Enhancing shared and surrogate decision making for people living with dementia: A systematic review of the effectiveness of interventions

Abstract Background Dementia can have a profound impact on decision making. People living with dementia (PLwD) often need to make decisions about health care, and, as dementia progresses, decisions may need to be made on their behalf. Specific interventions may support this process. Review Question What interventions are effective in improving shared decision making or surrogate decision making on the health care of PLwD? Methods A narrative systematic review of existing literature was conducted. Seven databases, grey literature and key journals were searched. After exclusion by title, abstracts then full texts were reviewed collaboratively to manage any disagreements. Results Eight studies met the inclusion criteria. Two articles, including one RCT, evaluated decision aids regarding the use of enteral feeding in advanced dementia. Six further articles, including five RCTs, were found which evaluated the effectiveness of interventions supporting patients or carers with advance care planning. Conclusion Decision‐making interventions typically consist of multiple components which aim to establish preferences for future health care. Advance care planning interventions supported aspects of the decision‐making processes but their impact on decision quality was rarely evaluated. Interventions did not increase the concordance of decisions with a person's values. The decision‐specific interventions are unlikely to produce benefit in other decision contexts. Patient Involvement Two caregivers, a public stakeholder group and a carer group were consulted in the design of the wider study to which this review relates. Six PLwD refined the research questions addressed in this paper.


| INTRODUC TI ON
Dementia is a term encompassing multiple neurodegenerative conditions, including Alzheimer's disease and vascular dementia. 1 Dementia is highly prevalent, with estimates predicting the condition will affect Examples of more complex decisions include those related to end-oflife care, 3 moving into residential care 4 or dental care. 5 Shared decision making (SDM) is a model for clinical practice which incorporates patients' values and preferences and supports them to make decisions about their own care in collaboration with clinicians. [6][7][8] This approach has been promoted in policy 9-11 as it emphasizes a patient's autonomy 6,7 through information sharing in both directions between patient and clinician. 12,13 There is a broad spectrum of perspectives on what constitutes SDM for PLwD. 14 As dementia progresses, communication and decision making may require more support from family members or carers yet studies have identified a dissonance between person's preferences and the perceptions of their caregivers. 15 There is a need for their knowledge of a person to be considered holistically alongside other factors relevant to decision making.
Where dementia progresses a specific decision may arise where decisional capacity is determined to be lacking. In this situation, decisions are made by a surrogate or substitute decision maker, typically a family member. The point at which decisional capacity is lost; the person who makes a decision; and the bioethical approach used to inform a decision varies based on international legislation. 16 Substituted Judgement is commonly used to refer to the trying to take the course of action a person would have chosen themselves. 17 This approach considers a person's values and preferences but fails to identify how they might feel in light of any new or worsening cognitive impairment. 18 The best interests standard is an alternative approach and is the cornerstone of the Mental Capacity Act (MCA) for England and Wales. 19 When making a decision in somebody's best interests, a holistic view of a person's welfare alongside their expressed values and preferences should be used to inform any decisions made for their care. 20 The Mental Capacity Act does not define best interests explicitly 21 though much has been written about both its scope and its perceived ethical superiority over substituted judgement. 18,20 Legal roles specified in the MCA, such as Lasting Power of Attorney or a Deputyship, allow named individuals to act as a decision makers. 21 In the absence of formal roles, a family member may support clinical decision making by providing insight into the PLwD's values and preferences. 19 The potential for the views of the PLwD, family and clinicians to differ can complicate best interests decision-making process. PLwD can exercise their autonomy by making their wishes known in an Advance Care Plan (ACP) before any loss of decisional capacity. An existing review, however, has found that ACPs are only limited in their effectiveness for PLwD. 22 Decision aids (DAs) are one approach commonly described in the literature yet other approaches can also support SDM. 23 DAs are tools that support patients or surrogate decision makers by providing information and supporting them to make the decision that is right for them . 24 DAs have been reviewed systematically both in general 24 and in relation to PLwD. 25,26 However, the effectiveness of DAs or other interventions are not comprehensively understood in supporting shared or surrogate decision making with or for PLwD. It could be suggested specific approaches to optimize communication, modify expectations or support value elicitation could all potentially enhance decision making for PLwD separately or alongside DAs. This review aims to explore the effectiveness of interventions that aim to support shared decision making with or for PLwD facing health-care decisions. Doing so should help inform the development of future DAs or other interventions to support this patient group and those involved in their care.

| Review question
What interventions are effective in improving shared decision making or surrogate decision making in relation to the health care of PLwD?

| Design
A systematic review was undertaken to identify studies that assessed the effectiveness of interventions relevant to health-care decision making with or for PLwD. Working with PLwD and carers Patient Involvement: Two caregivers, a public stakeholder group and a carer group were consulted in the design of the wider study to which this review relates. Six PLwD refined the research questions addressed in this paper.

