Background

Inequities in the provision of palliative care remain globally, whilst palliative care should be available to all who need it regardless of their diagnosis [1]. Global ageing and the changes in the prevalence of diseases imply that most needing palliative care worldwide are older people living with noncancer conditions [1,2,3,4]. However, there is consistent evidence that older people with noncancer conditions experience inequitable access to palliative care with low rates of referral or late referral in the last days or weeks of life [5, 6], such as in dementia [7] and heart failure [8].

There are several major barriers to referral in patients with noncancer conditions. A systematic review reported that one of the barriers to access and referral to palliative care is ‘a lack of national standardised referral criteria’ for screening patients with chronic noncancer disease regarding their need for palliative care [9]. A questionnaire survey showed that the highest barrier perceived by specialist palliative care service providers was ‘the unpredictable noncancer disease trajectory’ [10]. In primary care settings, the uncertainty of the illness trajectory was also identified as a barrier to effective primary palliative care provision for noncancer patients [11]. As a result, in clinical practice, key triggers or prompts for older adults with noncancer conditions to access palliative care are based on variable professional opinions or experiences [12, 13]. This means referral triggers are typically informed by differences in education, interest, and understanding on the intended outcomes of palliative care.

Referral criteria are needed to address the inequity of access to palliative care for older adults with noncancer conditions. Systematic review on referral criteria for noncancer patients aged over 65 years identified predictor variables to aid clinicians’ prognostic estimation [5]. However, the inherent uncertain disease trajectory for older adults with noncancer conditions requires the provision of palliative care to be based upon need, rather than time-based criteria, such as disease trajectory and prognostic criteria [13, 14]. Although systematic reviews identified referral criteria for palliative care among patients with heart failure [15, 16], dementia [17], and Parkinson’s disease [18], there has been no referral criteria based on palliative care needs for older general noncancer populations and the reviews assert the lack of consensus on palliative care referral criteria for adults with noncancer conditions.

Eligibility criteria for participation in clinical trials on palliative care can inform a needs-based criteria for palliative care. This approach was used and advocated by Hui et al. [12] and others [15, 17, 18] in systematic reviews investigating eligibility criteria for trials to inform triggers for outpatient palliative cancer care. Hui et al.’s review identified six themes for referral, two time-based, including cancer trajectory and prognosis, and four needs-based, including physical symptoms, performance status, psychosocial distress, and end-of-life care planning [12]. As appropriate trial eligibility criteria are designed to measure efficacy of an intervention, the criteria include populations considered likely to benefit from the intervention compared with the control. Eligibility criteria in palliative care trials seeks to identify patients with palliative care needs and considered likely to benefit from palliative care. Although eligibility criteria for randomised controlled trials (RCTs) may not be specifically designed for referral, they can inform a needs-based set of triggers for timely referral to palliative care.

Aim

This systematic review aimed to identify, appraise the applicability, and synthesise patient eligibility criteria in published trials on palliative care service level interventions for older adults severely affected by noncancer conditions. The findings intended to construct a needs-based set of triggers for timely referral to palliative care.

Methods

Study design

A systematic narrative review of the published literature on palliative care interventions for older adults with noncancer conditions to identify and synthesise the criteria used to indicate eligibility for palliative care provision [19]. The review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) 2020 guidelines [20]. The PRISMA 2020 checklist is shown in Additional file 1. The methodological quality of the included studies was assessed using a revised Cochrane risk of bias tool for randomized trials [21]. The protocol was registered in PROSPERO in November 2018 (CRD42018095845, https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42018095845845). We were not required to seek an institutional ethics approval because we only used publicly accessible documents.

Data sources and searches

Relevant articles were identified from six electronic searches: MEDLINE (Ovid), EMBASE (Ovid), CINAHL (EBSCOhost), PsycINFO (Ovid), the Cochrane Central Register of Controlled Trials (CENTRAL), and ClinicalTrials.gov. Search strategies were informed by previous systematic reviews related to palliative care and older adults [12, 13, 22]. A full search strategy can be seen in Additional file 2. All searches were conducted from database inception to September 2018 and updated on June 2022. We supplemented the electronic searches with reference chaining and citation tracking, and handsearching two palliative medicine textbooks [23, 24] and conference abstracts [Research Congress of the European Association for Palliative Care (EAPC), 2018]. All identified studies were managed in EndNote. There was no language restriction in the selection of studies.

