Evaluation of reconfigurations of acute stroke services in different regions of England and lessons for implementation: a mixed-methods study

Policy and research: major system change and innovation

There have been significant changes in the provision of clinical care within the English NHS since the turn of the century, with proposals (some now implemented) to concentrate specialist services, such as major trauma, cardiac surgery, vascular and specialist paediatrics, in fewer centres. 1–4 Major system change (MSC) in health care is seen as having the potential to increase the provision of evidence-based care and improve clinical outcomes at scale. 1 A review of the literature on implementing MSC defines it as ‘interventions aimed at coordinated, system-wide change affecting multiple organisations and care providers, with the goal of significant improvements in the efficiency of healthcare delivery, the quality of patient care, and population-level patient outcomes’. 5 MSC involves the reorganisation of services (sometimes termed ‘reconfiguration’6) at the regional level and may include significant alterations to care pathways. One such change is service centralisation, whereby some aspects of service provision across a given region are concentrated in a reduced number of hospitals. 7–12 It may involve many stakeholders across multiple organisations, and – when implemented successfully – is hypothesised to optimise the balance between quality of care, access, workforce capacity and cost. 1

A useful way to understand these reconfigurations is as processes of innovation. Reviews of the literature on the diffusion of innovations and MSC in health care draw attention to the need for more research on the processes by which such innovations are initiated (e.g. key drivers for change), implemented and sustained (or not), and in what particular contexts. 5,13 A review of evidence on the diffusion of innovations13 suggests that sustainability relates to the nature of the innovation (e.g. the benefits it offers, how complex it is, how it is led, how stakeholders are involved, and the use of evaluation and feedback) and the context into which it is introduced (local staff and organisational structures, interorganisational networks, external pressures). This evaluation aims to contribute to the development of this evidence base by studying in depth the implementation and sustainability of major service reconfiguration, using the example of stroke services.

The impact of centralisation on outcomes has been demonstrated in several specialist health-care settings, including trauma,14–16 cardiac surgery17 and neonatal intensive care. 18 However, when this study was commissioned, there was little evidence on the impact of centralisation in the context of acute stroke care. Furthermore, evidence on how changes of this scale are implemented and the relationship between implementation approaches and the impact of changes on quality of care, outcomes and costs was limited. 1

Policy and research: reconfiguration of acute stroke services

Stroke is a leading cause of mortality and disability worldwide. 19 In England, there are an estimated 125,000 cases of stroke and 40,000 deaths from stroke each year. 20 Organised inpatient Stroke Unit (SU) care is associated with better quality care21 and reduced death and dependency. 22 The case for MSC in acute stroke services was strong, with clear evidence of unacceptable variations in the quality of care, and many patients denied access to evidence-based care. 23 The Department of Health and Social Care’s National Stroke Strategy for England recommended MSC for acute stroke services, identifying that SU care was the single biggest factor that can improve outcomes following stroke. 24 However, evidence on the impact of centralisation of acute stroke services,25,26 and how best to centralise stroke services, was limited.

The NHS in London and Greater Manchester (GM) led the way in reconfiguring the acute stroke pathways across their regions. Before reconfigurations, in both London and GM, suspected stroke patients were taken to the nearest A&E department to receive stroke care (Figure 1). Both of the reconfigurations aimed to centralise services into ‘hub and spoke’ models, consisting of a reduced number of services providing acute stroke care up to 72 hours following stroke (hubs), with a larger number of services providing care beyond this acute phase (spokes). However, the two models differed significantly in relation to the degree of centralisation; the model in London could be characterised as a more ‘radical’ change and resulted in five stroke services closing, whereas no services closed in GM.

FIGURE 1.

Simplified pre- and post-reconfiguration models in London and GM. (a) Before, London and GM; (b) after, London; (c) after, Greater Manchester A; and (d) after, GMB. CSC, Comprehensive Stroke Centre; DSC, District Stroke Centre. Parts a–c reproduced from Ramsay et al. 26 with permission. Effects of centralising acute stroke services on stroke care provision in two large metropolitan areas in England. Stroke 2015;46(8):2244–51. Stroke is published on behalf of the American Heart Association, Inc., by Wolters Kluwer. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0) License, which permits use, distribution, and reproduction in any medium, provided that the original work is properly cited, the use is non-commercial, and no modifications or adaptations are made. See https://creativecommons.org/licenses/by-nc-nd/3.0/. Permission to adapt this material has been agreed with Wolters Kluwer.

Parts a–c reproduced from Ramsay et al. 26 with permission. Effects of centralising acute stroke services on stroke care provision in two large metropolitan areas in England. Stroke 2015;46(8):2244–51. Stroke is published on behalf of the American Heart Association, Inc., by Wolters Kluwer. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0) License, which permits use, distribution, and reproduction in any medium, provided that the original work is properly cited, the use is non-commercial, and no modifications or adaptations are made. See https://creativecommons.org/licenses/by-nc-nd/3.0/. Permission to adapt this material has been agreed with Wolters Kluwer.

Reconfiguration of acute stroke services in London

London covers an area of 1570 km² with a population of 8.17 million people,27 and has approximately 8000 annual hospital admissions following stroke. 25 The London reconfiguration was conducted at the request of the London Strategic Health Authority (SHA). 28,29 Commissioners agreed an additional £20M per annum to be paid through an enhanced tariff, providing that quality standards set by a multidisciplinary steering group were met. The model was developed with the support of a Joint Committee of Primary Care Trusts (PCTs) representing all commissioners in London. 30,31

Hospital trusts participated in a bidding process to host Hyperacute Stroke Units [(HASUs) offering care over the first 72 hours following stroke, including assessment by specialised stroke medical teams, brain imaging and thrombolysis, if appropriate], SUs (offering acute specialist stroke rehabilitation) and transient ischaemic attack (TIA) services.

Following this process, of the stroke services provided by 32 London hospitals pre reconfiguration, eight services were designated as HASUs [admitting suspected stroke patients 24 hours per day, 7 days per week (24/7)] and 24 as SUs and TIA services, and five services were decommissioned. The reconfigured London model was implemented in July 2010; since then, all suspected stroke patients have been eligible for treatment in a HASU, then repatriated to a SU, a nursing home or their own home (see Figure 1b).

Reconfiguration of acute stroke services in Greater Manchester

Greater Manchester covers 1276 km² with a population of 2.68 million people,27 and has approximately 4000 hospital admissions following stroke per year. 25

In 2007, the Greater Manchester and Cheshire Cardiac and Stroke Network (GMCCSN) was charged by the Greater Manchester Association of PCTs to reconfigure services to allow universal access to hyperacute stroke treatment in the area. 32

In early 2008, local hospital trusts submitted bids to host the new acute stroke services. An External Advisory Group (EAG) composed of local and national stakeholders assessed these bids, and the new acute stroke services were awarded on the basis of the EAG’s recommendations. The reconfiguration was implemented in a number of stages, commencing in December 2008 and ending in April 2010. 32 Three hyperacute services were designated: one Comprehensive Stroke Centre (CSC) and two Primary Stroke Centres (PSCs). A total of 11 District Stroke Centres (DSCs) provided post-4-hour care and ongoing acute rehabilitation services. No stroke services were decommissioned. This service model is referred to as Greater Manchester A (GMA) (see Figure 1c). GMA differed from the London model in two important ways. First, any suspected stroke patient presenting within 4 hours of developing stroke symptoms was transferred to either the CSC or PSC for hyperacute care; once stable, he/she was repatriated to a DSC, a nursing home or their own home. If presenting outside this ‘4-hour window’, stroke patients in GM were taken directly to the nearest DSC, much as they would have been prior to reconfiguration. This ‘4-hour window’ represented a contrast with London, where all suspected stroke patients were eligible for treatment in a HASU. Second, although the CSC admitted patients 24/7, the PSCs admitted stroke patients only between the hours of 07.00 and 19.00, Monday–Friday (see Figure 1c); this contrasted with London, where all HASUs admitted patients 24/7. GM initially chose a service model in which all patients presenting at hospital within 24 hours of onset of stroke symptoms would be treated in a CSC/PSC (i.e. similar to the model implemented in London). However, because some hospitals raised concerns about the impact of centralisation on hospital resources and patient safety, the ‘4-hour’ model was adopted.

Further reconfiguration in Greater Manchester

When the GM reconfiguration was first designed, it was agreed that a formal review of performance should be conducted 12 months post implementation, and, based on this review, the EAG concluded that the GMA model had not fully delivered on its aim to provide local populations with equal access to high-quality acute stroke services. Therefore, further changes were considered, with an agreement to implement a revised model reached in September 2013; in March 2015 a revised model [Greater Manchester B (GMB)] was implemented.

Under the new model (see Figure 1d), any suspected stroke patient was taken directly to either the CSC or a PSC; PSC hours were extended to cover 07.00–23.00, 7 days per week, and DSCs were no longer designated to receive suspected acute stroke patients, meaning that all suspected stroke patients in GM were designated to be treated in a hyperacute unit, similarly to the London model.

Reconfiguration activity in the Midlands and the East of England

The Midlands and East of England covers an area in excess of 48,000 km² with a population of 15.5 million people. 27 This region has 20,000 hospital admissions following stroke every year. 33 NHS Midlands and East SHA identified variation in acute stroke service performance and outcomes, both across the region and in comparison with other parts of the country. There was interest in the improvements in performance and outcomes achieved in London as a result of its major stroke reconfiguration in 2010.

The SHA recognised substantial differences between the Midlands and the East of England and London, not least in terms of demography and the predominantly rural nature of large parts of the Midlands and the East of England. Consideration was given to how, with differing geography, demography and economic circumstances, the Midlands and the East of England could achieve a step change improvement in stroke outcomes. A review was commissioned to identify the arrangements necessary to achieve this. An important influence on the review (e.g. in terms of timescale) was that it took place during the final year of the SHA, which, along with PCTs and stroke networks, were abolished as part of the NHS reforms brought about by the Health and Social Care Act 2012. 34

An External Expert Advisory Group (EEAG) developed a detailed best-practice specification to guide local service provision, covering the whole care pathway. Within local health systems proposals were developed, co-ordinated by the nine local Stroke Networks, detailing how they would meet this specification. Subsequently, the final recommendations were shared with participating areas in March 2013, shortly before the NHS reforms were implemented (and the SHA, PCTs and stroke networks were abolished). The service changes were anticipated to be implemented by March 2014. However, by December 2015, the reconfigurations either had not commenced or had been halted after making progress in planning and engagement, with several areas opting to carry out improvements within individual hospitals, rather than changing the acute stroke system more radically at a regional level.

Timeline for changes studied

Table 1 presents a timeline of the changes studied in this evaluation, covering changes implemented in GM in 2010 (GMA) and 2015 (GMB), changes implemented in London in 2010 and changes planned in the Midlands and the East of England.

Research questions

1. What are the key processes of and factors influencing the development and implementation of the acute stroke service reconfigurations?

2. To what extent have system changes delivered process and outcome improvements?

3. Have changes delivered improvements that stakeholders (e.g. commissioners, staff, patients and the public, and reconfiguration leads) think are worthwhile?

4. Have changes delivered value for money?

5. How is service reconfiguration influenced by the wider context of major structural change in the NHS?

Understanding what works and at what cost

This component of the evaluation analysed documentary evidence to establish the models applied in London and GM; it also analysed routinely collected data (e.g. national databases) to determine whether or not these changes were associated with any changes in care provision, clinical outcomes and cost-effectiveness (see Figure 2).

In assessing the nature and results of the reconfigurations, we applied a controlled before-and-after design. 45 This compared the participating regions in terms of the impact they had on delivery of clinical interventions associated with improved clinical outcomes (e.g. SU within 4 hours) (see Chapters 4 and 11), clinical outcomes (see Chapters 3 and 11) and cost-effectiveness of care (see Chapters 5 and 12). In addition to comparing these sites pre and post reconfiguration, we made wider comparisons with the rest of England (RoE); this approach facilitated analysis of these impacts in the context of changes that took place in the RoE over this period.

Understanding development, implementation and sustainability

To develop lessons for future reconfigurations, it is important to establish not just whether or not process and outcome changes took place, but also how and why they occurred; furthermore, it is important to analyse whether or not and how changes and their impact were sustained (defined as ‘the process through which new working methods [e.g. a new referral pathway], performance goals [e.g. improvements in clinical outcomes and delivery of evidence-based clinical interventions] and improvement trajectories are maintained for a period appropriate to a given context’46). 46–49 To study these, we used a range of qualitative methods (documentary analysis, stakeholder interviews and non-participant observation) (see Figure 2).

These data were used to explore themes drawn from the evaluation’s conceptual framework and thus establish the relationships between activities in support of change, the context, the complex interactions between stakeholders, and perceived process and outcome changes. In doing so, we analysed factors influencing the decision to change (or not to change) (see Chapters 6, 8, 10 and 13), the decision on which model to implement (see Chapters 6, 8 and 13), approaches to implementation (see Chapters 6, 7 and 13), and how these factors influenced the patient and carer experience (see Chapter 9) and sustainability of change (see Chapter 14).

Synthesis of approaches

We used a mixed-method case study approach to draw together the learning from the approaches described above (the cases being each MSC planned and/or implemented, i.e. London, GM and the Midlands and East of England). This facilitated the development and testing of theories on how efforts to bring about change interacted with the context in which they were implemented. 50–52 The qualitative component was designed to allow change to be evaluated, first, in relation to how the reconfigurations of acute stroke services were planned and governed at the regional level and, second, in relation to how services within each studied region experienced the changes. The services we studied were selected to reflect the main forms of change experienced by organisations participating in reconfigurations of this kind, including developing new services and refocusing or decommissioning existing services. We drew together findings from the evaluation’s quantitative and qualitative components to develop and test theories on the relationships between the models selected, implementation approaches applied, the degree to which the model was ‘successfully’ implemented, and how these contributed to the outcomes observed (see Chapter 7) in terms of provision of evidence-based care and clinical outcomes (such as mortality and patient LOS).

Ethics and research governance approvals

We recognised that taking part in the qualitative aspects of this research could potentially cause participants distress, for example patients and carers discussing personal experiences of stroke services, or staff discussing the reconfiguration and/or closure of services. We obtained full ethics approval for this study in September 2011 from the National Research Ethics Service Ethics Committee London-East, which set out how we would minimise this potential distress. Following the November 2012 extension to this project, a substantial amendment to ethics approval was obtained in February 2013, reflecting the additional participating sites and methods. Following the March 2015 extension, a substantial amendment (version 1.9, 23 November 2015) and a non-substantial amendment (version 2.0, 15 January 2016) were obtained from the South East Coast – Surrey NHS Research Ethics Committee. In support of data collection in our studied areas, we obtained local research governance permissions for all relevant organisations (see Appendix 1).

Patient and public involvement

Patient and public involvement informed and enhanced our research throughout the project. We provide a detailed description of our approach to PPI and its impact in Appendix 2.

What was already known?

• Organised inpatient SU care is associated with higher quality care and reduced death and dependency.

• Acute stroke services were being centralised in several countries as a means of improving access to organised inpatient SU care, but it was not known if this affected mortality and LOS.

What this chapter adds

• In London, where all patients were eligible for treatment in a HASU, there was a reduction in mortality and LOS.

• In GMA, where CSC/PSC care was provided to patients presenting within 4 hours of developing stroke symptoms, there was no impact on mortality but LOS was reduced.

Data

We obtained patient-level data from the Hospital Episode Statistics (HES) database57 for all patients in England with a primary diagnosis of stroke defined using International Classification of Diseases, 10th Revision (ICD-10)58 codes I61 (intracerebral haemorrhage), I63 (cerebral infarction) or I64 (stroke, not specified as haemorrhage or infarction) between 1 January 2008 and 31 March 2012. We excluded subarachnoid haemorrhage (ICD-10 code I60) because it is managed through a different clinical pathway. 59 The data were linked to mortality data supplied by the Office for National Statistics (ONS)60 using an anonymised unique patient identifier to identify deaths from any cause and at any place of death (hospital or otherwise) at 3, 30 and 90 days after hospital admission. LOS was measured in days as the difference between date of admission and date of discharge, including same-day transfers between hospitals.

January 2008 was defined as the start of our analysis period following the publication of the National Stroke Strategy for the English NHS in December 2007,24 leading to better emergency responses to stroke and acute stroke care around the country. 61 Our data cover a 27-month period before the changes in GM (which occurred in April 2010) and a 24-month period afterwards. In London they cover a 30-month period before the changes (July 2010) and a 21-month period afterwards. In both areas some hospitals began to reconfigure their services before these dates and we controlled for this in our analysis using hospital and time fixed effects.

Our main analysis was confined to patients living in urban areas (defined as ‘urban-less sparse’ using the urban/rural classification for England;62 95% of stroke patients in GM and London lived in these areas compared with 75% in the RoE). We did not restrict the analysis to any type of hospital (we included hospital fixed effects to allow for hospital differences) and we did not impose a minimum number of patients to be treated at each hospital (observations in the hospital-level regressions were weighted by the number of patients). Patients were treated at 11 hospitals in GM, 38 in London and 405 in the RoE over the period. Data were available for 258,915 admissions, of which 17,650 were in GM (9413 prior to reconfiguration, 8237 afterwards) and 33,698 were in London (18,672 prior to reconfiguration, 15,026 afterwards).

Statistical analyses

We evaluated whether or not centralising acute stroke services in GM and London had an impact on mortality and LOS using a between-region difference-in-differences regression analysis,63 comparing the changes over time in GM and London to the change over time in the RoE. The analysis was carried out at hospital level using quarterly observations of risk-adjusted mortality and LOS; the risk adjustment was conducted at the patient level on all patients in the data. The approach was consistent with the Medical Research Council guidelines for using natural experiments to evaluate population health interventions,64 and a similar method was used in an evaluation of the Advancing Quality initiative in the north-west of England. 65

We calculated expected risks of death at 3, 30 and 90 days after admission using patient-level logistic regressions, including binary indicators for sex and age interactions (age measured in 5-year bands), stroke diagnosis using the first four digits of the primary ICD-10 code (19 categories), Charlson Comorbidity Index66 derived from secondary ICD-10 codes, the presence of 16 comorbidities included in the Charlson Comorbidity Index, ethnic group (18 categories), and deprivation quintile67 and urban/rural classification62 (eight categories) of the area in which the patient lived (of 32,482 lower layer super output areas in England). The patient-level regressions were run only on patients who had a stroke before the reorganisations in GM and London so that the risk adjustment was not contaminated by the changes. The regression coefficients (derived from the logistic regressions for the pre-implementation period) were used to predict the probability of mortality for every patient (in both pre- and post-implementation periods). These were aggregated to create a data set of the actual percentage of patients who died and the expected percentage by admitting hospital and quarter. We tested whether or not the reconfigurations had an impact on mortality using least squares regression of the actual minus expected mortality percentage (because we are modelling differences) against interaction terms between GM and the post-reconfiguration period and London and the post-reconfiguration period. We included binary indicators for each of the 454 admitting hospitals (hospital fixed effects) and the 17 quarters (time fixed effects), and each observation was weighted by the number of patients treated at that hospital in that quarter. Standard errors were corrected for heteroscedasticity.

We used the same approach for LOS, but our risk adjustment equation was estimated using a generalised linear model (GLM) with gamma family and log-link to account for data skewness. 68 We experimented with other GLM specifications and a log-transformation but the selected model gave the best fit in terms of residual plots and Akaike’s information criterion. We added binary indicators for mortality at 3, 30 and 90 days after stroke to the risk equation and used the regression coefficients to predict expected mean LOS.

We undertook pre-trends tests to examine whether or not risk-adjusted mortality and LOS had a different linear trend in GM and London compared with the RoE before the reconfigurations. We reran the models on every quarter prior to the reconfigurations and included linear time trends instead of binary indicators for quarter. We added interaction terms between GM and the linear time trend and London and the linear time trend and tested the individual significance of the interaction terms. In every case they were non-significant (p > 0.05).

Principal findings

Risk-adjusted mortality and LOS fell in GM, London and the RoE during the study period. In London there was a significant reduction in mortality at 3, 30 and 90 days after admission over and above the reduction seen in the RoE; at 90 days the reduction in mortality was 1.1 percentage points. There was also a significant reduction in LOS of 1.4 days over and above the reduction seen in the RoE. In GM there was no impact on mortality over and above the change seen in the RoE but there was a significant reduction in LOS by 2.0 days. Significant reductions in mortality and LOS were largely achieved among patients with ischaemic stroke.

