Objectives To identify the patterns of teamwork displayed by interprofessional teams during simulated management of medical deterioration in pregnancy and examine whether and how they are related to clinical performance in simulated practice.
Design Exploratory observational cohort study.
Setting Interprofessional clinical simulation training with scenarios involving the management of medical deterioration in pregnant women.
Participants Seventeen simulated scenarios involving 62 qualified healthcare staff working within the National Health Service attending clinical simulation training (midwives (n=18), obstetricians (n=24) and medical physicians (n=20)).
Main outcome measure(s) Teamwork behaviours over time, obtained through detailed observational analysis of recorded scenarios, using the Temporal Observational Analysis of Teamwork (TOAsT) framework. Clinician rated measures of simulated clinical performance.
Results Scenarios with better simulated clinical performance were characterised by shared leadership between obstetricians and midwives at the start of the scenario, with obstetricians delegating less and midwives disseminating rationale, while both engaged in more information gathering behaviour. Towards the end of the scenario, better simulated clinical performance was associated with dissemination of rationale to the team. More delegation at the start of a scenario was associated with less spontaneous sharing of information and rationale later in the scenario. Teams that shared their thinking at the start of a scenario continued to do so over time.
Conclusions Teamwork during the opening moments of a clinical situation is critical for simulated clinical performance in the interprofessional management of medical deterioration in pregnancy. Shared leadership and the early development of the shared mental model are associated with better outcomes.
- interprofessional teams
- team-based care
Data availability statement
No data are available. Due to ethical restrictions, the data cannot be shared.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
What is already known on this subject
Managing medical deterioration in pregnancy requires healthcare providers from different professional backgrounds (eg, midwives, obstetricians and medical physicians) to work together as a team to save two lives.
Poor interprofessional teamworking and communication are a leading cause of maternal deaths following medical deterioration in pregnancy.
Research suggests that leadership and having a shared mental model contribute to successful management of medical deterioration, yet little is known about how teamwork unfolds over time or how the actions of one team member may influence subsequent teamworking or how patient care is delivered.
What this study adds
Shared leadership, and development of the shared mental model at the very start of the clinical interaction were associated with better team simulated performance, suggesting that these behaviours were associated with teams working together effectively.
Delegation at the start of a scenario was associated with poorer clinical performance in simulated practice.
Overall, the type of leadership behaviour displayed in the opening moments of a clinical interaction was related to the subsequent teamwork, and to the effective simulated clinical performance of that team; these findings develop our understanding of the temporal dynamics of teamwork in complex clinical scenarios and provide evidence to inform current educational training and future definitive studies testing cause and effect relationships.
Teamwork is a critical component of effective healthcare. The treatment and management of a hospital inpatient are rarely achievable by one individual but rely on the expertise and skills of multiple specialty doctors, nurses, midwives and many other allied health professionals to work as a cohesive, interprofessional team. Although the importance of teamwork for high quality care is well recognised,1 2 evidence suggests that interprofessional working may be challenging,3–5 and the absence of effective interprofessional teamwork can compromise patient safety.2 6–9
Teamwork is particularly important in the care of pregnant women with coexisting medical conditions. A recent report, Mothers and Babies: Reducing Risk through Audits and Confidential Enquiries (MBRRACE), found that the majority of UK maternal deaths are ‘in-direct deaths’ occurring due to pre-existing or new-onset medical conditions (eg, neurological or cardiac diseases), rather than directly related to pregnancy (eg, pre-eclampsia).10–12 The successful management of medical deterioration in a pregnant woman requires multiple sources of expertise and healthcare providers from different professional backgrounds (eg, midwives, obstetricians and medical physicians) working together as a team to manage a potentially life-threatening situation and save two lives. These professional groups do not routinely work together but must form a cohesive interprofessional team during such rare but critical situations. Numerous recent reports, including MBRRACE,10–12 specifically identify poor interprofessional teamworking and communication as leading causes of maternal death in such situations.10–15 As such, improving teamwork in this specific context is a requirement for improving patient care and reducing maternal death.
