Effectiveness and cost‐effectiveness of home palliative care services for adults with advanced illness and their caregivers

Barbara Gomes; Natalia Calanzani; Vito Curiale; Paul McCrone; Irene J Higginson; Maja Brito

doi:10.1002/14651858.CD007760.pub2

. 2013 Jun 6;2013(6):CD007760. doi: 10.1002/14651858.CD007760.pub2

Effectiveness and cost‐effectiveness of home palliative care services for adults with advanced illness and their caregivers

Barbara Gomes ^1,^✉, Natalia Calanzani ², Vito Curiale ^3,⁴, Paul McCrone ⁵, Irene J Higginson ⁶, Maja Brito ⁷

Editor: Cochrane Pain, Palliative and Supportive Care Group

PMCID: PMC4473359 PMID: 23744578

Abstract

Background

Extensive evidence shows that well over 50% of people prefer to be cared for and to die at home provided circumstances allow choice. Despite best efforts and policies, one‐third or less of all deaths take place at home in many countries of the world.

Objectives

1. To quantify the effect of home palliative care services for adult patients with advanced illness and their family caregivers on patients' odds of dying at home; 2. to examine the clinical effectiveness of home palliative care services on other outcomes for patients and their caregivers such as symptom control, quality of life, caregiver distress and satisfaction with care; 3. to compare the resource use and costs associated with these services; 4. to critically appraise and summarise the current evidence on cost‐effectiveness.

Search methods

We searched 12 electronic databases up to November 2012. We checked the reference lists of all included studies, 49 relevant systematic reviews, four key textbooks and recent conference abstracts. We contacted 17 experts and researchers for unpublished data.

Selection criteria

We included randomised controlled trials (RCTs), controlled clinical trials (CCTs), controlled before and after studies (CBAs) and interrupted time series (ITSs) evaluating the impact of home palliative care services on outcomes for adults with advanced illness or their family caregivers, or both.

Data collection and analysis

One review author assessed the identified titles and abstracts. Two independent reviewers performed assessment of all potentially relevant studies, data extraction and assessment of methodological quality. We carried out meta‐analysis where appropriate and calculated numbers needed to treat to benefit (NNTBs) for the primary outcome (death at home).

Main results

We identified 23 studies (16 RCTs, 6 of high quality), including 37,561 participants and 4042 family caregivers, largely with advanced cancer but also congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), HIV/AIDS and multiple sclerosis (MS), among other conditions. Meta‐analysis showed increased odds of dying at home (odds ratio (OR) 2.21, 95% CI 1.31 to 3.71; Z = 2.98, P value = 0.003; Chi² = 20.57, degrees of freedom (df) = 6, P value = 0.002; I² = 71%; NNTB 5, 95% CI 3 to 14 (seven trials with 1222 participants, three of high quality)). In addition, narrative synthesis showed evidence of small but statistically significant beneficial effects of home palliative care services compared to usual care on reducing symptom burden for patients (three trials, two of high quality, and one CBA with 2107 participants) and of no effect on caregiver grief (three RCTs, two of high quality, and one CBA with 2113 caregivers). Evidence on cost‐effectiveness (six studies) is inconclusive.

Authors' conclusions

The results provide clear and reliable evidence that home palliative care increases the chance of dying at home and reduces symptom burden in particular for patients with cancer, without impacting on caregiver grief. This justifies providing home palliative care for patients who wish to die at home. More work is needed to study cost‐effectiveness especially for people with non‐malignant conditions, assessing place of death and appropriate outcomes that are sensitive to change and valid in these populations, and to compare different models of home palliative care, in powered studies.

Keywords: Adult, Female, Humans, Male, Attitude to Death, Caregivers, Cost-Benefit Analysis, Critical Illness, Critical Illness/nursing, Home Care Services, Home Care Services/economics, Palliative Care, Palliative Care/economics, Palliative Care/methods, Patient Preference, Randomized Controlled Trials as Topic, Treatment Outcome

Plain language summary

Effectiveness and cost‐effectiveness of home‐based palliative care services for adults with advanced illness and their caregivers

When faced with the prospect of dying with an advanced illness, the majority of people prefer to die at home, yet in many countries around the world they are most likely to die in hospital. We reviewed all known studies that evaluated home palliative care services, i.e. experienced home care teams of health professionals specialised in the control of a wide range of problems associated with advanced illness – physical, psychological, social, spiritual. We wanted to see how much of a difference these services make to people's chances of dying at home, but also to other important aspects for patients towards the end of life, such as symptoms (e.g. pain) and family distress. We also compared the impact on the costs with care. On the basis of 23 studies including 37,561 patients and 4042 family caregivers, we found that when someone with an advanced illness gets home palliative care, their chances of dying at home more than double. Home palliative care services also help reduce the symptom burden people may experience as a result of advanced illness, without increasing grief for family caregivers after the patient dies. In these circumstances, patients who wish to die at home should be offered home palliative care. There is still scope to improve home palliative care services and increase the benefits for patients and families without raising costs.

Summary of findings

Summary of findings 1. Summary of findings.

Outcomes: home palliative care vs. usual care
Patient or population: adult patients with a severe or advanced disease (malignant or non‐malignant) Settings: Canada, Italy, Norway, Sweden, UK, US Intervention: home palliative care Comparison: usual care, which could include community care (primary or specialist care at home and in nursing homes), hospital care (inpatient and outpatient) and in some instances palliative or hospice care (or both)
Outcomes	Number needed to treat to benefit (NNTB)^a (95% CI)	Odds ratio (95% CI)	No of participants (studies)	Level of evidence^b (adapted fromVan Tulder 2003)	Comments
Home death follow‐up: 3 to 24 months Analysis 1.1 and Analysis 1.2	With study population control risk (307 home deaths/1000 deaths) NNTB 5 (3 to 14), meaning that for one additional patient to die at home five more would need to receive home palliative care as opposed to usual care With low home death population assumed control risk (ACR) (128 home deaths/1000 deaths) NNTB 9 (5 to 26) With medium home death population ACR (278 home deaths/1000 deaths) NNTB 6 (3 to 15) With high home death population ACR (454 home deaths/1000 deaths) NNTB 5 (3 to 13)	OR 2.21 (1.31 to 3.71)	1222 (7 studies, 3 of high quality; 5 RCTs and 2 CCTs)	Strong	The majority of patients had cancer but 3 trials also included non‐cancer conditions. 3 interventions provided specialist palliative care and 4 provided intermediate palliative care The direction of the effect was consistent across all studies but did not reach statistical significance in 3; ORs ranged from 1.36 (95% CI 0.80 to 2.31) to 2.86 (95% CI 0.78 to 10.53) Sensitivity analyses showed that exclusion of the 2 CCTs (both of Swedish hospital‐based services with a pooled OR 3.44, 95% CI 0.60 to 19.57) and inclusion of only high quality RCTs resulted in a reduction of the OR to 1.28 (95% CI 1.28 to 2.33) and 1.75 (95% CI 1.24 to 2.47) respectively, with more precision and less heterogeneity
Symptom burden follow‐up: 1 month from enrolment to the week of death Table 2	Not calculated, data were not pooled due to the high degree of heterogeneity. See comments		2107 (4 studies, 2 of high quality; 3 RCTs and 1 CBA)	Strong	Strong evidence of a positive effect on symptom burden: statistically significant reduction of symptom burden in 3 studies (one UK RCT of high quality) further to marginally significant positive effect among 209 patients in Bakitas 2009 (US RCT of high quality; P value = 0.06) Effect sizes were small (ranging from difference in mean scores of 0.08 in a 0 to 7 scale to a difference of 2.1 in a 0 to 20 scale). All studies used different measures 1 study evaluated a specialist palliative care intervention for patients with MS. The other 3 included only patients with cancer (1 evaluated a specialist service and 2 evaluated intermediate palliative care)
Pain follow‐up: 1 week from enrolment to week of death Table 3	Not calculated, data were not pooled due to the high degree of heterogeneity. See comments		2735 (9 studies; 4 high quality; 8 RCTs and 1 CBA)	Conflicting	2 UK RCTs (one of high quality) and Greer 1986 found statistically significant positive effects (the latter favouring the hospital‐based intervention); a marginally significant positive effect was found among 83 patients in McKegney 1981 in the last month before death (high quality RCT; P value = 0.06). The remaining 6 trials (including 1 high quality RCT) found no statistically significant group differences High variability in outcome measures (only the McGill‐Melzack Pain Questionnaire was used more than once)
Physical function follow‐up: 1 month from enrolment to week of death Table 4	Not calculated, data were not pooled due to the high degree of heterogeneity. See comments		2408 (7 studies, 3 high quality; 6 RCTs and 1 CBA)	Inconclusive	Statistical significance unknown in 2 out of 7 studies hence the evidence was deemed inconclusive 3 RCTs in the UK, Norway and the US (all of high quality) found no statistically significant group differences, while two RCTs of intermediate palliative care services in the US (McCorkle 1989; Aiken 2006) detected significantly better physical functioning trajectories in the intervention group through longitudinal analysis up to 9 months following enrolment
Quality of life follow‐up: 1 month from enrolment to week of death Table 5	Not calculated, data were not pooled due to the high degree of heterogeneity. See comments		2487 (7 studies; 3 of high quality; 6 RCTs and 1 CBA)	Inconclusive	Statistical significance unknown in 2 out of 7 studies hence the evidence was deemed inconclusive. 3 RCTs (2 of high quality) found no statistically significant group differences 2 US RCTs, 1 of a specialist service (high quality; Bakitas 2009) and 1 of an intermediate service (Aiken 2006) detected significantly better quality of life through longitudinal analysis up to the month of death Effects were statistically significant both forwards from enrolment and backwards from death in analyses by Bakitas 2009; they were statistically significant in physical functioning, general health and vitality but not in pain‐related, social, emotional and mental health dimensions of quality of life in Aiken 2006
Caregiver burden follow‐up: 1 month from enrolment to the patients' "last weeks of life" Table 6	Not calculated, data were not pooled due to the high degree of heterogeneity. See comments		1888 (3 studies; 2 of high quality; 2 RCTs and 1 CBA)	Conflicting	Conflicting findings from 2 high quality RCTs of specialist home palliative care interventions, 1 in the US with cancer patients (Bakitas 2009 reported no group main effects or group by time interactions 1‐10 months after enrolment) and 1 in the UK with MS patients (Higginson 2009 found differences in change scores from baseline at 12 weeks' follow‐up (P value = 0.01) Greer 1986 found a small but significant difference in the last weeks of the patient's life, with higher caregiver burden in the community‐based intervention
Caregiver grief follow‐up: from moment the patient died to 13 months after Table 7	Not calculated, data were not pooled due to the high degree of heterogeneity. See comments		2113 (4 studies, 2 of high quality; 3 RCTs and 1 CBA)	Strong	Strong evidence of no effect on caregiver grief: no statistically significant differences in three RCTs in the UK and the US (two of high quality) Outcome measures varied (only the Texas Revised Inventory of Grief was used more than once but scored in different ways) Greer 1986 found significant higher emotional distress as measured by the modified Grief Experience Inventory among caregivers in the community‐based intervention assessed 90 to 120 days after the patient died
Satisfaction with care follow‐up: 1 month from enrolment to approximately 4‐6 months after the patient died(caregiver report) Table 8	Not calculated, data were not pooled due to the high degree of heterogeneity. See comments		2497 (6 studies; 4 of high quality; 5 RCTs and 1 CBA)	Conflicting	3 RCTs (2 of high quality) found statistically significant positive effects; the other 2 RCTs (bothhigh quality studies in the US) reported no statistically significant differences. Positive effects were related to a hospital‐based specialist service in Norway (Jordhøy 2000) and 2 intermediate services in the US (Brumley 2007; Hughes 1992) Greer 1986 found significant higher satisfaction with care among caregivers in the hospital‐based intervention assessed 90‐120 days after the patient died

Study and country	Measure	Analysis	Follow‐up	Significance and direction	Details
Bakitas 2009 US (high quality)	Edmonton Symptom Assessment Scale (ESAS) measure of 9 symptoms (rated from 0 to 9): pain, activity, nausea, depression, anxiety, drowsiness, appetite, sense of well‐being, shortness of breath; scores: from 0 to 900, higher scores equal greater symptom intensity; patient report	Forwards from enrolment	1 month	Marginally significant difference favours intervention^a Mean treatment effect (intervention‐control) −27.8 (SE 5); P value = 0.06	Intervention (n = 109): LSM 241.81 (95% CI 216.35 to 267.28) Control (n = 100): LSM 288.53 (95% CI 262.03 to 315.03)
			4 months		Intervention (n = 73): LSM 254.67 (95% CI 224.55 to 284.78) Control (n = 76): LSM 271.87 (95% CI 242.11 to 301.64)
			7 months		Intervention (n = 62): LSM 238.77 (95% CI 206.60 to 270.95) Control (n = 54): LSM 268.59 (95% CI 234.34 to 302.83)
			10 months		Intervention (n = 48): LSM 271.57 (95% CI 235.83 to 307.31) Control (n = 45): LSM 294.20 (95% CI 257.27 to 331.12)
			13 months		Intervention (n = 28): LSM 295.56 (95% CI 250.65 to 340.47) Control (n = 31): LSM 251.66 (95% CI 208.51 to 294.82)
		Backwards from death	Third last assessment	n.s.^a Mean treatment effect (intervention‐control) −24.2 (SE 20.5) P value = 0.24	Intervention (n = 52): LSM 262.76 (95% CI 222.61 to 302.91) Control (n = 48): LSM 263.90 (95% CI 222.13 to 305.68)
			Second last assessment		Intervention (n = 81): LSM 274.69 (95% CI 240.63 to 308.76) Control (n = 75): LSM 304.93 (95% CI 269.53 to 340.33)
			Last assessment		Intervention (n = 80): LSM 322.29 (95% CI 288.08 to 356.51) Control (n = 74): LSM 353.90 (95% CI 318.33 to 389.47)
Higginson 2009 UK (high quality)	Palliative care Outcome Scale MS Symptoms subscale(POS‐MS‐S5) measure of 5 symptoms (rated from 0 to 4): pain, nausea, vomiting, mouth problems and sleeping difficulty; scores: from 0 to 20, higher scores equal greater symptom intensity; patient report	Forwards from enrolment	6 weeks	n.s.^b ES ‐0.5 F = 1.08 P value = 0.31	M change from baseline Intervention (n = 24): M ‐0.7 (SD 2.3; 95% CI ‐1.7 to 0.3) Control (n = 20): M 0.6 (SD 3.2; 95% CI ‐1.0 to 2.1)
Higginson 2009 UK (high quality)		Forwards from enrolment	12 weeks	Favours intervention^b ES ‐0.8 F = 4.75 P value = 0.04	M change from baseline Intervention (n = 25): M ‐1.0 (SD 2.7; 95% CI ‐2.1 to 0.1) Control (n = 21): M 1.1 (SD 2.8; 95% CI ‐0.2 to 2.4)
McCorkle 1989 US	Symptom Distress Scale measure of 13 symptoms (not stated which); scores: from 13 to 65, higher scores equal greater symptom distress; patient report	Forwards from enrolment	6 weeks	Favours intervention^c F = 5.01 P value = 0.03 Graphs showed that the entire sample experienced increased symptom distress over time but control2 (i.e. those receiving usual outpatient care) experienced elevated distress 1 occasion sooner (at 6 weeks) than the intervention and control1 (i.e. those receiving cancer home care)	Adjusted estimates Intervention: M 26.1 Control1 (cancer home care): M 24.88 Control2, (usual outpatient care): M 24.32
			12 weeks		Adjusted estimates Intervention: M 24.23 Control1 (cancer home care): M 24.71 Control2 (usual outpatient care): M 26.79
			18 weeks		Adjusted estimates Intervention: M 25.42 Control1 (cancer home care): M 26.14 Control2 (usual outpatient care): M 26.70
Greer 1986 (CBA)	Composite symptom severity scale modified from Melzack‐McGill Questionnaire measure of symptoms including nausea or vomiting, constipation, dizziness, fever or chills, dry mouth, breathlessness; scores: from 0 to 7, higher scores equal greater symptom severity; caregiver report	Backwards from death	3 weeks	Favours hospital‐based intervention^d "patients in HB hospices were likely to experience fewer symptoms than HC or CC patients, although at one week prior to death this difference was statistically significant only in the HB‐CC comparison. Subgroup analyses revealed that statistically significant differences persisted regardless of the level of symptoms at intake" (Greer 1986)	Adjusted estimates^d Community‐based intervention: M 2.89 (SE 0.09) Hospital‐based intervention: M 2.46 (SE 0.13) Control (conventional care): M 2.97 (SE 0.16)
Greer 1986 (CBA)		Backwards from death	1 week	Favours intervention regardless service based^d see comment above	Adjusted estimates^d Community‐based intervention: M 3.05 (SE 0.08) Hospital‐based intervention: M 2.78 (SE 0.12) Control (conventional care): M 3.38 (SE 0.15)

