In recent years there has been growing interest in the use of patient-reported outcome measures (PROMs) and patient-reported experience measures (PREMs) in healthcare settings. Such usage includes routinely collected measures for the evaluation of services such as the National Health Service (NHS) in the UK, although there are challenges concerning their effective implementation. Reference Bull, Teede, Watson and Callander1–Reference Bull and Callander3 PROMs focus on targeted outcomes Reference Gelkopf, Mazor and Roe4–Reference Nelson, Eftimovska, Lind, Hager, Wasson and Lindblad6 whereas PREMs focus on the experience of receiving the service. Reference Jamieson Gilmore, Corazza, Coletta and Allin7 PROMs can focus on individual conditions and/or symptoms, or on the whole person, measuring either their recovery or quality of life.
DIALOG
DIALOG is an 11-item scale that measures patient satisfaction across 11 domains. Reference Priebe and Bird8–Reference Bhattacharya, Priebe and Bird11 The first eight items measure satisfaction with different life outcomes (PROMs), whereas the final three measure satisfaction about treatment/support received (PREMs). DIALOG therefore encompasses both PROM and PREM components, in a manner intended to make routine patient–clinician meetings more effective. Reference Priebe, McCabe, Bullenkamp, Hansson, Lauber and Martinez-Leal12,Reference Priebe, Golden, McCabe and Reininghaus13 The DIALOG scale forms the basis of a wider care-planning tool, called DIALOG+, which is an intervention to support patients and clinicians in co-producing a care plan focusing on each of the quality-of-life domains, with the relative importance of domains determined by the patient using the DIALOG scale. Reference Jubokowa9,10,14 For DIALOG+, in addition to rating their satisfaction for each domain using the DIALOG scale, patients are invited to discuss goals and co-produce an action plan with the clinician, based on the principles of solution-focused therapy, to enable an improvement in the domain being discussed. Reference Jubokowa9,15
East London NHS Foundation Trust (ELFT) was one of the first NHS trusts to incorporate DIALOG and DIALOG+ care planning as a part of their Care Programme Approach, soon after the development of the tool in 2017. Reference Mosler, Priebe and Bird16 The use of DIALOG (including DIALOG+) spread across a number of NHS trusts, including its acceptance as the preferred PROM for adult mental health services in London. 17 The community mental health (CMH) transformation was proposed and implemented as a part the NHS Long Term Plan (2019), 18 and was intended to improve care pathways with locally integrated multidisciplinary teams, using whole-person and whole-population health approaches. DIALOG+ co-produced care planning resonated with the CMH transformation care planning ethos. Subsequently, NHS England identified DIALOG as one of the three recommended outcome measures for CMH. 19 NHS England stated that services should work towards the routine use of DIALOG+ to support care planning, and of DIALOG for ongoing monitoring within mental health services. This dual purpose – outcome measurement and care planning – has advantages for implementation and reducing clinical burden, but potentially complicates interpretation of outcome scores.
Existing evidence
Although some studies have been conducted on DIALOG and DIALOG+, there is only limited understanding of their broader utility in the evaluation of mental health services through routinely collected data. For example, several trial-based studies have looked at DIALOG in populations with psychosis and severe mental illness. Reference Omer, Golden and Priebe20–Reference Priebe, Kelley, Omer, Golden, Walsh and Khanom22 A separate study from ELFT – studying the period from January 2017 to December 2019 – evaluated routinely collected DIALOG data and found a trend of improving scores over time. Reference Mosler, Priebe and Bird16 However, this study analysed DIALOG scores over the course of treatment (using five time points) but did not distinguish the beginning and end of the patient journey (e.g. in terms of initial assessment or interim scores captured during review or at discharge). In our study, we took a different approach and looked at changes in pooled DIALOG scores through the stages of treatment (assessment, review and discharge), to capture a clear beginning and end to the patient journey at the population level.
Additionally, we aimed to compare routinely collected data pre- and post-CMH transformation, to assess their impact on DIALOG scores. This also offered us the opportunity to explore whether the pooled DIALOG scores detected any changes in the needs or satisfaction levels of the population pre-treatment (i.e. at the assessment stage before and after the COVID-19 pandemic).
