Barriers and enablers of recognition and response to deteriorating patients in the acute hospital setting: A theory- driven interview study using the Theoretical Domains Framework

Aim: To explore barriers and enablers of recognition and response to signs of patient deterioration by nursing staff in an acute hospital. Design: A theory- driven interview study underpinned by the Theoretical Domains Framework of behaviour change. Methods: Between 07/01/2019 and 18/12/2019 a purposive sample of registered nurses and healthcare assistants was recruited to participate in a semi- structured (audio- recorded) interview, to explore the determinants of seven specified behaviours of the afferent limb. Anonymised transcripts were deductively coded (using the 14 Theoretical Domains Framework domains as coding categories) and then extracts within each domain were inductively analysed to synthesise belief statements and themes. Prioritisation criteria from published literature were applied. Results: Thirty- two semi- structured interviews were conducted. From 1,888 quotes, 184 belief statements and 66 themes were synthesised. One hundred and forty- six belief statements, represented by 58 themes, met prioritisation criteria. Nine domains of the Theoretical Domains Framework were of high importance: Knowledge ; Social , Professional Role and Identity ; Beliefs about Consequences ; Reinforcement ; Intentions ; Goals ; Memory , Attention and Decision Processes ; Environment , Context and Resources and Social Influences . Conclusions: Barriers and enablers most likely to impact on nursing staff afferent limb behaviour were identified in nine domains of the Theoretical Domains Framework.


| INTRODUC TI ON
Hospitalised patients who deteriorate in a ward setting without recognition or an appropriate response are at risk of a serious adverse event (SAE) such as unplanned admission to the Intensive Care Unit (ICU), cardiac arrest or death (Tirkkonen et al., 2013;Trinkle & Flabouris, 2011). To optimise responses to deteriorating patients, rapid response systems (RRS) have been implemented internationally within acute hospitals (DeVita et al., 2006). While RRS broadly include an 'afferent limb' (the recognition arm) and an 'efferent limb' (the response arm; Figure 1), how RRS are implemented varies across providers (DeVita et al., 2006;Lyons et al., 2018).
Deleterious changes to vital signs (e.g. heart rate, respiratory rate, blood pressure) are frequently seen in patients preceding a SAE (Andersen et al., 2016;Kause et al., 2004). Consequently, track-andtrigger tools have been implemented as part of the afferent limb of the RRS. These tools, which may be paper based or electronic, allow healthcare professionals (typically nursing staff) to record routinely measured vital signs, providing a signal when the vital signs fall outside of acceptable parameters. In these circumstances, staff are prompted to increase the frequency of subsequent monitoring and to consult a practitioner with expertise in the management of acute/ critical illness (Grant, 2018). In some regions, including Australasia and North America, track-and-trigger tools typically include dichotomous criteria, that is, when any vital sign crosses a specific threshold (e.g. a respiratory rate >30 or <10 breaths/min) the patient is considered to be at risk and care should be escalated (Davies et al., 2014;Sprogis et al., 2017). Within the United Kingdom and parts of Europe, early warning scores are more common, particularly the National Early Warning Score (NEWS), which was developed to standardise practice between organisations (Royal College of Physicians, 2017). The NEWS signals patient risk based on six routinely recorded vital signs, each of which accrues a score (range 0-3) that is combined to produce the aggregate NEWS (range 0-20). The higher the aggregate score, the greater the risk to the patient and the more senior the practitioner to whom care should be escalated (Royal College of Physicians, 2017; supplementary file 1). The use of early warning scores and an accompanying escalation protocol are associated with improved patient outcomes (Credland et al., 2020).
Like the track-and-trigger tools themselves, the nomenclature and composition of efferent limb response teams also differ internationally, with nurse-led Critical Care Outreach Teams in the UK and more medically driven or multi-disciplinary Medical Emergency Teams and Rapid Response Teams common in Impact Rapid response systems have been implemented internationally including an afferent and efferent limb. Behaviours of the afferent limb include monitoring vital signs and escalating care.
Despite global uptake of rapid response systems, there is evidence that nursing staff do not consistently enact afferent limb behaviours according to policy (afferent limb failure). New insights into the complex and pervasive problem of afferent limb failure have been offered. Use of theory will permit mapping of these identified domains of high importance to precisely targeted behavioural intervention strategies, and subsequent evaluation of how these strategies may best be operationalised in different clinical settings.  (Churpek et al., 2017;Hughes et al., 2014;Priestley et al., 2004). Actions common to all of these response teams include patient assessment, initiation of definitive treatment or supportive care and facilitation of transfer to a higher-care setting, for example, an ICU (Bannard- Smith et al., 2016). In the current research, the focus is on the behaviours of the afferent limb.
Review findings suggest that escalation to a designated response team is associated with reduced in-hospital cardiac arrest and mortality (Maharaj et al., 2015;Rocha et al., 2018). However, patients will only benefit from the additional expertise provided by these teams if they are activated and mobilised to the patient's location (Lyons et al., 2018). Consequently, patient benefit is contingent on the precursory afferent limb behaviours of the RRS being enacted.
Despite the widespread implementation of RRS and availabilty of track-and-trigger tools, there is evidence that nursing staff do not consistently follow guidance (Credland et al., 2018). This lack of compliance has been termed 'afferent limb failure' (ALF; Johnston et al., 2014;Trinkle & Flabouris, 2011).