Study	Measure	Analysis	Follow‐up	Significance and direction	Details
Higginson 2009 UK (high quality)	Palliative care Outcome Scale (POS) pain item score: from 0 to 4, higher score equals greater pain; negative change equals reduction; patient report	Forwards from enrolment	6 weeks	n.s.	Mean change from baseline Intervention (n = 25): ‐0.23 (95% CI ‐0.66 to 0.20) Control (n = 23): 0.09 (95% CI ‐0.36 to 0.54)
			12 weeks	Favours intervention F = 5.15; P value = 0.028	Mean change from baseline Intervention (n = 26): ‐0.46 (95% CI ‐0.98 to 0.05) Control (n = 24): 0.30 (95% CI ‐0.16 to 0.76) Adjusted for baseline scores, the difference between scores was 0.56 (95% CI ‐ 0.75 to 1.19)
Jordhøy 2000 Norway (high quality)	EORTC QLQ‐C30 2‐item pain scale transformed score: from 0 to 100, higher score equals greater pain; patient report	Forwards from enrolment	1 month	n.s.^a P value = 0.35	Intervention (n = 153): M 36 Control (n = 116): M 36
			2 months		Intervention (n = 108): M 38 Control (n = 93): M 37
			4 months		Intervention (n = 71): M 41 Control (n = 65): M 37
			6 months	Differences and statistical significance not stated	Intervention (n = 56): M 39 Control (n = 52): M 34
McKegney 1981 US (high quality)	Sternbach Pain Estimate score score: from 0 to 100; higher score equals greater pain; patient report	Backwards from death	180 to 150 days	Authors stated there were no differences but statistical significance was not stated	"The two groups had essentially the same mean pain scores until the last 90 days before death." (McKegney 1981); this statement is corroborated by graph of mean pain scores in the 2 groups
			150 to 120 days
			120 to 90 days
			90 to 60 days	Authors stated there were differences but statistical significance was not stated	"The 'Intensive' group of patients has lower mean pain scores than the 'non‐intensive' group over the last 90 days before death. In these last 90 days, the mean pain scores in the non‐intensive group of patients continued to rise until death, whereas the mean pain scores in the intensive group of patients plateaued" (McKegney 1981). The difference in the 30 to 0 days period was marginally significant (P value = 0.06)
			60 to 30 days
			30 to 0 days	Marginally significant difference favours intervention P value = 0.06
Rabow 2004 US (high quality)	Brief Pain Inventory measure with 6 items: worst pain, least pain and "average" pain in last 24 hours (from 0 to 10); 'right now' pain (from 0 to 10); relief (from 0 to 100); interference with activities (from 0 to 70); higher scores equal greater pain; patient report	Forwards from enrolment	6 months	n.s.^b P values ranged from 0.94 (ANCOVA between groups for interference with activities) to 0.10 (ANCOVA between groups for least pain in last 24 hours)	Mean adjusted for baseline differences Intervention (n = 50) vs. control (n = 40) worst pain: 5.9 vs. 5.5 least pain: 2.7 vs. 3.9 "average" pain: 4.8 vs. 4.9 'right now' pain: 3.0 vs. 3.1 relief: 58.7 vs. 60.7 interference with activities: 43.1 vs. 39.9
			12 months		Mean adjusted for baseline differences Intervention (n = 50) vs. control (n = 40) worst pain: 4.8 vs. 5.6 least pain: 1.8 vs. 2.8 "average" pain: 3.6 vs. 4.5 'right now' pain: 2.3 vs. 2.1 relief: 68.7 vs. 59.8 interference with activities: 36.4 vs. 40.8
Aiken 2006 US	SF‐36 2‐item bodily pain subscale transformed score: from 0 to 100; lower score equal greater pain; negative slope equals reduction; patient report	Forwards from enrolment	3 months	n.s.^c	Growth modelling analysis (separate for COPD and CHF patients) COPD slope: intervention: 2.98 vs. control: ‐0.45 CHF slope: intervention: ‐0.57 vs. control: ‐0.45
			6 months
			9 months
Grande 1999 UK	Cartwright/Addington Hall surveys pain item 4‐point item, score range not stated; higher score equal greater pain; caregiver report 6 weeks after death	Backwards from death	Last 2 weeks	Favours intervention P value < 0.05	Intervention (n = 107): M 2.52 (SD 0.93) Control (n = 21): M 3.00 (SD 1.10) Although analysis used Mann–Whitney U‐tests, authors reported Ms and SDs for clarity
McCorkle 1989 US	McGill‐Melzack Pain Questionnaire score: range not stated; higher score equal greater pain; patient report	Forwards from enrolment	6 weeks	n.s.	"The three groups did not differ significantly with respect to McGill‐Melzack Pain Questionnaire" (McCorkle 1989); no data provided to support this statement
			12 weeks
			6 months
McWhinney 1994 Canada	McGill‐Melzack Pain Questionnaire score: range not stated; higher score equals greater pain; patient/caregiver report through diary	Forwards from enrolment	1 month	n.s.	"There were no clinically or statistically significant differences between the experimental and control groups on any of the measures at one month" (McWhinney 1994); no data provided to support this statement High attrition (53 /146) mainly due to death; 2 months data not analysed due to further attrition
Greer 1986 (CBA)	McGill‐Melzack Pain Questionnaire score range not stated; higher score equals greater pain; patient report	Forwards from enrolment	1 week	n.s.^d	"the average level of pain for all three patient groups was between mild and discomforting with no statistically significant differences among the groups" (Morris 1986, Greer 1986); no data provided to support this statement
			5 weeks
	1) Composite pain index modified from Spitzer Quality of Life Index score: from 0 to 4; higher score equals greater pain; caregiver report 2) Item on being pain‐free score: yes/no; caregiver report 3) Item on persistent pain score: yes/no; caregiver report	Backwards from death	3 weeks	Composite pain index n.s.^d Pain‐free n.s.^d Persistent pain favours hospital‐based intervention^d P value < 0.01	Adjusted estimates^d Composite pain index Community‐based intervention: M 1.41 (SE 0.08) Hospital‐based intervention: M 1.10 (SE 0.10) Control (conventional care): M 1.53 (SE 0.16) Patients pain‐free Community‐based intervention: 7% (SE 0.02) Hospital‐based intervention: 12% (SE 0.02) Control (conventional care): 9% (SE 0.04) Patients with persistent pain Community‐based intervention: 7% (SE 0.02) Hospital‐based intervention: 3% (SE 0.02) Control (conventional care): 14% (SE 0.04)
			1 week	Composite pain index n.s.^d Pain‐free n.s.^d Persistent pain favours hospital‐based intervention^d P value < 0.001	Adjusted estimates^d Composite pain index Community‐based intervention: M 1.61 (SE 0.06) Hospital‐based intervention: M 1.48 (SE 0.07) Control (conventional care): M 1.65 (SE 0.12) Patients pain‐free Community‐based intervention: 9% (SE 0.01) Hospital‐based intervention: 10% (SE 0.02) Control (conventional care): 16% (SE 0.04) Patients with persistent pain Community‐based intervention: 13% (SE 0.02) Hospital‐based intervention: 5% (SE 0.02) Control (conventional care): 22% (SE 0.05) Patient self reports failed to confirm these findings, but at 1 week to death 80% patients could not report

Economic data: home palliative care vs. usual care
Patient or population: adult patients with a severe or advanced disease (malignant /or non‐malignant) Settings: Italy, UK, US Intervention: home palliative care Comparison: usual care, which could include community care (primary or specialist care at home and in nursing homes), hospital care (inpatient and outpatient) and in some instances palliative or hospice care (or both)
Economic data	No of participants (studies)	Level of evidence^a (adapted fromVan Tulder 2003)	Comments
ED visits Time horizon: 6 months following enrolment; from enrolment to death, transfer to hospice care or study end; last 2 weeks before death Table 10	1103 (6 RCTs; 3 high quality)	Moderate	Moderate evidence of no statistically significant effect on ED visits: consistent across 4 RCTs. In addition, subanalysis of last 2 weeks of life for 33 patients that died in Zimmer 1985 found no ED visits in either group A significant reduction in ED use as a result of receiving home palliative care was found only in Brumley 2007 (high quality RCT conducted with patients with cancer, CHF and COPD in the US), where 20% of intervention patients had ED visits during the study period as opposed to 33% of those in usual care (P value = 0.01)
Total care costs time horizon: 12 weeks or 6 months following enrolment; from enrolment to death, transfer to hospice care or study end; last 2 weeks before death Table 11	2047 (6 studies; all high quality economic evaluations; 5 RCTs and 1 CBA)	Inconclusive	All studies reported lower costs in the intervention group with differences ranging from 18% to 35% (in Greer 1986 costs under the hospital‐based intervention were 2% lower than usual care as opposed to 32% lower under community‐based intervention). However, differences were statistically significant only in Brumley 2007 (the study with the largest sample size and only slightly underpowered by 3 patients as planned by authors to detect cost differences) Statistical significance not reported in 3 RCTs
Cost‐effectiveness time horizon: 6 months following enrolment; from enrolment to death, transfer to hospice care or study end; last 2 weeks before death Table 11	2047 patients and 1678 caregivers (6 studies; all high‐quality economic evaluations; 5 RCTs and 1 CBA)	inconclusive	Home palliative care were cost‐effective compared to usual care in Brumley 2007 (297 people with cancer, CHF and COPD) and Higginson 2009 (50 people with MS and their caregivers). However, cost‐effectiveness is unclear in the other 4 studies, as there were positive, null and negative clinical effectiveness results while costs did not differ (Hughes 1992) and the statistical significance of differences in outcomes or costs, or both, was not reported (2 trials and Greer 1986) Only Tramarin 1992 calculated a summary cost‐effectiveness measure ("average" cost‐effectiveness ratio reported in 1990 USD was USD482 per well‐week in intervention group and USD791 in control group) but with unknown statistical significance of difference or uncertainty around the estimates. In addition, Higginson 2009 plotted cost‐effectiveness planes for 2 of their outcomes. The plane for overall palliative care outcomes showed 33.8% replications in the quadrant indicating better outcomes and lower costs in the intervention group, and 54.9% in the quadrant indicating worse outcomes but lower costs. In contrast, the plane on caregiver burden showed 47.3% replications in the quadrant indicating lower costs and better outcomes and 48.0% in the quadrant showing higher costs and better outcomes

Study	Analysis	Time horizon	Significance and direction	Details
Bakitas 2009 US (high quality)	Forwards from enrolment	During study period	n.s. Wilcoxon rank sum test P value = 0.53	Intervention: 0.86 visits Control: 0.63 visits Note: not clear if the figures are means or medians
Brumley 2007 US (high quality)	Forwards from enrolment	During study period	Reduced ED use in intervention group Cramer's V 0.15; P value = 0.01 linear regression adjusted for survival, age and severity of illness showed intervention reduced ED visits by 0.35 (P value = 0.02)	Intervention: 20% had ED visits Control: 33% had ED visits
Rabow 2004 US (high quality)	Forwards from enrolment	During study period	n.s. t ‐0.24 P value = 0.81	Intervention (n = 50): M 1.6 visits (SD 2.2) Control (n = 40): M 1.7 visits (SD 2.8)
Aiken 2006 US	Forwards from enrolment	During study period	n.s. overdispersed Poisson regression model predicted number of ED visits during enrolment from group, diagnosis and their interaction, controlling for total number of days in study and number of pre‐enrolment ED visits; authors stated there was no significant intervention effect. Neither was there an effect on ED visits for a subgroup of participants identified as being at high risk for ED utilisation	Intervention: M visits/month 0.11 (SD 0.34) Control: M visits/month 0.10 (SD 0.31) Note: authors stated the number of ED visits per month remained "essentially unchanged" from 6 months prior to enrolment to period from then until the end of study
Hughes 1992 US	Forwards from enrolment	6 months following enrolment	VA ED visits "n.s." t 1.14 Non‐VA ED visits "n.s" t < 1	VA ED Intervention (n = 86): M 0.57 visits (SD 0.8) control (n = 85): M 0.72 visits (SD 0.9) non‐VA ED Intervention (n = 86): M 0.10 visits (SD 0.3) control (n = 85): M 0.08 visits (SD 0.3)
Zimmer 1985 US	Backwards from death	Last 2 weeks before death	No differences (there were no ED visits in either group)	Intervention (n = 21): 0 visits Control (n = 12): 0 visits

Study and sample analysed	Clinical effectiveness	Impact on resource use	Impact on total care costs	Cost‐effectiveness
Brumley 2007 US N intervention = 145 N control = 152 high quality economic evaluation (mean score 0.7)	+ death at home^a + death in hospital^a + patient satisfaction with care at 30 days + patient satisfaction with care at 90 days Ø patient satisfaction with care at 60 days Ø death in nursing home^a Ø death in inpatient hospice^a Ø survival^b	↓ ED visits ↓ hospital admission ↓ hospital inpatient days Ø referral to hospice care	↓ total adjusted mean costs per patient^c USD7552 lower in intervention group (33% lower; 95% CI ‐ USD12,411 to ‐ USD780; P value = 0.03; R² 0.16) unadjusted difference: t 3.63; P value < 0.001 time horizon: from enrolment to death, transfer to hospice care or study end (mean survival of 196 days in intervention group and 242 days in control group; 73% patients died) currency: 2002 USD Adjusted mean costs per patient^c Intervention USD2670 ± 12,523 Control USD20,222 ± 30,026 Adjusted mean costs per patient per day^c Intervention USD95.30 Control USD212.80 t ‐ 2.417; P value = 0.02 Total costs included those associated with physician visits, ED visits, hospital days, skilled nursing facility days, and home health or palliative days	+ no summary measure, but the intervention was cost‐effective as it resulted in statistically significant improved outcomes (no negative findings), reduced resource use (no negative findings) and a statistically significant reduction in total costs
Higginson 2009 UK N intervention = 26 N control = 24 high quality economic evaluation (mean score 0.8)	+ symptom burden at 12 weeks + pain at 12 weeks + caregiver burden at 12 weeks Ø palliative care outcomes (primary outcome; at 6 and 12 weeks) Ø symptom burden 6 weeks Ø pain at 6 weeks Ø MS psychological impact at 6 and 12 weeks Ø MS physical impact at 6 and 12 weeks Ø caregiver burden 6 weeks Ø caregiver mastery (learning new skills) at 6 and 12 weeks Ø caregiver positivity at 6 and 12 weeks	? authors reported the use of a range of health, social and voluntary services but the statistical significance of differences was not stated	Ø total mean costs per patient GBP1789 lower in intervention group (29% lower; bootstrapped 95% CI ‐ GBP5224 to GBP1902; n.s.); excluding inpatient care and informal care, mean service costs were GBP1195 lower in the intervention group (50% lower; bootstrapped 95% CI ‐ GBP2916 to GBP178; n.s.)^d time horizon: 12 weeks from enrolment (only 4 deaths) currency: 2005 GBP Mean costs per patient Intervention GBP4294 Control GBP6084 Total costs included those associated with a range of health, social, and voluntary services (inpatient care, respite care, day centre, contacts with district/practice nurse, MS nurse, palliative care nurse, other nurse, general practice, specialist at home, in hospital, in a ward and in other places, occupational therapist, physiotherapist, dietician, chiropodist, dentist, speech therapist, social services) and informal care	+ the intervention was cost‐effective as it improved caregiver burden (ZBI) with no statistically significant differences in palliative care outcomes (POS‐8) and total costs The authors plotted cost‐effectiveness planes for the two above‐mentioned outcomes (ZBI and POS‐8). These planes plot costs against outcomes forming four‐quadrants to visualise if the intervention has better outcomes and higher costs, better outcomes at lower costs, worse outcomes at higher costs or worse outcomes but at lower costs. The authors accounted for uncertainty around the cost‐effectiveness estimates by generating 1000 resamples using bootstrapping and computing cost and outcome differences for each, which were then plotted on the cost‐effectiveness planes. The point estimates in cost‐effectiveness planes suggest that the intervention was cost saving, with equivalent outcomes on overall palliative outcomes and improved outcomes for caregiver burden. The POS‐8 plane showed 33.8% replications in the lower right quadrant, indicating that intervention patients had better outcomes and lower costs than controls, and 54.9% in the quadrant indicating worse outcomes but lower costs. By contrast, in the ZBI plane, 47.3% replications were in the quadrant showing lower costs and better outcomes and 48.0% in the quadrant showing higher costs and better outcomes The authors also conducted a sensitivity analysis testing different imputation methods for dealing with missing data (last value carried, forward, next value carried backwards, and mean value), reporting similar results in nonimputed and imputed data, for all imputation methods
Zimmer 1985 US N intervention = 21 N control = 12 high quality economic evaluation (mean score 0.7)	Ø death at home^a Ø survival	? authors reported the use of a range of out‐of‐home and in‐home services but the statistical significance of differences was not stated	? total mean costs USD716 lower in intervention group (31% lower; statistical significance and/or uncertainty not reported) time horizon: last 2 weeks before death (subanalysis of deaths within the study) currency: USD, date not stated (study conducted in 1979‐1982) Mean costs of last 2 weeks before death per patient Intervention USD1577 Control USD2293 Total costs included out‐of‐home costs (hospital days, clinic visits, nursing home days, MD office or ED visits, ambulance or chairmobile rides) and in‐home costs (MD visits, nurse visits, RN/LPN hours, aide/homemaker visits, social worker visits, laboratory technician visits, meals‐on‐wheels visits)	? no summary measure, and it is unclear if the intervention was cost‐effective as there were no statistically significant differences in outcomes, and although total costs were lower in the intervention group, the statistical significance of this difference was not reported
Hughes 1992 US N intervention = 85 N control = 86 high quality economic evaluation (mean score 0.7)	+ patient satisfaction with care at 1 month + caregiver satisfaction with care at 1 month ‐ caregiver morale at 6 months Ø patient satisfaction with care at 6 months^e Ø caregiver satisfaction with care at 6 months Ø caregiver morale at 1 month Ø morale Ø cognitive functioning Ø physical function Ø survival	↓ hospital inpatient days^f ↓ VA outpatient clinic visits ↓ non‐VA community nursing visits Ø ED visits (VA and non‐VA) Ø ICU days Ø nursing home days Ø hospital admission Ø non‐VA community nursing visits Ø non‐VA private home care visits Ø extended care days ? length of last hospital admission before death	Ø total mean costs per patient USD769 lower in intervention group (18% lower; t 1.05; "n.s.") time horizon: 6 months from enrolment (mean survival was 76.2 days in intervention group and 67.1 days in control group; 79% and 78% patients died within the study, respectively) currency: 1985 USD Mean costs of 6 months following enrolment per patient Intervention USD3479.36 Control USD4248.68 Total costs included those associated with institutional care (VA and private hospitals, nursing homes) and non‐institutional (outpatient clinic visits, intervention team's visits, community nursing)	? no summary measure, and it is unclear if the intervention was cost‐effective as there were positive and negative results in clinical outcomes and the difference in total costs was not statistically significant
Tramarin 1992 Italy N intervention = 9 N control = 30 high quality economic evaluation (mean score 0.8)	? quality of life	? authors reported on hospital admission, length of hospital admission, hospital inpatient days and outpatient clinic visits but the statistical significance of differences was not stated	? total '"average" costs per person‐year USD7595 lower (35% lower; statistical significance or uncertainty, or both, not reported) time horizon: costs per person‐year (6 months from enrolment multiplied by 2; 22 deaths within the study) currency: 1990 USD (converted from 1990 ITL using healthcare‐specific purchasing power parities) "Average" total costs per person‐year Intervention stage 2/3 patients USD 14, 259 stage 2 patients only USD 11,321 stage 3 patients only USD 17,237 Control stage 2/3 patients USD 21,854 stage 2 patients only USD 15,944 stage 3 patients only USD 27,764 Total costs included more than 500 items including inpatient, outpatient clinics and home care (including intervention service), hotel and general services, diagnostic examinations and therapy, treatment items, medication and personnel salaries	? cost‐utility ratios calculated only for stage 3 patients ("average" cost‐effectiveness ratio of USD482 per well‐week in intervention group and USD791 in control group; statistical significance or uncertainty, or both, around estimates not reported) and more appropriate incremental ratios could not be calculated from the data; hence it is unclear if the intervention was cost‐effective
Greer 1986 (CBA) N intervention = 1457 (833 in community‐based intervention, 624 in hospital‐based intervention) N control = 297 (conventional care) high quality economic evaluation (mean score 0.7)	+ patient at home as long as wanted (favours community‐based intervention vs. other groups) + symptom severity at 3 weeks to death (favours hospital‐based intervention vs. other groups) + symptom severity at 1 week to death + persistent pain at 3 and 1 week to death (favours hospital‐based intervention vs. other groups) + hours of social visiting at 3 weeks to death + caregiver satisfaction with care 90 to 120 after death (favours hospital‐based intervention vs. control) + quality of death referring to 3 days before death ‐ social quality of life at 1 week to death ‐ caregiver burden in last weeks before death (higher in community‐based intervention vs. other groups) Ø patient report of pain at 1 and 5 weeks Ø survival Ø physical function at 3 and 1 week to death Ø social quality of life at 3 weeks to death Ø hours of social visiting at 1 week to death Ø hours of chatting with household members at 3 weeks to death Ø caregiver pre‐bereavement psychological well‐being (distress, use of medication for anxiety and depression, increased drinking) Ø patient satisfaction with care at 3 and 1 week to death Ø caregiver regret at 90 to 120 days after death concerning the medical care the patient received ? death at home ? caregiver satisfaction with place of death ? caregiver report of patient pain at 3 and 1 week to death (composite pain and pain‐free) ? quality of life at 3 and 1 week to death ? emotional quality of life at 3 and 1 week to death ? hours of chatting with household members at 3 weeks to death ? spiritual well‐being in the 3 days before death ? patient awareness at 3 and 1 week to death ? grief at 90 to 120 days after death ? caregiver post bereavement psychological well‐being in first 90 to 120 days after death (use of medication for anxiety and depression, increased drinking)	↑ receipt of social services ↑ general counselling in study period ↑ paperwork assistance ↑ analgesics prescribed and taken at 1 week to death (increased in hospital‐based intervention vs. other groups) ↑ oral route of analgesics ↓ analgesic consumption on a pro order ↓ aggressive interventions (radiotherapy, surgery, chemo or hormonal therapy) ↓ diagnostic tests (blood tests, x‐rays, scans) ↓ respiratory support interventions (oxygen, respiratory therapy) ↓ radiotherapy for patients with primary brain cancer or brain metastases 0 general counselling in last 2 weeks before death Ø legal/financial counselling Ø help getting services Ø self care training Ø caregiver post‐bereavement absenteeism from work in first 90‐120 days after death Ø analgesics prescribed and taken at 3 weeks to death Ø level of analgesics used Ø mean daily OME consumption Ø thoracentesis Ø palliative radiation therapy for patients with bone metastases with bone pain ? institutional days ? physician and outpatient visits ? home nursing visits ? home health/home worker visits ? hours of direct informal care caregiver post‐bereavement healthcare use (physician visits, hospitalisation in first 90 to 120 days after death)	? total costs per study day USD48 lower in community‐based intervention (32% lower) and USD3 lower in hospital‐based intervention (2% lower) compared to the control group (conventional care); statistical significance not reported) time horizon: from enrolment to death or study end (length of survival not stated; all patients died within the study) currency: 1982 USD Total costs per study day Community‐based intervention USD101 (SE 9.1)^g Hospital‐based intervention USD146 (SE 10.0)^g Control (conventional care) USD149 (SE 11.7)^g Note: not clear if the figures are means or medians Authors stated that total costs per study day were "substantially lower" in the community‐based intervention group than in the hospital‐based or control groups and that total costs were "comparable" for the first 2. Statistical significance was not stated Total costs included those associated with inpatient and home care, physician visits, outpatient clinic visits, drugs, supplies and equipment expenditures. Total costs combine "costs" and "charges" since only charges were available for physician services, drugs, supplies, and equipment purchased at home	? no summary measure, and it is unclear if the intervention was cost‐effective as there were positive and negative results in outcomes but also others where statistical significance was not reported; in addition, it is unclear if differences in costs were statistically significant between intervention groups and the control (conventional care)