Current study
We analysed anonymised, routinely collected data from RiO, ELFT’s electronic healthcare record system. We evaluated two time periods (financial years 2018–2019 and 2021–2022), capturing data pre- and post-CMH transformation. The CMH transformation was implemented in autumn 2019. These two time periods were selected to avoid data collection challenges during the height of the COVID-19 pandemic lockdown restrictions. We evaluated the data by conducting a quantitative pre–post observational analysis of the mental health service data from RiO. Apart from comparing the two time periods, the evaluation allowed us to understand whether DIALOG scores had changed over time and along the patient journey. We observed 7568 DIALOG scores in this study; this large number resulted, in part, because ELFT was an early adopter of DIALOG as both a PROM/PREM and a care-planning tool.
Our evaluation explored patient outcomes and experiences over time (in terms of both financial year and treatment stage) while controlling for several other variables (including demographic variables and protected characteristics: age, ethnicity, gender and index of multiple deprivation). This also helped us to understand the strengths and limitations of DIALOG as a routinely collected measure embedded within services.
This evaluation required close collaboration with ELFT clinicians, managers, patients, carers and data analysts to develop the analyses plan, understand data availability and quality, pathways for data input and procedures for data governance. This paper reports and discusses the findings related to treatment stage and financial year. A separate linked paper will report and discuss the results related to demographic variables and protected characteristics. This separate, linked paper will include the ways in which findings relate to the patient and carer race equality framework. 23
Method
ELFT commissioned the University of Plymouth to evaluate routinely collected community mental health team (CMHT) data, to assess the impact of the CMH transformation 24 as recommended by NHS England. As a part of the service evaluation, routinely collected DIALOG scores from CMHT services were analysed. ELFT’s business analysis team carried out a search of electronic patient records stored in RiO for DIALOG scores recorded for the identified patient group within the periods under investigation.
Design and data sources
Our evaluation used a quantitative pre–post observational design, with two cross-sectional time periods (financial years 2018–2019 and 2021–2022). The purpose of this evaluation was to assess population-level scores and changes in pooled outcome or quality-of-life measures and experience, rather than changes in individual patients. As part of the CMH transformation there was a greater focus on integration with primary care, and the post-transformation community teams had a broader scope, including what was previously the remit of primary care liaison teams. DIALOG scores from CMHTs and primary care liaison teams for 2018–2019, and from the ‘transformed’ community or neighbourhood mental health team in 2021–2022 (which offered the function of both the previous teams), were considered in scope. The patient group were all adults aged 18 years and above by the start of the first financial year (2018–2019) and with at least one DIALOG score in one of the two study periods (regardless of when they were first assessed or finally discharged).
The data were collected from three London boroughs serviced by ELFT: City and Hackney, Tower Hamlets and Newham. We analysed 11 198 DIALOG scores, 5294 for 2018–2019 and 5904 for 2021–2022. The number of unique patients (i.e. individual patients) in the study was 5007; 2693 unique patients were analysed in 2018–2019 and 3161 in 2021–2022 (because some patients were present in both data periods, the total number of unique patients is lower than the sum of unique patients in each financial year).
Of note, not all domains were filled in at each submission. We defined ‘stage of treatment’ as ‘assessment’ for new referrals, ‘review’ for ongoing treatment and as ‘discharge’ for end of treatment. We pooled analysis by stage of treatment at the DIALOG domain level for some analyses.
Measures
We compared pooled, pseudo-anonymised DIALOG scores along with protected characteristics and other demographic variables (age, ethnicity, gender and index of multiple deprivation (IMD) decile). We analysed data within and across the two time periods described above, and by stage of treatment (assessment, review and discharge).
Materials
The DIALOG scale is presented in Table 1. Each DIALOG domain item is scored using a Likert scale ranging from 1 (totally dissatisfied) to 7 (totally satisfied).
DIALOG scale. The items are scored on a scale of 1–7, where 1 is totally dissatisfied and 7 is totally satisfied
Analyses
Two types of analyses were carried out using the R statistical package (2024 version) for Windows and Ubuntu (R Core Team, The R Foundation, Vienna, Austria; https://www.r-project.org/). 25 Because the analytical framework Reference Bhattacharya, Priebe and Bird11 does not specify any statistical analyses of pooled DIALOG data, we developed the following approach as part of our evaluation.
-
(a) We performed descriptive statistical analyses of pooled DIALOG scores, including means and 95% confidence intervals, across a range of variables for each DIALOG domain, financial year and stage of treatment. We report means and CIs for the two financial years and three stages of treatment.
