nature and extent of service user involvement in critical care research and quality improvement; a scoping review of the literature

The importance of involving patients and the public in health care research is globally recognised, but how best to do this in critical care is unclear. The aim of this first published review was to explore the extent and nature of evidence on service user involvement in critical care research and quality improvement. Using the scoping review framework described by Arksey & O’Malley (2005) a team of service user and critical care researchers searched eleven online databases, reviewed relevant websites, conducted forward and backward citation searching and contacted subject experts. Extracted data were subjected to a narrative synthesis based on the objectives of the review. Findings from a broad range of evidence support that involvement is becoming more commonplace and that experiences are generally positive. Data extracted from 34 publications identify that involvement is most commonly reported at the level of consultation or participation in project teams, however, the extent to which involvement impacts on projects output remains unclear. Key barriers and facilitators relate to the challenge of recruiting a diverse group of service users, dealing with power hierarchies, being adaptable and effective consideration of the resource requirements. More research is required to identify the most effective methods to support the opportunity for involvement and more thorough reporting of service user involvement practices is strongly recommended. degrees extent service user processes; impacts; barriers to and ABSTRACT Structured summary 2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number.


Introduction
Patient and public involvement (PPI) in healthcare is strongly endorsed by the National Institute for Health Research (NIHR) in England (Denegri, 2015) and is a common pre-requisite for obtaining funding in the United Kingdom (UK), across Europe and Worldwide (Petit & Locock, 2013). There is a growing acceptance that health research and quality improvement (QI) work benefits from the personal insights and experiential knowledge of service users; that people have a right to be involved in work that may impact on their health or the services they receive, and that PPI helps improve the quality, relevance and impact of projects. (Beresford, 2005;Boote et al., 2011;Thompson et al., 2009). Implementation science literature also highlights the importance of engaging all stakeholders for successful knowledge transfer and service improvement (Ogden & Fixsen, 2014;Boaz et al., 2016). However, although PPI in both research and QI has expanded considerably in recent years, there remains an insufficient understanding of the most effective ways of doing it or the difference it makes to improving the health service (Nilsen et al., 2006;Petit & Locock, 2013;Staniszewska et al., 2008;Staniszewska et al., 2013).
Describing three levels of involvement (Table 1), the NIHR funded body that promotes public involvement in England (INVOLVE), defines PPI as projects carried out 'with' or 'by' members of the public rather than 'to ', 'about' or 'for' them (Denegri, 2015). Within critical care, PPI usually refers to the involvement of former patients and their family members, who are described as service users.  (Denegri, 2015) Previous reviews have focused on PPI related to patient safety initiatives (The Health Foundation, 2013), health policy and commissioning (Conklin et al., 2012;Crawford et al., 2002), and the research process itself (Boote et al., 2012;Boote et al., 2015;Brett et al., 2014;Staley, 2009;Viswanathan et al., 2009) in a range of population groups, but the involvement of people who have experienced critical illness has not been specifically explored. Having a critical illness and being in an intensive care unit (ICU) frequently leaves the individual and their family with ongoing physical and/or psychosocial sequelae (Griffiths et al., 2013;NICE, 2009). Involving former ICU patients and their family members in research or QI can be challenging due to the chronic health problems patients experience after discharge from hospital and because ICUs often admit out of area patients for speciality care. Some minority groups can also be difficult to access or engage (Dawson et al., 2015).
The aim of this review was to synthesise current evidence around best practice for PPI within critical care. The primary research question was: "How have former ICU patients and their families been involved in critical care research and/or QI projects?" As advised by Armstrong et al. (2011) a series of sub-questions were additionally identified to enable a clearly articulated scope of enquiry:

What levels of involvement have been described?
3. What is the impact of involving critical care service users?

What are the barriers and/or facilitators related to PPI in critical care?
The review questions emerged from local work with critical care service users and service providers (Bench et al., 2012;Goulding et al., 2015;Bench et al., 2016). These experiences identified a current lack of understanding around how best to involve former patients and their families and concerns regarding the impact of not doing it right.

Method
Scoping reviews involve the narrative synthesis of a broad and diverse range of evidence, commonly supplemented with consultative, consensus-building methodologies (Arksey & O'Malley, 2005;Davis et al., 2006). They can help identify gaps in the existing evidence, determine whether or not further reviews are justified and have the potential to influence policy and practice developments (Arksey & O'Malley, 2005;Armstrong et al., 2011;Davis et al., 2006). This review followed the six-stage framework described by Arksey & O'Malley (2005) and was further informed by the recommendations of Levac et al., (2010), Armstrong et al. (2011) and Daudt et al. (2013).