Strong	Findings from meta‐analysis or consistent findings across all studies including at least 2 high‐quality RCTs
Moderate	Consistent findings across all studies including at least 2 low‐quality RCTs/CCTs or 1 high‐quality RCT
Limited	1 RCT/CCT not reaching high quality
Conflicting	Inconsistent findings among at least 2 studies with at least 1 RCT/CCT
Inconclusive	Statistical significance of differences unknown in > 25% of all studies
No evidence from trials	No RCTs or CCTs
Consistent (conflates assessment of direction and precision): statistically significant effect in same direction in ≥ 75% of all studies High‐quality RCTs/CCTs: ≥ 3.5 methodological quality score (ranging from 0 to 6)

Study, country, setting	Type of care	Service base	Core team, responsibility	Theoretical framework	Additional component	Contacts, duration
Harding 2004 UK, London, metropolitan	Specialist	2 home palliative care teams (base not stated)	Composition of home palliative care teams not stated; additional component delivered by a facilitator‐led multiprofessional group (1 facilitator per site); weekly facilitators' peer supervision (intervention consistency)	—	"90 Minute Group" new brief caregiver group intervention (research‐based, grounded on literature and qualitative study) * aimed to promote self care * informal teaching * group support (caregiver information and support) * maximum of 12 carers/group (2 at each site) * 4 principles: 1) lack of identification with the label of 'carer', 2) safe and legitimate time/space away from caring, 3) time limited, 4) respectful of existing coping strategies * facilitator introduced multiprofessional input from a changing weekly speaker (welfare benefits advice, occupational and physical therapy, clinical nurse specialist, aromatherapy) and subsequently facilitate group discussion of the weekly themes * group initially focused on patient issues in order to provide legitimacy for attending, and to address the patient‐oriented sources of distress * detailed programmed objectives for each session * transport and patient‐sitting when needed	Additional component was a 6‐week intervention (6 weekly 90‐minute sessions); survival and mortality not stated
Hudson 2005 Australia, Melbourne, Victoria, metropolitan	Specialist	2 home palliative care teams (base not stated)	Home palliative care teams of specialist nurses, doctors, and allied health professionals (including nurses, social workers, medical consultants, pastoral care workers, volunteers, bereavement counsellors); additional component delivered by nurse	—	"Psychoeducational intervention" new brief caregiver psychoeducational support (research‐based, grounded on literature and focus groups with nurses and caregivers) * aimed to enhance support/guidance for caregivers * visits structured around and complemented by audiotape with reflections from caregivers, self care strategies, structured relaxation exercise, and guidebook with information on typical aspects of caring for dying person (link provided in paper) * caregivers read guidebook sections and noted questions prior to visits and call: * first visit ‐ prepared for caregiver role and aspects of caregiving (section 1 and 2 of guidebook) * telephone call ‐ focused on evaluating plans, self care (section 3), new issues and caregiver desire to go on to next section * second visit ‐ focused on evaluating previous strategies, new issues and preparing for dying phase (final section of guidebook)	Additional component was a 2‐week intervention (2 fortnightly home visits with follow‐up call between visits); survival and mortality not stated
McMillan 2007 US, Tampa, Florida, metropolitan	Specialist	Large non‐profit community‐based hospice	Composition of home hospice care team not stated (cared for 850 patients/day); additional component delivered by newly hired nurses (1 full‐time, 1 back‐up) and home health aides with hospice experience; 4‐day training on intervention with role play and audiotapes of sessions reviewed monthly to ensure protocols were followed	Additional component grounded on stress process models and problem solving training and therapy	"COPE Intervention" new brief caregiver psychoeducational support * aimed to teach a problem solving method to assess and manage patient symptoms (focus on pain, dyspnoea and constipation) * 4 components: 1) creativity (viewing problems from different perspectives to develop new strategies for solving caregiving problems) 2) optimism (positive but realistic attitude toward problem‐solving process, including communicating realistic optimism to patient) 3) planning (setting reasonable caregiving goals and thinking out, in advance, the steps necessary to reach those goals, e.g. family holiday) 4) expert information (about nature of problem, when and where to get professional help, and what to do on one's own to deal with problem) * Home Care Guide for advanced cancer (for patients and caregivers) with 23 patient problems with easy reference for caregivers * caregiver use of assessment scales for pain, dyspnoea and constipation, findings recorded in symptom diary; ratings were foundation for action to call hospice staff and manage symptoms at home * nurse pager contact * intervention manual (available from authors)	Additional component was a 9‐day intervention with 3 visits (first 5 min, second and third 30 min) with continued support in between with calls to assess problems, offered support and answer questions; survival and mortality not stated
Walsh 2007 UK, London, metropolitan	Specialist	7 home palliative care services covering 3 regional cancer networks (2 hospital‐based, 2 community‐based, 3 inpatient hospice‐based teams)	Home palliative care teams of clinical nurse specialists, specialist medical support and sometimes social work; additional component delivered by 2 part‐time carer advisors experienced in community nursing and social work (1 month's training with fieldwork in palliative care in home, hospice, hospital settings); weekly meetings between advisors and researchers for debriefing, advice on emerging issues and to ensure all domains of need were covered; half‐day training session on intervention after 1 year	—	"Carer advisor" new brief caregiver‐focused intervention (research‐based, grounded on literature and pilot with caregivers on preferred mode of extra support) * aimed to meet caregiver alone (outside home, at workplace, evenings or weekends if needed) * mainly face‐to‐face but sometimes on phone (enabled flexibility and helped approach issues difficult to discuss face‐to‐face) * comprehensive needs assessment, discussion and advice on past, present and future issues * information and emotional support * covered physical and psychological health, need for time away from patient in short and long term and to plan future, relationships and social networks, relationships with services, finances * kept to giving advice and support rather than taking action on behalf of caregivers (advice went as far as helping to calculate benefit entitlements) * bereavement support and advice * intervention manual (available from authors)	additional component was a 6‐week intervention (6 visits over 6 weeks), introduced at second/third contact with team and if patient died before sixth contact, visits took place after death; mean 5 contacts and mean 3.6 contacts up to death; 13 weeks median survival in intervention group; 47/137 intervention patients died during study period

Study	Methods of collecting resource use information and calculating costs
Rabow 2004 US	At study completion, a trained researcher assistant (blinded to group allocation) reviewed the medical center's computerised medical records (these contained information as entered in written medical records by treating physicians) for all patients during the study period. Charges were identified from the computerised billing record. The authors recognised that analysis of cost data would be preferable but they were able only to acquire data on charges. Costs were USD; currency date and year(s) to which resource use referred were not stated (study first published in 2003)
Grande 1999 UK	Information on home nursing and inpatient care in the last year of life was collated from electronic health service databases. Year(s) to which resource use referred were not stated (study first published in 1999)
Bakitas 2009 US	Data on resource use were collected by chart review until death or end of study (2003‐2007)
Brumley 2007 US	Resource use data for each participant were collected retrospectively from the non‐profit HMO mainframe database, from the time the patient enrolled in the study until the time of death or end of study (2002 to 2004). Costs were calculated using actual costs for contracted medical services (services provided by non‐HMO contracted facilities in Colorado) and proxy cost estimates for all services provided within the HMO. The authors explained that because services provided within the HMO were not billed separately, it was necessary to use proxy costs. Hospitalisation and ED cost estimates were calculated using aggregated data from more than 500,000 HMO patient records and included ancillary services such as laboratory and radiology. Costs of physician office visits included nurse and clerk expenses. Home health and palliative care visits were calculated using "average time spent on each visit" multiplied by the cost for each discipline's reimbursement rate. The authors stated that proxy costs generated for hospital days and ED visits were significantly lower than the actual costs received from contracted providers. Costs were in 2002 USD
McCorkle 1989 US	A Medical Record Review Instrument was developed to collect systematic information about resource use from the patient's medical record. Participants were asked at each interview if they had been hospitalised, received medical treatment or visited their physician within the last 6 weeks. This information was used as a guide to identify the institutions in which medical records were to be reviewed. Trained research assistants audited the records from the time of diagnosis through the entire 24 weeks (6 months) of home care treatment (1983‐1985)
Higginson 2009 UK	Research assistants (not blinded to group allocation) collected resource use data in the last 3 months from patients in face‐to‐face interview using an adapted version of the Client Service Receipt Inventory; data were collected at baseline (before randomisation) and at 6 and 12 weeks, at 18 weeks (the latter only for the control group, after receiving the service) and 24 weeks. Costs were assessed using a broad perspective including costs to health, social and voluntary services, and informal caregivers. Service costs were calculated by combining resource use data with nationally applicable unit costs. Informal care costs were estimated by assuming that in the absence of a caregiver, the help would need to be provided by a home care worker; the unit costs of the latter were, therefore, used as a "shadow price". Costs were in 2005 GBP
Aiken 2006 US	MCOs provided resource use data on each participant for the duration of his/her enrolment in the study (1999‐2001), as well for the 6 months prior to enrolment (or all the time that the patient had been enrolled in the MCO prior to enrolment in the study). The authors had planned to measure hospital admissions as discrete events to compute length of stay for each admission as well as the cumulative number of inpatient days. However, they found that inadequately reported hospital claims data did not permit calculation of length of stay: "In a notable number of instances, admission date/discharge date pairs could not be unambiguously established from streams of MCO claims data; such pairs are required for each hospitalization episode for computation of the LOS of that hospitalization. Beyond this, an extensive line‐by‐line review of medical claims data by the PhoenixCare Medical Director and Project Director revealed that hospitalizations could not be sorted into those directly related to the PhoenixCare target diagnoses (CHF, COPD) versus those related to other ancillary conditions, because all hospitalizations were identified by the primary diagnosis regardless of the treatment provided" (Aiken 2006)
Zimmer 1985 US	Utilisation data were derived primarily from a health services utilisation diary. The diary was designed for the study and was a complete record of the patient's use of health services for as long as 6 months following study entry (1979‐1982). It was kept on a daily basis by the patient, if able, or otherwise by the caregiver (family member or friend). The diaries were monitored and collected biweekly by study interviewers. The diary recording of hospitalisations was validated, and corrected where necessary, by comparison with billing forms for all hospitalisations during the 6‐month period and also cross‐checked against a sample of Visiting Nurse Service records and records of the intervention team (the latter for the intervention group only). Resource use data were weighted to estimate costs "using current average charges for services in the community at the main provider agencies" where rates were constant (e.g. visiting nurse service, health department, ED, nursing homes). The actual mean of hospital daily charges (USD300) was obtained from their billing forms and was used as the hospital per diem weight. The authors explained that since third‐party reimbursements for physician house calls were well below actual costs, a more realistic physician home visit weighting factor was developed from estimates of total costs of a visit, including overhead and travel time, derived from a previous time‐motion study of the intervention team. Costs were in USD (currency date not stated)
Jordhøy 2000 Norway	Limited description of data collection methods. The authors stated that sociodemographic and medical data were collected by research assistants from patients at baseline and that these data were recorded partly by interview and partly by consulting patients' medical recorded. Use of healthcare services by patients in both groups was recorded (study took place from 1995 to 1997)
Hughes 1992 US	Use of healthcare services within and outside the VA was monitored by participants. To improve recall, patients were provided with a healthcare diary and were asked to record services for the 6‐month period of their participation in the study. They were contacted monthly by research staff to retrieve information and diaries were also examined by research staff in patients' homes during the 1 month and 6 month interviews. Self reported use of VA health care services was confirmed using VA records, files and computer databases. Self reported use of private sector health services was confirmed through letters or telephone calls to hospitals, doctors, clinic, nursing homes and home care agencies. Confirmation by a provider was obtained in 99% of cases and only confirmed utilisation data were included in the analysis. All VA hospital cost figures were derived from the hospital except VA ED and outpatient clinics (derived from national VA figures). Hospital costs were based on "average accounting costs per day in the hospital" (the authors believed that average cost data were the best available approximation of marginal costs referring to prior literature). Yearly "average per diem costs" for non‐government not‐for‐profit hospitals in the respective state (Illinois) provided by the American Hospital Association were adjusted based on Medicare national data to include physician costs (an additional 33%), resulting in an adjusted private hospital per diem rate of USD 665.49. ED visit base costs for all private hospitals in Illinois was provided by the Illinois Cost Containment Council and the "average base cost" for hospitals located in the area where the patient sample resided was calculated from this list. Home health agency costs were obtained from the Federal Register's Medicare schedule of maximum limits on home health agency costs for metropolitan Chicago. Medicare nursing home reimbursement rate in Illinois (USD 54.41) were provided by a survey from the Illinois Health Care Association. Actual charges confirmed by providers were used to impute the cost of private ambulatory care and home care services. Costs were in 1985 USD as 40% of patients were enrolled during that year (study took place from 1984 to 1987)
McKegney 1981 US	Although group differences in hospital inpatient days and costs are stated, the methods of data collection were not described. year(s) which resource use referred to were not stated (study first published in 1977)
Tramarin 1992 Italy	Limited description of data collection methods. The authors stated this was conducted prospectively using all available clinical documentation; costs were calculated with a specially designed software using a dBIII Plus database with more than 500 items. The authors stated that a top‐down approach was used to calculate costs because it was not possible to identify the separate use of the hospital facilities by each patient in the study (in contrast, the costs of the intervention were calculated for individual patients). The costing model was based on the analytical accounting system in use at the study hospital. Inpatient and outpatient clinical costs were estimated in 4 steps: 1) costs were obtained from the 1989 hospital general ledger (hospital overheads, ancillary services, direct inpatient services and non‐inpatient services) were mapped into initial cost centres (costs of physically discrete patient or support services, e.g. laboratory, laundry ward, radiography unit costs, according to existing hospital accounting schemes); 2) costs from the overhead cost centres were allocated, on the basis of individual cost centre allocation statistics, to the final cost centres; 3) final cost centres were adjusted on the basis of the specific inpatient factions they contained to exclude any costs for outpatient treatment; 4) inpatient and outpatient costs within the final cost centres were allocated to individual patients on the basis of their relative consumption of each final cost centre service. To obtain cost per person‐year, costs calculated during the 6 months following entry in the study (1990) were multiplied by 2. Costs were in 1990 ITL and in order to compare these costs with estimates from studies in other countries, the authors converted them to 1990 USD using health‐care‐specific purchasing power parities
Ahlner‐Elmqvist 2008 Sweden	Limited description of data collection methods. The authors stated that a full‐time employed research nurse collected medical baseline data; hospital stays were prospectively recorded for the entire study period (1995‐1999)
Buckingham 1978 US	Data were obtained from existing records, self completed questionnaire items, hospice routine data retrieval system (for the intervention group only) and medical records kept by the primary physician (study took place from 1975 to 1977)
Axelsson 1998 Sweden	Intervention and control patients received upon discharge from hospital a folder with diary sheets for the registration of their use of different resources at home. At the same time their GP and district nurse were notified by separate letters that the patient was being included in the study and that they should record any contact with the patient in the folder. The time spent on home visits by different categories of staff for the care of these patients was continuously monitored on separate sheets. This included travelling time both ways and was registered in 15‐minute units. In addition, medical records were examined to assess medical care utilisation. This was done prospectively for both groups. When a patient died, the data regarding their nursing requirements at home and in hospital were compiled. Intervention patients died between 1991 and 1993; control patients enrolled from 1992 to 1993
Greer 1986 (CBA)	Resource use data were obtained from caregivers (at each interview they presented a record of all services used by the patient); these data were checked with Medicare and other reimbursement records whenever feasible. Information on treatment prior to study enrolment was obtained from medical records (study took place between 1981 and 1983). Hospice inpatient and home care unit cost coefficients were developed using 1982 cost report data compiled either by the Health Care Financing Administration or evaluation staff accountants. Cost reports separately allocated all pertinent agency costs to a hospice cost centre. All inpatient costs were nationally adjusted based on Medicare hospital reporting data; hospice home care costs were not nationally adjusted since national standards did not exist. Total costs combine 'costs' and 'charges' since only charges were available for physician services, drugs, supplies, and equipment purchased at home. Costs include only direct health care costs. Costs were in 1982 USD

