The demonstration of a theory-based approach to the design of localized patient safety interventions

Patient safety is a global priority. An estimated 3.7% to 17.7% of patients are inadvertently harmed as a result of hospital care [1‐3]. Some harm is caused by healthcare professional error arising from factors such as poor system and equipment design, and high workload [4, 5]. Other harm results from deviations from guidelines and policies; only between 50 and 70% of patients receive recommended care [6, 7]. Interventions to change professional behavior have modest and variable effects [8]. This variability is problematic because it limits the ability to predict with any confidence whether an intervention will work for a given problem and context [9]. The reported modest effects may be the result of problems with the ways by which interventions are selected. First, selection may be based more upon habit and disciplinary perspective than an explicit rationale that takes targeted behavior and context into account [10]. Second, the theory underpinning healthcare professional behavior change interventions is seldom explicated, thereby limiting the ability to generalize from one context to another [11]. Third, standardized 'top down’ interventions may lack flexibility to respond to local barriers and circumstances [12, 13].

Behavior change occurs within complex social and environmental systems that demonstrate local variations [14]. It could be argued then that interventions to improve patient safety will be most effective when developed by those with local 'expertise’ and tacit knowledge [15‐17], but which take account of evidence and external expertise. This paper describes how we co-designed interventions with local stakeholders, guided by both behavior change (impact) theory and implementation (process) theory [18].

We selected the theoretical domains framework (TDF) of behavior change [19] as the 'impact’ theoretical approach for this work. The TDF was developed to rationalize and reconceptualize the theoretical constructs from multiple psychological and organizational theories of behavior and behavior change using an expert consensus and validation process. The resulting framework includes the description of the nature of the behavior to be targeted and eleven domains of behavior change: knowledge, skills, social/professional role and identity, beliefs about capabilities, beliefs about consequences, motivation and goals, memory attention and decision processes, environmental context and resources, social influences, emotion, and behavioral regulation. The framework has been used to understand barriers and levers to change in a range of contexts c.f, [20‐26]. We also drew upon principles derived from implementation ('process’) theories [19], which provide insights into the necessary conditions for optimizing adoption and spread of interventions in practice (Table 1). This combined approach will be referred to as the Theoretical Domains Framework Implementation (TDFI) approach.

The TDF has been used for designing interventions to change clinician behavior in primary care [41], but not in an acute setting. We describe how the TDF was operationalized using co-production and implementation principles, and applied to three UK hospitals to improve the implementation of a patient safety guideline promoting safe nasogastric feeding. We addressed two questions: How important is local context in identifying barriers and appropriate interventions to implement safety guidance? How feasible and acceptable is the TDFI approach for implementing patient safety guidance?

We have demonstrated a process for developing and implementing theoretically derived, co-designed and context-specific interventions in healthcare organizations. This is the first study to use the TDF to directly inform intervention design in an acute setting, and to outline how specific implementation principles can facilitate the use of the TDF for eliciting behavior change in healthcare settings. This study highlights the importance of local context in identifying barriers and designing and implementing appropriate interventions, and the feasibility and acceptability of the TDFI approach.

There were differences in the key barriers reported among the three organizations, suggesting that local context might have affected perceptions about the challenges faced in complying with a patient safety guideline. We found significant differences between hospitals for three of the domains (knowledge, skills, and emotion); however, there was overlap in the top three barriers identified. For all hospitals, the mean scores for barriers were low, despite poor compliance with recommendations in practice—this could represent a tendency for staff to underestimate the barriers to behavior change, or to respond in a socially desirable way; nonetheless, the relative values between each domain within each organization demonstrate differences in perceptions of barriers, which is important for the purpose of tailoring interventions.

Although some key barriers identified were the same across each Trust (e.g., social influences, environmental context and resources), the detail provided about these barriers in focus groups varied somewhat (e.g., for 'environmental context and resources,’ some staff referred to a lack of adequate documentation, others mentioned how they were unable to easily access pH paper), as did the interventions implemented to overcome these barriers (e.g., to combat the pH paper issue, H2 designed an NG pack that included all the equipment, and H1 worked very closely with procurement to ensure the correct pH strips were available on every ward). The time and resources available to dedicate to this area of improvement in each organization also affected the type of interventions implemented and the degree to which they were developed. For instance, H2 generated an interactive e-learning resource which included video clips of medical staff modelling the entire NG tube procedure—this was achieved due to a dedicated implementation team, who formed working relationships with non-clinical departments (e.g., IT, medical illustrations). In H3, however, IT services were limited and, as such, generating an e-learning package was not possible—instead, to encourage sharing, staff were directed to the H2 e-learning package through posters and screensavers implemented within H3. These examples demonstrate the advantages of adopting the perspective of the target group c.f, [50], using a flexible approach c.f, [39], and incorporating interventions into established structures c.f, [27, 51]. Although these implementation principles were used and can be identified in the results of the reflective log analysis (Table 8), using the approach did not come without difficulties, such as resistance to change or a perceived lack of time.