Improving teamwork requires in-depth knowledge of which aspects of teamwork are most effective for ensuring good team outcomes. Theoretical teamwork models suggest that successful teamwork requires effective leadership, development of shared mental models, adaptability, backup behaviour, team orientation and mutual performance monitoring.16 Two studies have conducted video analysis observations of teamwork in the management of emergencies in pregnancy. However, both have focused on emergencies directly related to the pregnancy (ie, postpartum haemorrhage17 and pre-eclampsia18) rather than medical deterioration due to pre-existing or new medical conditions (eg, asthma or epilepsy). Observational analysis of real-world incidents of postpartum haemorrhage suggest that effective leadership is a critical predictor of performance.17 Analysis of simulated scenarios involving pre-eclampsia suggest that teams that take time to coordinate and share the mental model of the problem and the actions required have better simulated outcomes.18 These studies have analysed teamwork as a static phenomenon, assigning a total score for the whole team. Such studies confirm the importance of leadership and developing a team shared mental model, but do not provide details of how team members interact and coordinate their work.
In this study, we used a novel analytical technique to analyse teamwork as it unfolds temporally. Teamwork is dynamic, evolving over the course of an interaction. The demands of clinical care unfold temporally, and so clinical teamwork must also have a temporal aspect. Teamwork is achieved by team members acting both individually and in response to others’ behaviour moment to moment, while responding to the fluctuating demands of patient care. Examining teamwork as it unfolds over a clinical episode is important for understanding how teams coordinate and interact. However, temporal dynamics in organisational teamwork has received relatively little research investigation to date.19
Conducting a fine-grained analysis of how interprofessional teams manage medical deterioration in pregnant women, due to a new or pre-existing medical condition, is needed to identify the specific behaviours used in this context, and their corresponding temporal patterns. This requires detailed analysis of video recordings of clinical care using interaction analysis research methods.19 Previous studies have conducted such analyses in the context of anaesthesia and surgery. For example, in anaesthesia, observational analysis of simulated anaesthesia scenarios identified that speaking up behaviours, team monitoring and development of shared mental models predicted team simulated performance.20 21 Observational analysis of leadership behaviours of surgical teams during actual operations revealed that, during more complex phases of an operation, having one central clear leader rather than shared leadership resulted in better attainment of team goals.22
This study is to our knowledge the first to conduct a detailed analysis of teamwork behaviour over time, during simulated medical deterioration due to medical conditions in pregnant women. Simulated practice is a promising avenue for research, which has been under used to date.23 The study had two aims: first, to explore the relationship between teamwork behaviours that occur at different timepoints during an episode of clinical care, and second, to explore the relationship between teamwork and simulated clinical performance.
The corpus consisted of 17 audio visually recorded simulated scenarios recorded as part of the Multi-Disciplinary Simulation Training for Medical Emergencies in Obstetrics (MEmO) course.23 Recordings were randomly selected from a set of standardised MEmO courses delivered at a large simulation training centre in London, UK (May to July 2015 (n=9 scenarios) and December 2019 to January 2020 (n=8 scenarios)). Five different scenarios were recorded based on MBRRACE report cases,10 or real incidents within the institution. Scenario topics are listed in table 1. In each scenario, a pregnant woman with deteriorating health due to a medical condition not related to her pregnancy (eg, epilepsy) is cared for by an interprofessional team in an acute (hospital) setting. Scenarios mean duration was 622.67 s (SD: 186.76; range: 287.40–1060.59). Each scenario involved a high fidelity Maternal Simulator (Noelle24), which is a sophisticated full-size female manikin, and an embedded simulation practitioner was also present playing the role of a student midwife providing prompts if the scenario stalled.
Recordings included 62 qualified healthcare professionals working in the UK National Health Service (midwives (n=18), obstetricians (n=24) medical physicians—doctors that are not obstetricians but from other medical specialities (n=20)). Each scenario involved a minimum of 3, or a maximum of 4, unique participants, most of whom had not worked together previously; This is similar to clinical practice, where team members often do not know each other.