Strengths and weaknesses

The main strengths of the analyses are the large national data set we have used, which contains detailed information on outcomes and patient characteristics, and the robust quasi-experimental framework we have employed, which allowed us to control for trends in the RoE and other factors that could affect outcomes during the same period.

There are several weaknesses, which mean that our findings should be treated with caution – these call into question whether or not the findings of our analyses can be interpreted as causal effects. First, we consider the main limitation of our analysis to be that the HES database does not include information on stroke severity, which has been shown to be an important predictor of mortality. 69 Post-reconfiguration data from the Stroke Improvement National Audit Programme (SINAP)70 show that indicators of stroke severity such as ‘worst level’ of consciousness in the first 24 hours after stroke and neurological deficits on admission varied between GM, London and the RoE (Table 5), but there was no discernible trend over all the indicators. In spite of this, and even though our outcomes were risk-adjusted for a number of patient-level factors and we accounted very flexibly for differences between hospitals and trends over time, we cannot rule out the possibility that the differences in outcomes may be due to variations in stroke severity over time between GM and London and the RoE.

Second, we were unable to assess the impact of the reconfigurations on other outcomes, such as quality of life (QoL), disability or neurological and functional impairment, because these measures were not collected in the HES database.

Third, the HES database includes only patients admitted to hospital; it does not include any information about patients who died before they reached hospital, nor does it include information on the time of stroke, and, hence, our analyses of mortality were based on time from admission. If stroke patients in London were more likely to die before reaching the hospital owing to longer distances to HASUs, then the effects of the reconfigurations on mortality would be overestimated. Evidence suggests that this is unlikely, because ambulance journey times for stroke patients did not increase appreciably after the reconfiguration in London, with mean scene-to-hospital times of 14 minutes from January 2005 to March 200821 and of 16 minutes from April 2011 to March 2012. 73 In addition, stroke severity in London post reconfiguration was similar to the severity in the RoE (see Table 5); if more severely ill stroke patients died in London before reaching the hospital, the level of severity in the audit data for London would be lower than elsewhere.

Fourth, LOS was measured as the difference between date of admission and date of discharge. We assumed that when patients were discharged from one hospital and readmitted to another hospital on the same day that it was a transfer related to the original stroke, capturing the movement between components of the stroke care pathway (e.g. between a HASU and a SU in London). Conversely, we assumed that when a subsequent admission occurred ≥ 1 day after discharge that it was a recurrent stroke (there is a risk of stroke recurrence in the first month after discharge of 1.1% to 15%). 74

Fifth, there was a higher than expected number of stroke patients per month in London during the post-reconfiguration period. One possible reason is that after the reconfiguration London treated more patients from surrounding areas who might previously have gone to their local A&E department. It may also have been because greater public awareness of stroke in London meant that more people presented acutely in London rather than staying at home. This could have biased the results in favour of the London reconfiguration if the additional admissions were less severe strokes, but there were no discernible differences in severity between London, GM and the RoE (see Table 5).

Comparison with other studies

Our findings with respect to mortality in London are consistent with a cost-effectiveness analysis of the reconfiguration of acute stroke services in London,53 which also found a significant improvement in survival in London at 90 days after admission. In that study the impact on survival was calculated using survival models estimated before and after the reconfiguration, and an adjustment was applied to the difference in survival to account for national trends in mortality. The present study has a larger sample size, uses national individual-level data and has more patients in the post-reconfiguration period in London, provides a more robust analysis of national trends via the quasi-experimental difference-in-difference design and examines the reconfiguration in GM. Our findings are also consistent with a previous analysis based on national stroke audit data, showing that patients admitted to stroke services with higher levels of organisation were more likely to receive high-quality care and to have a reduced risk of death 30 days after stroke. 21

Implications

Our findings showed that mortality outcomes were different in the two systems. In London, where all patients were eligible for treatment in a HASU, ambulance data indicated that in 2011/12, 98.7% of London stroke patients were transported to the appropriate service: 95.7% were taken appropriately to a HASU, 3% were taken appropriately to an A&E department and only 1.3% were taken to an A&E department when they should have been taken to a HASU. 73 A review of the first year of the new GM model reported that, when only those patients presenting within 4 hours of developing stroke symptoms were eligible to be treated in a CSC/PSC, of the patients who presented with stroke within 4 hours, 36% were not taken to a CSC/PSC. 32 Hence, a higher proportion of patients than planned were admitted to district hospitals where access to specialist expertise and care was more limited. In addition, data on the achievement of processes measuring the quality of care that patients with stroke receive during the first 72 hours of care were collected between April 2011 and December 2012 as part of the SINAP in England. 70 A significantly higher proportion of patients in London received care that was compliant with the care processes than in GM and the RoE (which were broadly similar) (Table 6).

This suggests that the centralised model of care in London was more closely adhered to and achieved greater compliance with clinical interventions. There is evidence that better compliance with these measures is negatively correlated with mortality. 21 It is also noteworthy that, based on multivariate analyses, these measures independently affect mortality,21 suggesting that different aspects of specialist care provided throughout the HASU can separately affect patient outcomes. The upshot is that differences in mortality may be explained by the lower level of adherence in GM or differences between the two systems in terms of the access to hyperacute care for patients presenting after 4 hours of developing stroke symptoms. This suggests that the type of system redesign and the extent of its implementation can affect patient outcomes and needs to be taken into account by those who are reorganising services. In addition, although there is evidence that organised inpatient stroke care is beneficial to patients with intracerebral haemorrhage as well as ischaemic stroke,55 the improvements in outcomes from centralisation found in this study were largely among patients with ischaemic stroke. On the one hand, this may reflect the poorer prognosis among patients with intracerebral haemorrhage;75 we found some evidence in London that mortality was reduced at 3 days, suggesting that the reconfiguration may have delayed death among these patients but did not change the ultimate outcome. On the other hand, point estimates of the reductions in mortality in London were higher for intracerebral haemorrhage than for ischaemic stroke but the effects were non-significant; the wider CIs may be due to the smaller numbers of patients in the data suffering this type of stroke. On a different point, although the results were consistent when patients living in rural areas were included, they may be less relevant to services operating in rural settings. The greater travel times in rural areas make centralisation challenging and may necessitate other solutions, such as telemedicine, whereby consultation and triage may be conducted remotely by a stroke physician in a specialist SU. 76–78 Finally, our findings may also inform the centralisation of other care services, such as vascular surgery79 and specialised surgery for cancer and other conditions. 80

What was already known about this subject?

• Provision of evidence-based stroke care is associated with better patient outcomes.

• Centralisation of acute stroke services in London and GM had a significantly different impact on clinical outcomes, with only the London changes associated with significantly greater reductions in mortality (see Chapter 3).

• There was limited evidence on the impact that centralisation of acute stroke care has on provision of evidence-based clinical interventions, and whether or not this might explain the changes in clinical outcomes observed.

What this chapter adds

• In London, where almost all patients were treated in a HASU, patients were more likely than elsewhere to receive evidence-based care in the first hours following arrival in hospital.

• GMA’s CSC/PSCs performed as effectively as HASUs in London, and significantly better than London on several important clinical interventions, but treated only 39% of stroke patients. This difference is explained in part by differing eligibility criteria in GM and London, but also because adherence to the model in GM was lower, with two-thirds of eligible patients treated in hyperacute units.

• As a result, only patients in London were significantly more likely than patients elsewhere to receive evidence-based care; stroke patients in GM were overall no more likely to receive evidence-based care in the first hours following arrival in hospital than patients in areas where no equivalent centralisation had taken place.

Design

This study used a controlled before-and-after design. It analysed risk-adjusted likelihood of stroke patients receiving evidence-based clinical interventions in GM and London, pre and post centralisation, compared with urban areas of England where acute stroke services had not been centralised (hereafter referred to as the ‘comparator’).

Data

Patient-level data were drawn from two national audits organised by the Royal College of Physicians: (1) pre centralisation, the National Sentinel Stroke Clinical Audit (Sentinel 2008), conducted from April to August 2008, was used; and (2) post centralisation, the SINAP, which ran from April 2010 to December 2012,72,84 was used. Reflecting the implementation dates for the centralisations, the GM post-centralisation period was April 2010 to December 2012 inclusive, whereas London’s was July 2010 to December 2012 inclusive. Data collected in the two audits differed: Sentinel 2008 collected a ‘snapshot’ of up to 60 patients per participating stroke service, whereas SINAP collected data for all patients receiving stroke care. Consequently, post-centralisation data cover significantly more patients.

The analysis included data submitted by all hospitals providing acute stroke care in GM, London and a comparator area formed of hospitals providing acute stroke care in two parts of England (north-west England, excluding GM, and north-east England), where local documents showed that no equivalent centralisation had occurred. The comparator was limited to hospitals in urban settings equivalent to GM and London (classified as ‘major urban’ by the UK ONS85); it covered 1.8 million people27 and its level of participation in national audits was equivalent to GM and London (details available in Appendix 3). Although Sentinel 2008 had uniformly high participation across England,84 participation in SINAP was variable in several areas of England, with many hospitals submitting few or no data. 72 These differing participation levels meant that the RoE could not act as the comparator. Consequently, data for approximately 56,100 stroke patients (7300 ‘before’, 48,700 ‘after’) were excluded. Data for all patients diagnosed with stroke (intracerebral haemorrhage or cerebral infarction) were included, both those occurring in hospital and those occurring outside hospital. Patients with invalid data were excluded.

Measures

We analysed all evidence-based clinical interventions that had been measured consistently in both audits. 59,86 These measures were calculated from arrival at hospital (or symptom onset if occurring in hospital), and assessed whether or not patients had their first brain scan within 3 hours and 24 hours of arrival (cut-off points were identified in the baseline audit national report reflecting the time to scan to support administration of thrombolysis, and national guidance to scan within 24 hours84); were admitted to a SU within 4 hours; received antiplatelets within 48 hours (if ischaemic); and underwent physiotherapist, nutrition and formal swallow assessments within 72 hours (all if eligible).

Statistical analysis

Descriptive statistics

Data for 38,623 acute stroke cases submitted to national audit were analysed, covering 51 hospitals pre centralisation (from a total of 189 hospitals participating in the audit across England) and 44 hospitals post centralisation (from a total of 171 hospitals across England). Table 7 presents the unadjusted data for GM and London compared with the comparator. Patient characteristics were similar in GM, London and the comparator in both pre- and post-centralisation time periods, and any potential effects of patient characteristics were controlled for in the regression analyses. Post centralisation, the proportion of patients receiving evidence-based clinical interventions increased in all three areas. It should be noted that denominators for these indicators varied from measure to measure owing to variable eligibility of patients or availability of data. Increases were most pronounced in care provided in the first hours following arrival at hospital (brain scan within 3 hours, admitted to a SU within 4 hours). The proportion of stroke patients receiving these clinical interventions was higher in London than in GM and the comparator, both pre and post centralisation, but the absolute difference was similar. Proportions of patients receiving evidence-based clinical interventions 24–72 hours after admission also increased in all three areas, but in all areas pre-centralisation levels commonly exceeded 80%, and post centralisation all proportions exceeded 90%. Generally, London had higher pre-centralisation levels and post-centralisation levels were similar – approaching the maximum – in each area.

Hospital-level variation

Pre centralisation, there was substantial between-hospital variation in the proportion of patients receiving evidence-based clinical interventions in all three areas (see Appendix 3). Post centralisation, hyperacute units in GM (CSC, PSCs) and London (HASUs) treated a higher volume of patients than elsewhere (see Table 7), and provided evidence-based clinical interventions to a higher proportion of their patients (Figure 4 and see Appendix 3, Figure 20).

FIGURE 4.

Between-hospital variations in the proportion of patients admitted to a SU within 4 hours by area, post centralisation. (a) GM; (b) London; and (c) comparator. Adapted from Ramsay et al. 26 Effects of centralising acute stroke services on stroke care provision in two large metropolitan areas in England. Stroke 2015;46(8):2244–51. All measures reflect time from ‘clock start’ (i.e. when patient first arrives at hospital or when symptoms are identified in in-patients). Stroke is published on behalf of the American Heart Association, Inc., by Wolters Kluwer. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0) License, which permits use, distribution, and reproduction in any medium, provided that the original work is properly cited, the use is non-commercial, and no modifications or adaptations are made. See https://creativecommons.org/licenses/by-nc-nd/3.0/. Permission to adapt this material has been agreed with Wolters Kluwer. Figure title and numbering updated for report.

Adapted from Ramsay et al. 26 Effects of centralising acute stroke services on stroke care provision in two large metropolitan areas in England. Stroke 2015;46(8):2244–51. Stroke is published on behalf of the American Heart Association, Inc., by Wolters Kluwer. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0) License, which permits use, distribution, and reproduction in any medium, provided that the original work is properly cited, the use is non-commercial, and no modifications or adaptations are made. See https://creativecommons.org/licenses/by-nc-nd/3.0/. Permission to adapt this material has been agreed with Wolters Kluwer. Figure title and numbering updated for report.

Although the proportion of patients receiving evidence-based clinical interventions increased in GM and London’s non-hyperacute units and in the comparator area overall, patient volume increased less, and the proportion of patients receiving evidence-based clinical interventions tended to be lower and more variable than in the hyperacute units. When GM PSCs operated as DSCs, they performed in line with other DSCs (Figure 4a and see Appendix 3, Figure 20).

Risk-adjusted likelihood of receiving evidence-based clinical interventions

Trends in risk-adjusted proportions of patients receiving evidence-based clinical interventions (Tables 8 and 9) reflected the unadjusted findings (see Table 7). Post centralisation, on all clinical interventions analysed, London patients were overall significantly more likely to receive clinical interventions than comparator patients. GM patients were significantly more likely than comparator patients to receive two interventions (brain scan within 3 hours and 24 hours), significantly less likely to receive three interventions (admission to SU within 4 hours, and physiotherapist and swallow assessments within 72 hours), with no significant difference between GM and comparator patients on the two remaining interventions. London patients were overall significantly more likely than GM patients to receive six of the seven interventions (brain scan within 3 hours and 24 hours, admission to a SU within 4 hours, and physiotherapist, nutrition and swallow assessments within 72 hours), with the magnitude of differences ranging from 1.2% to 10.4% and with no significant difference on antiplatelets within 48 hours. Patients treated in hyperacute units, in both GM and London, were significantly more likely to receive clinical interventions than patients treated either in non-hyperacute units or in the comparator (with one exception, where comparator patients were significantly more likely to receive a physiotherapist assessment than GM hyperacute patients). Patients treated in GM hyperacute units were significantly more likely than London HASU patients to receive four clinical interventions: brain scan within 3 hours, admission to SU within 4 hours, brain scan within 24 hours and antiplatelets within 48 hours (the magnitude of the differences ranged from 2.0% to 13.8%). Patients treated in GM hyperacute units were significantly less likely to receive a physiotherapist assessment within 72 hours than patients treated in London HASUs (magnitude 2.2%). There was no significant difference between GM and London hyperacute units in the number of patients receiving nutrition and formal swallow assessments within 72 hours. To examine whether or not these differences reflected the greater proportion of patients in GM arriving within 4 hours, we reran this analysis focusing only on patients presenting within 4 hours of symptom onset and found that the differences between GM and London hyperacute units reduced substantially (see Appendix 3).

Access to care in hyperacute units in Greater Manchester and London

Post centralisation, 39% of GM patients were treated in a hyperacute unit, whereas 93% of London patients were (see Table 7). In addition, only 66% of GM stroke patients who presented within 4 hours of symptom onset were admitted to a hyperacute unit (see Table 7), meaning that 34% of patients who were eligible for hyperacute unit care were not admitted to one.

We reran our analyses using all available data for the RoE as the comparator; the results did not change appreciably (see Appendix 3).

Principal findings

Post centralisation, the risk-adjusted likelihood of patients receiving evidence-based clinical interventions increased significantly in all areas. London patients were overall significantly more likely than patients elsewhere to receive the interventions. Importantly, hyperacute units in both GM and London were significantly more likely to provide interventions than non-hyperacute units in these areas, and in the comparator area overall.

Fewer than 40% of GM patients were admitted to a hyperacute unit (in line with an analysis conducted by the GMCCSN as part of their 12-month review of the centralisation32) whereas 93% of London patients were admitted to a hyperacute unit; this is likely to explain the overall differences in provision over the two areas and suggests a clear effect of the models selected. This effect partly relates to the different eligibility criteria for admission to a hyperacute unit: in GM, only stroke patients presenting within 4 hours of developing symptoms were eligible for treatment in a hyperacute unit, whereas all London stroke patients were eligible. However, in GM, only two-thirds of patients eligible for care in a hyperacute unit were admitted to one; this suggests that differences between GM and London derived not only from the eligibility criteria, but also from how reliably they were followed, with both contributing to the lower overall likelihood of patients receiving interventions in GM. The fact that GM hyperacute units were significantly more likely than London HASUs to provide four of the interventions analysed may relate to the ways in which London and GM hyperacute units were designed and specified (e.g. in terms of staffing and availability of allied services, GM’s use of the 4-hour eligibility criterion, and the extent to which services had to operate a 24/7 service), and how these specifications were implemented. The contrast in performance by PSCs in-hours and out-of-hours is likely to reflect the differing specifications for hyperacute and non-hyperacute services applied at these times, including staffing levels and access to diagnostic services. In both areas, non-hyperacute units were less likely to provide interventions, reflecting an important consideration when centralising services. In London, only 7% of London patients were treated in SUs; this had a negligible effect, but may reflect issues where the referral protocol was not followed. In GM, DSCs were less likely to provide the interventions analysed; as DSCs treated 61% of GM patients, GM and comparator services performed similarly overall, suggesting that the benefits of this centralisation were limited. Finally, it is unlikely that geography influenced GM patient eligibility significantly, given that the maximum distance from any point in GM to the 24/7 CSC is < 25 miles.

Comparison with other studies

Little is known about how large-scale service centralisation influences provision of evidence-based stroke clinical interventions, but there is substantial evidence that such clinical interventions are associated with better clinical outcomes, for example through dedicated SUs with specialist staff providing rapid assessment, treatment and therapies. 21,59,81,86,87 Recent research indicates that the London centralisation had a significantly greater impact on patient mortality than in the RoE, whereas the GM centralisation did not. 25 The current analysis suggests that this difference might result from better access (on average) to evidence-based clinical interventions on arrival at hospital in London, owing to a higher proportion of patients being admitted to HASUs. This in turn suggests that the model implemented is important: centralising acute stroke services into a small number of hyperacute units can increase access to evidence-based clinical interventions, which may in turn lead to better clinical outcomes. Furthermore, the model selected may influence the extent to which it is followed; the GM model may have been followed less reliably owing to complexity associated with the 4-hour eligibility criterion for admission to hyperacute units and the limited hours of PSCs. This in turn suggests potential advantages of service models that aim for relative simplicity. Therefore, our findings indicate that there may be significant benefits to models in which all stroke patients are eligible for hyperacute unit admission; offering hyperacute unit admission selectively, for example based on time of stroke onset, may limit these benefits. As the centralisations studied were implemented in large metropolitan areas, these findings are of greatest relevance to urban settings.

General increases in the likelihood of patients receiving interventions, reflected in the comparator, should also be noted. These may reflect the influence of national drivers to improve stroke care across England, for example the National Stroke Strategy24 and the ‘Stroke Best Practice Tariff’,88,89 increasing hospitals’ prioritisation of clinical interventions measured in national stroke audits. This may contribute to the overall lower likelihood of GM patients undergoing four of the interventions than comparator patients, as our data indicate comparatively limited improvements in DSC services, whereas general improvements occurred in many of the comparator services.

Strengths and weaknesses

To study the impact of centralising acute stroke services, we have been able to analyse two different models of centralisation implemented within the same national health-care system at approximately the same time, using data from two comprehensive national audits. This allowed us to compare findings between the two centralisations and with a large, urban comparator, thus enhancing external validity. Our study has several limitations. First, because of variable participation in SINAP, it was not appropriate to use the RoE as our comparator; this limits the extent to which we may use these results to explain previous research on the impact of these centralisations on mortality. Second, it is possible that data completeness may have varied across hospitals in the GM, London and comparator areas, thus limiting the confidence with which we interpret the data. Third, owing to data availability, several important clinical interventions could not be analysed. Provision and timing of thrombolysis were excluded because in 2008 thrombolysis rates were not reported with sufficient frequency to permit reliable analysis. However, the stroke services analysed aim to treat all stroke patients, only a proportion of whom (approximately 16%90) are eligible for thrombolysis. Other interventions excluded were anticoagulation of atrial fibrillation (data were not collected in either audit), and occupational therapy assessment and continence assessment (criteria assessed changed over time). Fourth, it was not possible to adjust for stroke severity [e.g. using the National Institutes of Health Stroke Scale (NIHSS)] because these data were not collected in either audit analysed. However, it is recommended that stroke patients receive these interventions regardless of severity (unless contraindicated); therefore, controlling for severity should have little impact on whether or not these interventions are provided. However, the latest version of the audit collects NIHSS for all patients on arrival at hospital, and future research might usefully analyse whether or not stroke severity influences the likelihood of receiving interventions. Fifth, because the areas studied varied in overall performance levels pre centralisation, and several of the selected indicators approached ceiling both pre and post centralisation, our analysis of change over time was limited. Sixth, it was not possible to control for hospital/unit effects because the number and function of stroke services changed significantly over the period studied. Finally, the analysis of patients presenting within 4 hours of their symptoms appearing was derived from SINAP data on time of stroke symptom onset; although we made every effort to exclude unreliable data, it remains that some may have been recorded unreliably.