Study	Analysis	Time horizon and currency	Significance and direction	Details
Rabow 2004 US	Forwards from enrolment	Study period currency: USD, date not stated (date when study was conducted also unknown)	All medical centre services charges n.s. t 0.25; P value = 0.80 Clinic visits n.s.; t ‐0.34; P value = 0.73 Urgent care visits n.s.; t ‐1.06; P value = 0.29 ED visits n.s.; t ‐1.01; P value = 0.32 Inpatient services n.s.; t 0.01; P value = 0.10 Other charges n.s.; t 0.13; P value = 0.89	All medical centre services charges Intervention (n = 50): M USD47,211 (SD 73,009) Control (n = 40): M USD43,338 (SD 69,647) Clinic visits: Intervention (n = 50): M USD7311 (SD 10,880) Control (n = 40): M USD8068 (SD 9055) Urgent care visits: Intervention (n = 50): M USD749 (SD 2210) Control (n = 40): M USD1342 (SD 2909) ED visits: Intervention (n = 50): M USD754 (SD 1138) Control (n = 40): M USD1313 (SD 3281) Inpatient services: Intervention (n = 50): M USD31,294 (SD 54,285) Control (n = 40): M USD31,225 (SD 66,611) Other charges: Intervention (n = 50): M USD1619 (SD 7973) Control (n = 40): M USD1427 (SD 4714) "Although analysis of cost data is preferable, we were able only to acquire data on charges using the medical center's computerized billing system" (Rabow 2004)
Higginson 2009 UK	Forwards from enrolment	12 weeks following enrolment Currency: 2005 GBP	Statistical significance not stated (descriptive data only)	District/practice nurse Intervention: M GBP224 (SD 420) Control: M GBP398 (SD 922) MS nurse Intervention: M GBP33 (SD 62) Control: M GBP13 (SD 22) Palliative care nurse Intervention: M GBP46 (SD 72) Control: M GBP0 (SD 0) Other nurse Intervention: M GBP451 (SD 1573) Control: M GBP922 (SD 2078) General practice Intervention: M GBP48 (SD 87) Control: M GBP68 (SD 96) Specialist (home) Intervention: M GBP93 (SD 258) Control: M GBP0 (SD 0) Specialist (hospital) Intervention: M GBP29 (SD 44) Control: M GBP78 (SD 76) Specialist (ward) Intervention: M GBP18 (SD 37) Control: M GBP260 (SD 707) Specialist (other) Intervention: M GBP16 (SD 39) Control: M GBP19 (SD 39) Occupational therapist/physiotherapist Intervention: M GBP290 (SD 425) Control: M GBP389 (SD 897) Dietician/chiropodist/ dentist Intervention: M GBP41 (SD 65) Control: M GBP33 (SD 29) Speech therapist Intervention: M GBP16 (SD 58) Control: M GBP25 (SD 122) Social services Intervention: M GBP69 (SD 152) Control: M GBP46 (SD 89) Informal caregivers Intervention: M GBP2288 (SD 2254) Control: M GBP2620 (SD 2247) Day centre Intervention: M GBP110 (SD 321) Control: M GBP115 (SD 290) Inpatient centre Intervention: M GBP906 (SD 3173) Control: M GBP2377 (SD 6265) Respite care Intervention: M GBP39 (SD 137) Control: M GBP110 (SD 255)
Zimmer 1985 US	Backwards from death	Last 2 weeks before death currency: USD, date not stated (study conducted in 1979‐1982)	Statistical significance not stated (descriptive data only) "although the standard costs 'weight' of $300 was used in this analysis for all hospital days, the actual charges per day, taken from the hospital bills of those who were hospitalized, was higher for control than for team patients" (Zimmer 1984, Zimmer 1985)	Mean costs per patient of last 2 weeks before death Hospital costs Intervention (n = 21): USD942 Control (n = 12): USD1824 Intervention as % of control: 52% Mean actual hospital charges per day of patients hospitalised during the last 2 weeks before death Intervention (n = 5): M USD282; median USD292 Control (n = 6): M USD347; median USD313 Total out‐of‐home costs Intervention (n = 21): USD946 Control (n = 12): USD1847 Intervention as % of control: 51% Total in‐home costs Intervention (n = 21): USD631 Control (n = 12): USD446 Intervention as % of control: 141%
Hughes 1992 US	Forwards from enrolment	6 months following enrolment currency: 1985 USD	Institutional VA hospital increased in control group; t 2.47; P value = 0.02 Private hospital t < 1; "n.s." Total hospital costs increased in control group; t 2.09; P value = 0.04 Total institution costs increased in control group; t 1.99; P value = 0.05 Non‐institutional Outpatient clinic increased in control group; t 2.76; P value = 0.01 HBHC (intervention) intervention difference (tests not performed) Community nursing increased in control group; t 4.86; P value < 0.01 Total home care costs increased in intervention group; t ‐5.10; P value < 0.01 Total VA costs t < 1; "n.s." Total non‐VA costs t < 1; "n.s."	Institutional VA hospital Intervention (n = 86): M USD1795.07 Control (n = 85): M USD3434.38 Private hospital Intervention (n = 86): M USD457.56 Control (n = 85): M USD289.68 Total hospital costs (VA and private) Intervention (n = 86): M USD2251.25 Control (n = 85): M USD3724.06 Total institution costs (all hospital and nursing home) Intervention (n = 86): M USD2341.79 Control (n = 85): M USD3757.37 Non‐institutional Outpatient clinic Intervention (n = 86): M USD26.46 Control (n = 85): M USD100.42 HBHC (intervention) Intervention (n = 86): M USD999.28 Control (n = 85): ‐ Community nursing Intervention (n = 86): M USD1.97 Control (n = 85): M USD343.29 Total home care costs (HBHC plus community nurse) Intervention (n = 86): M USD1001.24 Control (n = 85): M USD343.29 Total VA costs Intervention (n = 86): M USD2934.52 Control (n = 85): M USD3602.37 Total non‐VA costs Intervention (n = 86): M USD544.84 Control (n = 85): M USD646.31 "The average general bed costs for HBHC was USD 1,310 as compared to USD 2,807 for controls (t=3.26, df=169, p<0.02)" (Hughes 1992)
Tramarin 1992 Italy	Forwards from enrolment	1 year (calculated by multiplying resources used in the 6 months following enrolment by 2) currency: 1990 ITL (million) and 1990 USD (converted from ITL using healthcare‐specific purchasing power parities)	Statistical significance not stated (descriptive data only	Inpatient care Intervention: stage 2 patients: ITL5.826 million (95% CI 2.8 to 8.6; USD3795) stage 3 patients: ITL19.744 million (95% CI 5.4 to 34; USD12,863) Control: stage 2 patients: ITL17.586 million (95% CI 14.2 to 21.4; USD11,457) stage 3 patients: ITL39.578 million (95% CI 32 to 47; USD25,784) Outpatient care Intervention: stage 2 patients: ITL9.184 million (95% CI 8.2 to 9.8; USD5983) stage 3 patients: ITL2.106 million (95% CI 1.8 to 2.2; USD1372) Control: stage 2 patients: ITL6.888 million (95% CI 6.6 to 7; USD4487) stage 3 patients: ITL3.040 million (95% CI 2.8 to 3.2; USD1980) Home care (intervention) Intervention: stage 2 patients: ITL2.368 million (95% CI 2 to 2.4; USD1543) stage 3 patients: ITL4.608 million (95% CI 3.8 to 5.4; USD3002) Control: stage 2 patients: ‐ stage 3 patients: ‐ "Analysis of cost distribution revealed that personnel salaries accounted for 71 and 19.9% of the total expenditure of the inpatient and outpatient clinics, respectively, for both groups. In the outpatient clinic, most expenses were associated with direct care of patients. Diagnostic examinations and therapy accounted for 48.5% and zidovudine for a further 29.5%" (Tramarin 1992)
Greer 1986 (CBA)	Forwards from enrolment	Study period currency: 1982 USD	Inpatient costs reduced in intervention groups "Inpatient and physician costs per study day were significantly higher in CC than in either HC or HB settings" (Greer 1986) Home care costs statistical significance not stated Physician visit costs reduced in intervention groups (see above) Outpatient costs statistical significance not stated Drugs, supplies and equipment expenditures statistical significance not stated	Inpatient costs Community‐based intervention: USD46 (SE 8.8) Hospital‐based intervention: USD99 (SE 9.6) Control (conventional care): USD135 (SE 11.6) Home care costs Community‐based intervention: USD54 (SE 4.5) Hospital‐based intervention: USD46 (SE 4.9) Control (conventional care): USD6 (SE 1.1) "Home care costs per study day in CC were approximately 10% of hospice costs"; "HC patients had more home visits than HB patients, but the difference in home care costs per day was not large enough to counterbalance the large inpatient cost difference" (Greer 1986) Physician visit costs Community‐based intervention: USD9 (SE 1.7) Hospital‐based intervention: USD8 (SE 1.9) Control (conventional care): USD18 (SE 1.6) Outpatient costs Community‐based intervention: USD1.8 (SE 0.69) Hospital‐based intervention: USD1.2 (SE 0.75) Control (conventional care): USD3.0 (SE 0.84) Drug, supplies and equipment expenditures Community‐based intervention: USD1.8 (SE 0.51) Hospital‐based intervention: USD0.7 (SE 0.56) Control (conventional care): USD0.1 (SE 0.60) Note: not clear if the figures are means or medians "Inpatient utilization increased as death approached in each of the systems of care [graphs provided]. Home service use increased dramatically in the last week of life only in the HC sample; during the last week of life, HC patients received an average of one home service visit per day in contrast to one per week for CC and 3.5 per week for HB patients" (Greer 1986)

Date	Event	Description
14 March 2016	Amended	Contact details updated.
10 November 2015	Review declared as stable	This review will be assessed for further updating in 2017. See Published notes.
19 March 2014	Amended	Reference added to Other published versions of this review.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Death at home	7	1222	Odds Ratio (M‐H, Random, 95% CI)	2.21 [1.31, 3.71]
1.1.1 RCTs	5	886	Odds Ratio (M‐H, Random, 95% CI)	1.73 [1.28, 2.33]
1.1.2 CCTs	2	336	Odds Ratio (M‐H, Random, 95% CI)	3.44 [0.60, 19.57]
1.2 death at home with only high quality RCTs	3	614	Odds Ratio (M‐H, Random, 95% CI)	1.75 [1.24, 2.47]
1.3 Death in hospital	6	1179	Odds Ratio (M‐H, Random, 95% CI)	0.31 [0.12, 0.79]
1.3.1 RCTs	4	843	Odds Ratio (M‐H, Random, 95% CI)	0.29 [0.08, 1.13]
1.3.2 CCTs	2	336	Odds Ratio (M‐H, Random, 95% CI)	0.32 [0.07, 1.46]
1.4 death in hospital with only high quality RCTs	3	614	Odds Ratio (M‐H, Random, 95% CI)	0.19 [0.03, 1.11]
1.5 Death in nursing home	5	899	Odds Ratio (M‐H, Random, 95% CI)	0.64 [0.40, 1.03]
1.5.1 RCTs	4	843	Odds Ratio (M‐H, Random, 95% CI)	0.63 [0.38, 1.02]
1.5.2 CCTs	1	56	Odds Ratio (M‐H, Random, 95% CI)	1.11 [0.11, 11.53]
1.6 death in nursing home with only high quality RCTs	3	614	Odds Ratio (M‐H, Random, 95% CI)	0.66 [0.39, 1.10]
1.7 Death in inpatient hospice/palliative care unit	5	1123	Odds Ratio (M‐H, Random, 95% CI)	1.46 [0.51, 4.19]
1.7.1 RCTs	4	843	Odds Ratio (M‐H, Random, 95% CI)	1.98 [0.36, 10.98]
1.7.2 CCTs	1	280	Odds Ratio (M‐H, Random, 95% CI)	1.30 [0.77, 2.19]
1.8 death in inpatient hospice/palliative care unit with only high quality RCTs	3	614	Odds Ratio (M‐H, Random, 95% CI)	3.71 [0.25, 54.92]