| BACKG ROU N D
The determinants (i.e. barriers and enablers) of nursing staff enacting best practice behaviours of the afferent limb have been broadly described in a number of published review papers (Massey et al., 2017;Olsen et al., 2019;Treacy & Stayt, 2019;Wood et al., 2019).
Despite acknowledgement that ALF is a problem characterised by inconsistent staff behaviour (Credland et al., 2018;Ede et al., 2019), no reports of studies were found where behaviour change theory had been applied to explore determinants. Furthermore, from the modest body of literature reporting interventions to address ALF (Bucknall et al., 2017;Connell et al., 2016;Duff et al., 2018;Liaw et al., 2016), no explicit reports of theory being applied during intervention development were identified. There is evidence that using theory to elucidate determinants and drive the selection of intervention content increases efficacy (Noar et al., 2007;Taylor et al., 2012;Webb et al., 2010) and replicability (Little et al., 2015;Michie et al., 2008) of the resultant intervention compared to pragmatic (i.e. intuition based) or non-theoretical approaches.
Theories of behaviour change attempt to understand the context in which desirable behaviours occur (or do not occur) as well as mechanisms of action and moderators of change along various causal pathways (Michie et al., 2016). There are numerous theories of behaviour and behaviour change available (Davis et al., 2015) making the selection of a suitable theory challenging for non-specialists . The Theoretical Domains Framework (TDF) was developed to overcome this challenge by identifying a parsimonious set of broad theoretical domains drawn from behavioural theories (Cane et al., 2012;Michie et al., 2005). The revised TDF (v2) specifies 14 theoretical domains ( Figure 2) that each represent between 3 and 11 conceptually related constructs. The 84 constructs of the TDF were obtained from 33 different behaviour change theories (Atkins et al., 2017;Holdsworth et al., 2015). In addition to the accessibility of the framework, benefits of the TDF include its versatility, enabling its application to a range of behavioural problems and F I G U R E 2 The domains of the Theoretical Domains Framework (TDF). Taken from: Atkins et al. (2017) TDF domain Content of the domain 1. Knowledge An awareness of the existence of something 2. Skills An ability or proficiency acquired through practice 3. Social/Professional role and identity A coherent set of behaviours and displayed personal qualities of anindividual in a social or work setting 4. Beliefs about Capabilities Acceptance of the truth, reality or validity about an ability, talent or facility that a person can put to constructive use 5. Optimism The confidence that things will happen for the best or that desired goals will be attained 6. Beliefs about Consequences Acceptance of the truth, reality, or validity about outcomes of a behaviour in a given situation 7. Reinforcement Increasing the probability of a response by arranging a dependent relationship, or contingency, between the response and a given stimulus 8. Intentions A conscious decision to perform a behaviour or a resolve to act in a certain way 9. Goals Mental representations of outcomes or end states that an individual wants to achieve 10. Memory, Attention and Decision Processes The ability to retain information, focus selectively on aspects of the environment and choose between two or more alternatives 11. Environment, Context and Resources Any circumstance of a person's situation or environment that discourages or encourages the development of skills and abilities, independence, social competence and adaptive behaviour 12. Social Influences Those interpersonal processes that can cause individuals to change their thoughts, feelings, or behaviour 13. Emotion A complex reaction pattern, involving experiential, behavioural, and physiological elements, by which the individual attempts to deal with a personally significant matter or event 14. Behavioural Regulation Anything aimed at managing or changing objectively observed or measured actions extensive coverage of the determinants of behaviour change (Atkins et al., 2017;French et al., 2012).

