A Systematic Review of Nursing Home Palliative Care Interventions: Characteristics and Outcomes

Abstract

Background:

Despite recommendations to integrate palliative care into nursing home care, little is known about the most effective ways to meet this goal.

Objective:

To examine the characteristics and effectiveness of multicomponent nursing home palliative care interventions.

Design:

Systematic review

Methods:

We searched MEDLINE via PubMed, Embase, CINAHL, and Cochrane Library’s CENTRAL for clinical trials from inception through January 2019. We assessed the type of intervention, outcomes, and the risk of bias in included studies.

Results:

We screened 1167 records for eligibility and included 13 articles. Trials used usual care and or an active comparator as the control condition. Most interventions focused on staff education/training strategies and implementing a palliative care team. Several integrated advance care planning initiatives into the intervention. We found that palliative care interventions in nursing homes may enhance palliative care practices including processes to assess and manage pain and symptoms. However inconsistent outcomes and high or unclear risk of bias among the majority of studies requires results to be interpreted with caution.

Conclusions and Implications:

Our systematic review revealed that evidence for nursing home palliative care interventions is sparse.

Brief Summary:

We conducted a systematic review to examine the characteristics and effectiveness of multicomponent nursing home palliative care interventions. We found 13 unique clinical trials. In six trials, interventions enhanced nursing home palliative care processes including increased use of care pathways, higher hospice referrals, and better documentation of pain and symptom assessment and management. However, in most trials there was mixed or insufficient evidence about the effect of palliative care interventions on reducing potentially burdensome care, resident/surrogate symptom ratings, advance care planning discussions and documentation, and improving family members’ evaluation of care. Our findings suggest there is limited evidence that palliative care interventions improve nursing home end-of-life care outcomes.

Introduction

Over the past two decades, palliative care services have grown dramatically to meet the healthcare needs of an aging population and the growing burden of serious illness. Palliative care is an interdisciplinary, patient-and family-centered approach to serious illness that incorporates symptom management, open communication, and informed decision making to elicit patients’ and families’ goals of care and preferences for treatments. Its goal is to relieve suffering and improve quality of life by addressing the myriad of physical, psychosocial, and spiritual needs of persons with serious, progressive illness.¹ An expanding body of literature demonstrates that palliative care programs improve patient outcomes, optimize the use of healthcare resources, and enhance family and caregiver evaluation of care.^2-4 However, much of the research has focused on home, hospital, and outpatient clinic settings.^5,6

Evidence-based nursing home palliative care services are essential to enhance high quality end-of-life care. Close to 30% of older adults receive care in nursing homes until their death⁷ and experience untreated pain and other physical symptoms.⁸ Dying residents often report loneliness, depression, and limited support with spiritual needs.⁹ Many undergo potentially burdensome treatments with little benefit.^10-12

The majority of published literature about palliative care programs in nursing homes is anecdotal or observational.^13-16 A 2010 Cochrane Collaborative systematic review found few studies of palliative care interventions in nursing homes that were of high quality to assess outcomes.¹⁷ Continued growth of nursing home palliative care programs in the last several years^18-20 requires analysis of new literature to expand on the earlier review by Hall et al.¹⁷ We therefore conducted a systematic review of multicomponent palliative care clinical trials in nursing homes to examine the characteristics and effectiveness of program interventions.

Methods

Search Strategy

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) in this review.²¹ We searched MEDLINE, Embase, CINAHL, and Cochrane CENTRAL from inception to January 1, 2019 using (""palliative nursing"" OR ""palliative care""[Mesh] OR ""palliative""[tiab] AND ""care"" [tiab] OR ""palliative care"" [tiab]) AND ""nursing homes""[Mesh]). A health sciences librarian assisted us as we developed, piloted, and executed the searches. Literature searches excluded pediatric, non–English-language articles, commentaries, conference abstracts, letters, notes, and news. We hand searched the reference lists of all included articles to identify additional trials that met our eligibility criteria.

Study Eligibility

We defined study eligibility using the PICO (Patient, Problem or Population; Intervention; Comparison, Control or Comparator; Outcome) framework (Table 1).²² We included clinical trials of palliative care interventions focused on individuals receiving care in a nursing home and/or their family caregivers. Initially, we planned to include interventions if they incorporated at least one of eight palliative care domains as defined by the National Consensus Project’s Clinical Practice Guidelines for Quality Palliative Care.¹ However, we later modified our inclusion criteria to include only multiple component interventions (defined as clearly integrating more than one palliative care domain).^6,23 Trials had to incorporate a comparison condition (e.g., usual care, attention control, historical control, waitlist). We only included trials that reported on at least one of six outcomes: healthcare use, symptoms, resident or family evaluation of the quality of care, advance care planning, quality of life, and costs/expenditures.

Table 1.

Patient, Problem or Population; Intervention; Comparison, Control or Comparator; Outcome (PICO) Criteria

Population	Inclusion: Individuals receiving care in a nursing home that provides healthcare services, skilled nursing, and medical care.
	Exclusion: Assisted living residential care, continuing care retirement communities
Intervention	Inclusion: Two or more of the following 8 domains: Structure and Process of Care: use of an interdisciplinary team, comprehensive resident assessment, communication, or care coordination; Ethical and Legal Aspects of Care: advance care planning strategies or discussions about goals for medical care, values, and preferences; Physical Aspects of Care: assessment and management of acute and chronic symptoms; Care at the End of Life: referral to hospice or integrating evidence-based care guidelines in the days leading up to death; Psychological and Psychiatric Aspects of Care: mental health assessment, treatment, and/or specialist referral; Social Aspects of Care: assessment of factors that affect resident quality-of-life and coping strategies; Spiritual, Religious, Existential Aspects of Care: address residents’ and families’ spiritual values and beliefs; Cultural Aspects of Care: integrating a culturally sensitive intervention
	Exclusion: Observational study, interventions focused on a single component of palliative care (e.g., advance care planning or pain management)
Comparator	Inclusion: Any comparison group including usual care, attention control, waitlist, historical control
	Exclusion: No comparator
Outcomes	Inclusion: One or more of the following patient/family focused outcomes: healthcare use, symptom management, resident or family evaluation of the quality of care, advance care planning, quality of life, and costs/expenditures.
	Exclusion: Staff focused outcomes; practice environment changes

PC=palliative care; ACP = advance care planning; NH = nursing home; ED= emergency department

Authors screened each article abstract for eligibility, then two reviewers independently appraised the full text of each eligible article using our inclusion/exclusion criteria to ensure appropriateness for the study. We resolved disagreements by consulting with a third author.