At the start of each training day, participants were provided with a study information sheet, given an opportunity to provide their written consent to participate and offered their right to withdraw. Participants were informed that their participation during the day would be audio visually recorded for subsequent video analysis. Participants were instructed to respond to events in the simulation as they would during routine clinical care. Participants were briefed on the simulator and environment, and were debriefed after each scenario using a structured debriefing approach25 by two trained and experienced debriefers. They participated in one scenario during the day and observed for the rest of the time. MEmO course details are presented elsewhere.23
The same interprofessional team of clinicians involving obstetricians, medical physicians and midwives designed the course and delivered the course on all occasions. This team was led by one of the coauthors, AB. The course was highly protocolised to ensure both the scenarios and the standard of course delivery remained consistent. The course manual included detailed scenario content, the role of the embedded practitioner (EP) and predefined essential clinical actions that trainees were expected to complete during each scenario.
Analysis of teamwork behaviours
Teamwork behaviours displayed by scenario participants were identified through microanalysis of behaviour. Observations used the Temporal Observational Analysis of Teamwork (TOAsT) framework. This is presented in figure 126 and is an observational framework consisting of five overarching teamwork domains (eg, Leading the team). These teamwork domains are operationalised though 23 observable behaviours, which are the specific verbal and non-verbal behaviours that can be identified during observations of teams (eg, setting a task for the team to complete). Observable behaviours are then grouped conceptually based on the function they have within the interaction. For example, the observable behaviour of ‘task setting’ would fall into the behavioural functional category of ‘planning’. The observable behaviours are grouped into 13 behavioural functions (figure 1). TOAsT is a framework for identifying observable behaviours. Although attitudes are not directly observable, the expression of attitudes in behaviour to team members is observable. For example, negative attitudes are inferred from behaviours such as ignoring team members when they occur in the context of a difficult team interaction. Although it is possible that ignoring others might be due to factors such as task fixation, the inference of negative attitudes was only made if the quality of the interaction was inhibited or conflicted. Although precision in such inferences is difficult to establish, a structured observational tool, review by research team members, and multiple trained raters were employed to increase the reliability of coding across scenarios.
Using the TOAsT framework allows observable behaviours (n=23) to be coded from video, these can then be categorised based on their ‘behavioural function’ (n=13) and these can then be further classified into the teamwork domain that they belong to (n=5; figure 1).
Two trained TOAsT raters (a psychologist, ML, and a research student under supervision of ML) coded TOAsT observable behaviours in the linguistic annotation software ELAN.27 ELAN allows raters to apply codes directly to video, providing a behavioural time series frame by frame. To allocate codes, raters viewed each scenario multiple times, rewinding and reviewing small sections of video footage. Observable behaviours of all participants in each scenario were coded. Two annotators achieved 85% agreement in behaviours identified (Cohen’s Kappa=0.79, SE=0.03). Differences in identified behaviours between raters were reconciled. There were no differences between raters in event durations. The resulting detailed behavioural time series was exported into the statistical software package SPSS28 for statistical analysis.
Clinical performance in simulated practice
The MeMO course is highly protocolised and follows a course manual. Each scenario is described in detail within the manual with predefined essential clinical actions that trainees were expected to complete. For scenario topics, and the essential clinical actions associated with it, see table 1. Two clinicians, who were involved in development and delivery of the course (one obstetrician and one medical physician from outside obstetrics), watched each scenario independently and decided whether each clinical action was performed spontaneously by participants (ie, unassisted by EP). Actions performed spontaneously were coded as one, while actions not performed, or performed only after assistance by the EP, were coded as zero. Clinical raters reached 100% agreement on all scenarios. For each scenario, the number of actions spontaneously performed was calculated as a percentage of the total number of possible actions (eg, three actions spontaneously performed out of a possible five gave a score of 60%). This percentage was used as an index of simulated clinical performance, with a higher percentage indicating better performance.
Statistical analyses were conducted using SPSS V.24.28 The behavioural time series of each scenario was aggregated into 30 s windows, to facilitate analysis of temporal dynamics.19 This time unit was chosen as no behaviour lasted longer than 30 s. The frequency and total duration of behaviours (in seconds) displayed by each participant within each 30 s window were identified. Behaviours crossing 30 s boundaries were counted in both adjacent windows, with the precise duration (in seconds) within each window recorded.
To enable analysis of sections of interaction, while adjusting for the variations in scenario duration, each scenario was divided into three sections of equal duration based on overall scenario duration. This is an accepted approach for standardising recordings of naturalistic behaviour for analysis.19 29 The time divisions were calculated as follows: (total scenario duration (in seconds)/3=duration of each time section (in seconds)): scenario start (T1), middle (T2) and end (T3). The mean duration of each time section across all 17 scenarios was 190 s (SD=62.18; range: 60–330 s).
