key: cord-0313957-r8dr8j03
authors: Endacott, R.; Dall'Ora, C.; Richardson, A.; Griffiths, P.; Pattison, N.; Pearce, S.; team, The SEISMIC research
title: The organisation of nurse staffing in intensive care units: a qualitative study
date: 2022-01-21
journal: nan
DOI: 10.1101/2022.01.18.22269459
sha: 18b49e4fcb8e86ea8c01375806284835617872db
doc_id: 313957
cord_uid: r8dr8j03

Aims: To examine the organisation of the nursing workforce in intensive care units and identify factors that influence how the workforce operates. Background: Pre-pandemic UK survey data show that up to 60% of intensive care units (ICUs) did not meet locally agreed staffing numbers and 40% of ICUs were closing beds at least once a week because of workforce shortages, specifically nursing. Nurse staffing in ICUs is based on the assumption that sicker patients need more nursing resource than those recovering from critical illness. These standards are based on historical working, and expert professional consensus, deemed the weakest form of evidence. Methods: Focus groups with health care professionals working in ICUs (n= 52 participants) and individual interviews with critical care network leads and policy leads (n= 14 participants) in England between December 2019 and July 2020. Data were analysed using framework analysis. Findings: Three themes were identified: the constraining or enabling nature of ICU and hospital structures; whole team processes to mitigate nurse staffing shortfalls; and the impact of nurse staffing on patient, staff and ICU flow outcomes. Staff made decisions about staffing throughout a shift and were influenced by a combination of factors illuminated in the three themes. Conclusions: Whilst nurse: patient ratios were clearly used to set the nursing establishment, it was clear that rostering and allocation/re-allocation during a shift took into account many other factors, such as patient and family nursing needs, staff wellbeing, ICU layout and the experience, and availability, of other members of the multi-professional team. This has important implications for future planning for ICU nurse staffing and highlights important factors to be accounted for in future research studies.

Pre-pandemic UK survey data show that up to 60% of intensive care units (ICUs) did not meet locally agreed staffing numbers and 40% of ICUs were closing beds at least once a week because of workforce shortages, specifically nursing (FICM, 2018) . Nurse staffing in ICUs is based on the assumption that sicker patients need more nursing resource than those recovering from critical illness. These standards are based on historical working, and expert professional consensus, deemed the weakest form of evidence (Howick et al 2012) . For a fuller description see Endacott et al (2021) .

In England, critical care services are supported by 15 Critical Care Networks (NHS Commissioning Board, 2012) . Critical care networks support the coordination of patient pathways between health services to ensure access to specialist support at a regional level and benchmark services to ensure consistency, including peer review (NHSE 2021) . Critical care networks also monitor workforce levels and provide support to individual ICUs, particularly to nurse and medical critical care leads. It is timely to examine how ICU nurse staffing is organised in this resource-constrained climate and what factors influence how the workforce operates.

A recent systematic review of 55 studies found significant associations between low ICU nurse staffing and worse outcomes for patients, staff and health services (Rae et al 2021) . Odds of nosocomial infection were 3.28-3.60 times higher in ICUs with lower levels of nurse staffing and mortality odds were 1.24-3.50 times greater (Rae et al 2021) . West et al (2014) found that significant associations between number of nurses and mortality had the greatest impact on patients at higher risk of death. Rae et al (2021) highlighted the wide variation in the way nurse staffing is defined and organised. However, the most common approach was some form of ratio (nurse:patient or nurse:bed). From their systematic review of ICU nursing workload instruments, Greaves et al concluded that no workload instrument has been adequately validated for this purpose, or found to be superior to the clinical judgment of an experienced ICU nurse manager (Greaves et al 2018) . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Persistent registered nurse and ICU-qualified nurse staffing challenges, with vacancy rates at 10% across the UK (CC3N, 2020), have led to consideration of alternative staffing models. Coupled with the development of new roles, such as critical care nursing associates (Bates, 2019) , and physician assistant/assistant practitioner roles in ICU, the workforce operates in an increasingly diverse way across the UK. The evidence around these changes is, as yet, lacking but a need to better understand what influences how this new workforce is deployed has been articulated (Henshall et al 2019) , not least as there are several implications around role boundaries that have a direct impact on staffing decisions.

