Objective Multidisciplinary team (MDT) meetings are assumed to produce better decisions and are extensively used to manage chronic disease in the National Health Service (NHS). However, evidence for their effectiveness is mixed. Our objective was to investigate determinants of MDT effectiveness by examining factors influencing the implementation of MDT treatment plans. This is a proxy measure of effectiveness, because it lies on the pathway to improvements in health, and reflects team decision making which has taken account of clinical and non-clinical information. Additionally, this measure can be compared across MDTs for different conditions.
Methods We undertook a prospective mixed-methods study of 12 MDTs in London and North Thames. Data were collected by observation of 370 MDT meetings, interviews with 53 MDT members, and from 2654 patient medical records. We examined the influence of patient-related factors (disease, age, sex, deprivation, whether their preferences and other clinical/health behaviours were mentioned) and MDT features (as measured using the ‘Team Climate Inventory’ and skill mix) on the implementation of MDT treatment plans.
Results The adjusted odds (or likelihood) of implementation was reduced by 25% for each additional professional group represented at the MDT meeting. Implementation was more likely in MDTs with clear goals and processes and a good ‘Team Climate’ (adjusted OR 1.96; 95% CI 1.15 to 3.31 for a unit increase in Team Climate Inventory (TCI) score). Implementation varied by disease category, with the lowest adjusted odds of implementation in mental health teams. Implementation was also lower for patients living in more deprived areas (adjusted odds of implementation for patients in the most compared with least deprived areas was 0.60, 95% CI 0.39 to 0.91).
Conclusions Greater multidisciplinarity is not necessarily associated with more effective decision making. Explicit goals and procedures are also crucial. Decision implementation should be routinely monitored to ensure the equitable provision of care.
- Chronic Disease Management
- Decision Making
- Health Services Research
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 3.0) 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 and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/3.0/
Statistics from Altmetric.com
Multidisciplinary team (MDT) meetings are central to the management of chronic disease and they have become widely established across the National Health Service (NHS)1–3 and internationally.4 Their purpose is to optimise decision making and improve outcomes.5 Evidence for their effectiveness, however, is mixed6 ,7 and the degree to which MDTs have been absorbed into clinical practice varies widely across conditions and settings.8 Research into the critical factors which impact on MDT effectiveness has identified clear leadership6 and explicit, shared goals with measurable outcomes.9 A positive team atmosphere or ‘climate’, defined as shared perceptions of policies, practices and procedures,10 has also been found to improve performance in some studies11 while others have found no association.12 Other fundamental determinants of the effectiveness of MDTs have not been established, and so, basic questions, such as the importance of ‘multidisciplinarity’ in practice, have not been addressed.
Uncertainty about some of the features associated with effectiveness has resulted in substantial diversity in the guidance provided regarding MDT meeting structure and processes.1 ,13 ,14 In the UK, policy for cancer and community mental health teams (CMHTs) lists the professional skills required. Cancer teams follow national standards and are audited to ensure they fulfil requirements including the routine recording of decisions.8 In other specialties, such as dementia and cardiovascular disease, the format is locally determined. It is unknown whether such flexibility is appropriate or undermines the productivity of meetings. Patient-related factors may also impact upon MDT decisions. There is evidence that the socioeconomically disadvantaged are less likely to receive healthcare15 despite the focus of NHS policy on equity.16 If patients’ sociodemographic characteristics influence MDT effectiveness, then explanations such as the influence of patient preferences and comorbidities should be sought.17
Given the widespread presence of MDTs, the opportunity costs for the NHS of unwarranted variations in team membership and processes and the consequences for patients of inequitable care, we undertook a prospective mixed-methods study of a range of chronic disease MDTs to investigate determinants of effectiveness.
