Risk factors for delirium in adult patients receiving specialist palliative care: A systematic review and meta-analysis

Background: Delirium is common and distressing for patients receiving palliative care. Interventions targetting modifiable risk factors in other settings have been shown to prevent delirium. Research on delirium risk factors in palliative care can inform context-specific risk-reduction interventions. Aim: To investigate risk factors for the development of delirium in adult patients receiving specialist palliative care. Design: Systematic review and meta-analysis (PROSPERO CRD42019157168). Data sources: CINAHL, Cochrane Database of Systematic Reviews, Embase, MEDLINE and PsycINFO (1980-2021) were searched for studies reporting the association of risk factors with delirium incidence/prevalence for patients receiving specialist palliative care. Study risk of bias and certainty of evidence for each risk factor were assessed. Results: Of 28 included studies, 16 conducted only univariate analysis, 12 conducted multivariate analysis. The evidence for delirium risk factors was limited with low to very low certainty. Potentially modifiable risk factors: Opioids and lower performance status were positively associated with delirium, with some evidence also for dehydration, hypoxaemia, sleep disturbance, liver dysfunction and infection. Mixed, or very limited, evidence was found for some factors targetted in multicomponent prevention interventions: sensory impairments, mobility, catheter use, polypharmacy (single study), pain, constipation, nutrition (mixed evidence). Non-modifiable risk factors: Older age, male sex, primary brain cancer or brain metastases and lung cancer were positively associated with delirium. Conclusions: Findings may usefully inform interventions to reduce delirium risk but more high quality prospective cohort studies are required to enable greater certainty about associations of different risk factors with delirium during specialist palliative care.


What this paper adds?
• • A number of studies have examined delirium risk factors for patients receiving palliative care but the overall evidence is limited with low to very low certainty. • • Potentially modifiable risk factors reported to be positively associated with delirium included opioids and lower performance status, with some evidence of association for dehydration, hypoxaemia, sleep disturbance, liver dysfunction and infection. • • The evidence for several risk factors commonly targeted in preventative interventions in other settings was mixed or reported in a single study only.

Implications for practice, theory or policy
• • There is a need for more high quality prospective cohort studies to enable greater certainty in the association of different risk factors with delirium. • • The limited evidence may be used to inform interventions to reduce the risk of delirium in the palliative care context.

Introduction
Delirium is a distressing condition that is common in patients with advanced illness and associated with serious adverse outcomes. [1][2][3] It is a manifestation of underlying and multifactorial pathophysiological abnormalities and is characterised by acute and fluctuating disturbances in attention, awareness and cognition. 4,5 Multicomponent interventions targetting modifiable risk factors reduce the risk of delirium by around one in three cases during hospitalisation. 6,7 Systematic reviews suggest that although many risk factors for delirium are common across different patient populations and settings, their types and strengths of association vary. Reviews in older hospitalised patients 8 and Intensive Care Units 9 both found strong associations with older age, dementia and illness severity, but other significant risk factors differed and some were settingspecific, such as mechanical ventilation in intensive care. Vasilevskis et al. 10 and Ahmed et al. 8 have highlighted this variability between clinical settings and patient populations. Therefore, it is important to identify specific delirium risk factors, and the strengths of their association, for patients receiving palliative care. For example, palliative care patients' risk of delirium may be increased due to factors related to advanced disease and multi-organ dysfunction 11 and their frequent exposure to iatrogenic risk factors such as opioids. However, to our knowledge, no systematic review of delirium risk factors in patients receiving palliative care has yet been reported. 12 Synthesis of evidence on potentially modifiable risk factors is important to inform the design of interventions to reduce the risk of developing delirium. The term, 'potentially modifiable' is used in this review as the modifiability of some delirium risk factors in the palliative care context may vary, depending on factors such as the patients' illness trajectory and goals of care. This evidence may also be used, in conjunction with that on non-modifiable risk factors, to develop predictive models of patient groups most likely to develop delirium. 11 This systematic review aimed to investigate risk factors for delirium in adult patients receiving specialist palliative care and address the research questions: 1. What possible risk factors for delirium have been measured in studies of adult patients receiving specialist palliative care? 2. What is the strength of association of identified risk factors with delirium incidence or prevalence in adult patients receiving specialist palliative care?