Data Extraction and Bias Assessment

We developed and piloted a structured abstraction tool to record evidence from each study. Two reviewers independently evaluated each report and extracted data (Table 2). We also rated risk of bias using a tool informed by the Cochrane Collaboration that included five categories: random sequence generation, allocation concealment, blinding of outcomes assessment, incomplete outcomes data, and selective reporting.²⁴ If a study scored low on 4-5 of 5 categories, it was judged to have an overall low risk of bias; studies that scored high on 2-5 of 5 categories were judged overall high risk of bias; studies with unclear risk of bias or if assessment was not applicable (such as random sequence generation in a trial comparing an intervention to a historical control group) on 2-5 of 5 categories were judged overall unclear risk of bias.

Table 2.

Table of Evidence

Reference	Intervention Domain	Design	Description	Main Findings
Agar et al, 2017 PLOS one (Australia)	Structure and processes of care Ethical and legal aspects of care	Design: Two arm, parallel cluster RCT Randomized at the NH level Population: People with advanced dementia and their families and professional caregivers in 20 NHs Dates: Feb 2013 - Dec 2014 Primary Measures Surrogate rated Comfort Assessment in Dying with Dementia; CAD–EOLD, Symptom Management at the End of life in Dementia; SM–EOLD, Satisfaction with Care at the End of life in Dementia; SWC–EOLD Secondary Measures: Proxy Nurse-rated CAD-EOLD, SM-EOLD, and Quality of Life in Late-stage Dementia (QUALID) Scale, rates of potentially inappropriate non-palliative interventions, rates of inappropriate acute care, nursing home staff ’s attitudes to, knowledge of and confidence in providing palliative/EOL care to residents with advanced dementia.	Intervention: NH(n=10) ∣ Residents(n=156) Facilitated family case conferences (FCC) and staff palliative care training. Palliative Care nurse Planning Coordinator in facility for 16 hours per week over 18 months Control: NH(n=10) ∣ Residents(n=130) Usual care	Healthcare Use Physician input during the month prior to death were significantly more common in the intervention group compared to control Symptoms Intervention group significantly more likely than control to have staff documentation of pain/discomfort (p=0.04), restlessness (p=0.02), constipation (p=0.002), skin tears (p=0.005) and other symptoms: resistive to care, agitation, pressure areas, wounds, respiratory infection and chest rattles (p<0.001) Pharmacological and non-pharmacological changes to management during the month prior to death were significantly more common in the intervention group compared to control Equal numbers (47%) in both groups had formal pain assessments, but more documentation of pain assessment frequency was seen in the intervention (97%) than control group (79%) No significant intervention effect on surrogate report of comfort during the last 7 days of life and symptom management in the last 90 days of life Quality of Life No significant intervention effects were observed on quality of life outcomes Ratings of the person's quality of life during the 90 days prior to death were significantly associated with staff knowledge in the intention to treat model; better quality of life associated with nursing homes with higher staff knowledge scores (coefficient −1.04, P = 0.036) Family Evaluation of Care No significant group effect Satisfaction with Care at the End of life in Dementia (SWC–EOLD)
Berkowitz et al, 2011 JAGS (Boston, MA, USA)	Structure and processes of care Physical aspects of care Ethical and legal aspects of care	Design: Two arm prospective Quasi-experimental trial (historical control group) Population: Residents of one 50-bed skilled nursing unit (short term care) Dates: Control: June 2008 to May 2009 before the intervention started; Intervention Jun 2009 - May 2010 Measures: Distribution of discharge dispositions, including readmission, transfer to long-term care facility, discharge to home, and death.	Intervention: (n=862) Standardized physician admission template Goals of Care discussion template Automated Palliative Care consult if >3 hospitalizations in last 6 months and Interdisciplinary team conference every 2 weeks (cases of rehospitalization deemed potentially avoidable were reviewed and “lessons learned” email was shared with all staff after each meeting) Control: (n=863) Retrospective cohort	Healthcare Use (intervention vs. control) Discharge dispositions differed significantly between groups (P=.03) Increase discharge to home (73.0% vs 68.6%) Reduced discharge to LTC (11.5% vs 13.8%) Residents more likely to die on NH unit during intervention (OR: 2.45) Reduced rehospitalization rate (16.5% vs 13.3%)
Boogaard et al, 2018 Palliative Medicine (The Netherlands)	Structure and processes of care Care of the patient nearing the end of life	Design: Three arm parallel cluster-RCT; triplets of NHs matched on number of psychogeriatric beds, presence of a palliative care unit, and urban/nonurban location; Randomized at the NH level Population: Family caregivers of deceased residents with dementia in 18 Dutch NHs psychogeriatric wards Dates: Pre-intervention phase January 2012- October 2012; Intervention: November 2012-July 2014 Measures: Bereaved family member rated End-of-Life in Dementia-Satisfaction With Care scale (EOLD-SWC) and the End-of-Life in Dementia-Comfort Assessment in Dying scale (EOLD-CAD)	Intervention: Data feedback reports (derived from two dementia-specific family-reported tools) coupled with evidence-based suggestions to improve end-of-life care (1) Generic feedback strategy NH n=6∣Residents (n=123) NH teams receive and discuss targeted suggestions for care improvement strategies based on cumulative family-reported feedback (2) Patient-specific feedback strategy: NH n=4∣Residents(n=71) NH teams receive and discuss targeted suggestions for care