K E Y W O R D S
cognitive impairment, decision making, dementia, proxy decision making, shared decision making helped to focus the research question and the need to know 'what actually works in the real world'. It was anticipated that studies of effectiveness would typically be quantitative yet qualitative studies were not explicitly excluded as these may shed light on the processes of implementation and the factors contributing to effectiveness. The review aligns with guidance from PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analysis. 27 The protocol was prospectively registered on PROSPERO (CRD42019154707), accessible at: https://www.crd.york.ac.uk/PROSP ERO/displ ay_record. php?Recor dID=154707.

| Inclusion and exclusion criteria
The inclusion and exclusion criteria are shown in Table 1. Systematic reviews were excluded but were read to identify other relevant studies. There were no limitations on date or geographic region.

| Outcomes of interest
Fixed outcomes of interest were not specified due to the wide range of measures used to evaluate shared or surrogate decision making and the lack of standardization in assessment. 29 The process by which a decision is made and the quality of a decision itself can be evaluated by multiple measures. 29,30 These include assessments of satisfaction, value concordance and patient involvement such as by the OPTION scale. 31,32 It was anticipated that the effectiveness of interventions would be evaluated using any of these measures, additional measures of communication or a variety of combinations, thus preventing formally specified fixed outcomes of interest.

| Search strategy
Seven databases were searched with a strategy using MeSH headings and keywords with support of a subject-specific librarian (LE).
The strategy was informed by a P-I-C-O design in which Population:

Justification for criteria
Inclusion Criteria 1) The paper assesses effectiveness of an intervention aiming to support shared or surrogate decision making for PLwD The review is aiming to determine the effectiveness of interventions, not simply to describe a list of untested interventions.
2) The paper examines decisions about health-care interventions For non-health-care decisions, such as those relating to finance or social care, a range of different factors could be identified that do not translate to healthcare contexts.
3) The decisions studied are actual decisions made as opposed to hypothetical decisions The effectiveness of an intervention cannot be definitely established by testing it against hypothetical or simulated decisions. Studies (from excluded list) that identify additional interventions discussed in narrative review: n = 13

Records excluded after review of full text (n = 32)
Primary reason for exclusion: † Inadequate detail of methods n = 3 Evaluation using hypothetical decisions n = 6 Intervention not evaluated for effectiveness n = 11 Intervention for capacity assessment: n = 1 Non-healthcare decisions: n = 4 Not an intervention: n = 1 Not specific to people with dementia: n = 5 Protocol of study not yet completed: n = 1

| Study selection
After removal of duplicates, study titles were screened against inclusion and exclusion criteria by one author (AGR). Studies were included for review at the subsequent stage when there was any uncertainty about inclusion. For the remaining studies, two authors (AGR and LE) reviewed all abstracts using Rayyan. 33 After screening the first 100 abstracts, the interpretation of inclusion and exclusion criteria was reviewed to establish agreement in study inclusion. Any uncertainty or disagreement about inclusion was discussed in-depth and, where necessary, resolved by discussion with a third reviewer (RT). Full texts were reviewed by AGR, and any uncertainty about inclusion was discussed with RT. Pilot or feasibility analyses identified were read to assess whether or not the study aimed to assess effectiveness formally. Figure 1 shows the PRISMA flow diagram.

| Data extraction and quality assessment
The outcomes specified in the protocol were extracted onto a specific data-extraction spreadsheet which was used to populate

| Analysis and synthesis processes
The heterogeneity of studies and measures used meant a metaanalysis was not possible. A narrative review was employed to broadly reflect a thematic summary approach 35 whereby studies are categorized into groups and discussed based on their key features. This approach allowed a detailed narrative exploration and a comparison of studies presenting interventions with similar aims. 36