Generally, perceptions of the feasibility and acceptability of the TDFI approach appeared positive; interviewees indicated that the outcomes so far for their organization (e.g., engaging in patient safety work, spreading knowledge) were beneficial. Challenges of this approach included having to gain commitment from staff (e.g., building an implementation team, eliciting contributions from non-clinical departments). However, it is possible that these types of challenges are not exclusive to this approach, but are more common observations of work aiming to produce large scale change within a complex organization [52]. Interviewees expressed appreciation for the HIEC team support throughout the process, and recognized the benefits of holding focus groups to co-design interventions, suggesting that the blend of theoretical support and clinician context expertise [38] worked well. While this feedback suggests that it is feasible and acceptable to use the TDF with healthcare professionals to drive behavior change for patient safety, the support required to ensure teams maintained momentum was resource intensive, and this is highlighted in the results of the reflective log analysis. Irrespective of this point, the positive aspects of this approach for sustainability were evident, including the forming of intra- and inter-organization networks working together on areas of patient safety [53], and the spread of enthusiasm for patient safety work by champions [54].

Collection of time trend data was also feasible. Although we could not perform formal statistical analysis (i.e., time series analysis) given data limitations mentioned earlier, the Trust A audit suggested that 'shifts’ [40] occurred for improved documentation of practice and use of pH first line, and decreased use of X-ray first line, following intervention implementation. Furthermore, the pre-post intervention data indicated clinically significant improvements in practice for all three measurement outcomes. These improvements may be associated with the implementation of the intervention strategies, e.g. brief practical training provided for junior doctors (skills; beliefs about capabilities), the presentation of information to senior members of staff at clinical audit meetings (social influences), screensavers and posters (emotion; social influences), and the radiology system change (environmental context and resources; social influences). More formal evaluation methods are needed to establish causation [49, 55]; however, these findings indicate potential for co-designing theoretically underpinned interventions to address specific barriers to behavior change for patient safety.

There are two main limitations to our methods. First, participating hospitals were volunteers and therefore may have been more likely to complete the process. However, simply receiving agreement for participation by a medical director did not automatically lead to continued involvement from front line staff. Furthermore, audit data indicated the participating trusts were experiencing issues with guideline implementation that are similar to those faced by others. We also examined a single patient safety alert so, as yet, the extent to which it is possible to use the TDFI approach for other alerts/guidance is unknown. In addition, the number of exit interviews undertaken to understand perceptions of feasibility and acceptability was small (n = 5). Finally, the time trend data was collected only for a single Trust. Pre-post intervention implementation data is being collected for all three Trusts, which we are attempting to compare against retrospectively collected control data.

Second, identifying the successful components of this approach and of the specific interventions on behavior change will also be challenging. For example, at this early stage of development and feasibility testing, it will be difficult to understand the extent of the benefits of using the TDF as part of an approach to implement patient safety guidance in an acute setting, compared with simply providing additional support for implementation. Furthermore, given the range of interventions used to address key barriers, it will be difficult to identify which BCTs have led to change, and whether any change occurred as a result of mediating perceived barriers. Nonetheless, the interventions used have been designed using underlying theory and reported explicitly to enable replication.

The outcomes of this work include a framework for the implementation of patient safety guidelines which consists of a) a set of tools to identify context-specific target behaviors to address, barriers to improvement, and theoretically underpinned strategies to overcome barriers, alongside b) a set of implementation principles to guide use of these tools with organizations. The TDFI approach and the associated resources may be of use in other healthcare organizations and to guideline implementation teams and policy makers, especially if the post-intervention data from the three Trusts indicate changes in behavior in comparison to control sites.

It is feasible and acceptable to combine theory-driven and co-design approaches in the development of strategies to support the implementation of an evidence-based patient safety guideline. The impact of local context and value of local expertise should not be under-estimated. Future work should replicate or adapt theory-driven, co-designed interventions and evaluate their effects within rigorous designs.

NT and RL have previously worked on projects that involve using the TDF framework to identify barriers and design interventions using theoretically underpinned behavior change techniques to design tailored interventions to address key barriers for a range of health behaviors. RL conceived the original idea for using the TDF to implement NPSA guidelines. NT developed the TDFI approach. BS is an organizational psychologist who has drawn upon her knowledge and experience of implementation science to contribute to this work. RF is a general practitioner and implementation researcher, and is also the Deputy Editor of Implementation Science; all decisions on this manuscript were made by another editor.