Eligibility criteria

Types of studies

We included RCTs, including cluster randomised trials, pilot, and feasibility trials. We sought to identify and collate trial eligibility criteria for patient participants and appraise what patterns of eligibility criteria were successful in terms of recruitment, attrition, attaining sample size, and effect on the primary outcome. Feasibility and pilot trials were included, as intention is to evaluate if they can recruit patients to the trial using the stated eligibility criteria. We excluded trials that focused exclusively on the economic evaluation of palliative care as not evaluating the effect of palliative care on patient outcomes, and non-experimental studies (observational studies) as our interest was patient eligibility criteria for palliative care intervention trials. We excluded opinion pieces including editorials, commentaries, letters, and dissertations.

Types of participants

We included adults (aged ≥ 65 years) severely affected by chronic noncancer illness, including chronic heart failure (CHF), chronic obstructive pulmonary disease (COPD), chronic kidney failure, cirrhosis of the liver, stroke and long-term neurological conditions, including dementia and Parkinson’s disease. These conditions can cause considerable distressing symptoms and concerns [25]. Receipt of palliative care is shown to relive suffering [26, 27]. We included studies where > 50% of the population was aged ≥ 65 years and > 50% of the population were people with noncancer conditions.

Types of interventions

We included palliative care trials to identify individuals who could benefit from palliative care. Specialist and general palliative care interventions were included that aimed to promote quality of life for adults aged older adults ≥ 65 years severely affected by noncancer conditions. We defined a palliative care intervention as a model or service of palliative care, not a discrete aspect of palliative care, such as oxygen therapy. We defined a palliative care model or service as comprising four key elements [28,29,30,31], namely:

All levels of palliative care in any setting

We referred to the model of a three-level structure: palliative care approach in all settings, general basic palliative care, and specialist palliative care with adequate skills for each level [32]. All study settings were included: community health services, including clinics and health centres, outpatient and ambulatory care settings, and inpatient units.

An intervention providing direct palliative care to older adults

Interventions that did not directly deliver care to patients were excluded (e.g., interventions to caregivers, education programs to healthcare professionals, or evaluations of assessment tools). We considered a palliative care service intervention if the authors described it as 'palliative' anywhere in the manuscript.

An intervention had multi-component services

A palliative care service is a multidimensional and holistic approach to meet the physical, psychological, social, and spiritual needs of patients. Interventions that delivered only one component of palliative care (e.g., medication, psychotherapy, complementary therapy, decision aid) were not considered as palliative care service.

An intervention was provided by a multidisciplinary team

We defined 'palliative care services' as multidisciplinary services providing comprehensive care aiming at different physical and psychosocial components of palliative care. We excluded interventions provided by only one professional (e.g., nursing intervention).

Outcome

Types of outcome measures were not restricted.

Study selection and data extraction

The review author A.K. screened and assessed the identified titles and abstracts according to the inclusion criteria, followed by assessing all relevant full-articles by A.K. and E.A.D.P., independently. For the update search, A.K. and R.T. assessed full-articles, independently. Disagreements were resolved by consensus and discussed with C.J.E. The inter-rater reliability between the first author A.K. and E.A.D.P. and between A.K. and R.T. were assessed with a percentage of agreement. The selection process was presented in a PRISMA 2020 flow diagram (Fig. 1) [20]. Data were extracted by A.K. The PRISMA guideline [20] informed the data extraction detailing trial eligibility criteria, target population, impact on clinical outcome, and stated limitations, study design, study aim, including intervention, participant eligibility criteria, participant characteristics, screening to recruitment rate and main outcomes.