improvement strategies based on individual family-reported feedback Control (3) Usual care : NH n=6 ∣ Residents(n=113)	Symptoms No clinically significant pre to post intervention changes in End-of-Life in Dementia–Comfort Assessment in Dying (EOLD-CAD) in any of the three groups. EOLD-CAD total change scores for patient-specific feedback group were statistically significantly higher compared to the control condition only in unadjusted analysis (B = 2.20, CI = 0.15; 4.39). EOLD-CAD “Dying Symptoms” subscale change scores for patient-specific feedback group were statistically significantly higher than control in unadjusted and adjusted analyses (B = 0.86, CI = 0.43; 1.68 and B = 1, CI = 0.38; 2.00, respectively). Family Evaluation of Care Non-significant differences in scores of patient specific feedback group versus control No clinically significant pre-to post intervention changes in End-of-Life in Dementia–Satisfaction with Care in (EOLD-SWC) scores in any of the three groups. Significantly lower scores in EOL-SWC for generic feedback intervention group compared control (adjusted B = −2.41, 95% CI = −4.07; −0.76)
Brännström et al, 2016 Palliative Medicine (Sweden)	Structure and processes of care Care of the patient nearing the end of life Physical aspects of care	Design: Two arm quasi experimental exploratory, controlled before- and-after trial Population: Deceased residents in 19 residential care homes Dates: Jun 2009 – Oct 2011 Primary Measures: Edmonton Symptom Assessment System (ESAS) Secondary Measures: Views of Informal Carers – Evaluation of Services (VOICES)	Intervention: (usual care n=64; intervention n=71) Usual care x 15 months, then Liverpool Care Pathway (LCP) for the Dying Patient (education and training for staff included 1) a 3-hour session about the LCP and EOL care for all assistant nurses, RNs and GPs; 2) a nurse who completed a 35-hour web-based LCP train-the-trainer course) Control: (usual care n=61; usual care n=64) Usual care X15 months, then usual care	Symptoms Intervention group showed significant improvement in shortness of breath compared to control group ( −2.46; 95% CI −4.43; −0.49) via Edmonton Scale and nausea (−1.83; 95% CI −3.12; −0.54) via Edmonton Scale Intervention group showed significant improvement in breathlessness (−0.47; CI=−0.85; − 0.08) p=0.02 via Views of Informal Carers – Evaluation of Services (VOICES) questionnaire No significant improvement in pain, tiredness, drowsiness, loss of appetite, anxiety, depression, wellbeing
Casarett et al, 2005 JAMA (Phila, PA, USA)	Structure and processes of care Care of the patient nearing the end of life Ethical and legal aspects of care	Design: Two arm parallel RCT; randomized at patient level Population: NH residents and their surrogate decision makers in 3 NHs Dates: Dec 2003 – Dec 2004 Measures: Deaths, hospitalizations, new orders to limit life-sustaining treatment, hospice enrollment, bereaved family survey (Afterdeath Survey), surrogate rated: care in the last week of life, whether a member of the health care team discussed the resident’s prognosis with them and what to expect during the dying process, whether the resident had pain or shortness of breath that was not managed as well as it could have been, and whether the resident died where he/she would have wanted.	Intervention: (n=107) Structured interview to identify residents whose goals for care, treatment, preferences, and PC needs are appropriate for hospice care → hospice informational visit Control: (n=98) Usual care	Healthcare Use Intervention group more likely to (intervention vs control): Enroll in hospice within 30 days (20% vs 1%; p<0.001) Enroll in hospice during the 6 month follow up period (25% vs 6%, p<0.001) Undergo fewer acute care admissions (mean: 0.28 vs 0.49, p=0.04) Spend fewer days in acute care (mean: 1.17 vs 2.99, p=0.03) Intervention group not more likely to (intervention vs control): Be enrolled in hospice at the time of death (36% vs 38%; P=.93) Die in the nursing home rather than in an acute care setting (60% vs 62%;P>.99) Family Evaluation of Care Family ratings of care in the last week of life significantly higher in the intervention group (mean on 1-5 scale; 4.3 vs 2.2, p=0.01) Surrogates of residents with a longer length of stay in hospice rated the resident’s end-of-life care more highly (Spearman correlation coefficient, 0.75; P=.05).
Chapman and Toseland, 2007 (Northeastern USA)	Structure and processes of care Physical aspects of care Psychological and psychosocial aspects of care Social aspects of care	Design: 2x2 randomized partial crossover trial; randomized in cohorts at patient level; crossover results not reported Population: NH residents with advanced dementia in 2 NHs Dates: Unclear; Random assignment done in cohorts over a 1 year period Measures: Cohen-Mansfield Agitation Inventory, Faces Legs Activity Cry Consolability, Cornell Scale for Depression in Dementia, Pain in Advanced Dementia (PAINAD)	Intervention:(n= 57; after crossover n=118) Advanced Illness Care Teams (AICT) consisting of staff from medicine, nursing, social work, psychology, physical and occupational therapy, and nutrition. AICT met 5 times during an 8 week period to address 1) medical issues, 2) meaningful activities, 3) psychological problems, 4) behavioral concerns. ACP explicitly was not included in the intervention b/c investigators assessed that participating NHs “thoroughly addressed the ACP needs of residents during the admission and care planning process” Control: (n=61) Usual care; residents in usual care received the intervention after the 8-week measures were administered.	Symptoms Pain: Faces Legs Activity Cry Consolability (FLACC) Behavioral Pain Scale: Significant decrease in pain for residents on both conditions over time (F=8.39, p<0.01) but not significant between intervention and control Pain: Pain in Advanced Dementia (PAINAD) Significant decrease in pain for residents on both conditions over time (F=5.38, p<.05) but not significant between intervention and control Depression: Cornell Scale for Depression in Dementia: Significant decrease in Behavioral Disturbances for residents on both conditions over time (F=5.14, p<.01) but not significant between intervention and control Physically nonaggressive behaviors via Cohen- Mansfield Agitation Inventory: Intervention experienced significantly greater decrease compared to control (F=4.22, p<.01).