*Study characteristics*
Methods	Design: CCT (patient and caregiver) Methodological quality: 1.5/6^a (EPOC RCT/CCT checklist)
Participants	Country and regions: Sweden, Malmö, metropolitan area Recruitment (1995‐1998, 30 months): physician referrals or self referrals from departments of Oncology, Respiratory Medicine, Haematology, Surgery, Otorhinolaryngology, Urology and Gynaecology at Malmö University Hospital; patients living in the city of Malmö, with cancer (histological verified), life expectancy from 2 to 12 months (not stated who estimated), informed about diagnosis, about the palliative care situation and possibility to get care at home and additional hospital treatment if needed. Exclusion criteria: non‐Swedish speaking, mentally or physically unable to fill in questionnaires Number of patients (allocated according to their preference): 297 (119 intervention and 178 control) Diseases: cancer (297) Patient characteristics: median age 67 years intervention, 68 years control; 48% female intervention, 54% female control; 29% lived alone Number of caregivers (consented): 204 (102 in each group) Caregiver characteristics: median age 63 years intervention, 65 years control; 64% female; 77% spouses Deaths at end of study: 117 intervention (98%) and 163 control (92%) Withdrawals/other drop‐outs: "no patients are lost for follow‐up"
Interventions	Home palliative care vs. usual care "Advanced Home Care (AHC)" Type: intermediate palliative care Service base: hospital oncology department (Malmö University Hospital) Team: 9 experienced nurses, oncologist, social worker, physiotherapist, secretary; priest associated on consultation basis; all with long experience in advanced cancer care with no formal training but attended palliative care training programme prior to service start; nurses worked day and evening shifts and on night emergency services; other staff worked daytime hours; weekly meetings; able to care for 25 patients at a time Responsibility: not stated Description: new 7‐days‐a‐week service to complement existing inpatient and community care; included symptom treatment, counselling and emotional, social and family support; team home visits planned according to patient's needs and with degree of flexibility, could include injections, intravenous fluid therapy, blood transfusions, chemotherapy, nasogastric intubation and catheterisation of the urine bladder and other forms of technical support; access to on‐call service from service base (evenings, nights and weekends with on‐call oncologist home visits if need be); 3 'back‐up' beds available (2 at hospice and 1 at the oncology unit) Duration: not stated but likely to be from enrolment to death or study end (median survival for intervention group was 94 days, only 2 patients were still alive at study end) Control: CC provided jointly by hospital and community services, including inpatient and outpatient hospital care, primary care and access to inpatient hospice care (public health system); basic ideology was for patients to be cared for at 'lowest' possible level, implying that patients who do not need hospital care were cared for at home by primary care physicians and district nurses; generally, patients had to visit primary care centres and the primary care physician only visited at home in exceptional circumstances; HC on a 24‐hour basis was difficult due to organisational limitations in primary care thus if patients need advanced medical and nursing care, they had to be admitted to hospital or possibly offered a bed in a hospice
Outcomes	Death at home (primary outcome) Death in hospital Death in inpatient hospice Survival Caregiver pre‐ and post‐bereavement outcomes (quality of life; general health; vitality; physical, psychological and social functioning; bodily pain) Assessment points: caregivers were assessed every second month until death and then 1, 3, 6 and 13 months after death Analysis: backwards from death and forwards from death
Resource use/costs	Hospitalisations (admission, proportion of time in hospital) Time horizon: enrolment to death or study end (admission and proportion of time in hospital); month before death (proportion of time in hospital)
Notes	Power considerations: "the planned sample size was 150 patients in each group. (...) This number of patients is adequate to detect a 14% absolute change in place of death (...) power 90%, significant level 0.05" Return of the first completed questionnaire was regarded as written consent. Those with a preference for HC who refused to participate still received HC. No patients changed from intervention to control or vice‐versa
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	High risk	Allocation based on patient preference
Follow‐up (RCT / CCT)	Unclear risk	117 intervention (98%) and 163 control (92%) for primary outcome and survival (patients alive at the end of study were excluded from survival analysis); 122/204 (60%) for caregiver pre‐bereavement outcomes (first assessment) and 130/204 (64%) for caregiver post‐bereavement outcomes (1 month after death)
Blinding (RCT / CCT)	Unclear risk	Source of primary outcome not stated; blinding not stated
Baseline measurement (RCT / CCT)	High risk	Intervention patients preferred to be cared for at home (variable influencing in same direction as results). In addition, there were significant differences in physical function (intervention poorer KPS), primary cancer diagnosis (controls had more often lung and gastrointestinal) and weeks from diagnosis to inclusion (intervention longer time)
Reliability of outcome measurement (RCT / CCT)	Unclear risk	Source of primary outcome not stated, although it was objective; others were self completed
Protection against contamination (RCT / CCT)	High risk	Patient allocation

*Study characteristics*
Methods	Design: RCT (patient) Methodological quality: 3/6^a (EPOC RCT/CCT checklist)
Participants	Country and regions: US, Phoenix, Arizona, metropolitan area Recruitment (1999‐2001, 20 months): from community agencies, hospitals, 7 HMOs, physicians, family/friends and self referral; patients aged ≥ 18 years with CHF and COPD disease‐specific criteria of life expectancy ≤ 24 months (expert judgement) with ED or urgent care facility or hospital admissions in last 3 months and marked limitation in physical function with activity resulting in fatigue, palpitations, dyspnoea or angina Number of patients (randomised): 192 (101 intervention and 91 control) Diseases: CHF (130), COPD (62) Patient characteristics: "average" age 68.5 years; 64% female Deaths at end of study: 16% intervention and 13% control Withdrawals/other drop‐outs: 41% intervention and 54% control (entered hospice care or skilled nursing facilities, declined continued participation, left the HMO)
Interventions	Home palliative care vs. usual care "PhoenixCare intervention" Type: intermediate palliative care Service base: Hospice of the Valley – largest community‐based hospice care provider in the US Team: physician (medical director), 2 or 3 nurses (RN case managers with 30‐35 patient caseload), half‐time social worker, half‐time pastoral counsellor; staff training (2 weeks on FairCare communication model and other monthly training) Responsibility: team's nurse (with primary care physician and HMO case manager); nurse went with patient to physician visits to discuss progress and care options Description: intervention grounded on patient empowerment model and behaviour change theory (shift from crisis‐response to self empowerment and prevention); without forgoing curative treatment; included disease and symptom management, patient and caregiver education (included self monitoring, home safety, equipment and medication), psychosocial and spiritual support, co‐ordination of care and case management; needs assessment (including psychological and spiritual); referral to counselling and community resources (custodial care, transportation, help with finances); management protocols by phase of illness (initially unstable, stable, unstable following exacerbation); individualised emergency response plan at home (used in hospice care) and on‐call nurse contact; advanced care planning with patient and family; patient notebook at home (with goals patient was working towards, emergency plans, information material, reviewed in calls and visits); nurses telephone calls and home visits ("average" 6, 5, 4, 4, 3 and 3 in first 6 months and stable from then onwards; half calls, half visits); visits from other staff (45% saw social worker, "average" 6 times); on‐call nursing Duration: from enrolment to death, transfer to hospice care or skilled nursing care facility, or study end (length of stay not stated) Control: usual care focused on technical care (medication and laboratory monitoring, weight/blood pressure and blood glucose monitoring, implementation of prior authorisation mechanisms) and most patients had HMO case management (51% intervention vs. 62% control; P value = 0.12). 7 HMOs had telephone monitoring and 5 had occasional home visits. Other services included disease and symptom education, nutrition and psychological counselling, transportation and care co‐ordination. During the study, HMOs greatly narrowed criteria for case management and reduced services, due to financial pressures. 5 HMOs ended use of telephone calls and all terminated home visits and other support services. 3 HMOs reinstated some services in last study quarter, when enrolment was sparse
Outcomes	Physical symptoms (pain, most troublesome symptom) Quality of life General health and vitality Physical, psychological and social functioning Self management of illness and knowledge of resources Advance care planning (preparation for end of life) Assessment points: baseline and 3, 6 and 9 months after Analysis: forwards from enrolment
Resource use/costs	ED visits HMO case manager assigned to patient Time horizon: from enrolment to death or study end
Notes	Power considerations: "Statistical power to detect moderate effect size differences between intervention and control participants fell from 0.93 to 0.74 to 0.64 at time 0, 3, and 6, respectively" Data at 9 months were not analysed except for quality of life (reason not stated). Data on hospitalisations (presence and number of days) were assessed but could not be analysed because of "inadequately reported hospital claims data" related to problems in pairing admission and discharge dates and to reporting based on primary diagnosis regardless of treatment provided
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	Low risk	"Randomization was carried out within diagnosis, in blocks of 30 patients (15 intervention, 15 control) by a member of the project administration staff. Sealed envelopes, colour‐coded by diagnosis and containing the assignment to condition, were shuffled and assigned to participants in order of shuffling"
Follow‐up (RCT / CCT)	High risk	62% intervention and 54% control follow‐up at 3 months
Blinding (RCT / CCT)	Unclear risk	Single‐blind: "interviewers were blind to condition and diagnosis", but participants not stated
Baseline measurement (RCT / CCT)	Low risk	Statistical tests of differences in sociodemographics not reported. Trend towards significant differences at baseline SF‐36 subscales (same direction as treatment effect): "PhoenixCare participants tended to have higher scores on the SF‐36 Social Functioning Scale, F(1,189) = 3.45, P = 0.07, g = 0.27, and on the SF‐36 Mental Health Scale, F(1,190) = 3.35, P = 0.07, g = 0.26." Analysis was adjusted for baseline scores and authors reported all effects were maintained when controlling for SF‐36 baseline differences
Reliability of outcome measurement (RCT / CCT)	Unclear risk	Collected from patients by interviewers on the telephone. "Proxy respondents could read or translate for patients. Proxy responses were not accepted"
Protection against contamination (RCT / CCT)	High risk	Patients randomised

*Study characteristics*
Methods	Design: cluster CCT (patient) Methodological quality: 2.5/6^a (EPOC RCT/CCT checklist)
Participants	Country and regions: Sweden, Östersund, County of Jämtland, metropolitan area (sparsely populated region) Recruitment (1991‐1993, 30 months): mainly from Department of General Surgery at County Hospital of Östersund but also from other departments and other GPs; patients with symptomatic incurable cancer (disease within the realm of general surgery), who wished to stay at home and had a caregiver; patients living within 40 km of the hospital (9 primary care health centres) were allocated to the intervention; controls met all inclusion criteria but the latter (unknown number of primary care health centres) Number of patients (allocated): 72 (57 intervention and 15 control; authors referred to the control group as "reference group") Diseases: cancer (72) Patient characteristics: median age 72 years intervention, 71 years control; 46% female Deaths at end of study: 72 Withdrawals/other drop‐outs: 17 intervention (1 actively dying, 16 who were not matched to a third historical control group, see notes)
Interventions	Home palliative care vs. usual care "Palliative Support Service (PSS)" Type: specialist palliative care Service base: Department of General Surgery at County Hospital of Östersund Team: full‐time nurse (15 years' experience of practical cancer care on a surgical ward, trained in pain relief and symptom control, worked weekdays 8 a.m. to 5 p.m.), surgeon (half‐day per week basis); leasing car and mobile telephone; weekly conference to review cases; collaboration with the 9 local primary care centres Responsibility: not stated Description: new programme; nurse role ‐ acted as link between home and hospital, supported patient, caregiver and district nurse, district nurse education on management of infusors, epidural catheters, central venous catheters and other technical devices, administered blood transfusions, low‐dose chemotherapy and parenteral nutrition at home when needed; surgeon role ‐ scrutinised all referrals to select patients for whom support could imply improved palliation and extended time at home, reviewed patients at weekly conference with nurse, occasional home visits if needed (supported by 6 interested colleagues); patient dossier – given to the patient on discharge to improve communication between the hospital and primary healthcare service, included copies of essential information from medical records, with 3 specific fact sheets (1. sheet listed all names and telephone numbers of staff involved, namely surgeon, ward, GP, district nurse, HC assistant, consultant nurse, emergency telephone numbers for nights and weekends; 2. all prescribed drugs, information on effects and administration instructions; 3. exact wording used by doctor informing the patient of their disease and its dissemination, stating clearly whether patient had been informed of a lump, a tumour or a cancer and which metastases the patient knew of, also noted the family member(s) who had received the same information) Duration: median length of stay in intervention 70 days Control: usual care, mainly home care and conventional primary and hospital care
Outcomes	Death at home Death in hospital Death in nursing home Percentage of time and days spent at home Survival Assessment points: prospectively for intervention and retrospectively for control group, through medical records
Resource use/costs	Hospitalisations (number of admissions, terminal hospitalisation days) Institutional days (in hospital and nursing homes) HC (hours of contact with intervention consultant nurse and surgeon, district nurse, assistant nurse, GP, home service, night sitting) Informal care (days next of kin off work) Time horizon: from enrolment to death; from diagnosis to death (institutional days); last 6 months before death (hospital admissions, institutional days)
Notes	Power considerations: none stated Analysis involved only part of the intervention group, those matched to a third historical control group who died locally prior to service start (1990); patients aged ≥ 58 years were matched for age, diagnosis, place of residence and as far as possible for gender and marital status (younger patients could not be adequately matched). This excluded 16 of 57 eligible patients, resulting in a sample of 41 in the intervention group Longer recruitment for intervention patients (30 months) than for controls (4 months)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	High risk	Allocation based on area proximity to service (not random)
Follow‐up (RCT / CCT)	High risk	56/72 (78%)
Blinding (RCT / CCT)	Unclear risk	Not stated
Baseline measurement (RCT / CCT)	Unclear risk	No significant differences in age, sex distribution, marital status and KPS; however, groups differed in distance to hospital
Reliability of outcome measurement (RCT / CCT)	Unclear risk	Patient and staff filled sheets in patient's dossier
Protection against contamination (RCT / CCT)	Low risk	Allocation was by area

*Study characteristics*
Methods	Design: RCT (patient and caregiver) Methodological quality: 3.5/6^a (high quality, EPOC RCT/CCT checklist)
Participants	Country and regions: US, New Hampshire and Vermont, rural areas Recruitment (2003‐2007, 42 months): from weekly management meetings of tumour boards for gastrointestinal, genitourinary, breast and thoracic cancers at the Norris Cotton Cancer Center in New Hampshire (National Cancer Institute‐designated comprehensive cancer centre) and affiliated outreach clinics and the VA medical centre in White River Junction, Vermont; newly diagnosed (from 8 to 12 weeks) gastrointestinal, lung, genitourinary or breast cancers with specific clinical markers of advanced illness and life expectancy of approximately 1 year (assessor not stated). Exclusion criteria: impaired cognition, axis I psychiatric disorder, active substance use Number of patients (randomised): 322 (161 intervention and 161 control) Diseases (outcome sample): cancer (279): gastrointestinal (119), lung (93), genitourinary (37), breast (30) Patient characteristics (outcome sample): mean age 65.4 years intervention, 65.2 years control; 39.8% female Number of caregivers ( ): 220 (116 intervention and 104 control) Caregiver characteristics (outcome sample of 198 caregivers): mean age 58.0 years intervention, 59.9 years control; 77% female; 71% spouse/partner Deaths at end of study: 231 (72%); 112 intervention (70%) and 119 control (74%) Withdrawals/other drop‐outs: 57 patients (18%): 29 intervention (18%) and 28 control (17%) (reasons not stated); 134 caregivers (61%): 66 intervention (57%) and 68 control (65%) (due to withdrawal, refusal or completion of initial form of the questionnaire)
Interventions	Home palliative care vs. usual care "Project ENABLE II" Type: specialist palliative care Service base: palliative care programme, Dartmouth‐Hitchcock Medical Center Team: certified palliative care physician, advanced practice nurses with high speciality training in palliative care (acting as case managers with caseload balanced by diagnosis and gender); staff training (12‐20 hours on problem solving and group medical appointments provided by study psychologist; methods included didactic presentations, written treatment manuals, role‐playing with feedback – training materials available from authors); biweekly reviews of audio‐taped educational sessions and feedback on difficult patient management issues Responsibility: team's nurse (but final responsibility rested with primary clinical team for medical decisions including medication and inpatient care management, although intervention nurses, in consultation with primary team, could facilitate referrals to ancillary resources) Description: grounded on chronic care model and previous demonstration project; without forgoing curative treatment; included case management, crisis prevention, and patient and caregiver education to encourage activation, self management and empowerment; telephone‐based format (for rural population); 4 initial structured educational and problem‐solving modular sessions (31‐40 minutes each) and at least monthly telephone follow‐ups; initial full needs assessment on practical problems (e.g. work or school), family, emotional, spiritual or religious and physical problems (made use of Distress Thermometer following National Comprehensive Cancer Network guidelines – if distress rated > 3, sources of distress and problem solving explored); education manual with 4 modules (problems solving, communication and social support, symptom management, advance care planning and unfinished business) and list of supportive care resources (available from authors); referral to other resources (e.g. palliative care team, hospice, spiritual counsellor); nurse encouraged or offered to contact oncology and palliative care team when problems were identified (e.g. unrelieved pain); nurse readily available by telephone; monthly group shared medical appointments with physician and nurse for patients and caregivers to ask about medical problems and related issues, more in‐depth than clinic visits; follow‐up call in bereavement Duration: from enrolment into bereavement or study end (median survival 14 months in intervention group) Control: all oncology and supportive services without restrictions including referral to the institutions' interdisciplinary palliative care service; 1 of the sites (VA Medical Center in Vermont) had an advanced illness co‐ordinated care programme that provided consultation to oncology staff for inpatients with life‐limiting illness
Outcomes	Quality of life (primary outcome) Symptom intensity (primary outcome) Depression Death at home Death in hospital Death in nursing home Death in inpatient hospice Survival Caregiver satisfaction with care Caregiver pre‐bereavement outcomes (burden) Caregiver post‐bereavement outcomes (grief intensity) Assessment points: baseline, 1 month after, and then every 3 months until death of study end; approximately 4‐6 months after death Analysis: forwards from enrolment and backwards from death
Resource use/costs	Hospital inpatient days ED visits Intensive care unit days Referral to palliative care Referral to hospice care Time horizon: from referral to death or study end (median 14 months intervention and 8.5 months control)
Notes	Power considerations: "Original target sample size of 400 was chosen to provide 80% power to detect treatment effects of at least 0.35 SDs for scores on the Functional Assessment of Chronic Illness Therapy for Palliative Care, ESAS, and CES‐D based on a t test comparing the treatment groups with respect to the last observed value with a 2‐sided α of.01. However, at the planned study completion date, the final total study enrolment was 322 due to slightly slower accrual than anticipated" Authors conducted semi‐structured interviews of a subsample of intervention and control participants and caregivers, as well as oncology clinicians, to obtain in‐depth data about their experience with intervention and usual care
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	Low risk	"stratified randomization scheme developed for each of the 2 primary sites (...) stratified by disease and blocked within strata (block lengths of 2 and 4 varied randomly)"
Follow‐up (RCT / CCT)	Unclear risk	108 intervention (67%) and 97 control (60%) 1 month follow‐up for quality of life; 109 intervention (68%) and 100 control (62%) for symptoms; 102 intervention (63%) and 98 control (61%) for depression; 112 intervention (70%) and 119 control (74%) for place of death; 161/161 intervention (100%) and 161/161 control (100%) for survival; 50/116 intervention (43%) and 36/104 control (35%) for caregiver burden
Blinding (RCT / CCT)	High risk	Blinded at baseline but not subsequently: "research assistants notified the participant of group allocation when the baseline assessment was returned"
Baseline measurement (RCT / CCT)	Low risk	No statistically significant differences at baseline between intervention and control for demographic and clinical characteristics, the 3 patient reported outcomes and care use. "Each analysis was adjusted for the respective baseline instrument score"
Reliability of outcome measurement (RCT / CCT)	Low risk	Patients and caregivers completed the baseline questionnaires upon enrolment and the follow‐up questionnaires mailed to them
Protection against contamination (RCT / CCT)	High risk	Patients/caregivers randomised