| Study aim
The aim of this interview study was to explore determinants (barriers and enablers) of recognition and response to signs of patient deterioration by nursing staff in an acute hospital. Specific objectives were as follows:

| Design
This was a qualitative semi-structured interview study informed by the TDF.

| Sample
A purposive sample based on seniority (employment grade or role) and experience (duration of time in role) of nursing staff was recruited from two acute floors (four wards) within a UK metropolitan teaching hospital. In the UK context, unregistered Healthcare Assistants (HCAs) are frequently involved in enacting behaviours of the afferent limb, particularly the monitoring of patients' vital signs (Ede et al., 2019;Smith et al., 2020). Consequently, both registered nurses (RNs) and HCAs were recruited.
In 2018, it was confirmed that the hospital would be switching from paper-based patient records to Electronic Health Records (EHR). Part of this process was migration from a paper-based NEWS chart to an electronic version. It was identified that this period of transition would provide a unique opportunity to explore determinants of afferent limb behaviour in both paper and EHR contexts.
Consequently, participants were recruited pre-and post-EHR ac- tivation. An acclimation period of 3 months (Bedoya et al., 2019) was allowed following EHR implementation when no data were collected.

| Data collection
TDF topic guides (supplementary file 2) were developed to explore the determinants of seven specific behaviours of the afferent limb (referred to hereafter as the target behaviours). The target behaviours (Table 1) were shortlisted from a longer list of behaviours identified through an extensive period of focused ethnography in an earlier phase of this programme of work (Smith et al., 2020). A minimum of one question for each of the 14 TDF domains was included. Interviews were carried out by a single researcher [DS], in a room adjacent to, or away from, the ward in which the participants worked. Interviews were audio-recorded, transcribed verbatim, checked for accuracy and anonymised. DS, a clinical-academic nurse with 11 years of experience of working in critical care outreach roles, received specific training on in-depth/complex interviewing prior to data collection.
As the pre-EHR period of data collection was finite (a period of 3 months), sampling continued until the EHR was implemented. In the post-EHR period (an indefinite period), sampling continued until the point of theoretical saturation which was determined as follows: (1) an initial analysis sample of 10 interviews was conducted with nursing staff; (2) data from the initial analysis sample was deductively coded (into the 14 TDF domains) and within each domain the text inductively analysed; (3) a stopping criterion of three was used, meaning that theoretical saturation was achieved when no new themes (synthesised from inductive analysis of coded data) were identified from three subsequent consecutive interviews (Francis et al., 2010).

| Ethical considerations
Permission to conduct this research was granted by the National Health

| Data analysis
First, using Framework method (Gale et al., 2013), interview data were systematically and deductively indexed and charted using the 14 TDF domains as the coding categories (Cane et al., 2012).
Second, inductive content analysis (Elo et al., 2014)  That is, beliefs about behaviour influence whether the behaviour is performed or not, and how consistently. Therefore, while domain-level data are typically used in TDF studies to select intervention content (Cadogan et al., 2015;Patton et al., 2018), belieflevel data are required in order to prioritise the most important determinants.
The approach used [by DS] to synthesise participants' beliefs was as follows: quotes from each of the charts developed during deductive coding were read and re-read to ensure familiarisation; quotes reflecting similar beliefs were grouped and categorised using a simple label (i.e. a brief description of content); quotes were scrutinised further and 'belief statements' were synthesised to represent beliefs held by (a minimum of two) participants (e.g. RNs and HCAs believe that their professional responsibility ends/does not end, when the next clinician along the escalation pathway is notified; McBain et al., 2016;Roberts et al., 2017)). Where participant beliefs were discordant, that is, a barrier for some while an enabler for others, this was reflected in the wording of the statement (see bold text in example above). Belief statements representing overlapping or related content were grouped and a suitable theme heading synthesised Presseau et al., 2017). In this study, theme-level data were used to establish theoretical saturation in keeping with reported methods (Francis et al., 2010).
To identify TDF domains of particular importance; first, four criteria (Table 2) with binary assessments were selected from the TDF literature (Atkins et al., 2017;Goddard et al., 2018;Islam et al., 2012;Patey et al., 2012) and applied at belief statement level. Second, these criteria were used to categorise the TDF domains as being of high, moderate or low importance based on the number of criteria met. Domains with any belief statement that meet 3 or 4 of the criteria were considered of high importance; 2 criteria of moderate importance and 1 or 0 criteria of low importance (Goddard et al., 2018).