Spearman’s correlations explored the relationship between behaviours designed to lead the team and develop a shared mental model over time. Correlations between behaviours at T1 and T2, and between behaviours at T2 and T3 were investigated.
The relationship between teamwork behaviours and simulated clinical performance was explored using generalised estimating equations (GEE). This was limited to the teamwork behaviours for which there was sufficient data to enable statistical analysis. GEE was selected as it provides the parameters of a general linear model for data that are clustered. For example, these data are clustered by scenario, meaning that we would expect correlations between behaviours that occur within the same scenario, because they involve the same people and may be related to each other. GEE adjusts for intrascenario correlations, and accounts for the fact that data are organised sequentially.
In order to identify which teamwork behaviours were associated with simulated clinical performance overall, a GEE explored the relationship between simulated clinical performance (dependent variable: % actions correct) and teamwork behaviours (predictor variables: leading the team and developing the shared mental model behaviours) aggregated across all professional groups and time points. Behaviours significantly associated with simulated clinical performance were further analysed using GEE models to identify behavioural predictors of simulated clinical performance (DV) as they are displayed by each professional group at each of the three time points. Significance levels were set at p<0.05. Effect sizes were calculated as Phi=√(χ2/n).
Simulated clinical performance
Performance scores ranged from 0% to 100% with a median of 80% (mean 71% SD=27).
Overview of behavioural durations
Across the sample of 17 scenarios, 3639 distinct behaviours were observed, with a mean duration of 2.6 s (SD=2.7, range:.02–30 s). Behavioural frequencies and durations were highly correlated for all behaviours (Rho >/=0.9), and so the mean duration of behaviour was used for all subsequent analyses.
The mean durations of behavioural functions (see TOAsT—functions column, figure 1) (per 30 s interval) by professional group (obstetricians, medical physicians and midwives) are displayed in figure 2. Participants spent most time displaying behaviours in the teamwork domains of leading the team (delegating, planning, disseminating rationale and information gathering), and developing the shared mental model (information clarification and information sharing). Comparatively less time was spent displaying behaviours in the domains of assisting (requesting and providing help), monitoring the team (implicit and explicit performance monitoring) and team attitudes (positive attitudes, negative attitudes and disagreement), which precluded further statistical analysis. A descriptive analysis of the behaviours that were used less will be presented below. The remainder of the results section will focus on behaviours in the domains of leading the team and sharing the mental model.
Descriptive analysis of teamwork behaviours excluded from the statistical analyses
Obstetricians displayed most help requests, while midwives provided the most help. All professional groups spent very little time displaying explicit performance monitoring behaviours and midwives did not display it at all. However, midwives spent comparatively more time than other professional groups, displaying implicit performance monitoring (ie, implicit communication designed to get others to identify their own mistakes/errors). The use of positive attitude (eg, providing positive encouragement to team members) and negative attitude behaviours (eg, ignoring others) did not appear to differ across professional groups. Midwives rarely displayed disagreement.
Temporal dynamics of teamwork behaviours across all participants
Correlations between behaviours designed to lead the team (delegation, information gathering, planning, and disseminating rationale) and develop a shared mental model (information clarification and information sharing) across all participants are displayed in table 2. More delegation at T1 (start) was associated with less information sharing at T2 (middle) (p<0.01). More planning and disseminating rationale at T1 were associated with more dissemination of rationale at T2 (p<0.05). Information clarification at T1 was associated with more information clarification at T2 (p<0.05).
More delegation at T2 was significantly associated with more information clarification at T3 (end) (p<0.01), while more planning at T2 was significantly associated with less delegation at T3 (p<0.05). More dissemination of rationale at T3 was significantly associated with more dissemination of rationale (p<0.01), less information clarification (p<0.01) and more information sharing at T2 (p<0.05). More dissemination of rationale at T2 was also associated with more information gathering at T3 (p<0.01).