-
(b) We used multiple logistic regressions on the DIALOG domains, where we converted the raw scores into a binary variable of ‘satisfied’ (scores of 4–7) and ‘dissatisfied’ (1–3). Thereafter, for each DIALOG domain we estimated a multivariable logistic regression, with the odds of reporting a ‘satisfied’ DIALOG domain score as the outcome variable and the following explanatory variables: stage of treatment, financial year, age, ethnicity, gender and IMD decile (we cross-referred DIALOG scores from individuals with the IMD scores generated from postcode information, using openly available government data 26 ). The interaction between treatment stage and financial year was also analysed. If models including the interaction term performed significantly better at explaining the data than those without, the former were selected and reported – otherwise the latter were selected and reported. The model selection was conducted using analysis of deviance significance tests.
There was clinical importance in understanding whether patients had moved from being dissatisfied to satisfied (or vice versa) over the course of the patient journey from assessment to discharge. Logistic regression measures the odds of these satisfaction changes at the population level. This provides a more clinically useful measure than more abstract changes in level of satisfaction along an ordinal scale, while also being a more statistically strict form of analysis. However, the original raw DIALOG scores are also reported in the descriptive statistics (means and 95% confidence intervals), to provide a full report of the data-set. The 95% confidence intervals provide an indication of where mean scores are significantly different. Additional significance testing on the descriptive statistics was deemed redundant, because the logistic regressions provide a more statistically sophisticated and robust analysis of potential effects.
Results
We analysed routinely collected data from 5007 patients with at least one DIALOG score and reported at least in one domain: a total of 11 198 DIALOG assessments across both years. This was then split, because 2693 patients had a total of 5294 assessments in financial year 2018–2019 and 3161 had a total of 5904 in financial year 2021–2022. The DIALOG scores were collected routinely for patients receiving adult CMH treatment from the CMHT and primary care liaison teams in the east London boroughs of City and Hackney, Tower Hamlets and Newham in 2018–2019, and the ‘transformed’ CMHTs in the same boroughs in 2021–2022, with the care being delivered by ELFT. Routinely collected DIALOG scores linked to treatment stage were obtained from electronic patient records, anonymised and pooled for the analysis.
First, we pooled both time periods to analyse differences in scores across stages of treatment. Figure 1 shows that DIALOG scores across the board tended to improve with duration within the service; this was true both pre- and post- pandemic, and is statistically significant (according to 95% confidence intervals).
Mean DIALOG scores for each domain, split by treatment stage across 2018–2019 and 2021–2022 combined. Error bars are 95% confidence intervals; bars with non-overlapping confidence intervals can be interpreted as significantly different. There were n = 4193 sets of DIALOG assessment scores, n = 6764 sets of DIALOG review scores and n = 240 sets of DIALOG discharge scores. MH, mental health; PH, physical health; JS, job situation; AC, accommodation; LA, leisure activities; RS, relationship with partner/family; FS, friendships; PS, personal safety; MD, medication; PR, the practical help you receive; MP, meetings with mental health professionals.
These results show that patient satisfaction improved over the categorical stages of the patient journey, from assessment (time of referral), through reviews (mid-treatment) and, finally, to discharge (end of treatment). Because collection of DIALOG scores at discharge was carried out less routinely, the number of discharge scores (240) was lower than that at both assessment (4193) and review (6764). This may also indicate that most patients remained within the services during the evaluation period.
Figure 2 shows analyses of these same scores by year, allowing for comparison of pre- and post-CMH transformation. As in Fig. 1, Fig. 2 shows a trend of increasing satisfaction by treatment stage, but also some marginal evidence of a decrease in satisfaction from pre- to post-transformation. We identified an increase in mental health need in those referred to ELFT services from the community in 2021–2022 compared with 2018–2019, as evident from poorer satisfaction with mental health at the time of referral. There was also an apparent reduction in satisfaction for physical health at assessment from 2018–2019 to 2021–2022. However, these differences were not significant at the 95% confidence interval level. The small number of observations at discharge increases uncertainty in the estimates, such that differences in mean scores at discharge were not statistically significant between the two time periods. Additionally, these results suggest an increase in satisfaction post-transformation for people in contact (reviews) with ELFT mental health services (for mental health, physical health, leisure activities, friendships, personal safety and medication).
DIALOG scores by stage (assessment, review and discharge) and year (2018 refers to 2018–2019 and 2021 refers to 2021–2022). Error bars are 95% confidence intervals; bars with non-overlapping confidence intervals can be interpreted as significantly different. In this subgroup, there were n = 1515 and 2678 sets of DIALOG assessment scores in 2018 and 2021, respectively, n = 3643 and 3121 sets of DIALOG review scores in 2018 and 2021, respectively and n = 136 and 104 sets of DIALOG discharge scores in 2018 and 2021, respectively. See Fig. 1 for list of abbreviations used.