Comart et al, 2013 Gerontologist (Boston, MA, USA)	Structure and processes of care Physical aspects of care Psychological and psychosocial aspects of care Ethical and legal aspects of care Social Aspects of Care Spiritual, Religious, Existential Aspects of Care	Design: Two arm Quasi-experimental trial (matched historical control group) Population: Deceased residents in 1 long term care facility Dates: 2006 – 2009 Control: Residents who died when the PC service didn’t exist (2006). Intervention: residents who died during the period of the PC service (2007–2009) Measures: Physician orders, medication use, hospital stays, ER visits, falls, shortness of breath, weight loss, depression, pain, delirium, pressure ulcers.	Intervention: (n=125) Interdisciplinary Palliative Care Team: PC physician (0.6 fte), clinical nurse specialist (0.6 fte), chaplain (0.3 fte), social worker (0.5 fte), and a psychologist who also serves as the lead administrator for the program (1.0 fte): helping the family define the goals of care, providing expert pain management, or assisting families come to agreement on treatment options Control: (n=125) Retrospective cohort; usual care	Healthcare Use Intervention showed significant reduction in emergency department visits compared to control (p<0.001) Intervention did not show a significant difference in hospitalizations, number of medications, and physician orders compared to control. Symptoms Compared to control, the intervention showed significant reduction in depression (p=0.031); but did not significant reduction delirium, falls, pain, skin ulcers, shortness of breath, weight loss
Hanson et al, 2005 J Palliative Med (NC, USA)	Structure and processes of care Ethical and legal aspects of care Physical aspects of care	Design: Two arm Quasi-experimental pre-post trial; “control” sites used to examine temporal trends. Measures at baseline, 3, 6 months post-intervention Population: 9 NHs in North Carolina Dates: Jan 2001 – Dec 2002 (3 month recruitment/planning phase, 9 month intervention) Measures: Percent of all residents enrolled in hospice or receiving hospice palliative care consultation services, pain, treatments for pain, advance care planning discussions, orders limiting life-sustaining treatments, living wills, and health care powers of attorney	Intervention: NH n=7∣Residents(n=345) 1) Identification of NH-based interdisciplinary Palliative Care Leadership Team who all attended 1-day conference covering hospice services, pain management, ACP, communication, quality improvement ; 2) Monthly on-site education and technical assistance on quality improvement, 3) 6 in-services for clinical staff (live & videotaped); 4) Monthly strategy meetings w/ investigators & hospice providers to design/implement stepwise organizational changes w/feedback on performance measures and action plans Control: NH=2∣ Residents(n=113) Usual Care	Healthcare Use Intervention NHs showed increase in residents on hospice from baseline to 6 months: (4.0% vs. 6.8%, p=0.01) Symptoms Intervention sites showed increases in completed pain assessments (18% at baseline to 60% after intervention (p = .001) Orders for nonpharmacologic pain treatments increased significantly from 15% baseline to 34% after intervention (p = .001). Orders for pain medication was not statistically significant Prevalence of pain (measured by pain despite treatment) on MDS did not change over time Advance Care Planning Intervention nursing homes had a significant increase from baseline to post-intervention residents with 1) a documented discussion (4% to 17%) of (p < .001); 2) DNR orders (58% to 65%, p =0.04) Use of living wills (31% vs. 30%) and health care powers of attorney (27% vs. 33%) were unchanged
Hockley at al, 2010 Palliative Medicine (Midlothian, Scotland)	Structure and processes of care Ethical and legal aspects of care Physical aspects of care Care of the patient nearing the end of life	Design: Two arm quasi-experimental trial (historical control group) Population: Deceased residents in 7 NHs Dates: Control: those who had died a year previous to the study (3/2006-2/2007) Intervention: those who had died during/following the Intervention (3/2007-unclear) Measures: Medical record evidence of documentation for advance care planning and do not attempt resuscitation orders, hospital admissions in the last 8 weeks of life, place of death and the use of the Liverpool Care Pathway	Intervention (n=133) Gold Standards Framework for Care Homes and an adapted Liverpool Care Pathway for Care Homes implemented over an 18 month period with a model of high facilitation every 10–14 days; incorporated “key champions”, a nurse facilitator and key champion and NH staff training. Control (n=95) Retrospective cohort; usual care	Healthcare Use Use of the Pathway: 3% control v 30% intervention (p<0.001) Hospital death (pre/post intervention): 15% to 8% Hospital admissions in the last 8 weeks of life (pre/post intervention): 31% to 24% Advance Care Planning Do Not Attempt Resuscitation Orders increased control 15% v intervention 72%; p<0.001 Written evidence of advance care planning; control 4% v intervention 53%; p<0.001