Fig. 1
figure 1

PRISMA 2020 flow diagram

Full size image

Quality appraisal of included studies

The author A.K. assessed the risk of bias in all included RCTs, as described in the Cochrane Handbook for Systematic Reviews of Interventions [33]. A revised Cochrane risk-of-bias tool for randomized trials is composed of the following five domains of bias: risk of bias arising from the randomization process; risk of bias due to deviations from intended interventions; risk of bias due to missing outcome data; risk of bias in measurement of the outcome; and bias in selection of the reported result. The summary judgements of the level of risk of bias for each domain are presented in Table 1. We used the robvis which is a visualization tool for tabulating a table of risk of bias and categorised them as, ‘low risk’, ‘high risk’, ‘some concerns’, or ‘no information’ [34].

Table 1 Characteristics of the included studies
Full size table

Data analysis and synthesis

Descriptive analysis and narrative synthesis provided descriptions of the patterns and appraised the applicability of included trial eligibility criteria to identify patients likely to benefit from receiving palliative care. We sought to identify and collate eligibility criteria and appraise what patterns of eligibility criteria were successful in terms of recruitment, attrition, attaining sample size, and effect on the primary outcome. Eligibility criteria were summarized by frequency counts of domains and synthesized by intended sample size, attrition rate, causes of attrition, and limitation that reflected on the eligibility.

To assess the recruitment, we appraised whether: 1) the study recruited the relevant population to answer the study aim, 2) the actual sample size was larger than the intended sample size determined by sample size calculation, and 3) the explanation for revision of the target sample size was given.

We assessed the attrition rate in accordance with the implementation of sample size estimation, cause of attrition, and anticipated attrition rate. Rates of attrition in the included trials were assessed to explore levels of attrition and if high attrition, to consider the appropriateness of the trial eligibility criteria to identify patients for palliative care (or not). To describe causes of attrition, we used the MORECare classification of attrition to describe causes of attrition: attrition due to death (ADD), attrition due to illness (ADI), and attrition at random (AaR) [62, 63]. While there is no standardised level of loss to follow-up which attrition related bias was identified as a problem, Schulz and Grimes noted that the readers should be concerned about the possibility of bias when the attrition rate was 20% or greater [64]. However, the weighted average attrition across palliative care trials involving adults with serious illness and increasing nearness to end of life in a systematic review was 29% [63]. For example, a review of interventional palliative oncology trials stated that the attrition rate was 26% for the primary endpoint and 44% for the end of the study [65]. Therefore, we considered high attrition rate when the rate was more than 25%. Although dropouts due to symptom progression or death were not considered as protocol failures in palliative care trials, we sought to appraise what patterns of eligibility criteria were successful in terms of recruitment and the effect on the primary outcome.

We identified for each trial the level of statistically difference on the primary outcome between the intervention and control groups and the effect size. We then explored the respective eligibility criteria and target population to map and identify criteria associated with effect on the primary patient outcome. This intended to explore further the appropriateness (or not) of the patient eligibility criteria used. Because the timing of outcome measurement could influence the attrition rate and the effect on the primary outcome, we also considered the impact of length of intervention and time points of data collection.

Results

Study selection

The electronic search strategy identified total 9,584 papers (6,720 in 2018 and 2,850 in the update search). An additional 21 papers were identified by hand searching and citation tracking. After removing duplicates, 7,134 studies were screened at title and abstract, and 67 were assessed as full-text articles. 27 met eligibility (20 studies identified by electronic searches and seven from hand search) (see Fig. 1) [35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61]. All included studies were written in English. The reasons for study exclusion are reported in Fig. 1 [66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103]. The inter-rater agreement between independent reviewers of full text screening were 89% in the initial search and 86% for the update search.

Quality appraisal

Only five studies were considered low risks of bias [36, 37, 50, 52, 53]. 20 studies had high risks of bias, mainly due to incomplete outcome data. Two study were assessed at some concerns because of unstated data and lack of information [46, 56]. The risk of bias plots are presented in Additional files 3 and 4. The average attrition rate was 23%, ranging from 0 to 52%. The major cause of attrition was death. Eleven papers stated that small sample size was one of the limitations. What we valued more than attrition rate when assessing incomplete outcome data was advance estimations of attrition, descriptions of reasons for missing data, and whether they integrated these into sample size calculation.