*Study characteristics*
Methods	Design: cluster CCT (patient and caregiver) Methodological quality: 3/6^a (EPOC RCT/CCT checklist)
Participants	Country and regions: US, New Haven, metropolitan area Recruitment (1975‐1977, 21 months): referrals from oncologists; patients aged ≥ 18 years with terminal cancer (primary site lung, colon or breast) with life expectancy ≤ 6 months and ≥ 14 days survival after enrolment, living with friend or relative with potential for being primary caregiver; areas within hospice geographical region allocated to intervention and other areas allocated to control (allocation according to area where patient lived); control group matched to age (within 10 years), gender and cancer primary site of intervention group Number of patients (allocated): 70 (35 intervention and 35 control) Number of caregivers (allocated): 70 (35 intervention and 35 control) Diseases: cancer (70); primary site lung, colon or breast Patient and caregiver characteristics: not stated Deaths at end of study: not stated Withdrawals/other drop‐outs: not stated
Interventions	Home palliative care vs. usual care "Hospice home care program" Type: intermediate palliative care Service base: not stated Team: not stated Responsibility: not stated Description: hospice HC programme of terminal care, 'experienced services', "providing for many needs that often cannot be met in acute care settings" Duration: not stated Control: usual care with "differing therapeutic philosophies of organizations charged with their care"
Outcomes	Anxiety and depression Social adjustment Caregiver pre‐bereavement outcomes (anxiety, depression, social adjustment) Assessment points: not stated Analysis: forwards from enrolment
Resource use/costs	Time spent in an acute hospital or institutional setting Hospitalisations HC Time horizon: not stated
Notes	Power considerations: none stated
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	High risk	Allocation based on area proximity to service (not random)
Follow‐up (RCT / CCT)	Unclear risk	Not stated
Blinding (RCT / CCT)	Unclear risk	Not stated
Baseline measurement (RCT / CCT)	Unclear risk	Not stated
Reliability of outcome measurement (RCT / CCT)	Unclear risk	Self report questionnaires (unclear who provided)
Protection against contamination (RCT / CCT)	Low risk	Allocation by area

*Study characteristics*
Methods	Design: CBA (patient) Methodological quality: 3/6^a (EPOC CBA checklist)
Participants	Country and regions: Spain, all 19 country regions, urban and rural areas Recruitment (2007, 3 months): consecutive new referrals in 5 days to 105 palliative care services, patients > 18 years with advanced or terminal cancer, not receiving chemotherapy or radiotherapy in previous 30 days. Exclusion criteria: unavailable on the telephone Number of patients (included): 265 (158 intervention and 107 control) Diseases: cancer (265) Patient characteristics: mean age 72.2 years; 39% female Deaths at end of study: 43 (drop‐outs) Withdrawals/other drop‐outs: 19 (14 due to severe impairment, 5 reasons not stated)
Interventions	Home vs. hospital palliative care Type: specialist palliative care Service base: 62 services, base not stated Team: typically included doctor, 2 or 3 nurses and psychosocial worker (social worker or psychologist) with advanced training in palliative care; listed in national directory of palliative care services; according to a previous study, each team cared for a mean of 112 advanced or terminal patients/year Responsibility: not stated although services aimed to advise other conventional teams in the community (primary physician/team likely to remain in charge) Description: no specific information about intervention aims, components and availability of on‐call; according to previous study the teams performed a mean of 9 interventions per patient and telephone calls were the second most frequent activity Duration: not stated (median length from day of first clinical visit to death or last monitoring visit within study period was 42 days) Control: specialist palliative care provided in hospitals (43 services) including palliative care units in acute bed hospitals, palliative care units in medium‐term stay facilities, hospital support teams and outpatient clinics
Outcomes	Symptoms (pain, anorexia, nausea/vomiting, constipation, insomnia, breathlessness, anxiety, depression) Assessment points: baseline, 7 and 14 days after
Resource use/costs	Not assessed
Notes	Power considerations: none stated Of 318 services invited to take part, 223 accepted (70%) and 105 (33%) were able to recruit patients; excessive clinical workload was main reason for refusal
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Baseline measurement (CBA)	Unclear risk	Baseline characteristics for intervention patients and controls not reported separately
Characteristics of intervention and control providers (CBA)	Unclear risk	Limited information on providers' characteristics
Blinded assessment of outcomes (CBA)	Unclear risk	Source not stated
Protection against contamination (CBA)	Low risk	Second sites used as controls (hospital palliative care services)
Reliable outcome measurement (CBA)	Unclear risk	Source not stated
Follow‐up of patients/caregivers (CBA)	High risk	203 (77%) patients followed‐up (drop‐out reasons: 69% death, 23% severe impairment, 8% reasons not stated)

*Study characteristics*
Methods	Design: CBA (patient and caregiver) Methodological quality: 3.5/6^a (EPOC CBA checklist)
Participants	Country and regions: US, Southern New England, Northern Midwest and Southern California, metropolitan and rural areas Recruitment (1981‐1983, 19 months): trained staff interviewers screened records from 39 hospices (20 community‐based and 19 hospital‐based, both groups provided hospice home care and are hereby called community‐based intervention and hospital‐based intervention, respectively) and 14 conventional oncology care settings (outpatient clinics and oncology units) representing "good" oncology care according to knowledgeable area physicians (control, hereby called conventional care ‐ CC); Medicare and non‐Medicare patients aged ≥ 21 years with cancer confirmed by tissue diagnosis (except for brain and pancreatic cancer), remote metastasis (except for lung, brain and pancreatic cancer), presence of caregiver (this requirement excluded nursing home patients); additional criteria for controls KPS ≤ 50 (i.e. requiring assistance in daily activities) and poor life expectancy in the relatively short‐term (judged by referring physician) Number of patients (outcome sample): 1754 (833 community‐based intervention, 624 hospital‐based intervention, 297 control) Diseases: cancer (1754) Patient characteristics: median age band 65‐74 years; 52% female; 6.7% community‐based intervention, 16.1% hospital‐based intervention, 16.6% controls lived alone ("average" 11.7%) Number of caregivers (outcome sample): 1754 (833 community‐based intervention, 624 hospital‐based intervention, 297 control) Caregiver characteristics: not stated Deaths at end of study: 1754 (100%) Withdrawals/other drop‐outs: 4.4% with no differences among settings
Interventions	Home palliative care vs. usual care "Hospice care" Type: intermediate palliative care Service base: 20 community‐based hospices without inpatient beds (intervention1 ‐ HC), 19 hospital‐based hospices with inpatient beds (intervention2 ‐ HB) Team: medically supervised interdisciplinary teams composed of several individuals with appropriate skills including professionals and volunteers Responsibility: not stated (although team is medically supervised) Description: intervention grounded on hospice and humanistic philosophy of caring for terminally ill patients and their families (as per National Hospice Organization's 1979 standards of hospice programme of care); palliative, preventive, supportive, less resource‐intensive and with more informal care (family and volunteers) than CC; repertoire of services including palliative care and control of symptoms, physical, emotional and spiritual distress, with emphasis placed upon care in the home, individualised care plan and provision, choices about treatment and place of death; supportive care for the family during the dying process and bereavement (this may include follow‐up visits), maximum accessibility to care regardless time of day or ability to pay, and treating the patient and caregivers as a unit; inpatient and HC services closely integrated to ensure continuity and co‐ordination of care; care available 24/7; education programme (for patient, family and team) Duration: "average" length of stay ‐ intervention 1: 72 days and intervention 2: 63 days, half of all patients dead within 35 days of admission, nearly 20% stayed 10 days or less Control: CC "promotes aggressive care rather than palliative care; is directed by a physician with limited input from other professionals, focuses treatment on only the identified patient; generally ceases to provide services after death occurs; and varies greatly in quality depending upon the financial resources of the patient and the time at which services are most needed"
Outcomes	Quality of life (primary outcome) Death at home Patient at home as long as wanted Caregiver satisfaction with place of death Pain Symptom severity Psychological well‐being (emotional quality of life) Social functioning (social quality of life, hours of social visiting, chatting with household members) Spiritual well‐being Quality of death Awareness Physical function Survival Patient and caregiver satisfaction with care Caregiver pre‐bereavement outcomes (anxiety/depression, caregiver burden, increased drinking, medication use for anxiety and depression) Caregiver post‐bereavement outcomes (grief intensity, medication use for anxiety and depression) Assessment points: baseline, 7 days after and 14 days thereafter until death (last measure "on average" 7 days before death and penultimate measure approximately 21 days before death with no differences among settings); 90‐120 days after death for post‐bereavement outcomes Analysis: forward from enrolment, backwards from death and forwards from death
Resource use/costs	Institutional days Home nursing visits, home health/homemaker visits, social services/other therapies visits, physician and outpatient clinic visits Social services (general counselling, legal/financial counselling, paperwork assistance, help getting services, self care training) Aggressive interventions (radiotherapy, surgery, chemotherapy or hormonal therapy, thoracentesis) Diagnostic tests (blood tests, x‐rays, or scans) Respiratory support intervention (oxygen, respiratory therapy) Analgesics (for a subsample of 181 patients) Palliative radiation (for a sub‐sample of 2 clinical profiles) Informal care (hours of direct care help from primary care person) Caregiver post‐bereavement healthcare use (hospitalisation, number of physician's visits) and bereavement absenteeism from work Time horizon: from enrolment to death; last 3, 2 and 1 week before death Costs: inpatient, HC, physician visits, outpatient clinic visits, drugs, supplies and equipment expenditures, total costs per study day (included all the former) Currency: 1982 USD
Notes	Power considerations: "Sample sizes are predicated upon a 95% confidence limit, or at α = 0.05. The differences between samples are expressed as percentages of the mean on the QLI scale which ranges from 0 to 10. A difference of at least 15% on this scale will be accepted as a true difference. A sample size of 304 conventional care patients will be required to assure that 80% of the time we will be correct in assuming a true difference actually exists"; "a minimum of a 20% difference in the terminal care health costs associated with the two samples will be accepted as a true difference. At a desired power level of 0.8, a sample size of 404 patients in the conventional care group will be required" Considerations about hospice care sites (i.e. intervention): initial number of hospices involved was 40 but 1 hospital‐based hospice was excluded from all analyses since it provided virtually no HC service; 26 were demonstration sites (receiving Medicare demonstration waiver allowing payment for normally non‐covered services) but analyses of patterns of care and of quality of life showed no differences between demonstration and non‐demonstration hospices thus these were aggregated Only patients who died during the study period were included in analyses since outcomes were assessed in relation to proximity to death Self report data are from first and second patient interview rather than those interviews closest to death; most patients were unable to communicate as they approached death (e.g. patient self reports on pain could not be completed by 50% patients at 2 weeks before death and 80% at 1 week before death) thus caregiver reports were used
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Baseline measurement (CBA)	High risk	Significant differences in age, family income, living alone, caregiver employment, KPS (hospital‐based intervention group was more functionally impaired than community‐based intervention), hospitalisation in 2 months prior to study (control and hospital‐based intervention more often than community‐based intervention); analysis adjusted for differences in patient case mix
Characteristics of intervention and control providers (CBA)	Unclear risk	Limited information on providers' characteristics
Blinded assessment of outcomes (CBA)	Unclear risk	Blinding not stated; source of some information not stated; mix of self and proxy reports
Protection against contamination (CBA)	Low risk	Second sites used as controls
Reliable outcome measurement (CBA)	Unclear risk	Source of some information not stated (e.g. place of death); mix of self and proxy reports
Follow‐up of patients/caregivers (CBA)	Low risk	4.4% drop‐outs with no differences among settings

*Study characteristics*
Methods	Design: phase II RCT (patient and caregiver) Methodological quality: 3.5/6^a (high quality, EPOC RCT/CCT checklist)
Participants	Country and regions: UK, South East London, metropolitan area Recruitment (2004‐2005, 11 months): from local health and social professionals and in a few instances via voluntary groups and self referrals; patients with MS living in South East London, deemed (by clinicians) to have specialist palliative care needs (i.e. 1 or more unresolved symptoms, psychosocial concerns, end of life issues, progressive illness or complex needs); referrals screened by consultant in palliative medicine independent to study. Exclusion criteria: very urgent needs or rapid deterioration (immediate referral to service offered) Number of patients (randomised): 52 (26 intervention and 26 control) Diseases: MS (52) Patient characteristics: mean age 53 years; 69% female; 17.3% lived alone Number of caregivers (randomised): 52 (26 intervention and 26 control) Caregiver characteristics: not stated Deaths at end of study: 4 (1 in the intervention and 3 in the control) Withdrawals/other drop‐outs: 2 controls (1 protocol violation, 1 severely ill)
Interventions	Home palliative care vs. usual care (fast‐track design with a 12‐week wait) Type: specialist palliative care Service base: King's College Hospital (general public teaching hospital) Team: part‐time consultant in palliative medicine with training in neurology and special interest in MS, part‐time clinical nurse specialist (working in neurology and with some palliative care training), full‐time administrator; easy access to consultant neurologist through weekly multi‐professional MS clinic (in addition to telephone and email contact) to joint review by neurology and palliative care and other relevant professionals; team worked closely with the existing hospital palliative care team, using time from their psychosocial worker (on maternity leave for 7 months during project); weekly meetings to discuss caseload and for consultant to input into management of all patients Responsibility: not stated, "the service aimed to complement and not to replace existing services" Description: new short‐term intervention grounded in previous work (pre‐clinical and phase I comprising systematic reviews and qualitative studies, as per the Medical Research Council framework for the evaluation of complex interventions); referral form – referrers were asked to identify the main issues requiring team's involvement and record in a standard form with criteria based on palliative care need criteria and specific triggers identified by team (hydration and nutrition, advance directives, competency and consent); initial comprehensive assessment – conducted by team member including demographics, ability to communicate, main symptom issues, current medication, psychological concerns, social issues (including care package and agencies involved), caregiver concerns and advance care planning (this took an "average" 110 minutes of direct contact and 90 minutes of liaison/administration); action plan and follow‐up – following assessment, communicated to primary team and other professionals as appropriate, and followed through calls or visits depending on need; specialist welfare benefits advice; bereavement support; crisis prevention; education and support for primary and secondary care; patients were seen in their place of choice (often at home and sometimes at outpatient clinics, nursing home, hospital); 1‐3 contacts per patient (visits or telephone calls or both); referral to longer‐term specialist community palliative care – identification of specialist need by consultant based on WHO definition of palliative care (specifically physical symptoms, integration of psychological and spiritual aspects of care, need for ongoing support for patients and families, terminal care and bereavement support) and referral to local teams for those in need (12%) Duration: 1‐3 contacts Control: usual care, including community and hospital services (e.g. neurologists, MS nurses, district nurses, GPs, rehabilitation, neurological, social services) in first 12 weeks after enrolment, topped up with access to intervention from then onwards; a few patients received home physiotherapy, occupational therapy, specialist rehabilitation services, continence advice, psychological and psychiatric care; emergency access to intervention in case of very urgent needs or rapid deterioration
Outcomes	Palliative outcomes (primary outcome) Pain Symptom burden Physical and psychological disease impact Caregiver pre bereavement outcomes (burden, mastery, positivity) Assessment points: baseline, 6 and 12 weeks after (primary analysis), 18 weeks (only controls – after receiving intervention) and 24 weeks after Analysis: forwards from enrolment
Resource use/costs	Formal care (included a range of health, social and voluntary services – inpatient care, respite care, day centre, contacts with district/practice, MS nurse, palliative care nurse, other nurse, general practice, specialist at home, in hospital, in a ward and in other places, occupational therapist, physiotherapist, dietician, chiropodist, dentist, speech therapist, social services, day centre, inpatient care, respite care); informal care (hours of care) Time horizon: 12 weeks following enrolment Costs: mean costs for each of the categories above; difference in total costs per patient: a) including formal and informal care, and b) excluding inpatient and informal care as above Currency: 2005 GBP Cost‐effectiveness: cost‐effectiveness planes for palliative outcomes (primary outcome) and caregiver burden
Notes	Power considerations: "We estimated that a sample of more than 25 patients in each arm would enable us to detect differences of >2 on the POS‐8 at P < 0.05, power 80% (with a standard deviation [SD] of 2.25) at 12 weeks"
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	Low risk	"The researcher e‐mailed relevant data to independent statisticians who conducted the randomization using the minimization method to give an equal balance of gender, age, date of diagnosis, and according to whether patients could or could not communicate. (...) The statistician informed researchers who then informed patients of their allocation"
Follow‐up (RCT / CCT)	Low risk	46/52 (89%) at study end
Blinding (RCT / CCT)	High risk	"We were unable to blind the interviewers or participants from group allocation"
Baseline measurement (RCT / CCT)	Low risk	No significant baseline differences: gender, age, ethnicity, type of MS, education, caregiver characteristics, functional status and outcomes (analysis adjusted for baseline scores)
Reliability of outcome measurement (RCT / CCT)	Unclear risk	Face‐to‐face interview with patients using standard measures; self complete questionnaires with caregivers
Protection against contamination (RCT / CCT)	High risk	Patients/caregivers randomised

