Using Behaviour Change Theories to Understand Pediatric Emergency Discharge Communication Interventions

Interventions to improve discharge communication in emergency practice settings are inherently complex and incorporate multiple strategies aimed at patient, caregiver and healthcare provider behaviours. Behaviour change theories, including the capability-opportunity-motivation-behaviour (COM-B) model and Behaviour Change Technique Taxonomy version 1 (BCTTv1), provide consistent language for identifying and evaluating the active components of interventions. The objectives of this study were to: (1) identify common barriers and enablers to discharge communication interventions and (2) describe behaviour change techniques (BCTs) present in pediatric emergency discharge communication interventions aimed at patients, caregivers and health care providers.

Pediatric emergency department (ED) settings can be noisy and chaotic and the discharge process can be a stressful for patients and families. Interventions to improve discharge communication are often complex and can incorporate multiple strategies aimed at health care provider, caregiver and patient behaviours [19,20]. Effective discharge communication interventions include standardizing content delivered through both verbal and written modes, and ensuring patient comprehension and recall through strategies such as demonstration of the behaviour, prompts, and social support [21,22]. Findings from a recent systematic review and narrative synthesis of pediatric discharge communication found that the majority of interventions across both acute and chronic illness presentations were aimed at improving caregiver comprehension of health information [23]. Further, many of these discharge communication interventions include multiple intervention components (e.g., education, reminders) and targets (e.g., patient and family, health care provider and system), making it di cult to discern which intervention components, or interaction of multiple components, resulted in the measured outcome [2,23,24].
To date, behaviour change frameworks have yet to be used to identify the barriers, enablers and active ingredients in emergency discharge communication interventions. Using the results from a recent systematic review and narrative synthesis of pediatric emergency discharge communication, the objectives of this study were to: (1) map identi ed intervention barriers and enablers identi ed by authors in the experimental and quasi-experimental studies to the COM-B model, and (2) describe common BCTs included ED discharge communication interventions.

Methods
We conducted a secondary analysis of 35 experimental and quasi-experimental studies included in a systematic review and narrative synthesis of discharge communication in pediatric emergency care to gain a better understanding of the behaviours related to discharge communication and the characteristics of existing interventions [23]. A full description of the narrative synthesis protocol [25] and ndings [23] can be found elsewhere. Coders were blinded to the intervention function classi cation completed during the narrative synthesis.

Identifying and mapping intervention barriers and enablers
All included studies were analyzed by two reviewers (JAC and AJG). First, the full text articles were imported into NVivo 11 qualitative software [27], and the study reports were reviewed to identify barriers and enablers to intervention implementation and effectiveness as reported by the study authors. The reviewers met after coding every 5-7 studies to compare consistency and discuss any discrepancies. Once barriers and enablers were identi ed they were mapped onto relevant domains within the COM-B model. The two reviewers independently mapped the barriers and enablers to the components of the Page 5/27 COM-B model and then met to compare ndings of the mapping exercise. Discrepancies in mapping were resolved through discussion.

BCTTv1 coding
Similar to other research [16], we developed rules a priori to guide BCT coding and reviewer consensus. The rst rule was to code only those intervention components with su cient detail to clearly identify a BCT. This prevented over-coding of interventions by limiting coder assumptions that a BCT could be present but poorly described. The second rule was speci c to education interventions, which are common in discharge communication. We assumed that components of education intervention included at a minimum instruction on how to perform the behavior. Other BCTs were coded in education interventions only when su cient details were provided.
Textual data from study reports were managed using NVivo 11 [27]. All reviewers completed the online BCTTv1 training course [28] and received a copy of the 93 individual BCT de nitions and examples from the BCTTv1 [7,10] prior to coding. Four reviewers (AJG, AB, JE, CC) independently coded the intervention descriptions in all included studies. The study objectives were reviewed rst to obtain a sense of the intervention target and outcomes of interest. The methods sections were then read in their entirety and analyzed to identify BCTs in the interventions. BCTs were only identi ed in content related to the intervention groups, as the control groups received either no intervention or standardized care.
Following independent coding, a consensus meeting has held to compare coding and resolve discrepancies. Coding discrepancies were discussed as a team and resolutions were achieved by consensus. When coding discrepancies could not be resolved, a fth team member (JAC) was included in the discussion to help achieve consensus. We chose to reach consensus rather than use inter-coder reliability statistics as inter-coder agreement has not been well established in the literature [29]. All coding resolutions were then noted in the NVivo le. Descriptive statistics were used to determine the frequency of the 93 individual BCTs and 16 BCT categories for patient, caregiver and health care provider interventions. Our study report adheres to the Enhancing Transparency in Reporting the Synthesis of Qualitative Research (ENTREQ; [26]) checklist (see additional le 1).