| Study characteristics
Eight studies met the criteria for inclusion (Table 2). Two studies assessed interventions related to enteral feeding in advanced dementia, a controlled before and after study 37 and a randomized trial. 38 Six studies evaluated interventions related to ACP with either caregivers alone or PLwD-caregivers dyads; these included one controlled before and after study 39 and five randomized trials. [40][41][42][43][44] All were undertaken in either the USA, the UK or Belgium. DAs featured in both studies regarding enteral feeding decisions 37,38 but only in one study related to ACP. 40 The remaining studies detailed interventions aiming to support multiple decisions as opposed to tackling one specific issue.
The sharp divide between the type of decision being made in the included studies warranted a separate discussion for each category. This differs from the protocol which intended to separate studies by decisions made by or for the PLwD. The heavy emphasis on surrogate decision making and future care planning made it necessary to redefine how the studies would be separately examined.

| Overview of DAs for feeding decisions
Decisions aids for enteral feeding decisions were targeted at surrogate decision makers in advanced dementia. Both studies 37,38 used a modified version of the DA initially described by Mitchell et al 45 The studies were led by the same researcher with the later study 37 describing itself as undertaking 'baseline interviews' as part of a larger trial, citing the RCT. 38 Though one study precedes the other, it was published at a later time and both provide separate insights into the effectiveness of the intervention in different contexts.

| Quality assessment of studies of DAs for enteral feeding decisions
Reviewed against the MMAT, the RCT 38 has the potential for bias as the method of randomization is unclear and incomplete outcome data are presented. The use of cluster randomization was necessary, and a control was employed. Assessors were not blinded potentially leading to bias. The before and after study 37 had complete follow-up, though the methods were not comprehensively described in either quantitative or qualitative components. This limits the understanding of the specific approaches used. Neither study provided a power calculation yet sample sizes were similar. The distribution of participants across intervention and control groups was also comparable.  In the intervention group dyadic functioning improved in only one dimension: decreased emotional disruption. There was no increase in care-related agreement, in positive interactions or in negative interactions.

TA B L E 2 (Continued)
However, this concept arose from the interviews undertaken and the authors note this is not a demonstration of effectiveness. Measured quantitatively, surrogates were significantly less likely to expect artificial feeding to provide benefits to the PLwD. Table 3 shows the impact of the interventions on the outcome measures common to both studies. Decisional conflict was reduced and knowledge increased; though both changes were statistically significant the real-world extent of changes was minimal. Though the studies demonstrate effectiveness, the magnitude and impact of this effectiveness was limited.

| Overview of ACP interventions
Only one study focused specifically on participants living with dementia at an early stage, where they were paired with family members in dyads. 44 Other studies included people living with later stages of dementia, also as dyads, 40,41 and one study included only caregivers as participants for those with late-stage dementia. 42 The remaining two studies 39,43 detail approaches that could be employed with PLwD, caregivers or both as the course of dementia progressed.
A range of evaluation measures were used, including decisional conflict, 42  tioning. 44 The variation in outcome measures reflects the variation in study design, the primary outcome measure chosen and the specific features of each intervention.

| Quality assessment of studies of ACP interventions
The assessment of the five RCTs against the MMAT is shown in

| Summaries of studies of ACP interventions
Whitlatch et al, 44

| Interventions Identified that failed to fulfil inclusion criteria
Within the 32 papers excluded at full-text stage, a further 10 interventions related to decision making for PLwD were described across 13 publications. These are summarized in Table 5 Other studies have examined separate approaches to support surrogate decision makers for PLwD. [49][50][51]58 One example is 'Fact Boxes' 49 which present factual information to support decision making for specific decisions. These were evaluated using hypothetical scenarios meaning their effectiveness in in real decision-making contexts cannot be established. 49 Paralleling the included study from Hanson et al 41 the study from Sampson et al, 50 aimed to assess an intervention designed to support carers to produce ACP following unplanned hospital admissions for PLwD. In this feasibility study, decisional conflict decreased in the intervention group though the extent and statistical significance of this is not clear. This study suggests that specific palliative care assessments and support did not increase the frequency of ACP production. Similar approaches are employed by Reinhardt et al, 58 whose intervention contained structured conversations with palliative care teams. This randomized trial considered satisfaction with care as its primary outcome without emphasis on any decision-related outcomes.

| Components of interventions
Within the interventions in the included studies (Table 2) and those excluded ( Table 5) the most common components appear to be