*Study characteristics*
Methods	Design: RCT (caregiver) Methodological quality: 2.5/6^a (EPOC RCT/CCT checklist)
Participants	Country and regions: Australia, Melbourne, Victoria, metropolitan area Recruitment (date not stated, 12 months): from 2 home palliative care services (similar annual admissions and similar to national "average" length of stay of 75 days); new patient referrals (within 1 week) with ECOG Performance Status from 0 to 3, i.e. at least capable of minimal self care; up and about more than 50% of waking hours (criteria to minimise attrition). Exclusion criteria: no caregiver living with patient, patient with intellectual or psychiatric illness or lack of English language skills Diseases: cancer (106) Number of caregivers (randomised): 106 (54 intervention and 52 control) Caregiver characteristics: mean age 60.78 years; 65.1% female Deaths at end of study: not stated Withdrawals/other drop‐outs: 34 intervention and 27 control (non‐response to questionnaire)
Interventions	Reinforced vs. standard home palliative care Additional component: "psychoeducational intervention" Type: specialist palliative care Service base: 2 home palliative care teams (base not stated) Team: additional component delivered by nurse; standard home palliative care team comprised specialist nurses, doctors, and allied health professionals (similar range of disciplines in each service, including nurses, social workers, medical consultants, pastoral care workers, volunteers and bereavement counsellors) Responsibility: for additional component ‐ nurse; for standard home palliative ‐ team's nurse (not clear if responsibility rested with primary physician/team) Description: additional component grounded on literature and focus groups with nurses and caregivers (current and bereaved); aimed to top up home palliative care with a psychoeducational component to enhance support and guidance for caregivers – this consisted of 2 fortnightly home visits with follow‐up call between visits, structured around and complemented by audiotape with reflections from caregivers and self care strategies and structured relaxation exercise, and guidebook with information on typical aspects of caring for dying person (link provided in paper); caregivers read guidebook sections and noted questions prior to visits ‐ first visit prepared for caregiver role and aspects of caregiving (section 1 and 2 of guidebook); telephone call focused on evaluating plans, self care (section 3), new issues and caregiver desire to go on to next section; second visit focused on evaluating previous strategies, new issues and preparing for dying phase (final section of guidebook) Duration: additional component lasted 2 weeks; standard home palliative care likely to be provided from enrolment to death (length not stated) Control: specialist home palliative care including information and access to services outside the team and spiritual guidance; emergency visits from nurses in addition to pre‐scheduled home visits from team members; 24/7 phone advice
Outcomes	Caregiver pre‐ and post ‐bereavement outcomes (preparedness, competence, self efficacy, caregiving rewards, anxiety) Assessment points: baseline, 5 weeks after (questionnaire posted 28 days after baseline), and 8 weeks after death Analysis: forwards from enrolment and forwards from death
Resource use/costs	Not assessed
Notes	Power considerations: "There was insufficient power to detect differences based on a simultaneous comparison of data obtained at all three assessment times (...) only moderate to large effects were likely to be found."; "Owing to time and financial constraints, sample size was not calculated on the basis of a predetermined power and effect size"; "Cohen's power tables were used to determine the power associated with each statistical test based on the pre‐specified effect and sample size (alpha levels were set at 0.05; all tests were two tailed). Power values ranged from 0.70 (t‐test for two independent samples) to 0.99 (t‐test for related samples)"
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	Low risk	"Randomization to group occurred via a computer‐generated software system"
Follow‐up (RCT / CCT)	High risk	40 intervention (74%) and 35 control (67%) follow‐up 5 weeks after baseline; 20 intervention (37%) and 25 control (48%) 8 weeks after death
Blinding (RCT / CCT)	Unclear risk	Blinding not stated
Baseline measurement (RCT / CCT)	Unclear risk	"Chi‐squared tests and t‐tests for independent samples were used to assess the possibility of pre‐existing differences between the standard care and intervention groups based on baseline data"; no results stated, descriptive data not shown
Reliability of outcome measurement (RCT / CCT)	Unclear risk	Collected from caregivers by research assistants at baseline and 8 weeks after death (home); self completed by caregivers 5 weeks after baseline (mail)
Protection against contamination (RCT / CCT)	High risk	Patients randomised

PERMALINK

Effectiveness and cost‐effectiveness of home palliative care services for adults with advanced illness and their caregivers

Barbara Gomes

Natalia Calanzani

Vito Curiale

Paul McCrone

Irene J Higginson

Maja Brito

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Summary of findings

Summary of findings 1. Summary of findings.

1.1. Analysis.

1.2. Analysis.

1. Symptom burden: home palliative care versus usual care.

2. Pain: home palliative care versus usual care.

3. Physical function: home palliative care versus usual care.

4. Quality of life: home palliative care versus usual care.

5. Caregiver burden: home palliative care versus usual care.

6. Caregiver grief: home palliative care versus usual care.

7. Satisfaction with care: home palliative care versus usual care.

Summary of findings 2. Summary of findings.

8. Emergency department use.

9. Cost‐effectiveness analyses with total care costs.

Background

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

1. Primarily for patients with a severe or advanced disease (malignant or non‐malignant), no longer responding to curative/maintenance treatment or symptomatic (or both), or their family caregivers, or both

2. Aiming to support patients or family caregivers, or both, outside hospital and other institutional settings as far as possible and to enable patients to stay at home

3. Providing either specialist or intermediate palliative/hospice care, as defined in a previous systematic review

4. Providing comprehensive care and aiming at different physical and psychosocial components of palliative care

Types of outcome measures

Primary outcomes

Secondary outcomes

Economic data

Search methods for identification of studies

Electronic searches

Searching other resources

Handsearching

Reference lists

Correspondence

Language

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of methodological quality of included studies

Measures of treatment effect

10. Levels of evidencea.

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Results

Description of studies

Results of the search

1.

Included studies

Design

Sample sizes

Setting

Participants

Intervention

Home palliative care

11. Nineteen models of specialist and intermediate home palliative care.

Reinforced home palliative care

12. Four models of reinforced home palliative care.

Controls

Outcomes

Economic data

10. Levels of evidence^a.

*Study characteristics*
Methods	Design: cluster RCT (patient and caregiver) Methodological quality: 3.5/6^a (high quality, EPOC RCT/CCT checklist)
Participants	Country and regions: Norway, Trondheim, 6 metropolitan and 2 rural areas Recruitment (1995‐1997, 32 months): from hospital departments, particularly gastrointestinal surgery, urological surgery, oncology and gynaecology, lung and gastrointestinal departments of internal medicine (through professionals, lead nurse and researchers screening visits once or twice per week) and primary care professionals in 8 healthcare districts; patients aged > 18 years with incurable cancer, life expectancy 2‐9 months (not stated who estimated) living in 8 community healthcare districts (clusters) stratified into 3 pairs according to number of inhabitants > 60 years and urban/rural status (2 smaller urban districts merged with larger ones) and randomised to intervention or control. Exclusion criteria: haematological malignant disorders other than lymphomas, patient not informed about their diagnosis and that the disease was incurable, participation in other trials with quality of life as outcome, unable to fill in questionnaire Number of patients (randomised): 434 (235 intervention and 199 control) Diseases: cancer (434): gastrointestinal (181), lung (52), breast and female genitals (67), prostate and male genitals (41), kidney or vesica (29), lymphomas (13), skin (12), others (39) Patient characteristics: median age 70 years intervention, 69 years control; 47% female; 32.5% lived alone Number of caregivers (consented): 313 (183 intervention and 130 control) Caregiver characteristics: median age 57.6 years intervention, 53.9 years control; 68.3% female Deaths at end of study: 395 (219 intervention (93%) and 176 control (88%)) Withdrawals/other drop‐outs: 6 intervention and 10 control (reasons not stated)
Interventions	Home palliative care vs. usual care Type: specialist palliative care Service base: palliative medicine unit at University Hospital of Trondheim (12 beds, outpatient clinic and consultant team in and out of hospital) Team: 1 full‐time physician; 2 palliative care nurses, social worker, priest, nutritionist, part‐time physiotherapist; staff worked daytime hours only; weekly meetings Responsibility: consultant nurse was the care co‐ordinator; primary family physician and community nurse maintained as main professional carers Description: new intervention grounded on holistic philosophy, included multidisciplinary approach to the patient's needs and co‐ordination of care; consultant team who co‐ordinated care; initial care plan set up in meeting of patient, caregivers, family physician, community nurse, and consultant nurse or physician from team; follow‐up routine consultations by community staff supervised by team (available for joint home visits); no additional out of hours or on‐call service; service linked the palliative medicine unit to community services under pre‐defined guidelines to keep interaction at optimum level; education programme (including bedside training and 6‐12 hours of lecturers every 6 months on symptoms and difficulties in palliative care); inpatient and outpatient services provided at unit except when required for medical reasons (e.g. surgery or acute intensive care for intercurrent disease) Duration: from enrolment to death or study end (median survival 99 days in the intervention group) Control: usual care shared among University hospital departments and community services according to diagnosis and medical needs (no well‐defined routines existed and poor communication between levels of services had been addressed as general problem); despite minor variations, community services were similar in all districts, including family physicians, HC nursing (nurses and nurse‐assistants) and nursing homes (when extensive need for attention or basic nursing care but not specialised hospital care is required); all except smallest urban district had 24 hours HC; limited night coverage (generally only short visits or telephone consultations); HC and nursing home services co‐ordinated at district level (common community nursing office ‐ decisions based on patient's medical and social needs and current availability of community resources); no specialist palliative care service available other than the intervention
Outcomes	Quality of life (primary outcome) Physical symptoms (pain – primary outcome; breathlessness, nausea/vomiting, constipation, diarrhoea, sleep disturbance, fatigue, appetite loss) Physical function (primary outcome) Psychological well‐being (primary outcome; emotional functioning and psychological stress) Death at home (primary outcome) Death in hospital Death in hospital palliative care unit Death in nursing home Percentage of time spent at home General health and vitality Social functioning Survival Caregiver satisfaction with care Caregiver pre‐bereavement outcomes (quality of life, general health, vitality, physical, psychological and social functioning, bodily pain) Caregiver post‐bereavement outcomes (quality of life, general health, vitality, physical, psychological and social functioning, bodily pain, grief intensity) Assessment points: baseline and monthly until death (or up to 2 years to avoid research burden); satisfaction with care measured 1 month after death; caregivers assessed 1, 3, 6 and 13 months after death Analysis: forwards from enrolment, backwards and forwards from death (caregivers' outcomes)
Resource use/costs	Hospitalisations (admission, number of admissions, mean length of admission, inpatient days, proportion of time in hospital) Care in nursing homes (admission, number of admissions, nursing home days, proportion of time in nursing homes) Time horizon: from enrolment to death or study end; last month before death
Notes	Power considerations: "Pre‐planned sample size was 200 patients in each group, based on quality of life."; "Because there was uncertainty about the likely difference, the impact of a variety of effect sizes was explored. Between 50 to 75 patients per group would be realistic to detect differences as small as 0.5 SDs in an ordinary randomized clinical trial (RCT), an effect size that is commonly regarded as a moderate change, and for which the EORTC QLQ‐C30 scores are roughly comparable to a change of 10 units. This represented a plausible and realistic effect of the intervention policy. However, for a cluster randomized design, the statistical power will be reduced because of within‐cluster correlation. This can be taken into account by increasing the total number of clusters, or to a lesser extent, by increasing the number of subjects per cluster. In this trial, for practical and economical reasons, participation had to be restricted to the clusters (health care districts) located within close reach from the palliative medicine unit, and it was decided to include 200 patients in each treatment group" Authors discuss issues related to recruitment attrition, and compliance in a methodological paper.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	Unclear risk	"Cluster randomised design"; unit of allocation was health district but random process not explicit
Follow‐up (RCT / CCT)	Unclear risk	Patients: 153 intervention (65%) and 116 control (58%) follow‐up at 1 month on questionnaire‐based measures (including primary outcomes); 219 intervention (93%) and 176 control (88%) on place of death Caregivers: 113 (49%) intervention and 70 control (36%) 1 month after death
Blinding (RCT / CCT)	Unclear risk	Not stated
Baseline measurement (RCT / CCT)	Unclear risk	Significant differences in diagnosis (intervention more gastrointestinal/lung, less breast/prostate/kidney/vesica/lymphoma/skin), housing (intervention higher proportion living in villa/apartment), access to informal help (intervention higher) and receipt of HC nursing (intervention lower); unclear if all are in same direction as results. No significant baseline differences in quality of life (1 of primary outcomes)
Reliability of outcome measurement (RCT / CCT)	Unclear risk	Questionnaire‐based measures were self completed; source of data on place of death not stated
Protection against contamination (RCT / CCT)	Low risk	Health districts randomised

*Study characteristics*
Methods	Design: cluster RCT (patient) Methodological quality: 4/6^a (high quality, EPOC RCT/CCT checklist)
Participants	Country and regions: US, Vermont, rural area Recruitment (date not stated, "four‐year study"): from radiotherapy and medical oncology clinics at Vermont Regional Cancer Center; new patients with cancer (histologically confirmed) aged ≥ 16 years with life expectancy from 3 months to 1 year (judged by experienced oncologist) living in Vermont counties; counties randomly allocated to intervention and control (number of counties not stated, paired based on population density, distance from centre, socioeconomic status, local medical facilities, referral patterns and local social services). Exclusion criteria: patients mentally incompetent Number of patients (allocated): 199 (98 intervention and 101 control) Diseases: cancer (199) Patient characteristics (primary outcome sample of 83 patients): mean age 58.7 years intervention, 56.1 years control; 42% female Deaths at end of study: 139 (70%) Withdrawals/other drop‐outs: not stated
Interventions	Home palliative care vs. usual care "Intensive home care" Type: intermediate palliative care Service base: Vermont Regional Cancer Center (department not stated) Team: trained oncology nurse practitioners with "extensive experience in care of the patients with advanced cancer" backed up by usual multidisciplinary team of nurses, medical and radiation oncologists, dieticians, enterostomal therapist, physiotherapists, occupational therapists, psychiatrists, social workers and clergymen at the cancer centre; formal interdisciplinary training on extended assessment skills (sophisticated approach to physical, psychological, nutritional and financial assessments of both patients and families); weekly team conferences Responsibility: team's nurse (acting as link between the multidisciplinary team and care co‐ordinator responsible for following patients through home visits, clinic visits and hospitalisations to ensure continuity of care; primary physician informed and collaborating in complex medical problems) Description: intervention designed to improve patient's quality of life and communication between the conventional cancer rehabilitation team and the patient by assigning oncology nurses to closely monitor and co‐ordinate patient care; assumed nurses would anticipate and solve medical and psychosocial problems and that their expertise would improve quality and extend survival; initial interdisciplinary assessment (by each team member) followed by care plan build by nurse; frequency of home visits defined based on prognosis (life expectancy < 3 months – biweekly; ≥ 3 months – monthly); home visits (35‐40 minutes) focused on patient needs ("incidental interactions with family"); physical care, talking with patient about illness and implications, mobilising family and social resources and co‐ordinating with patient's physician; protocol for pain management (in varying degrees of pain), nausea, vomiting and administration of intravenous chemotherapy drugs at home (available from authors); pain killers given regularly to ensure continuous control Duration: not stated but likely to be from enrolment to death or study end (mean survival in intervention group was 345 days) Control: usual care provided by multidisciplinary team at cancer centre and patient's local/private physician
Outcomes	Pain (primary outcome) Physical activity/function Nutrition Psychological well‐being (optimism) Survival Assessment points: baseline and then timed with nurse visits (biweekly for patients with life expectancy < 3 months; monthly for those with ≥ 3 months) Analysis: backwards from death
Resource use/costs	Hospital inpatient days Time horizon: not stated
Notes	Powerconsiderations: none stated Potential for measurement bias: less frequent assessments for patients with longer life expectancy (assessment timed with nurse visits, frequency of visits decided based on patient estimated life expectancy)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Concealment of allocation (RCT / CCT)	Unclear risk	"paired counties were randomly separated into two groups"; process not stated
Follow‐up (RCT / CCT)	Unclear risk	38 intervention (39%) and 45 control (45%) follow‐up (minimum of 3 time points on pain backwards from death, primary outcome); not stated for other outcomes
Blinding (RCT / CCT)	Unclear risk	Not stated
Baseline measurement (RCT / CCT)	Low risk	"A comparison of patient characteristics for these two groups demonstrated similarities in cancer diagnosis, sex, age, social class, and religious preference. The initial, on‐study scores on the CMI, I‐E, and KPS did not differ significantly"; "When pain levels (…) were compared over time, in 30‐day periods before death, the two groups had essentially the same mean pain scores until the last 90 days before death"
Reliability of outcome measurement (RCT / CCT)	Unclear risk	"by trained independent raters, using structured interviews in the patients' homes, done at the same frequency as the nurses' visits, which were based upon the patients' prognosis. Follow‐up interviews were carried out in the patients' homes because earlier data had indicated greater validity of home vs. clinical evaluations for some measures. Initial and follow‐up rater training took place during the pilot phase of the project using videotapes. At the end of the first year the raters had achieved acceptable interrater agreement on major scale items (Pearson correlation coefficient 0.70)"
Protection against contamination (RCT / CCT)	Low risk	Counties randomised

