The search strategy was devised using a PICOS grid, terms were selected according to our inclusion and exclusion criteria and using variations found in major known key texts. These terms were discussed and agreed among members of the research team (LS, TG, AG, MS) Due to the small evidence base on this topic and the need for a broad search only variations for terms for the “Problem” (bipolar) and “Intervention” (advance decision making) were entered into the PICOS grid. All comparators, outcomes and study designs were eligible for inclusion.

The basic search string was then tailored according to the capabilities of each database. Where available sub-headings were used and, when relevant terms were listed, options to explode subheadings were employed. Search limits were also tailored with a focus on maintaining a wide search. All years, publication types (e.g. books and papers, published and unpublished) were allowed. The searches were cross checked for re-producibility among the research team (LS and TG for initial search and LS, AG, and MS for updated search).

The texts resulting from the electronic search were compiled into a central database. Duplicates were removed. After research team members discussed and confirmed understanding of the inclusion/exclusion criteria, the titles, then abstracts, then full texts of the documents were independently searched for relevance by two team members (TG and LS) with discussions held at each stage to determine which articles should be included. Any disagreements were discussed (by TG, LS, GO) until consensus was reached.

The final list of documents for inclusions was sent to international experts in the field to request suggestions for other relevant documents that were not located in the search. The references of the included documents were also followed to ensure full coverage.

Contact with study authors of RCTs which trialled the use of ADM with people who have SMI was made to ascertain whether sub analyses of results for people with bipolar were available.

The following electronic search strategy was used for PubMed: ((((advance directive OR advance care planning[MeSH Terms])) OR (“advance decision”[Title/Abstract] OR “self-binding”[Title/Abstract] OR “psychiatric advance directive”[Title/Abstract] OR “advance agreement”[Title/Abstract] OR “advance statement”[Title/Abstract] OR “mill’s will”[Title/Abstract] OR “voluntary commitment contract”[Title/Abstract] OR “nexum contract”[Title/Abstract] OR “crisis plan”[Title/Abstract])) AND (mental disorder OR Bipolar and related disorders OR Mood disorder OR psychotic disorders OR depression OR mental health OR mentally ill persons[MeSH Terms])) NOT (dementia OR terminal care OR palliative care[MeSH Terms]).

Databases were searched across all available dates: SCOPUS (1966–2019), CINAHL (1937–2019), Cochrane (1996–2019), EMBASE (1974–2019), Medline (1946-2019), PsychINFO (1806–2019) and PubMed (1865–2019). Where it was possible to specify, published and unpublished papers were included in the search filters. Initial searches were conducted in 2017 (LS, TG) and updated in 2019 (LS, AG and MS).

Standardised data extraction pro formas were designed by team members (TG, GO, LS). The data was extracted by one researcher (LS) and reviewed by two senior researchers (TG and GO). Like for like comparison of quality between articles was challenging due to the heterogeneous nature of the included papers. The Mixed Methods Appraisal Tool (MMAT) was used to categorise the body of literature identified, further clarifying the current state of the available evidence base (10).

Neither quantitative or qualitative meta-analysis was appropriate for the body of literature identified during the searching process. Instead, data was tabulated, arranged and synthesised in a narrative fashion (11).

A total of 3,067 articles were identified by electronic database searches, citation searches and consultation with experts in the field. After the removal of duplicates, and title and abstract screen, a total of 37 articles remained. Following full text screening a total of 13 articles detailing 11 studies were included in the final review. Figure 1 shows the flow diagram.

The range of study designs included 2 mixed methods studies, 7 quantitative descriptive studies and 1 randomised controlled trial. The results are summarised in Table 1 . The focus of the studies can be grouped according to the following outcome domains for study participants with bipolar: interest in ADM, preferred type of ADM, barriers to engaging in ADM and impact of using ADM. These outcome domains were determined after a series of discussions among 3 members of the research team (LS, TG, GO) during which consensus was reached.

This first systematic review of the literature on ADM for bipolar aimed to: establish the state of the current evidence base, clarify key features of ADM which are relevant to people with bipolar, identify gaps in the evidence base, and refine future research directions. We found a small evidence base which is at an early stage of developments and comprises largely heterogeneous observational study designs exploring a range of ADM types. Given the current state of the evidence it is not possible to draw any firm conclusions about the effectiveness of ADM for people with bipolar in achieving outcomes stakeholders desire. However, the available research appears to most strongly confirm a high level of interest in ADM among people with bipolar. ADM features that seem to be important to this group include support for collaboratively made, instructional directives which facilitate treatment requests as well as refusals. There is some inconsistency between studies about whether or not people with bipolar are likely to support “self-binding”, which merits further exploration. Several studies identify multiple and significant barriers to ADM for bipolar and there is a far greater focus on factors which are barriers, compared to those that facilitate ADM. However, it may be that people with bipolar face comparatively fewer barriers. ADM appears to hold promise for people with bipolar in reducing coercion and improving the experience of using mental health services in crises. Again, this is not a guaranteed outcome, but it seems that those with bipolar may have better chances of making and successfully using ADM documents than those living with other SMI for reasons that are, as yet, unclear. Gaps in the current evidence base include: understanding outcomes that service users with bipolar wish to achieve using ADM, understanding causal mechanisms and features of ADM that allow these outcomes to be achieved, clarifying service user opinion on self-binding ADM and developing a model of ADM and materials tailored for people with bipolar. Developing models of collaborative ADM could be helpfully informed by existing research on shared decision making (SDM) in mental health.