Implications

Our data indicate that both the centralisation model selected and the degree to which it was followed may have influenced the likelihood of stroke patients receiving evidence-based clinical interventions, and that these factors may be interrelated. Therefore, qualitative analysis of how model selection and implementation processes influenced the development of hyperacute and non-hyperacute services, and the reliability with which patients were transferred to these services, is required to generate valuable lessons for future changes of this kind.

What is already known about this subject?

• A previously published model showed that the reconfigurations in stroke care in London were cost-effective. The model did not take into account changes that occurred in the RoE over the same time period.

• In London, where hyperacute stroke specialist care was provided to all patients, there was a reduction in mortality and LOS (see Chapter 3).

• In GM, where specialist hyperacute stroke care was provided only to patients presenting within 4 hours of developing stroke symptoms, there was no impact on mortality but LOS fell (see Chapter 3).

What this chapter adds

• There was a high probability that the centralised models of stroke care in London and GM were cost-effective. In London this was as a result of improvements in mortality and morbidity and in GM as a result of reduced cost of stroke care due to reduced LOS.

Overview

We estimated difference-in-differences in costs and outcomes to estimate NMBs and to evaluate the effect of the reconfigurations in London and GM on costs, mortality and QALYs compared with metropolitan areas in the RoE. Metropolitan areas were chosen so that population density and blue-light ambulance travel time to hospital were as comparable as possible with London and GM. The analysis was undertaken from the perspective of the English NHS and personal social services. Eight decision–analytic models were developed, each with 1000 hypothetical stroke patients. The time periods were as in Morris et al:25

1. London before the reconfiguration (January 2008 to January 2010)

2. London after the reconfiguration (July 2010 to March 2012)

3. RoE/excluding GM for the same time period as (1)

4. RoE/excluding GM for the same time period as (2)

5. GM before the reconfiguration (January 2008 to November 2008)

6. GM after the reconfiguration (April 2010 to March 2012)

7. RoE/excluding London for the same time period as (5)

8. RoE/excluding London for the same time period as (6).

Total costs and QALYs per 1000 patients were calculated for each of the eight models at 90 days and 10 years after admission, using a discount rate of 3.5% for costs and QALYs. 93 Costs are for the year 2013/14 and are in GB pounds sterling (£). The difference in costs and QALYs for (1) and (2) compared with (3) and (4) was then compared with the difference in costs and QALYs for (5) and (6) compared with (7) and (8).

Data

We used routinely collected hospital data (HES), audit data (Sentinel 200884 and SINAP;72 see Chapter 4) and data from the published literature to construct the model (Table 10). The data were linked to data from the ONS to obtain the date of death. 60 The South London Stroke Register (SLSR), a population-based stroke prospective registry recording all first-ever strokes in patients of all ages living in an area of South London,31 was used to calculate changes in disability before and after the reconfigurations. Further details on the data sets used can be found in Chapter 325 for HES and ONS data, Chapter 426 for SINAP and Sentinel 2008, and Hunter et al. 53 for SLSR.

Statistical analyses

Model structure

Each of the eight models has two components:

1. A 90-day discrete event simulation (DES) of daily acute hospital ward movements, discharge destinations and mortality using data from HES, Sentinel and SINAP (Figure 5).

2. A 10-year Markov model30 with 90-day cycles using information from the 90-day model plus HES, the SLSR and published data to calculate costs and QALYs (see Appendix 4, Figure 27).

All patients entered the 90-day model as an admission to hospital following a stroke. The proportion of patients in each health state in the first cycle of the 10-year model was determined by the final destination of patients in the 90-day model, including still on an acute hospital ward.

FIGURE 5.

Ninety-day DES model structure. ITU, intensive therapy unit; CCU, coronary care unit. Adapted from Hunter et al. 91 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Adapted from Hunter et al. 91 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

The flow for the 90-day model is illustrated in Figure 5, excluding the state of death, patients had a probability of death every day of the 90-day model regardless of where they were in the model. The same flow was used for all eight models and differed from the stylised diagrams in Figure 1. This is because data from SINAP show only if the first ward of admission was a SU or not, not whether it was a HASU/CSC/PSC or a SU/DSC. 26

Base case

The characteristics of patients are reported in Table 10 and detailed resource use and costs for the model are shown in Table 11. Compared with the RoE, at 90 days London had a greater improvement in the reduced risk of mortality at 90 days from before to after the reconfiguration. Per 1000 strokes, the adjusted number of deaths in London at 90 days was 100 before the reconfiguration and 75 after, a reduction of 2.7 percentage points in the number of deaths. During the same period in the RoE, there was an adjusted reduction in deaths of 1.8 percentage points (114 deaths before compared with 97 deaths after) (Table 12). This represents a relative reduction of deaths in London compared with the RoE of 0.9%, or nine deaths per 1000 patients. In 90% of iterations of the model, London had a greater reduction in deaths after the reconfigurations than the RoE. There was no reduction in deaths in GM at 90 days (Table 13). Both areas had a reduction in LOS relative to the RoE; 2 days fewer in GM and 0.6 days fewer in London.

At 10 years, both reconfigurations resulted in more QALYs than their RoE comparator (see Tables 12 and 13). London resulted in 58 more QALYs at 10 years per 1000 patients at an additional cost of £1,014,363. This was equivalent to an incremental cost-effectiveness ratio of £17,452; hence, at a WTP for a QALY gained of £20,000, the London changes had a higher NMB than the RoE changes over the same time period. At 10 years, the reconfigurations that occurred in GM dominated what happened in the RoE over the same time period in that there were more QALYs (18 QALYs per 1000 patients over 10 years compared with the RoE) and less cost (–£470,848 per 1000 patients compared with the RoE over 10 years).

Cost-effectiveness acceptability curve

The probability that the changes that occurred in GM as a result of the reconfigurations had a higher NMB than in the RoE over the same time period and peaked at a WTP for a QALY gained of £7000 and 82% probability (Figure 6).

FIGURE 6.

Cost-effectiveness acceptability curve of the probability that the reconfigurations in London and GM resulted in a higher NMB than in the RoE over the same time period. Adapted from Hunter et al. 91 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Adapted from Hunter et al. 91 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

London had a higher probability of being cost-effective than GM at WTP values > £30,000. GM had a higher probability of being cost-effective at lower values of WTP for a QALY; this is shown in Figure 7, where 35% of iterations of the difference in costs and QALYs in GM compared with in the RoE over the same time period fell in the south-west quadrant of the cost-effectiveness plane. The changes in GM cost less than in the RoE but there was only a 60% chance of a reduction in mortality compared with in the RoE. For London, the majority of iterations of the model (80%) fell in the north-east quadrant, whereby London cost more but also had a higher probability that there was an improvement in mortality (81%) (Figure 7).

FIGURE 7.

Difference-in-difference cost-effectiveness plane of the adjusted difference in 10-year costs and QALYs between London before and after reconfigurations minus the difference in the RoE over the same time period and the difference in costs and QALYs in GM compared to the RoE over the same time period. Adapted from Hunter et al. 91 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Adapted from Hunter et al. 91 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Sensitivity analyses

The results of the sensitivity analysis are reported in Appendix 4, Tables 45 and 46. None of the sensitivity analyses changed the conclusions drawn as a result of the analysis, other than to say that once the cost of a HASU increased by 50% per night, the London reconfigurations were no longer cost-effective at a WTP of £30,000 per QALY. If it is assumed that improvements in functioning (Barthel index94) occurred only in London and GM as a result of the reconfiguration but not in the RoE, there was a 97% chance in both London and GM that the reconfigurations were cost-effective at 10 years at a WTP of both £20,000 and £30,000 for a QALY (see Appendix 4, Figure 28).

Costs of implementation

In addition to the funding for the increased tariff, additional one-off financial investments were made to bring about the reconfigurations in London and GM. It was estimated that an investment of £9M was made in London to meet the requirements for the new HASUs and SUs, covering capital, equipment and premises refurbishment. 104 In GM it was estimated that an initial investment of £2.79M was required for similar items. 105 Other costs are likely to have been incurred in both London and GM to pay for research to identify the optimal configuration of services and to consult staff and public, but evidence of these costs was not available.

Principal findings

Both London and GM models had a higher NMB than changes that occurred over the same time period elsewhere in England at 10 years at a WTP of £20,000 to £30,000 for a QALY. The GM model appeared to be cost saving at 10 years but did not have the same health benefits as the London model and hence had higher values of WTP for a QALY, with London having a higher probability of being cost-effective. The differences between the two regions were small compared with the size of the CIs; therefore, they were not significant.

The results were in line with the findings on the impact of London and GMA on clinical outcomes25 (see Chapter 3) and clinical interventions26 (see Chapter 4) but also provided additional information on the value for money of the reconfigurations in two metropolitan areas in England compared with the rest of the country over the same time period. The London model reiterated findings that the centralisation of stroke care in London corresponded with clinical improvements that did not occur in the RoE over the same time period, but also showed that there was a high probability that the benefits were worth the additional costs. The results of the GMA analysis suggest that MSC of stroke care is possible with the investment of limited additional resources and potentially cost saving, although resulting in fewer clinical improvements.

Comparison with other studies

Evaluations of major system reconfigurations using decision modelling are becoming increasingly common. Although trials can provide information on what system changes should be implemented, decision models and population-level data can provide information on the implications for real clinical settings. 106 Previous evaluations of the centralisation of ovarian cancer treatment have used a combination of decision modelling of clinical trials prior to implementation107 and Dutch registry data following implementation to establish the cost-effectiveness of changes to the patient care pathway. 108 Synthesising audit or routinely collected data from the whole population in addition to data on costs has been used previously in New Zealand109 to show that admission to a SU is more cost-effective than admission to a general ward. In Scotland, population-based data have been used to evaluate the clinical effectiveness of admission to stroke wards,106 and analyses by our own team have used population data to establish the clinical effectiveness of MSC for stroke care in GM and London. 25 This analysis was the first to use whole population data, HES and audit data to evaluate the MSCs that occurred in stroke care in London and GM compared with in the RoE.

Strengths and weaknesses

The key strength of this analysis was that it provided information about the real-world implications of the changes made and hence can inform decisions about MSC in the future.

There are weaknesses to using observational, routinely collected hospital data, most notably coding errors. For example, in one London hospital 3% of 1300 stroke patients had a discharge destination code for discharge to a mental health hospital, compared with 0.13% of patients for hospitals across all of England. It is unlikely that such a large percentage of patients were being discharged to this location; instead, it is more likely that the wrong code had been used and that they were being discharged elsewhere. There is no way to know where and, hence, the only option was to remove the data prior to estimating discharge destination. It is likely that the data set contained other errors in relation to discharge destination. We assumed, however, that those errors would occur equally at random across the three regions. It was also hard to know how ‘stroke mimics’, that is patients presenting with stroke-like symptoms but who do not have a final diagnosis of stroke, appear in the data. They may be screened out of the data sets as not having stroke even though they have a SU admission, or the final diagnosis may potentially be missing from our data set. It was hard to know from the data what proportion were mimics and what impact they had on the cost and outcomes for the care pathway and there were no comparable data from other sources to estimate this.

It is important to note that the values reported in the model for LOS and mortality were values adjusted for differences between patients within the regions in age, ethnicity, sex and comorbidities. As a result they did not represent true values per 1000 patients but adjusted values, to allow for comparison between areas over the same period of time. As the changes occurred in GMA and London at different time points, no direct comparison between GMA, London and the RoE was possible as it would not have taken account of differential changes over time. One of the key weaknesses in the model is that South London was the only area for which there were data available on the impact of the changes on functioning using the Barthel index. 94 Although there were no changes in the proportion of patients discharged home before and after the reconfigurations, after the reconfigurations there may have been some improvements in Barthel scores for people who were discharged home, based on data from the SLSR (see Appendix 4, Table 43). Although GM and London had some information on functioning available for time periods after the reconfigurations, there was no information on functioning available before the reconfigurations with which to compare it, other than in the SLSR. Instead, we had to make the conservative estimate that the same improvement in functioning seen in people discharged home in the SLSR was seen throughout the whole country over the same time period. If we assume that, instead, the improvements were the direct result of the reconfigurations, the sensitivity analysis showed that the London and GMA models had a significantly higher probability of being cost-effective.

In addition, a limited number of data were available on which ward patients were admitted to, how long they spent on each ward and which wards they moved between prior to the reconfigurations. Instead, assumptions were made using the data collected in SINAP after the reconfigurations took place. Previous evaluations in this area suggest that admissions to the ITU reduced in London after the reconfigurations, but as no data were available for GMA or the RoE, no analysis could be done to account for this, only the sensitivity analysis, which also shows that if these improvements can be accounted for there is again a higher probability that the changes were cost-effective.

Implications

Centralised models of stroke care across an entire metropolitan area had a high probability of being a cost-effective way to improve outcomes for stroke patients. In London, this was as a result of improvements in mortality and morbidity and in GM as a result of reduced cost of stroke care attributable to reduced LOS.

What was already known about this subject?

• Evidence on implementing MSC in health care was limited.

• A realist review of the MSC literature identified five ‘rules’ for enhancing implementation, linking ‘success’ to (1) multiple forms of leadership, (2) learning from history, (3) improvement through feedback, (4) physician engagement and (5) service user involvement.

What this chapter adds

• By combining qualitative analysis of processes of change and drawing on the quantitative analysis of the impact on clinical outcomes and delivery of interventions, we were able to adapt and extend these lessons for MSC.

• Both system (top-down) and clinical (bottom-up) leadership were identified as necessary to align multiple stakeholders and thus overcome resistance to change.

• System leadership can:

  • provide political authority and power to co-ordinate multiple local stakeholders to agree to change services over a wide area

  • capitalise on clinical leadership to develop further support for the goals of change.

• Change in both areas involved clinical leadership, but although system-wide authority was used in London to align stakeholders, it was not applied in GM.

• Combining feedback with other tools (e.g. use of audit data) was important to build the case for change and to assess its impact.

• It was necessary to involve a range of stakeholders beyond physicians in planning MSC.

• Consideration should be given to how system-wide structures can be used to enable the joint planning and implementation of MSC.

Major system change of stroke services

Through the qualitative analysis of stakeholder interviews and documentary evidence, the aim of this chapter is to explain why different models were implemented in London and GM and, in doing so, enhance the Best et al. 5 framework (rules for implementing MSC based on a review of the literature) for future use in planning MSC.

In GM, the case for centralising services was made locally by health professionals to commissioning and provider leads. In London, centralising acute stroke services was recommended as part of a region-wide review of services. 28 In both areas, the process of centralisation was characterised by overlapping phases of planning and development, consultation, and implementation (Figures 8 and 9).

FIGURE 8.

Governance arrangements for centralising acute stroke services in London. HfL, Healthcare for London. Adapted from Turner et al. 110 This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 3.0 License (www.creativecommons.org/licenses/by-nc/3.0/), which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access page (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Adapted from Turner et al. 110 This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 3.0 License (www.creativecommons.org/licenses/by-nc/3.0/), which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access page (https://us.sagepub.com/en-us/nam/open-access-at-sage).

FIGURE 9.

Governance arrangements for centralising acute stroke services in GM. Adapted from Turner et al. 110 This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 3.0 License (www.creativecommons.org/licenses/by-nc/3.0/), which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access page (https://us.sagepub.com/en-us/nam/open-access-at-sage).

Adapted from Turner et al. 110 This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 3.0 License (www.creativecommons.org/licenses/by-nc/3.0/), which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access page (https://us.sagepub.com/en-us/nam/open-access-at-sage).

In London, the previous system for delivering acute stroke services involved 32 providers and the local ambulance service. In GM, nine providers and the ambulance service were involved. In both areas, a stroke project board, including providers, commissioners and patient representatives, was established to develop new service models and manage implementation. Change was governed using a ‘top-down’ approach in London, led by the pan-regional health authority (Figure 8), whereas a more ‘bottom-up’ network-based approach was used in GM, led by local providers and commissioners (Figure 9).

Informed by the EAG’s recommendations, commissioners in GM initially chose a service model in which all patients presenting at hospital within 24 hours of onset of stroke symptoms would be treated in one of three HSUs (one 24/7 CSC and two ‘in-hours’ PSCs); post-hyperacute care would be delivered by 11 local DSCs. Shortly before implementation, some hospitals raised concerns about the impact of centralisation on hospital resources and patient safety (e.g. repatriation of frail and elderly patients). Consequently, the model was revised: only patients presenting at hospital within 4 hours of stroke symptoms appearing would be admitted to a CSC/PSC; all others would be taken as before to the nearest DSC. Acute stroke services were not entirely withdrawn from any hospital. In London, the project board’s clinical subgroup recommended 10–12 HASUs, but the project board proposed fewer HASUs (eight), reflecting the earlier pan-London review’s recommendations. Subsequently, 24 providers became SUs and stroke services were withdrawn from five hospitals.

This chapter used qualitative data related to the leadership of change to analyse why different models for centralising services were chosen in London and GM, and to assess how the different approaches to leading change contributed to different centralisation models being implemented.

Data

We conducted 45 semistructured stakeholder interviews and analysed 316 documents (see Table 2 and Chapter 2, Understanding development, implementation and sustainability). Interview topic guides (see Appendix 5), based on MSC literature, covered governance approaches, proposal development, model agreement and implementation. Documentary evidence was used to develop narratives of the transformations (e.g. public consultation arrangements). The interviews represented a retrospective account of the changes and took place when there was awareness of how the services were performing in each area post implementation. To mitigate risk of bias as a result of the length of time since interviews we drew on different types of evidence,111 including interviews with diverse stakeholders and analysis of documentary evidence.

Analysis

Data analysis from interviews and documents was inductive and deductive,112 as coding was informed by themes emerging from the empirical data and a realist review of MSC literature. 5 Data relating to London and GM were initially coded separately to produce narratives of transformation. MSC across the two areas was compared using cross-cutting themes from the narratives, including receptivity to change, stakeholder involvement and experiences of service transformation. Finally, the coded data were analysed using the five rules of Best et al. 5 to compare mechanisms influencing MSC. The research team met regularly during the analysis process to discuss and agree interpretations of the data and to identify questions iteratively for further exploration. To enhance validity, emerging findings were shared with relevant stakeholders from the two study areas, including professionals and service user representatives.

Rule 1: combining designated and distributed leadership

The first rule highlights the importance of combining formal or ‘designated’ leadership with shared or ‘distributed’ responsibility for improvement. In GM the leadership of change was mainly ‘distributed’, whereas in London distributed and designated leadership were more fully combined. This meant that when there was resistance to centralising services from some stakeholders in both areas, there was a lack of system-wide leadership in GM to challenge resistance and align stakeholders (as Figure 9 shows, the pan-regional health authority was not directly involved in the programme), resulting in a less radical transformation of services in GM relative to London.

Rule 2: establish feedback loops

This rule refers to the importance of measuring outcomes that are trusted by stakeholders and incentivising improvement. Performance data were collected in both areas. In London, the designation process for hospitals was linked to achieving standards (e.g. minimum staff numbers), meaning that providers had to comply in order to receive accreditation. Providers also received a financial incentive for performing well, as ongoing performance data were monitored by the stroke networks and payments were made only if quality standards were met. In GM, a local payment system was required to split the costs of providing services between CSC/PSCs and DSCs, although payments were based on patients treated by each form of provider, not achieving standards. Thus, financial incentives were stronger in London than in GM for improving clinical standards (in GM financial penalties were considered punitive by planners and not used).

