If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
The Effectiveness of Advance Care Planning in Improving End-of-Life Outcomes for People With Dementia and Their Carers: A Systematic Review and Critical Discussion
Address correspondence to: Josie Dixon, MSc, Personal Social Services Research Unit (PSSRU), London School of Economics and Political Science (LSE), Houghton Street, London WC2A 2AE, United Kingdom.
End-of-life care for people with dementia can be poor, involving emergency hospital admissions, burdensome treatments of uncertain value, and undertreatment of pain and other symptoms. Advance care planning (ACP) is identified, in England and elsewhere, as a means of improving end-of-life outcomes for people with dementia and their carers.
Objective
To systematically and critically review empirical evidence concerning the effectiveness of ACP in improving end-of-life outcomes for people with dementia and their carers.
Methods
Systematic searches of academic databases (CINAHL Plus with full text, PsycINFO, SocINDEX with full text, and PubMed) were conducted to identify research studies, published between January 2000–January 2017 and involving statistical methods, in which ACP is an intervention or independent variable, and in which end-of-life outcomes for people with dementia and/or their carers are reported.
Results
A total of 18 relevant studies were identified. Most found ACP to be associated with some improved end-of-life outcomes. Studies were predominantly, but not exclusively, from the U.S. and care home-based. Type of ACP and outcome measures varied. Quality was assessed using National Institute of Health and Care Excellence quality appraisal checklists. Over half of the studies were of moderate to high quality. Three were randomized controlled trials, two of which were low quality.
Conclusion
There is a need for more high-quality outcome studies, particularly using randomized designs to control for confounding. These need to be underpinned by sufficient development work and process evaluation to clarify the appropriateness of outcome measures, explore implementation issues and identify “active elements.”
Advance care planning (ACP) is a voluntary process of discussion and review concerning preferences for future care, in the event that a person no longer has capacity to make decisions for himself or herself. There is growing interest, in England and elsewhere, in ACP as a way of supporting people's autonomy even when they can no longer speak for themselves and improving the quality and appropriateness of end-of-life care,
on behalf of the European Association for Palliative Care (EAPC) White paper defining optimal palliative care in older people with dementia: a Delphi study and recommendations from the European Association for Palliative Care.
House of Representatives Standing Committee on Health and Ageing Thinking ahead: Report on the inquiry into dementia. Early diagnosis and intervention.
The Parliament of the Commonwealth of Australia,
Canberra (Australia)2013
ACP conversations may take place with health care or other professionals and will generally involve a person's relative or carer (unpaid caregiver). In the case of dementia, ACP conversations may also be conducted with a carer within a “best interests” framework. The outcomes from such discussions are usually documented. In England, an advance statement sets out general preferences, an advance decision (sometimes known as an advance decision to refuse treatment) documents a decision to refuse specific medical treatment in certain circumstances and is usually legally binding, and a person may also assign Lasting Power of Attorney for Health and Welfare (Lasting Power of Attorney for property and finance can also be assigned). Similar provisions apply in comparable countries, such as the United States (U.S.), Canada, Australia, and many European countries, albeit within slightly different legal frameworks. In the U.S., preferences for future care are generally set out in advance directives, an ACP conversation is (as of January 2016) included as a separate and billable Medicare service and the Patient Self-Determination Act (1990) requires health and care facilities in receipt of Medicare or Medicaid funds to provide information to patients about advance directives and to enter existing advance directives in patients' medical records.
Dementia, Advanced Illness, and End of Life
A palliative approach in the care of people with dementia, particularly in the advanced stages, is recommended but not always taken.
The typical dwindling health trajectory associated with dementia makes it difficult to identify a terminal stage. Professionals and families may lack confidence, and be uncomfortable, in discussing end of life. They may be reluctant to limit life-prolonging treatments to avoid experiencing a sense of professional failure or, in the case of families, of “abandoning” a loved one.
As a result, many people with dementia, toward the end of life, have numerous emergency hospital admissions and, despite the likelihood of poor outcomes,
Behavioral and psychological symptoms of dementia (BSPD) are often exacerbated by hospitalizations and invasive treatments, with a consequent increase in the risk of restraint.
At the same time, there is evidence that people with dementia are often under-treated for pain and other symptoms, and that they and their carers frequently have unmet need for emotional, social and spiritual support.
ACP may help to improve end-of-life outcomes by creating an opportunity for people with dementia (or, if in later stages, their carers) to consider, in advance and in a supported way, the circumstances in which they would prefer to limit life-prolonging treatments and instead opt for an approach that prioritizes quality of life. Importantly, ACP limits the need to make difficult decisions in crisis and allows a person with dementia to, if they choose, participate directly in decision-making about their end-of-life care. ACP may have special relevance in dementia given inevitable loss of capacity and ability to communicate. We know that people who engage in ACP (in general, rather than just those with dementia) commonly use it to limit life-prolonging treatment in specified circumstances, generally where the risks, burdens, and benefits of active treatment are finely balanced, and they are particularly likely to do so when well informed about the risks and benefits of potential treatments.
on behalf of Project PRISMA Priorities for treatment, care and information if faced with serious illness: a comparative population-based survey in seven European countries.
We would, therefore, expect ACP with (or for) people with dementia to be associated with more frequent use of a palliative approach, fewer hospitalizations, and unwanted and burdensome treatments and with less emotional distress. We would also expect to see improved congruence between people's preferences and the care provided in practice, as well as greater, patient or carer, satisfaction with care.
Challenges of Advance Care Planning in Dementia
Despite the likely benefits, however, implementation of ACP in dementia presents a range of special challenges. Timing discussions is difficult, with the point of diagnosis generally considered too early but with pressure to not leave the discussion so late that the person with dementia is unable to participate meaningfully.
These include planning for a future unknown self and, for care providers, balancing a person with dementia's apparent wishes or best interests with previously expressed preferences.
These challenges are especially acute in the case of dementia, where the person will retain the ability to feel emotion and experience the outside world after losing capacity and where the period of incapacity is likely to be an extended one.
as well as complicated emotions, including anticipatory grief and, potentially, a sense of relief, and associated guilt, when the person they care for finally dies.
There is also specific evidence that people with dementia, while potentially willing to have ACP conversations, are often reluctant to set their wishes down in writing for fear of constraining the discretion of their carers and health care providers.
and considerable resources and training have been put into supporting generalist health care professionals to undertake such discussions in clinical practice.
However, evidence for the effectiveness and cost-effectiveness of ACP for people with dementia is currently unclear while implementation challenges appear to be significant. This review aims to systematically identify, and critically assess, outcome studies indicating the potential effectiveness and cost-effectiveness of ACP for people with dementia.
Given the considerable challenges associated with implementing ACP in dementia, establishing when and how ACP can be best undertaken with people with dementia has been identified as a key research priority, by the James Lind Alliance/Palliative and end-of-life care Priority Setting Partnership
Palliative and end of life care Priority Setting Partnership and James Lind Alliance Putting patients, carers and clinicians at the heart of palliative and end of life care research.
on behalf of the European Association for Palliative Care (EAPC) White paper defining optimal palliative care in older people with dementia: a Delphi study and recommendations from the European Association for Palliative Care.
Although it is not within the scope of this study to review all relevant evidence about “what works,” this review does consider what the identified outcome studies can tell us, if anything, about successful approaches for undertaking ACP with people with dementia and/or their carers.
This review was undertaken as part of the Modeling Outcomes in Dementia (MODEM) study. This four-year (2014–2018) project is led by a team from the London School of Economics and Political Science in partnership with the Universities of Newcastle, Southampton, and Sussex and the International Longevity Centre-UK. It has the aim of comprehensively modeling the outcome and cost impacts of evidence-based interventions in dementia in England in the period of 2040.
Although the MODEM study has a focus on England, this review covers international evidence.