Kinley et al, 2014 Palliative Medicine (UK)	Structure and process of care Ethical and legal aspects of care Care of the patient nearing the end of life	Design: 3 arm parallel cluster RCT; stratified according to geographical location; randomized at NH level Population: Deceased residents in 38 NHs Dates: Jun 2008 – May 2011 preliminary stage (June 2008–August 2009); implementation (September 2009–May 2010); consolidation (June 2010–May 2011) Primary Measures: Place of death of resident Secondary Measures: Medical record documentation of advance care planning, having a cardiopulmonary resuscitation decision, and using end-of-life care plan documentation	Intervention: Implementation of the GSF (Gold Standards Framework for Care Homes) High facilitation and Action Learning (NH n=12∣ Residents n=805); high facilitation and managers receiving action learning High Facilitation only (NH n=12∣ Residents n=703) high facilitation only Control: (NH n=14∣ Residents(n=936) Usual GSF care (received standard facilitation of the program available in their locality).	Healthcare Use Place of death: no significant effect of intervention although the difference by Year 3 was in the expected direction Using end-of-life care plan: significant effect of the High facilitation and Action Learning intervention v High Facilitation only (p = 0.036) Advance Care Planning Resuscitation decision in place: no significant effect of intervention Completed advance care plan: no significant effect of intervention
Strumpf et al Annals of Long-Term Care, 2004 (MD, USA)	Structure and processes of care Ethical and legal aspects of care Physical aspects of Care Psychological and psychosocial aspects of care	Design: Two arm quasi-experimental trial Population: Residents at 6 community NHs; with terminal or end-stage disease, or those who had experienced significant functional decline, recent multiple hospitalizations, or major alterations in nutritional status Dates: Dec 1998-Dec 2001 Measures: Medical record documentation of physical and emotional symptoms, recording events at the time of death (eg, pain, hospital transfer, family present, etc), recording advance directives, surveying family satisfaction	Intervention NH (n=4) PC Delivery Process training on advance care planning, pain and symptom management, and psychosocial concerns with ongoing guidance from a nurse with expertise in palliative care; Two of the four intervention homes also developed a dedicated interdisciplinary PC team with weekly meetings. Control NH (n=2) Usual Care	Healthcare Use (intervention v control) Higher rates of physician visits and x-rays in the last weeks of life (P < 0.01) Symptoms (intervention v control) More residents with pain managed by narcotics (P < 0.01) Higher rates of symptoms (documented in medical record) in the last weeks of life (P < 0.05) Rural homes had a significantly higher percentage of physical symptoms with documented assessment and treatment than those in urban area (P< 0.01) Advance Care Planning (intervention v control) More residents with advance care plans as measured by recording advance directives (P < 0.01)
Temkin-Greener et al, 2018 Medical Care (NY, USA)	Structure and processes of care Care at the end of life Physical aspects of care Psychological and psychosocial aspects of care	Design: Two arm RCT with difference-in-difference analysis; randomized at the facility level after stratified by bed size Population: Deceased Medicare beneficiaries over age 65 at New York state NHs Dates: Pre intervention Oct 2012- Sep 2013; Intervention Oct 2013 – Apr 2016 Measures: Minimum data set documentation of place of death, number of hospitalizations within the last 90 days of stay, pain, depression	Intervention: (NH n=14) Facility level intervention: Implemented PC teams and training-education program for staff, then 2-month period of active intervention phase with rounds w/ gero-palliative nurse practitioner, then 8-month passive phase with as needed coaching. Control: (NH n=11) Usual care Non-randomized control: (NH n=609) Usual care (used to assess placebo effect)	Healthcare Use No significant impact of the intervention compared to control condition on residents’ outcomes Intervention NH with “working teams” (n=6) for the entire duration of the study compared to those with “nonworking teams” (n= 8) had significant reductions of in-hospital death (OR: 0.4). Symptoms No significant impact of the intervention compared to control condition on residents’ outcomes Intervention NH with “working teams” (n=6) for the entire duration of the study compared to those with “nonworking teams” (n= 8) had significant reduction in depressive symptoms (OR=0.2)
Teo et al, 2014 Palliative Medicine (Singapore)	Structure and processes of care Physical aspects of care Ethical and legal aspects of care	Design: Quasi-experimental trial (matched historical control group) Population: Residents at 7 NHs at risk of dying in the next 6 months-1 year (identified through a clinical algorithm) Dates: Control group: deceased residents who were at the NH for at least 3 months Oct 2007-Sept 2009; Intervention group: Sept 2009-Feb 2011 Measures: hospitalization length of stay, nursing home LOS, emergency department visits, specialist outpatient clinic visits, costs of healthcare use, nursing home cost	Intervention: (n=48) Project CARE: ACP (Respecting Choices framework) to 1) identify residents who choose comfort care; 2) palliative care staff training; 3) access to specialist physician and nurse visits Control: (n=197) Retrospective cohort; usual care	Healthcare Use 3 months prior to death (intervention vs control) significant differences between groups in terms of health-care usage: Hospitalization: 42% vs 68%; Days in the NH 90 vs 84 ED visits: 3 months prior to death (intervention vs control) (1% vs 1%); Final month of life (1% vs 2%). Costs/Expenditures Hospital costs (intervention vs control): last 3 months of life: 42% vs 68%; final month of life 51% vs 80% (P<.05); Within the last 3 months of life, more than 52% of the savings in the healthcare system comes in the final month of life; After controlling for differences in baseline characteristics, economic impact remained highly significant at 3 months (45% savings) and 1 month (48% savings)