Methods
The protocol is registered on PROSPERO on 11.11.2019 (CRD42019157168). 13 The review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. 14 Review guidance specific to observational studies of aetiology also informed its conduct. 15 The definition of a risk factor as, 'A factor that is causally related to a change in the risk of a relevant health process, outcome or condition' 16 was used.

Eligibility criteria
Eligibility criteria were framed according to PECO (Participants, Exposures, Comparators, Outcomes) criteria. 15 Eligible participants were adult patients (⩾18 years) receiving specialist palliative care services, including inpatient care in a hospice or hospital palliative care unit, palliative care consultation teams and specialist outpatient and community services. Studies needed to report the association of a possible risk factor with delirium incidence or prevalence, or data permitting its calculation. Studies that included only risk markers per se (noncausal predictive factors e.g. educational level) were excluded. Eligible studies needed to report delirium diagnosis using Diagnostic and Statistical Manual (DSM) or International Classification of Diseases (ICD) diagnostic criteria or a diagnostic cut-off score on a validated assessment (diagnostic or screening) tool. Studies needed to compare occurrence of delirium in those exposed and not exposed to the potential risk factor. Study designs eligible for inclusion included prospective and retrospective cohort studies, case-control, crosssectional studies and randomised controlled trials. Case studies, case series, qualitative studies, opinion pieces and reviews were excluded. Studies published from 1980 onwards were included (when delirium was first included in DSM-III). 17 No language restrictions were imposed. Abstracts were excluded if no full text was available.

Study retrieval and selection
The following databases were searched: CINAHL, Cochrane Database of Systematic Reviews (CDSR), Embase, MEDLINE and PsycINFO. The search strategy was developed with a Health Sciences Information Specialist. Key terms were: 'Palliative Care' and 'Delirium'. A validated search filter for palliative care was used with minor adaptations. 18 Delirium search terms were derived from recent relevant reviews. 12,19 The search was conducted in June 2019 and updated in April 2020 using the full search and screening strategy (Supplemental File 1). A final rapid update (March 2021) was conducted using an accepted method 20 (single reviewer, MEDLINE only, as it was the source of 83% of included studies in previous searches). Reference lists of included studies, relevant reviews identified through the search and delirium guidelines were examined for additional eligible studies.
Search results were screened in two stages: title and abstract and full text screening, using Covidence software. 21 Two reviewers independently screened each result (IF, RW, JB, AH, PL) using Google Translate for non-English papers. At full text stage, reviewers selected a reason for exclusion from a hierarchical list. Where needed, translators, contacted through the Centre for Reviews and Dissemination (University of York), assisted in clarifying eligibility and enabling data extraction and risk of bias assessment. Disagreements were resolved through discussion to reach consensus or consultation with a third reviewer. When no full-text paper could be found, study authors were contacted.

Data extraction
Relevant data were extracted by one reviewer (IF) using a pre-piloted data extraction table in Excel 22 and checked by a second reviewer (RW) (See Supplemental File 2 for data extraction items). Delirium occurrence was recorded as: point prevalence (delirium at a time point, such as admission); period prevalence (delirium on admission and during study period); or incidence (new cases during study period).

Risk of bias assessment
Validity and risk of bias were assessed independently by two reviewers (IF, GR, RW) using a modified version of the Quality in Prognosis Studies (QUIPS) tool. 23 The following domains were assessed: study participation; attrition; risk factor measurement; confounding measurement and account; outcome measurement; analysis and reporting. Consensus on ratings was reached through discussion. QUIPS assessments were conducted only in relation to study data relevant to this review, therefore they should not be interpreted as assessments of the studies' overall risk of bias per se. This informed decisions about whether study results should be included in meta-analysis; the narrative synthesis of results; and the GRADE (Grading of Recommendations Assessment, Development and Evaluation) 24 ratings of the evidence for each risk factor.