Barriers and enablers
Barriers and enablers were identi ed across patient/caregiver, health care provider, intervention and health system levels. All six domains of the COM-B model were identi ed across the barriers and enablers, with psychological capability being the most commonly identi ed domain across levels. The full list of barriers and enablers and identi ed COM-B domains are outlined in Table 1.
Patient/caregiver barriers ranged from low health literacy levels, feelings of anxiety and stress, and poor understanding of discharge instructions. Health care provider barriers included resistance to change and poor relationships with patients/caregivers. Complex interventions that required a substantial allotment of resources to implement was also identi ed as a key barrier. Finally, two important health system barriers included the chaotic nature of the ED context and limited access to primary care providers.
Enablers were also identi ed across studies. Enablers included providing caregivers with information early in the ED visit, having health care providers use pre-formatted discharge communication templates as part of the intervention, establishing links between the ED and primary care providers, and personalizing intervention content to the patient's needs.

BCT Coding
Two [60,61] of the 35 experimental studies did not include identi able BCTs as components of the interventions and were therefore not included in the analysis. These interventions either did not target behaviour change or did not provide enough detail in the methods sections to identify potential BCTs. Table 2 provides an overview of study characteristics and identi ed BCTs from the remaining 33 studies.

Patient/Caregiver targeted interventions
Of the 33 studies that included BCTs, 29 interventions targeted patients/caregivers. A total of 20 out of the 93 individual BCTs, from 12 of the 16 BCT categories, were identi ed in the interventions ( Figure 1). The number of BCTs identi ed in each study ranged from 1-9 BCTs, with a mode of 1 BCT per intervention. The three most commonly coded BCTs in patient/caregiver interventions were: 4.1-Instruction on how to perform the behaviour: Fourteen studies included this component as part of their interventions [30-34, 37, 38, 42, 43, 47, 50, 53, 55, 62]. Interventions that included this BCT were most often aimed at asthma, fever, and gastroenteritis illness presentations. e.g., "the video discusses methods for taking a temperature, outlines indications for contacting a physician, refutes common parental misconceptions about fever, and identifies methods to comfort a febrile child" [37]. [30, 33, 39, 47, 48, 50, 55-57, 63, 64]. This BCT was identi ed across numerous studies, including interventions aimed at pain management, mental health, and wound care. e.g., "the instructions were legible, were written to be understood by a patient with a seventh-grade education, and included information concerning suture removal, wound care, signs of infection, and methods to reduce scarring" [30].

5.1-Information about health consequences: Eleven interventions included information about health consequences
3.2-Social support (practical): Ten studies used practical social support in their interventions [33,40,43,47,48,50,58,59,62,63]. In one study [62], this BCT was used in two components of the intervention. Six of these interventions were aimed at asthma or mental health presentations [43,47,50,58,62,63]. e.g., "the educator also asked about and reinforced use of the asthma action plan, provided a referral for any identified obstacles, tried to help the family set goals to assure follow-up, and addressed any new concerns or questions as appropriate" [43].