Study Intervention Name Description of Intervention
Bonner et al (2014) 56 Advance Care Treatment Plan (ACT-Plan) for African American Family Caregivers A group-based education intervention, with dementia caregivers where a number of end-of-life discussions were had to inform future decisions from the patient's perspective.
Dassel et al (2019) 55 The LEAD Guide (Life-Planning in Early Alzheimer's and Dementia) A dementia-focused instrument that can be used by those with early dementia, family members and clinicians to document the persons with dementia's preferences and values to inform current or future care.
Górska et al (2016) 54 Family Group Conferencing A five-stage process in which service users, family members, and health/social care professionals come together into a family-led decision-making forum.
Loizeau et al (2019) 49 Fact Boxes Fact Boxes: short paper-based tools to support a variety of decision makers, presenting the benefits and harms of treatment approaches using simple language. The examples studied are artificial hydration and antibiotic therapy in pneumonia in PLwD.
Murphy and Oliver (2012), 53 Reitz   Two separate systematic reviews have specifically examined decision aids related to decisions with or for PLwD. 25

| Advance Care Planning and the 'here and now'
In evaluating ACP interventions longer-term follow-up becomes more methodologically challenging. ACPs relate to anticipated events yet the anticipated scenario expressed in the ACP may differ to the reality being faced by a substitute or surrogate decision maker. Here, a decision maker may need to deviate somewhat from the wishes expressed in the ACP. 22 The anticipated but variable progression of dementia over time may explain why the intervention detailed by Hanson et al,40 only led to greater concordance between decision makers and healthcare providers at a nine-month follow-up and why no change in carerelated agreement arose in the study presented by Whitlatch et al 44 Family decision makers at an earlier stage may have more optimistic expectations of their relative's cognitive decline than medical professionals. With time, the expectations of the family member may converge with those of professionals hence why alignment in perspectives occurs only when significant time has passed.

| Decision Making by Clinicians and the Absence of Family Support
The focus on family decision makers detracts from the role of clinicians who may need to make decisions on behalf of patients.
Specific interventions could support clinicians to obtain informa- preferences can be gained. In these situations, a time-specific evaluation of the person's best interests is made, and independent assessment and support can prove valuable. 62 No interventions were identified to support decision making where the views and preferences of a PLwD cannot be established.

| The role of study context
Reflecting on the studies included it becomes clear that many are highly specific to the context in which they were devised and tested.
This may support implementation but can also limit the transferability of interventions to new settings. If any of the interventions were to be modified and implemented into a new context, their effectiveness would need to be demonstrated in this new setting.
Ideally, effectiveness would be determined by measuring of decision process and decision quality related to the PLwD, surrogate or both separately based on the PLwD's decisional capacity. If new interventions are designed for specific decisions, settings or individuals, the involvement of PLwD, carers and health-care professionals will be crucial in ensuring any interventions produced are acceptable to all those they aim to support. 63

| Strengths and limitations
The search criteria were intentionally broad but specifically limited to dementia. A broader search using terms such as 'older adults' may have identified additional studies yet produced an unmanageable number of results. Full-text reviews were primarily undertaken by one author only; however, this was mitigated by in-depth discussions about suitability of inclusion with an additional author. The search found six studies focusing on ACP; though these are valuable, ACP was not the primary intended focus of the review. No studies demonstrated effectiveness in supporting decisions beyond tube feeding that may arise in the here and now for PLwD.

| CON CLUS ION
Certain interventions can improve various aspects of the decisionmaking process with, or for, PLwD. Establishing goals of care formally can improve concordance of goals between family members and clinicians over time, hence improving the quality of a decision made. The interventions focus primarily on a substitute decision makers as opposed to a PLwD and no approaches were effective in improving decisions made by PLwD themselves or making decisions align with their values. There is a lack of evidence demonstrating that preference elicitation at earlier stages of dementia positively impacts the quality of any real-world decisions made at a later stage either with or for a person living with dementia.
Interventions need to be designed and evaluated that support the PLwD in contributing to decisions throughout the progression of dementia.

AUTH O R CO NTR I B UTI O N S
AGR led the project, drafted the protocol and revised it following input from the remaining authors, acted as primary reviewer for all articles, and drafted the manuscript. LE designed the search strategy and acted as second reviewer screening papers by title and abstract.
RT supported the design of the review and methods, and supported decisions of study inclusion where AGR and LE were uncertain. CE, RRW and CA supported the protocol development and provided oversight on the wider study. All authors commented on drafts of the manuscript, supported the response to peer reviewer's comments and approved the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analysed in this study.