Data synthesis
Clinical and methodological heterogeneity were assessed to decide whether to conduct meta-analyses. Studies evaluating the same risk factor were assessed to determine the extent of variability in relation to participants and settings; measurement of the risk factor and delirium outcome, including type of delirium occurrence; study design; risk of bias; and whether only univariate or multivariate analysis was conducted. 25 Statistical heterogeneity was assessed using the I 2 statistic. 26 The principal summary measures used in the review were odds ratios (95% confidence interval (CI)) for dichotomous risk factors and mean differences (95% CI) for continuous risk factors. 27 Data were transformed to present a common statistic across studies where possible.
Revman software 28 was used to conduct meta-analyses using a random effects model (DerSimonian and Laird inverse variance) due to heterogeneity between studies. 25 Where studies provided only the odds ratio and 95% confidence interval, the log odds ratio and standard error was calculated for all studies in the comparison and generic inverse variance was used for meta-analysis. 25 Forest plots were used to present the results. Subgroup analyses were conducted to explore heterogeneity in relation to the type of delirium occurrence measured (point prevalence, period prevalence, incidence).
Narrative synthesis was used when meta-analysis could not be conducted. Studies examining each possible risk factor were grouped, their characteristics tabulated, and a common statistic presented across studies where possible. Potential moderator variables, including measurement of the risk factor and delirium outcome, were examined and a narrative interpretive summary was produced. 29

GRADE assessment
The GRADE approach, 24 with guidance on its adaptation for reviews of prognostic factor research, 30,31 was used to assess the certainty of the body of evidence for the association of each risk factor with delirium. Downgrading of the level of certainty was based on assessment of five domains (risk of bias, inconsistency, indirectness, imprecision and publication bias) and upgrading was based on large effects or exposure-response gradient. 24

Study characteristics
Supplemental File 3 shows detailed study characteristics.

Risk of bias within individual studies
Routine data on risk factors were often used and their measurement not clearly defined. The timing of exposure in relation to outcome measurement was unclear in many studies (cross-sectional or measuring delirium period prevalence). 39,47,57,59,61,62 All included studies used validated delirium assessment tools or criteria but the methods and frequency of assessment were variable. Sixteen studies included only univariate data on the association between risk factors and delirium, so made no adjustment for confounding. 37,[39][40][41][42]44,48,[51][52][53][54][55]57,58,61 Many of the studies which conducted multivariate analysis had adjusted for some, but not all, important potential confounding factors. 39,43,[45][46][47]49,50,59,60,62 Several included studies selectively reported results, including six studies which only reported multivariate analysis results that were statistically significant. 39,49,50,56,59,60 Potentially modifiable risk factors. Additional forest plots for study results included in meta-analyses are presented in Supplemental File 4. Due to heterogeneity between studies, most results could not be combined and are tabulated in Supplemental File 5.
Medications and treatment risk factors. Opioids (n = 12): Three studies reported statistically significant positive associations between opioid use and delirium in univariate analysis, 46 59 reported no statistically significant association with morphine use.
Regarding opioid dose (n = 5), two studies 54,56 reported higher mean opioid dose in the delirium group and one 57 reported a higher median dose, although these results were not statistically significant. Morita et al. 47 reported that higher opioid use (cut off unclear) was significantly associated with delirium and Mercadante et al. 33 reported a statistically significant correlation between opioid dose and delirium on admission.
All three studies examining 'opioid toxicity' 32,39 reported a statistically significant positive association with delirium.
Steroids (n = 6): Four out of five studies examining steroid use (yes/no) in univariate analysis reported a positive association with delirium which reached significance in three of the studies, 33,46,47,50,61 Two studies used multivariate analysis and found no statistically significant association. 33,46 In Matsuo et al. 49 no statistically significant association was found between higher dose of corticosteroid treatment (>3 mg initial daily dose as betamethasone equivalent) and delirium incidence (unadjusted OR 0.90, 95% CI 0.43, 1.88).
Anticholinergics (n = 6): Five studies 42,47,49,59,61 reported mixed results from univariate analysis of association between anticholinergic medications/load and delirium. Zimmerman et al. 45 reported that delirium incidence following hospital admission was significantly higher in participants whose Anticholinergic Risk Scale score increased from baseline in multivariate analysis (OR 1.43, 95% CI 1.04, 1.94).
Anxiolytics/hypnotics (n = 4): Results of four studies examining anxiolytics/ hypnotics 49,56,59,61 were mixed, but overall suggested a small positive association with delirium. These results could not be pooled due to differing definitions of medication types.  59 and antihistamine drugs 46 were each examined in one study with statistically significant associations with delirium reported in univariate analysis. Significance was retained in multivariate analysis for beta blockers (OR 3.95 95% CI 1.642, 9.479) but not for antihistamine drugs (OR 1.28 95% CI 0.14, 11.37). Plaschke et al. 59 found no association between antibiotic use and delirium.
In univariate analysis, Şenel et al. 61 reported a statistically significant positive association between polypharmacy (more than three drugs) and delirium period prevalence (unadjusted OR 26.39, 95% CI 12.70, 54.82).
Other treatments: One study found no significant association of delirium with blood transfusion. 50 Matsuo et al. 49 examined the association between indwelling urinary catheter use and delirium incidence in univariate analysis and found a positive association close to statistical significance (OR 2.25, 95% CI 0.97, 5.22).
Physiological risk factors. Nutrition-related risk factors (n = 11): The pooled mean difference in the Edmonton Symptom Assessment System (ESAS) lack of appetite score between those with and without delirium on admission from three studies 32,33,56 (Figure 2) was statistically but not clinically significant (less than 1 point difference in ESAS score). 63 Results from two further studies of ESAS lack of appetite score, not included in meta-analysis, were mixed, 33,37 as were the results of two studies examining the association of malnutrition 61 Two studies reported statistically significant positive associations between fever and delirium. 47,49 One study found a statistically significant positive association between sepsis and delirium period prevalence 62 ; but no association with pneumonia. 62 Dehydration (n = 5): Five studies examining the association between dehydration and delirium in univariate analysis, 33 33 ) reported positive associations.