Health care provider targeted interventions
Only eight of the 33 studies that included BCTs targeted health care provider behaviours. Four interventions solely targeted health care providers [35,41,44,51], while the other four interventions primarily targeted caregivers but also included health care providers [42,[48][49][50]. Four of the potential 93 individual BCTs were identi ed from these eight interventions across four BCT categories: shaping knowledge, natural consequences, associations, and repetition and substitution ( Figure 1). The four identi ed BCTs in health care provider interventions were: 4.1-Instruction on how to perform the behaviour: Four studies included this BCT as a component of their interventions [44,[49][50][51]. Two of the interventions were focused on asthma, one on mental health and one otitis media illness presentations. Further, one study [44] used this BCT at two different points in the intervention. e.g., "…all nurses and Respiratory Care Practitioners (RCP) were instructed on the use of the plan to ensure that all healthcare providers delivered the same messages to the patients and parents" [51]. [41,42,44,48]. Two of these interventions targeted asthma, one targeted otitis media illness presentations, and the nal intervention used prompts/cues to encourage families to purchase and use helmets. e.g., "physicians were provided, during orientation sessions, with a small plasticized pocket cue card outlining a number of points to serve as the basis for counseling the family about the risks of cycling and the bene ts of helmets" [48]. [35,50]. One of the studies targeted asthma and the other targeted fever. e.g., "each ED was introduced to the protocols through educational sessions outlining the importance of compliance with asthma care guidelines" [50].

5.1-Information about health consequences: Two studies included this BCT
8.1-Behaviour practice/rehearsal: Only one intervention included this BCT [49]. This intervention targeted mental health presentations and also included numerous intervention components targeted at caregivers and patients to increase use of outpatient mental health services and to decrease suicide attempts. e.g., "clinicians with graduate mental health training received didactic training with role playing, observed intervention sessions, and were observed until a senior clinician certi ed them as pro cient" [49].

Discussion
The application of the BCTTv1 was useful for identifying the active ingredients present in pediatric emergency discharge communication interventions targeting caregivers and health care providers. We identi ed 20 out of a possible 93 BCTs across all studies, indicating that less than a quarter of the available BCTs are commonly used in interventions to improve ED discharge communication. This low percentage of identi ed BCTs is consistent with other reviews of BCTs in diabetes care [17,66] and medication adherence literature [67].
Changing patient/caregiver behaviours The majority of ED discharge communication interventions from our analysis solely targeted caregivers. Instruction on how to perform the behaviour was the most commonly identi ed BCT in interventions across all illness presentations that aimed to improve caregiver compliance, comprehension, or recall of discharge instructions. This is not surprising as discharge communication is meant to instruct patients and caregivers on how to perform behaviours speci c to managing care of their child at home. This nding is also consistent with BCTs identi ed in the diabetes [17,66], cardiovascular disease [68], and physical activity [69] literature. Interestingly, many of the common BCTs used to address other health behaviours,, including action planning [67], goal setting [66,68,70], self-monitoring of behaviours [67,[69][70][71], problem solving [70,72] and reducing negative emotions [68,70] have not been widely used in discharge communication interventions to date. While education is a commonly employed behaviour change strategy, without the addition of BCTs like action planning and/or goal setting, it is unlikely to lead to changes in behavior [73,74]. There is further evidence to suggest that interventions that incorporate barrier mitigation and problem solving along with action planning can be even more e cacious [75,76]. As such, there is an opportunity to leverage the strengths of these BCTs to examine their utility in conjunction with educational interventions for ED discharge communication.
We identi ed a subset of studies which incorporated practical social support as the only BCT [40,58], or combined with other BCTs [33,43,47,48,50,59,62,63] to increase follow-up care after discharge from the ED to reduce inappropriate return visits. This is consistent with other studies which found that practical social support techniques such as phone call reminders and customized asthma care plans, can improve the likelihood of follow-up care with primary care providers [77]. Including social support in health-related interventions has been associated with long-term positive outcomes, especially in behaviour change studies [78,79], as well as having an overall positive impact on health [80]. However, no signi cant impacts on system level outcomes, such as ED return rates or unscheduled care, have been found [77].
Finally, interventions targeting chronic illness presentations, like asthma and mental health, were more likely to include multiple BCTs compared to acute illness presentations. Of the 13 asthma interventions targeting patients/caregivers, ve interventions included at least three BCTs [36,42,43,50,62]. Each of the three mental health interventions included between four and nine BCTs targeting patients/caregivers, signi cantly higher than the mode of one BCT per intervention in the majority of other studies [47,49,63].
As chronic disease often involves multiple interacting risk factors, management of these conditions may require more multi-component, complex interventions.
The identi ed barriers to effective ED discharge communication provides some guidance on which BCTs should be prioritized and evaluated moving forward [81]. BCTs associated with increasing familiarity and con dence in performing behaviours following ED discharge, such as demonstration of the behaviour and behavioural practice/rehearsal, may help to overcome barriers associated with improved teaching and demonstration. Future research should consider the possibility of a wider range of important patient and caregiver level barriers to effective discharge communication when designing intervention strategies Changing health care provider behaviours Few interventions in our analysis targeted ED health care provider behaviour. This is concerning given that effective discharge communication requires an interactive process between health care providers and patients/caregivers that includes assessment of comprehension [21]. By primarily focusing interventions on patients/caregivers, half of the discharge communication equation is missing.
Coding the barriers and enablers to intervention effectiveness allowed us to identify factors at the health care provider level that may affect uptake of the intervention. Some of the barriers identi ed at this level could be addressed by incorporating BCTs speci cally targeting health care providers into the interventions. For example, health care providers' resistance to intervention uptake could be addressed by including habit forming, habit reversal, or restructuring the physical environment components in the intervention. These BCTs may help facilitate the uptake of new clinical behaviours. Additionally, health care providers' rapport with caregivers could be addressed by use of prompts/cues, which could support providing more consistent and comprehensive discharge communication. Other studies detailing health care provider-focused interventions have also identi ed prompts/cues and instruction on how to perform the behaviour as commonly coded BCTs [66]. Reminders have also been shown to be an effective strategy to change health care provider behaviours across health care settings [82], although factors such as patient preference may ultimately override their decision-making processes [83]. As only four studies in our analysis included BCTs targeting both caregivers and health care providers, greater evaluation of health care provider-focused behavioural interventions are needed.