*Study characteristics*
Methods	Design: ITS with nested CBA Methodological quality: 5/8^a (EPOC ITS checklist); 3.5/6^a (EPOC CBA checklist)
Participants	Country and regions: UK, Trent and Yorkshire health regions, metropolitan and rural areas Recruitment (date not stated): all deaths from neoplasms (ICD 140 to 209) in 3 years before and 3 years after 8 Macmillan HC nursing services started (identified from Office of Population, Censuses and Surveys Classification of Surgical Operations and Procedures); CBA included patients newly referred to the 8 services during 1 year or until 100 patients were enrolled Number of patients: 40,072 for ITS (17,864 in 3 years before and 22,208 in 3 years after the introduction of the 8 services); 957 for CBA Diseases: cancer (40,072) for ITS; not stated for CBA but likely to be all cancer as Macmillan nurses provide care to people with cancer Patient characteristics: not stated Deaths at end of study: 100% (ITS); not stated for CBA Withdrawals/other drop‐outs: none stated
Interventions	Home palliative care vs. usual care "Macmillan home care nursing care" Type: intermediate palliative care Service base: 4 inpatient hospice; 4 non‐hospice base (1 hospital, 1 non‐inpatient hospice, 1 in grounds of future in‐patient hospice, 1 in primary care community team offices) Team: members ranged from 2 to 8; 7 teams were multidisciplinary and 1 was of nurses only; all had nurses (from 81 to 182 patients per nurse/year for hospice based services vs. from 33 to 70 for non‐hospice based), 7 had medical input, 5 had social workers; 2 had case conferences (1 daily, 1 weekly) Responsibility: not clear although services' aim was to augment but not replace care from district nursing, GPs and families (primary physician/team likely to be in charge) Description: services varied in provision but also had common components; all were in operation for ≥ 18 months; mean from 2.3 to 7.7 home visits per patient/month for hospice‐based services and mean from 7.9 to 12.5 for non‐hospice based) – visits aimed to monitor current situation, reassure/support, monitor pain and other symptoms, assess need for community care, inpatient or day unit, give advice, deliver equipment/ medication, practical nursing care; arrangements for out‐of‐hours crisis ‐ all had 24/7 coverage (e.g. hospice out of hours, bleeps, radio paging, nursing message systems and nurses home number, coverage by inpatient hospice staff for advice); out of hours visits ranged from 0 to 9 except in 1 hospice‐based service (40 visits) and in 1 non‐hospice based service (81 visits); patients from hospice‐based services could access hospice beds, day centres, family relief and other hospice services Duration: from enrolment to death (median 32 days, range 21.5‐44.5 days across services) Control: usual care including district nursing and care from GPs
Outcomes	Death at home Preference for place of death met
Resource use/costs	Terminal institutional days (before death in hospital or hospice) Time horizon: variable (identified backwards from death)
Notes	Power considerations: none stated For 2 of the 8 services, the authors could only provide 2 or 1 year data prior to their introduction; as this did not meet our ITS inclusion criteria we analysed outcome data only for the remaining 6 services (this is referred to in the review as the Trent and Yorkshire analysis, with 31,890 deaths) The authors collected other outcome data on a small sample (40 patients): on pain, activities of daily living, quality of life, patient preference for place of death and satisfaction with care ‐ but did not use these to compare services
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Baseline measurement (CBA)	Unclear risk	Characteristics not reported
Characteristics of intervention and control providers (CBA)	Unclear risk	Providers' characteristics in text but no data presented
Blinded assessment of outcomes (CBA)	Unclear risk	Source of information on place of death not stated
Protection against contamination (CBA)	Low risk	Second sites used as controls (non‐hospice‐based services)
Reliable outcome measurement (CBA)	Unclear risk	Source of information on place of death not stated
Follow‐up of patients/caregivers (CBA)	Unclear risk	Not stated if some were still alive
Intervention dependent of other changes (protection against secular trends) (ITS)	Unclear risk	Not specified
Data analysed appropriately (ITS)	High risk	Data described only (no ARIMA models or time series regression models used)
Reason for number of points pre‐ and post‐intervention (ITS)	High risk	Rationale for choice of 3 years pre‐ and post‐intervention not given; sample size calculation not reported
Shape of intervention effect specified (ITS)	Unclear risk	Rationale for shape of intervention effect not reported
Intervention likely to affect data collection (protection against detection bias) (ITS)	Low risk	Sources and methods of data collection were the same before and after intervention
Blinded assessment of outcomes (ITS)	Low risk	Objective outcome variable (using standard and official statistics)
Completeness of data set (ITS)	Low risk	100% of total number of deaths in study
Reliable outcome measurement (ITS)	Low risk	Outcome obtained from standard and official statistics

Study	Reason for exclusion
Addington‐Hall 1992	Intervention does not meet criteria for palliative care; co‐ordinators did not provide "practical nursing care" or "specialist palliative care advice"; co‐ordination only
Brumley 2003	CBA not meeting criteria, conducted in only 1 intervention site and 1 control site
Feldman 2011	Intervention focused on staff education and mentoring with the main goal of increasing hospice admissions; not specifically aimed to support patients or caregivers at home
Hebert 2006	Televisits provided in addition to standard home palliative care; additional component procedural only and non‐comprehensive addition to care
Kane 1984	Intervention focused on inpatient care with a component of home care; however, inpatients and patients at home are shown together (both part of the hospice group); no subanalysis and frequent transits between inpatient and home care mentioned
McCorkle 2000	Study conducted with post‐surgical patients and only a few died, thus population was not considered terminal/advanced
Morita 2012	Study designed to be a cluster RCT but authors decided to adopt a mixed‐method approach because "1) the intervention itself should be applied to all populations over the country and clear distinction between intervention and control groups is difficult, 2) a concealment problem is likely to occur, and 3) the most important mission at a national level is not to clarify if one specific intervention actually changes outcomes, but to obtain comprehensive insights into how to disseminate palliative care throughout the country". The resulting study is an ITS that does not meeting the criteria for our review, where data collected in the two years prior the program was implemented (2007‐2008) was compared to data collected in the two years after program implementation (2010‐2011)
Smeenk 1998a	Intervention does not meet criteria for palliative care; "care specially tailored to meet his [patients] individual needs, and provided by professional caregivers from primary and hospital teams"
Temel 2010	Intervention focused on outpatient meetings with patients at a large academic medical centre; not specifically aimed to support patients or caregivers at home

Methods	Design: CCT (awaiting confirmation from authors)
Participants	Country and regions: South Korea (included regions to be confirmed by authors) Recruitment: to be confirmed by authors Number of patients: 46 (24 intervention and 22 control) Diseases: cancer (46): digestive system (22), lung (11), liver, biliary or pancreatic (10), others (3) Patient characteristics: mean/median age not stated, 59% male, 76% married Deaths at end of study: to be confirmed by authors Withdrawals/other drop‐outs: to be confirmed by authors
Interventions	Type: to be confirmed by authors Service base: to be confirmed by authors Team: to be confirmed by authors Responsibility: to be confirmed by authors Description: several components for patients, caregivers or both: pain and symptom control (drug and non‐drug therapy, care co‐ordination); empowering (providing information, promoting control of emotions and supporting patient and family); finding the meaning (thinking about life and facing death); and networking (making and maintaining relationships, connecting and referring). Duration: to be confirmed by authors Control: to be confirmed by authors
Outcomes	Pain Symptom burden (decreasing symptom experience) Psychological well‐being (improving psychological QoL) Caregiver pre bereavement social well‐being (improving family‐related and social QoL) Spiritual well‐being (relieving spiritual pain and improving spiritual QoL) Caregiver burden Assessment points: to be confirmed by authors Analysis: to be confirmed by authors
Notes

Study name	Palliative Care Trial (PCT)
Methods	Design: cluster RCT (patient and general practitioner)
Participants	Country and regions: South Australia Recruitment (2002‐2004, 26 months): every patient referred to SAPS was screened by a triage nurse who assessed eligibility. Inclusion criteria: patients mentally competent at enrolment as documented by an MMSE score ≥ 24, or who had a GP‐identified caregiver or legal healthcare proxy who could provide informed consent. Patient and subsequent GP consent were further required for enrolment and randomisation. Exclusion criteria: patients not living within the region served by the palliative care services or expected to die within 48 hours of referral. Number of patients: pending for full sample Diseases: pending for full sample Patient characteristics: pending for full sample Deaths at end of study: pending for full sample Withdrawals/other drop‐outs: pending for full sample
Interventions	"GP Educational Outreach visiting in addition to standard palliative care" (intervention 1); "Structured patient and caregiver educational outreach visiting in addition to standard palliative care" (intervention 2) "Case conferencing in addition to standard palliative care" (intervention 3) vs. "standard palliative care"' "Palliative Care Trial (PCT)" Type: specialist palliative care Service base: organised as regional whole‐population networks with consultative specialist medical, nursing and allied health support for GPs and community nurses Team: at least the GP, a palliative care nurse and a trained educator (GP) employed for the study, plus 1 educator for the GP intervention (intervention 1); weekly palliative care team meetings. Medical specialists from other disciplines, district nurses, domiciliary care representatives, social worker, volunteers, pharmacists and pastoral care workers were also available according to patients' needs. Responsibility: not clear Description: intervention 1: GP Educational Outreach Visiting in addition to standard palliative care. Educational sessions took place in GP's office, 2 x 20‐ to 30‐minute sessions with GPs 2‐4 weeks apart. Evidence‐based messages on palliative care pain management derived from structure literature review. Key message focused on knowledge and attitude deficits. Intervention based on the "educational outreach visiting" or "academic detailing" model that "demonstrated to change physician practice". Intervention 2: standard palliative care plus structured patient and caregiver educational outreach visiting. Trained educator conducted 2 x 30‐ to 40‐minute sessions with patient with or without their caregivers in place chosen by patients. Evidence‐based messages on palliative care pain management derived from structure literature review. Key message focused on knowledge and attitude deficits. Intervention derived from a blend of "patient coaching" and "educational outreach visiting". Intervention 3: case conferencing in addition to palliative care. Used a case conferencing model, minimally including the patient, the caregiver, or both, the GP and the palliative care nurse. Other professionals took part based on patients' needs. Conferences organised by the palliative care nurse and conducted within 28 days of randomisation. Agenda set by patients and caregivers by identifying functional, physical, or emotional goals and concerns. GPs paid for participation. Duration: not clear Control: described as standard palliative care
Outcomes	Place of death Pain Symptom burden Physical function Quality of life Assessment points: from baseline until death Analysis: forwards from enrolment
Starting date	April 2002
Contact information
Notes	Both intervention and control groups received standard palliative care

Study name	Promoting Effective Advance Care for Elders (PEACE) study
Methods	Design: RCT (patient)
Participants	Country and regions: Ohio, United States Recruitment (pending): New PASSPORT (Ohio's community‐based, long‐term care Medicaid waiver program) enrollees randomised either to the intervention or the usual care. Inclusion criteria: > 60 years old who passed a mental status screening (the Mental Status Questionnaire) and had one of the following: congestive heart failure and being actively treated (American Heart Association stage C); chronic obstructive pulmonary disease and on home oxygen; diabetes with renal disease, neuropathy, visual problems, or coronary artery disease; end‐stage liver disease or cirrhosis; cancer (active, not history of) except skin cancer; renal disease and actively receiving dialysis; Amyotrophic Lateral Sclerosis with history of aspiration; Parkinson’s disease stages 3 and 4; or pulmonary hypertension. Exclusion criteria: active alcoholics (i.e., those who drink >2 drinks per day "on average") and illegal substance users as well as clients who have schizophrenia or are psychotic; consumers who could not pass the Mental Status Questionnaire and those were already enrolled in hospice Number of patients: pending for full sample (80 in pilot study; 40 intervention and 40 usual care) Diseases: pending for full sample Patient characteristics: pending for full sample Deaths at end of study: pending for full sample Withdrawals/other drop‐outs: pending for full sample
Interventions	Intervention ['Promoting Effective Advance Care for Elders (PEACE)'] versus usual care Type: pending Service base: pending Team: PASSPORT care manager, interdisciplinary team (details pending) and primary care physician Responsibility: not clear Description: Intervention based on Wagner's Chronic Care Model and includes comprehensive interdisciplinary care management for frail elders with chronic illnesses. It emphasises patient activation, and integrates with community‐based long‐term care and other community agencies. Intervention involves an in‐home geriatrics/palliative care needs assessment by a trained PASSPORT case manager. The case manager presents the findings to an interdisciplinary team who develops and individualised care plan based on the consumer’s goals and best practice guidelines. The case manager then implements the plan with the consumer. Family and primary care physician. Duration: not clear Control: described as usual care
Outcomes	Symptom management Quality of life Mood Decision making/care planning Spirituality Assessment points: pending Analysis: pending
Starting date	pending
Contact information
Notes

Study name	Palliative Network for Severely Affected Adults with MS in Italy (PeNSAMI) study
Methods	Design: RCT
Participants	Country and regions: Northern, Central and Southern Italy Recruitment (pending): patients with MS, details pending Number of patients: pending Diseases: MS Patient characteristics: pending Deaths at end of study: pending Withdrawals/other drop‐outs: pending
Interventions	Palliative care integrated with standard MS care versus standard MS care alone Type: pending Service base: pending Team: pending Responsibility: pending Description: Contents of the intervention will be obtained from a Phase I study composed of individual semi‐structured interviews of adults with severe MS, three focus group meetings with caregivers of people with severe MS and two focus group meetings with health professionals caring for MS patients. A phase II RCT will follow to examine the acceptability, feasibility and effectiveness of the intervention. In order to cover a range of experiences, participants will be selected from Northern, Central and Southern Italy. Duration: pending
Outcomes	Pending Assessment points: pending Analysis: pending
Starting date	Phase II RCT planned for 2013
Contact information
Notes

Study name	Hospice rapid response community service
Methods	Design: cluster RCT
Participants	Country and regions: East Kent, England Recruitment (pending): All new referrals to the hospice who are assessed by a member of the hospice team during the study period are potentially eligible for inclusion in the study, but only those referred who die within the intervention or control period will be included in the analysis. Number of patients: pending (441 patients needed according to sample size calculations) Diseases: pending Patient characteristics: pending Deaths at end of study: pending Withdrawals/other drop‐outs: pending
Interventions	Rapid response service ('Hospice rapid response community service') plus usual care vs usual care only fast‐track design applied to three centres (intervention rolled out sequentially in the three centres with six months between the start of provision in each site) Type: specialist palliative care Service base: hospice‐based (three hospice centres with inpatient beds, community outreach service and day centre) Team: Rapid response teams (details below) Responsibility: pending Description: Intervention provided by rapid response teams was developed in line with best practice and following a literature review, Its main features are: 1) available to patients in their homes (or care homes); 2) has a robust ‘hospice standard’ assessment which takes account of: patient preferences, caregiver/family preferences, patient needs, and patient prognosis; 3) provides hands on care; 4) responds rapidly to crises using human and material resources available 24/7 with access to health care assistants, service coordinator, palliative care nursing, medical advice, and small pieces of equipment which can be carried by car; 5) works in coordination with other community services Duration: pending
Outcomes	Dying in the preferred place (primary patient outcomes) Preferences and change of preferences for place of death Actual place of death Caregiver quality of life (primary caregiver outcome) Caregiver anxiety and depression (HADS) Caregiver health status (EQ‐5D) Caregiver demand Satisfaction with care (caregiver) Service utilisation costs (including intervention costs such as staff time, mileage travelled to patient homes and consumables; general practice and community resources, outpatient, inpatient stays, out‐of‐hours service, Marie Curie nurse visits, and social care packages) Assessment points: baseline until death, timepoints not stated (patients); baseline and eight months after enrolment, and six to eight months after the patient died (caregiver) Analysis: pending
Starting date	pending
Contact information
Notes

Study name	TRAPADO study
Methods	Design: CCT
Participants	Country and regions: Lyon, France Recruitment (2000‐2002, 24 months): Inclusion criteria: ≥ 18 years, cancer diagnosis, prognosis (estimated by clinician) of < 3 months, plus 2 of the following 3 criteria: performance status > 2, LDH levels higher than normal, albumin < 30 g/L. Participants also needed to be living in the Lyon area for home care intervention and have someone available to answer the questionnaires. After being informed of their prognosis, patients were invited to take part in the study by their oncologist, hospital doctor or the home care physician co‐ordinator. Informed consent was done in writing and patients expressed their choice of cohort (TRA, PA or DO), knowing that a change was possible at any time Number of patients: 52 (at the time of interim analysis ‐ aimed to have 100 patients, this was achieved in 2005 but complete data not yet published) Diseases: cancer (52) Patient characteristics (interim analysis): median age 61 years, 56% male Deaths at end of study: 46 (at the time of interim analysis) Withdrawals/other drop‐outs: 1 (at the time of interim analysis)
Interventions	Standard hospitalisation (TRA), Palliative Care Unit (PA) (presented together) vs. home care or hospitalisation at home (DO) 'TRAPADO study' Type: pending Service base: hospital‐based Team: the project is co‐ordinated by a nurse or a psychologist, further information pending Responsibility: pending Description: main objective to measure the quality of care and quality of life in palliative patients in the 3 cohorts. The goal was to examine the settings instead of comparing them. The secondary objective of the study was to describe the patient's choices, the characteristics of the 3 populations and the impact on families Duration: pending
Outcomes	Quality of care Quality of life Anxiety Depression Pain General health Assessment points: baseline, 15 days after baseline, 30 days after baseline and every month until death, for a maximum of 4 months Analysis: pending
Starting date	2002
Contact information
Notes