Quantitative data collection.

Data collected to determine the fidelity, dose and reach of the intervention included:

1. Daily adherence to conducting the intervention components (multidisciplinary ward round; COMFORT assessment; screening for readiness for a spontaneous breathing trial; undertaking an SBT when screening criteria were met)
2. Numbers of staff trained in the intervention
3. Numbers of children admitted requiring IMV and numbers screened for trial eligibility

Trained research nurses collected and recorded the daily adherence data onto case report forms (CRF). Staff training data were recorded on study training logs, and admission and recruitment data were collected from the study recruitment logs, and the paediatric intensive care national database.

Quantitative data analysis.

Adherence was measured by the proportion of each intervention component performed and captured daily; the number of staff trained; and intervention reach (admissions screened divided by IMV admissions during the trial period). Due to sequential recruitment of clusters over time, we did not measure adherence over time, but provide the final mean adherence proportions for each ICU, ranked from highest to lowest. The trial’s primary outcome was the duration of IMV, measured from initiation of ventilation to the first successful extubation event. The estimate of the treatment effect was a time and cluster adjusted hazard ratio (HR) with 95% confidence intervals. The duration of IMV was analysed using a Cox proportional hazards model, with a frailty term for clustering by ICU. The time to event outcome was censored at the transition from usual care to the intervention training period, hospital discharge, 90-days, death, or if the child received a tracheostomy. We derived an absolute measure of effect by calculating the median of the model-based prediction of survival duration at all 22 periods in the stepped wedge trial and the difference between the intervention and usual care periods, and we summarised the extent of variability over the 22 periods using the interquartile range.

Qualitative data collection.

A senior ethnographer (JJ), who was not involved in the clinical trial, visited participating sites to undertake face-to-face individual and focus group interviews. All clinical staff involved in delivering the intervention were eligible for interviews. Recruitment entailed an initial email from the researcher comprising an introduction, with attached Participant Information Sheet (PIS). The email was distributed to all eligible staff ahead of site visits. From amongst the staff working on the days of site visits, a range of clinicians and grades were purposively selected for interview. Prior to the interview, participants were given an opportunity to read the PIS again and ask questions. Once satisfied that informed consent had been completed, participants were asked to sign a Consent Form.

No prior relationship between JJ and participating staff existed, and all were made aware of her impartiality and need to gain candid insight into their experiences of trial delivery. Guided by a semi-structured interview schedule, discussions explored participant understanding and experience in relation to the five areas theorised as underpinning intervention effectiveness: training; flexibility; proactivity; consistent weaning and sedation care and management; and MDT working. The interview guide was developed by JJ and the Chief Investigator (BB). It was piloted with another author (LMcI), who is a paediatric intensive care nurse. Interviews were undertaken in a quiet location close to the paediatric ICU, lasting between 30–90 minutes. They were audio-recorded and fully transcribed, with anonymisation. Only JJ had access to information that could identify individual participants during and after data collection.

Qualitative data analysis.

Data were analysed deductively and inductively. Deductively, the four components of the intervention and associated processes were used as a ‘point of entry’ into data. Inductively, Braun and Clarke’s thematic content analysis framework [15] was used to generate themes that crosscut the entirety of the qualitative dataset [S3 Appendix]. Individual and group interview data were combined [16, 17]. The same ethnographer led data analysis. To promote confirmability and trustworthiness, a 15% sample of data was sent to an independent qualitative researcher (CK) to check the analysis, identifying key differences and similarities in pursuit of an agreed final analysis [18]. Interview findings are presented according to the five proposed processes of how the intervention was envisaged to achieve the trial outcome outlined in the logic model.

Training promotes understanding and skill set.

Overall, the training was considered integral to staff understanding of the purpose and content of the intervention and the effective discharge of their respective roles and responsibilities. Training was regularly discussed as time-consuming, particularly the online component. Face-to-face training was considered more enjoyable and effective due to its practical nature and focus, including as participants could ask questions and resolve emerging issues. In contrast, the online component could be described as lacking an immediate “real-world” relevance.

Intervention acceptability.

The intervention was consistently validated as technically straightforward to use and accommodate within routines of care. Although bedside nurses were aware of increased documentation, overall, they endorsed the ease and flexibility with which COMFORT scoring and SBT readiness screening could be undertaken. That the intervention explicitly accommodated independent clinical decision-making was considered particularly important for buy-in from unit consultants, who expressed no sense of being constrained in their decision-making regarding either the conduct of SBTs or patient extubation. The same independence was highlighted by nursing staff, primarily in relation to the management of patient sedation.