The research team
King's Improvement Science (

Study selection
The final selection of studies was guided by predetermined inclusion and exclusion criteria (Table 3).

Table 3: Inclusion and exclusion criteria
A single researcher screened all titles (n=4654), with 10% of those marked for exclusion reviewed by a second person.
All researchers reviewed the remaining 39 citations, 34 of which were included in the review. Disagreements were resolved using a consensus approach. A decision was taken to exclude protocols and focus instead on work that had already been conducted and written up in some form. Where a full text version of a study was available, additional published abstracts were also excluded. Figure I provides an overview of the search process.

Data charting, collation and analysis
Data were extracted, collated on an EXCEL spreadsheet and cross checked by all three researchers. Using a predetermined framework, a deductive approach focused on extracting and synthesising data with reference to the study objectives. In accordance with scoping review practice, no quality assessment of studies was performed (Arksey & O'Malley, 2005). Data were synthesised narratively to summarise current knowledge and practice around the nature and extent of PPI in critical care and its potential impacts.

Results
The 34 papers included in the review were all published in the period 2003-2016. There were 19 full text papers and 15 conference abstracts/project summaries, representing 29 QI or research projects (Table 4). Five papers reported work focused on a neonatal or paediatric, as opposed to an adult patient population. The majority of projects were conducted in the UK (n=18), however, seven papers reported work from the United States (US) and two reported data from a project conducted in Australia. Only two projects generated research data evaluating PPI approaches. Locock et al. (2014a) used ethnography to evaluate an Accelerated Experience Based Co-Design (AEBCD) project undertaken in two ICUs. Experience Based Co-Design is a six-stage process, which includes collating patient and staff experience data, followed by staff and service users working together to agree and develop QI priorities. The papers included in this review report an accelerated version of this process (AEBCD), using already available trigger films as an alternative to developing them during the project period. The ethnographic evaluation involved 155 hours of observations, 33 individual interviews and two group interviews with staff, patient and carer participants and facilitators. One hundred and sixty-six evaluation forms were also completed. Data were thematically analysed and a comparative framework developed based on key themes. Boaz et al. (2016) report data specific to the contribution of service users in identifying, designing and implementing quality improvement activities in the context of the AEBCD project and the discussion papers by Robert et al. (2015) and Tollyfield (2014) add further reflective information from the same AEBCD project regarding the PPI activities, their impact on project outcomes, and the barriers and facilitators associated with their use. Researchers initially observed a one-day workshop event. Data from the group discussions were audio recorded, interactions and processes observed and field notes kept. Data were synthesized using qualitative thematic analysis.
Two years later, focus groups and interviews were conducted during the destiny phase of the project and data were subjected to a process of inductive thematic analysis (Trajkovski et al. 2015b).
The remaining papers included in this review detail how service users were involved in research studies or QI initiatives, with some reflections on how involvement impacted on project outcomes and experiences. These papers do not include any evaluative/research data. Findings from all publications are summarised under four overarching themes: types, degrees and extent of service user involvement; processes; impacts; barriers to and facilitators of PPI.