| Rigour
To achieve trustworthy interpretation of the data, a number of recommended methods were applied. To ensure credibility of findings, audio-recordings of two pilot interviews were listened to by [DS] and other researchers not involved in the study, permitting selfreflection and peer debrief on both topic guide content and questioning approaches (Morse, 2015). Reflection continued throughout the period of data collection and analysis, facilitated by the recording of reflexive notes and regular debrief with other members of the research team (Forero et al., 2018;Koch & Harrington, 1998).
The interview data were collected over a period of 8 months as part of a bigger study that included direct observation of ward staff in situ (Smith et al., 2020) and brief, unrecorded interviews.
This prolonged and varied engagement with participants increases the credibility of our data set (Forero et al., 2018); while the use of methodological triangulation increases confirmability (Morse, 2015).
To enhance dependability of the data, a codebook (supplementary file 3) was developed to ensure a clear audit trail and to enable reliable coding (Forero et al., 2018;Morse, 2015

Criterion Description
Frequency a The belief (a barrier or an enabler) was reported by more than a third of the sample

Personal importance
The belief was expressed using emphatic language in one or more illustrative quote/s Direction of effect There were discordant views between participants about the belief operating as a barrier or enabler Professional discordance The belief was held by RNs but not by HCAs or vice versa a Frequency, in this context, relates to the number of different participants who express a belief rather than the number of times it is mentioned.

| Goals
Goals related to the measurement of vital signs were often described by participants as being of higher priority than other clinical tasks.

| Environment, context and resources
Numerous participants reported a mismatch between human resource (i.e. the number of nursing staff on duty) and patient dependency as a barrier to staff reviewing NEWS charts or taking further timely measurements of vital signs. Similarly, some RNs and HCAs believed that they often did not have sufficient time to undertake these actions during the course of the shift due to unpredictable nature of the working environment.

I think I can do them [vital signs] if I want them.
It's just that we don't have the time. Generally, to go around and do six obs and six sets of meds and six discharges or whatever you're doing that day is quite a heavy job. I think, in some wards, RNs do it and think it's great, but I think our ward is just too busy. (RN10) When escalating to practitioners external to the ward-based team (e.g. medical staff or CCOT), using the hospital's pager system, some participants reported experiencing long delays before a response, particularly when enacting this behaviour at night. were underpinned by belief statements that met all four of the prioritisation criteria. In a further four domains (Knowledge; Intentions; Goals; Social Influences) underlying belief statements met three of the prioritisation criteria. These nine domains of the TDF were identified as highly important determinants of RNs and HCAs enacting seven specified target behaviours of the afferent limb of the RRS.
In our research, decisions to 'normalise the abnormal' and tolerate elevated NEWS were reported by both RNs and HCAs. These decisions were typically informed by the patient's medical history and how persistent the abnormality appeared, that is, how long the NEWS had been elevated. There is currently a paucity of research related to the adjustment of escalation criteria for deteriorating patients. In a retrospective cohort study conducted in Australia (Ganju et al., 2019), modifications in calling criteria were found to be relatively frequent (63% of the patients had modified criteria) but did not reduce the number of rapid response activations. Furthermore, an increased mortality was reported in patients who had modified calling criteria, specifically where the adjustments resulted in a more conservative approach than the standard guidance (Ganju et al., 2019). In a more recent study (also conducted in Australia) similar findings were reported, whereby patients with adjusted criteria more frequently triggered an efferent limb activation, more frequently had a cardiac arrest and more frequently died in hospital compared to patients with standard (i.e. unmodified) criteria (Crouch et al., 2020). These findings highlight the potential vulnerability of the sub-group of patients likely to have response criteria modified, the potential safety implications of reducing the level of response and the complexity of the clinical decisions that underpin these adjustments. Consequently, it is concerning that nursing staff in our study reported making individualised adjustments to response thresholds without consulting senior personnel. It is plausible that there may be a 'spill over' effect (Michie et al., 2014) (Nursing & Midwifery Council, 2015). Notwithstanding the difference in context, similar discrepancies were reported in a qualitative study conducted in Singapore (Chua et al., 2019). Here, the researchers reported inadequate direction and supervision of Enrolled nurses (licensed practitioners who are educated at a lower level than a RN) by RNs, when vital signs were being monitored (Chua et al., 2019).
The belief that RNs do not need to delegate to junior colleagues is particularly problematic given the reported association between poor delegation and aspects of nursing care being delayed or missed entirely (Kalisch, 2006). On this basis, we echo the suggestions in other work (Chua et al., 2019;Kalisch, 2006), and recommend that attention be given to raising the importance of delegation as a safety critical aspect of the RN role. We also encourage educators to equip registrants with the necessary communication and leadership skills to delegate care effectively in increasingly complex clinical environments.