Figures 1 and 2 present descriptive univariate analyses, and do not address potential confounding variables. Below, we present results from multivariable regressions that estimate DIALOG domain scores, controlling for age, ethnicity, gender and index of multiple deprivation. For ease of interpretation, the regression results are split across three separate figures (see Supplementary Fig. 1 for a combined plot). Figure 3 reports the regression results for discharge compared with assessment, comparing 2021–2022 scores against 2018–2019.
Results of multiple regressions on DIALOG scores by year (2018 refers to 2018–2019 and 2021 refers to 2021–2022) and treatment stage. Higher values indicate higher odds of satisfaction on each DIALOG domain. Whiskers are 95% confidence intervals; if whiskers are >1, this variable is significantly associated with greater odds of being satisfied; and if <1, associated with lower odds of being satisfied; intermediate values are not significant. Predictor variables are all labelled to show what they are being compared against, e.g. ‘2021 v. 2018’ means that 2021–2022 is the predictor variable level and 2018–2019 is what we are comparing it against. See Fig. 1 for list of abbreviations used.
Results controlling for demographic characteristics are largely consistent with the findings from descriptive univariate analyses by treatment stage. The odds of patients being satisfied were higher at discharge than assessment, for all DIALOG domains apart from personal safety. However, for most variables there was no difference in satisfaction between the two time periods, although we did observe a statistically significant increase in satisfaction with friendships and personal safety between 2018–2019 and 2021–2022.
Figure 4 shows that the odds of satisfaction were higher at review (midway through treatment) compared with those at assessment (initiation of treatment) for all 11 DIALOG domains.
Results of multiple regressions on DIALOG scores by treatment stage. Higher values indicate higher odds of satisfaction on each DIALOG domain. Whiskers are 95% confidence intervals; if whiskers are >1, this variable is significantly associated with greater odds of being satisfied; and if <1, this is associated with lower odds of being satisfied; intermediate values are not significant. See Fig. 1 for list of abbreviations used.
We also estimated models, including an interaction term for year × treatment stage, to test whether differences across treatment stage exhibited different patterns across the different years. There were only three DIALOG domains where the regression model that included an interaction term, between treatment stage and year, performed significantly better than the regression model with no interaction term (i.e. explains significantly more of the data). These are reported in Fig. 5. The results show a significant negative interaction between review (versus assessment) and year (2021–2022 versus 2018–2019) for friendships, practical help received and meetings with mental health professionals. This means that, in 2021–2022 there was a reduced increase in satisfaction, from assessment to review, for these domains (patient satisfaction, however, still improved from assessment to review – see Fig. 4).
Results of multiple regressions on DIALOG scores by year and treatment stage. Higher values indicate higher odds of satisfaction on each DIALOG domain. Whiskers are 95% confidence intervals. The interactions show how changes in satisfaction by treatment stage differ between the two years. The predictor variables in the legend are labelled to show what they are being compared against (assessment for review and discharge: 2018–2019 for 2021–2022). Again, the plots are interpreted as greater odds for whiskers >1, and lower odds for those <1. See Fig. 1 for list of abbreviations used.
Discussion
Our results are consistent with earlier findings by Mosler et al, Reference Mosler, Priebe and Bird16 and show that routinely collected DIALOG data can provide a sensitive and useful tool for detecting changes in patient satisfaction associated with contact with mental health services. Our results do not merely replicate the Mosler et al study – they show that DIALOG patient satisfaction measures improved with treatment stage. In other words, the treatment journey through mental health services was associated with improvements to patient satisfaction for both outcomes and services. Moreover, patients were more satisfied across the various quality-of-life domains, not just mental health, once they were receiving support (i.e. already assessed and had had at least one review, compared with initial contact of assessment). There was further improvement in satisfaction at the point of discharge (compared with review stage). The only exception to this was for personal safety at the point of discharge in 2021–2022, in which the observed improvement was not statistically significant. We observed improvement across the eight PROM domains as well as the three PREM domains. Improvement was observed particularly in the practical help domain throughout the stages of treatment (PREM scores were not reviewed in the study by Mosler et al).