Methods
Review Strategy
The strategy adopted by reviews conducted as part of the MODEM study is to ensure that, where a high-quality systematic review already exists, rather than doing a new one, the existing review is updated and, where necessary, extended. The process for our systematic review was therefore two-stage. First, a “review of reviews” was conducted, involving identification and quality assessment of the most relevant systematic review(s). Relevant individual studies in the systematic review(s) were then identified and abstracts and, where relevant, full articles were assessed against inclusion and exclusion criteria (see the section Inclusion and Exclusion Criteria). The second stage was to systematically search academic databases for further articles to update and/or supplement those already identified. Our review meets all PRISMA criteria, and our search strategy was developed with the assistance of a research librarian.
Inclusion and Exclusion Criteria
We sought studies of any design using statistical methods, with ACP as the intervention or an independent variable in a correlation study and with reported end-of-life outcomes for people with dementia and/or their carers. ACP was taken to mean having a written advance directive or other documented advance care plan, having a facilitated ACP discussion or having participated in an ACP intervention. We included only articles published in English in peer-reviewed journals. Our date range ran until the end of January 2017, with the start date determined by that of any relevant identified systematic review(s), but to be no later than 2006 to allow at least 10 years of publications to be considered. In the event, our review covered the period from January 2000 to January 2017 inclusive. We only included studies where the sample was exclusively or predominantly comprised of people with dementia or where results for people with dementia were separately reported. We excluded studies about the acceptability of ACP, purely descriptive studies (e.g., only reporting prevalence), studies about attitudes and intentions, studies specifically measuring the effects of medical orders (e.g., do-not-hospitalize orders), and study protocols. The same inclusion and exclusion criteria were applied to all studies, whether identified through a systematic review or through searches of research databases.
Identifying Relevant Systematic Reviews
We systematically searched for systematic reviews of ACP that potentially included individual studies in which ACP was an intervention or independent variable and in which end-of-life outcomes for patients and/or carers were reported. The systematic reviews, themselves, did not have to be dementia specific, as long as they potentially contained individual studies involving people with dementia and their carers. Articles for these studies were then separately retrieved and reviewed individually. The following databases were searched across all fields: CINAHL Plus with full text, PsycINFO, SocINDEX with full text and PubMed, using the search terms (“advance care plan*” OR “advance directive” OR “advance decision*” OR “living will” OR “end of life decision*” OR “end-of-life decision*”) and either (“review”) or using the filter ‘systematic reviews’ depending on the database.
A total of nine unique reviews were identified as potentially containing studies that met our criteria.
Abstracts, or full articles where necessary, were screened for relevance, by the lead author and, independently, by a second researcher. A systematic review of the noneconomic effects of ACP, for all populations not just people with dementia, by Brinkman-Stoppelenburg et al.
was identified as the most recent and comprehensive for our purposes. The other reviews were less relevant or complete. One focused specifically on the effectiveness of ACP in dementia but was more restricted in the types of study designs it included.
This review identified four studies, three of which were also included in the review by Brinkman-Stoppelenburg et al., whereas one was a wide ranging palliative care intervention that did not meet our inclusion criteria. Two further reviews reported solely on the economic outcomes associated with ACP.
These identified no studies involving people with dementia published between 2000 and 2012 that were not already included in the review by Brinkman-Stoppelenburg et al., reflecting the fact that studies reporting the economic effects of ACP tend to do so alongside other outcome measures. One further review focused on facilitators and barriers to ACP in dementia, rather than end-of-life outcomes,
In line with other systematic reviews conducted as part of the MODEM study, the review by Brinkman-Stoppelenburg et al. was assessed for quality against the Health Evidence Quality Assessment Tool for Review Articles (McMaster University) and graded as “strong,” the highest rating possible. In the interests of rigor, we reviewed abstracts and full articles where necessary, for all studies that could possibly include people with dementia and not just those identified as involving people with dementia by the authors. A total of 13 relevant individual articles were identified and retrieved in full for data extraction.
Using the review by Brinkman-Stoppelenburg et al. (covering 2000–2012 inclusive) as a baseline, additional systematic searches were conducted for the 49-month period from January 2013 to January 2017 inclusive. We used a more targeted search strategy than that used by Brinkman-Stoppelenburg et al., because of our focus on ACP for people with dementia and their carers and because we were interested in a narrower definition of ACP than the review by Brinkman-Stoppelenburg et al. (e.g., we did not include studies that only looked at medical orders). The following databases were searched across all fields; CINAHL Plus with full text, PsycINFO, SocINDEX with full text, and PubMed. The search terms used were (“advance care plan*” OR “advance directive” OR “advance decision*” OR “living will” OR “end of life decision*” OR “end-of-life decision*”) AND (“dementia” OR “cognitive impairment” OR “Alzheimer*”). A total of 164 unique references were retrieved. Abstracts, and full articles where necessary, were reviewed by the lead author and, independently, by a second researcher. Five studies met the inclusion criteria.
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
The references within these articles were also handsearched and, as a result, one further relevant study was identified, although this was eventually excluded since sample attrition meant that no statistical analysis was possible.
For each of the 18 included articles, the country in which the study was conducted, the type of ACP intervention or variable, the setting, sample composition and size, and a summary of the statistical results were extracted (see Table 1). We also identified the research design used and, in line with other reviews conducted as part of the MODEM study, each article was assessed against the National Institute of Health and Care Excellence (NICE) Quality Appraisal Checklist, either for quantitative intervention studies (NICE, 2012, Appendix F) or for quantitative studies reporting correlations and associations (NICE, 2012, Appendix G).
These checklists are specifically designed to identify rigor across the wide range of research designs that are commonly used in public health research, ranging from “small (experimental) randomized controlled trials through to large-scale observational studies” (p.192). They are appropriate for situations where randomized controlled trials are rare or difficult to conduct and helpful for identifying rigor in observational studies that explore and generate hypotheses about causal relationships before these are tested using more expensive and/or experimental designs.
Consequently, therefore, a well-conducted observational study may be rated more highly than a poor quality randomized controlled trial. The assessment resulted in two summary scores, one for internal validity and one for external validity (see Table 1).
Table 1Description of Included Studies and Reported Effects of Advance Care Planning
Authors, Date, Country, ACP Intervention, or Variable
All interventions/outcomes are included. (✓) or () indicates whether there were statistically significant positive results. Effect sizes are only reported where statistically significant.
Assessed using the National Institute of Health and Care Excellence (NICE) Quality Appraisal Checklist, either for quantitative intervention studies (Appendix F) or for quantitative studies reporting correlations and associations (Appendix G) as published National Institute of Health and Care Excellence (NICE). Process and methods guides: methods for the development of NICE public health guidance (third edition). 2012. Available scores are as follows: (++) All or most of the stated criteria have been fulfilled, and where they have not been fulfilled the conclusions are very unlikely to alter; (+) some of the stated criteria have been fulfilled, and where they have not been fulfilled, or not adequately described, the conclusions are unlikely to alter (−). Few or none of the stated criteria have been fulfilled and the conclusions are likely or very likely to alter. In the current review, assessments that were considered to fall between two different scores were given both scores separated with a forward slash.
Care homes
Garden et al., 2016 U.K. ACP intervention: ACP facilitation by two specially trained registered general nurses, supported by a consultant-liaison psychiatrist. ACP could include DNACPR. Completed plans were sent to care homes, carers, general practitioners, and secondary care. Small-group training for care home staff in management of end-stage dementia.
Seven care homes, 283 residents with dementia (107 of these and/or their carer agreed to ACP). 250 care home staff trained.
(✓) Hospital admissions: decrease from 202 to 91 (55%) admissions over two years. (✓) Place of death: 67/68 residents died in preferred place. (✓) Carer satisfaction: 64/80 carers who participated in ACP were surveyed (80% response rate); 92% rated service 9/10.