Study characteristics

This review included 27 RCTs written in English. Most studies were conducted in US. This is important to understand the context of the work and the applicability of the proposed triggers for which settings. The first study was conducted in US in 2000 [35]. (Table 1).

Study design

We included 18 phase III RCTs and nine feasibility trials. The 15 parallel RCTs compared the palliative care intervention with usual care. Five studies used fast-track design and compared a fast-track group with a waiting list group. Higginson et al. [57] used a parallel group fast-track trial design. Finally, five studies used a mixed methods trial design [49, 50, 53, 54, 57].

Participants

The studies included 3,663 participants ranging from 14 to 517 per study [54, 55]. The mean age ranged from 65.5 years with heart failure [36] to 85.7 years with chronic noncancer conditions and frailty [53]. The female percentage ranged from 9.1% [41] to 81.8% [35]. Eleven studies described the ethnicity of the participants; the majority of participants were White, followed by African Americans.

Of the included 27 papers, eight were conducted with patients with heart failure (HF) [36, 43,44,45, 47, 50, 60, 61]. Six included participants with respiratory disease, two with COPD [58, 59], two with interstitial lung disease (ILD) [37, 49], one with idiopathic pulmonary fibrosis (IPF) [41], and one with COPD/chronic obstructive airway disease (COAD) [54]. Other diagnoses were neurological diseases, two with Parkinson’s disease (PD) [42, 52], one with dementia [35], and one stated long-term neurological conditions. Four studies [38, 48, 51, 57] included multiple diseases, three included both cancer and noncancer conditions [38, 51, 57], and one included both CHF and COPD [48]. Three studies stated general chronic/advanced noncancer populations [46, 53, 55], and one included intensive care unit (ICU) populations [40].

Intervention and control

Eleven different models of home palliative care in 14 studies were identified [48,49,50,51,52,53,54,55,56,57,58,59,60,61]. All eleven models were composed of multi-components of palliative care intervention, such as symptom management, self-management education of disease, end-of-life discussions, case conferences, documentation, regular home visits, or a telephone hotline. Seven studies [35, 38, 40, 43,44,45, 47] provided inpatient care and six [36, 37, 39, 41, 42, 46] implemented outpatient palliative care services. Inpatient palliative care services in seven studies were developed based on the standard referral process of the hospital palliative care team or developed for the trial. As for control group, usual care differed across studies due to the wide variety of health systems and local service provisions. Several studies followed national or government guidelines.

Primary outcomes

20 studies [36,37,38, 41, 42, 44,45,46,47,48,49,50, 52,53,54,55,56,57, 59, 60] set quality of life (QOL) or symptom burden as a primary outcome, with marked heterogeneity in the measurements used. (Table 1). Eight used disease-specific QOL measurements, two used HF-specific QOL measures; The Kansas City Cardiomyopathy Questionnaire (KCCQ), four used respiratory diseases specific measures; a QOL domain in the Chronic Respiratory Disease Questionnaire (CRQ), the St. George’s Respiratory Questionnaire (SGRQ), or Maugeri Respiratory Questionnaire (MRQr), and two used measures for neurological conditions; the 39-Item Parkinson's Disease Questionnaire (PDQ-39) or the Quality of Life in Alzheimer’s Disease (QoL-AD). Other primary outcomes were medical service use, documentation of care preferences, and patient satisfaction.

Summary of eligibility criteria of included studies

Six main domains for eligibility criteria were identified, including diagnostic criteria (n = 26 out of 27 included studies, 96%), medical history-based (n = 15, 56%), symptoms (physical and psychological) (n = 14, 52%), prognostic criteria (n = 9, 33%), functional status (n = 8, 30%]), QOL (n = 2 [7%]), and other criteria (n = 4, 15%). We categorised these domains into three major criteria themes: needs-based, time-based, and medical history-based.