Abbreviations: PC=palliative care; ACP = advance care planning; NH = nursing home; ED= emergency department

Results

We identified 1148 records through our database searches and an additional 19 articles by hand searches of reference lists. After reviewing the abstracts for all 1167 records, we identified 67 articles for full text review. Following review of the full text, we retained thirteen articles, each reporting results from unique trials, published between 2004 and 2018. Figure 1 details the screening and inclusion/exclusion process.

Figure 1: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Flow Diagram:

Results of Searches to Identify Clinical Trials of Palliative Care Interventions

Study Characteristics

About half of the studies were conducted in the United States (n=7).^25-31 Three studies focused specifically on residents with dementia.^27,32,33 Five studies used parallel group design with randomization at the facility (n=4)^31-34 or resident level (n=1)²⁶ and compared interventions to usual care^26,31,32 or an active comparator.^33,36 Seven studies used quasi-experimental designs in which intervention groups were compared to historical control groups (n=4)^25,28,35,36 or usual care (n=3).^29,30,37 One study used a randomized partial cross over design.²⁷ Overall, nine of the 13 trials were judged as having high (n=4) or unclear (n=5) risk of bias (table 3).²⁴

Table 3.

Bias Assessment

Author, Year	Random sequence generation	Allocation concealment	Blinding of Outcomes assessment	Incomplete outcomes data	Selective reporting	Overall Rating
Agar et al, 2017	●	⊗	●	●	●	●
Berkowitz et al, 2011	n/a	n/a	n/a	●	⊝	⊝
Boogard et al., 2018	●	●	⊗	⊗	●	⊗
Brännström et al, 2016	n/a	n/a	n/a	⊗	●	⊝
Casarett et al, 2005	●	●	⊗	●	●	●
Chapman and Toseland, 2007	⊝	●	⊝	●	●	⊝
Comart et al, 2012	n/a	n/a	n/a	●	●	⊝
Hanson et al, 2005	⊝	⊗	⊗	●	●	⊗
Hockley at al, 2010	n/a	n/a	n/a	●	●	⊝
Kinley et al, 2014	●	●	●	●	●	●
Strumpf et al 2004	n/a	n/a	⊝	●	●	⊝
Temkin-Greener et al, 2018	●	⊝	●	●	●	●
Teo et al, 2014	n/a	n/a	n/a	n/a	●	⊝

●= low; ⊗= high; ⊝= unclear

Palliative Care Intervention Characteristics

The multi-component interventions comprised a median of three of eight National Consensus Project domains (range 2-6). All interventions incorporated Structure and Process of Care and over half integrated Ethical and Legal Aspects of Care (n=9) and Physical Aspects of Care (n=9). Several interventions addressed Care at the End of Life (n=6) and Psychological and Psychiatric Aspects of Care (n=4). Two of the studies specifically described Social Aspects of Care.^27,28 Only one intervention explicitly incorporated Spiritual, Religious, Existential Aspects of Care.²⁸ None of the interventions addressed Cultural Aspects of Care.

Interventions focused on three major strategies: (1) advance care planning, (2) staff education (e.g., general concepts taught in a class or seminar) or training (e.g., mentoring palliative care skills in practice) and (3) integrating a palliative care team. Berkowitz et al. and Teo et al. integrated all three strategies into interventions.^25,35

Staff education and/or training was the most common strategy used in the interventions. One trial used an audit and feedback strategy in which they regularly shared results of bereaved family care evaluations with staff.³³ After discussing the reports, staff participated in quality improvement activities aimed at increasing evaluation scores. Several education/training programs also incorporated staff champions or facilitators to act as palliative care resources.^29,34,37 Others implemented staff training in end-of-life care alongside facility based palliative care teams^30,31 or consultative teams.²⁸

Of six trials that implemented an interdisciplinary palliative care team; four clearly incorporated multiple disciplines (medicine, nursing, chaplains, social worker, and/or psychologist).^25,27,28,30 The others utilized two disciplines: medicine and nursing³⁵ or nursing and social work.^31,38 Three interventions explicitly described creating palliative care teams from existing nursing home staff.^27,30,31 Others implemented consultation services.^25,28,35

Five studies integrated advance care planning into the palliative care intervention. Advance care planning strategies most often screened for individuals that would benefit from an advance care planning conversation. Four used illness prognosis and acute changes in clinical status to identify residents who would benefit from an advance care planning discussion^{25,36 35 32} One intervention implemented resident and surrogate interviews to identify residents with goals for care that focused on comfort.²⁶

Outcomes

Healthcare use and symptoms were the most common outcomes evaluated in the trials and measured in various ways. Four trials described evidence of resident/surrogate advance care planning discussions and documentation. Three assessed family evaluation of care and only one trial each evaluated resident quality of life and costs/expenditures (Table 4).

Table 4.

Strategies, outcomes, and overall risk of bias

Author, year	Strategies			Outcomes	Overall Risk of Bias
	Advance Care Planning	Staff training and education	Integrating a palliative team
Agar et al, 2017	X	X		Healthcare Use  Enhance NH palliative care (+) Symptoms  Assessment/Management Process: pain/discomfort (+), restlessness (+), agitation (+), constipation (+), skin issues (+), resistive to care (+),respiratory issues (+) Resident/surrogate report: Comfort (−) Quality of life  Resident/surrogate ratings at the end-of-life (−) Family Evaluation of Care  Satisfaction with care (−)	Low
Berkowitz et al, 2011	X	X	X	Healthcare Use  Reduce potentially burdensome care (+)  Enhance NH palliative care (+)	Unclear
Boogard et al., 2018		X		Symptoms  Resident/surrogate report: Comfort (−)  Dying symptoms (+), physical distress (−), emotional distress (−), wellbeing (−) Family Evaluation of Care  Satisfaction *	High
Brännström et al, 2016		X		Symptoms  Resident/surrogate report: shortness of breath (+), nausea (+), pain (−), tiredness (−), drowsiness (−), loss of appetite (−), anxiety (−), depression (−), wellbeing (−)	Unclear
Casarett et al, 2005	X			Healthcare Use  Reduce potentially burdensome care*  Enhance NH palliative care* Family Evaluation of Care  End-of-life care ratings (+)	Low
Chapman and Toseland, 2007			X	Symptoms  Resident/surrogate report: pain (−), depression (−),agitation (−), physically nonaggressive behaviors (+)	Unclear
Comart et al, 2012		X	X	Healthcare Use  Reduce potentially burdensome care*  Enhance NH palliative care (−) Symptoms  Resident/surrogate report: depression (+), delirium (−), falls (−), pain (−), skin ulcers (−), shortness of breath (−), weight loss (−)	Unclear
Hanson et al, 2005		X		Healthcare Use  Enhance NH palliative care (+) Symptoms  Assessment/Management Process: pain*  Resident/surrogate report: pain (−) Advance Care Planning  Documented discussions and orders (+)  Advance Directives (−)	High
Hockley at al, 2010	X	X		Healthcare Use  Reduce potentially burdensome care (+)  Enhance NH palliative care (+) Advance Care Planning  Documented discussions and orders (+)	Unclear
Kinley et al, 2014		X		Healthcare Use  Reduce potentially burdensome care (−)  Enhance NH palliative care (+) Advance Care Planning  Documented discussions and orders (−)	Low
Strumpf et al, 2004		X	X	Healthcare Use  Enhance NH palliative care (+) Symptoms  Assessment/Management Process: pain (+) Advance Care Planning  Advance Directives (+)	Unclear
Temkin-Greener et al, 2018		X	X	Healthcare Use  Reduce potentially burdensome care (−) Symptoms  Resident/surrogate report: pain (−), depression (−)	Low
Teo et al, 2014	X	X	X	Healthcare Use  Reduce potentially burdensome care* Costs/Expenditures  Reduce costs (+)	Unclear