Relationship between teamwork and simulated clinical performance
Table 3 shows the results of the GEEs analysing differences in the duration of leadership and shared mental model behaviours between scenarios with good and poor simulated clinical performance (% correct actions) across all professional groups and all time points. Overall, better simulated clinical performance was significantly associated with less delegation, more information gathering, dissemination of rationale and information sharing (see table 3). As such, these teamwork behaviours were included in the subsequent GEE analysis exploring the relationship between simulated clinical performance and teamwork behaviours of each professional group at each time point (T1-3) (see table 4).
Significant results are reported in table 4. At the start of scenarios (T1) with better simulated clinical performance, obstetricians display less delegation (p<0.01, Phi=0.24) and spend more time gathering information (p<0.01, Phi=0.26) and sharing information with the team (p<0.01, Phi=0.28), while midwives gather information (p<0.01, Phi=0.35) and disseminate rationale (p<0.01, Phi=0.39). No significant relationships between teamwork and simulated clinical performance were identified at T2. At the end of scenarios (T3) with better simulated clinical performance, dissemination of rationale is displayed by both medical physicians (p=0.03, Phi=0.20), and midwives (p=0.02, Phi=0.22), while medical physicians also display more information sharing (p=0.04, Phi=0.20). Effect sizes ranged from 0.20 to 0.39, indicating small-to-medium effects, with the largest effect sizes for midwives' information gathering and rationale at T1.
The findings of this study demonstrate the importance of teamwork for good clinical performance in simulated scenarios requiring management of medical deterioration due to new or pre-existing medical conditions in pregnant women. It provides clear evidence of which teamwork behaviours may be associated with better performance in simulated practice, and therefore which behaviours to potentially target in future research. Better simulated performance was predicted by teamwork behaviour in the opening and closing moments of a clinical patient care event. Specifically, scenarios with better simulated clinical performance were characterised by shared leadership between obstetricians and midwives at the start of the interaction, with obstetricians delegating less, midwives disseminating more rationale and both engaging in more information gathering behaviour. Towards the end of the interaction, better simulated clinical performance was associated with medical physicians and midwives disseminating their rationale to the team. Exploring the temporal dynamics of teamwork, we found that delegation at the start of a scenario is associated with team members not sharing information and rationale later in the scenario. However, when behaviours designed to develop a shared mental model (information sharing, information clarification) and to share leadership thinking (rationale and planning) were used at the start of the scenario, they were associated with continued use as the scenario progressed.
The opening moments of scenarios with poorer simulated clinical performance were characterised by obstetricians spending less time gathering or sharing information with their team and more time delegating to others. This fits with the literature showing that emergency teams that act ‘too fast’ may fail to make an initial assessment of the situation and decision making may suffer as a result, leading to poorer clinical performance.30 Although a speed accuracy trade-off is one explanation for this pattern, there are several other possible explanations. One hypothesis is that early delegation by the doctor in the scenario may define a hierarchical dynamic from the start of the interaction, disempowering other team members (eg, midwives) from performing at the peak of their professional ability in the context of the team. Our findings demonstrating a relationship between delegation at the start of the scenario and less spontaneous sharing of information and rationale later in the scenario, further support this theory. An alternative explanation is that teams with poorer simulated clinical performance, may have less experienced, or less capable, team members who require more explicit delegation and instruction. An experienced team leader may recognise this early and engage in more delegation. Thus, the relationship between poorer performance and early delegation may be indicative of the lower baseline performance of these teams.
The pattern of early delegation may be problematic for several reasons. First, it could mean that the person delegating does so without having assimilated the knowledge they need to make accurate decisions (eg, acting too fast30). Second, team members are physically or cognitively engaged in tasks they have been delegated at the start of the interaction, and therefore may not attend to clinically relevant discussions that occur during this time. This may hamper the teams’ ability to develop a shared understanding of the clinical picture. This, in turn, may result in staff being less able to use their own initiative to begin necessary subsequent follow-up tasks. However, if the team members are less capable or experienced, early delegation by an experienced leader may be necessary.
In line with previous studies that have examined the relationship between teamwork and clinical performance in managing emergencies in the context of pregnancy, this study highlights the critical roles of shared leadership31 and building a shared mental model17 18 in facilitating good outcomes. This is one of the first empirical studies to identify a link between the temporal dynamics of team communication and simulated clinical performance. The findings are supported by research in surgical teams, which found that shared leadership was effective during certain phases of surgery and not beneficial at other times.22 In this study, the type of leadership behaviour in the opening minutes of the scenario was critical to the subsequent teamwork, and to the effective performance of that team. Although effect sizes were small to medium, they are in line with those expected in a study of this nature.32 In making this link, we demonstrate the value of sequential, temporal analysis of teamwork behaviour for revealing interpersonal dynamics that may be masked by exploring the interaction as a whole, or by focusing on only one team member.
