Implementing a toolkit for the prevention, management and control of carbapenemase-producing Enterobacteriaceae in English acute hospitals trusts: a qualitative evaluation

The Behaviour Change Wheel (BCW), a comprehensive framework for developing and evaluating behaviour change interventions, was used to theoretically inform the evaluation [16]. It describes seven broad policy level options through which behaviour can be influenced: fiscal measures, regulations, service provision, legislation, communication and marketing, environmental and social planning, and guidelines. To target capabilities, opportunities and motivation to perform behaviours (COM-B model), policy interventions support specific intervention functions, for example education, persuasion, incentivisation, coercion, training, restriction and environmental restructuring. Mapping onto the COM-B model, the Theoretical Domains Framework (TDF) breaks down the capability, opportunity and motivation domains further, outlining 14 areas of influence on professional behaviour, for example improving skills, changing beliefs about consequences or social influences [17, 18]. We examined which influences were relevant to the CPE toolkit implementation and what behaviour change pathways applied according to frontline staff. The evaluation built on empirical findings from a 2016 online survey of executive acute trust staff on organisational level perceptions of the CPE toolkit implementation, which suggested awareness of the CPE toolkit, but a lack of confidence in its practical value [19]. Additional file 1 illustrates how the evaluation frameworks informed the following research process.

A qualitative design was used to gain a contextual understanding of CPE awareness amongst frontline staff and their experiences of the CPE toolkit implementation [14, 20, 21]. The evaluation protocol was reviewed by the University of Bristol Research Ethics Committee, classifying this health service evaluation as an audit of existing NHS procedures [22].

The evaluation was conducted in acute hospital trusts in England, the setting for which the CPE toolkit was developed. At the time of this research, there were 154 English acute trusts. Trusts that had given their consent to be approached for further evaluation activities (59 of 99 respondents) in the prior online survey [19] were purposively sampled to ensure a diverse sample [23‐25]. The sampling frame focused on capturing all potential combinations of two main sampling criteria, CPE colonisation rates and time point of introducing local CPE plans, and on including large, medium and small trusts from each commissioning region. All trusts had local CPE plans in place, but the time point of initial introduction varied from prior to the CPE toolkit publication, as early as 2014, to late adoptions (2015/16). CPE colonisation, a precursor for infection [26], correlated with CPE infection rates (χ² = 116.49; p < 0.001) and ranged from no cases, to some (1–10), and many CPE colonisation rates (> = 11) [19].

A personal access strategy [27, 28] was employed, gaining access through gatekeepers such as the local infection prevention and control teams (IPCT) and antimicrobial resistance (AMR) leads or through executive members of staff that participated in the survey. Following a maximum variation sampling approach to explore different views and experiences and to maximise the chance of discovering differences between staff with a range of roles and responsibilities [29‐31], gatekeepers identified interview participants from two staff groups targeted by the CPE toolkit: 1) staff with specialist IPC responsibilities (e.g. IPC directors, lead IPC nurses or consultant microbiologists) as key informants [32], hereafter called IPC frontline staff, and 2) non-IPC frontline staff (e.g. staff nurses, ward managers and doctors) caring for patients, hereafter called general frontline staff. Settings included low CPE risk but high patient throughput wards, such as admission and acute medicine, as well as specialised, high CPE risk wards, like intensive care and surgical wards. Depending on trust size and speciality, the number of staff available for interviews varied, ranging from two to five participants per trust (Additional file 2) with at least one person from each group (Table 1). Out of the 14 trusts approached, 12 agreed to participate and 44 interviews were conducted (Table 2). Inability to spare staff time was given as reason for non-participation.

An interview topic guide (Additional file 3) was developed by the multidisciplinary research team, piloted with participants at two acute trusts and iteratively adapted over the course of the interviews as new insights emerged [21, 33, 34]. Semi-structured interviews ensured detailed information relevant to the evaluation questions were captured while allowing participants to raise additional topics of importance [24, 25]. Participants were asked to reflect on any ‘significant changes’ [35, 36] concerning CPE prevention, management and control since 2014 when the CPE toolkit was launched nationally. Questioning not referring explicitly to the CPE toolkit initially was aimed at identifying changes in the wider context and how much importance participants ascribed to the toolkit, if they were aware of it. Aspects related to potential behaviour change pathways and context factors were explored with questions based on the TDF [18]. As visual stimulus to prompt further insight into participants’ experiences and views on recommended procedures [14, 37], participants were asked to comment on the patient management steps specified in a flowchart included in the CPE toolkit. Interviews were conducted in accordance with data protection and confidentiality regulations, by briefing participants and obtaining informed consent. As this evaluation was commissioned by the same group at PHE who commissioned the CPE toolkit, interviewers emphasised their independence of the evaluation group to counter response biases. Depending on participants’ preferences and availability, interviews were conducted between April and August 2017 either face-to-face during one-day site visits or by telephone over a maximum period of one month (Table 1). Both interviewers (AS, CC) were women with a PhD in health science (health psychology, epidemiology) with prior qualitative research experience. Interview times ranged from nine to 108 min with an average length of 26 min. The average interview length was similar for face-to-face (27 min) and telephone interviews (24 min).