Limitations of this review include the inherent difficulties of evaluating and synthesising results from a range of study types many of which are pilot-like. It is of note that none of the RCT studies and meta-analyses which explore outcomes for ADM in SMI (25–31) provide sub analyses according to individual diagnoses. Instead, a broad division is frequently made between psychotic and affective disorders. This is of concern given the differences in psychopathology between various SMI and the varying patterns of assistance individuals with bipolar are likely to require from health services (32). In addition, the included studies do not specify whether participants had a diagnosis of type 1 or type 2 bipolar. It is likely that a range of severity is represented in this review as the included two service user surveys relied on self-report diagnosis by service users with any level of mental health input (13, 14, 17). Whereas the other studies recruited service users with a diagnosis of bipolar who were actively in touch with specialist psychiatric services (requiring outpatient visits/emergency contacts/hospital admissions) (12, 16, 18, 19, 21–24).

Therefore, given the studies that were eligible for inclusion only a low level of confidence in the results is possible. However, a strength of this review is the comprehensive search and wide inclusion criteria which aimed to establish the state of the evidence base.

The results of this review focussing on ADM in bipolar are largely consistent with the literature on ADM in general SMI. This literature also suggests that ADM is supported by service users (33) and that service users tend to produce clinically feasible documents (34, 35). Despite this enthusiasm, there are often low levels of uptake (33).

A key difference is that in the wider literature on ADM in SMI there have been larger scale RCT studies of the outcomes of using ADM. The purpose and primary outcome of ADM in these RCT studies is generally considered to be avoiding coercion, particularly in the form of compulsory inpatient admissions. Studies on this topic have had conflicting results, although meta-analyses suggest that overall ADM reduces coercion for people with SMI. Given that people with bipolar may face comparatively fewer barriers to completing ADM (24) and that their ADM preferences may be more likely to be followed (22) it could be hypothesised that this group is more likely to achieve success.

It remains unclear exactly what service users and professionals believe ADM are for i.e. which outcomes we might expect and want from their implementation. One outcome which has been empirically explored with people who have bipolar is the reduction in coercion (largely distilled into meaning reduction in the number of compulsory admissions) from mental health services. Related to this is the emergence of the subjective experience in the reduction of coercion or increase in self-determination as an outcome in itself.

Among the wider literature on ADM for SMI, RCT studies have used the reduction or avoidance of hospital admission as the primary outcome (25–28, 31). Avoiding admission and coercion has been understood to be crucial in increasing self-determination and therefore affording people with SMI equal rights. Redressing inequality in this area is a key concept of the UN Convention on the Rights of Person with Disabilities (36).

However, this perspective that a reduction in coercion leads to increased self-determination is not without challenge, particularly in the bipolar population. In their survey, Hindley et al. found a third of respondents with bipolar would use an ADM document to request admission. In another survey of over 500 service users with experience of mania and carers, most respondents (the majority of whom were service users) supported the use of coercion when unwell (37). Gergel and Owen (4) argue that earlier admission, even if facilitated through coercion, will ultimately be more effective in promoting self-determination. The person is thus empowered to harness the strength of the available legal provision and wield it as they see appropriate in order to gain more control of their illness and the consequences that may have occurred during episodes of mania/depression in the past.

Therefore, it seems prudent to challenge the assumption that ADM may be automatically judged as operating successfully if admissions are reduced through their use. If an ADM document expresses the choices of an individual it is likely that different people and groups will wish to achieve or remedy different things. Future research should focus on the outcome of harm avoidance in a broader sense, and as defined by the service user. For some, who experience compulsory admission as the main harm resulting from a crisis, this may mean early intervention to prevent admission, for others this may mean early admission to avoid other harms e.g. social or financial.

Implementation of ADM has jumped ahead of the research on outcomes. One reason for this is the ethical imperative to maximise an individual’s autonomy, particularly when autonomy is compromised. For those with bipolar, autonomy may be threatened on multiple levels: firstly, they may experience periods of illness as inauthentic i.e. not arising from their true sense of self; secondly, during periods of illness an individual may experience a loss of self-control, engaging in behaviours they would never normally consider; thirdly, attempts at illness management may elicit coercive measures from other parties such as family, friends or health professionals. Given the purported centrality of autonomy as a stated driver for implementing ADM (38) there is a strong argument for exploring whether using ADM increases service users’ sense of control over their illness and treatment as a primary outcome.

The Medical Research Council (MRC) defines complex interventions as those that: have “a number of interactions between components within the experimental and control conditions”, require “a number and difficulty of behaviours required by those delivering or receiving the intervention”, target “a number of groups or organisational levels”, have “a number and variability of outcomes” and require “flexibility or tailoring of the intervention” (39). This is undoubtedly the case for ADM in bipolar. Particular dimensions of complexity include: Creating an ADM document is a highly individualised process (40), successfully implementing ADM is likely to depend on an organised response from several services e.g. mental health teams, Accident and Emergency departments, specialist social workers, inpatient wards and the desired outcomes are (as yet) unclear and likely to include subjective measures.

This review has confirmed that more research on ADM as a complex intervention is required. MRC guidelines specify that developing and evaluating a complex intervention usually involves developing a conceptual model, followed by feasibility testing, evaluation and then implementation. Evaluation commonly takes the form of randomised control trials to determine the effectiveness of an intervention and therefore justify larger scale implementation. However, in multiple international jurisdictions (soon to include England and Wales following their government’s acceptance of the recommendation of the Independent Review of the MHA) ADM has already been introduced largely on the basis of stakeholder consensus and strong ethical and human rights grounds. Therefore, the important outstanding research questions for jurisdictions which already have statutory provision and those who are considering introducing it are less around examining effectiveness to build a case for introducing ADM and more around achieving high quality implementation and identifying potential harms. This will involve refining understanding of enabling conditions (including training and materials), key features of process and desired outcomes. For jurisdictions where legislation is in force further research will be essential to determine whether legal changes have been effective in achieving valued outcomes for stakeholders.

Future research directions could include the following: