Do patient engagement interventions work for all patients? A systematic review and realist synthesis of interventions to enhance patient safety

Abstract Background Patients are increasingly being asked for feedback about their healthcare and treatment, including safety, despite little evidence to support this trend. This review identifies the strategies used to engage patients in safety during direct care, explores who is engaged and determines the mechanisms that impact effectiveness. Methods A systematic review was performed of seven databases (CINAHL, Cochrane, Cochrane‐Central, Embase, ISI Web of Science, Medline, PsycINFO) that included research published between 2010 and 2020 focused on patient engagement interventions to increase safety during direct care and reported using PRISMA. All research designs were eligible; two reviewers applied criteria independently to determine eligibility and quality. A narrative review and realist synthesis were conducted. Results Twenty‐six papers reporting on twenty‐seven patient engagement strategies were included and classified as consultation (9), involvement (7) and partnership (11). The definitions of ‘patient engagement’ varied, and we found limited details about participant characteristics or interactions between people utilizing strategies. Collaborative strategy development, a user‐friendly design, proactive messaging and agency sponsorship were identified as mechanisms to improve engagement about safety at the point of direct care. Conclusions Agency sponsorship of collaboration between staff and patients is essential in the development and implementation of strategies to keep patients safe during direct care. Insufficient details about participant characteristics and patient–provider interactions limit recommendations for practice change. More needs to be learned about how patients are engaged in discussions about safety, particularly minority groups unable to engage with standard information. Patient or Public Contribution Review progress was reported to the CanEngage team, including the consumer steering group, to inform project priorities (PROSPERO CRD42020196453).


| BACKGROUND
Patients are often the only constant element in their healthcare journey and provide important contextual information for designing safe healthcare services. 1 The fact that patients can retrospectively identify unsafe events that occurred during their care is well established. [2][3][4][5] They also play an active role in their own safety by raising concerns or flagging inconsistencies and inaccuracies during healthcare interactions. 4 Over the past 20 years, interventions that encourage patients to discuss or raise concerns about inaccuracies relevant to their care have been implemented. [6][7][8] These interventions occurred simultaneously with evaluations of patient involvement in system-and service-level patient safety programmes, such as patient-led incident reporting systems. 9 The use of patient-centred tools and strategies to enhance safety has increased despite limited evidence about their effectiveness.
Limited research about the effectiveness of patient engagement, and the depth of engagement needed to promote safe care, is reflective of wider inconsistencies. There are various definitions of patient engagement, involvement and participation in the literature. 10 Carman et al. 10  and refers to the period when clinicians deliver healthcare services or treatments to patients; this can be hospital or community based. 12 With growing recognition of the value of engaging patients in healthcare design and delivery, and the susceptibility that some specific population groups have to adverse events, the need for better data about facilitating engagement is imperative. The literature identifies various system, service and clinical factors that support effective patient engagement such as education about their condition, 13,14 empowerment to engage 15,16 and the willingness and ability of clinicians and patients to communicate about safety. [17][18][19][20] The extent to which an organisation is committed to patient engagement is a measure in most organisational safety culture surveys, but there is little evidence of the enablers and system prerequisites to facilitate effective engagement. 21,22 Evidence of the enablers and system prerequisites for effective engagement have not been synthesized to support the implementation of such interventions. 22 Similarly, evidence about the nature and extent to which patients are engaged in safety is fragmented and lacks information about approaches for diverse populations, such as people from culturally and linguistically diverse (CALD) backgrounds or other communication needs. 8,21,23,24 This systematic review aims to address the knowledge gaps identified above using a realist synthesis 25 to explore the following questions: (1) What interventions have been used to engage patients in safety during direct care and what is the mode of intervention (e.g., video, paper chart, electronic portal) and extent of engagement (e.g., number of opportunities, with whom)? (2) What types of patients and their contexts are described in the interventions? (3) What are the mechanisms that influence the effectiveness of consumer engagement approaches in enhancing safe care and treatment?

| METHODS
A systematic review and realist synthesis were undertaken and reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) statement. 26

| Prospero registration number: CRD42020196453
Inclusion criteria: Studies published between January 2010 and December 2020 in English were included. All research designs were eligible, including qualitative, quantitative, multi-and mixed-method studies. All studies included participants who were healthcare consumers, patients, family members or other caregivers. Safety outcomes in the clinical encounters described encompassed increased notifications of or the prevention of safety breaches, errors, accidents, incidents, complications and infections. Selected interventions had to use patient engagement designed to minimize harm.
Studies that did or did not include a comparator intervention were eligible.
Exclusion criteria: All studies outside the date range or published in a language other than English were excluded. Systematic or other literature reviews were not included, but their reference lists were searched. Studies that focused on methods beyond direct care, for example, to enhance governance or inform improvements to organisational safety for example, adverse event reporting systems or service governance, service planning, self-management or improving health, such as self-management for people with chronic conditions, were beyond the scope of the present review. Studies about patient involvement in training medical or nursing staff, patient attitudes towards safety or willingness to participate and studies about participatory research or codesign methods unrelated to safety were not included.

| Study identification
The key concepts of patient engagement and unsafe healthcare were used to generate keywords, synonyms and phrases to inform a comprehensive search strategy (see File S1). The search strategy was

| Study selection and data extraction
Search results were exported to Endnote (X10) and duplicates were removed. Articles were then extracted to Covidence systematic review management software (Veritas 150 Health Innovation). Two reviewers (J. L. and B. J.) completed the initial title and abstract review, followed by an independent screening by a third reviewer (B. N). The inclusion criteria were then independently applied to full-text articles by two reviewers (B. N. and R. H.), with disagreements or uncertainty resolved through discussion. The following data were extracted: author, year, country, aims/ objectives, setting, number of participants, participant characteristics, inclusion of diverse populations, method of data collection and samples, intervention/method of patient engagement, main findings and what worked (enablers, barriers).

| Assessment of study quality
Due to the heterogeneity of the study types, the Quality Assessment Tool for Diverse Studies, a validated quality appraisal tool, 27 was used. Two reviewers (B. N. and K. J.) independently applied the 13 criteria to the included studies. The κ test was used to determine inter-rater reliability, and substantial reliability was confirmed (κ = .726). 28

| Data synthesis
Findings were synthesized using a narrative approach and the realist framework to explore which interventions worked, in what conditions and with whom. 25 Realist evaluation was selected because it examines the conditions that facilitate success rather than just information about whether the outcome or intervention 'worked'. 29 Key findings relevant to the review questions were extracted, including barriers and enablers to implementation. Carman et al.'s 10 engagement framework was used to determine the extent of engagement in the interventions, and engagement strategies are described in relation to the three levels of engagement.

| RESULTS
The systematic search produced 3029 papers, with 2706 studies excluded and 217 duplicates removed. A total of 82 full texts were reviewed and 55 were excluded (28 included outcomes not related to safety, 14 study designs and 9 intervention types did not fulfil the inclusion criteria, 3 were in nonhealth settings and 1 more recent paper available), leading to 26 included publications describing 27 strategies. Figure 1 shows the search and selection process.
Characteristics of the included studies: Studies originated from the United States (13), the Netherlands (4), the United Kingdom (3), Canada (2), Vietnam (1), Australia (1), Korea (1) and Norway (1). Seventeen of the twenty-six studies were focused on inpatient safety, three on specific clinics or treatments and six were focused on treatment between face-toface visits. The studies were conducted in a range of clinical areas including inpatient adult general medical services, 30 52,53 and homecare visits, appointments, discharge and pharmacy community. 32,52,54 Study quality: A score of 0-3 was assigned in the 13 categories used to assess quality (see File S2). Of the 26 included papers, most achieved high scores (2 or 3) in the categories of statement of aims (23), appropriate study design to address research aims (22) and format of data collection tool to address research aims (23). Fourteen studies contained limited details in the participation data provided (Criteria 9) and eight described involvement of 'consumers or stakeholders' (Criteria 12) in the process of study design and conduct.   Table 1). The engagement activities are described in relation to approaches that were focused on consultation, involvement or partnership relative to the Carman Framework. 10 Evidence from each study of the effectiveness of a strategy in improving safety is presented in Table 1.  Table 1 conceptualize engagement as consultation. What is distinct in this phase of engagement is that patients were consulted, or invited to provide input, about a specific safety issue/s within parameters of engagement set by health practitioners.
Four strategies involved staff-initiated engagement about a specific treatment or potential adverse event. Kim et al. 44 describe how direct questioning by staff to patients about the site of their X-ray at an orthopaedic clinic in Korea led to a significant decrease in X-ray site errors. Bergal et al. 40 describe a similar strategy implemented to reduce wrong-site surgery in the United States of America with less definitive findings, primarily due to few incidents of wrong-site surgery. van Gaal et al. 39 described a programme focused on reducing poor outcomes by staff providing education and opportunity for engagement in three areas (ulcers, urinary tract infections and falls) in 10 wards across four hospitals and aged care facilities in the Netherlands. 39 Rochon and Salazar 42 described a four-stage falls reduction process implemented in medical/surgical wards in a USA hospital. Although both van Gaal et al. 39 and Rochon and Salazar 42 reported decreased falls and fewer adverse events, limited details about interactions between staff and patients were reported. These two interventions have been classified as consultation due to the focus on patient education and staff-directed interaction. 39,42 One strategy sought to create engagement about a safety event by driving patientinitiated contributions through a feedback mechanism: the Patient Reporting and Action for a Safe Environment (PRASE) Tool. PRASE was trialled in 33 wards across 5 UK hospitals, which demonstrated a decrease in preventable harm at the ward level. 50 Four strategies described staff adapting existing engagement tools to promote interaction by staff with patients in hospital settings and were relevant to this category due to the focus on patient education and staff-directed interaction. 31 Table 1 for effectiveness data).

Involvement:
The involvement phase of engagement 10 indicates that patients were asked about their preferences and concerns, with the opportunity to interact and engage with practitioners about a specific health or treatment issue. This stage of engagement contains strategies devised by staff, offering opportunities for increased ongoing interaction between staff and patients that were not evident in strategies classified as consultation. Six strategies sought to enhance safety by involving patients (Table 1).
Only one strategy in the involvement phase of the continuum was related to face-to-face interactions between staff and patients, reporting on a strategy used in an outpatient interventional radiology clinic in the Netherlands. 45 Clinic patients were invited to attend an additional appointment before their interventional radiology visit to discuss their queries and concerns about the procedure, risk and consent, hence the classification as involvement. This strategy enhanced the relationship between the practitioner and the patient, led to increased informed consent and a reduction in deviations from process (Table 1). Two further online strategies were used to facilitate communication between patients and health practitioners 35,38 about specific areas of care. de Jong et al. 38 Table 1).
Partnership/leadership: Strategies that create a partnership between healthcare providers and patients are at the endpoint of the continuum of engagement. 10 Almost half of the strategies (12) sought to provide patients with the opportunity to raise concerns about their treatment and 'work' with practitioners to improve the safety of their care and treatment, often with strategies using person-centred tools or designed to empower patients to alert practitioners of concerns. 30,32,36,43,[46][47][48][49]51,52,54 Although all strategies in this classification enhance partnership, only six strategies included patients in the inception, design or evaluation of strategies. [46][47][48][49]52,54 Of the range of strategies included, six described collaboratively developed tools [46][47][48][49]52,54 and processes designed to encourage and facilitate patient communication and feedback. All six studies reported positive impacts on patient safety, including decreased adverse events and increased identification of errors that would have resulted in harm (see Table 1). Four of the six strategies were 'bedside' tools collaboratively developed with patients designed to enhance quality and included 'safety' as one of many goals. [46][47][48]52 Dykes et al. 46 Table 1 for effectiveness data). The included studies were predominantly conducted in inpatient settings (13 studies) or after discharge from inpatient stay (10); the remainder were conducted in outpatient clinics or community settings, including an aged care facility (4). Twenty-four studies were conducted in countries classified as 'developed' by the United Nations, 25 studies in countries classified as high income and one developing economy. 51 Participants were most often recruited from university or teaching hospitals (22), predominantly city based (17), and in the United States of America (13). Twenty-five studies recruited male and female participants; one study recruited only women. 43  The nature of engagement across the strategies spanned the patient engagement classifications of consultation (nine strategies), involvement (seven strategies) and partnership (twelve strategies). 10 Working in partnership with patients and families is central to devising suitable engagement approaches for specific populations or settings. 10 It was notable that publications provided varied levels of detail in data about the type and degree of patient engagement in strategy development or implementation. In some instances, researchers identified limited inclusion of diverse patients as an issue to address; however, it was difficult for researchers to ascertain whether it was the patient engagement strategy or research data collection tools that precluded engagement. 32,38,45 While some papers included details about codesigning strategies with patients, [46][47][48][49]52,54 this aspect of engagement is most often absent, undefined or un- reported. Insufficient information about such elements as the patient role in strategy design reflected limited evidence that the strategies described were theoretically informed. Lawton et al. 50 provided a theoretical background to engagement; similarly, the patient-centred strategies embedded engagement in such approaches. [46][47][48] However, the theoretical justification for strategy design presented in most papers was on content (e.g, falls prevention or wrong-site intervention), technical production (e.g., videos 37 ) or staff implementation, 42,44 rather than on the nature or details of engagement.
Attributing changes in patient safety outcomes to a particular type of patient engagement was challenging due to the variation in the definition of engagement in the included papers, which is reflected in the wider literature. 10  Patient-professional collaboration, user-friendly strategies, proactive messaging and agency sponsorship were all recognized as enablers of patient engagement. Findings regarding the facilitators of patient engagement between papers in this review were consistent, confirming recent research seeking to empower patients to raise safety issues within a supportive culture. 15,16,22,61,62 The importance of agency sponsorship of a collaborative culture for engagement has long been emphasized in the change management and personcentred care literature. 63 Staff identified that agency support is required to address the potential impact of engagement on the patient/ provider relationship and workload. 34  Few studies have utilized theory-informed approaches or robust study designs to evaluate current techniques.
There are implications for health services in the challenges posed to scaling and spreading the adoption of potentially useful patient engagement strategies. There is a danger of unintended harmful impacts for those for whom the intervention may not be suitable.
There are resource, financial and ethical implications, given the additional time and technologies required by patients and staff to take part in such interventions. This review reinforces the need for a multifaceted approach to patient engagement, incorporating agency culture, practices and appropriate engagement strategies. 9 Therefore, researchers need to work collaboratively with health services to establish more robust evidence of (a) what the intervention mechanisms are in current strategies and (b) information about (1) the feasibility and acceptability of the strategies for all parties, (2) the end-users and (3) cost-effectiveness.

| Strengths and limitations
The capacity to explore varied engagement strategies by using a The findings of this review must be understood in the context of the limitations of the included studies. We identified limited geographical diversity in the countries where the research originated and a lack of studies that sought to compare groups, or samples that were sufficiently powered. By including only published material, valuable insight from nonpublished and nonempirical work may have been missed. 65 An additional limitation arises from the wide range of terms used to describe patient engagement in safety and the many different types of journals used to house patient safety research. The lack of evidence regarding the theoretical underpinning of the interventional approaches and their intended impact on patient engagement creates barriers to determining the intervention mechanism/s responsible for identified changes. The diverse purposes of papers included also created challenges, particularly papers that reported on a single element of a bigger project, multiple interventions across several sites or safety outcomes reported among a number of interventions carried out simultaneously.
The levels of sensitivity and precision of bibliographic databases vary and can also affect the number of articles returned. We used several databases in addition to manual searching to broaden coverage, but there may have been omissions.

| CONCLUSION
Despite the growing number of patient-centred tools and safety engagement strategies, evidence about use and effectiveness is limited. More details about how they are used and with whom are required to enable patients and practitioners to engage effectively.
More clarity is needed to consistently define patient engagement along with further research to determine which strategies are effective. Little evidence exists about people from minority or vulnerable backgrounds in patient safety, which needs to be addressed due to acknowledged disparities in healthcare safety and engagement.