Types, degrees and extent of service user involvement
Data about the number of service users involved in projects, their demographic details or the extent of their involvement were generally poorly reported, with no detail provided in some cases (for example, Gooding et al., 2012).
The number of service users involved ranged from one to 181. Thirty-three former ICU patients and family members were included in the AEBCD project (Locock et al. 2014a) whereas the AI project by Trajkovski et al. (2015a;2015b) only included eight family members in total. Most other projects included approximately six to eight people, most commonly as members of advisory groups.
A desire to involve people who were cognitively intact, had recent experiences and who lived locally was reported (Kingsbury, 2010;Menzies et al., 2011 Similarly, Locock et al. (2014a) report seeking to include a range of people, although details of those who took part was limited to a statement that all "were adults able to give informed consent…they or their relative had received care in one of the four participating services during the 6-month period of the fieldwork" (page 12).
From the available details, adult men and women aged 23-87 years including former patients with varying degrees and types of illness and their family members (partners, children, siblings, parents, aunts) were reported to have taken part in included projects. There was also evidence of the involvement of children and young people, aged 8-16 years (Menzies et al., 2011;Menzies et al., 2014) and of bereaved family members (Menzies et al., 2011). Demographic details were very limited but typically reported white British middle class involvement (Higginson et al., 2013;Wilson et al., 2012) with zero-five years most commonly reported as the length of time since the ICU experience (Locock et al., 2014a;Reay et al., 2014;Trajkovski et al., 2015a;Trajkovski et al., 2015b;Willis et al., 2011). A number of people had been previously involved in charity work (Hamil & Heslop, 2010), volunteer roles (Anderson & Finley, 2010) Table 1), and sometimes used generally to describe a project in which some aspects would be considered consultation and others collaboration (Denegri, 2015).
Most publications described consultation activities (Table 5), where commonly service users' experience data informed the design of an intervention or service improvement initiative. There were also examples of collaboration, with service users named as co-authors or co-presenters, acting as service user representatives on project steering groups or helping to design or deliver solutions and/or interventions. Several studies seem to have utilised both consultation and collaboration processes and strategies (Table 5).

Table 5: Examples of consultation and collaboration
Findings from the ethnographic evaluation by Locock et al. (2014a) identify a variety of roles undertaken by patients and carers in their AEBCD study: sharing experiences, identifying priorities for quality improvement, co-developing potential solutions and helping to implement and evaluate agreed interventions. Other authors also report patient involvement in data collection and analysis (Table 5), but involvement in the early stages of designing projects or reviewing study documentation was rarely described.

Processes
In some cases, there was an attempt to follow standard research procedures to recruit people for involvement. For example, Goulding et al. (2015) describe how they approached patients during their critical care follow up clinic appointment. Locock et al. (2014a) also report using a purposive sampling technique. Darbyshire (2015b) used targeted adverts, placed on user group websites and in follow up clinics, to invite service users to become part of a research team. In other cases, patients and family members were contacted directly and invited to take part in a survey or interview or to be part of a departmental project. Alternatively, service user representatives were accessed via existing groups or via contacts from previous projects. In contrast, Hamil & Heslop (2010) report that the relative of a former patient involved in their QI work directly approached the unit offering to volunteer within the department.

Duration of involvement
For some service users, involvement was limited to a one-day event whereas for others, involvement lasted for two years or more. Locock et al. (2014a) report that, in their study, a few (n=3) patients remained actively involved after the AEBCD process had finished. Many of the publications based on service improvement also refer to ongoing relationships with service user representatives, particularly where advisory groups or 'collaboratives' were set up or where volunteer roles within the department were established.

Payment/incentives
Some project groups reported covering travel expenses and offering refreshments, whilst others provided incentives to compensate people for their time and participation. These included discounted or free services such as counselling and spa sessions, Christmas presents, gift certificates and crèche facilities. Only Ramsay (2013) reported offering a £50 payment, noting that this was often declined or given to charity. In some cases, involvement was reportedly viewed by service users as a way of paying back the health service for care received (Hamil & Heslop, 2010;Ramsay 2013;Wilson et al., 2012) and frequently described as including an element of therapeutic support for those involved.

Impacts
A wide range of service improvements was reported to have resulted from the QI projects (Table 6).  Boaz et al. (2016) note that patient involvement "led to priorities and solutions that would not otherwise have occurred to frontline staff immersed in day-to-day service delivery" (page 11). Service user involvement also led to the establishment of ongoing advisory/stakeholder /taskforce groups in a number of cases. For example, Trajkovski et al.
(2015a) report the formation of a family centred care working party that met monthly.
Service user involvement was also reported to have impacted on research priorities (Reay et al., 2014) and on the design of a range of research based interventions and study methods. For example, Ramsay (2103) describes how a website designed as part of a wider research project was altered due to the involvement of service users stating "One patient fell asleep when trying to read qualitative quotes on website, so web designers built in a function to enable it to be spoken rather than read. Also, one lady's husband was visually impaired, which led web designers to use different colours" (telephone transcript). Darbyshire (2015b) explain that their research title and data collection plans were amended in light of service users' views and Menzies et al. (2014) report that PPI gave them a better understanding about how best to capture the perspectives of children with differing experiences. that took place generated a "social capital that creates and improves social provider-consumer networks" (page 1).