In Table 2, we used the initial letter of each domain to show which category the eligibility trials were categorised. The letter D stands for ‘Diagnostic Criteria’, P for ‘Prognostic’, S for ‘Symptoms’, Q for ‘QOL’, F for ‘Functional Status’, M for ‘Medical History and Treatment’, and O for ‘Other’. The number of domains was calculated by adding the number of domains used in the trial as eligibility criteria. Since some standardised measures covered several domains, we analysed the domain in the measurements and counted the number of domains. For example, Bekelman et al. [36] used a score of HF-specific health status (KCCQ) to assess eligibility. As KCCQ is a reliable and valid measure of symptoms, functional status, and QOL, the number of domains counted was three [104]. Table 3 gives an overview of the different criteria and use by respective disease groups. Table 4 gives a systematization of the eligibility criteria which shows major categories for referral criteria.

Table 2 Comparison of trial eligibility criteria and rates and cause of attrition
Full size table
Table 3 Summary of trial eligibility criteria by disease type
Full size table
Table 4 Criteria for palliative care referral
Full size table

Need-based criteria

Focused on three main areas of symptoms, function, and quality of life, including:

Symptoms

Around half of the 27 included studies set an existence of physical/psychological symptoms as eligibility criteria [36, 42, 45, 46, 48, 50, 53,54,55,56,57, 59,60,61]. As for physical symptoms, Aiken et al. [48] included HF or COPD patients suffering from fatigue, palpitation, dyspnoea, or angina with any activity. For participants with HF, Brännström et al. [50] checked the presence of cardiac cachexia with involuntary non-oedematous weight loss ≥ 6% of total body weight within the preceding 6–12 months, and Bekelman et al. [36] confirmed reporting at least one of the target symptoms of fatigue, shortness of breath, pain, and/or depression. Only six studies [36, 42, 53, 56, 60, 61] contained psychological symptoms as eligibility criteria. Four studies that provided breathlessness intervention/support service examined whether breathlessness existed in spite of optimisation of the underlying illness [46, 50, 54, 57]. Among them, Higginson et al. [57] used the Medical Research Council (MRC) dyspnoea scale to assess the degree of refractory breathlessness. Four studies [42, 53, 56, 59] conducted comprehensive screening of complex symptoms in palliative population.

Functional status

Eight studies included criteria that assess functional or performance status in their eligibility [35, 36, 40, 42, 46, 51, 53, 59]. In trials of patients with HF or COPD, the Palliative Performance Scale (PPS) and KCCQ were employed to assess functional status alongside prognosis, symptoms, or QOL [36, 51]. Regarding neurological conditions, Ahronheim et al. [35] used the Functional Assessment Staging Test (FAST) for systematic examination of the functional changes occurring in patients with dementia. Helgeson et al. [40] considered admission of patients with dementia from nursing care facilities as a pre-existing functional dependency. The trial eligibility criteria of Kluger et al. [42] on Parkinson's disease contain the Palliative Care Needs Assessment Tool (PC-NAT) and their criteria were based on a broad range of potential palliative care needs rather than time-based criteria.

Evans et al. [53] assessed the existence of frailty with the clinical frailty scale score. Shunk et al. [46] set the capability to participate in physiotherapy as a functional criterion because of the nature of the intervention programme.

Quality of life

Only two trials used QOL for trial eligibility [36, 50]. Bekelman et al. [36] used KCCQ in their trial as a measurement of the patient’s perception of their health status which includes how their heart failure impacts their QOL within a 2-week recall period. Brännström et al. [50] measured QOL using a Visual Analogue Scale (VAS). VAS is commonly used to rate subjective experiences [105].

Time-based criteria

Diagnostic criteria

Diagnostic criteria were a set of signs and tests for use in routine clinical care to guide the care of individual patients. In the 12 studies that included HF, six studies [43, 44, 48, 50, 60, 61] used the New York Heart Association (NYHA) classification of II-IV [43, 44], or NYHA III-IV [47, 48, 50, 60]. Brumley et al. [51] included not only participants with HF, but also COPD and cancer, and used the Palliative Performance Scale to assess disease severity. Rogers et al. [45] measured signs of volume overload in accordance with the HF diagnosis. But, three HF studies [38, 47, 57] used no diagnostic criteria.