Hypoxaemia (n = 5):
Of the five studies which examined association between hypoxaemia and delirium in univariate analysis, 39,46,49,57,61 four reported positive associations, of which three reached significance. However, in the one multivariate analysis 46 this was no longer associated (OR 0.93, 95% CI 0.28, 3.13).

Liver and renal dysfunction (n = 4):
Four studies reported on risk factors related to liver and renal dysfunction. 47,49,61,62 In univariate analysis, delirium was positively associated with several markers of liver dysfunction including raised bilirubin 47,49,61 and low serum albumin levels. 47,49 One study reported a non-significant positive association of delirium with liver failure 62 and one reported a statistically significant association with renal and/or liver failure. 61 In multivariate analysis, one study 62 reported no association between delirium period prevalence and chronic renal disease (OR 0.81, 95% CI 0.26, 2.53), but a statistically significant positive association with acute renal failure (OR 6.79, 95% CI 1.06, 43.41).

Leucopoenia (n = 4):
In univariate analysis, three out of four studies reported positive associations that did not reach significance between delirium and high leucocyte count 47,49,50,59 , two studies 49,50 found no association with low lymphocytes.
Calcium/sodium abnormalities (n = 2): Two studies reported no significant associations of calcium and sodium abnormalities with delirium in univariate analysis. 47,49 Oedema (n = 3): In univariate analysis, no significant association was found between lung oedema and delirium in one study. 62 In one study, no association was found with peripheral oedema, 49 but another found very increased odds of delirium for those with oedema in the lower leg or upper arm (OR 10.92, 95% CI 5.21, 22.89). 47 Pleural effusion (n = 2): No significant associations were found between delirium and pleural effusion in two studies. 49,62 Anaemia (n = 2): Two studies reported non-significant positive associations of anaemia with delirium period prevalence. 46,61 Sensory impairment (n = 1): One study reported statistically significant associations of hearing impairment 62 (OR 3.52, 95% CI 1.721, 7.210) and vision impairment 62 (OR 3.15, 95% CI 1.765, 5.607) with delirium period prevalence in multivariate analysis.
One study each reported statistically significant associations of numbness 47 and pressure sores 62 with delirium.

Symptom risk factors. (Reported in order of ESAS, except loss of appetite reported in nutrition-related risk factors above)
Symptom burden (total ESAS score): Two studies examined the association of symptom burden (total ESAS score) and delirium with mixed results. 32,33 Pain (n = 8): Pooling three studies 32,33,56 comparing the ESAS pain scores of patients with and without delirium on admission found no difference (Figure 3). Two studies 37,59 reported a higher mean pain score in participants without delirium during the study whereas two studies 53,62 reported a positive association of pain with delirium period prevalence in univariate analysis, which was close to significance in Seiler et al. 's 62 multivariate analysis (OR 2.42, 95% CI 0.95, 6.23, p = 0.07). Overall, the evidence on the association of pain with delirium was mixed.  Fatigue/weakness (n = 5): Overall, the evidence from five studies did not find a significant association of fatigue/weakness with delirium 32,33,37,49,56 (Appendices 4 and 5).