Rule 3: attend to history

The Best framework suggests the importance of learning from previous transformation attempts, including ‘failures’. 5 In London, past failures to achieve MSC meant that programme leaders focused on implementing ‘a small number of absolute priorities’ (Lon14, pan-regional health authority). Change leaders were aware that dealing with stakeholders’ differences was critical during planning meetings:

. . . not letting people go out the room if I thought actually they were disagreeing but they weren’t disagreeing in the room.

Lon17, service commissioner

In GM, decision-making based on unanimity was preferred in this case as collective decisions were historically made by consensus among commissioners and providers. One programme leader believed retrospectively that the approach taken ‘introduced an awful lot of risk that we needn’t have played into it’ (GM04, service commissioner).

In both areas, planning utilised stroke network members’ experiences of an earlier reorganisation of acute cardiac services. This highlighted a need to encourage dialogue between the ambulance service and other stakeholders. In London, insistence by the ambulance service informed the decision to take a ‘big bang’ approach to transformation whereby centralised services ‘went live’ on a single date. GM’s ambulance service expressed a similar preference, but a decision was made to implement changes to services in stages instead. A barrier to addressing the ambulance service’s preference was the need to accommodate providers’ concerns about transferring patients in the new system:

The worry was that if you suddenly changed the system, the whole system, you could become completely overwhelmed.

GM05, stroke physician

In summary, programme leaders attended to history by recognising that the existing system for delivering stroke care was unlikely to be receptive to change. In GM, programme leaders attempted to mitigate potential resistance by making decisions through consensus. However, this approach involved bowing to resistance from some providers and resulted in less radical transformation (i.e. hospital resistance trumped ambulance preferences). Conversely, the political authority with which London’s programme leaders acted was critical in being able to overcome resistance.

Rule 4: engage physicians

This rule highlights the need to engage physicians as actors who have historically had the power to influence MSC, as literature on the sociology of professions shows. 113 Although physician engagement was important in planning transformation, our analysis suggested that the involvement of a range of stakeholders was required. Ambulance services, needed to assess and transport the majority of patients with suspected stroke, were critical:

. . . it wouldn’t have happened if the ambulance service hadn’t been fully on board with it.

Lon06, Stroke Network Board

Obtaining the agreement of hospitals’ senior management was necessary, as changes to stroke provision would affect hospital income and other departments. In GM, one hospital did not bid to become a hyperacute service, despite physician support, on account of senior management concerns about A&E pressures:

. . . senior management had told [the consultant] not to, that the bid shouldn’t go in.

GM05, stroke physician

The wider financial impact of full centralisation was also understood by programme leaders. To meet both physicians’ and hospital managers’ expectations, the model for centralising services needed to achieve ‘clinical consensus’ and be a viable ’business proposal’ (GM01, Stroke Network Board).

As well as affecting stakeholders within health services, MSC in London and GM was affected by local politics. In London, the change programme included establishing a committee of local politicians to scrutinise the proposal’s public interest. The implications of services being discontinued in some areas owing to centralisation caused resistance from local government representatives:

. . . issues which caused the most angst was the removal of facilities from a certain local authority area. Every elected councillor wants to protect their area.

Lon21, local politician

In summary, although stroke physicians were key stakeholders in MSC, the geographic scale and public interest in the changes proposed meant that other stakeholders, both within and outside the health service, needed to be engaged to avoid derailment of change. Champions included senior stroke physicians, who exerted social influence over other clinicians, as the Best framework suggests. Rather than relying on physician engagement, success was enabled through dialogue between, and the alignment of, different stakeholders (e.g. among hospital managers and clinicians and ambulance staff).

Rule 5: involve patients and families

This rule suggests that leaders of MSC should include service users’ perspectives and priorities in the change process. Attempts were made to represent their views in both areas (e.g. patient organisations sat on committees for governing changes). Patients’ perceived priorities informed other stakeholders’ decision-making. In London, the initial pan-London proposal to centralise stroke care underwent public consultation. As the majority of respondents (67% of 3464) agreed with introducing specialist centres,35 this was seen to justify implementing a centralised model:

Our mandate for doing what we were doing came from that public consultation in which ‘about seven’ had been supported, and there were going to be about seven.

Lon03, pan-regional health authority

In GM, patients’ perceived needs were used to reach consensus during decision-making about centralising services. However, some interviewees doubted whether or not service users’ views, despite being sought through consultation on proposed service models, influenced transformation:

I don’t think it really changed anything . . . but at least people felt that they had a voice.

GM13, service user representative

In summary, patients and the public were consulted about, rather than involved in, the decision-making. Furthermore, their perceived views were used instrumentally by leaders to lend support to the implementation of well-defined models of care. In London, as public support for ‘about seven’ specialist centres had been quantified, this aided agreement to implement more fully centralised services. Public involvement had a political dimension as engagement was structured in particular ways (i.e. programme leaders’ framing of the options for consultation and use of the outcomes to legitimise changes to services). In GM, there was no equivalent process to establish service users’ priorities.

Implications

The combination of bottom-up leadership and top-down co-ordination of MSC is vital in navigating the complex process of its implementation. Engaging local stakeholders in planning processes is important to ensure that change builds on their experiences, is relevant to their needs, and motivates staff. 119 However, as multiple stakeholders are often involved in MSC, a co-ordinating body with political authority to bring together those different interests is needed. Regional, system-wide leadership – underpinned with the political authority to align stakeholders – should assume a leading role in supporting innovation by co-ordinating processes of MSC.

Since the transformations of stroke care described in this chapter were implemented, pan-regional health authorities have been abolished and many purchasing and performance management duties in England have been transferred to more localised commissioning consortia. 120 Consideration should be given to the collective capacity commissioners and providers have to pursue MSC. Substantial work has been conducted to devolve budgets and decision-making authority back to regions, with GM taking a leading role. 121,122 In relation to these different approaches to decision-making, consideration should be given to how system-wide structures can be used to enable the joint planning and implementation of MSC. Finally, against a background of austerity where many health systems face financial pressure, some ‘rules’ may need to be prioritised over others. Formative research could explore the feasibility of using an explicit framework of rules to inform programme leaders’ decision-making about allocating resources to the activities signified by each ‘rule’ and highlight perceived barriers and enablers to their use in MSC.

What was already known about this subject?

• Little was known about the best way to make improvements to health services across a whole geographical area or ‘system’.

• Research and theory indicated a number of interrelated components of MSC.

• Following centralisation in GM and London:

  • in both areas, LOS reduced more than in the RoE; only in London did mortality reduce more than in the RoE (see Chapter 3)

  • after centralisation, London stroke patients were significantly more likely to receive evidence-based care than in GM or elsewhere; this was in part because more patients were treated in a HASU in London (93%) than in GM (39%) (see Chapter 4).

• Analysis of planning and implementation of the changes suggested significant contributions of combining top-down, system-wide authority with bottom-up clinical leadership to engage and align multiple stakeholders across the system (see Chapter 6).

What this chapter adds

• A theory-based framework enabled an analysis of the relationships between planning, implementation and outcomes of MSC by linking quantitative outcomes with qualitative findings on process of change.

• Referral pathway: in London, where all patients were eligible for HASUs, and all HASUs admitted patients 24/7, the pathway was reported to be more straightforward and inclusive. It was more likely to be understood and followed by both hospital and ambulance staff, maximising the proportion of patients who were treated in a HASU. In GMA, the referral pathway where only a selection of patients were eligible for treatment in a HASU was found to be less inclusive and more complex than in London; this reduced the proportion of patients treated in a HASU, in part through limited adherence to the pathway.

• Phases of implementation: in London, the single launch date was identified as facilitating clear understanding of and adherence to the pathway. In GMA, phased implementation caused uncertainty among hospital and ambulance staff, both during and post implementation.

• Use of service standards linked to financial incentives: in London, standards were linked to financial incentives and services could not launch until accredited against the standards. This was reported to have increased the likelihood of services having capacity to provide evidence-based care. In GMA, service standards were not linked to incentives and there was no accreditation process, which may have led to greater variation across stroke services.

• Facilitation: the local stroke network in London was described as providing substantial co-ordination and hands-on facilitation. In GM, the network facilitated implementation by acting as a platform for sharing learning across sites.

Understanding outcomes of implementation of change

The field of implementation science articulates the need for a nuanced approach when evaluating the outcomes of change. An important distinction is drawn between ‘implementation outcomes’ (i.e. the adoption of, fidelity to and sustainability of a given intervention)5,41–44,124–127 and ‘intervention outcomes’, for example changes in the provision of care or patient outcomes. 41 This enables the study of factors that influence implementation (including the nature of the intervention and how its implementation is facilitated), and the potential relationships between these and intervention outcomes,41,42 thereby allowing insights into the ‘black box’ of implementation.

Understanding how evidence-based practice is implemented in complex settings such as health care is enhanced when its various components are considered (decision to change, intervention selection, planning and implementation of change, and outcomes). 40,42,44,125 The value of theory, as represented through conceptual frameworks, is recognised as benefiting the design, application and understanding of implementation approaches. 44,128–130 Such frameworks provide, first, an analysis of how contextual factors, such as national policy or a ‘burning platform’, can influence the decision to change and the type of intervention that is implemented. 5,128,130 Second, they indicate how characteristics of the intervention (e.g. a new service model), such as its complexity or its compatibility with local context, might influence the outcomes of implementation. 40,124,126,127,130 Third, they show how the implementation approaches employed (i.e. how change is facilitated, managed and led) can influence implementation outcomes. 1,40,43,44,128,130

However, research exploring the relationships between implementation approaches, implementation outcomes and intervention outcomes remains limited. 41

Developing a framework to analyse major system change

Drawing on the literature on implementation and MSC described above, we developed a schematic framework that identifies key components of MSC and how they might interact (Figure 10). The framework distinguished between implementation outcomes and intervention outcomes.

FIGURE 10.

Key components of MSC. Adapted from Fulop et al. 123 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Adapted from Fulop et al. 123 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

The decision to change [e.g. the drivers for change, governance and leadership of the decision-making process; see Component (C) 1, Figure 10] may influence the nature of the model (i.e. the intervention) that is implemented (C2). 5,40 Through processes of adaptation, both contextual factors (e.g. managerial capacity to lead change) and the model selected (e.g. the scale of change required) may influence the implementation approaches used (e.g. the degree to which local staff may require hands-on support in managing change) (C3). 43,127 Through both its complexity and its compatibility with the context of its introduction, the model selected may also influence implementation outcomes in terms of uptake and fidelity. 41,126 The model may influence intervention outcomes directly, although it is important that the extent to which the effects of the model are mediated through the process of implementation be considered. 41 Implementation approaches, such as how change is facilitated and local staff are supported (C3), have the potential to influence implementation outcomes (C4). 5,43 Implementation outcomes (C4) are likely to influence overall intervention outcomes, including provision of evidence-based care, clinical outcomes, patient and carer experience and cost-effectiveness (C5). 41 Finally, assessment of implementation outcomes may prompt a decision to change again and implement amended or alternative models. 127 The relationships between these components are unlikely to be linear; some (e.g. C1–3) may occur simultaneously and some components may be bypassed [e.g. model characteristics (C2) may influence implementation outcomes (C4) directly].

Major system change in Greater Manchester and London acute stroke services

This analysis was conducted having established the following (Figure 11): the drivers for MSC in both regions included national policy and local awareness of variations in and overall quality of acute stroke care provision;110 there were important differences in how the decision to change was led and governed, how local resistance was managed (see Figure 11, C1)110 and how these influenced the models selected (C2);110 the London and GM changes were associated with different intervention outcomes [London patients were significantly more likely to receive evidence-based care than patients in GM (C5)];26 and only London was associated with significantly greater reduction in stroke patient mortality than other urban regions of England (C5). 25

FIGURE 11.

Previous findings on MSCs in London and GM stroke services. NSD, no significant difference. Adapted from Fulop et al. 123 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Adapted from Fulop et al. 123 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

In this chapter, we present a comparative study of these two MSCs, examining the relationships between implementation approaches employed and the implementation outcomes (see Figure 11, C3 and C4). We address how implementation approaches and implementation outcomes were influenced by differences in the model selected (see Figure 11, C2), and how they influenced the differing intervention outcomes (see Figure 11, C5). This analysis contributed to an understanding of the relationships between the service models selected and implementation approaches applied, and how these each influenced implementation outcomes and intervention outcomes (see Figure 11, C3–5). 131

Study design

We focused on GM and London as the only examples of changes of this kind being implemented at such a scale at the time. 36 Qualitative fieldwork, combining documentary analysis and interviews, was undertaken at ‘governance’ and ‘service’ levels to compare the implementation of the changes in the two regions.

Data

We combined analyses of semistructured stakeholder interviews and documents. We conducted 125 semistructured interviews with stakeholders at governance (n= 45) and service (n = 80) levels over the period April 2012 to December 2013 (see Table 2, and Chapter 2, Understanding development, implementation and sustainability). Interviews at governance level covered background to the centralisations (including drivers for change), governance, developing the proposal for change, agreeing the model, implementing changes, the impact of centralisation and reflections on the changes (see Appendix 5). Interviews at service level covered background to changes, processes of service development, the impact of centralisation and reflections on changes (see Appendix 5). In addition, 653 documents were collected from governance and service levels (see Table 2 and Chapter 2, Understanding development, implementation and sustainability).

Analysis

We compared the London and GM changes in terms of the implementation approaches employed and the implementation outcomes. Findings were considered in relation to our previously published findings, that is, the different models implemented110 and their differing impact on intervention outcomes (likelihood of patients receiving evidence-based care26 and patient mortality25).

Data analysis from interviews and documents combined inductive and deductive approaches,112 as themes were drawn from our framework (see Figure 10) and emerged from the empirical data. Documents were analysed to identify various aspects of the changes, including drivers, key events and activities, and overarching chronology. Interviews were analysed to draw out similar information and to understand why and in what ways aspects of implementation were influential, in order to compare the two regions. Analysis took place in two phases, building on the narrative summaries and timelines of the changes developed from documentary analysis used in a previous analysis. 110 In phase one, service-level narrative summaries were developed, using the constant comparative method,132 from documentary evidence and initial readings of interviews. These were developed separately for the changes in London and GM (by the authors AIGR and CP) and covered a number of cross-cutting themes: service-level context, service development processes (including thrombolysis and repatriation protocols, recruiting and training staff), launching new services and perceived impact of changes. In phase two, we used the overall timelines and summaries and service-level summaries to identify key tasks in implementing the models in each region, and contrasts in how these tasks were accomplished. In phase three, a subgroup of the authors (CP, AIGR, SM and NJF) applied the framework (see Figure 10) to a cross-region analysis that sought to test explanations of the differing implementation outcomes identified in previously published quantitative analyses. This phase drew on further thematic analysis of interview and documentary data to identify factors influencing the contrasting implementation approaches, and how the approaches may have influenced the resultant outcomes.

To enhance reliability, emerging findings from each phase were shared and discussed regularly with other co-authors until agreement was reached. To enhance validity, an interim version of this analysis was shared with people who had been involved in the planning and implementation of the changes in London and GM (some of whom we had interviewed for this study).

Factors influencing implementation approaches

The approaches to implementation (Figure 12, C3) differed across the two regions on the basis of the degree to which it was phased, the degree to which it was linked to standards set out in service specifications and financial incentives, and the degree to which networks provided hands-on facilitation.

Factors influencing implementation outcomes

As previously established (see Figure 11), there were differences, first, in implementation outcomes (i.e. greater fidelity to the referral pathway in London than in GM) and, second, in the greater likelihood of providing evidence-based care in London than in GM (with provision equally high in London HASUs and GM CSC/PSCs, but lower in GM DSCs). 26 We first discuss factors influencing fidelity to the referral pathway (model complexity, ‘big bang’ vs. phased implementation, and the degree of ‘hands-on’ facilitation by networks) (Figure 12, C4). Second, we discuss factors influencing service development (use of service specifications and ‘hands-on’ facilitation). As set out in the preceding section, many of the factors influencing implementation outcomes related to implementation approaches, including the degree to which implementation was phased, the use of standards and financial incentives and the degree of hands-on facilitation.

Understanding outcomes of major system change

In this section, we bring together the current findings with those from previous analyses to illustrate how components of MSC (see Figure 10) contributed to the significantly different outcomes associated with the changes to acute stroke services in GM and London. These relationships are summarised in Figure 12, and described below.

The changes in London and GM appeared to be influenced significantly by the degree to which change leaders ‘held the line’ on the models to be implemented (see Figure 12, C1 and C2). 110 The models implemented and the approaches employed played an important role in the outcomes observed.

London’s inclusive 24-hour model (i.e. all suspected stroke patients were eligible for HASU), requiring relatively few referral decisions to be made, increased the likelihood of staff following the referral pathway. In contrast, GM’s 4-hour model was significantly more selective (limiting the number of patients who were transferred to a CSC/PSC) and complex, increasing uncertainty among staff about where suspected stroke patients were to be treated (see Figure 12, C2). Furthermore, these models were implemented differently, reflecting a contrast in the degree to which implementation was actively managed in the two regions (see Figure 12, C3). London adopted a ‘big bang’ approach whereby the new system was launched on a single date, increasing the likelihood of the referral pathway being followed. This launch was dependent on services being accredited against standards linked to financial incentives, increasing the likelihood of services providing evidence-based care. Significant ‘hands-on’ facilitation was provided by the London network to ensure that services met the required standards. In GM, services launched in multiple phases, which limited confidence in the referral pathway, and service specifications were not linked either to service launch or to financial incentives; this may in part have limited development of DSCs.

Implementation outcomes (see Figure 12, C4) had a significant influence on intervention outcomes (see Figure 12, C5). Almost all London patients were treated in a HASU, and all HASUs were likely to provide evidence-based care; this meant that London patients were overall more likely to receive evidence-based care and in turn had a larger reduction in mortality than patients in GM and RoE. In contrast, GM patients were far less likely to be treated in a CSC/PSC, with two-thirds treated in DSCs, which were significantly less likely to provide evidence-based care; as a result, GM patients’ likelihood of receiving evidence-based care, and associated mortality, did not differ significantly from in the RoE. 25,26 The 12-month review in GM noted national audit data indicating that DSCs were providing evidence-based care less frequently than CSC/PSCs. 32 Based on this information, and discussion with an EAG, it was agreed that further reconfiguration of acute stroke services should be explored. 110

Principal findings

This chapter examined the complex, non-linear relationships between the type of model selected, implementation approaches, implementation outcomes and intervention outcomes. By analysing the centralisation of acute stroke care in two regions, distinguishing between implementation outcomes and intervention outcomes (following Proctor41), we made a significant contribution to the understanding of MSC,5 specifically in terms of the factors influencing outcomes. 1 We demonstrated a number of inter-related factors that potentially influenced such outcomes (as detailed in Chapters 3 and 425,26). Certain characteristics of the intervention and implementation approaches were associated with more positive implementation outcomes and intervention outcomes, and many of these reflected existing implementation and diffusion theories, described in Comparison with other studies below. 5,40–44,126–128 In turn, the service model and implementation approaches were influenced significantly by leadership and engagement factors described in Chapter 6. 110

Strengths and weaknesses

A key strength of this analysis is that it took a theory-driven approach to analyse two cases of MSC or MSC across two geographical regions in order to establish the relationships between the planning, implementation and outcome of MSC. This analysis had a number of limitations. First, because this research was retrospective in nature, interviewees were looking back on the changes, with some awareness (e.g. by monitoring national audit data) of the degree to which changes had succeeded in influencing the provision of evidence-based care. This may have affected the way in which they articulated their views of the implementation of the changes. Future research on changes of this kind would benefit from being carried out contemporaneously, ideally from the pre-implementation stage, and extending over a sufficient time period to allow a formal evaluation of the impact on intervention outcomes, such as patient mortality. Second, although we believe that our data indicate that the identified implementation strategies played a significant part in the implementation outcomes observed, the relative contribution of each component cannot be established. Third, we sampled only a proportion of services in this study, and other factors may have been important to implementation in other services in the reconfigured systems. However, we believe that by conducting interviews at the pan-regional level, and sharing findings with local stakeholders, our findings provide a strong representation of implementation in both regions. Finally, studies of MSC in other acute (and non-acute) care settings would be of value to aid identification of potentially generalisable lessons.

Comparison with other studies

In terms of intervention characteristics, we found that in this case ‘simpler’ and more inclusive referral pathways (such as London’s 24-hour model) were more likely to be understood and followed by both hospital and ambulance staff. This effect might reflect such established concepts as ‘feasibility’,41,128 ‘compatibility’126 and ‘complexity’,40,126 whereby an intervention is more likely to be adopted if it is readily incorporated into existing or standard activities.