Intervention study Service evaluation, no comparator. Internally valid (−) Externally valid (−)
Vandervoort et al., 2014 Belgium ACP variables: 1) Advance directive documented; 2) nurse-led end-of-life conversation with resident; 3) nurse-led end-of-life conversation with carer, relative, or friend.
Stratified random sample of nursing (care) homes (n = 69, 58% response rate), from which 101/205 residents (49% response rate) with dementia, all levels of severity, dying over three months, were recruited. An advance directive was present for 17 (17.5%) residents. Nurse-led end-of-life conversation with 13 (13.7%) residents and 57 (60.6%) carers, relatives, or friends. Approximately 50% of residents were not in an advanced stage of dementia a month before dying.
Advance directive documented (✓) Emotional Distress Subscale, Comfort Assessment in Dying with Dementia (CAD-EOLD): AOR: 2.99 (1.1–8.3, P ≤ 0.05) () Total score and physical distress subscale, dying symptoms, and well-being subscales of CAD-EOLD. Nurse-led end-of-life conversation with resident () CAD-EOLD, total score, and all subscales of CAD-EOLD. Nurse-led end-of-life conversation with carer, relative, or friend (✓) Physical Distress Subscale, CAD-EOLD (AOR: 0.28; CI: 0.08–0.98, P = ≤ 0.05) (✓) Dying Symptoms Subscale, CAD-EOLD (AOR: 0.26; CI: 0.1–0.6, P ≤ 0.05) () Total score and well-being subscale, CAD-EOLD
Correlational study Retrospective cohort Internally valid (++) Externally valid (+/++)
Livingston et al., 2013 U.K. ACP intervention: A 10-session manual-based, interactive staff training program, emphasizing preferred place of care, how to have difficult conversations and ACP (n.b. delivered alongside Gold Standards Framework training for care home managers).
120-bed nursing (care) home, providing care recognizing Jewish traditions, beliefs, and cultures for people of various religions. Fifty-six residents with dementia preintervention and 42 residents with dementia postintervention were recruited. Qualitative interviews with relatives (n = 8 postintervention).
(✓) Place of death: increase in proportion dying in care home (rather than hospital), 47%–76%, χ2(1) = 5.3, P = 0.02. (✓) Carer satisfaction: 7.5/10 (SD = 1.3) preintervention to 9.1/10 (SD = 2.4) postintervention; t = 17.6, P = 0.06. (✓) Concordance: End-of-life care consistent with advance wishes regarding hospitalization or DNR, increased (71% to 100%, P = 0.04). () Hospital days in last three months
Intervention study Small-scale exploratory study, no comparator Before and after, no control. Included qualitative interviews Internally valid (−) Externally valid (−/+)
Kiely et al., 2012 U.S. ACP variables: 1) Health care provider spent >15 minutes with carer discussing ACP, 2) physician counseled carer about health problems resident may experience later on, and 3) physician counseled carer about resident life expectancy.
323 residents with severe dementia, 60+ years and their carers from nursing (care) homes with 60+ beds, < 60 miles of Boston, recruited as part of CASCADE study.
Satisfaction with Care at End of Life in Dementia (SWC-EOLD) scale (✓) >15 minutes discussing ACP, yes 32.8 (32.2, 33.3) vs. no 30.5 (29.9, 31.0), 2.3 difference in score, P = 0.0001 (✓) Counseled re: future health problems, yes 32.3 (31.6–33.1) vs. no 31.1 (30.5–31.6), 1.2 difference in score, P = .0.002 (✓) Counseled re: life expectancy, yes 32.8 (31.8–33.8) vs. no 31.2 (30.6–31.8), 1.6 difference in score, P = 0.002
Correlational study Prospective cohort, to test responsiveness of carer satisfaction scale Internally valid (++) Externally valid (+)
Vandervoort et al., 2012 Belgium ACP variable: Advance directive in records.
764 residents with dementia in 345 nursing (care) homes in Flanders (58% response rate) dying over a two-month period. Three percent had an advance directive.
() Place of death, quality of life, perceived “mildness” of death and Edmonton Symptom Assessment Scale
Correlational study Retrospective cohort Internally valid (++) Externally valid (+/++)
Gozalo et al., 2011 U.S. ACP variable: Written advance directive.
474,829 nursing (care) home residents identified using Minimum Data Set (MDS) data for all care (care) home residents in U.S. and Medicare claims from January 1, 2000, to December 31, 2007. Eligible residents had advanced cognitive impairment and were age over 66 years, enrolled in a Medicare fee-for-service plan, resident in a nursing (care) home 120 days before death, in need of extensive assistance with activities of daily living. 91,759 had an advance directive, 19% had at least one burdensome transition in last 3 days of life, 6.8% had a do-not-hospitalize order.
(✓) Burdensome transition 1) transition in last three days of life; 2) lack of continuity in nursing (care) homes after hospitalization in last 90 days of life; and 3) multiple hospitalizations in last 90 days of life. Lack of advance directive = ARR: 1.15 (1.14–1.17)
Correlational study Retrospective cohort Internally valid (+/++) Externally valid (++)
Engel et al., 2006 U.S. ACP variables: 1) Health care provider spent >15 minutes with carer discussing ACP, 2) physician counseled carer about health problems resident may experience later on, 3) comfort is a primary goal of care, and/or 4) had a living will/advance directive
148 residents with severe dementia, 60+ years and their carers from nursing (care) homes with 60+ beds, < 60 miles of Boston, recruited as part of CASCADE study.
Satisfaction with Care at End of Life in Dementia (SWC-EOLD) scale (✓)>15 minutes discussing ACP, regression coefficient 2.39 (95% CI: 1.16–3.61, P < .001) () Physician counseled carer about health problems resident may experience later on, comfort is a primary goal of care or had a living will/advance directive
Correlational study Prospective cohort, to test responsiveness of Carer Satisfaction Scale Internally valid (++) Externally valid (+)
Caplan et al., 2006 U.S. ACP intervention: Training care home staff, local health care providers, residents, and families about dementia and about a “hospital in the home” service. Offering facilitated ACP using the Let Me Decide advance directive to residents with capacity/carers of residents without capacity
Intervention area comprised two hospitals and 21 surrounding nursing (care) homes (19 agreed to participate), control area comprised hospital and 13 surrounding nursing (care) homes. Referrals were received for 63 residents in intervention area. Of these, 45 (71%) agreed to ACP. Five had capacity and undertook ACP. Of 40 without capacity, one already had an ACP, 20 residents had discussion and preferences recorded in notes but no formal plan, 10 residents had discussion but not documented in notes, three carers completed a “plan of treatment” and six carers declined.
(✓) Emergency ambulance calls: change from 2002 to 2003 in intervention −1% vs. control +21% P = 0.0019. (✓) Hospital day: increase in RR of hospital day per nursing home bed, from (preintervention), RR: 1.01 (0.98–1.04, P = 0.442, 95% CI) to (postintervention) RR: 0.74 (0.72–0.77, P < 0.0001, 95% CI). (✓) Hospital admission increase in RR of hospital admission per nursing home bed, from (preintervention), RR: 1.07 (1.03–1.11, P = 0.0005, 95% CI) to (postintervention) RR = 0.89 (0.85–0.93; P < 0.0001, 95% CI)
Morrison et al., 2005 U.S. ACP intervention: Training for social workers in ACP using small-group workshops and role play/practice sessions. ACP discussions with residents and carers at admission, after change in clinical status, and yearly. Completion of treatment directive (could include DNACPR). “Flagging” advance directives on care home charts. Formal review at team meetings. Feedback to physicians on concordance of care provided.
A nursing (care) home with eight long-term care units, four social workers, each assigned to two of the units. 139/201 residents (70% response rate). Of these, 96 were in a control group, of which 76 (79%) had dementia, and 43 in an intervention group, of which 40 (93%) had dementia.