Similarly, diagnostic eligibility criteria were used in studies on lung disease and neurological conditions. Aiken et al. [48] in a study on COPD used measures of hypoxemia, oxygen saturation, pO2, and oxygen requirements. Three studies [37, 41, 49] on ILD or IPF used high-resolution computed tomography of lung or a composite physiologic index. Janssen et al. (2019) [58] and Scheerens et al. [59] used the Global Initiative for Chronic Obstructive Lung Disease (GOLD) system to categorize airflow limitation into stages of COPD.

Regarding neurological conditions, Gao et al. [56] employed the Hoehn and Yahr scale and the Expanded Disability Status Scale (EDSS) to describe the progression of each neurological disease. Hanson et al. [39] used the Global Deterioration Scale (GDS) to assess the severity of dementia. The other seven studies did not clearly state the measurements of diagnostic eligibility criteria [40, 46, 51, 53,54,55, 57].

Prognostic criteria

Nine studies included prognostic eligibility criteria [37, 38, 40, 45, 48, 50, 51, 60, 61]. The ‘surprise question’ was used in four studies [38, 51, 60, 61]. The question is, “Would I be surprised if this patient died in the next 12 months?”, which has been used to identify patients at a high risk of death who might benefit from palliative care services [106]. Three [45, 60, 61] used HF-specific standardised prognostic measures. Ng et al. [60] and Wong et al. [61] and used multi-components of the prognostic indicator guidance to identify end-stage heart failure (ESHF) [107]. The indicators are constituted by three steps, which initiate intuitive surprise questions, followed by general and specific clinical indicators. In the three steps, they used only the last step, heart disease-specific clinical indicators. Rogers et al. [45] used the North American Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE) risk score which uses clinical information to derive a discharge model for six months risk of rehospitalisation and mortality [108]. Bassi et al. [37] used the gender-age-physiology (GAP) index to estimate prognosis and enrol participants with advanced ILD [109]. Helgeson et al. [40] used APACHE and SOFA scoring models to measure severity of critically ill patients admitted to ICU and to predict their mortality [110].

Medical history-based criteria

We identified 15 articles that included medical history and treatment criteria. Among 15, eleven trials included criteria related to repeated unplanned hospital admissions due to the deterioration of illness [35, 39, 40, 43,44,45, 50, 58,59,60,61]. The study of dementia by Ahronheim et al. [35] used medical history-based eligibility criteria for hospitalization for acute illness occurred with advanced dementia. However, as many participants died when they were admitted to a hospital, the authors assert these criteria limited study recruitment and attainment of sample size. Treatment-based criteria included previous/current administrative data, such as intravenous therapy support (e.g., diuretics), intubation or non-invasive ventilation, required diuretic dosing, required long-term oxygen therapy, post-cardiac arrest, and results of heart function [36, 48, 51]. Evans et al. [53] considered increasing health service use as a concern caused by severely affected non-malignant chronic conditions.

Other criteria

We found four other criteria which can be considered as psychosocial eligibility and needs perceived by healthcare professionals. Higginson et al. [57] asked patients their willingness to engage with a breathlessness support service. Farquhar et al. [54, 55] assessed whether patients might benefit from a self-management programme. Although Helgeson et al. [40] considered the medical ICU perceived need, they did not clearly state the measurements of the criteria.

Discussion

Main findings of the study

We systematically reviewed and narratively synthesised the eligibility criteria of 27 RCTs in palliative care. The findings of our review inform development of needs-based triggers for timely referral to palliative care for older adults severely affected by noncancer conditions. The results showed the list of potential needs-based triggers which were composed of three criteria; symptoms, functional status, and QOL. Eligibility criteria that were ‘successful’ tended to utilize at least one domain of needs-based criteria. Six studies used successful eligibility criteria according to the recruitment, attrition rate, and effect on primary outcomes in each study [50, 51, 53, 54, 57, 61].