The aims of the study were to examine the organisation of the nursing workforce in intensive care units and identify factors that influence how the workforce operates.

A qualitative in-depth exploration of the organisation of nurse staffing in ICUs from the perspectives of multi-disciplinary ICU teams, critical care networks and policy leads. Focus Groups were conducted with a mix of uni-professional (nursing) and multi-professional teams in ICUs and individual interviews were conducted with network leaders and policy makers.

Purposive sampling was used to recruit nurses and other members of the ICU multi-professional, critical care network leaders, national and regional policy makers to the study. The concept of information power (Malterud et al., 2015) underpinned sample size, the premise being that the larger the information power of the sample, the smaller the sample required to achieve saturation. It was anticipated that the network and policy participants would provide broad insight across a number of ICUs. Policy-makers were drawn from NHS England and Improvement, the critical care Clinical Reference Group and clinicians responsible for commissioning ICU services at regional and national level. Representing most regions in England, 14 individual interviews were conducted with network leaders and policy makers. All bar two of the interview participants had an ICU clinical background. The network participants were responsible for networks across England covering between eight and 21 ICUs, a total of 145 ICUs. Six Focus Group interviews, lasting on average just over 80 minutes, were conducted with 52 healthcare professionals (nurses, allied health professionals, doctors and nursing assistants) from four hospitals and eight ICUs across England. The number of participants at each focus group ranged from 4-12 and participants worked in general ICUs (n=4), cardiac ICUs (n=2) and neurosurgical ICUs (n=2) in hospitals of varying bed size and patient population. Professional roles for all 66 participants are provided at Table 1 . 

Study design and conduct were underpinned by principles of trustworthiness (credibility, transferability, dependability and confirmability) (Lincoln & Guba 1985) . The interview topic guides were developed with an expert advisory group external to the research team (the UK Critical Care

Nursing Alliance, which is the alliance uniting all critical care nursing organisations in the UK).

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Individual interviews were conducted by RE and Focus Groups were led by SP with support from one other member of the study team. RE and SP are highly experienced qualitative researchers with extensive backgrounds in research related to the organisational aspects of care delivery. The interviewers (RE & SP) had regular meetings during data collection and analysis. Early analysis was reviewed by the whole research team and analytical memos were shared. To ensure a strong connection between the analysis and clinical perspectives, emerging themes were discussed with a clinical stakeholder group. Focus Groups were conducted at the health services from December 2019 to February 2020 (pre-pandemic), and the individual interviews were conducted online between July-September 2020, the early stages of the COVID-19 pandemic. The individual interviews had two components: pre-pandemic staffing (reported here) and the impact of COVID-19 on staffing models (reported in Endacott et al, in press ).

Framework analysis was used, a process comprising five stages: familiarisation, defining a thematic framework, indexing, charting and mapping/ interpretation (Pope et al., 2000) . The Focus Groups were conducted first and the individual interviews built on the initial themes. Early analysis identified that nurse staffing was commonly discussed in relation to safety, and nurse staffing was described in terms of structures, processes and outcomes. Hence, we used the Donabedian model of structures, processes and outcomes to structure the a priori framework for the analysis (Donabedian 1988 ); Donabedian's model is one of the most dominant quality evaluation instruments used in healthcare and has been used in the ICU context, for example, to develop ICU patient safety indicators (Wu 2020) .

Participants were contacted prior to planned interviews and given sufficient time to consider participation. Audio-recorded verbal informed consent was sought prior to each interview. Focus Group participants provided written consent in person. Focus groups and interviews were audio-. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. 