Drowsiness (n = 5):
In univariate analysis in four studies, 32,33,37,56 the evidence of an association between drowsiness and delirium was mixed but multivariate analysis in three studies 33,49,56 found statistically significant positive associations (Appendices 4 and 5).
Depression, anxiety and decreased wellbeing: Overall, the evidence from five studies examining depression and anxiety 32,33,37,51,56 did not support a clinically significant association with delirium. Four studies examining the association of decreased wellbeing with delirium had mixed results. 32,33,37,56 Sleep disturbance (n = 5): Combined mean difference in ESAS sleep disturbance score between those with and without delirium on admission from two studies was statistically, but not clinically, significant ( Figure 4). 32,56 One study 37 reported a non-significant negative association between ESAS sleep disturbance and delirium period prevalence. One study 49 reported a non-significant positive association of insomnia with delirium incidence. Slatore et al. 43 reported a hazard ratio for developing delirium for each 1 point worse sleep quality in the last month as 2.37 (95% CI 1.50, 3.74).

Constipation (n = 4):
Two studies 39,61 reported statistically significant positive associations between constipation and delirium in univariate analysis and two studies 39,49 found no association.
Behavioural disorders. Association of smoking, 56 alcohol misuse 32 and substance use disorder 62 with delirium were examined in one study each and no statistically significant associations were found.
Two studies found no significant difference in Charlson Comorbidity Index scores. 52,62 One study 45 found no statistically significant association between illness severity (APACHE III) and delirium incidence; while one study each found that admission for end of life care 32 and higher palliative prognostic index scores 61 were associated with delirium.
Non-modifiable risk factors. Additional forest plots for study results included in meta-analyses are presented in Supplemental File 6. Results are tabulated in Supplemental File 7.
Seven studies reported positive associations between other neurological/ cerebral risk factors and delirium. 39,40,44,46,51,61,62 As these were defined in different ways they could not be combined for meta-analysis. One study each reported a statistically significant association of 'neurological diseases' 61 ; 'cerebral disease' 39 ; 'focal neurological deficit' 44 and 'myoclonus' 40 with delirium.

GRADE assessment of the certainty of the evidence for risk factors' association with delirium
The certainty of the evidence for the following risk factors was graded as low: gender, lung cancer, metastases (overall), primary brain cancer, performance status, opioid use, dehydration, infection and drowsiness. For all other risk factors, certainty was rated as very low (Supplemental File 8). The evidence for most risk factors was downgraded for risk of bias, based upon the QUIPs assessments and lack of adequate adjustment for confounding factors in analysis. The evidence for many risk factors was also downgraded for inconsistency, due to unexplained heterogeneity between study results, and for imprecision of the results. 24

Main findings
We found that the evidence from 28 included studies regarding the association of possible risk factors with delirium in adults in palliative care settings was limited, particularly for many potentially modifiable risk factors, and was of low or very low certainty.
In relation to potentially modifiable risk factors, there is evidence of positive associations between delirium and both opioids and lower performance status and suggestive evidence for dehydration, hypoxaemia, sleep disturbance, liver dysfunction and infection. Positive associations with delirium were reported in only one study for each of the following: polypharmacy, sensory impairment, immobility, pressure sores and indwelling catheter use.
There was more mixed evidence about the association of delirium with pain, dyspnoea, anticholinergics, drowsiness, anxiolytics/ hypnotics, steroid use, constipation and nutrition-related risk factors.
In relation to non-modifiable risk factors, the evidence suggests positive associations of delirium with older age; male sex; primary brain cancer or brain metastases and lung cancer.