The concept of ‘execution’ (i.e. where implementation is achieved as intended40) was highly relevant to this analysis. In terms of timeliness of implementation, the advantages of a ‘big bang’ launch and associated planning, and the disadvantages of phased implementation, were clear: a single launch date gave clear understanding across all stakeholders when implementation was complete. This finding was potentially counter-intuitive, given previous research indicating risks related to ‘big bang’ implementation. 133 However, in changes such as these, when service models have multiple interdependent components, a ‘big bang’ approach appears to be beneficial and represents an example of ‘adaptation’,43,127 whereby an implementation approach is selected to reflect the scale of the task and the complexity of the new system. Further research is required to establish the extent to which this finding applies to similar changes in different health-care settings. Also associated with ‘execution’ was the use of standards: linking the launch of the new model to achieving standards appeared to increase the likelihood of there being uniform capacity to provide evidence-based care and also gave a shared understanding of what had to be delivered in all services. By associating the achievement of standards with financial incentives, the London changes reflected the hypothesised benefits of altering remuneration to encourage adoption of the intervention. 127 The contribution of ‘hands-on’ facilitation, for example by external change agents, to implementation outcomes is acknowledged by various implementation frameworks. 40,43

These differences in relation to a ‘big bang’ versus phased launch, the use of standards and hands-on facilitation reflect an underlying contrast in the implementation approaches adopted in our studied regions, with implementation in London facilitated significantly more actively than in GM.

Implications

This chapter used a framework that draws on key features of change identified in existing implementation theory to analyse two examples of MSC in acute stroke care. We found that the model selected (in terms of simplicity and inclusivity) and implementation approach (launch date, prioritisation of standards, financial incentives and hands-on facilitation) made significant contributions to implementation outcomes observed, and in turn intervention outcomes. 25,26 We explore these factors further in Chapter 10 (in relation to efforts to implement change across the Midlands and East of England), Chapter 13 (in relation to further centralisation in GM) and Chapter 14 (in relation to the sustainability of the London system).

We believe that this analysis has demonstrated the value of considering the interdependencies between intervention, implementation approach and outcomes when planning and evaluating MSC. However, the particular relationships identified in this analysis may vary depending on the nature of the change being implemented. The framework described in this chapter would be strengthened through further use in evaluating MSCs conducted in other health-care settings.

What was already known about this subject?

• Patient and public involvement is required where changes to care provided by the UK NHS are proposed. The manner in which involvement should be put into practice is not prescribed, and there is little evidence about which methods should be used in which circumstances. Therefore, involvement is characterised by ambiguity about its rationales, methods and impact.

• Our analysis of the planning and implementation of the GM and London changes (see Chapter 6) indicated that PPI was used instrumentally by programme leaders to demonstrate support for proposals rather than with the intention of developing co-designed services.

What this chapter adds

• The study adds new empirical knowledge to the relatively small body of literature on involvement practices in MSC.

• Lay involvement was enacted through consultation exercises, lay participation in governance structures and the elicitation of patient perspectives.

• Interviewees’ views of involvement in these MSCs varied, reflecting different views of involvement per se and of implicit quality criteria.

• The value of involvement was felt to lie not in its contribution to acute service redesign but in how involvement practices facilitated its implementation. Our analysis identified three types of processes – agitation management, verification and substantiation – through which this was achieved.

Participants

In-depth interviews were conducted with 17 people in GM and 26 in London who were involved in the governance of the changes (see Table 2).

Overall, participants were aware of the need to involve patients and public in the proposed changes, confirming the view that PPI has acquired normative status among professionals in the NHS. Interview data also suggested that there were concerns about the changes from the outset. Hospitals and professionals were concerned about implications of the new models of care in terms of loss of resources and, in London, about the closure of five acute stroke wards. There were concerns about patient safety and acceptability because the new models meant journeys for some patients beyond their locality, as well as transfers from one site to another. The need to engage all types of stakeholder was recognised if the proposed changes were to be implemented.

Involvement practices

At both sites, multiple strategies were used to involve patients at a range of levels. Existing resources were accessed (such as NHS involvement managers and individual voluntary sector organisations); social and other media were used to invite patients to participate. In London, an involvement strategy was developed that outlined plans to access a range of public and patient groups and highlighted the need to access ‘hard to reach’ groups. Interviewees in both areas identified different types of activity through which patients and public were involved, namely stakeholder events and consultations, lay membership of governance structures and the formal elicitation of patient perspectives. No similar single strategy document was developed in GM but project documents, such as that detailing the governance framework, identified the organisational structures through which PPI, and public awareness and education, would be developed and implemented. 159

The quality of involvement

Our interviews did not ask participants to comment on whether involvement was meaningful or tokenistic, but many offered their own views of the quality of what was done, with wide variations in their appraisals.

Interviewees suggested that in GM there was little formal consultation of patients and the public. Public events were described as designed to provide information about the planned changes rather than to elicit views. For example, a commissioner said:

I don’t think at any stage we said to the public of Greater Manchester, if there’s such a thing – and there isn’t – we didn’t say, ’Do you want three of these, or five of these?’ We never said that.

GM01, Commissioner, GM

In London, where there were sustained efforts to hold formal consultations, two of three NHS managers interviewed were positive about the consultations because of the efforts made to encourage high levels of participation and to be inclusive. An interviewee from the voluntary sector challenged this view, arguing that consultation was not sufficiently accessible to people with stroke-related aphasia, for example. Others suggested that the consultation was useful because it secured buy-in from the public or ‘political legitimacy’. However, one commissioner, despite conceding that consultation was necessary because of societal expectations of transparency in public services, also expressed concerns about the limits of the public’s knowledge to comment on proposed new plans. Others, including a commissioner and two doctors, either were dubious, expressing doubt about the usefulness or validity of consultation, or considered it ‘a complete waste of time’ (L02) because consulting the public had eclipsed the need to consult more widely with professionals providing long-term services to stroke survivors.

There were also different evaluations of individuals’ contributions to governance structures. Several interviewees from GM spoke about the lone activist who was appointed to the project board because of previous professional political experience, his ability in committee work, his history as a campaigner for stroke service quality and even his willingness to challenge others. Others rehearsed well-known arguments about the limits of individuals’ contributions based on the fact that they were self-selected, or the ‘usual suspects’ and, therefore, unrepresentative, or because they lacked experience in formal committee work.

The limits of how actively involved lay people could be in this involvement process were also recognised. For some, this was related to the nature of MSC itself. It required knowledge of a range of complex problems such as population needs, resource implications and political implications. Because the designs had been worked out by a core group of professionals, there was limited opportunity for lay people to influence the service design. For some interviewees this meant that there was ‘no real involvement’, whereas others conceded that, although this might not be patient-led involvement, consultation processes allowed patients to become ‘advocates for the model’.

The trope of the patient voice also figured in interviewees’ accounts. Involvement was seen as an opportunity for the patient voice to be articulated, represented by interviewees as an important corrective to the dominant perspectives of professionals and organisations. For example, one said:

People (professionals) have to, really have to be brought back to what’s best for the patients and an awful lot of what gets discussed in the NHS is not about that, it’s about what’s best for my organisation.

GM05, stroke physician, GM

However, a minority of interviewees argued that although patient voices may have been heard, they could not be acted on within the scope of the consultation activities, which were specifically limited to the MSC projects’ redesign of acute stroke services. Nevertheless, these interviewees noted that patients took the opportunity to articulate another concern, namely the quality of rehabilitation services. One interviewee clearly made this point:

. . . every single question which was asked in the half hour or so that the meeting was thrown open for questions from the floor was about rehab [rehabilitation] . . . Nobody asked a question or protested about the decisions we were making about the hospital service but all the members of the public and their representative questions were about rehab.

Lon03, project team, London

Patients may have been content to leave acute service design to professionals who carefully made a convincing case that centralisation was able to deliver better quality acute care for all. Yet the question of rehabilitation services was outside the brief of change leads, and even when involvement permitted dialogue between stakeholders, the question of rehabilitation and longer-term care was inadmissible.

Constructing value

It could not, therefore, be argued that involvement in these examples of MSC influenced or improved the design of the acute stroke services. As we have previously argued, involvement was used instrumentally by programme leaders to gain support for change, the case for which had already been made, and for service models already developed. 110 For a minority of interviewees, this indicated a failure to achieve an ideal of patient-led involvement but, for most, even if flawed, the practices of involvement had intrinsic value for the implementation of MSC. We identify three types of value suggested by the interview data; these are summarised in Table 15 and discussed in detail in the following sections.

Principal findings

This study drew on project documents and qualitative interview data with a wide range of professionals engaged in complex and protracted processes to redesign acute stroke care in two English cities. The case for MSC was constructed by professionals drawing on clinical and managerial experience and an examination of population-level patient data that demonstrated the need to improve access to best evidence care. This was consistent with NHS strategic guidance that requires service-level changes to be clinically led and underpinned by clinical evidence. 166 This set the parameters of involvement from the outset; the case for change was professionally led but the co-operation and approval of a wide range of stakeholders including clinical staff, NHS managers and local politicians was required. PPI was a tool to facilitate implementation of the changes. In this sense, involvement could be seen as instrumental, achieving the outcomes desired by professionals. 110 However, we would further argue that rather than either ‘tokenistic’, suggesting a cynical position, or ‘meaningful’, implying conformity with some a priori agreed definition of what involvement means, involvement here was enacted in strategic ways.

Interviewees’ accounts varied widely in how they evaluated involvement. Consultations were seen as, at worst, a waste of time to, at best, wide reaching, inclusive events in which the patient voice could be heard and professional transparency be demonstrated. The contribution of individuals taking part in governance structures was also differently viewed. They were variously portrayed as powerful voices reminding professionals – at risk of promoting their own interests – of their true purpose, as making a limited contribution because of their limited competence in meeting behaviour and as self-selected and unrepresentative.

This implied that professionals controlled not only the agenda but also the manner in which involvement was enacted. 117,153 This may be inevitable. The model of involvement that dominates the NHS requires professionals to invite lay people to participate in activities that professionals design, focused on questions that they identify. This differs from PPI’s political antecedents: self-organising patient movements, which drew on principles of social justice and emancipatory practice to challenge biomedical definitions of illness and solutions, counter discrimination and stigma and call for action into emergent health problems. 167 Although involvement may promise a transformation in relations between patients and professionals, as Komporozos-Athanasiou et al. 168 have argued, the ritual nature of PPI activities constitutes ‘a conservative form of engagement in health’ that serves to reinforce existing statuses, neutralising the transformational potential of involvement.

Strengths and weaknesses

Our study provided professionals leading MSC with the opportunity to reflect on involvement practices in two MSC projects in the broad context of service change, rather than focusing on involvement alone. Participants were from a wide range of backgrounds and professions, and so a number of perspectives were captured. This is one of a relatively small number of studies investigating involvement in MSC. The data are limited in that they are retrospective, rather than contemporaneous accounts. The diversity of interviewees and their role in MSC means that each provides something of a partial view of which activities were undertaken and by whom. What emerges is a rather complex picture of diverse activities from information-giving events through to research to collect accounts of patient experience, framed by participants as involvement. The data do not necessarily provide an accurate historical record of PPI in the two MSC projects but they offer a moral account of implementing PPI in the projects.

Implications

Our study offers lessons for thinking about involvement in general and in relation to MSC. In particular, the findings represent a challenge to contemporary concerns that the literature reports processes of involvement but fails to report on impact, and that improved methods to demonstrate impact are required. 137,169 Our interviewees’ accounts did not suggest that it was possible to demonstrate impact of involvement in a linear way, because involvement was designed not to effect but to support change. Involvement here was a strategically symbolic process that served to use the moral authority of the imagined but substantiated patient to support change implementation. Thus, involvement as enacted also reiterated the significance of involvement itself. Conklin et al. 157 and Li et al. 156 have suggested that, rather than focus on impact, we should consider the quality of involvement in processual terms, either as democratic acts in their own right or as strategic acts of informing and legitimising.

Our study also found limitations to involvement as a democratic process. The MSC sought to effect change in acute stroke care and this set the parameters of what was admissible; thus, patients’ concerns about the quality of care needed after discharge from hospital were rendered irrelevant. In London, the need to consider the whole stroke pathway was acknowledged in the Healthcare for London report,160 but this acknowledgement was made after the event and it did not lead to a sustained effort to effect MSC in the priority area identified by patients.

Nevertheless, most participants in the study believed that involvement activities had intrinsic value, facilitating the implementation of MSC. The value attributed to involvement sustained the idea of involvement itself because, as the anthropologist David Graeber has remarked, value can be considered as the way in which specific activities are made meaningful to those involved. 170 Investigating how value is produced – and for whom – through involvement might offer a way of rethinking impact assessment in involvement, which Edelman and Barron158 have faulted for treating as if it were an intervention in its own rather than something integral to a larger process. As these authors suggest, rethinking impact requires revisiting the goals and purpose of involvement. This study further suggests a need to identify those goals and purposes that are shared by different constituencies, because we do not know if patients and the public who were involved in MSC in London and GM would have recognised the value that emerged from our interviewees’ accounts.

What was already known about this subject?

• Patient experience of stroke care has been reported as varied. Patients treated on acute SUs are more satisfied with their care than those on general wards and, more broadly, well-organised stroke care is associated with more positive patient experiences.

• The impact of centralised acute stroke care pathways on patient experience has not been studied in depth.

What this chapter adds

• Although the referral pathways differed, similar patient and carer experiences were reported in the two regions.

• Patients and carers on the centralised acute stroke care pathways reported many positive aspects of care.

• Participants were impressed with emergency services and the initial reception at hospital; disquiet about travelling further than a local hospital was allayed by clear explanations.

• Participants described that they knew who was treating them, were involved in decisions and had adequate specialist stroke care.

• Difficulties for families visiting hospitals a distance from home were raised. Repatriation to local hospitals was not always timely, but no detrimental effects were reported. Discharge to the community was viewed less positively.

Centralised care pathways

Patients treated on acute SUs are more satisfied with their care than those on general wards,172–174 and, more broadly, well-organised stroke care is associated with more positive patient experiences. 175 However, the impact of centralised acute stroke care pathways on patient experience has not been explored in depth. Centralised services may affect patient experience in a number of ways. Services are likely to be relatively high volume, and patient satisfaction with stroke services has been reported to be lower in larger stroke services. 176 Care may be provided in an unfamiliar environment, with travelling distances increased for patients and families. 177 Payne et al. 178 reported that travel for cancer treatment had been described as inconvenient and could be perceived as a barrier to treatment. However, for another clinical issue, Sampson et al. 179 concluded that, although it may be perceived as inconvenient, people would travel further in order to access centralised angioplasty services. One survey study of the experience of patients and carers of the newly centralised stroke care pathways in London reported that the majority of stroke patients and carers were either happy or did not mind being treated in a more distant HASU, and that although some concern was expressed about repatriation, only 6% reported any negative effect of the transfer on recovery or outcome. 177

The importance of patient and carer experience

The definition of quality in health care has expanded to include patient experience,180 and the concept is prominent in the measurement of health service performance. 181 Patient experience is now a recognised component of high-quality care in the English NHS, alongside patient safety and clinical effectiveness. 182 It is mandatory for NHS providers to gather patient experience data. 183

Although there is no single universal definition of patient experience,180 many definitions reflect that of The King’s Fund Point of Care Programme: ‘the totality of events and interactions that occur in the course of episodes of care’. 184 Patient experience is more than ‘patient satisfaction’, and asking patients ‘what happened’ during a specific episode of care is more valid in judging quality of care than just asking about ‘satisfaction’. 175 Good patient experience includes a focus on individualised care, and is tied to patient expectations and whether or not they are positively realised and to the principles of patient- and family-centred care. 180 Positive patient experience is associated with patient safety and clinical effectiveness183 and understanding how patients experience care can highlight substandard care. 185 In 2012, the National Institute for Health and Care Excellence (NICE)186 produced a quality standard to provide the NHS with clear commissioning guidance on the components of a good patient experience: 14 quality statements against which patients’ experience can be measured are presented in Table 16.

What was already known about patient and carer experiences of acute stroke care pathways?

The literature on patient and carer experience of acute stroke care provides evidence in relation both to the various stages of the acute stroke care pathway, and also to a number of cross-cutting issues that relate to all stages of care.

Sample

Patients were recruited from the research case study sites: three sites in GM (the sole 24/7 CSC, one of two in-hours PSCs and one of 11 local DSCs) and four sites in London (two of eight 24/7 HASUs and two of 24 local SUs). Any patient diagnosed with stroke was eligible for inclusion provided that they had adequate cognitive function. Purposive sampling ensured that patients experiencing different elements of the centralised pathways, such as repatriation to a local SU/DSC or direct discharge home from a HASU, were included.

Participant recruitment and data collection

Recruitment and data collection occurred between April 2013 and May 2016. Potential participants approached shortly before discharge from hospital by a research nurse or clinician were asked if they were willing to speak to a researcher. The researcher explained the study, and if patients were willing to participate, permission to contact them after their discharge from hospital was obtained. Patients were interviewed at home within 3 months of discharge, with fully informed written consent. Carers were included if the patient wished, or if they were incidentally available at the time of the interview and the patient was happy for them to contribute; they were asked about their perceptions of care received by the stroke patient. Most interviews lasted between 45 minutes and 1 hour. A semistructured interview schedule was used, covering experiences from onset of symptoms to experience once home. This interview guide was developed with reference to the literature reviewed, to established recommendations such as the NICE quality standards for patient experience,186 and in relation to the new care pathways. A patient coinvestigator assisted with development of the guide, which was also discussed with the Study Steering Committee (including patient representatives) and a stroke patients’ research group. With the permission of participants, interviews were digitally recorded and then professionally transcribed.

Data analysis

A thematic analysis was undertaken, initially using a deductive approach guided by a baseline framework developed from the literature (see Appendix 6, Table 47). Two of these themes (explanation and information and person-centred approach) were cross-cutting in nature and reflected many of the NICE quality standards for patient experience. 186 Two people (CP and IP) used the baseline framework to analyse early interviews, with some transcripts analysed by both to ensure consistency in data coding. As analysis continued, an inductive approach was used and the final framework was developed to mirror the centralised stroke care pathway (Table 17).

The analysis framework had five main themes: (1) initial transfer to hospital, (2) reception at hospital, (3) in-hospital care, (4) repatriation to local hospital and (5) discharge home. The cross-cutting issues identified in the original framework were evident throughout the stages of the pathway. In order to enhance reliability, the emerging analysis was discussed with a subgroup of the authors (AIGR, NJF, CM and RB). To enhance validity, interim versions of the analysis were presented to stroke patient support groups.

Initial transfer to hospital

Most people who experience stroke are transported to hospital by ambulance. Participants in this study reported that ambulances arrived quickly and ambulance staff gave clear information about likely diagnosis, which served to reduce anxiety. Transferring people to a more distant HASU/CSC/PSC, thus bypassing a local hospital, was voiced as a potential concern by those planning the reconfiguration and concerns were expressed to ambulance staff, particularly by relatives:

We’re going further, that’s going to take longer, what happens if it gets worse on the way there?

GMHp03, family member

However, once explained that transfer was to a specialist unit, fears were usually allayed. One patient described that they were told they would ‘go to the right place that would sort me out’ (LonAp05, patient, London) and another stated that:

They said we’re taking you to [CSC/PSC] because they’ve got a specialist Stroke Unit there, effectively, and I said, ‘well that’s fine’.

GMHp03, patient

The experience with the ambulance service of a woman who took herself to the A&E at a local DSC was less smooth, however. Local DSC staff considered she was eligible for, and would benefit from, CSC/PSC care, and arranged for her transfer by ambulance to a CSC/PSC. The ambulance staff were not clear about what was happening, which was unsettling to the patient who said she had to explain her transfer to them. She commented:

I’d never been in an ambulance before, which was daunting in itself, and then the ambulance man was sort of talking and saying, ‘Well we’ve not had a proper handover, we don’t know what’s going on’.

GMHp07, patient

Transfer from a local SU/DSC to a HASU/CSC/PSC happened relatively infrequently, which may explain the lack of adequate handover to ambulance staff and why they were unclear about the procedure.

Reception at hospital

Transfer from ambulance staff to hospital teams was perceived as smooth. Participants were impressed with the reception they received on arriving at hospital. In the centralised care pathway, HASUs were organised so that stroke teams met the patient on arrival, following a pre-alert from the ambulance service, and this enhanced participants’ experience of care. Participants reported receiving timely investigations and treatment and that the teams who treated them knew what they were doing. Their perception was that stroke was treated as a priority and a medical emergency:

You went in and they were so ready for him, I know they’d radioed through, I know they were prepared for him.