(✓) Concordance: 2/49 (5%) intervention residents received treatment in conflict with stated wishes vs. 17/96 (18%) control residents (P = 0.04).
Mitchell et al., 2003a U.S. ACP variable: “Living will” (advance directive), as recorded in the National Repository Resident Assessment Instrument Minimum Data Set (MDS)
186,835 residents of Medicare- or Medicaid-certified U.S. nursing (care) homes, with advanced cognitive impairment.
(✓) Enteral (“tube”) feeding: Characteristics associated with enteral (“tube”) feeding include the absence of a living will OR 1.32 (1.21–1.35, 95% CI)
Correlational study Cross-sectional, administrative data Internally valid (++) Externally valid (++)
Mitchell et al. 2003b U.S. ACP variable: Proportion of residents with advance directive, defined in OSCAR data set as having a living will or durable power-of-attorney for health care.
1057 licensed nursing (care) homes in six U.S. states; all residents 65 + with advanced cognitive impairment and feeding tube placed in a one-year period
(✓) Enteral (“tube”) feeding: Characteristics associated with above median level (>10%) of enteral feeding include less than median level (<69%) of residents with advance directives OR: 1.66 (1.27–2.17, 95% CI)
Correlational study Cross-sectional, administrative data Internally valid (++) Externally valid (++)
Ahronheim et al., 2001 U.S. ACP variable: “Living will” (advance directive), as recorded in the National Repository Resident Assessment Instrument Minimum Data Set (MDS).
57,029 nursing (care) home residents across all U.S. states, age 65+ years, with severe cognitive impairment.
(✓) Enteral (“tube”) feeding: Characteristics associated with enteral (“tube”) feeding include having a living will; OR: 0.68 (no CI provided); P < 001
Correlational study Cross-sectional, administrative data Internally valid (++) Externally valid (++)
Gessert et al., 2000 U.S. ACP variable: “Living will” (advance directive), as recorded in the National Repository Resident Assessment Instrument Minimum Data Set (MDS).
4997 nursing (care) home residents with severe and irreversible cognitive impairment, identified using data in U.S. Minimum Data Set (MDS) reports.
(✓) Enteral (“tube”) feeding: Characteristics associated with enteral (“tube”) feeding include having a living will; 1.287 (1.01 1–1.637); P = 0.0402
Correlational study Cross-sectional, administrative data Internally valid (++) Externally valid (++)
Molloy et al., 2000 Canada ACP intervention: Training care home staff, local health care providers, residents, and families. Offering facilitated ACP using the Let Me Decide advance directive to residents with capacity (49% completed)/carers of residents without capacity (78% completed).
1292 residents from six pair-matched nursing (care) homes (three of which were randomized to intervention). These were selected from survey of 171/215 (80% response rate) nursing (care) homes in a single province. Of the 636 residents in the intervention homes, 408 (64%) were assessed as incompetent. In the control group, of the 656 residents, 389 (59%) were assessed as incompetent. At outset, all homes had less that 25% residents with advance directives and at end control had 70% and control had 57%.
(✓) Hospital days, intervention 2.61 vs. control 5.86, P = 0.01 (✓) Risk of hospitalization, intervention 0.27 vs. control 0.48, P = 0.001 (✓) Hospital costs only, Can$1772 vs. control $3869, P = 0.003 (✓) Health care + intervention costs, Can$3490 vs. control Can$5239, P = 0.01 () Satisfaction
Meier et al. 2001 U.S. ACP variable: Advance directive in records (covering living will, an appointed health care proxy or clear oral, or written evidence of the patient's wishes)
99/192 (52% response rate) patients with advanced dementia and their carers recruited in single hospital over a three-year period, 51 of whom left hospital with new feeding tube. 15 (15.2%) had AD.
() Enteral (“tube”) feeding: HR = 0.32 (0.04–2.8, 95% CI). Enteral feeding also found not to be associated with survival (HR = 0.97, 0.5–1.9, 95% CI, P = 0.9)
Correlational study Prospective cohort Internally valid (++) Externally valid (+)
Community
Hilgeman et al., 2014 U.S. ACP intervention: Preserving Identity and Planning for Advance Care (PIPAC) intervention; in-home sessions by trained staff over four–six weeks; identity-maintaining activities; and ACP discussion.
18 people with mild dementia and a family contact (carer), referred by local community partners. Blocked randomization stratified by ethnicity and sex.
Person with dementia (✓) Depression (CSDD): F (1, 15) 5.51, ηp2 0: 27 (large effect). (✓) Quality of life (BASQID): F(1, 15) 1.13, ηp
0:07 (medium effect) (✓) Coping (IMMEL): F(1, 16) 3.35, ηp2 0.17 (large effect) (✓) Decisional conflict: F (1, 14) 3.74, ηp2 0.21 (large effect) (✓) Mobility factor EQ-5D: F (1,15) 2.72 ηp2 0.15 (large effect) () Self-reported anxiety (CSDD), meaning, social engagement, emotional and anticipated support, and quality of life as measured by the QOL-AD; and self-care, usual activities, pain/discomfort, anxiety/depression, and EQVAS on EQ-5D Family contact (✓) Depression (CSDD): F(1, 14) . 1.72 ηp2 0.11 (medium effect) (✓) Quality of Life (QoL-AD): F(1, 14). 5.41 ηp2 0.28 (large effect) (✓) Self-care on EQ-5D: F(1, 11). 1.44 ηp2 0:12 (medium effect) (✓) Usual activities on EQ-5D: F(1, 10). 2.52 ηp2 0.18 (large effect) (✓) Anxiety/depression on EQ-5D: F(1, 11). 2.03 ηp2 0.16 (large effect) () Self-rated anxiety (CDSS), social engagement, and mobility, pain/discomfort, and EQVAS on EQ-5D
Intervention study Feasibility trial using block randomization. Internally valid (−/+) Externally valid (−)
Population (all settings)
Nicholas et al., 2014 U.S. ACP variable: Treatment-limiting advance directive
3876 Health and Retirement Study (HRS) participants, who died age 65+ 1998–2007; grouped by cognition (severe dementia/cognitive impairments but no dementia [CIND] and mild dementia/normal cognition) and setting (community/nursing [care] home)
(✓) Medicare spending in last six months: severe dementia in community $32,200 without advance directive vs. $20,700 with (P ≤ 0.01); CIND/mild dementia in nursing (care) home without advance directive $34,700 vs. with 30,600 (P ≤ 0.05); no association in other subgroups. (✓) Hospital death: severe dementia in community without advance directive 31.8% vs. with 13.9% (P ≤ 0.001); normal cognition in nursing (care) home without advance directive 27.2% vs. with 17.9% (P ≤ 0.01); severe dementia in nursing (care) home without advance directive 20.6% vs. with 14.6% (P ≤ 0.05); no association in other subgroups. (✓) ICU use: severe dementia in community without advance directive 19.6% vs. with 10.1% (P ≤ 0.01); no association in other subgroups. () Life-sustaining treatments.
Correlational study Retrospective cohort Internally valid (++) Externally valid (++)
Volicer et al., 2003 U.S. ACP variable: Written advance directive
156 recently bereaved carers of people with dementia (27% estimated response rate), accessed through Alzheimer Association Chapters (38/201) Geriatric Research Education Clinical Center (GRECC) programs (12/16), and Alzheimer's Disease Centers (ADCs) (5/27).