NH= Nursing Home

^*= mixed results

Changes in Clinical Practice and Healthcare Use

Six studies measured nursing home clinical practices aimed to enhance palliative care for residents. Seven studies examined the effects of palliative care interventions on potentially burdensome end-of-life healthcare use such as hospitalizations and emergency department visits.

In total, six studies reported positive changes in nursing home palliative care practices. Four found greater physician involvement in residents’ care during the last weeks of life^30,32 and increased use of a pathway to assist staff in providing end-of-life treatments in nursing homes.^34,36 Two studies found that interventions resulted in more discussions about hospice and higher hospice referral and enrollment rates.^26,29

The effect of nursing home palliative care on residents’ acute care outcomes varied. Five studies found that palliative care interventions reduced resident emergency department visits,²⁸ rehospitalizations,^25,35 acute care admissions,^26,36 days in acute care,^26,35 and in-hospital death.³⁶ In contrast, four interventions showed no significant effect on emergency department use in the last three months of life,³⁵ rehospitalization,^28,31 or in-hospital death.^31,34

Effects of palliative care on nursing home death and hospice use were conflicting. Berkowitz et al. reported that residents receiving a palliative care consult intervention were more likely to die in the nursing home.²⁵ Conversely, Casarett et al. found that an advance care planning intervention was not associated with greater likelihood of dying in the nursing home or death while receiving hospice services.²⁶

Symptoms

Across eight studies, pain and symptom measures varied. To better synthesize the results, we categorized outcomes according to how they were measured: nursing home processes or resident/surrogate report. One study used process measures only (e.g., documented pain assessment and management),³⁰ five studies evaluated resident or surrogate reported symptom severity,^{27-29,31,33,37} and two studies used both.^29,32 Of the eight studies, three focused exclusively on residents with dementia.^27,32,33 Most studies focused on measuring pain or discomfort, respiratory issues, depressive symptoms, and agitation.

Two studies found an improvement in pain and symptom assessment and management processes. After implementing staff education, Hanson et al. reported pre- to post-intervention increase in pain assessments (18% to 60%) and nonpharmacologic treatments (15% to 34%).²⁹ Using education and training in conjunction with a facility-based palliative care team, Strumpf et al. reported intervention nursing homes had higher rates of documented symptoms and more residents with opioid pain management in the last weeks of life.³⁰ One study that focused exclusively on residents with dementia found that staff education and training improved symptom management practices including documentation of pain, shortness of breath, agitation, nausea and constipation.³²

In contrast to studies that documented improved pain and symptom clinical processes, evidence of enhanced patient symptom outcomes was uneven. Four interventions did not improve resident pain ratings.^28,29,31,37 Effects of palliative care interventions on non-pain symptoms were reported in three trials with inconsistent results. Braanstrom et al. found residents in intervention nursing homes experienced less shortness of breath after integrating an end-of-life care pathway.³⁷ In contrast, Comart et al. found no intervention effect on shortness of breath²⁸ after implementing staff education and training and a palliative care team. Comart et al. found depression improved²⁸ but trials using a palliative care team³¹ and an end-of-life care pathway³⁷ did not report an improvement in depressive symptoms.

Of three studies that reported on surrogate-reported symptoms, none found that palliative care interventions decreased pain or improved comfort at the end of life for residents with dementia.^27,33,34 Two trials that measured non-pain symptom outcomes had mixed results; residents assigned to a palliative care intervention showed a decrease in physically non-aggressive behaviors²⁷ and improved dying symptoms;³³ but no improvement in agitation or depression.

Advance Care Planning

There was mixed evidence among four studies examining the effect of palliative care interventions on advance care planning discussions and documentation. All the interventions incorporated ongoing staff support for palliative care with onsite education and expert assistance over a nine to 18-month period. Three quasi-experimental studies found increases in documented discussions of residents’ end-of-life treatment preferences, ²⁹ “do not attempt resuscitation” orders,³⁶ and presence of advance directives.³⁰ In contrast, Kinley et al. found no significant difference in documented “do not attempt resuscitation” orders between intervention and control groups³⁴ and Hanson et al. showed no change in documented use of advance directives pre- to post-intervention.²⁹

Family Evaluation of Care

Three randomized clinical trials examined the effectiveness of interventions on bereaved family ratings of end-of-life care with varied outcomes. Casarett et al. reported that an advance care planning intervention designed to promote communication about hospice showed higher bereaved family ratings of end-of-life care, a difference that also was associated with greater numbers of days receiving hospice services.²⁶ In contrast, two studies focusing on residents with dementia found no significant improvement in family members’ satisfaction with end-of-life care.^33,32

Quality of Life

Only one study reported on the effects of a palliative care intervention on quality of life. In a trial of interactive, person-centered palliative care training and facilitated case conferencing, Agar et al found no intervention effects on quality of life in the 3 months prior to death for residents with advanced dementia and their family caregivers.³² However in additional analyses, better quality of life ratings were associated with nursing homes reporting higher staff knowledge and confidence providing palliative care.