GMGp02, family member

In GM, patients admitted to a local DSC rather than a CSC/PSC also reported timely care, stating that admission to hospital was ‘very expeditious’ (GMHp01, patient) and that there was ‘no waiting . . . attention was very quick’ (GMHp04, patient).

In-hospital care

Generally, patients were happy with in-hospital care, with most indicating that they knew who was treating them, they received clear explanations and they were involved in decisions about their care. In addition, patients reported receiving adequate therapy, which they perceived had aided their recovery. There was some lack of continuity reported in the evenings and at weekends. For example, participants commented that staff in the evenings and at weekends were unfamiliar to them, and that some were not so clear about their/their family member’s care, which caused some anxiety.

Although patients did not mind being admitted to a more distant CSC/PSC, visiting for families and friends was raised as an issue which worried them:

It was a bit awkward being so far away.

GMHp07, patient

I can imagine it would affect people if they were in Kent or something.

LonAp04, patient

Carers recounted difficulties in visiting a more distant hospital. One said that ‘it was so expensive . . . well I were [sic.] there twice a day’ and also explained the impact that the distance to travel had on her:

Back home again, you have no time. I think I’d get home, took the dog out, come back and go again. There was just no time and, you know, you couldn’t just not go.

GMGp02, family member

Repatriation to local hospital

The repatriation of patients from a HASU/CSC/PSC to local SUs/DSCs was something about which concern was expressed by those planning the reconfigurations. As part of the centralised stroke pathway, patients who were admitted to a unit that was not their local hospital were returned to their local SU/DSC after they had received their acute care, if they were not well enough for discharge home. For participants in this study, repatriation happened relatively smoothly and patients were mostly happy about the way in which they were transferred from the HASU/CSC/PSC to their local SU/DSC:

Once they told me yes there’s a bed available they then came and said we’ve ordered an ambulance and it will take between 1 and 4 hours to come, I remember them saying that. But it came well within 4 hours, under 4 hours.

GMHp04, patient

Being kept informed about what was happening contributed to a smooth transfer:

They kept me informed of what was going on, that was . . . I think is the most important thing. Rather than just leaving you laying there or sitting there as the case may be, not knowing whether it’s daytime or night time.

LonCp01, patient

Most people perceived that their care was continued smoothly once they were transferred and that staff were aware of what had happened to them, and that repatriation did not have any impact on the trajectory of their recovery. Some commented favourably on the increase in therapy input once they had been transferred (an increase which would be expected as local units were focused on rehabilitation).

However, some difficulties were described with repatriation. It was reported that hospital staff were not always sure which hospital a patient should be repatriated to; this may reflect the fact that staff were learning to work with a new care pathway and had initial uncertainties. Some people also described delays in obtaining a bed at a local SU/DSC. Although this could be frustrating, patients generally accepted this situation if it was explained to them that they were waiting for a bed to become available, as long as they were kept informed about what was happening. This emphasises the importance of clear information. A family member explained that:

We were waiting for a bed to become available at [local unit], that was the reason he was in [CSC/PSC] a bit longer.

GMHp02, family member

This can be contrasted with the experience of another family member who did not understand what was happening:

Somebody told us she would definitely be going at one time, then she didn’t go and then somebody else said no . . . you know it was a little bit confusing.

LonDp02, family member

Once it was decided that a patient was to be repatriated and a time for transfer was given, some delays in transport to a local SU were described. This kind of delay was not tolerated well by patients and their families:

We weren’t very happy if you recall at the time with transfer from HASU to local Stroke Unit, because it took 6 hours, which left both of us in a very het up and upset state.

LonDp01, family member

For some people, delay in transportation to a local unit resulted in transfer happening later in the evening, which was another situation patients were unhappy about. One patient described his experience:

The next day they said they wanted to send me to [hospital], which was the nearest hospital to home. I set out, I didn’t set out, they said you'll be going later on in the day and an ambulance would come for me. Well I sat around all day and nothing happened and by half past nine at night no ambulance had arrived so I said well I'm not going, I'm not going to be carted in the middle of the night through a big city and you know it was snowing.

GMHp09, patient

This person also described being transferred to a bed on another ward for one night because of the pressure on CSC/PSC beds.

Discharge home

With centralised acute stroke care pathways some HASU/CSC/PSC patients would be discharged home to a different area from where they had received their acute care. This potentially posed difficulties to hospital teams that did not know the local processes of care, or the teams to which they were discharging people. In terms of transfer between hospital and community, most participants thought that communication between hospitals and GPs happened effectively and that their GP was aware of their stroke and the care they had received. However, some people were not clear about their follow-up once home and were unsure about when, whether or how this was to happen, or experienced some delay. For example:

It’s unclear even to me today what’s going to happen with physiotherapy in the future because apparently there is . . . a waiting list and I’ve not heard much from them.

GMHp03, patient, GM

Being discharged at weekends was also problematic in terms of arranging adequate care once home, and in practice rarely happened. This meant that some people stayed in hospital for longer than necessary, or that they had to wait to receive ongoing care once home, which was perceived to slow down recovery.

Principal findings

This chapter has explored patient and carer experience of centralised acute stroke care pathways in two metropolitan areas. Similar experiences were reported by participants from the two different geographical regions, which is perhaps unsurprising: although the care pathways differed in terms of who was eligible for HASU/CSC/PSC care (those presenting within 4 hours of symptom onset in GM and all patients in London), patients went through similar stages of care in both locations following stroke. The findings contribute to knowledge about patient experiences of acute stroke care services and also to the wider body of knowledge relating to the centralisation of services in general. The data also demonstrate how patient experience can provide valuable information about how a service is operating, what is working well and what is not. 185 For example, one patient’s observation that ambulance staff were not sure why they were transferring her to a different hospital indicates that the pathway was not operating optimally.

Strengths and weaknesses

Strengths include that this was a rare opportunity to compare patient and carer experiences of two centralisations of acute stroke care, which differed in terms of service model implemented and impacts on care provision and clinical outcomes. A range of experiences were represented in terms of whether people were admitted to a HASU/CSC/PSC or a local SU/DSC, were repatriated to a local SU/DSC or discharged straight home. Interviews took place within 3 months of discharge from hospital to enhance recall of events. In addition, interim findings of the analysis were presented to stroke patient support groups for their comment.

In terms of weaknesses, only stroke patients who were cognitively able to participate in an interview were recruited into the study, and it is possible that the experience of those who had a less positive outcome after their stroke was different. The study was of centralisation of stroke care pathways in two metropolitan areas of England; centralised services in more rural areas may well be experienced differently by patients and carers. Finally, some patients taken onto the centralised acute stroke care pathways in GM and London were ultimately not diagnosed with stroke. These so called ‘stroke mimics’ were thus transferred to a hospital more distant from their homes with no particular benefit for themselves and were not part of this study. It is important that the experience of this group of patients is analysed in any overall evaluation of centralised acute stroke care pathways.

Comparison with other studies

Participants in this study were impressed with both their contact with the emergency services and their initial reception at hospital. Their experience of timely investigations and initial treatment suggested that stroke was treated as a priority and a medical emergency. This was in line with the National Stroke Strategy24 and in contrast to some earlier studies. 189 Once admitted to hospital, patients described that they knew who was treating them, they received clear explanations about their care, were involved in decisions about care and had adequate access to therapy. This was again in contrast to much published literature on patient experience,178,192,195 but reflected what is known about the relationship between well-organised stroke care and more positive patient experience. 175 The extent to which timely investigation and treatments, and the availability of therapy, can be attributed solely to the centralised acute stroke care pathways is difficult to discern, as national initiatives such as the National Stroke Strategy24 were current at the time of the centralisations in GM and London and would have driven such improvements in care. However, the centralisations introduced HASUs/CSCs/PSCs, which are associated with a greater likelihood of receiving evidence-based care interventions. 26

Other findings from this study of stroke care pathway centralisation are relevant to the centralisation of any services where patients are taken to more distant care settings and discharged to the community from there, or repatriated back to a local hospital. There was some evidence in our data of these processes of care impacting upon patient experience. Patients, and particularly family members, expressed some disquiet on being informed that they were going further than their local hospital, but were generally reassured once an explanation was given. This emphasised the importance of clear explanations by the paramedic team on initial contact with stroke patients, and reflects NICE quality standards regarding communication. 186 The importance of effective and timely information provision, as shown in the literature, was clear at this stage of the pathway. Difficulties for families visiting hospitals a distance from their homes were discussed, in terms of time and financial costs, but patients and carers broadly prioritised quality of care and outcomes over the issues presented by being cared for at a more distant site. This was similar to the survey findings of Moynihan et al. 177

Repatriation was not always smooth in terms of happening in a timely manner (within 72 hours), but on the whole worked well from the perspective of patients. When there were issues, the importance of communication with patients and carers was emphasised. If patients were informed about what was happening, knew where they were being repatriated to and when, and had the reason for any delays explained, they were more likely to report a good experience. This reflected the NICE quality standards for patient experience186 and the stroke-specific literature in which the importance of clear information, communication and explanation about care are emphasised. In both study areas a lot of effort was put into ensuring that patients understood the care pathway from initial admission to a HASU/CSC/PSC and knew about the possibility of moving to a more local SU for ongoing care.

Repatriation also involved the transition of care from a HASU/CSC/PSC to a local SU. The NICE quality standards186 suggest that care should be well co-ordinated between different health-care professionals. The experience of patients in this study was that care was handed over smoothly, and nobody perceived that the transfer had any adverse effect on the trajectory of their recovery, similar to the findings of Moynihan et al. 177 The most difficult transition for patients was discharge to care in the community, for example patients’ reports of not being clear about follow-up care. Clarity about addressing ongoing care needs is one of the NICE patient quality standards. 186 Although in general people being discharged from a specialist stroke ward are more likely to have adequate follow-up care arranged than those from a general ward,174 patients in this study, who were all discharged from a specialist ward either at a HASU/CSC/PSC or local SU/DSC, experienced some difficulties. This may reflect the focus of the stroke care pathway centralisations on hyperacute care, and known variations in early supported discharge (ESD) and community therapy services across GM and London, as well as the difficulties experienced by hospital staff in trying to arrange care for patients to be discharged to different geographical areas.

Implications

Patients and carers on the centralised acute stroke care pathways in GM and London reported many positive aspects of care and it is evident that they often experienced standards of care in line with the NICE quality standards for patient experience. 186 Taking a broader perspective, the findings from this study suggest that the centralisation of care pathways in general can offer patients and carers good experiences of care. The disadvantages of travelling further were perceived to be outweighed by the opportunity to receive the best-quality care. This chapter has demonstrated the need for the delivery of clear and understandable information to patients and their families about every stage of the care pathway, to maximise their experience of care, and highlighted the importance of wide stakeholder engagement in the provision of centralised services (e.g. the importance of the role of the ambulance service in explaining the centralised pathway on initial contact with stroke patients).

What was already known about this subject?

• Several factors, including approaches to leadership, stakeholder involvement, use of measurement and data, and learning from previous experiences of change, play a significant part in the planning and implementation of MSC in acute stroke services (see Chapters 6 and 7). However, these lessons were generated in relation to changes conducted in urban areas and there may be value in studying similar changes conducted in more rural contexts.

• Stroke care across the Midlands and East of England was reviewed as part of a programme led by the SHA (a source of system-wide leadership) over the period 2012–13. This led to recommendations of MSC in acute stroke services across the nine network regions covered by the SHA, representing an opportunity to study MSC in less urban contexts.

• However, following these recommendations, no MSC to acute services was implemented.

What this chapter adds

• Several factors known to be associated with the successful implementation of MSC were either absent or severely hampered; this was perceived to have influenced progress of this programme.

• Although system-wide leadership of the programme was evident in the beginning, support from senior leaders in the SHA reduced over time and following the NHS reforms implemented in April 2013, the complete loss of this top-down leadership was reported to have made it easier for local commissioners to withdraw than if the SHA had remained.

• Data were used to build the case for reconfigurations and proposals for MSC. However, local stakeholders did not feel sufficiently engaged in the process, resulting in limited local ownership of recommendations, prompting local areas to repeat similar modelling exercises.

• The programme sought to engage local networks and stakeholders throughout the review process, but the impending reforms of NHS commissioning made this challenging.

• The programme used lessons from previous MSC to make their case. However, local stakeholders did not engage with these lessons because they related to work conducted in what were perceived to be very different contexts (i.e. in terms of rurality).

• Underlying these issues, the NHS reforms implemented in 2013 had a significant influence on the progress of this programme. Key examples of this included (1) disrupting system commissioning and governance across the English NHS, (2) introducing significant distraction throughout the system and (3) limited time to develop reconfiguration proposals because recommendations had to be delivered before the SHA was abolished.

Obstacles to implementation of major system change

As discussed in Chapters 1, 6 and 7, MSC may be seen as particularly complex, interacting across multiple levels – the macro (policy), meso (organisation) and micro (team/individual) – and multiple boundaries (sectoral, organisational, professional). 5,6,110,201 Given its scale and complexity, there are many factors that might influence the way in which MSC is developed and implemented, and indeed whether or not it is implemented.

Evidence suggests that a purely rational approach, based on resource use or potential outcomes, is unlikely to be sufficient to build an effective case for MSC. 6,202 Instead, active engagement of a range of stakeholders, using system-wide leadership to align these stakeholders, must be sustained throughout and beyond the change process. 6,110,202,203

Integral to these complex and dynamic relationships is the context in which the change is to be implemented. Context, for example in terms of national policy or local organisational, financial, political or resource factors, plays a significant part in how MSC develops and is implemented. 6,110,203 The process of implementation is likely to reflect an ongoing interplay between these intervention, implementation and contextual factors. 6,201 Therefore, there are many ways in which MSC might be adapted,110 delayed204 or halted over the course of planning and implementation.

The NHS Midlands and East review of stroke care

The review of stroke services across the Midlands and East of England (serving a population of 16 million people), conducted over the period 2012–13, recommended MSC of acute stroke services across the participating regions. Despite a year of detailed planning and engagement to agree the recommendations, and significant local expenditure of time and effort, MSC was not implemented in any of these areas.

Elsewhere in this report we have analysed examples of MSC in acute stroke services that were implemented and that illustrated the influence of these complex dynamics (see Chapters 6 and 7). The focus of this chapter was a case in which MSC was planned but not implemented. In this chapter, we have addressed the following RQ: which factors influenced progress of implementation of MSC in acute stroke services across the Midlands and East of England? Based on our findings, which we organised with reference to lessons for ‘successful’ implementation of MSC (in general,5 and adapted in relation to acute stroke services110), we aimed to illustrate how plans for MSC were developed in a context of significant contextual disruption (i.e. the reforms to the NHS in England implemented in 2013).

Design

This was a case study analysis, focusing on (1) the overarching service review and (2) local programmes to implement the recommendations resulting from the review (including implementation of MSC); the four local case studies were categorised in terms of their degree of rurality, with one area classified as ‘urban’ [a large conurbation (Area A)], two as ‘rural’ [predominantly rural areas (Areas C and D)] and one as ‘mixed’ [i.e. encompassing both urban centres and rural areas (Area B)] (Table 18).

Data

For this analysis, we conducted 33 interviews and 12 non-participation observations (covering approximately 30 hours) and collected > 200 documents over the period May 2013–December 2014 (see Table 2). We interviewed stakeholders involved in the service review (including members of the Programme Board and Advisory Group); we also interviewed representatives of the teams that led local implementation, including representatives of the project teams, the local network, hospital, ambulance and commissioning organisations, and patient representatives. We observed meetings related to planning the changes in Area A, including the Programme Board and the Clinical Advisory Group. We analysed documents relating to the review and local implementation programmes, in each case covering project plans, meeting minutes, supporting documents and modelling reports.

Analysis

This was a single case study analysis guided by the adapted rules for implementing MSC in acute stroke services (see Chapter 7). 110 Throughout, we considered factors that the literature suggests might work against implementation, with a particular focus on the influence of context at macro and meso levels.

The NHS Midlands and East review

Progress of local implementation

Table 18 summarises how implementation of MSC progressed in the four areas we studied.

It is noteworthy that in two areas that aimed to implement MSC (Areas A and B) the review recommendations were viewed as a starting point for planning local changes to services, rather than a model to be implemented. In Areas A and B, local project boards were set up to develop local implementation plans, drawing together a range of stakeholders. Much work was conducted to model local capacity, transport and finances, building towards new recommendations for MSC. These local programmes did not progress from planning to public consultation owing to a range of factors, including CCG concerns about changes in activity (whether in terms of losing services or in meeting increased demands). As a result, CCGs were unable to sustain a shared direction, and decisions to implement MSC were put on hold.

In Area C, the recommendations for MSC were seen as not appropriate until significant improvements had been made in the service felt most appropriate to be a HASU.

In Area D, during the three waves of proposals there was debate between the local network and the Programme Board over the need for further centralisation (services had already been reorganised in 2009). Following a site visit by members of the EEAG, it was agreed that no further reorganisation of acute services should take place, which permitted greater prioritisation of other parts of the stroke pathway, including development of ESD services:

We’ve not changed our reconfiguration, we fought quite hard not to, and we were very much, ‘We’re not, we’ve done that piece of work’. We gave quite strong justifications for why we didn’t want to change our reconfiguration, but it has pushed in terms of looking at quality again and giving a big re-impetus to stroke.

MED02, network representative, Area D

In the sections that follow, we present our findings to reflect the adapted lessons for MSC in acute stroke services, covering leadership, use of feedback loops, stakeholder engagement, and learning from history. 110 We summarise these findings in Table 19.

Leadership

As noted in Chapter 6, combining top-down (system-wide) and bottom-up leadership can help to ensure that plans for change are agreed and maintained through to implementation. However, system-wide authority may be especially important in sustaining alignment of stakeholders and managing resistance to proposals for change.

The Programme Board, in combination with the EEAG, represented a wide range of experts, many with national and regional remits (including the National Clinical Director for Stroke), and senior representatives of the SHA. The approach to developing recommendations was consensual, based on a dialogue of local proposals and central responses. This approach was questioned by some members of the project team, who suggested a stronger steer may have been valued locally:

I think people struggled, they almost wanted direction, they almost wanted to say, no, there was only going to be one hyperacute centre in this geography.

ME03, project team

Although senior leaders of the SHA launched the programme, there was a view that SHA ownership of the programme reduced over the course of the review, as a result of its impending abolition and the departure of some senior staff:

Public health was changing, the network was changing, the SHA was changing, and so that change [of SHA leadership] . . . was a real problem to the project. I felt that at the end the SHA were thinking, ‘Oh goodness, there was that stroke review by that guy: I suppose we ought to finish that off, really’. So I don’t think they had any ownership of what they were doing.

ME08, Programme Board

The SHA’s reduced ownership of the programme over time may have limited the influence of the final recommendations at the time of delivery. However, shortly after the recommendations were delivered the SHA was abolished, resulting in a more pronounced loss of system leadership.

From April 2013, in the absence of the SHA, and the associated loss of system leadership, local implementation faced a number of obstacles. Change was to be led by the newly formed CCGs, which were still establishing their role. There was significantly reduced support for work of this kind; for example, although some stroke network representatives were retained as part of the new NHS England Strategic Clinical Networks (SCNs), they no longer had an operational support role and their capacity to engage and facilitate leadership of change had been much reduced. As discussed above, it is likely that changes in leadership and loss of capacity to support change limited sustained local ownership of proposals; this in turn is likely to have contributed to local implementation either not commencing or halting when challenging decisions required a united response from leaders.

Feedback loops

Measuring system performance on various levels (including care provision and outcomes, service capacity, travel times and finance) and communicating it effectively to stakeholders can be an important facilitator of MSC, including agreeing to change, agreeing the focus and priorities of change, assessing progress and sustainability of changes.

A number of data were used in support of the review process. This included available evidence, for example citing local service performance on the national stroke audit. The programme also generated evidence that would help with developing viable proposals by conducting a series of modelling exercises to analyse current activity and capacity in services, and travel times between services. In support of this work, a subgroup (working with a commissioned management consultancy) developed a number of processes, including a tool for local systems to model activity levels.

However, despite substantial work, some of the modelling results were not seen as convincing locally. Local staff suggested that greater involvement in modelling would have permitted more meaningful discussion of local variations in performance of ambulance services and transport infrastructure:

It was only when they really got the ambulance service modelling, the remodelling our region where anybody in the region would . . . could put their name to understanding the geography, the travel times and the potential consequences of reconfiguration. I didn’t see a single piece of [Review Programme] modelling that looked like it was realistic or sensible . . . they’re paying this company a lot of money to do this and we’re not getting information on which we could base decisions.