Stasis in the absence of explicit direction.

Although the intervention’s strategic flexibility was widely endorsed, it was also discussed as inadvertently working against optimal intervention delivery. Three main issues were identified. First, the flexibility inherent in the COMFORT assessment, which was categorised into a range of scores indicating sedation ‘status’ for weaning. The categorisation could be problematic due to an overlap in ranges (10–12 consider weaning sedation; 12–17 comfortable and can be safely extubated). A child who scored in the range 12, or just above, could fail a SBT screen and remain longer on the ventilator. Second, the option of running a SBT for up to two hours. This period of time could allow attention to move away from a patient’s weaning progress as other priorities of patient care took precedence. It could also diminish the perceived importance of the need to extubate, as staff were aware a patient could remain on SBT settings for two hours. On the occasions when a patient remained on a SBT for longer than two hours, the end goal of extubation could be undermined in that the patient could tire and need to be returned to higher levels of support. Third, the intervention did not stipulate that extubation should follow within a set time after a successful SBT, in recognition that this might not be practically possible. However, the lack of such a requirement was linked to an indefinite postponement of extubation in some cases as other unit pressures took priority and/or relevant staff were unavailable.

Promotes proactivity in weaning.

Effectiveness of ward round planning. The morning ward round was the primary forum of decision-making regarding daily plans for individual patient care in all units. Building on this opportunity, the intervention included explicit guidance on the use of ward rounds to set targets for patient sedation and weaning. Where discussion occurred, participants gained clarity concerning plans. Conversely, deficits in ward round discussion, and thus planning, were associated with uncertainty and potential delay in patient weaning progress. Overall, accounts suggested inconsistent adherence both within and between units. Three main factors were implicated. First, that those leading the ward round ensured relevant discussion took place. Second, discussion was more likely to occur if a patient was considered to be approaching readiness for extubation. Third, weaning plans could be de-prioritised if other information took precedence within a tight time frame for discussion.

Development of an underpinning momentum towards extubation. To varying degrees, participants in all units considered adoption of the intervention to have encouraged an overall momentum towards patient extubation. Several inter-related processes were identified. First, staff were prompted to consider weaning and keep opportunities for progress at the forefront of thinking. Second, staff were enabled to advance weaning according to their respective roles and responsibilities without explicit direction. Third, inconsistency was reduced as staff who might otherwise adopt a more extended approach to weaning were discouraged from doing so. Finally, anticipatory patient preparation was encouraged in respect of what might ‘come next,’ e.g. stopping patient feeds ahead of a SBT and potential extubation.

Promotes consistent weaning related care.

Enhanced prioritisation and conduct of patient sedation. Across all professions, participants upheld the positive impact of COMFORT assessment in promoting appropriate and consistent sedation management. Consistency was linked to the shared multidisciplinary language provided by the intervention. Resulting improvements included increased attention to other factors impacting on patient comfort (e.g. pain, and improved drug regimes), which were associated with enhancing extubation readiness.

Impulse to adhere to preferred sedation practice. Inconsistencies in the performance of COMFORT assessment and attendant sedation management were associated with two issues. First, the prioritisation of other care, particularly in the context of heavy workloads and unwell patients. Second, a preference amongst bedside nurses, particularly the inexperienced, to maintain patient safety by keeping children ‘well’ sedated. The intervention itself could be implicated in over-sedation in that it could be followed ‘technically’, but still allow for a child to remain too well sedated.

Expediting patient extubation. Participants confirmed that units incorporated SBTs with daily readiness screens into their practice, indicating that both were valuable in advancing extubation. Units created times for both activities to fit with established routines. SBT readiness screens scheduled for the early morning, typically towards the end of the night shift, were described as most consistently undertaken for two reasons: night nurses had the requisite time available, and it ensured the result was available for the morning ward round discussion. Compared to SBTs scheduled for later in the day, participants confirmed that those scheduled for the morning were also more consistently performed and led to patient extubation due to better staffing resources and established routines of patient care.

Both medical and nursing staff discussed the valuable learning gained concerning the validity of the SBT readiness screen/SBT for advancing patient weaning. Overall, this involved acceptance of the value of moving away from a gradual “step-down” approach to one of weaning patients more dynamically. Thus, one of the most frequently articulated aspects of learning concerned the potential for weaning to be safely expedited by taking patients from a relatively high pressure support to a lower pressure support during the SBT.