(✓) Hospital days without advance directive 12.8 days +/− 20.9 SD. vs. with 5.7 days +/− 12.6 SD; f = 5.09, P = 0.026) (✓) Place of death: Hospital death (without advance directive 25% vs. with 18%) and nursing (care) home death (with advance directive 54% vs. without 31%) χ2 = 9.766, P = 0.045 () Comfort Assessment in Dying (CAD-EOLD) score
ACP = advance care planning; ICU = intensive care unit; DNR = do-not-resuscitate; AOR = adjusted odds ratio; ARR = adjusted risk ratio; RR = risk ratio; CASCADE = Choices, Attitudes, and Strategies for Care of Advanced Dementia at the End-of-Life; OR = odds ratio; HR = hazard ratio; DNACPR = Do-not-attempt-cardiopulmonary resuscitation; CSDD = Cornell Scale for Depression in dementia; BASQID = Bath Assessment of Subjective Quality of Life in Dementia; IMMEL = Index for Managing Memory Loss; EQ-5D = EuroQol 5 Dimesion Scale; QOL-AD = Quality of Life in Alzheimer's Disease; EQ-VAS = EuroQol Visual Analogue Scale; CAD-EOL = Comfort Assessment in Dying.
a All interventions/outcomes are included. (✓) or () indicates whether there were statistically significant positive results. Effect sizes are only reported where statistically significant.
b Assessed using the National Institute of Health and Care Excellence (NICE) Quality Appraisal Checklist, either for quantitative intervention studies (Appendix F) or for quantitative studies reporting correlations and associations (Appendix G) as published National Institute of Health and Care Excellence (NICE). Process and methods guides: methods for the development of NICE public health guidance (third edition). 2012. Available scores are as follows: (++) All or most of the stated criteria have been fulfilled, and where they have not been fulfilled the conclusions are very unlikely to alter; (+) some of the stated criteria have been fulfilled, and where they have not been fulfilled, or not adequately described, the conclusions are unlikely to alter (−). Few or none of the stated criteria have been fulfilled and the conclusions are likely or very likely to alter. In the current review, assessments that were considered to fall between two different scores were given both scores separated with a forward slash.
A key aspect of our critical discussion is to consider what light the identified studies shed on when and how ACP can be successfully conducted with people with dementia and/or their carers. It was not our intention to identify all relevant research into processes of ACP (which would have been far too large an undertaking and not within the scope of this review). Rather, the aim was to identify relevant information provided within the included articles, or companion articles produced as part of the same overall study, about exactly how the results reported in this review were achieved. Information about process, particularly where positive results are reported, is important, both to allow studies to be replicated, potentially using more rigorous research designs, and for facilitating the further development of successful approaches. It is not our intention, where information about underlying processes is limited, to criticize the scope of individual studies (which are delivered within the constraints of methodological and other limitations), but rather to systematically map what is known about underlying processes in studies for which there is currently some outcome evidence.
In this, we were particularly interested in information about implementation. In the case of intervention studies, we considered how complete the description of the ACP intervention was and identified any discussion of implementation challenges and how these were overcome. We were also particularly interested in information about causal pathways and “active elements.” Here, we considered whether studies had a clear and sufficiently detailed “intervention logic,” mapping the expected causal pathways between the ACP intervention or variable and end-of-life outcomes, and whether or not this was tested or evidenced as part of the study. These data were extracted by reading each of the articles in full. For intervention studies, companion articles produced as part of the same study were identified by carrying out targeted searches using author name and/or key identifying terms from the study (such as the name of the research study or intervention program). We identified one companion article reporting qualitative research
These were also read in full. The data extracted on process are summarized in Table 2.
Table 2Summary of Information About Processes Provided
Authors, Date, Country, ACP Intervention, or Variable
Information About Processes Provided
Garden et al., 2016 U.K. ACP intervention: ACP facilitation by two specially trained registered general nurses, supported by a consultant-liaison psychiatrist. ACP could include DNACPR. Completed plans were sent to care homes, carers, general practitioners, and secondary care. Small-group training for care home staff in management of end-stage dementia.
•
A full description of the intervention is provided
•
The authors state that “assuming the reduction in hospital admissions observed in this project was due to the BCHS [the Bromhead Care Home Service], we cannot be sure whether it was advance care planning, staff education or the combination that was responsible.”
Vandervoort et al., 2014 Belgium ACP variables: 1) Advance directive documented; 2) nurse-led end-of-life conversation with resident; 3) nurse-led end-of-life conversation with carer, relative, or friend.
•
The authors note that they “cannot disentangle the particular mechanisms responsible” for the identified associations
•
The authors, however, hypothesize that advance directives may reduce fear and anxiety (but not physical distress or dying symptoms) because it promotes communication (this would be relevant for the approximately 50% who die before dementia is advanced). Alternatively, advance directives may lead carers (or staff, who in turn discuss with carers) to interpret emotional well-being at end of life more positively
•
Nurse-led conversations with carers had a negative association with quality of dying. The authors hypothesize that this is because most conversations are occurring late, when residents are already experiencing clinical complications at the end of life
•
The authors hypothesize that a lack of association between nurse-led conversations with residents and quality of dying is because “conversations between resident and nurses were limited in frequency, not repeated over time or not intense or thorough enough.”
•
States that, whatever the causal relationships, evidence suggests there is a “strong relationship between patient-centered planning and quality of dying which should be unraveled in more detail in further research.” (p.5)
Livingston et al., 2013 U.K. ACP intervention: A 10-session manual-based, interactive staff training program, emphasizing preferred place of care, how to have difficult conversations and ACP (n.b. delivered alongside Gold Standards Framework training for care home managers).
•
Intervention is briefly described
•
The study includes qualitative research interviews with family carers (n = 8 postintervention/death of resident) and with staff (n = 58 preintervention and postintervention). Results from family carer interviews are reported very briefly (including a large table of direct quotes) and focus on satisfaction with care rather than process. Interviews with staff are more fully reported in separate article and focus on experiences prior to, and need for, the intervention (see Livingston et al., 2012), although improved staff confidence after training is briefly reported in the current study
•
The authors hypothesize that positive associations are related to the training addressing staff fears and concerns, as well as increasing knowledge.
Kiely et al., 2012 U.S. ACP variables: 1) Health care provider spent >15 minutes with carer discussing ACP; 2) physician counseled carer about health problems resident may experience later on; and 3) physician counseled carer about resident life expectancy.
•
SWC-EOLD is a validated measure. Study was designed to demonstrate that the SWC-EOLD (and SM-EOLD) measures are sensitive to clinically meaningful changes in number of relevant factors (including the ACP variables). These were selected a priori from the literature as likely to show differences in carer satisfaction (see also Engel et al., 2006, below).
Vandervoort et al., 2012 Belgium ACP variable: Advance directive in records.
•
The discussion notes the low rate of ACP, reflecting their relatively late introduction into Flemish care homes, and predominance of GP orders at late stage in place of ACP.
Gozalo et al., 2011 U.S. ACP variable: Written advance directive.
•
Having an advance directive is included as one of a number of variables measuring health status in a multiple regression model. No additional discussion about process or likely causal pathways.
Engel et al., 2006 U.S. ACP variables: 1) Health care provider spent >15 minutes with carer discussing ACP; 2) physician counseled carer about health problems resident may experience later on; 3) comfort is a primary goal of care; and/or 4) had a living will/advance directive
•
Factors (including ACP variables) thought to be associated with family satisfaction with care (SWC-EOLD) were selected based on the literature and clinical experience.
•
The discussion >15 minutes was conducted with physician (n = 23, 18.2%); nurse providing direct care (n = 45, 35.7%); social worker (n = 95, 75.4%); nurse administrator (n = 21, 16.7%); other administrator (n = 6, 4.2%); and other; physical therapist and ombudsman (n = 2, 1.6%), with some having a conversation with more than one type of provider
•
Provider type was not associated with whether advance directives were discussed for >15 minutes
•
The authors hypothesize that “time spent discussing specific directives may be a marker for overall better communication and shared decision-making between HCPs and health care professionals.”
Caplan et al., 2006 U.S. ACP intervention: Training care home staff, local health care providers, residents, and families about dementia and about a “hospital in the home” service. Offering facilitated ACP using the Let Me Decide advance directive to residents with capacity/carers of residents without capacity.