Experiences of involvement
Experiences of involvement were generally described as positive and rewarding. Service users reported feeling listened to, having their views valued and feeling as though they were equal partners (Locock et al., 2014a;Trajkovski et al., 2015b;Menzies et al., 2011). Despite some initial concerns that involvement would lead to little or no change, service users noted that, "It's great to see changes happening -it's not just lip service" (Tollyfield, 2014, page 140). Service providers and researchers also reported positive experiences, with one staff member involved in an AEBCD project stating that, "In twenty years of nursing I have never interacted with patients in this way before" (Tollyfield, 2014, page 140).
Some initial feelings of apprehension about working together were reported, however, as project teams became more familiar with PPI, individuals reported feeling more comfortable and confident. Some service users also viewed the experience as a way of 'giving something back' (Locock et al., 2014a;Williams et al., 2013) with one patient saying "I volunteer in the ICU as a token of gratitude and sincere thanks for being given another chance to carry on enjoying life, and to be able to continue giving a little back" (Wilson et al., 2012, slide 11).

Barriers to and facilitators of involvement
Trying to recruit a 'representative' and diverse sample of service users was reported as challenging and rarely achieved.  (Bench et al., 2014;Ramsay, 2013). However, Darbyshire (2015b) reflects on the challenges their research team faced, trying to find people for every project and points out that, "There seems to be a limited cohort of 'professional patients' within the ex-ICU patient community and they are at the point where they cannot take on more responsibilities" (page 2).
The importance of involving people early and sustaining relationships was evident. However, this was not always possible or desired by service users. Locock et al. (2014a) report that patients ended their participation because they did not feel they had anything more to contribute or because they felt that implementation was not their responsibility.
Some service users also became too ill to continue or found it too distressing to sustain involvement. Flexibility was considered vital, and being able to adapt to suit individual preferences and health status associated with more successful involvement (Boaz et al., 2016;Tollyfield, 2014). report providing reassurance to those involved that participation would "have no effect on future associations with the health service" (page 242).
Providing adequate time, space and support to work together was reported to be vital for successful PPI (Boaz et al., 2016;Wilson et al., 2012). One nurse involved in an AI project noted the importance of creating "a physical and mental space that encouraged dialogue, built trust and created links between health professionals and parents" (Trajkovski et al., 2015b, page 4). A physical space suited to the purpose was reported as necessary. In their AI project, Trajkovski et al. (2015b) describe achieving this by using a location that was away from clinical ward distractions, easy to travel to and that offered free parking.
Organisational support emerged as a further facilitator of PPI. Tollyfield (2014) recommends that local staff be involved as much as possible and states that "Buy-in from senior staff and the practice education team is absolutely necessary" (page 137). Willis et al. (2013) also report the value of unit champions "serving as the contact person for questions or concerns" (page 87).

Training and resources
Providing sufficient time at the start of a project for service users to understand their role and to agree ground rules was described as important for success (Boaz et al., 2016). Ensuring adequate resources, such as costs associated with room hire, catering, parking and training are allocated for the duration of the project were also identified as facilitators of effective PPI (Dunn et al., 2006;Trajkovski et al., 2015b;Wilcock et al., 2003).
No reference was made to the provision of training for patients and family members. The need for staff training was supplemented with email and telephone support from researchers was described.

Types and levels of involvement
People become involved in critical care research and QI projects for many different reasons and take part in numerous different ways. Although descriptions of PPI used by authors did not always reflect the INVOLVE definitions (Denegri, 2015), we did not identify any examples of a user led/user controlled approach to PPI, a finding consistent with literature in the area of mental health where substantive and widespread service user/survivor leadership in research is discussed as something yet to be attained (Callard & Rose, 2010;Jones & Shattel, 2016).
Consultation was the main PPI approach described, predominantly using service users' experiences to inform service improvement. There is some evidence that involvement is growing in other areas-for example, participation in research project steering groups. However, where collaboration was described in the reviewed papers, it appeared to be invariably initiated by health professionals or researchers.
Blurring of roles was also evident, with patients sometimes participating as research subjects as well as carrying out involvement activities, a finding also reported by Mathie et al. (2014). The impact of these combined roles and responsibilities on projects and individuals warrants future consideration.