The categories initially used for the framework were: the staffing establishment, rosters, staff allocation structures and decisions, multi-professional teamwork, communication and outcomes for patients and staff. These were iteratively refined as coding and categorisation developed, in line with framework analysis methods (Gale et al 2019) . The findings are reported using Donabedian's model of structures, processes and outcomes. Three themes were identified: the constraining or enabling nature of structures; whole team processes to mitigate nurse staffing shortfalls; and the impact on patient, staff and ICU throughput outcomes. Further categories emerged during the analysis; the final framework with themes and categories is provided at Figure 1 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10. 1101 /2022 Data excerpts presented below and in tables are annotated as follows: FG denotes Focus Group and participant type, by profession; Int refers to interview and participant category (Network Lead or Policy Maker).

Many of the organisation-wide and ICU-specific structures described by participants related to the staffing models used to set the nursing establishment and the rostering template used to provide nurse staffing for each shift. The other categories in this theme are decision-making structures and ICU factors (see Table 2 ). These were all described in terms of the extent to which they enabled or constrained the organisation of ICU nurse staffing. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. Discussions about the staffing establishment focused on number of staff and the skill-mix required;

there was evidently a complexity and interaction between roles. The inclusion of support staff (referred to by participants as Band 4, health care assistants (HCAs) or nurse associates) in the staffing establishment has been standard practice for a long time: 'we've had band 4 probably on our shop floor taking patients alongside a registered nurse for over 10 years' (FG4/Sister). These roles were also deemed important: 'we didn't used to have any but, in a way, it's probably made more of a difference (having an HCA) than having another nurse' (FG5/Staff Nurse).

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10. 1101 /2022 The skills of support (non-registered) staff were particularly valued in the care of long-term patients 'band 4 (senior) HCAs have a more holistic view [than junior support staff], they've built a relationship with the family so they know the patient a bit better ' (Int5/Network Lead) and patients with delirium (FG7/Staff Nurse). The importance of supernumerary roles (for example, shift leader, practice educator) was also emphasised, although they were also described as 'on the substitutes' bench -when it's busy they have to take a patient, which happened about 25% of the time over the past year (Int 1/Policy Maker).

The number and skill mix of the multi-disciplinary team (MDT) also had an impact on the is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10. 1101 /2022 There was a clear sense of a shift away from organ support as the rationale for nurse staffing 'we need 1:1 nurse patient ratios for different reasons now (than when ratios were first introduced 20 years ago)' (Int 8/Network Lead). This was also referred to as needing 'more columns' in the decision-making algorithm:

However, whilst it was suggested that 'most units don't measure against the ratios, day to day' (Int 10/Network Lead), they were deemed to provide a safe standard that was easily communicated from junior ICU nurse right through to Board Members. As a policy lead (and ICU clinician) The executive-level hospital decision-making structures were also perceived to be important; however, the inclusion of non-clinical managers in staffing decisions caused frustration and anger:

'as senior nurses we have lost the ability to manage our staff within intensive care …the key frustration is when it's the Housekeeping Supervisor, who happens to be the on-call manager, telling you to move staff to the ward' (FG1/Sister) Having 'a champion for ICU at executive level' (Int 9/Network Lead) was deemed to make a positive difference, as well as twice-daily meetings to look at hospital-wide staffing (FG2, FG5). A longer-term impact of hospital staffing policies related to promotion criteria with nurses in some sites having to leave ICU to achieve promotion (Int 4, Int 8, FG4). This was also reflected in the nursing staff impact category, under the Outcomes theme (see Table 2 ). P1: …..the rehab patients are not an organ failure in the strictest sense of the word. They're not on filtration (renal support). They're on minimum amounts of oxygen but they are requiring the services of one nurse just to look after them in the way that they should be looked after. (FG5/Doctor) . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. Individual ICU factors such as visiting policies (FG6), admission policies (FG2), the size of the ICU (Int4) and the mix of emergency and elective patients admitted (FG7) influenced how nurse staffing was managed. The latter point had an important influence on processes, in particular the frequency of workarounds, articulated more frequently in ICUs with fewer elective patients. However, the layout of the ICU was the most frequently discussed factor across Focus Groups and individual interviews. This was important for structure and process factors as it affected whether the staffing establishment was correct for that particular ICU but also affected how staff were allocated and reallocated during the shift (see Table 2 ).