What this study adds

Decisions about informing palliative care should be based on individual needs related to symptoms, functional status, and QOL. Few studies used standardized measurements with specific cut-offs for symptoms, functional decline, or QOL assessment.

Symptoms

Comprehensive assessment aligned to philosophy of palliative care is essential to identify patients likely to benefit from receiving palliative care. In this review, trial eligibility criteria are limited to mainly physical symptoms with little consideration of psychological symptoms e.g. anxiety, depression. Outcome measures of specific physical symptoms such as pain are well developed, but psychosocial symptoms are liable to be considered less serious than physical symptoms. Furthermore, only one study used standardised measures to assess symptom control: MRC dyspnoea scale [57]. Four studies [54, 55, 60, 61] measured physical symptoms such as breathlessness in spite of optimisation of underlying illness, though the measurements are not standardised tools. It implies that using a validated measure for palliative care referral is uncommon.

The best practice of symptom assessment is patient self-report outcomes for example using patient-reported outcome measures (PROMs/PROs) rather than clinician assessment due to the subjective nature of symptoms [111]. However, considering the illness trajectory and deterioration in physical/cognitive abilities in palliative care populations and the potential burden of completing PROMs, reporting by proxies such as relatives or healthcare professionals is important, especially for older adults [112,113,114]. For older adults with noncancer conditions, loss of mental capacity is common with advancing age associated with for example severe dementia and nearness to end of life. Therefore, measures used in palliative care need to be validated for the population and clinical practice, and for both self and proxy reporting [115]. Some outcome measures include a proxy version, for example the Integrated Palliative Care Outcome Scale (IPOS) [116]. This allows for the adjusting of proxy ratings if the patient is not able to complete the measure as their disease progresses.

Functional status and quality of life

Functional status and QOL are important needs-based triggers for older adults with noncancer conditions. Functional status is defined as the level of ability to do “activities performed by an individual to realize the needs of daily living in many aspects of life including physical, psychological, spiritual, intellectual, and roles [117]. Three of the included trials indicate that using standardised measures to assess functional status of patients is important to identify individuals likely to benefit from palliative care [35, 36, 51]. Most older adults severely affected by noncancer conditions experience progressive functional disability and subsequent health decline during the course of their disease. Moreover, some studies reported that functional status is significantly associated with health-related QOL (HRQOL) in people with noncancer conditions [118, 119]. Although disease-specific functional assessment measures can be available in some noncancer diseases, the Australia-modified Karnofsky Performance Status (AKPS) is a modified version for palliative care that is widely used and appropriate for multiple care settings in palliative care populations [120]. Collecting and evaluating data on functional status during routine care could inform the need for palliative care for timely referral.

The primary goal of palliative care for older people is to improve QOL with provision based on their needs [13, 29]. Quality of life can be defined as a complex, multifaceted construct that requires multiple approaches from different theoretical angles [121]. Although physical and psychological symptoms and functional impairment can be related to decline in QOL, QOL can be a trigger for referral as it can reflect an unmet need. In this review, eligibility criteria related to QOL were uncommon. As QOL that has a broad multidimensional concept can be difficult to be used as a single referral criterion, it could be operationalized as referral criteria in conjunction with other needs-based criteria. There are few relevant assessment tools addressing functional status and HRQOL for populations with multiple chronic conditions [122].

Willingness to engage with intervention

Psychosocial eligibility criteria were uncommon, mostly limited to views on willingness to engage with the intervention [54, 55, 57]. One of the major differences between palliative care and other fields of healthcare is the holistic approach it takes, including psychosocial and spiritual dimensions in addition to physical suffering. Willingness to receive palliative care may reflect the patient's preference and could form a needs-based trigger for a referral on preference for palliative care. However, patients who have preferences for palliative care may differ in characteristics compared to those with reluctance to refer to palliative care. For example, low level of health literacy of illness may preclude understanding on benefit of receiving palliative care service and impede individuals’ access to palliative care services. The Health Literacy Skills conceptual framework introduced by Squires et al. [123] illustrates mediators between health literacy and health outcomes. According to the framework, lack of knowledge about available palliative care services means patients do not request or access these services [124, 125]. Educating individuals about the role and function of palliative care, and confirming the willingness to engage with the intervention, may be one of the simplest ways to assess needs-based triggers for a referral on preference for palliative care.