•
The intervention and the “hospital in the home (HITH)” scheme is only briefly described
•
Some qualitative data concerning staff and carer satisfaction with the intervention is reported (p.584), but unclear what methods used to obtain these data.
Morrison et al., 2005 U.S. ACP intervention: Training for social workers in ACP using small-group workshops and role play/practice sessions. ACP discussions with residents and carers at admission, after change in clinical status, and yearly. Completion of treatment directive (could include DNACPR). “Flagging” advance directives on care home charts. Formal review at team meetings. Feedback to physicians on congruence of care provided.
•
A full description of the intervention is provided
•
The rationale for delivery by social workers includes expertise in counseling and communication, and availability in most care homes
•
The authors state that “because of the size of the sample, it was not possible to determine the individual contributions of the key components of the intervention on the outcomes that were observed” and note that “larger studies will be required.”
Mitchell et al., 2003a U.S. ACP variable: Advance directive (“Living will”), as recorded in the National Repository Resident Assessment Instrument Minimum Data Set (MDS)
•
The authors note the need for future research to explore how ACP (having a “living will”) and DNR orders, which are both associated with lower rates of enteral (“tube”) feeding, relate to the broader process of shared decision-making.
Mitchell et al., 2003b U.S. ACP variable: Proportion of residents with advance directive, defined in OSCAR data set as having a living will or durable power of attorney for health care.
•
The authors state that the majority of advance directives do not specifically address artificial nutrition in patients with advanced cognitive impairment Therefore, they hypothesize that facilities that engage patients and carers in ACP are also more likely to engage them in discussions about tube feeding
•
The authors state that “because of the broad definition of advance directives in the OSCAR data set, the influence of specific types of directives could not be made.”
Ahronheim et al., 2001 U.S. ACP variable: “Living will” (advance directive), as recorded in the National Repository Resident Assessment Instrument Minimum Data set (MDS).
•
Analysis of differences between U.S. states, controlling for range of factors including advance directives. No additional discussion about process or likely causal pathways.
Gessert et al., 2000 U.S. ACP variable: “Living will” (advance directive), as recorded in the National Repository Resident Assessment Instrument Minimum Data set (MDS).
•
The authors note that “living wills” (advance directives) reflect the prior wishes of residents and DNR reflects current assessments of prognosis. Both are associated in this study with less likelihood of enteral (“tube”) feeding.
Molloy et al., 2000 Canada ACP intervention: Training care home staff, local health care providers, residents, and families. Offering facilitated ACP using the Let Me Decide advance directive to residents with capacity (49% completed)/carers of residents without capacity (78% completed).
•
A full description of the intervention is provided
•
The authors state that the Let Me Decide document has been previously evaluated in small-scale studies, which have found “consistent support from residents, families, and health care workers” although these studies had not found any effect on health care utilization
•
The authors hypothesize that the length and complexity of the ACP document may have deterred some from completing it
•
The authors hypothesize that the education components in the ACP intervention, rather than completion of advance directives, may have been responsible for associations identified
•
The authors hypothesize that the increased acknowledgment of patients' wishes might have lead to an overall higher threshold for hospital transfers.
Hospital
Meier et al. 2001 U.S. ACP variable: Advance directive in records (covering living will, an appointed health care proxy or clear oral or written evidence of the patient's wishes)
•
The authors hypothesize that the high evidentiary standard for decisions to forego artificial nutrition and hydration could influence the degree to which ACP and enteral (tube) feeding are associated in a hospital context.
Community
Hilgeman et al., 2014 U.S. ACP intervention: Preserving Identity and Planning for Advance Care (PIPAC) intervention; in-home sessions by trained staff over four–six weeks; identity-maintaining activities and ACP discussion.
•
A clear theoretical basis for the development of the intervention is provided
Population
Nicholas et al., 2014 U.S. ACP variable: Treatment-limiting advance directive
•
The authors note that data source (Health and Retirement Survey) provides no information on ACP processes
•
The authors discuss why ACP may influence end-of-life outcomes for people with severe dementia living in the community, while no similar association was found for people with cognitive impairments, mild dementia, or normal cognition living in the community; they hypothesize that people with severe dementia will not be able to contribute to decision-making and hence ACP is more likely to be activated and guide care
•
Severe dementia may also not be distinguished from delirium in an emergency situation, and so treatment “might not fully take into account the terminal nature of dementia.” ACP in this context is a way of improving clinicians' knowledge of a patient's diagnosis, as well as preferences.
Volicer et al., 2003 U.S. ACP variable: Written advance directive
•
No discussion of process
ACP = advance care planning; DNACPR = Do-Not-Attempt Cardiopulmonary Resuscitation; DNR = Do-Not-Resuscitate; SM-EOLD = Symptom Management at the End of Life in Dementia; DNR = Do-Not-Resuscitate.
Finally, to note, the studies use a range of terms for long-term care facilities, including “care home,” “nursing home,” and “residential care home.” These can have different meanings, particularly internationally. Given the population group that this review focuses on, it is likely that facilities will be able to provide some level of skilled nursing care. For consistent usage, however, we use the generic term “care home” throughout.
Results
In total, 18 relevant studies were identified. The review by Brinkman-Stoppelenburg et al. included 13 studies involving people with dementia and their carers.
These were not consistently identified as such in the review, with three of the studies identified only as involving care home residents but, upon closer inspection, were comprised primarily of people lacking capacity due to cognitive impairments and dementia.
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
Table 1 sets out, for each included study, the country that the study was conducted in, the type of ACP intervention or variable, the setting, sample composition and size, statistical results, the type of research design, and a quality rating (with separate scores given for internal validity and external validity).
Care Setting
Thirteen of the included studies are from the U.S.,
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
although there are also two general population studies, one community-based study and one hospital-based study.
ACP Intervention or Variable
A range of ACP interventions and variables were evaluated in the included studies. These include written advance directives, different types of end-of-life conversation, and various types of staff training intervention. Ten studies covered ACP undertaken with the person with dementia prior to losing capacity,
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
although in studies that use ACP documents as their ACP variable, face-to-face conversations may also occur.
Research Design and Quality
Twelve of the studies are correlational studies, testing the strength of associations between ACP variables and end-of-life outcomes, six of which used large, representative samples (ranging from n = 3876 to n = 474,829).
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
The majority of the intervention studies (including two of the randomized controlled studies) were of low to moderate quality. Overall, however, most studies were of moderate or high quality.
Participants
The studies involved people with dementia at different levels of severity, assessed using different diagnostic criteria. In three cases, the studies did not exclusively involve people with dementia, although people with dementia comprised the majority of the sample. In one of these cases, 93% of residents in the intervention group and 79% in the control group had dementia.
The third study was cross-sectional, covering a general population sample (n = 3876), but it reported results for people with severe and mild dementia (or cognitive impairments) separately and compared these to results for people with no dementia or cognitive impairment.
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
However, the outcomes measured in the studies were numerous and diverse, covering a range of health utilization and patient and carer outcome measures. Thirteen studies measured healthcare utilization outcomes including place of death, hospitalization, and specific treatments,
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
Two studies considered economic outcomes; a cohort study, which identified Medicare cost savings for people with severe dementia in the community (Nicholas et al.),
and a cluster randomized controlled trial, which identified cost savings across a range of health and hospital-based services for care home residents in Canada (Molloy et al.).
None of the studies, including the intervention studies, identified the cost of ACP, although Molloy et al. identified, in very broad terms, the amount of staff time used to implement the intervention.
Processes
The included studies were outcome studies, selected on the basis that they involved statistical analysis, ACP as an intervention or independent variable and end-of-life outcomes for people with dementia or their carers. They were not selected on the basis that they included in-depth analysis of process. Nonetheless, given that ACP is a complex intervention that can be conducted in various ways, we have identified available information given in the articles about underlying processes, seeking to clarify how positive results were obtained and the “active elements” involved.