At one health service, a new ICU layout, with mainly side rooms, resulted in all patients needing a 1:1 ratio because staff were in a side room and unable to oversee more than one patient due to the physical environment (Int 11/Network Lead). This led to an increased staffing establishment, particularly supernumerary staff, in order to manage breaks and provide adequate supervision; however, the nurse manager had to defend these decisions by developing 'a checklist of reasons why patients can't be left alone in side rooms, to justify staffing [to managers outside of ICU]' (Int 11/Network Lead).

The processes described were positive steps taken to continue care provision when staffing was challenging, often undertaken in rapidly changing situations, such as emergency admissions, patients . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10. 1101 /2022 deteriorating. The main categories were workarounds, communication and staff allocation decisions (see data excerpts at Table 3) . Table 3 Data excerpts related to whole team processes to mitigate inadequate nurse staffing theme

The physios and the techs, they will muck in. They will come and help. They will do turns. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. Table 3 ).

The risks of admitting a patient when there were not enough nurses were balanced against the resource drain of transferring the patient to another hospital: 'pulling a doctor and a nurse out of the staffing for two or three hours' (FG7/Doctor). It was clear that allocation decisions were not based solely on nurses' experience or education; confidence and trust in the nurses also had a key role to play. Tailoring allocations to staff wellbeing as well as patient need came through as key to getting these decisions right:

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. Decisions about staffing, mainly allocation and re-allocation throughout the shift, were emphasised:

'it's not that one allocation first thing in the morning. It's how that allocation works through the day.

It's quite an organic thing maybe (FG6/Sister). The acuity of the patient was the 'most important' driver (FG6/Sister) but care needs, such as bowel care, pressure area care (FG3/nurse) and the need for family support (FG5/Matron) were also factored in where possible. By contrast, there was a clear perception that decision-makers from outside of critical care focused on levels of care and the nurse:patient ratio (structure factors) to make decisions, usually about moving ICU nurses to the ward to fill gaps. A network participant also reported a 'lack of understanding (from managers outside of ICU) of what it means to be a 24/7 service' (Int 8/Network Lead). The importance of avoiding 'levels of care' terminology when discussing staffing with managers outside of ICU was evident in one setting: P1: that takes a bit of justification sometimes. But that's why, on our unit, when we report back to the powers that be, we don't refer to level 3, 2 or 1 (level of sickness). We refer to patients that are one to one (nurse:patient ratio), one to two or 'fit for the ward'. So, that they can be easily justified and they can understand why on this particular day we might be fully staffed but actually … we have no beds because of the acuity of the unit based on one to one, one to two or ward-fit patients….. were also moved from ICU to the wards either for a shift or, in one setting, for a longer secondment to the ward (FG7/Sister). This was also perceived as damaging for nurses: 'it breaks people (to be constantly asked to work on the wards)' (Int 4/network lead). There was also a perception that ICU . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 nurses were seen as a 'feeding station for the wards' (Int 8/Network Lead), providing a bank of staff to fill gaps in ward staffing.

Participants identified a number of outcomes that they perceived to be influenced by nurse staffing but also identified the challenges of demonstrating some of these outcomes in the data currently collected. Categories in this theme were: adverse events; staff support; ICU flow and nursing workforce (see Table 4 ). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. Most of the outcomes discussed were negative outcomes for patients or staff; safe staffing was most commonly described in patient safety terms. Network participants identified a number of outcomes that they see as 'red flags' (potential risks to safety), particularly adverse events such as nasogastric tube displacements (Int 4/Network Lead), and pressure injuries (Int 1/Policy Lead). Focus Group participants emphasised the lack of time to provide full care for the patient, such as brushing hair and teeth and planning for the next stage in the patient's care, and the lack of time for nurses to take breaks and remain hydrated (FG3/Staff Nurses and Sisters). Doctors also articulated an impact on medical workload of having experienced ICU nurses, for example in prioritising the actions needed after the ward round (FG1/Doctor; FG7/Doctor), but in a more general way they also emphasised 'the importance of the art and skill of the senior nurse' (FG6/doctor).