MEC04, stroke physician, Area C

However, the team leading the modelling process suggested that this was in part a result of how localities engaged with it:

I thought the networks would lead this and really help it, all they did was they cut and pasted it like a jigsaw, they asked the different providers to fill it in, and they didn’t even review what had been filled in. So, when we had one where their costs were double their income, and we said, ‘Are you sure this is right?’ they were like, ‘Oh, we didn’t even look at the piece of paper, we just cut and pasted the thing together and sent it off to you’.

ME03, project team

These views suggest that there were underlying difficulties in engaging with local stakeholders in relation to the purpose and the process of the review. The outcome of gaps in engaging stakeholders is likely to have contributed to local decisions to repeat modelling exercises and, furthermore, decreased local ownership of the review recommendations.

Stakeholder engagement

Engaging relevant stakeholders (including service providers, commissioners, stroke patients and carers, members of the public and local politicians) helps to ensure the development of viable proposals that are understood and supported by these key groups.

In the review, the Programme Board included a wide range of experts in stroke care, and the process had several points at which localities had the opportunity to feed into the process. However, there was a view that the review could have been strengthened by engaging more or differently with certain stakeholder groups. For example, some CCG representatives indicated that the stroke service specification was too ‘input-focused’, that is, prioritising what services should look like rather than the outcomes they achieved, and that greater involvement of commissioners in its development could have prevented this:

These are people you know with huge expertise, massive national reputations, you know brilliant guys, but they’ve never commissioned . . . When you’re making service change and service planning decisions like this, you know you need the right people in the room.

MEC08, commissioner, Area C

This may in turn have led to the programme focusing too strongly on MSC as the solution:

Rather than focusing only on ‘What can we achieve in terms of a better outcome?’, the conversation already got right into how can you change the configuration.

ME08, Programme Board member

However, other areas saw significant engagement from CCGs in the process; for example, shadow CCGs from Area B presented the third and final set of proposals for their area to the EEAG towards the end of the review process.

One important issue raised in relation to the overall engagement dynamic was that building relationships and negotiating how stakeholders participate in them can take substantial investment of time and effort. The impending NHS reforms meant that there existed a challenging ‘hard deadline’ of March 2012, by which time recommendations had to be delivered. Furthermore, the fact that the reforms would involve closure of many of the participating organisations meant that many people were distracted, whereas others did not see this period as appropriate for planning a change of this kind and indeed felt little ownership of its objectives:

The big headline for me is, don’t try and do a review on the whole of the NHS – and it was the biggest restructuring I’ve ever encountered – don’t do it in that year. Don’t restrict yourself to 12 months: it’s better to leave it slightly longer – and do it in more detail with more ownership.

ME08, Programme Board member

Learning from history

Applying lessons from previous change can be an important factor in how change develops and is implemented, for example in terms of addressing established contextual barriers in a given locality.

The Midlands and East of England review was driven and informed by lessons from previous implementation of MSC in acute stroke services. Key examples of this included the goal of achieving improvements in outcomes previously achieved in London and the approach to engagement (e.g. working with local networks, shadow CCGs and ambulance services). However, as discussed above, the NHS reforms meant that leaders were aware that the task of engaging key stakeholders was inherently more complex, especially in terms of ensuring sustained ownership of proposals beyond April 2013. Therefore, the lessons in relation to engagement may have required greater adaptation to this challenging context.

Furthermore, lessons from past changes were used as a means of engaging stakeholders across the Midlands and East of England in developing the proposals for change in these areas. Documents and summaries of how change was carried out in London and GM were used throughout developing the service specification, and examples of engagement strategies and care protocols were shared with local networks and on the programme website (which is no longer online). However, local stakeholders noted that many of the key lessons were drawn from experiences of implementing change in highly urban areas, such as London; furthermore, many people referred to the changed financial context in which changes were to be carried out:

I think that was one of the constant bits of feedback about you know well it’s all very well to say use the London model but London had however many millions of pounds to implement this with the additional nurses and now London is seeing the costs savings and probably cut back on some of the things but they had that initial amount of money to be able to do it. And yeah, I think that’s . . . that was really hard for those organisations.

ME01, Programme Board member

As a result, evidence from previous changes achieved limited purchase with local stakeholders. Indeed, it may have compounded for some a view that leaders of the review did not appreciate the context in which the proposed changes were to be implemented, which may in turn have contributed to the difficulties in engaging stakeholders, described above.

Principal findings

The review of stroke services in the Midlands and East of England between April 2012 and March 2013 entailed substantial activity, including stakeholder engagement, specification development and service/system modelling. However, although local systems did work to improve their services in line with some aspects of the service specification, the MSCs recommended were not implemented in any of the areas we studied.

System leadership – how it was used during the review and its absence during local implementation – had a significant role throughout. During the review, some questioned whether or not there was a need for a more top-down approach, noting that support from senior representatives of the SHA may have been important; during implementation the loss of system-wide leadership made it harder for local commissioners to sustain a united position when challenging decisions relating to service reorganisation. Programme leaders used national audit and locally collected data to build the case for change and develop proposals for how services might best be organised. However, there were challenges in engaging local stakeholders in the process, and there was limited local ownership of the findings and associated recommendations, prompting local areas to repeat similar exercises before considering implementing and changes. The programme sought to engage local networks and stakeholders throughout the review process, but because of the impending reorganisation of NHS commissioning it proved difficult to involve shadow CCGs sufficiently in order to gain their input on and ownership of the programme recommendations. Finally, although the programme was guided in its approach by lessons from previous changes, and indeed made use of these lessons as a means of strengthening the case for change, the fact that these lessons related to changes conducted in significantly different political and financial contexts may have reduced local stakeholders’ engagement with the process.

Underlying these issues, the NHS reforms implemented in 2013 had a significant influence on the progress of this programme. Key examples of this include (1) disrupting system commissioning and governance across the English NHS (thus removing established sources of system leadership through SHAs and operational support for MSC through clinical networks, (2) introducing significant distraction throughout the system (reducing the extent to which stakeholders could engage) and (3) setting a ‘hard deadline’ to deliver recommendations (which limited time to conduct the review process and engage these distracted stakeholders).

Strengths and weaknesses

A strength of this analysis was that we had the opportunity to study contemporaneously both the review and local efforts to implement its recommendations. This gave us access to data on implementation as it took place, rather than relying on retrospective accounts of the programme.

A potential weakness of this analysis is that it focused on only four of the nine areas covered by the Midlands and East review. Other factors may have played a part in changes not being influenced elsewhere in the Midlands and East of England. However, the area studied still represented a substantial proportion of the changes (covering a population of approximately 4.8 million people); furthermore, the study team engaged regularly with local network representatives to establish progress across the whole of the Midlands and East of England, these discussions confirmed that there was similarly slow progress across these areas, and suggested that many of the same factors had a significant role.

Comparison with other studies

This chapter has demonstrated that factors identified as important to planning and implementing MSC successfully [described by Best et al. 5 and modified by our evaluation (see Chapter 6)110] can also be used to explain how and why MSCs are not implemented. Our findings on the impact of both the 2013 NHS reforms and the wider financial issues faced by the NHS at the time of this analysis reflected past evidence on how financial and political context can influence efforts to implement changes to how care is organised and provided. 6,110,202,206 The case of Area D, where it was agreed over the course of the review process negotiations that no MSC would be recommended, may be seen as an example of data being used to guide decision-making in relation to whether or not and how services should develop.

Implications

The absence of system-wide leadership made it difficult to sustain stakeholder unity over the course of MSC. Effective communication of system and service data can support and guide change, but only if the relevant stakeholders have ownership of the evidence. Stakeholder engagement requires an understanding of which groups are likely to influence both implementation and sustainability of the intended changes, and ensuring that these stakeholders have sufficient time and capacity to participate effectively in planning and implementation processes. Finally, contextual factors will influence change throughout and beyond the implementation phases. This suggests that planners should attend not just to current local and national issues (including political and financial issues), but also consider the likely impact of factors that are still on the horizon.

What was already known about this subject?

• The initial reconfiguration of acute stroke services in GMA covering the period up to March 2012 was associated with no impact on mortality but LOS fell, whereas in London there was a significant reduction in both mortality and LOS (see Chapter 3).

• Post-reconfiguration patients treated at HASUs in GM and London were significantly more likely to receive evidence-based interventions than those treated at non-hyperacute units in these areas and in the RoE. However, a far higher proportion of patients were treated in a HASU in London than in a CSC/PSC in GM. This might explain the difference in mortality outcomes between the two areas (see Chapter 4).

What this chapter adds

• Following implementation of MSC in GMB, there was a reduction in mortality in GM of approximately 189 deaths per year.

• However, there were also reductions in mortality in the RoE, and most of the differences in mortality between GMB and the RoE were not statistically significant.

• In GMB there was a reduction in mortality over and above that seen in the RoE for people treated in CSC/PSCs (there were approximately 69 fewer deaths per year).

• The proportion of patients treated in CSC/PSCs in GMB increased to 86% in 2015/16; this was a higher proportion than in GMA (39%), but still lower than in London (93%).

• LOS fell in GMB, resulting in around 6750 fewer bed-days per year.

• There was no significant variation in mortality or LOS over time since the reconfiguration in London, indicating that the reductions in mortality and LOS following centralisation in London were sustained.

• These patterns were reflected by the analyses of clinical interventions in both areas. Following the implementation of MSC in GMB, the delivery of clinical interventions improved significantly, while also improving (although generally not to the same degree) in the RoE; in London the delivery of clinical interventions either improved or was sustained.

Analyses of clinical outcomes

Analyses of clinical interventions

Analyses of clinical outcomes

In the analyses for GM the findings are summarised as follows. First, across all stroke types combined, mortality at 3, 30 and 90 days declined over time in GM, although the decline was not smooth (see Appendix 7, Figure 29). Note that only 1 year’s worth of data were available for the GMB period.

Across all stroke types combined, comparing GMB with ‘Before’ plus GMA, adjusted mortality at 90 days in GM fell by 4.2 percentage points (see Appendix 7, Table 49). If there are an estimated 4500 strokes in GM each year, then this represents 189 fewer deaths per year (i.e. 4500 × –4.2/100). Running the same adjusted before-and-after analyses for the RoE gives a reduction of –2.6% (see Appendix 7, Table 49). The unadjusted between-region differences are broadly in line with these findings, showing an additional 1.1-percentage-point reduction in 90-day mortality in GMB compared with ‘Before’ plus GMA over and above the reduction seen in the RoE over the same time period (see Appendix 7, Table 51).

The adjusted between-region difference-in-differences across all types of stroke combined were negative and significant for mortality at 3 days, non-significant for mortality at 30 days and borderline significant (95% CI –2.7 to 0.01; p = 0.05) for mortality at 90 days (Table 21). The point estimates broadly reflect the results of the above analyses. When we reran the above analysis for all stroke types only for patients treated at the GM CSC/PSCs, there was a 1.8-percentage-point reduction in 90-day mortality (95% CI –3.4 to –0.2; p = 0.03; see Table 21) – the coefficient was larger (more negative) than the all-hospital model and significantly different from zero. Assuming that 86% of the 4500 strokes in GMB went to CSC/PSC (see Analyses of clinical interventions), then if this reduction was applied to the estimated 4500 strokes in GM each year this indicates 69 fewer deaths per year (i.e. 4500 × 0.86 × –1.8/100).

In the analysis of mortality in London there was little variation over time since the reconfiguration in terms of mortality at 3 days (see Appendix 7, Figure 32), 30 days (see Appendix 7, Figure 33) and 90 days (Figure 14). For example, in the analysis of mortality at 90 days, the p-value under the null hypothesis that the regression coefficient for every month after the reconfiguration is the same as the regression coefficient for July 2010 (when the reconfiguration occurred) was 0.09. This indicates that the reduction in mortality following centralisation has been sustained.

FIGURE 14.

Adjusted trends in mortality at 90 days in London. The dashed vertical line indicates when the reconfiguration in London occurred (July 2010). Adapted from Morris et al. 207 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. Figure title and numbering updated for report.

Adapted from Morris et al. 207 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. Figure title and numbering updated for report.

In terms of LOS, mean unadjusted LOS was 14.4 days in GMB and 17.4 days in GMA. Figures for the same time periods in the RoE were 15.7 days and 18.1 days, respectively (see Appendix 7, Table 54). Across all stroke types there was a significantly larger decline in risk-adjusted LOS compared with the RoE, by –1.5 days (95% CI –2.5 to –0.4; p < 0.01; Table 22). If this reduction was applied to the estimated 4500 patients with a stroke in GM each year this would result in 6750 fewer bed-days a year.

In London, LOS since the reconfiguration declined (Figure 15). The p-value under the null hypothesis that the regression coefficient for every month after the reconfiguration is the same as the regression coefficient for July 2010 was < 0.01; this appears to be due to a further decline in LOS since 2015.

FIGURE 15.

Adjusted trends in LOS in London. The dashed vertical line indicates when the reconfiguration in London occurred (July 2010). The p-value, under the null hypothesis that the regression coefficient for every month after the reconfiguration (which occurred in July 2010), is the same as the regression coefficient for July 2010, at a p-value of < 0.01. Adapted from Morris et al. 207 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. Figure title and numbering updated for report.

Adapted from Morris et al. 207 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. Figure title and numbering updated for report.

Analyses of clinical interventions

Clinical interventions

In GMB, achievement of evidence-based clinical indicators in year 3 either stayed the same or improved compared with years 1 and 2 (Table 23; see Appendix 7, Figure 34). In year 3 there were improvements of > 10 percentage points for several ‘front door’ processes (brain scan within 60 minutes, arrival in SU within 4 hours), and assessments over the first 72 hours (consultant assessment within 14 hours, physiotherapist assessment, occupational therapist assessment and speech and language therapist within 24 hours). Improvements and levels of achievement in the RoE comparator during the same time period were either similar or lower than those seen in GM. In London the ‘front door’ processes, use of thrombolysis and assessments during the first 72 hours all stayed high and constant over the 3 years, with no more than a ± 5 percentage point fluctuation each year (Figure 16; see Appendix 7, Table 61).

TABLE 23 - Risk-adjusted proportions of patients receiving clinical interventions in GM by year

SaLT, Speech and Language Therapist; tPA, tissue plasminogen activator.

All measures reflect time from ‘clock start’ (i.e. when patient first arrives at hospital or when symptoms are identified in in-patients). ‘Ramsay et al. (2015)’ refers to the post-centralisation (covering the period 2010–12) sample’s risk-adjusted likelihood of patients receiving interventions reported in Ramsay et al. 26 The comparator includes RoE, with all regressions adjusted for presence of London data.

Adapted from Morris et al. 207 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. Table title and numbering updated for report.

Intervention	Risk-adjusted proportions (95% CI)
	GM				Comparator
	Ramsay et al. (2015)26	2013/14	2014/15	2015/16	2013/14	2014/15	2015/16
Scan 60 minutes		40.7 (39.1 to 42.2)	49.2 (47.6 to 50.8)	60.8 (59.4 to 62.2)	43.0 (42.6 to 43.4)	44.8 (44.4 to 45.2)	47.6 (47.3 to 48.0)
Scan 180 minutes	65.2 (64.3 to 66.2)	74.1 (72.7 to 75.4)	81.7 (80.5 to 82.9)	86.4 (85.4 to 87.3)	70.9 (70.6 to 71.3)	74.0 (73.7 to 74.3)	77.0 (76.7 to 77.3)
Scan 24 hours	94.0 (93.5 to 94.4)	95.1 (94.4 to 95.8)	97.1 (96.5 to 97.6)	97.4 (96.9 to 97.9)	94.9 (94.8 to 95.1)	96.0 (95.8 to 96.1)	96.7 (96.6 to 96.8)
tPA to eligible patients		67.9 (63.6 to 72.1)	82.3 (78.1 to 86.5)	88.6 (85.5 to 91.7)	80.6 (79.8 to 81.3)	89.1 (88.5 to 89.7)	91.8 (91.3 to 92.4)
tPA within 60 minutes		74.4 (69.6 to 79.2)	70.5 (65.1 to 75.8)	74.6 (70.2 to 78.9)	53.1 (52.0 to 54.2)	57.1 (56.0 to 58.1)	59.7 (58.6 to 60.7)
Swallow screen in 4 hours		58.7 (57.1 to 60.3)	63.4 (61.8 to 65.0)	69.2 (67.8 to 70.6)	59.7 (59.3 to 60.1)	63.9 (63.5 to 64.2)	67.7 (67.3 to 68.0)
SU within 4 hours	55.9 (54.9 to 57.0)	57.4 (55.8 to 59.1)	58.8 (57.1 to 60.5)	79.1 (77.9 to 80.4)	51.3 (50.9 to 51.7)	52.8 (52.4 to 53.2)	53.4 (53.0 to 53.7)
Consultant assessment within 14 hours		57.6 (55.9 to 59.2)	58.9 (57.2 to 60.6)	79.2 (78.0 to 80.4)	50.6 (50.2 to 51.0)	52.1 (51.7 to 52.5)	52.6 (52.2 to 53.0)
Consultant assessment within 24 hours		78.2 (76.9 to 79.6)	80.2 (78.9 to 81.5)	93.0 (92.3 to 93.8)	81.1 (80.8 to 81.4)	82.8 (82.5 to 83.1)	84.3 (84.1 to 84.6)
Nurse assessment within 12 hours		86.4 (85.3 to 87.4)	86.6 (85.5 to 87.7)	91.2 (90.4 to 92.0)	86.5 (86.2 to 86.7)	88.1 (87.8 to 88.3)	89.3 (89.0 to 89.5)
Nurse assessment within 24 hours		93.4 (92.7 to 94.2)	92.8 (92.0 to 93.6)	94.6 (93.9 to 95.2)	93.2 (93.0 to 93.4)	93.9 (93.7 to 94.0)	94.5 (94.4 to 94.7)
Physiotherapist assessment within 24 hours		55.4 (53.7 to 57.0)	63.2 (61.6 to 64.8)	80.7 (79.5 to 81.9)	55.6 (55.2 to 56.0)	57.1 (56.7 to 57.4)	59.7 (59.3 to 60.1)
Physiotherapist assessment within 72 hours	92.1 (91.5 to 92.7)	95.1 (94.3 to 95.8)	97.0 (96.4 to 97.6)	97.5 (97.1 to 98.0)	93.6 (93.4 to 93.8)	93.4 (93.2 to 93.6)	93.7 (93.5 to 93.9)
Occupational therapist within 24 hours		50.5 (48.8 to 52.2)	61.2 (59.6 to 62.8)	79.4 (78.1 to 80.6)	45.2 (44.7 to 45.6)	47.5 (47.1 to 47.9)	52.3 (51.9 to 52.7)
Occupational therapist within 72 hours		94.4 (93.6 to 95.1)	96.4 (95.8 to 97.0)	96.9 (96.4 to 97.4)	87.3 (87.1 to 87.6)	88.5 (88.2 to 88.7)	89.7 (89.4 to 89.9)
SaLT swallow within 24 hours		53.2 (50.8 to 55.6)	56.9 (54.4 to 59.4)	55.7 (53.1 to 58.2)	48.4 (47.8 to 49.0)	49.9 (49.3 to 50.5)	54.0 (53.4 to 54.6)
SaLT swallow within 72 hours	94.0 (93.6 to 94.4)	85.5 (83.9 to 87.2)	91.3 (89.9 to 92.7)	91.9 (90.6 to 93.3)	78.8 (78.3 to 79.3)	82.2 (81.7 to 82.6)	84.9 (84.5 to 85.3)
SaLT communication within 24 hours		39.7 (37.3 to 42.2)	53.8 (51.5 to 56.2)	69.9 (68.0 to 71.7)	35.5 (34.9 to 36.1)	38.0 (37.5 to 38.6)	42.0 (41.5 to 42.6)
SaLT communication within 72 hours		86.7 (85.1 to 88.3)	93.4 (92.2 to 94.5)	96.6 (95.9 to 97.3)	78.3 (77.8 to 78.7)	81.6 (81.2 to 82.0)	85.1 (84.7 to 85.5)

FIGURE 16.

Risk-adjusted likelihood of patients receiving clinical interventions in London. Error bars, 95% CI; SaLT, Speech and Language Therapist; tPA, tissue plasminogen activator. All measures reflect time from ‘clock start’ (i.e. when patient first arrives at hospital or when symptoms are identified in in-patients). Scan 60, brain scan within 60 minutes of clock start; Scan 180, brain scan within 180 minutes of clock start; Scan 24 hours, brain scan within 24 hours of clock start. Adapted from Morris et al. 207 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. Figure title and numbering updated for report.