There is debate in the literature concerning the most appropriate indicator of MDT effectiveness. Health outcomes may not be the most suitable measures because they are affected by factors other than the quality of care, including, for example, the extent to which patients adhere to treatment.18 ,19 Additionally, intended health outcomes may not be evident in the short or medium term, and many outcomes are disease specific, making it impractical to compare MDTs for different conditions. Thus, in this context, the use of a process measure, rather than an outcome measure, is appropriate.20
However, the identification of an indicator of high-quality decisions is also difficult. While some decisions can be compared with national guidelines, these seldom specify recommended courses of action for every management decision made for a patient. Additionally, guidelines rarely consider how decisions should be modified for patients with comorbidities or other factors which will influence decision making, such as their social circumstances or preferences. Finally, guidelines are not available for all the conditions considered in MDT meetings. We identified treatment plan implementation (as documented in patient notes), as an appropriate process indicator of effectiveness, because it lies on the pathway to health improvement and reflects effective team decision making which has taken account of relevant clinical and non-clinical information, including patient preferences. Another advantage of this measure is that it can be compared across all MDTs.
Our aim was to identify key characteristics of chronic disease MDT meetings associated with decision implementation. To achieve this, we addressed the following objectives: (1) to examine chronic disease MDT meetings to quantitatively identify factors influencing their effectiveness in terms of decision implementation and (2) to explore the reasons for any factors identified through semistructured interviews and non-participant observation of meetings.
We undertook a prospective cohort study of MDT meetings in 12 chronic disease adult MDTs across London and North Thames between December 2010 and December 2012. We aimed to include teams which varied in terms of healthcare context, professional mix of team, conditions which affected patients from different parts of the adult life range, and fatal versus life-long conditions. We examined one skin cancer team, one gynaecological cancer team, two haematological cancer teams, two dementia teams, two heart failure teams, and four community mental health teams (including one early intervention service for psychosis).
Data were collected from three sources:
Source A: We undertook non-participant observation and audiotaped 370 weekly MDT meetings over 18–55 weeks (depending on patient numbers discussed at each meeting). Each team was observed for 2 weeks before data were collected in order to accustom team members to the observer's presence. For each patient discussed, we collected quantitative data on decisions made and diagnosis. We also recorded whether health behaviours (smoking, drinking, physical inactivity), other clinical factors (including comorbidities, medical and family history) and patient/carer treatment preferences were mentioned. For the logistic regression, these categories were collapsed into ‘health behaviours or other clinical factors’ (mentioned or not mentioned) and ‘patient/carer preferences’ (mentioned or not mentioned). A decision was defined as a resolution or recommendation about patient management.
After each meeting, field notes were written on MDT features (eg, agenda, attendee job titles, presence of an MDT coordinator), context (eg, reference to national policy/guidance and local resource issues) and process (eg, levels of participation, role of the chair, complexity and uncertainty of problem, and clarity and documentation of decisions).
Core MDT members (as formally defined in MDT documents, or by the clinical lead, and excluding observers and occasional visitors) completed the Team Climate Inventory (TCI) during the final month of observation.10 This validated measure21 assessed members’ perceptions across the four domains of ‘team vision’, ‘participative safety’ (ie, a facilitative atmosphere for involvement), ‘task orientation’ (eg, with respect to individual and team accountability) and ‘support for innovation’. A low TCI score reflects perception of poor team climate.
Source B: We conducted semistructured interviews with 53 MDT members recruited from all the MDTs observed, purposively sampled to include core professional groups and frequent and infrequent attendees. Topics discussed included perceptions of the purpose and value of MDT meetings as well as the individual's role and value within the MDT meeting.
Source C: Quantitative data on decision implementation, reasons for non-implementation and patients’ sociodemographic (gender, date of birth, and postcode) and diagnostic details were collected from medical records. Decision implementation was assessed 3 months after the MDT meeting unless the MDT explicitly recorded that implementation should be later (eg, ‘follow up in six months’).
The data were categorised and scored for analysis using the following summary measures:
Treatment plan (the unit of analysis)
We analysed each patient's first presentation at an MDT meeting within the observation period. This often resulted in more than one decision being made. These could not be assumed to be independent, and so, decisions were grouped into a ‘treatment plan’. Decisions were classified as ‘not implemented’ if there was no record in the notes indicating decision implementation. In the majority of cases (88%) implementation of the treatment plan was fully determined (by all component decisions being either implemented or not). In cases where some decisions in the plan were implemented and others were not, we classified a treatment plan as implemented if more than 50% of the component decisions were implemented. This definition was agreed to be reasonable by our study team but was investigated further in sensitivity analyses.