Limited availability of validated assessment tools

A barrier to using individual needs-based triggers for referral criteria is the limited availability of validated and brief standardised assessment tools encompassing symptom severity, functional status, and QOL for older people with noncancer conditions. Whilst generic measures are able to be used on a large range of health and in various health conditions and populations, specific measures specifically developed to measure outcomes in palliative care are more responsive to needs-based triggers than generic outcome measures. As palliative care focuses on providing holistic care, the outcome measure used to assess palliative care needs for people with noncancer diseases should be comprehensive and encompass multiple health domains [126, 127]. Validated comprehensive measures for palliative care are available and used in clinical care, for example Edmonton Symptom Assessment Scale (ESAS) [128], and the IPOS [116] with condition specific measure for dementia and multi-morbidity (IPOS-Dem) [129]. They encompass holistic domains including, physical, psychological, social and spiritual dimensions.

Implications for further research and practice

Our review produced the initial step toward developing standardized referral criteria for clinical practice for older adults severely affected by noncancer conditions. Although the results inform a needs-based set of triggers for timely referral to palliative care, further research is needed to examine the feasibility, outcome and processes to operationalise the needs-based triggers as referral criteria in clinical settings.

Future research is needed to develop an international consensus on referral criteria for older adults with noncancer conditions and investigate if the developed referral criteria can be used in clinical settings to identify patients likely to benefit from receiving palliative care. The provision of palliative care should be based on needs assessment [32]. Our findings indicate that needs-based criteria are more likely to suit older people with noncancer conditions. Suitable needs-based referral criteria should meet the varied needs of people with different illness trajectories and different complexities of need for palliative care [32]. We recommend using measurements that encompass symptoms, QOL and functional status. Simple comprehensive measures developed and validated for palliative care population are practical for quick assessment of the palliative care needs, for example IPOS [116] or ESAS [128]. As palliative care needs vary widely, continued assessment of needs-based triggers are advocated. Standardised measures that can aid clinicians to assess palliative needs and concerns should be easy to use and interpret for all healthcare professionals and short to accommodate time constraints in clinical settings [130].

Strengths and limitations of the study

A key strength of this review is the identification and analysis of trial eligibility criteria for noncancer conditions without restricting by diagnosis. This intended to identify referral criteria applicable across noncancer conditions and multimorbidity. However, most participants in the included studies had heart failure or chronic respiratory disease. This strengthens the applicability of the identified needs-based set of triggers for these population groups, but may limit wider application to all older patients with other noncancer conditions. There is great heterogeneity among older people aged over 65 driven by for example variable diagnosis and multimorbidity, compared to any other age group. The impact this of heterogeneity on the recommendations for palliative care referral should be carefully considered. Though some studies that included mixed diagnosis attempted to reduce the heterogeneity of multimorbidity by identifying disease combinations, future research should consider how to manage heterogeneity, including stratification by age, diagnostic group, and number of co-morbidities. In the study selection process, although we assessed all relevant full-articles by two reviewers independently, the titles and abstracts of studies retrieved in bibliographic searches were assessed by one reviewer. Single screening of the titles and abstracts can influence the number of studies missed. Finally, the included studies were predominantly conducted in high-income countries in Europe and the US. This limits generalisability to non-Western regions and low-middle income countries.

Conclusion

The findings of this systematic review and narrative synthesis inform development of needs-based triggers for timely referral to palliative care for older people severely affected by noncancer conditions. For older people severely affected by noncancer conditions, decisions about providing palliative care should be based on the present needs related to symptoms, functional status, and quality of life. Further research is needed to examine the feasibility, outcome and processes to operationalise the needs-based triggers as referral criteria in clinical settings and develop international consensus on referral criteria for older adults with noncancer conditions.