| Limitations
The semi-structured interviews that were conducted generated a large volume of data. Consequently, it was necessary to identify TDF domains with barriers and enablers most likely to impact on nursing staff afferent limb behaviour (i.e. the domains of high importance) to subsequently target for change. We used a number of reported criteria to identify the TDF domains of importance. However, there is currently no evidence that using these criteria will result in a more successful intervention (Goddard et al., 2018) reflecting a broader limitation of the methods employed.
The potential for social desirability bias is a limitation of our study. Participants were made aware that the researcher was a clinician with a background in acute/critical care, and a specific interest in the recognition and response to deteriorating patients. As such, it is plausible that participants may have answered questions in a way that they perceived would please the researcher. Interviews were conducted over a period of 8 months, during which time the researcher was frequently present on the clinical floors. This strong presence increased the likelihood of participants habituating to the researcher (Pope, 2005) and therefore may have mitigated the extent of social desirability bias.
A further limitation of the study is that our sample included only nursing staff and excluded all medical staff. Given the close working relationships between ward-based nurses and their medical colleagues (particularly, the primary medical teams responsible for the ward-based patients), we may have missed part of the picture by opting to focus only on nursing staff. Given the evidence that junior medical staff do not consistently escalate deteriorating patients to their senior colleagues (Callaghan et al., 2017), there is grounds for further theory-driven work to improve understanding of the determinants of medical staff behaviour in this space.

| CON CLUS ION
Through the use of structured methods, and the systematic application of theory, we identified a range of determinants (i.e. barriers and enablers) to RNs and HCAs enacting specified behaviours of the afferent limb. Consistent with other published research, we identified barriers related to lack of knowledge, fear of reprimand, high workload and lack of physical resources needed to enact these behaviours. We offer new insights into barriers and enablers relating to motivation and intentions, goals, professional role and responsibility, decision-making processes, social interactions with peers and colleagues and feedback received or anticipated from more senior staff. Having reported the TDF domains that appear to influence the target behaviours, further work is required to map determinants to intervention content as part of the development of a tailored behaviour change intervention (Baker et al., 2015).

ACK N OWLED G EM ENTS
The authors would like to acknowledge Professor Jill J. Francis, who played an integral role in designing the study within which this research is situated.

CO N FLI C T O F I NTE R E S T
No conflict of interest has been declared by the author(s).

AUTH O R CO NTR I B UTI O N S
All authors have agreed on the final version and meet at least one of the following criteria (recommended by the ICMJE*): (1) substantial contributions to conception and design, acquisition of data or analysis and interpretation of data; (2) drafting the article or revising it critically for important intellectual content. * http://www.icmje.org/ recom menda tions/

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/jan.14830.

T WIT TER
Duncan Smith @duncan281013