Consistent with calls to broaden the scope of simulation research,33 34 this work also shows how simulation can be used to study team dynamics and individual behaviour in ways that are much more difficult to do in clinical environments dedicated to patient care.35 By designing scenarios that provide opportunities for leadership within interprofessional teams, and for team members to practice leadership behaviours that are associated with positive team performance, these results can contribute knowledge to the simulation setting from which the research emerged and help simulation educators to design opportunities for positively reinforcing practice.36 Furthermore, an awareness of the potential for some leadership behaviours to be more beneficial at certain points of a clinical care situation than others, can be a valuable discussion point in debrief conversations.
Strengths and limitations
This study had several strengths. First, all behaviour of all the participants was meticulously examined and coded frame by frame to uncover both the overt and nuanced teamwork behaviours and their patterns over time. Second, an index of simulated clinical performance was developed based on the scenario aims. Thirdly, sophisticated statistical methods (ie, GEEs) were employed to adjust for intrascenario correlations and accommodate time series data.
A limitation of this study was the use of video-recorded simulated clinical interactions. Although we do not know how the teamwork behaviour patterns identified in the current study equate to real-world behaviour, the scenarios were based on real-world incidents and were developed by a team of experienced clinicians from different professional backgrounds. A second limitation was that the data we gathered were not collected specifically for the purpose of teamwork analysis; this is a limitation shared by other studies examining human behaviour in naturalistic settings. A third limitation was sample size, due to the time and labour-intensive nature of the analysis. However, the data produced by the analysis are rich, resulting in a dataset of over 3600 behaviours. Finally, we were not able to control for the order of the analysed scenarios in the training day and this may have affected the findings.
Future research should focus on those teamwork behaviours that this research and the broader literature have highlighted as critical for clinical performance. There is also scope to further explore the methodological challenges of analysing episodes of clinical care. For example, the choice of time segments to compare is potentially important. Dividing scenarios into equal time periods, as in this study, is one approach, but there may be merit in determining whether there are natural transition points which delineate task phases. This is not feasible for all clinical interactions, including the ones used in this study in which there were no transition points. In other settings, however, this may be possible and would provide a nuanced understanding of how task demands are related to teamwork behaviour.
In line with recent reports and research,10 11 13 14 17 18 this study confirms the central role of teamwork in simulated clinical performance, but suggests that the opening moments of a patient care encounter may be important for establishing effective teamwork and performance. Shared leadership, and development of the shared mental model at the very start of the interaction are associated with better simulated performance, suggesting these behaviours are associated with teams working together effectively. This study develops our understanding of the behaviours that are potentially effective for interdisciplinary teamwork and the relationship of teamwork to simulated performance in these complex clinical scenarios, and provides evidence to inform clinical education of interprofessional teams.24 37
Data availability statement
No data are available. Due to ethical restrictions, the data cannot be shared.
Ethical approval for this work was provided by King’s College London ethics committee (RESCMR-15/16–1561) in October 2015.
The authors would like to sincerely thank the Simulation and Interactive Learning Centre at Guy’s and St Thomas’ NHS Foundation Trust for their support of this work, particularly the centre leads Dr Peter Jaye and Colette Laws-Chapman.
Twitter @marylavelle55, @gabereedy, @anitaobsmed, @ProfAndersonHF
Contributors JEA and GBR designed the study. TS and AB collected the data, assessed clinical performance and provided clinical perspectives during interpretation of the findings. ML conducted the observational and statistical analysis under supervision of GBR and JEA. ML drafted the first version of the manuscript. All authors contributed to the manuscript and approved the final version.
Funding This work was supported by direct and in-kind funding from the Simulation and Interactive Learning Centre at Guy’s and St. Thomas’ NHS Foundation Trust; Maudsley Simulation at South London and Maudsley NHS Foundation Trust; and Health Education England.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.