In addition to reading the articles in full, for the six intervention studies, we also searched for any companion articles (e.g., from process evaluations or complementary qualitative studies). We found three additional articles; one qualitative study (that focused on experiences prior to, and the need for, the intervention)
Since none of the included articles involved extensive discussion of process and there were few companion articles, available information about underlying processes was limited.
Intervention studies varied in how well they described the intervention, with some studies including only brief descriptions.
No study discussed implementation challenges, explored how the intervention may have differed in practice from what was originally envisaged, or discussed any adaptations made. No study included a clear and detailed “intervention logic” (e.g., a flow diagram showing causal pathways between intervention and outcome) or any process measures. Some articles included descriptive results about implementation (e.g., how many ACP documents were completed, which professionals delivered ACP conversations, and so forth), and authors sometimes hypothesized about why they obtained the results they did, especially where these were unexpected.
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
A strength of our review is a clear focus on outcomes and effectiveness of ACP in dementia, including a critical discussion of the evidence identified. We employed a rigorous, multistage systematic search strategy, assisted by a research librarian. We used the systematic review by Brinkman-Stoppelenburg et al.
as a “baseline” for subsequent searches, and so, our review was dependent on the quality and comprehensiveness of this review for the completeness of records from 2000 to 2012. However, it was, in its design, clearly inclusive of the studies we wished to identify and we assessed its quality using a well-recognized quality-assessment tool and found it to be “strong,” the highest possible rating. We also referred to two other systematic reviews for economic outcomes of ACP
(the review by Brinkman-Stoppelenburg et al. covered only noneconomic outcomes) and were unable to find any articles for the same period that were not already included in the review by Brinkman-Stoppelenburg et al., reflecting the fact that economic outcomes have tended to be measured and reported alongside noneconomic outcomes. We further took the precaution of reviewing all articles in the review by Brinkman-Stoppelenburg et al. whether or not they were clearly identified by the authors as involving people with dementia and their families. This allowed us to identify three studies conducted in care homes as involving people with dementia which may otherwise have been overlooked.
We used a similarly rigorous strategy to search for articles published after 2012, including handsearching of references. There is a small risk that we may have missed studies in which results for people with dementia and their carers are reported but which do not include dementia terms in searchable fields. Thirteen of the studies are from the U.S. and three from other non-U.K. countries, with their different health systems and legal frameworks, potentially affecting generalizability. However, all the countries are comparable to England in socioeconomic terms. It is also possible that there might be publication biases. Quality assessment was undertaken using tools that are appropriate for research into complex interventions, where few randomized controlled studies are expected.
The Effectiveness of ACP in Dementia
The diversity of the included studies, while reflecting the inherent complexity of ACP for people with dementia, makes direct comparison between studies and overall interpretation of the evidence difficult. Most of the evidence was from care homes and, consequently, involved ACP at a late stage in the disease trajectory, including with carers within a “best interests” framework. These studies varied in quality, and results, particularly when looking just at studies of moderate to high quality, were mixed. Most of the care home–based studies considered advance directives, an ACP conversation of some sort or a multicomponent intervention. However, two moderate to high-quality studies with positive results used the narrower measure of a health care professional spending more than 15 minutes discussing ACP with a family carer.
In terms of outcomes, four high-quality care home–based studies with positive results, focused specifically on enteral “tube” feeding, a specific and burdensome treatment which, when delivered toward the end of life, has been shown to provide little to no benefit.
Of the remaining care home–based studies, two provide strong support for an association between ACP and end-of-life outcomes. One of these is a moderate/high-quality randomized controlled trial by Molloy et al.,
which found positive results for a series of health care utilization measures (but not carer satisfaction). The other is a high-quality correlational study using an administrative sample of nearly 500,000 people that found a correlation between advance directives and fewer burdensome transitions in the last three months of life.
Two other high-quality care home studies, based on survey data, found more mixed results for place of death and a range of patient experience measures.
There was less evidence concerning ACP in noncare home settings or about ACP initiated earlier in the illness trajectory. Two of the four studies in noncare home settings were of low quality.
The Bromhead Care Home Service: the impact of a service for care home residents with dementia on hospital admission and dying in preferred place of care.
The remaining two studies were high quality but focused on different populations and different end-of-life outcomes and produced different results. One, conducted in a hospital, found no association between having an advance directive and enteral (tube) feeding.
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
This found that people with severe dementia living in the community with a treatment-limiting advance directive (compared to those with severe dementia living in the community without an advance directive) experienced significantly reduced levels of hospital death and intensive care unit use and lower Medicare costs.
Choice of Outcome Measures
The studies used a diversity of outcome measures. This, in part, reflects methodological and data availability constraints, but it also reflects that ACP is thought to be associated with multiple potential benefits with little agreement about which are the most important. There was limited discussion about the choice of outcome measures in the included articles, particularly in the context of alternative measures. Exceptionally, Vandervoort et al.
stated that they had selected “quality of dying” outcome measures on the basis that previous studies had largely focused on health care utilization. The wide variety of outcome measures makes meaningful comparison between studies difficult. Morrison et al.,
for example, describe their study (involving a staff training intervention and measuring concordance of care) as “building on” previous research. The research referred to, however, not only involved different types of intervention (a system for translating preferences for care into medical orders, a staff training intervention, and a system-wide ACP program involving dedicated ACP facilitators) but also a wide range of different outcome measures (hospital transfers, intensive care unit use, cardiopulmonary resuscitation in care homes, carer satisfaction, and concordance of care).
In the included articles, we identified two main groups of outcome measures. One group is health care utilization measures. Often associated with economic analyses, these appear to function in these studies more as proxy measures of unwanted treatment or overtreatment. This group includes health care costs, place of death, hospital admissions, length of hospital stay, use of intensive care, ambulance call-outs, and specific treatments. The other group is patient and carer measures (covering a range of physical comfort and psychosocial measures). For the person with dementia, these include concordance of care, physical and emotional distress, anxiety, depression, quality of life, decisional conflict and coping, and for carers, satisfaction with the end-of-life care received by their relative or friend, quality of life, anxiety, depression, and self-care. Most studies, particularly those of moderate to high quality, focused on measuring health care utilization (e.g., hospitalization, length of stay, and hospital costs).
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
There was less high-quality evidence about wider patient/carer outcomes (e.g., carer satisfaction, quality of life, mildness of death, comfort in dying, and symptom burden).
Furthermore, it was unclear from the included studies whether reduced health care utilization was achieved while at the same time improving, or at least not compromising, patient and carer outcomes. Only five of the studies measured health care utilization and patient/carer outcomes together in the same study.
Two others found evidence of reduced health utilization, but no improvement (but also no reduction) in “comfort in dying” or satisfaction with care, respectively.
Where ACP is associated with reductions in health care utilization, questions also arise about what substitute care, if any, is needed and whether this is accessed in practice. The availability, appropriateness, and quality of this care are likely to influence end-of-life outcomes. Improved access to specialist palliative care (such as admission to hospice in the U.S.) is sometimes seen as a potential concomitant of ACP and has been included as an outcome measure in some evaluations (e.g., Nicholas et al.,).
However, none of the studies with health care utilization outcomes used this or any similar measure and substitute care was little discussed. Exceptionally, Molloy et al.
note, in their discussion, that care homes are often not equipped to provide adequate symptom relief or palliative care to residents at the end of life and they point to the need for consequent investment.
Causal Pathways and “Active Elements”
The causal pathways underpinning associations between ACP and end-of-life outcomes were generally not well articulated or explored in the included studies, and in no study were they actively tested, either quantitatively (using process measures) or in accompanying qualitative research. By looking across the studies, however, multiple possible mechanisms may be gleaned. Advance directives may directly guide care, being “activated” when someone without capacity requires relevant medical decision-making (e.g., Nicholas et al.).