2. Team Climate Inventory
Responses for each item on the TCI are given on a 5-point Likert scale. The overall team TCI score was obtained by averaging the scores of team members. We excluded TCIs with more than 25% of items missing in one of the TCI dimensions. A unit increase in TCI score indicates an improvement in team climate (eg, a change from an average response of ‘agree’ to ‘strongly agree’).
3. Skill mix (Adjusted Teachman's Index (ATI) and number of professional groups)
We used the ATI22 together with the number of professional groups represented (classified as diagnostic medical practitioner, surgeon, physician, MDT coordinator, nurse, researcher, social worker, allied health professional and psychologist) to measure skill mix. See online supplementary appendix Section 1. A higher ATI value reflects a more even spread across a greater number of groups.23 Use of both measures allowed consideration of whether the more complex index added predictive value over a simple count of professional groups. This was measured at each individual meeting, and core and non-core staff were included. Observing students were excluded.
4. Index of Multiple Deprivation
We used the Index of Multiple Deprivation (IMD) 2010 to measure socioeconomic circumstances.24 This is a widely used area-based measure that combines seven domains (income, employment, health and disability, education, skills and training, barriers to housing and services, and crime and living environment) into a single deprivation score for each small geographical area (ie, each ‘lower layer super output area’ (LSOA), covering about 1500 people) in England. We grouped IMD scores into quintiles, with higher quintiles denoting more deprivation. The postcode address of each patient was linked to the LSOA and the corresponding deprivation quintile.
The influence of MDT and patient-related factors on treatment plan implementation was investigated using random effects logistic regression models, allowing for clustering by MDT. We fitted unadjusted models for each factor of interest. We then undertook the following predefined selection process to obtain a final adjusted model. Initially, we fitted two separate models. Model 1 included patient characteristics (age, gender and IMD quintile), and Model 2 considered MDT and decision-related characteristics (TCI score, disease type, ATI score, number of professional groups, mention of patient preferences and other clinical factors (defined as health behaviours, comorbidities, medical and family history). Factors identified as having potential importance from these models (p<0.3)25 were then fitted in the final model (Model 3).
The suitability of models was investigated, including considering normality of the random effect, goodness of fit (Hosmer Lemeshow) and evidence of over-fitting (using bootstrapping). We examined the pattern and extent of missing data and considered its potential impact on our results. We also undertook exploratory investigations to understand further the associations observed in our model and whether these differed by disease type. We extended Model 3 to examine interaction terms between the number of professional groups and disease type and between IMD and guideline-driven cancer compared with non-cancer MDTs.
Planned sample size
The sample size calculation was based on decision implementation, assuming one decision for each patient.
As patients could have more than one decision when presented at an MDT, and these could not be assumed to be independent, we grouped these decisions to consider implementation of their overall ‘treatment plan’ for analysis. Where patients were discussed at more than one MDT, we analysed the treatment plan from their first presentation during the observation period.
We calculated the sample size using the conservative assumption (based on published research and clinical experience) that 18% of decisions would not be implemented.26 Using Peduzzi's rule of 10, we required 80 non-implemented decisions to estimate eight coefficients in our regression model; this would require approximately 440 patients.27 To allow for clustering by MDT, the sample size was inflated using estimates of the intracluster correlation coefficient (ICC) and the average cluster size. Directly relevant estimates for the ICC were not available, however, published models associating the ICC and outcome prevalence for a range of outcomes in community and health services settings suggested an ICC between 0.01 and 0.05.28 Thus, we assumed an ICC of 0.025, and that across MDTs the average number of patients with relevant decisions during the study would be approximately 230. Based on this information, we aimed to include 3000 patients with a decision from the 12 MDTs.