Intervention content descriptions
While the majority of interventions in our analysis included education components, education content and delivery were rarely reported in enough detail to identify additional BCTs that might have been present. Our ndings suggest there is a critical need for improvement in study reporting, particularly regarding details about intervention components and implementation strategies, clearly explicating assumptions about how the intervention is expected to work. This has been identi ed as an important factor to advance the science of behaviour change and improve replicability [84,85]. Articles included in our analysis did not use a checklist, such as the TIDiER checklist, which has been widely available since 2014 [4]. Further, none of the included articles speci ed a taxonomy or classi cation system, like the BCTTv1, to describe individual intervention components. Use of a reporting guideline to guide intervention descriptions would greatly strengthen not only the ndings generated from the evaluation of pediatric ED interventions, but also lead to greater ability to replicate studies and synthesize resulting evidence. Intervention delity not only assesses whether the intervention was delivered as planned, but provides important insight into how implementation varied and important barriers and facilitators to consider in future research [86].

COM-B Intervention barriers and enablers
To our knowledge, this is the rst instance of using the COM-B model to describe barriers and enablers of discharge communication interventions. Applying this model to analyze barriers and enablers has been completed in other research areas, including a systematic review regarding testing for sexual transmitted infections [9]. Our analysis mapped intervention barriers and enablers to all six domains in the COM-B model, with most categorized under the psychological capability domain. Instruction on how to perform the behaviour and information about health consequences were two of the most commonly identi ed BCTs in patient, caregiver and health care provider interventions. While these frequently used BCTs may address some of the barriers in the psychological capabilities domain, clearly they provide insu cient coverage to address all existing barriers within this domain.

Limitations
De ciencies in the reporting of implementation strategies and intervention descriptions made it challenging to identify all behaviour change content. It is possible that we may have missed some relevant intervention content during coding. While we developed coding rules to help prevent under or over-coding of BCTs, it is possible there were more behaviour change techniques associated with pediatric discharge communication interventions that we were unable to identify.

Conclusions
There is limited research identifying the active components of pediatric ED discharge communication interventions. The COM-B model and BCTTv1 provided a useful to analyse the barriers, enablers and common mechanisms used in these interventions. With only a fraction of the available BCTs used in the included studies, future research is needed to explore the utility of other BCTs to improve the effectiveness of discharge communication interventions. Further, the BCW presents an opportunity to prospectively design effective discharge communication interventions that incorporate prioritized BCTs using a systematic, theoretical approach. At a minimum, it is critical that intervention descriptions be standardized to ensure consistency in reporting, the ability to discern and evaluate successful intervention components, and to improve replicability moving forward.