Processes, methods and impact
The papers included in our review were predominantly descriptive, with very little evaluative/research data reported.
There was also very limited information provided with regard to the service users involved, their experiences of PPI processes and outcomes or the facilitators and barriers associated with PPI in critical care service improvement and research projects. These findings concerning the paucity of detailed publically available information about PPI processes, methods and impact are consistent with those reported in generic literature on PPI Conklin et al., 2012;Mathie et al., 2014;Mockford et al., 2012).
The absence of crucial information of this sort makes it difficult to fully understand and appreciate PPI processes and their likely impact (Conklin et al., 2012;Crawford et al., 2002;Nilsen et al., 2006). Findings from the present review do, however, support that PPI can influence the prioritisation of work streams, as well as the design of interventions and study methods. Data from our review also suggest that once people overcome their initial apprehension, service user involvement in critical care is associated with positive experiences.

Barriers to and facilitators of public and patient involvement
In agreement with the existing literature (Conklin et al., 2012;Staniszewska & Denegri 2013;Snape et al., 2014;Wilson et al., 2015), we found key facilitators associated with meaningful PPI to include availability of adequate time and resources/support for PPI activities; a physical and mental space conducive to collaborative working; organisational support and buy-in; flexibility and adaptability to suit different contexts, individual service user preferences and health needs; and creating a respectful environment in which people can work together in partnership. Findings also highlight the importance of making PPI routine, so that it becomes part of the organisational culture.
A key challenge reported in our review concerns the recruitment of 'representative' and diverse groups of service users to critical care research and QI projects. Service user 'representativeness' is a pervasive theme in the existing PPI-related health research literature (Jones & Shattel, 2016;Mathie et al., 2014). Achieving a 'representative sample' of service users in the domain of PPI may be impossible and furthermore, may not be what we should be aiming for. No single person is able to represent others, thus we should perhaps be focusing more on how best we can capture the collective sharing of experiences, views and values among service users. Some authors also report a negativity and a tokenistic response towards involving service users, arguing that those involved tend to be articulate and able to work with project teams (El Enany et al., 2013). El Enany et al. (2013) argues that these service users are also complicit in professionalising their involvement role. However, what has been discussed in the literature as the 'professionalisation' of service users in PPI (Thompson et al., 2012) involves a complex set of issues. This entails a need to consider with greater care the social and material conditions underpinning the professionalisation of service users in the PPI arena.
How to access 'hard to reach' population groups, such as those from minority backgrounds, those who have been bereaved, or those who have cognitive difficulties is an unresolved dilemma. Findings from this review suggest a desire to include bereaved family members, but limited examples of this actually happening provides little evidence with which to guide future practice. It is often assumed that bereaved family members may not want to take part, but this remains an assumption rather than an informed understanding of their wishes and views. When it comes to former ICU patients who may experience cognitive difficulties, it should not be a priori assumed that this group of patients would not want or would not be able to be involved in research or QI initiatives; researchers and QI practitioners need to be more innovative, flexible and adaptable with regard to how people with cognitive difficulties can be involved, at whatever level they are able and want to.

Limitations of the review
Data about PPI is often hidden, is frequently not reflected in study titles or abstracts and may only be uncovered by word of mouth, making it difficult to identify using standard review processes (Rogers et al. 2016). This review included published abstracts as well as full text publications and no quality appraisal of included studies was undertaken. These decisions were taken because of the limited available research data and a desire to reflect the most current evidence status. We also excluded protocols as they do not necessarily reflect what actually happened. However, this can possibly mean that this review does not reflect fully the current state of PPI practice in critical care.
Furthermore, studies in which service users were co-authors on the paper were in the minority, thus our findings are mostly reliant on professional reports of service users' views and experiences.

Conclusion
This is the first published review to detail the nature and extent of PPI in critical care research and QI. Limited research data were identified, but findings from a broad range of evidence support that PPI is becoming more commonplace. On the whole, we found experiences of PPI within critical care to be reported positively, and available data suggest that PPI does have a positive impact on both the design and conduct of both research and QI initiatives. The key barriers and facilitators of PPI that we identified relate to the challenge of recruiting a diverse group of service users, dealing with power hierarchies, being adaptable and effective consideration of the resource requirements. Findings lead us to the conclusion that we currently lack an evidence base with which to guide our practice. We offer the following recommendations and insights, which we hope will aid the future development of PPI in critical care in a systematic and evidence based way.