Impulse to adhere to preferred ventilator weaning practice. Participants highlighted the persistence of established practice, resulting in inconsistent intervention adherence and delayed weaning progression. Various manifestations were described. Doctors could insist on personally assessing all patients before moving to the ‘next step’ of the intervention. They could also continue to wean patients onto progressively lower ventilator settings before extubation, rather than initiate a SBT following a successful screen. In some cases, SBTs were performed only when low ventilator settings had been reached. In such cases, rather than being used to potentially expedite extubation, the SBT was being used as a final check. Following a successful SBT, but where extubation was contraindicated, on occasions, patients were returned to full ventilator support due to consultant aversion to leaving patients on low pressure support for any length of time. Finally, inexperienced medical and nursing staff were sometimes described as prematurely ending a SBT based on initial changes in a patient’s breathing rate, which may have settled as patients became accustomed to the new mode/level of support. Where non-adherence to the protocol occurred, doctors were required to provide a rationale for their decision-making. Bedside nurses, who were responsible for recording the reasons, could express frustration that none was given or that adequate detail was missing.

Variability of consistency relating to unit routines. Participants highlighted opportunities and challenges to consistent intervention delivery afforded by the daily routines of unit activity. One challenge concerned the morning ‘ward round’. In many units, off-unit rounds took place involving the medical team and senior nurse-in-charge, with the latter communicating plans ‘back’ to bedside staff. This practice was considered to negatively affect full multidisciplinary weaning decision-making. Potentially, relevant patient information might not be communicated, and decisions could be made without input from bedside nurses. Senior nurses highlighted frustration at being unable to achieve proper trial-related discussions during off-unit handovers, meaning they had to make plans based on non-specific discussions.

Many accounts suggested a diminished impetus to undertake SBT screening and SBTs in late afternoon or evening because unit routine practice meant patients were less likely to be extubated. Even when units operated with a formal ‘24/7’ extubation policy, participants acknowledged that policy implementation was selective. For example, if patients passed a SBT later in the day, ventilator pressures were reduced, but patients often remained intubated until a decision the next morning. Thus, instead of expediting ventilator liberation, the function of later screening and SBTs was seen as providing insight into patient progress.

Patient weaning or extubation could be subordinated to other core unit activity, resulting in patients remaining on ventilation for several hours. Examples included patient investigations or clinical procedures, off-unit investigations, and staff breaks. In some units, participants described a routine of stopping feeds in the early morning (typically 4 hours) in anticipation of a successful SBT as contributing to timely extubation. Conversely, in units where this routine was absent, extubation could not immediately follow a successful SBT because the feed-stopping period had not commenced.

Intervention adherence was most strenuously questioned when it was understood to be largely redundant in terms of patient care. For example, participants talked about non-performance of SBT and screening when staff were aware that patients were clinically inappropriate for extubation. Similarly, ward round discussions of ventilator and sedation targets more likely occurred when a patient was approaching readiness for extubation. An awareness of the adverse impact of such behaviour on trial delivery was often accompanied by avowals that selective performance of intervention-related tasks was rational, especially in the context of an already heavy workload. The tailoring of trial delivery to maximise relevance, as well as the possibilities for timely extubation, was commonly described.

Availability of appropriately skilled staff. The impact of the availability of medical consultants on weaning progress was repeatedly discussed by participants. Across all units, when the consultant was absent, responsibility for patient care fell to less experienced medical colleagues. Participants highlighted delays in weaning progression during such times, based on a combination of personal lack of confidence in their knowledge and skills, or fundamental disinterest amongst these staff; lack of confidence in their skill set on the part of senior colleagues; and the 3–6 monthly rotation of this staff group militating against relevant training in trial initiatives. A frequently described example of delayed weaning progress concerned their refusal to undertake a SBT until consultants were available.

Some units employed Advanced Nurse Practitioners (ANPs) with specialist training in patient weaning and extubation. At times, participants described the ANPs as central to the ability of the unit to progress weaning/extubation in the absence of consultant cover. In units where ANPs did not use these skills, a missed opportunity for trial delivery was highlighted. In some units, more experienced bedside nurses who had completed an advanced training course were authorised to perform SBTs, thereby increasing the availability of staff able to progress weaning. Differential physiotherapy involvement in weaning meant that only a few units authorised physiotherapists to extubate. In these units, physiotherapist participants highlighted the advantages of their involvement in terms of availability of staff to execute SBTs and/or extubate in the absence of consultant cover.