Having enough staff to provide support to colleagues during tough shifts was deemed important.

Examples included the reallocation of experienced nurses to work closely with, and support, less experienced staff (FG4/Staff Nurse) and the matching of nurse experience with patient level of care so that the situation was safe for patients and staff (FG5/Staff Nurse). Receiving good feedback from the nurse in charge and/or colleagues was reported to make a big difference for team morale and job satisfaction (FG 4/Staff Nurse and Sister). Important outcomes for nurses included having the time to provide fundamental care such as hair washing and nail care. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Patient flow outcomes such as delayed admission (FG 1/Staff Nurse), transfers for non-clinical reasons (Int 1/Policy Lead) and newly closed beds were also important outcomes suggesting to participants that nurse staffing may be inadequate. Weaning from sedation medication is important in preparing patients for ICU discharge but often delayed because of nurse staffing: 'we sometimes sedate patients for longer if there aren't enough staff' (FG 7/Doctor). Patient rehabilitation, usually led by physiotherapists, was also delayed, especially when the nurse was managing two patients and not able to assist physiotherapists (FG7/Physio). Other aspects of physiotherapy treatment were also affected by nurse staffing: 'we want to go in and bag [manually ventilate to loosen secretions] the patient but they're often cardiovascularly unstable so we need a nurse to be there in case they need a bolus [of medication] or something…. Often we can't get on with what we need to do (FG4/Physio).

Nursing workforce outcomes discussed by participants included retention of nurses, nurse vacancies, use of agency staff (Int 9/Network Lead), loss of education opportunities (FG1/Staff Nurse) and the number of shifts without supernumerary staff (Int10/Network Lead). The risk of allocating two high dependency patients to one nurse was also echoed across the interviews and Focus Groups: 'when one level 2 patient goes off (deteriorates) another level 2 patient is left unattended (Int 7/Network Lead). The impact of being allocated two level 2 patients was articulated by some nurses as inducing a feeling of failure because they could not spend enough time with either patient (FG7/ Staff Nurse).

The challenges of demonstrating outcomes, particularly of reduced staffing, were acknowledged:

'the patient case-mix programme measures things like length of stay and mortality but these are not qualitative enough for nursing workload, but there are flaws in subjective data too, as things like patients and family satisfaction can be skewed by gratitude [for surviving the ICU stay]' (Int 8/Policy Lead).

The interaction between the three themes is depicted at Figure 2 ; decisions about staffing were made throughout a shift and were influenced by a combination of factors illuminated in the three themes. Participants often described structural, process-related and outcome factors that influenced . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 staffing decisions. An example was narrated in one Focus Group, comprised mainly of Matrons, where decisions about staffing depended on the structure of the MDT staffing for that shift, whether workaround processes were feasible and the outcome of current staffing, with regards to patient rehabilitation (FG5). Participants described reallocating staff during a shift, where possible, if patients were not going to get the care they required. Similarly, the complex interplay between the structure, process and outcome factors could constrain or enable decisions about nurse staffing. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 

Previous studies provide evidence that outcomes such as patient mortality and increased risk of nosocomial infection are associated with nurse staffing in ICU (Rae et al. 2021) . Analysis of our qualitative data from focus groups and individual interviews with 66 healthcare professionals showed three key factors influencing the complex and dynamic organisation of nurse staffing: multiprofessional interdependence and workarounds; ICU geography; and staff wellbeing.