Policy, practice and research recommendations
PPI in critical care research/QI has received limited attention when it comes to the evaluation of its processes, methods and impact. More research is required to identify the best methods with which to recruit and involve people, as well as to explore the appropriateness of and timeliness for approaching and involving patient/family groups following hospital discharge.
Not being able to easily locate relevant literature is a significant barrier to the utilisation of available evidence. The development of a search filter for PPI, such as the one described in a study by Rogers et al. (2016) could aid the identification of literature on PPI in critical care, and its subsequent application to practice. However, as Rogers et al.
(2016) acknowledge, locating PPI research hidden in the body of the text will remain an issue until database indexing and searching facilities are improved.
The lack of reporting regarding methods of PPI does not necessarily mean PPI did not take place, but it does mean that others are unable to learn from previous positive or negative experiences. Some toolkits and reporting guidelines for PPI activity are now available (Bagley et al., 2016;Staniszewska et al., 2011). Further development and application of these to critical care projects should help to address this important oversight, adding to the developing body of improvement science knowledge.
Finally, the papers included in our review raise a number of crucial implementation questions that future research needs to consider: i) To what extent does PPI influence the adoption and implementation of QI programmes in critical care?
ii) How does PPI shape the complex environments (contexts) within which QI programmes in critical care are implemented?
iii) Does PPI impact on issues of fidelity to and adaptation of QI interventions in critical care, and if so, how?
The  Service users are actively and collaboratively involved as members of research or project implementation teams, as co-researchers, co-implementers and co-authors of academic publications Consultation Service users are consulted, asked for advice, and/or provide information that is used to inform decision making by others

Inclusion criteria Exclusion criteria
All documents that report data about:  Any type of service user engagement or planned engagement in research, service development and/or clinical audit activities within critical care (adult, paediatric, obstetric and neonatal)  The nature, extent and terminology used to describe public and patient involvement PPI*  Methods for capturing PPI data and measurement  The impact of PPI on health or process outcomes  Impact of the research on individual users or research team members, on groups and/or on organisations  Experiences of and/or reflections on PPI Documents that:  Only report service user satisfaction data, with no link to a service improvement/research initiative  Report data collected on projects conducted outside critical care  Do not report data collected/to be collected by the authors of the paper  Report study/quality improvement protocols *PPI: Public and patient involvement Reports an ethnographic evaluation of AEBCD in two intensive care units. Robert et al. (2015) Reports experiences of using EBCD in intensive care.   Service users consulted on study design and/or asked to review study documents (six reports)  Service users took part in stakeholder engagement activities such as workshops/research priority setting exercises (three reports)

Collaboration
 Service users on project steering group (nine reports)  Service users participated in workshops or other activities focused on co-designing/co-producing a project, intervention, solution or implementation strategy (16 reports)  Service users manage an area of service provision e.g. memorial service, care navigators, patient and family support roles, writing patient information, sourcing new equipment, teaching staff, chairing committees, representing service users on various committees, writing reports or other volunteer roles (seven reports)  Service user assisted with data collection and/or data analysis (six reports)  Service user co-authored publications/presentations or took part in other dissemination activities (five reports)  Robert et al.(2015)  Sign to enhance dignity and privacy  Promotion of family involvement in personal care  Washing of patients' hair more regularly  Changing time of patient washes  Trial of a new wash basins  Encouraging wards to send patients' wash bags to the intensive care unit  Clocks to aid patient orientation  Encouraging nurses to brush patients' teeth more regularly  New toothbrushes  Promoting correct application of continuous positive airway pressure masks  Booklet informing patients about potential for hallucinations  Changing waste removal to outside of the rest period  Redesign of the intensive care unit discharge summary form  Redesigned lighting systems  Tablet computer applications to assist ventilated patients communicate  V-shaped pillows for postoperative patients  New process for transfer of patients' belongings from theatre to recovery ward  Improved patient-doctor communication during ward rounds Bench et al. (2012;  Intensive care unit discharge summaries for patients  Discharge information for patients and relatives N/A Eligibility criteria 6 Specify study characteristics and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

5-6
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.

5-6
Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

5-6
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in review) 6 Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

7-8
Data items 11 List and define all variables for which data were sought.

7-8 Risk of bias in individual studies
12 Describe methods used for assessing risk of bias of individual studies and how this information is to be used in any data synthesis. 14-17 FUNDING