Satisfaction across the quality-of-life domains, as measured by DIALOG, were not markedly different between the two years evaluated. However, there were some subtle exceptions to this. Post-pandemic, there was a perceived reduction in satisfaction with mental health at the point of initial assessment; this can be understood potentially as an increase in mental health needs in the community during the pandemic. In the multiple regressions, after controlling for other variables, friendships and practical help received were the only two DIALOG domains in which there was an improvement in patient satisfaction between 2018–2019 and 2021–2022, whereas there was no evidence of any change for the other nine domains. We know that the pandemic and lockdowns altered social interactions, and that many community centres and resources were shut down.
It is difficult to draw clear inferences about the impact of the CMH transformation from these results, partly as a result of the COVID-19 pandemic producing a confound. The pandemic is known to have had a detrimental impact on mental health and well-being, Reference Pierce, Hope, Ford, Hatch, Hotopf and John27–Reference Chen and Wang30 while also limiting face-to-face contact in both healthcare settings and everyday social interactions Reference Anderson, Walsh, Anderson and Burnley31–Reference Schneiders, Mackworth-Young and Cheah34 (e.g. greater use of online support, less face-to-face contact, less community interaction). It is also due to the well-recognised limitations of uncontrolled whole-service evaluations detailed below. Measuring the impact of service reorganisation is challenging, and attempts to capture the impact of reconfiguration have often not been successful. Reference Giacco, Bird, Ahmad, Bauer, Lasalvia and Lorant35
We have identified several limitations in our evaluation. First, DIALOG response rates were low: scores were estimated to be available for less than 20% of ELFT CMHT patients for both 2018–2019 and 2021–2022. We cannot rule out the possibility of systematic differences in ELFT patients completing DIALOG questionnaires compared with those who did not. We also observed a comparatively low n figure at discharge compared with assessment and review, indicating attrition over time. From a broader perspective, working with routinely collected natural data is more prone to confounding effects than with formal research data, with less scope for incorporating appropriate controls. Reference Nørgaard, Ehrenstein and Vandenbroucke36,Reference Sauer, Chen, Hyland, Girbes, Elbers and Celi37 However, these limitations need to be balanced against the benefits of leveraging real-life data to evaluate and improve service provision, including larger sample sizes, better generalisability and lower costs. Reference Bull, Teede, Watson and Callander1,Reference Sauer, Chen, Hyland, Girbes, Elbers and Celi37,Reference Von Gerich, Peltonen, Séroussi, Weber, Dhombres, Grouin, Liebe and Pelayo38
We were also limited by the nature and intended purpose of DIALOG data, which are primarily intended to aid in care planning during active mental health treatment, and to judge the success of treatment in meeting an individual’s goals for care. These data are informative, but challenging to use, for causal inference or evaluation purposes at an aggregate level, because they have no standard study period, study population or data collection schedule. This makes it difficult to establish baselines and controls for studying differences in differences. Further longitudinal analysis of DIALOG is needed to better understand its utility as a tool for evaluation. Nevertheless, as our findings demonstrate, routinely collected mental health data can provide a useful tool for understanding the needs of the population being referred, changes through the patient journey and other aspects of population evaluation.
Although we did find that patients’ quality of life had improved overall, especially for those who continued to engage with treatment, our study does not provide any answers as to ‘why’. Our study looked at the overall change in PROM and PREM through the patient journey while receiving care from CMH services. Quality of life is affected by a multitude of variables both inside and outside the health system, and contact with services potentially forms only a fraction of this experience. The improvement in quality-of-life domains extended beyond mental health. There can be a range of hypotheses as to how and why this happened, including improvement in the mental health domain having a positive impact on other quality-of-life domains, as well as the impact of care not being restricted to one’s mental health. However, we cannot infer a direct causality between the care received and improvement in DIALOG scores.