Qualitative observational and interview data were used to explore possible reasons for variations in implementation across teams and disease types. Our approach used deductive themes based on the main quantitative findings and inductive coding to expand on the deductive codes and to explore relevant emerging issues.
For the interview analysis, all transcripts were coded, checked and analysed in a constant comparative manner by three field researchers using NVivo software (V9). For the observational analysis, we coded field notes for the first 16 meetings observed for each of the 12 MDTs (excluding 2 weeks of pilot observation with each team) and selectively transcribed portions of these. Having listened to all 370 meetings at least twice, the researchers were confident that the 192 meetings coded in this way were representative.
Having coded approximately 1/3 of the data, we found that our coding framework became stable. During analysis of the remaining data we did not need to add further codes to describe the data in relation to our aims and develop a meaningful and useful interpretation, suggesting a sufficiently high level of saturation.29
Analytic meetings involving the wider research team were held to review, refine and cluster the codes into organising themes, resolve disagreements and ensure consistency of interpretation. Conducting qualitative analysis as a team safeguarded against bias, and facilitated a richer conceptual analysis and interpretation.30 Qualitative data were then triangulated to examine the consistency and internal validity of findings, identifying patterns of convergence in order to develop a coherent framework for understanding the results overall. Further details are provided in the online supplementary appendix Section 2.
We observed 370 MDT meetings during which 3184 patients were discussed. There was considerable variability among the 12 teams in the number of patients discussed at each meeting. The average for each team ranged from 4 to 49 patients. The meeting size across the 12 MDTs ranged between 5 and 28 members, with 1–7 professional groups represented (a median of 4).
Quantitative results (objective I)
Overall, 2654 patients had a treatment plan with an average of 1.6 decisions per plan. Two thousand five hundred and twelve (95%) patients had adequate implementation data recorded, of whom 1967 (78%) had implemented plans. Implementation was highest in the gynaecological cancer team and lowest in the CMHTs (table 1). There was a trend for non-implementation with increasing patient deprivation.
Adjusted associations (table 2) showed no evidence of a relationship between treatment plan implementation and patient age or gender, or with discussion of patient preferences or health behaviours/other clinical factors. ATI and number of professional groups were highly correlated (correlation coefficient 0.8) and did not appear to have an association when fitted together (Model 2). When included individually in the model, however, each showed a relationship such that increases in team diversity or number of professional groups were associated with a reduced odds of treatment plan implementation (table 2 footnote).
Factors with an association (p<0.3)25 in Models 1 and 2 were fitted together in the final model (Model 3). The adjusted odds of implementation almost doubled (OR 1.96 (95% CI 1.15 to 3.31) for a unit increase in TCI score (indicating improved team climate). By contrast, the adjusted odds of implementation was reduced by 25% for each additional professional group represented and was lower for patients living in more deprived IMD quintiles. Adjusted odds of implementation also varied by disease type.
We had sufficient data on 2512 (92%) patients to include them in the final model. Data were missing for age (2%), IMD quintile (6%) and treatment plan implementation (5%), mainly because of missing patient notes. The only characteristic associated with missingness was MDT disease type. Given the small proportion of missing values, we did not consider it necessary to account for this in our analysis.
Sensitivity analyses for Model 3 included: adjusting for the number of decisions making up the first treatment plan (which varied between 1 and 11 decisions); examining the impact of using the first recorded treatment plan in analysis by refitting the model based on a randomly chosen treatment plan for each patient; collapsing the 5 IMD quintiles to two groups (IMD1-3 and IMD 4–5) to produce an 8 coefficient model; refitting the model with implemented treatment plans redefined as those with greater than 80% of component decisions implemented (a change which had little impact on the overall implementation rate reducing it from 78% to 74%); and refitting the model excluding the CMHTs (which made up a large proportion of the teams in the study). None of these sensitivity analyses substantially changed our conclusions.