The ICU care of critically ill patients is typically delivered using inter-professional care bundles, for example the ABCDE bundle for managing sedation, delirium and early mobility (Pun et al. 2019 ) and the ventilator associated pneumonia (VAP) reduction bundle (Rello et al. 2013) , with clear responsibilities for the different professions but, ideally at least, shared mental models (Botley et al. Liu et al. 2021) . Hence the blurring of roles evidenced in our findings should not be a surprise.

However, the examples, including workarounds, provided by our participants were not part of a planned inter-professional care bundle and carried an implicit assumption that they will understand each other's roles, e.g. that the physiotherapist will know whether s/he needs to alert the RN to the observations just recorded. There is also an implicit assumption that interdisciplinary working will take place; our participants indicate that patient safety relies on this. However, there is no infrastructure to support this, for example knowledge/sharing of rosters between professions or joint planning, with evidence that hierarchical issues and cognitive biases persist (Liu et al. 2021) .

Staffing guidance assumes that healthcare professionals work in silos, rather than in the multiprofessional models portrayed by some of our participants and reflected the work cited above (Pun et al. 2019 , Rello et al. 2013 , Botley et al. 2019 . The distinct contribution of individual professions to the management of the critically ill patient has been articulated, particularly in the many studies evaluating care bundles (Pun et al. 2019 , Rello et al. 2013 , Botley et al. 2019 ); our findings do not contradict this. However, there is unlikely to be a one size fits all solution and the potential opportunities for different staffing models are likely to be highly dependent on other professions. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10. 1101 /2022 Hence, any change to staffing models needs to take into account how different professions work together, and to formalise this in staff planning processes.

Processes are more amenable to change than structures (Mountford & Shojania 2012) , which may explain why workarounds were quite dominant in the Focus Group discussions. The number of processes, including decision-making, reported by our participants to make the ICUs 'safe' for patients and staff highlight the importance of undertaking process evaluation alongside implementation of a change to staffing models. Process evaluation, as a component of a research study, captures the context in which changes are introduced and other factors that might illustrate why change does or does not become embedded.

The impact of the workplace layout on nurse workload has been reported in previous studies. The increase in nurses' workload of changing from multi-occupancy to single room layout, has been reported in hospital wide and ICU settings (Maquire et al 2013, Lin et al. 2016 ). In Lin et al's (2016) exploration of an ICU relocation, participants voiced concerns about patient and staff safety with the move to a single room model, whilst Leon-Villapalos et al (2020) found that perceptions of safety in ICU were negatively correlated with higher patient numbers and higher percentages of patients managed in single rooms. These studies, together with our findings, suggest that patient safety in ICU may not be best served by blanket 'ratios' approaches to nurse staffing, intended to apply uniformly across health services.

Staff wellbeing has arisen as a major concern during the COVID-19 pandemic, with one study in nine UK hospitals showing evidence of 49% of ICU nurses meeting criteria for post-traumatic stress disorder (168/344) and/or depression (167/344) (Greenberg et al 2021) . Similar levels are reported in other studies (Rattray et al 2021 , Sampaio et al 2021 . Whilst these findings are important for future recruitment and retention of ICU nurses, our findings show that, among our participants, the shift-by-shift organisation of nurse staffing takes account of staff wellbeing were possible. Our Focus . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10. 1101 /2022 Group interviews were conducted before the COVID-19 pandemic, demonstrating that concerns about staff wellbeing are not purely pandemic-related.

Across the interviews and Focus Groups it was evident that a constant re-visiting of decision-making about staffing happened throughout the shift; this is reflected in the categories for structure, process and outcome themes. Most of the outcomes described are unintended consequences of the staffing model, or decisions in relation to the model. For example, the review of the number of level 2 and level 3 patients against the number of RNs on a shift, by a decision maker with limited understanding of ICU patient needs, might lead to an ICU RN being deployed to a ward to cover staffing shortages. However, the ICU shift leader cannot move the ICU RN back to ICU when a new patient is admitted (unintended outcome). Some of these unintended outcomes warrant exploration in empirical studies designed to examine staffing models.