In some units, staff recruitment issues impacted trial delivery. A lack of a full complement of consultants, and/or frequent changes to the medical team, were discussed as increasing consultant workload. This engendered a concentration on priorities of patient care to the detriment of trial related activity. Additionally, locum consultants employed to make up staff shortfalls were inevitably less versed in trial activity and likely to bring greater diversity in terms of individual practice. Problems with nursing recruitment and retention were also frequently identified. Many units were dependent on junior nurses or bank and/or agency nurses. Lack of confidence and/or a preoccupation with basic patient care meant these nurses were likely to be less engaged in SANDWICH.

Promotes inter-professional working.

Across all disciplines of participants, there was widespread agreement that the trial improved inter-professional working by: broadening the range of staff involved in sedation and weaning management; outlining key roles to these different staff; and providing a shared language for communication. Of particular importance was the enhanced role played by bedside nurses. Throughout, the crucial factor identified was the assignment of specific roles and the shared decision-making through which these roles could be discharged. Both COMFORT scoring and SBT screening criteria were consistently discussed as having promoted bedside nurse autonomous practice, as well as their engagement in meaningful discussion about this practice with colleagues. In most units, the SBT screen gave bedside nurses a designated role in ventilator weaning for the first time. Bedside nurses were aware of this greater involvement, as they talked about more senior colleagues actively seeking out their recommendations for patient care. Throughout, there was a clear sense that the intervention had enhanced colleagues’ confidence in these recommendations because they were premised on a shared evidence base.

A similar, but more understated, enhanced confidence amongst senior nursing staff was apparent. They considered themselves the “drivers”, but not technical implementers, of weaning and saw their assigned SANDWICH role (to commence a patient on a SBT) to have the potential to become embedded as part of standard unit practice. Such a role was seen as representing an important advancement of their professional authority. Amongst physiotherapists, the trial was seen as having facilitated greater involvement for their own team, as they learnt more about appropriate sedation, and contributed to the formulation of patient weaning plans. There was universal agreement amongst medical participants that the trial had enhanced nursing confidence and autonomy of practice, which was viewed as a positive development. At times, their reflections were couched in a strategic context of their vision for unit inter-professional working. At other times, accounts focused on an awareness that this working had improved precisely because they were required to reflect on, and account for, their clinical decision-making.

Although the positive impact of the SANDWICH trial on inter-professional working was widely endorsed, several important caveats were articulated. First, the full realisation of nurse autonomy ultimately depended on its validation by senior, especially medical, colleagues. This could involve consultant failure to communicate with bedside nurses and/or acceptance or rejection of information and/or advice given by them. In this context, consultant participants described differential confidence in nursing staff based on their perceived knowledge and expertise. At times, senior nurses were described as retaining control of patient care, thereby limiting the potential for junior colleagues to fully discharge their designated roles. Second, although experienced nurses could be self-assured and comfortable in their interactions with senior nursing and medical colleagues, this did not hold true of the profession more generally. Several bedside nurses commented on their own reticence. Failure on the part of nurses to question medical colleagues was seen as an impediment to consistency in intervention delivery in that it could result in the perpetuation of deviation from the protocol. A case in point concerns the recording of reasons for decisions taken regarding failure to perform a SBT when indicated. Although it was a requirement that doctors explicitly account for such non-performance, with bedside nurses being responsible for recording the reasons given, in practice this did not always happen. Although frustrated, nurses frequently commented that they did not feel able to press doctors to provide a rationale. More broadly, a culture of bedside nurses not being involved in any aspect of ventilator weaning could mean they were limited, even excluded, by others, and limit or exclude themselves, from discharging their trial roles.

Final authority rests with consultant. As foreshadowed in the preceding section, across all units, participant accounts underscored the final authority of consultants regarding patient care, and consequent vulnerability of the intervention to their own willingness to adhere to the protocol, and to support the same adherence by other staff. In some units, the medical team was described as greatly facilitating unit wide adherence by demonstrating manifest interest and support. In other units, the opposite was identified, with doctors being described as disinterested, even obstructive. In both contexts, as members of staff in positions of final authority, the approach of consultants was understood to have played an important role in either promoting or impeding the consistency with which a unit adhered to the protocol and delivered the intervention as intended.