The interviews were digitally recorded on an encrypted recording device and transcribed verbatim. The transcripts were imported into NVivo Version 12 [38] and analysed following thematic analysis steps [39] using the constant comparison technique, coding the material line-by-line [40]. To ensure reliability and validity of the analysis [21, 30, 41], the first five transcripts were independently double-coded (AS, CC). Potential discrepancies were discussed to reach consensus and a coding framework was developed using a hybrid approach [42], combining data-driven inductive codes and deductively developing categories based on the Medical Research Council process evaluation guidelines [12] and procedures outlined by the CPE toolkit [10]. The first author applied the agreed coding framework to the remaining transcripts. Themes were converted into thematic networks [43], mapping implementation barriers and facilitators according to COM-B model categories for behavioural influences [16]. Further procedures employed to ensure quality of the data analysis included discussions of findings within the multidisciplinary team, memo writing and keeping reflective and decision logs throughout the data collection and analysis phase, grounding findings in the data through anonymised participant quotes and transparent reporting in line with recognised standards [25, 44].

The extent to which frontline staff came into direct contact with the published CPE toolkit varied by staff group. All interviewed IPC frontline staff were aware of the CPE toolkit. This was rarely the case amongst general frontline staff (10%) and the few who recalled seeing it, acknowledged that they had not read it. General frontline staff, which referred directly to the CPE toolkit, usually did so in the context of CPE outbreaks. IPC frontline staff often could not remember how they initially heard about the CPE toolkit. If they did, this was typically through their professional networks and PHE mailing lists.

All trusts had local CPE plans in place and participants described CPE prevention, management and control procedures that aligned with the CPE toolkit guidance as much as possible considering local circumstances. Figure 1 provides an illustration of the role of the CPE toolkit in the typical development and implementation process of local CPE plans.

The process started with encountering CPE, either within their trust or becoming aware that it was a problem at other, particularly neighbouring trusts. CPE proximity seemed crucial in influencing the implementation process by raising awareness of CPE as a problem (82%). In response, IPC frontline staff would use the CPE toolkit to inform trust practice and translate it into local procedures and policies, making amendments to accommodate their circumstances and demands. Examples of local adaptations were enhancement of CPE precautions for high CPE colonisation rates or CPE outbreaks and large-scale patient cohorting at CPE isolation wards in response to insufficient isolation facilities. The three trusts that had local plans in place before the CPE toolkit was published reported their own hospital CPE outbreaks as triggers for developing CPE policies in the absence of national guidance. When the CPE toolkit was published, they referred to it to check if their policies aligned with recommended procedures, but did not necessarily find them useful.

The process of developing local CPE plans was often described as an interdisciplinary effort. Considering multiple perspectives helped to engage different staff groups and aimed to generate policies that were relevant, workable and acceptable for users. Communication channels for local CPE plans varied, including word of mouth, meetings and training sessions, intranet communications and written information leaflets and checklists. There was awareness that the implementation of procedures recommended in the CPE toolkit or locally-relevant adaptations promoted a general confidence in CPE prevention, management and control capabilities. This, however, did not necessarily co-exist with an in-depth knowledge of CPE, even in hospitals that experienced outbreaks.

The described process (Fig. 1) was iterative in nature without an ultimate endpoint, as responding to implementation experiences and situational factors (e.g. CPE outbreaks, availability of new evidence) local plans and procedures continued to change. Overall, developing local CPE plans was seen as less of a challenge than their implementation.

Table 4 provides an overview of factors mentioned by participants that acted as facilitators and/or barriers to implementing recommended CPE prevention, management and control procedures. They are grouped under the most relevant COM-B model component and illustrated by verbatim quotations. Factors influencing implementation often interacted as staff capabilities (e.g. knowledge and skills) and opportunities (e.g. social influences and available resources) had an impact on their motivation (e.g. beliefs and intentions) to engage in target behaviours.