We further explored the observed trend between implementation and number of professional groups to consider whether this might be dependent on disease type. Tabulations showed that for dementia teams and CMHTs this trend was apparent. For the other disease types, however, such a trend was not obvious. An interaction term added to Model 3 indicated some evidence of a differential effect by type of disease (p=0.06), however, we note that this exploratory result should be interpreted cautiously, as for some diseases, there was limited variability in the number of professional categories represented. We also considered whether the relationship between IMD and implementation differed for cancer and non-cancer specialties. In all cases, the observed general trend was for less implementation in more deprived groups. An interaction term included in Model 3 was not significant (p=0.13).
The reasons for non-implementation of treatment plans are summarised in table 3. Patient-led reasons were recorded most frequently in CMHTs (eg, due to patient non-attendance), while patient death was reported most frequently for heart failure MDTs (see online supplementary appendix. Section 3 summarises reasons for non-implementation by disease type). We also collapsed the 5 IMD quintiles to two groups (IMD1-3 and IMD 4–5), and found that reasons for non-implementation were similar, although patient death was a more frequent reason for non-implementation in the least deprived group (data not shown). The ‘other’ reasons were diverse, and included cases where the decision was implemented outside the 3-month follow-up period, or when new information or test results emerged after the MDT meeting. This category also included cases where it was subsequently decided that a referral had been inappropriate or that a patient should have been referred to a team in a different catchment area.
Qualitative results (objective II)
Features of teams with higher rates of treatment plan implementation
High implementing MDTs, regardless of disease type, had clear goals and more frequently referred to diagnostic or treatment protocols and national guidelines. All four cancer teams adhered closely to national guidelines, and so, had the dedicated administrative support members considered to be essential for meeting preparation and facilitation. High implementing teams also tended to have permanent chairs who maintained a focus on decisions, which were clearly recorded. They also assembled patient lists before meetings. This stimulated case presenters to consider the management options they wished to discuss before rather than during the meeting and led to more explicit decisions. The other MDTs included none or some of these features, and this varied by team rather than by disease type.
Features of teams with lower rates of treatment plan implementation
Those dementia and CMHT meetings with larger numbers of professional groups and lower implementation tended to discuss a more diverse range of issues (from medication to social issues, housing and benefits), and also used meetings to provide updates on ongoing cases and peer support. There was some evidence that these teams lacked clarity of purpose:
I am never quite sure what the purpose of the meetings are (psychiatric nurse, interview),
and were less focused on specific questions:
Thanks for sharing that story. Why did we hear it? (psychiatrist, observation).
Decision documentation was inconsistent, and CMHT members frequently discussed difficulties in engaging patients with the service which also contributed to the low implementation in these teams.
Chairing varied considerably between teams. Most meetings were formally chaired: in cancer teams, one dementia team and two CMHTs, this was a designated person who chaired the meeting each week (either a consultant or the team manager). In two mental health teams, the chairing role rotated between team members on a weekly basis. The remaining three teams did not have a predefined chairing system, and different senior members took the lead on different occasions, sometimes changing during a meeting.
Team members reported that they valued hearing a range of disciplinary perspectives, however, not all disciplines were perceived to have an equal ‘voice’. Cancer and heart failure teams were characterised by a strong medical dominance in terms of attendance and participation. While these teams typically made and implemented high numbers of treatment plans, those plans were less likely to have incorporated the full range of disciplinary perspectives:
It's rarely a multidisciplinary meeting…it often consists of a surgeon talking to the radiologist (gynaecological cancer oncologist, interview).
CMHT and dementia teams were characterised by more even levels of participation across the disciplines, however, interviews revealed that differences in the ways that professional groups approached problems could lead to difficulties:
They're in two camps… social workers very much do try to defend, you know, a person's right to be mad. I mean the nurse view is ‘you're not well so we need to get you on the path and let’s make the decisions you would make [if you were well]’ (psychiatric nurse, interview).
We explored the roles that patient preference, health behaviours and other clinical factors may have in explaining the socioeconomic variations in implementation that we found. However, most members considered that the extent to which patient behaviours and other factors were discussed was appropriate, although there was marked team variation in discussion of these factors. With regard to patient preferences, some argued that these were central to MDT decisions while others tended to elicit patient preferences after the MDT, when treatment options could be shared with them.