This national study recruited participants from ICUs of different size, with different heath service configurations from across England. The inclusion of staff from different professions has provided a more complete picture of factors influencing, and influenced by, nurse staffing in intensive care units. We conducted the focus groups before the pandemic, and the individual interview participants reflected on the organisation of staffing pre-pandemic, hence it is not known the extent to which perceptions about nurse staffing may have changed. However, our findings reveal a picture of under-staffing, at times, across all settings; this situation will not have improved during the pandemic.

There was a, sometimes implicit, focus on patient and staff safety across the interviews, hence Donabedian's model was a useful framework with which to structure the initial analysis. The Donabedian approach did not detract from inductive generation of themes, but provided an overarching framework for analysis, facilitating constant comparison across the data sources. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 21, 2022. ; https://doi.org/10.1101/2022.01.18.22269459 doi: medRxiv preprint

Whilst ratios were clearly used to set the nursing establishment, it was clear that rostering and allocation/re-allocation during a shift took into account many other factors such as staff wellbeing, the nursing needs of patients and family members, the ICU layout and the experience of all members of the multi-professional team. This has important implications for the future planning for ICU nurse staffing and highlights important factors to be accounted for in future research studies.

The findings have the potential to feed into discussions about the national tariff for critical care and quality metrics to be included in commissioning contracts.

This paper presents independent research funded by the National Institute for Health Research (Programme Development Grants, Safe staffing in ICU: development and testing of a staffing model, NIHR200100). The views expressed in this publication are those of the author(s) and not necessarily those of the National Institute for Health Research or the Department of Health and Social Care, neither of whom have had involvement in any aspect of the design, data collection, synthesis, interpretation or writing of, this review.

Developing the nursing associate role in a critical care unit

Ability to predict team members' behaviors in ICU teams is associated with routine ABCDE implementation

National Critical Care Nursing Workforce Survey: Overview Report

The quality of care: How can it be assessed?

and the SEISMIC study team (in press) How COVID-19 has affected staffing models in intensive care: a qualitative study examining alternative staffing models (SEISMIC)

Workforce Data Bank

Faculty of Intensive Care Medicine/Intensive Care Society (2019) Guidelines for the Provision of Intensive Care Services

Comparison of rapid vs in-depth qualitative analytic methods from a process evaluation of academic detailing in the Veterans Health Administration

Mental health of staff working in intensive care during COVID-19 Occupational Medicine

Nursing workloads and activity in critical care: A review of the evidence

The role of the assistant practitioner in the clinical setting: a focus group study BMC

Oxford CEBM Evidence Levels of Evidence (Introductory Document)

Exploratory study of staff perceptions of shift safety in the critical care unit and routinely available data on workforce, patient and organisational factors

Relocating an intensive care unit: An exploratory qualitative study

Naturalistic inquiry, 1st edn

Barriers and facilitators to interdisciplinary communication during consultations: a qualitative study

Clinician perceptions of a changing hospital environment

Sample size in qualitative interview studies: guided by information power. Qualitative Health research

Refocusing quality measurement to best support quality improvement: local ownership of quality measurement by clinicians

Developing operational delivery networks the way forward

Adult Critical Care Service Specification (DO5)

Analysing qualitative data

Caring for Critically Ill Patients with the ABCDEF Bundle: Results of the ICU Liberation Collaborative in Over

Outcomes sensitive to critical care nurse staffing levels: a systematic review

Work-related stress: the impact of COVID-19 on critical care and redeployed nurses: a mixed-methods study

A care bundle approach for prevention of ventilator-associated pneumonia

Impact of COVID-19 outbreak on nurses' mental health: A prospective cohort study

Nurse staffing, staffing, medical staffing and mortality in Intensive Care: An observational study

The Development of Patient Safety Indicators for Critically Ill Patients Admitted to the Intensive Care Unit (Unpublished doctoral thesis)

The SEISMIC research team comprises: Suzanne Bench, Fiona Boyd, Carole Boulanger, Chiara