Strengths and limitations of the review
To our knowledge, this is the first systematic review conducted of risk factors for delirium in palliative care settings. Rigorous review methods were used. Particular strengths were the inclusion of studies that only used validated delirium assessment tools or criteria; inclusion of non-English language papers; screening and risk of bias assessment conducted by two independent reviewers and GRADE assessment of the certainty of evidence for each risk factor. 24 Most included studies were not high quality cohort studies comparing delirium incidence between those exposed/ not exposed to the risk factor. Identifying delirium risk factors was not the primary aim of most included studies. Many provided only cross-sectional data, or measured delirium period prevalence, so timing of risk factor exposure in relation to delirium development was unclear. Little evidence was reported on several potentially modifiable risk factors recommended in delirium risk reduction guidelines. 64,65 Delirium can result from a complex interplay between multiple risk factors 5 and this was not adequately accounted for in the included studies which mostly used only univariate analyses. Due to the multiple risk factors for delirium, it is difficult to account for all potential confounding and, as Zaal et al. 9 (p. 69) noted, 'It is poorly established which confounders should be incorporated in multivariable risk factors models' for delirium. Multivariate analysis adjusted for some, but not all, potentially important confounding factors. For example, those examining associations between pain and delirium did not adjust for opioid use.
We separately reported the results for those risk factors that are clearly non-modifiable (e.g. age, sex) and those which may potentially be modifiable. However, what is modifiable is uncertain, for example, it might not be possible to modify risk factors such as liver and renal dysfunction and performance status. In palliative care contexts, the modification of many risk factors may not be possible due to factors such as the patient's stage of illness and their goals of care. 11 Most studies in this review were conducted in inpatient palliative care settings with participants with latestage cancer, so results may be less generalisable to other patients receiving palliative care services. The definition of the eligible population as patients in receipt of specialist palliative care services could be clearly operationalised to conduct this review. However, it led to the exclusion of some possibly relevant studies, such as those of delirium risk factors for terminally ill patients in the general hospital setting. 66,67 What this study adds Delirium guidelines, not specific to palliative care contexts, such as NICE 64 and SIGN, 65 and systematic reviews of multicomponent interventions, 6,7 recommend delirium risk reduction through targetting modifiable risk factors. The evidence from this review suggests positive associations between delirium and opioid use, higher opioid dose and toxicity in palliative care settings. This supports the importance of medication review and use of the minimum effective dose, in line with delirium guidelines. 64,65 The limited evidence of associations of other modifiable risk factors with delirium, provides some support for targetting risk factors included in guidelines and interventions from other settings, 6,7,64,65 including dehydration, hypoxaemia, sleep disturbance and infection, to reduce delirium in palliative care.
The low or very low certainty of the evidence must be taken into account in drawing clinical implications from our findings. Several potentially modifiable risk factors targetted in guidelines and interventions from other settings, 6,7,64,65 including sensory impairments, mobility, catheter use and polypharmacy, were each examined in only one study, which reported positive associations with delirium. A systematic review found these to be statistically significant risk factors for delirium incidence for older hospitalised patients. 8 We found mixed evidence regarding several other risk factors-pain, constipation and nutrition-which are commonly targetted in delirium prevention strategies. This highlights the need for further, high quality research focussed on the association of potentially modifiable risk factors with delirium in palliative care settings.
Although performance status was strongly associated with delirium in this review, it is not a single modifiable risk factor. It could be useful, in conjunction with evidence on other modifiable/non-modifiable risk factors, in developing predictive models of patient subgroups most likely to develop delirium. In relation to non-modifiable risk factors, this review's results support the association between older age and delirium found in other settings. 8,9 Associations between delirium and male sex, brain cancer/metastases and lung cancer were also identified in palliative care contexts. Notably, dementia, which is strongly associated with delirium in other settings, 8,9 was only examined by one study in this review. However, as delirium prevalence in palliative care patients is very high (58%-88% in the last weeks of life), 1 it is arguably more of a priority to develop population-level preventative interventions targetting modifiable risk factors, than to predict subgroups of patients at elevated risk in order to target interventions for them.
In treating an established episode of delirium, modifiable causes are also targetted (in accordance with the patient's goals of care), to attempt to reduce the length and severity of the episode. A study in an acute palliative care unit found 49% of delirium episodes could be reversed. 55 Although the primary focus of our review was on factors affecting the risk of developing delirium, its findings could also be used to inform further research to identify modifiable factors to target in delirium treatment. Several additional studies have investigated delirium reversibility through systematic investigation of multiple precipitating risk factors in palliative care patients who have developed delirium. 55,[68][69][70] Their data were excluded from this review because they did not include a comparator group without delirium, but may be informative in research to identify factors associated with delirium reversibility.

Conclusion
The limited evidence from this systematic review offers some support for the use of interventions that target modifiable risk factors to reduce the risk of delirium for palliative care patients. However, there is a need for high quality prospective cohort studies with more comprehensive and robust delirium risk factor measurement along with adequate adjustment for important confounding factors through multivariate analysis. This will enable much greater certainty in the evidence regarding the strength of the association of different risk factors with delirium, to inform the design of multicomponent interventions to reduce risk of delirium, and clinical decisionmaking regarding competing risks, in the palliative care context.
Research (NIHR) Doctoral Fellowship [DRF-2017-10-063] for this research project. This paper presents independent research funded by the National Institute for Health Research (NIHR). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.