Further quotes illustrating these findings are provided in Section 4 of the online supplementary appendix.
In this large study, the first to include MDTs for different conditions, we demonstrate the implications of MDT diversity for the NHS and patient care. We have shown that a good team climate and less diverse membership in the MDT were the most important predictors of treatment plan implementation across teams. There was variation within and between disease types with gynaecological cancer having the highest rates of treatment plan implementation, and CMHTs the lowest. There was less implementation for patients living in more deprived areas. Lack of implementation was commonly due to patient or family choice, and to difficulties in engaging patients with the service. The better implementing teams tended to have dedicated administrative support and clear goals and processes.
Contrary to previous reports and expectations, greater multidisciplinarity is not necessarily associated with more effective decision making. Previous research suggests that each team member brings relevant information for the development of a cohesive care plan.31 However, diversity may raise communication barriers due to differences in knowledge, skills and ability,9 misunderstandings of each other's roles,32 and identifying with subgroups rather than with the team as a whole.33 Our research suggests that the presence of many professional groups needs to be mediated by a clear purpose and agreed processes. This is supported by evidence which points to the benefits of shared objectives to guide and structure communication, focused leadership, and team reflexivity.9 ,34
Second, our results concur with others which demonstrated that a good team climate does translate to better decisions.11 It had been suggested that transformational leadership affects team performance (better decisions) by creating a positive climate.35 It is also possible that the mixed results in the literature are explained by the different outcome measures used to define effectiveness.12
Finally, patients from more deprived areas are less likely to have their treatment plans implemented, and this occurred despite the routine reference to treatment guidelines by cancer teams. Consideration of patient preference, comorbidities or other health-related factors do not seem to explain this.
Previous research reporting socioeconomic inequalities in use of adjuvant therapy for cancer suggests that non-implementation equates to non-uptake of care for cancer patients.36 This is likely to partly explain the continuing socioeconomic variations in cancer survival.37 There is little research on treatment inequalities for people with dementia, although a small study indicates that these may occur.38 Recent models demonstrated equitable provision of cardiovascular care which translated to similar proportions of deaths averted across deprivation quintiles.39 However, non-implementation may result in care delays which might impact upon morbidity rather than mortality—something that was not addressed in these cardiovascular models.
We recognise that the use of decision implementation as an indicator of effectiveness does not address the issue of whether or not the ‘right’ decisions have been made. It is possible that poor decisions are implemented. In order to examine whether implemented decisions are ‘good decisions’ that do indeed lead to patient benefit, it would be necessary to analyse carefully identified disease-specific health outcomes which can be plausibly and directly related to the MDT decision. Alternatively, a subset of decisions about which there are specific best practice guidelines could be assessed according to whether they conform to this guidance. However, many MDT decisions cannot be derived directly from treatment guidelines, and the detail provided in such guidelines varies by specialty.
Additionally, the factors we considered in our analyses were identified as potentially important based on previous research. We appreciate that there may be other, unmeasured factors that could also have influenced implementation. Furthermore, decision making is only one of a range of functions for some MDT meetings, and in particular for the CMHTs we studied. Our outcome measure was not designed to assess how well these additional functions were performed.
When deciding which and how many chronic diseases to include, in addition to the criteria described above, we were also bound by pragmatic imperatives to ensure feasibility of data collection. We recognise that this has implications for the generalisability of our results, in terms of applicability to other chronic conditions and with respect to the similarity of the MDTs we examined in comparison to other MDTs for the conditions under study. For this reason, we included more than one MDT wherever possible. Online supplementary appendix Sections 5 and 6 report team and meeting characteristics to assist readers in assessing the extent to which our results are likely to be applicable to other teams.
We combined decisions where implementation was ‘not documented’ with those decisions which were ‘not implemented’. This was because, on the basis of our clinical experience, non-implementation is commonly not explicitly recorded in patients’ records. Our conclusions are, therefore, limited by the accuracy of record keeping.