We propose the need for greater focus on routine outcome and experience data-gathering, with real-time analysis of the data shared at an individual patient, team or service level, as well as across the organisation and between healthcare providers. A significant focus of the CMH transformation work was around improving access and care coordination, which DIALOG is less able to capture. Access and other parameters need to be considered in conjunction with the pooled DIALOG scores in assessing overall quality of impact across the population. In other words, we are unable to gauge how many people might be missing out on ELFT mental health services, and what challenges they may face. Future research could investigate such potential underserved populations, with a particular focus on health inequalities. Another suggested avenue for future research is to compare DIALOG scores with other routine measures of improvement (e.g. Health of the Nation Outcome Scales), to check whether they align. This has already been carried out in a small-scale study at service level in the same organisation. Reference Butt, Walls and Bhattacharya39 Finally, our study is not a formal evaluation of the measurement properties of DIALOG, although it would be useful to investigate properties such as construct validity, responsiveness and reliability. Reference Mokkink, Prinsen, Bouter, Vet and Terwee40
In conclusion, we conducted a quantitative evaluation of patient outcomes and experiences within ELFT mental health services during the CMH transformation, using routinely collected DIALOG scores. Our analyses investigated the change in DIALOG scores by treatment stage for two different years. Our results showed that pooled, routinely collected DIALOG data can provide a useful measure of changes in patient satisfaction across the 11 domains (both PROM and PREM). The changes or improvements in DIALOG scores were similar across the two study years, and scores improved over the course of the patient journey from assessment to discharge in both time periods. Although this effect is not causal, it is suggestive of a positive impact of treatment on quality-of-life domains, including patient experience. Further research may consider the use of DIALOG as an assessment tool in a controlled study; comparisons of quality of life among people in mental health treatment, especially specific interventions, compared with those who are not; or variations in baseline scores or responsiveness to treatment in certain subgroups, including historically underserved populations. Overall, our results highlight both the strengths and limitations of routinely collected DIALOG data, as well as PROM/PREM data more broadly.
About the authors
S.G.S. is a senior research fellow in applied healthcare at the University of Plymouth Community & Primary Care Research Centre, Plymouth, UK and PenARC, Plymouth, UK. R.Bh. is a consultant psychiatrist and Clinical Lead for Mental Health Payment and Outcomes for East London NHS Foundation Trust, London, UK and Honorary Associate Clinical Professor at Warwick Medical School, University of Warwick, Coventry, UK. K.S. is a data analyst in healthcare and was a researcher at the University of Plymouth, Plymouth, UK at the time of this study. A.S. is a consultant psychiatrist at North East London NHS Foundation Trust, London, UK. P.S. was a researcher in psychology and applied healthcare at the University of Plymouth, Plymouth, UK at the time of this study. R.By. is Professor in Primary Care Research at the University of Plymouth, Plymouth, UK, Head of the University of Plymouth Community & Primary Care Research Centre, Plymouth, UK and Deputy Director, PenARC, Plymouth, UK.
Supplementary material
The supplementary material is available online at https://doi.org/10.1192/bjb.2026.10215.
Data availability
The data underlying this study are derived from anonymised patient records from East London NHS Foundation Trust. Due to the sensitive nature of NHS mental healthcare data and information governance restrictions, these data are not publicly available. Access to the data is subject to appropriate approvals from the Trust and relevant governance bodies, and may be considered on reasonable request, subject to data sharing agreements and ethical approval.
Acknowledgements
Prof. Stefan Priebe for his advice on the paper; Prof. Frank Rohricht, Medical Director for Research and Innovation from ELFT, who commissioned the initial analysis and agreed to further analysis; and Thomas Nicholas, Associate Director for Business Intelligence and Analytics, ELFT, for help with data-capture.
Author contributions
S.G.S., K.S., R.By. and R.Bh. conceived the study, including its rationale, aims and methodology. K.S., P.S. and S.G.S. conducted the initial analyses. S.G.S. completed the analyses with input from R.Bh. and A.S. S.G.S. drafted the first version of the manuscript, and S.G.S., R.Bh. and A.S. worked on the revised version. All authors had input into and approved the final version.
Funding
East London NHS Foundation Trust commissioned the University of Plymouth to conduct this evaluation. Funding was awarded to the University of Plymouth following a competitive tender process. S.G.S., R.By., K.S. and P.S. were additionally funded and supported by the National Institute for Health & Care Research Applied Research Collaboration South West Peninsula. R.By. and A.S. contributed to the evaluation as University of Plymouth partners; they are employed by East London Foundation Trust.
Declaration of interest
R.Bh. is a member of the BJPsych Bulletin editorial board; he did not take part in the review or decision-making process of this paper.
Ethical standards
East London NHS Foundation Trust commissioned the University of Plymouth to evaluate routinely collected CMHT data, to assess the impact of the CMH transformation, as recommended by NHS England. Because these routinely collected data were part of a commissioned local evaluation rather than research (i.e. classed as service evaluation rather than research under the UK Policy Framework for Health and Social Care Research), ethical approval was not required (as per Health Reimbursement Arrangement standards, and was agreed by the ELFT Ethics Committee). The data-sharing agreement for the evaluation was detailed within the terms of the contract.
Open access
eLetters
No eLetters have been published for this article.