Costs/Expenditures

Teo et al examined healthcare costs following staff education and training, a palliative care team intervention, and a clinical algorithm to identify residents at high risk of unmet palliative care needs.³⁵ The investigators found that the mean cost per resident at three months prior to death was lower in the intervention group than that of a historical control group; this difference persisted into the final month of life.

Discussion

We examined the results of 13 clinical trials of multi-component palliative care interventions for persons receiving care in nursing homes. In six trials, interventions enhanced nursing home palliative care processes including increased use of care pathways, higher hospice referrals, and better documentation of pain and symptom assessment and management. However, in most trials there was mixed or insufficient evidence about the effect of palliative care interventions on reducing potentially burdensome care, resident/surrogate symptom ratings, advance care planning discussions and documentation, and improving family members’ evaluation of care. Our findings suggest there is limited evidence that palliative care interventions improve nursing home end-of-life care outcomes. In addition, the high or unclear risk of bias among the majority of the clinical trials requires caution when interpreting the results.

Of the three studies that included only residents with dementia, none found that palliative care interventions improved surrogate-reported resident pain. Because persons with dementia have limited ability to self-report symptoms, researchers often rely on surrogate report and medical record data, which may be inaccurate. Previous research demonstrates that pain in persons with dementia is frequently unrecognized and under-reported in long-term care settings.^39,40 Additionally, small sample sizes in the included studies may have prevented researchers from achieving significant results.

Eleven of the studies in our review used an intervention that included palliative care education/training; four trials used this approach alone and had mixed or no effect on resident symptoms. Clinicians and researchers alike suggest that education is essential for change in clinical practice however a recent literature review found that existing palliative care education/training resources may be inadequate.⁴¹ Our findings demonstrate a movement toward interventions that integrate education and training with other approaches to improve palliative care delivery. Of the seven interventions that combined education with at least one other palliative care strategy, four enhanced nursing home palliative care practices.^25,30,32,36 This is consistent with prior work showing residents in nursing homes with higher palliative care knowledge and practice scores experience better care at the end-of-life.⁴³

The design of palliative care teams differed across six studies, with some creating an internal team from the preexisting facility staff and others creating a consultative palliative care team; neither model demonstrated consistently positive results. For example, three studies in our review that created an internal team and provided ongoing access to a palliative care expert showed an improvement in symptom management processes³⁰ but not in residents’ pain and depression.^27,31 We believe the existing nursing home interdisciplinary team is a strength that can be built upon for palliative care interventions. However, Norton et al. cite many challenges with this approach including the need for administrative support, financial resources, staff and leadership turnover, and competing priorities; all may interrupt or delay the uptake of palliative care interventions.³⁸

Three trials that used a palliative care specialty consultation team demonstrated a reduction in potentially burdensome care including emergency department visits²⁸ and hospital transfers.^25,35 These findings are complementary to previous work by Miller et al. who showed that integrating a palliative care consultant into routine care reduces potentially burdensome setting transitions.^15,19 It may be that using an “outside” resource captures the specialty training of consultants to address the unmet needs of seriously ill residents. However, current palliative care workforce projections suggest that palliative care needs cannot be met by specialists alone.⁴³

Based on our findings, we recommend that future research should be multifaceted and support nursing home care providers in delivering primary palliative care, allow for ongoing collaboration with experts, and identify appropriate residents for services. This is consistent with Hall et al. 2010 systematic review that revealed potentially promising results for interventions that assessed residents’ suitability for palliative care and focused on developing palliative care expertise in nursing homes.¹⁷ In addition, program expenses and staffing should be monitored to address facilities’ concerns about costs and resource utilization.

The evidence included in this review is limited by the complexity of the palliative care interventions. A lack of effect on outcomes may reflect intervention implementation failure rather than ineffectiveness.⁴⁴ A systematic process evaluation during and after a trial may help to identify implementation challenges.⁴⁵ Further, it is difficult to assess what parts of each intervention were most effective and what parts did not provide benefit. Future work should focus on examining components of palliative care interventions that residents, family, and nursing home staff value, especially those that improve outcomes.

Our review has a number of strengths, including a comprehensive search for palliative care clinical trials that integrate multiple domains recommended by national guidelines and reporting on resident and family-oriented outcomes. However, there are several limitations. First, we noted great heterogeneity across trial designs that limited our synthesis of study outcomes. For example, quasi-experimental designs that compare intervention groups to historical control groups and non-randomized pre-post studies are known to be confounded by temporal trends and the inability to control for variability among the groups being compared. Similar methodologies across studies would have enabled a deeper examination of inconsistent or mixed outcomes as few clinical trials of palliative care interventions in nursing homes exist. Second, our selection criteria may have resulted in the exclusion of some effective interventions because we did not include single component interventions that incorporated only one palliative care domain. Additionally, we acknowledge that the projects supported by the Centers for Medicare and Medicaid Initiative to Reduce Avoidable Hospitalizations among Nursing Facility Residents included palliative care componenects however they were not explicitly framed as palliative care interventions so were not included in our review.⁴⁶ Lastly, while we attempted to use a practical risk of bias assessment tool, it was subjective and may not be sensitive to the complexities of all the clinical trial designs in our study.

Conclusion

In this systematic review, we found that the evidence base for multi-component nursing home palliative care interventions is limited. We identified a small group of diverse clinical trials that enhanced nursing home palliative care and improved symptom assessment and management processes. Effects on other outcomes such as reducing potentially burdensome care, resident /surrogate reported symptoms, advance care planning discussions and documentation, and family members’ evaluation of care is lacking and warrants further study. There is a critical gap in evidence for quality of life and costs outcomes. Future research should focus on developing nursing home palliative care interventions that contain components important to staff, providers, residents and their family caregivers, provide access to experts, and identify residents who would most benefit from services. In addition, palliative care interventions are highly complex and outcome evaluations should explore how an intervention is implemented so that if successful, it can be tailored to nursing home environments.⁴⁴