There is no consensus on the best measure of skill mix or team diversity.23 We therefore categorised professionals into coherent groups on the basis of our clinical expertise. We used a well-established area measure of socioeconomic characteristics, commonly used where individual-level indicators are not available. However, this method rests on the assumption that individuals conform to the socioeconomic profile of their residential area. We recognise that misclassifications can cause underestimates or overestimates of the relation between socioeconomic circumstances and implementation.
We did not achieve the target sample size of 3000 individual patients with treatment plans. This is because fewer patients than anticipated had treatment plans agreed at each meeting. However, even with our smaller sample, investigation into model overfitting did not raise concerns.
We did not collect cost or waiting time data, and so, cannot comment on the implications of MDT diversity with respect to MDT cost effectiveness or consequences for patients.
It is possible that team members behaved differently when under observation, but this was mitigated by the 2 weeks of observation before data were collected. Additionally, MDT members were aware that decision making was being studied, but not that implementation was being examined.
MDTs are widespread across the NHS, but they are resource intensive, and so their value to the NHS and patients should be maximised. A team atmosphere that facilitates involvement, task orientation and support for improvement is key to ensuring team decisions are implemented. The benefits of multidisciplinarity per se should not be assumed. However, clear goals and procedures are crucial, and we also recommend routinely monitoring decision implementation to ensure the equitable provision of care.
We are indebted to all the North Thames MDT teams and patients who participated in this research and made it possible, and to the NHS staff who assisted with recruitment and access to medical records. We would like to thank Professor Ewan Ferlie for his advice, and contribution to the study design. We are grateful to Dr Sophie Bostock who assisted during the inception of the project, to Dr Khadija Rantell who assisted with data cleaning, and to Dr Mike Galsworthy who assisted with database building and cleaning and preliminary analysis. We would also like to acknowledge the support of the National Institute for Health Research, through the Comprehensive Clinical Research Network. This study was funded by the NIHR Health Services and Delivery Research Programme.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors RR, AL, AP, MK and JMB conceived the idea for this study. RR led the funding application and all components of the study. She prepared the first draft of the paper with assistance from JB, CN, IW and PX. All other authors contributed to subsequent drafts and approved the final draft. CN, IW and PX contributed to study design, review of the literature, recruitment of participants, data collection, analysis and interpretation, assisted in writing the first draft of the paper, and provided critical review of subsequent drafts. JB collaborated in the funding application and was responsible for statistical aspects of the study, including involvement in the study design, analysis, and interpretation. JMB, AC, SG, MK, AL, GL, SM and AP collaborated in developing the idea for the research and in the funding application. They contributed to study design and provided clinical expertise throughout the project. DA, MH, AC, AL, GL and AP contributed to participant recruitment. AL and AC contributed to the qualitative analysis and interpretation of the findings. DA and MH contributed to study design, and provided guidance from a patient perspective throughout the project. NA-P assisted with the literature review and data collection. As the chief investigator and guarantor, RR managed the study overall and had final responsibility for the analysis and manuscript content.
Declaration of transparency The lead author affirms that the manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study, as planned, have been explained.
Funding This project was funded by the NIHR Health Services and Delivery Research programme (project number 09/2001/04) and will be published in full in the Health Services and Delivery Research Journal. Further information available at: (http://www.netscc.ac.uk/hsdr/projdetails.php?ref=09-2001-04). In accordance with NIHR regulations, the findings reported here are to be referred to as initial until the NIHR final report has been published. The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, the writing of the report, or the decision to submit the report for publication.
Department of Health disclaimer This report presents independent research commissioned by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, MRC, CCF, NETSCC, the Health Services and Delivery Research programme or the Department of Health.
Competing interests All authors have completed the ICMJE uniform disclosure form at (http://www.icmje.org/coi_disclosure.pdf) (available on request from the corresponding author) and declare: support from the organisation described below for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years, and no other relationships or activities that could appear to have influenced the submitted work.
Ethics approval East London Research Ethics Committee (10/H0704/68) and NIGB (ECC 6-05 (h)/2010).
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.