The ACP interventions in care homes, however, were multicomponent programs involving (often system-wide) staff education and training, as well as sometimes education for patients and carers, and other elements such as flagging ACP on charts and feeding back to physicians on whether the care they have provided is congruent with people's stated preferences.
These interventions appear to have influenced end-of-life outcomes as much by influencing the policy and culture of the care home, as well as staff skills and knowledge, as through individual-level ACP conversations or completion of ACP documents. For example, Molloy et al.
found significant differences in end-of-life outcomes between intervention and control care homes. However, when looking only at intervention care homes, no significant differences were found between residents with and without advance directives. The authors hypothesized that the education and training for health and care staff increased acknowledgment of resident and carer preferences and, as a consequence, may have led to a generally higher threshold for transfer to hospital independent of the completion of individual advance directives (p.1442). Similarly, Caplan et al. (2006) found differences between intervention and control care homes, despite the fact that, in practice, the intervention had resulted in few written advance directives.
Some authors of articles reporting on nonintervention studies also suggested that the completion of advance directives may be associated with a broader culture of good practice and shared decision-making that is, in turn, associated with better end-of-life outcomes.
for example, hypothesize that “time spent discussing specific directives may be a marker for overall better communication and shared decision-making between [family carers] and health care professionals” (p.5). In care homes, advance directives were also often formalized by medical orders. The results from three studies in which both ACP and medical orders were studied are consistent with this being an important “active element.”
In care homes, ACP was also often carried out with carers within a “best interests” framework (as well as with people with dementia directly wherever possible). This may imply different causal pathways, although these were not explored in any of the studies. Unusually, Hilgeman et al.,
Preserving Identity and Planning for Advance Care (PIPAC): preliminary outcomes from a patient-centered intervention for individuals with mild dementia.
in their study, focused on the potential therapeutic effects of ACP.
Implementation Issues
Descriptions of the interventions varied in the level of detail provided. However, all the interventions involved multiple components. Consequently, even where interventions were well described, restrictions on word count in academic articles meant that there was inevitably limited detail provided about any of the individual elements. Only two of the studies had accompanying articles; for one study, there were two articles reporting on earlier small-scale pilot studies and, for the other study, one qualitative article exploring staff experiences prior to the implementation of the intervention. These did not provide much more detail about the intervention itself or its implementation. This meant that there was no, or very limited, discussion of implementation challenges. We do not know, for example, if or how implementation may have varied in practice from that initially envisaged and, if this occurred, how this was responded to. This is potentially important since, not only did the design of ACP interventions vary considerably, the acceptability and effectiveness of the individual-level ACP component appeared to sometimes be in question. In their discussion, Molloy et al.,
for example, noted that the length and complexity of the ACP document used in their intervention may have deterred some from wanting to complete it. It was not within the scope of their study, however, to explore this further.
In the nonintervention studies, commonly conducted using administrative or large-scale survey data, the authors sometimes noted the “black box effect” (i.e., having no information about underlying processes) as an explicit limitation.
They also sometimes speculated about implementation, particularly where there were unexpected results. For example, in their study of 69 care homes in Belgium, Vandervoort et al. found that nurse-led end-of-life conversations with residents were not significantly associated with improved end-of-life outcomes (unlike advance directives).
The authors suggest that “one possible explanation is that the conversations between residents and nurses were limited in frequency, not repeated over time or not intense or thorough enough” (p.5). They also found that nurse-led conversations with family carers were actually associated with poorer, rather than better, end-of-life outcomes and hypothesized that it could be that “these conversations take place reactively or ad hoc when residents are experiencing clinical complications at the end of life” (p.5).
Conclusion
While implementing ACP may be more challenging in dementia, it has also been identified as especially important, both because it allows people with dementia to have a say in their future care and because of the currently poor outcomes for people with dementia at end-of-life, which ACP may help to improve. Our review suggests that, although the evidence base is currently limited, ACP is likely to be relevant and applicable to people with dementia and that ACP may, in some circumstances, be associated with a range of positive end-of-life outcomes. The evidence, therefore, provides some support for the current emphasis on ACP in policy. However, the research evidence needs further development if policy and practice on ACP in dementia are to evolve effectively. There are a number of priorities for future research that can be identified.
One priority area is that of outcome selection. Currently, there is more evidence of an association between ACP and reduced health care utilization than there is between ACP and improvements in wider patient/carer outcomes. Health care utilization measures, in this context, are more than economic measures; they may function as proxies for reductions in unwanted treatment and for increased concordance between people's preferences for care and the care they receive in practice. Nonetheless, it is important that future studies include a spread of different types of measure to provide reassurance that reduced health care utilization is achieved alongside improvements in wider patient/carer outcomes or is at least not achieved at their expense.
Studies that measure reductions in health care utilization should also take account of the availability, quality, and appropriateness of substitute care. This is of particular importance in dementia, since we know that people with dementia are less likely to access specialist palliative care services
By including all important outcomes, studies will additionally be better placed to consider cost-effectiveness. Better information about the resources required to deliver ACP interventions is also needed.
A second priority concerns when and where ACP research is conducted. Currently, most of the evidence on ACP in dementia is from care homes. While more good quality research in this setting is needed, there is a particular need for more research in other settings and earlier on in the disease trajectory. While most people who die with dementia, do so in care homes, the numbers who die elsewhere should not be underestimated. Nicholas et al.
found (using survey-based cognition measures) that more than half of community-dwelling older adults in the last year of their lives were cognitively impaired to some degree (41.6% had cognitive impairments or mild dementia and 11.7% had severe dementia). They also found that people with cognitive impairments/dementia in the community tended to receive much more aggressive care at end of life than similar people in care homes. In this context, their finding, that people with severe dementia in the community who have an advance directive experience better end-of-life outcomes than similar people without an advance directive, suggests the special importance of ACP in community settings and of having early conversations.
A third priority concerns the use of more rigorous evaluation approaches. While randomization in this area is ethically and practically challenging, cluster randomized designs and blinding of outcome assessment may well be feasible. Well-designed and rigorously conducted observational studies are also likely to be valuable. The effectiveness and value of outcome research is likely to be limited, however, without accompanying research to better understand underlying processes.
ACP interventions vary and implementation can be challenging. Therefore, rich description of interventions, of challenges met and of how these were addressed, can help with the effective transfer of approaches shown to be successful. The “active elements” of ACP interventions, and associated causal pathways, should also be identified (e.g., in accompanying process evaluations). Realist approaches to evaluation
are likely to be of value here. In this way, more theoretically driven, evaluative research can be produced that fully takes into account the complexity of ACP for this population.
Disclosures and Acknowledgments
The support of the Economic and Social Research Council (ESRC) and National Institute for Health Research (NIHR) is gratefully acknowledged (ES/L001896/1). Following award of funding, the Funders had no role in the study design, in the collection, analysis and interpretation of data, in the writing of the report or in the decision to submit the article for publication. The authors would also like to thank the MODEM project team, particularly David McDaid and Adelina Comas-Herrera, and Andra Fry, the LSE librarian who assisted the database searches.
The authors have no conflict of interest to declare. Disclosure forms from the International Committee of Medical Journal Editors have been completed by all authors.
Ethical approval: All findings and data presented in this paper are already in the public domain. No new data were collected for this study. No ethical approval was required.
on behalf of the European Association for Palliative Care (EAPC)
White paper defining optimal palliative care in older people with dementia: a Delphi study and recommendations from the European Association for Palliative Care.
30333-0&id=gr1.jpg){kind=link}
) Total score and physical distress subscale, dying symptoms, and well-being subscales of CAD-EOLD.
30333-0&id=fx1.jpg){kind=link}