary to validate these results and enhance their applicability to varied clinical settings.
This study was conducted in a large teaching hospital with a high volume of oncology patients, also limiting generalisability. Additionally, two of the three trigger tools utilised in this study were developed within the context of US ICUs. Differences in healthcare delivery models between the US and UK—particularly around funding structures and referral pathways—may affect their applicability and implementation within the NHS.
The seven-month data collection period ended prior to the formal onset of the COVID-19 pandemic; nonetheless, early mortality trends related to the undetected spread of SARS-CoV-2 virus may have impacted the findings.
As a retrospective study, our findings are dependent on the accuracy and completeness of documentation within the EHR. At the time of data collection, the EHR system had been recently implemented, and it is possible that unfamiliarity with the new interface may have temporarily compromised documentation quality, leading to missing or incomplete data relevant to the study.
Some trigger tool criteria, such as “moderate or severe psychological distress,” are subjective, introducing potential interpretation bias. Additionally, defining “advanced or metastatic cancer” in haematological malignancies and patients receiving novel treatments may have proved challenging.
Despite these limitations, all three trigger tools identified an unmet need for PC referral. While this study did not assess the effectiveness of PC interventions, extensive evidence supports improved patient outcomes with early PC involvement. Importantly, we believe that valuable insights can be gained from examining the subset of patients who did receive early referral. Further work is warranted to elucidate the factors that contributed to timely referral in these cases and to identify best practices that may inform future care pathways.
5. Conclusions
The existence of multiple trigger tools underscores the complexity of real-world clinical decision-making, reflecting the clear need for such supporting frameworks. These tools offer an objective structure to aid in the recognition of dying patients and facilitate a ‘good death’.
This study demonstrates that the use of a trigger tool has the potential to increase PC referrals in the ICU. However, despite recognition of this approach as best practice, the findings reveal limited engagement with its implementation. It reinforces existing evidence supporting trigger tools in ICU settings and emphasises the evolving role of PC beyond end-of-life care, promoting a holistic, patient-centred approach. It is therefore imperative that in a patient population where mortality rates can be high and predictable, early PC involvement is considered. This shift has broad public health implications, fostering a less medicalised, more individualised approach to death, particularly in an ageing population, while encouraging open discussions on end-of-life planning.
Further research is needed to refine trigger tools for general ICU use, not only to identify those likely to die but also those who would benefit from PC in decision-making and long-term care. Crucially, studies must continue to gather evidence to establish whether increased PC consultation translates into tangible benefits for patients and families. In parallel, improving multidisciplinary team education in end-of-life and supportive care is essential to ensure that staff are equipped to recognise when PC input is appropriate and to deliver compassionate, coordinated care. PC remains an underfunded, resource-limited service, heavily reliant on charitable funding. Expanding PC access to ICUs will require significant investment, reinforcing the urgent need for increased funding to support this essential service.