Due to the novelty of the topic, a lack of awareness and knowledge especially amongst general frontline staff prevailed. Expert CPE knowledge went hand-in-hand with concerns about resulting risks and increased efforts to deal with CPE. Uncertainties about CPE risk factors and its management as well as receiving too much and too detailed information were problematic. Implementation was made easier, if new processes such as screening procedures were added to already existing standard operational procedures and routines. One way of doing this was making CPE part of a wider IPC plan. This however posed the risk of not paying enough attention to CPE but prioritising other more pressing issues.

Contextual factors, including resources and social influences, were overall key to implementation. Participants recognised multidisciplinary, trust-wide and across-trusts efforts and knowledge sharing as facilitators, even if these slowed down the process since involving various stakeholders took time. This approach promoted communication of relevant patient information and helped to creatively deal with important barriers, such as insufficient numbers of isolation facilities. A problem mentioned was communication and collaboration with community care providers, where anxieties and a lack of CPE knowledge hampered the discharge of colonised patients.

Further facilitative social opportunity factors included the presence of dedicated IPC staff and support from supervisors, modelling good practice, and trust management support. The CPE toolkit was seen as a credible source of information, which could help to secure management support for financial investments to address resource challenges such as a shortage of isolation facilities or funding for additional training. Comparing their own situation to other trusts seemed to encourage CPE prevention and control efforts as CPE outbreaks at other trusts highlighted the seriousness of the problem; some participants mentioned a desire to outperform other trusts. Patients’ lack of awareness and refusal to participate in recommended procedures, particularly refusing rectal swabs, was a barrier reported by some trusts but could be addressed with communication and reassurance.

Encountering CPE increased staff motivation to develop local plans and follow recommended procedures and led to raised awareness and CPE knowledge. On the other hand, if CPE was already a problem, procedures outlined in the CPE toolkit were often not seen as adequate or practicable, forcing staff to find new ways of managing the problem, including cohorting whole patient groups. Older hospital buildings were a barrier to ensuring appropriate cleaning of facilities and isolation. Staff shortages, time pressures and competing IPC demands complicated matters further. Training members of staff for example, the most frequently named dissemination strategy for local CPE plans, was difficult due to high staff turnover and work pressures. There was increased anxiety in the context of resource shortages as this increased the challenge of managing potential CPE cases. Being extremely concerned and anxious about CPE seemed to evoke helplessness and inaction.

Optimism and confidence in the capability to carry out recommended procedures was both a barrier and facilitator. A certain level was required for successful implementation, for example when carrying out rectal swabs. An over-confidence, assuming current practice based on clinical expertise and experience sufficed, however, prevented staff from acquiring further knowledge and skills. Some staff mentioned that the CPE toolkit helped reduce concerns and improved beliefs in their capacity to manage outbreaks as it provided clear guidance. Continuous improvement efforts, e.g. using monitoring, audit or feedback, helped adapting patient management procedures. IPC nurses typically took charge as they felt responsible for and took ownership of the issue and for passing on relevant knowledge. Overall, it was facilitative to change if staff perceived following recommended procedures and dealing with CPE as an important part of their role and identity.

Participants, who were familiar with the CPE toolkit, often could not remember their initial thoughts or recall its specific content, but reported views were mixed (Table 3). Some thought it was comprehensive and useful and much needed at the time, especially the flowchart, while others perceived the recommended procedures as impractical to implement locally. Participants criticised the CPE toolkit for not being specific enough, sometimes acknowledging this could be due to lack of a clear evidence base. The lack of scientific evidence led others to question the credibility of the information provided and to voice distrust in or confusion around the recommended procedures. While the toolkit did not include enough information for some, other participants criticised it for being too long, making it inaccessible. There was also some uncertainty about the CPE toolkit target group.

The most commonly voiced improvement suggestion was that the current version of the CPE toolkit should be updated regularly (23%). Participants were keen that an updated version would include new evidence and further clarifications on the difference between CPE and Carbapenemase-producing Organism (CPO), risk factors and screening procedures, including different implementation options to fit clinical practice. One way to address concerns around the evidence base of the CPE toolkit and strengthen its credibility could be linking it to related guidance and research publications or providing contact details of experts for further queries. A better fit for clinical practice might be achieved building on previous involvement of experts “on the ground”.

Participants felt it needed to be more explicitly highlighted that the CPE toolkit encouraged local adaptations and to provide tailored information to meet the needs of different staff groups and types of trusts, for example trusts with low and high colonisation rates. To improve accessibility and reduce information overload for general frontline staff, an abbreviated version was suggested. Participants also expressed a desire for further visualisations, particularly in the form of graphs and figures concerning epidemiological information. Those figures would not necessarily need to be provided in writing, but could be accessible online and should again be updated regularly. Table 5 summarises recommendations for improving the CPE toolkit based on participants’ suggestions and outlined implementation facilitators and barriers.