Adapted from Morris et al. 207 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. Figure title and numbering updated for report.

Principal findings

The principal findings of this analysis are as follows.

1. There was a decline in stroke mortality following the GMB reconfiguration, but reductions in mortality were also seen elsewhere in the RoE over the period; these national reductions appear to have been slightly smaller than the reductions in GMB, although most of the changes over time were not statistically significantly different between the two areas. We interpret this to mean that across all hospitals in GM there was not a significant impact on mortality over and above the changes seen in the RoE. There was, however, a reduction in all-cause mortality in any location (not just the hospital) seen in the GM CSC and PSCs during the GMB period that was statistically significantly different and larger than the reductions seen elsewhere in the RoE, and would amount to approximately 69 fewer deaths per year.

2. There was a significant reduction in LOS in GMB of 1.5 days per patient over and above the reduction seen in the RoE, which would amount to approximately 6750 fewer bed-days a year.

3. In GM a larger proportion of patients were treated in CSC/PSC following the GMB reconfiguration. Achievement of evidence-based clinical indicators either stayed the same or improved compared with the RoE comparator.

4. In the analysis for London there appears to be little variation in mortality, LOS and evidence-based clinical indicators over time since the reconfiguration in July 2010. This indicates that the improvements following the centralisation in London have been sustained.

Strengths and weaknesses

The strengths and weaknesses are similar to the previous analyses (see Chapters 3 and 4). The strengths related to the large national data sets that we used and the robust, quasi-experiment analytical framework we employed. Limitations included a lack of data on stroke severity in the HES database, a lack of data on QoL, a lack of data on pre-hospital processes, outcomes and limitations in measurement of LOS when patients may be repatriated between SU. Our analysis used only data for patients treated up to 1 year after the GMB reconfiguration, which may have limited the likelihood of demonstrating a significant impact of the further reconfiguration.

Implications

The increased number of patients treated in CSC/PSC in GM following further reconfiguration led to improvements in LOS and in clinical indicators on average across GM as a whole. There were reductions in mortality among patients treated at a CSC/PSC, but there was no overall reduction in mortality on all admitted stroke patients (i.e. including the proportion who were not treated at a CSC/PSC). The reduction in mortality seen in the GM CSC/PSCs resulted in 69 fewer deaths per year. This exceeded the ‘50 fewer deaths per year’ figure that we calculated based on the results of our previous analysis (see Chapter 3), if the 1.1-percentage-point reduction in 90-day mortality in London was achieved in GM. Although this reduction in mortality in the CSC/PSCs saved lives given that in London > 90% patients are treated in a HASU, this suggests that even more lives might be saved if a higher proportion of GM stroke patients were treated in CSC/PSC. We were able to analyse only data for patients treated up to 1 year after the GMB reconfiguration, and SSNAP data indicate that the proportion of patients treated in a CSC/PSC and receiving clinical interventions has continued to increase in GM. 213 It would, therefore, be beneficial for further research on and monitoring of the relationship between the proportion of patients treated at a CSC/PSC, delivery of clinical interventions, and reductions in mortality and LOS over a longer period. The impact of the London reconfiguration in July 2010 on mortality, LOS and evidence-based clinical indicators was sustained up until March 2016, which suggests that changes of this kind are sustainable.

What was already known about this subject?

• The reconfiguration of acute stroke services in GMA covering the period up to March 2012 was found to be cost-effective compared with changes that happened elsewhere in England. This was primarily as a result of cost-savings per patient attributable to reductions in LOS in GM compared with in the RoE (see Chapter 5).

• The GMB model was associated with significant improvements in care provision and LOS overall, and significant reductions in patient mortality for patients treated in CSC/PSCs (see Chapter 11).

What this chapter adds

• Over 90 days, GMB cost £932 less per stroke patient than GMA. This is compared with a cost saving of £635 in the RoE over the same time period.

• GMB also resulted in 4 additional QALYs per 1000 patients over 90 days compared with GMA. In the RoE there were 3 additional QALYs per 1000 patients over the same time period.

• At 90 days there was an 88% probability that GMB was cost-effective compared with changes that occurred in the RoE over the same time period at a WTP for a QALY of £20,000 to £30,000.

• At 10 years there was a 31% probability that GMB was cost-effective compared with changes that occurred in the RoE over the same time period at a WTP for a QALY of £20,000 (39% for a £30,000 WTP for a QALY). This was driven primarily by an apparent 3% increase in discharge to non-acute NHS inpatient providers, according to HES data. Further work is required to validate this and refine other values in the model, potentially with richer data sets such as SSNAP.

Data

We used the same data as the analysis of clinical outcomes and clinical interventions (see Chapter 11, Analyses of clinical outcomes and Analyses of clinical interventions for variables and data used).

Statistical analyses

The decision model described in Chapter 5 was updated with data for the reconfigurations in GM that took place in April 2010 (GMA) and April 2015 (GMB). Data for the same time period for the RoE (urban areas only and excluding London) were also used to update the model.

The cost of implementation was from a provider perspective only. An extra £2.4M in investment was provided to assist with the additional cost of the CSC/PSCs. This was given as an uplift to the tariff and included in the analysis as an additional cost of a CSC/PSC admission. All other costs remained the same as those reported in Table 11.

We report the NMB of the additional costs and QALYs of GMB compared with those of GMA, and compare that to the additional costs and QALYs in the RoE over the same time period. We report the results at 90 days and 10 years. Time to recurrent stroke and time to death were also updated in the 10-year model based on HES data. Costs and QALYs beyond 12 months were discounted at a rate of 3.5%.

We conducted sensitivity analyses to test the impact of different assumptions made in the model. The results for only those sensitivity analyses that had a significant impact on the results are reported in this chapter.

Analyses of clinical outcomes

The results of the 90-day and 10-year cost-effectiveness analysis are reported in Table 24. At 90 days GMB compared with GMA resulted in lower costs and more QALYs than in the RoE. There was an 88% chance, however, that the changes were cost-effective at a WTP for a QALY gain of £20,000 and £30,000. At 10 years there was a 31% probability that GMB is cost-effective compared with changes that happened in the RoE at a WTP of £20,000 per QALY (39% at a WTP of £30,000; Figure 17). Although GMB resulted in more QALYs than the RoE over the same time period (an additional 6 QALYs) this was also for a greater cost (£719,948), resulting in an incremental cost-effectiveness ratio of £130,667 per QALY gained.

FIGURE 17.

Cost-effectiveness acceptability curve of NMB of second reconfigurations in GMB compared with the RoE at 90 days and 10 years (discounted).

Sensitivity analysis

The location that patients are discharged to is a key driver of the 10-year cost-effectiveness of the model. In GMB compared with GMA and the RoE there was an increase from 6% (n = 966) to 9% (n = 319) in the percentage of patients discharged to a non-acute NHS inpatient provider, most commonly a community hospital. No other changes to discharge destination were observed, other than an equal reduction to the number of people discharged home. Following interrogation of the discharge destination, we identified two outlier hospitals and that the majority of the change occurred in the last quarter of 2015/16, potentially owing to incomplete data. As a result we conducted the following sensitivity tests for discharge destination and time to discharge:

1. removal of the two outlier hospitals for both GMA and GMB

2. removing the data for the last quarter of 2015/16

3. holding discharge destination constant from GMA to GMB.

The number and proportion of people discharged to each location in GMA and GMB used for each of the three sensitivity analyses and the base case (values used to generate the results in Analyses of clinical outcomes above) are reported in Table 25. The results of the sensitivity analyses are reported in Table 26. In all three instances, GMB is cost-effective at 90 days compared with the RoE, resulting in less cost for fewer deaths. In analysis 1 and 2, when outliers are removed, GMB is not cost-effective at 10 years compared with the RoE. In analysis 3 (holding discharge destination constant from GMA to GMB), GMB is cost-effective at 10 years at £30,000 WTP for a QALY gained.

Principal findings

At 90 days the reconfigurations in GMB dominated changes that happened in the RoE, with an 88% probability of being cost-effective at a £20,000 WTP for a QALY gained. At 10 years, both GMB and the RoE maintained their cost-savings and decrease in deaths, but there was only a 31% chance that GMB was cost-effective. This may be due to the short maximum of 1-year follow-up available for GMB rather than the changes in GMB not being maintained over longer follow-ups. The discharge destination recorded in HES also has implications for the results. Further work is required to validate the 3% increase in discharge to non-acute NHS inpatient providers given the impact it has on the results. If it is incorrect and is a coding error, GMB is potentially cost-effective at 10 years.

Strengths and weaknesses

The strengths and weaknesses were similar to those of the previous analysis (see Chapter 5). The strengths related to the large national data sets we used and the robust quasi-experiment analytical framework we employed. Limitations were similar to those of the clinical outcomes and interventions analysis (see Chapter 11). Additional limitations were the number of the assumptions made to the model, as discussed in Chapter 5.

The short follow-up duration following the second reconfiguration of stroke care in GM made it challenging to assess the long-term implications of the reconfigurations.

Implications

There was clear evidence that the second reconfigurations of stroke care in GM were cost-effective at 90 days. There was less evidence for this being maintained over 10 years, however. In Chapter 5, we reported that the first reconfiguration of stroke care in GM was cost-effective compared with changes that happened in the RoE over the same time period. This result was driven by the reduction in LOS only. The second reconfiguration of GM included an additional investment of £2.4M as an uplift to the stroke tariff and has resulted in additional reductions in LOS and reductions in mortality in CSC/PSCs (see Chapter 11). At 90 days the additional investment and reconfiguration translated to net cost-savings due to reduced LOS, although the cost-savings were not maintained at 10 years. GMB resulted in additional QALYs at 90 days and 10 years, although the additional health benefit of 6 QALYs per 1000 patients at 10 years was not sufficient to compensate the reduction in cost-savings compared with the RoE over the same time period.

The 3% increase in discharge to non-acute NHS inpatient providers in GMB is of concern, particularly given the implications for the findings. As a result this warrants additional investigation potentially using another data set such as SSNAP.

It is important to note that the cost-savings we modelled were realised by stroke care providers and not commissioners of stroke care services. Commissioners pay for stroke admissions as a tariff per admission, whereby additional costs per bed-day for LOS are incurred only after a trim point. In the case of stroke this was a LOS > 40 days. As a result, commissioners would realise cost-savings only as a result of the reconfigurations (1) if the reconfigurations prevented patients being readmitted for stroke or (2) if a lower percentage of patients were still in hospital after the trim point. We found no evidence that GMB prevented readmissions compared with GMA and changes that occurred in the RoE at the same time. This may be because the follow-up duration was not sufficient to detect a difference. Based on the model there was a reduction in the percentage of patients still in hospital at 41 days in GMB compared with GMA of 3.9% (2.2% of patients still in hospital at 41 days for GMB, compared with 6.1% in GMA). In the RoE over the same time period there was a reduction of 1.7% of patients in hospital at 41 days (2.1% of patients at 41 days after compared with 3.8% before). After the reconfigurations there was a reduction of 888 bed-days per 1000 patients after the trim point in GM compared with 338 in the RoE over the same time period. This translated to a reduction of 540 bed-days following the trim point for GMB compared with that for the RoE. Assuming a cost per bed-day of £200, this was a total cost saving to the commissioner of £108,000 per 1000 strokes for additional bed-days. Assuming 4500 strokes per year in GM, this was still £2M shy of the total investment made.

What was already known about this subject?

• Service centralisation in GM led to significant reduction in LOS, but not in mortality, compared with in the RoE. Therefore, further reconfiguration of services was planned and implemented.

• There is growing evidence about how approaches to leadership, use of feedback, stakeholder engagement and experience of previous changes influence implementation of MSC in acute stroke services (see Chapters 6, 7 and 10).

What this chapter adds

• The GM acute stroke system – like the rest of the English NHS – faced significant obstacles over the period studied and these affected the time it took to agree a new model, plan it and implement it. Despite these obstacles, change was implemented.

• Obstacles included turbulence caused by reforms to the NHS and national staffing shortages, which contributed to significant pressures on planning. Post implementation, these factors led to delays in the movement of patients through the system, in finding beds for stroke patients and in discharging stroke patients.

• Issues relating to leadership and governance, and the use of service and process reviews, including the earlier findings from this study (see Chapter 3) were reported as important in enabling implementation.

• Leaders of the GMB reconfiguration ‘held the line’ on approaches to implementation (e.g. the timing and degree of phasing of change). A key system enabler post implementation was identified as the Operational Delivery Network (ODN) governance model, funded by the providers, covering the whole stroke pathway. This enabled regular audits and a mechanism to facilitate the system-wide discussions needed to maintain effective system operation. Sources of pressure post implementation included not only staffing but also changes in the nature of workload at CSC, PSCs and DSCs caused by the reconfiguration.

Design

This chapter presents a single case study, based in GM, as described in Chapter 1. Comparison is made with GMA, reported in Chapters 6 and 7, and is discussed in relation to the quantitative data analysis of GMB (see Chapter 11) and sustainability in London (see Chapter 14).

Data

As detailed in Chapter 2, we conducted interviews with 78 stakeholders (governance and service level) and 59 non-participant observations of relevant meetings and events (approximately 120 hours). We also collected > 100 documents (meeting minutes and papers, key local documents, etc.). Data were collected between December 2013 and March 2017 (see Table 2 and Chapter 2, Understanding development, implementation and sustainability), enabling analysis of the 2-year post-implementation qualitative data.

Analysis

We identified those factors that influenced the decision to change, the planning and the implementation approach of GMB, and GMB post implementation. We combined analyses of interviews, meeting notes and documents in order to develop an understanding of MSC in relation to previous evidence (see Chapter 6).

To agreement

Activities that were undertaken during the ‘to agreement’ phase are outlined, as well as aspects of governance, which influenced the timescale.

Planning

Once agreement was reached in September 2013, formal planning for implementation began, although informal activities had already started. Issues of leadership, governance and feedback were highlighted as influencing process and timescale, with some other specific considerations during planning.

Implementation: a ‘big bang’ launch

Implementation of GMB took place on 30 March 2015, as a ‘big bang’ rather than a phased launch, as had been the case with GMA (see Chapter 7). Leaders held the line on this approach when phased implementation was proposed, because of the difficulties in phasing presented to the ambulance service and concern about capacity issues if a CSC and one PSC went ‘live’ while the second PSC was not fully operational (‘learning from history’).

Post implementation

Findings are presented here using the same framework for analysis as in Chapter 14, enabling us to compare the post-implementation periods in London and GM.

Principal findings

The time taken from the original recommendation to further reconfigure stroke services in GM and to obtain formal agreement to do so and implementation of the new model was considerable. However, this was a period of great turbulence in the NHS, and that such system-wide change was achieved in this context may be significant in itself.

Changes in NHS structures and leadership contributed to delays in planning and implementation, with other local changes in context having an impact on planning.

The importance of having a governance structure to oversee and co-ordinate the planning of changes to services was demonstrated. Delays to the planned implementation date meant that the sites were ready on the launch date and there were few teething issues following implementation. This may have led to the achievement of improved outcomes within a year (see Chapter 11).

Challenges identified in GMA, such as dealing with stroke mimics, the repatriation of patients from CSC/PSCs to DSCs and discharging people to the community, continued post implementation. The importance of working collaboratively to find local solutions to these problems was shown, enabled by the governance model, which provided support for regular audits and a mechanism to facilitate the system-wide discussions needed to maintain effective system operation, as well as ensuring that SSNAP data were entered consistently across GM, so that it could be used for reviews. It is likely that the co-ordinating role of the ODN contributed to the improvements in the provision of evidence-based care and the reduction of mortality among those patients treated in a CSC/PSC.

Strengths and weaknesses

Strengths of this analysis included the contemporaneous study of both planning and implementation. A wide range of interviewees participated, including those involved in the planning, governance and review of stroke services and staff from the CSC, both PSCs and two DSCs. Staff from across GM participated in observed meetings, so a wide range of views and experiences were captured. Emerging findings were shared with local stakeholders in order to enhance validity.

Comparison with other studies

The importance of the context215,216 within which changes to stroke services were planned and implemented was evident in this analysis. However, despite this turbulent background, MSC was achieved, with the data showing how the context delayed change, had to be taken into account and was worked around.

As in the Midlands and East of England (see Chapter 10) and London (see Chapter 14), the 2013 NHS reforms disrupted system commissioning and governance. Changes in leadership, both designated and distributed, meant that it was difficult to get decisions made, and there was a consequent loss of knowledge and ownership of the plans for reconfiguration. This suggests that continuity of leadership could be added to designated and distributed leadership as necessary for large-scale transformation,5 given that it will always take place over a period of time.

Feedback was used throughout the ‘to agreement’, ‘planning’ and ‘implementation’ phases. Following the review of GMA, other reviews were carried out, which fed in to the planning process and justified decisions made. Agreement for reconfiguration was based mainly on an analysis of SSNAP data, which continued to be important after implementation, to monitor activity on the stroke pathway. Resource was dedicated to ensuring that staff appreciated the importance of accurate and complete data collection and understood the analysis of the data in relation to their own practice.

Financial penalties were not used in GMA (see Chapters 6 and 7)110,123 but were introduced in relation to delayed repatriation in GMB. However, there were changes to the mechanism used, with plans for ‘top slicing’ of the tariff paid to a DSC if a stroke patient was not repatriated in a timely manner, after a previous method of doing this was not imposed.

The use by leaders of discursive power to promote change,203 with an emphasis on evidence and better patient outcomes, was evident in the planning of GMB. Leaders used SSNAP data and then the publication of research findings in order to keep to the fore the reasons that reconfiguration was necessary.

Attending to history, that is, learning from previous transformations, was evident. A phased approach to implementation was not undertaken, largely because of previous experience in GM.

Despite system-wide disruption resulting from NHS reforms (reducing the extent to which stakeholders could engage) there is evidence that in addition to engaging a range of stakeholders,5,110 ‘ownership’ of the changes was engendered in stakeholders. For example, CCGs funded a project manager, stroke staff championed the changes through training other staff in their locality and provider Trusts funded the ODN beyond its initial year post implementation. ‘Ownership’ can be conceptualised as a step further than engagement, and ultimately as necessary for sustainability.

Implications

• Major structural reorganisations have been shown to represent a significant obstacle to implementing MSC. However, GMB indicates that it is possible to mitigate or overcome such disruption.

• Network support organisations (such as the ODN) can facilitate collaboration, service audit and support the development of standards and protocols for organisation and provision of care. Such organisations might maximise their impact if they support the whole patient pathway rather than just the acute phase.

• MSC benefits from programme management support, within a framework of robust governance, which engages all key stakeholders.

• Workload profile and the related skill requirements are important during and after change and may be underestimated. Insufficient attention may lead to skills shortages as well as low staff morale.

• Continuity of leadership is important: steps should be taken to minimise the effect of contextual changes by the involvement of leaders with system-wide authority and collaborations of organisations where possible.

• Consideration must be given to harmonisation of terms and conditions for staff across the system. Working in a 24/7 service, for example, poses additional requirements for night and weekend working that may not automatically be rewarded equitably and can lead to staff shortages.

What was already known about this subject?

• Ongoing sustainability is a key measure by which the ‘success’ of organisational change might be assessed. Sustainability may be influenced by characteristics of the intervention itself, the context in which it was implemented, leadership and processes to support implementation and sustainability.

• Changes implemented in London led to a significantly higher likelihood of providing evidence-based care (see Chapters 4 and 7), and significantly greater reductions in mortality than elsewhere (see Chapter 3); analysis of data to March 2016 indicates that care provision and outcomes either improved or were sustained (see Chapter 11).

What this chapter adds

• The London acute stroke system – like the rest of the English NHS – faced significant obstacles over the period studied.

• Factors identified as obstacles to sustaining organisation and provision of acute stroke care in London included turbulence resulting from the 2013 NHS reforms, perceptions of conflicting national targets (e.g. A&E targets), national staffing shortages and significant pressures on social care services.

• These factors were felt to have contributed to significant pressures on service provision, for example through delays in the transfer of patients through the system and resultant difficulties in finding beds for stroke patients in HASU and SU wards.

• Key promoters of sustainability of the system included:

  • the characteristics of the model itself (in particular the service standards linked to the London stroke tariff)

  • processes of sustaining the model, including regular service reviews and use of national audit data

  • leadership of the model (in terms of continuity but also adaptability)

  • the prioritisation of generating and sharing independent evidence was key to ensure ongoing stakeholder ownership.

Design

This chapter presents a single case study, based in London and focusing on governance and service levels (see Table 2 and Chapter 2, Understanding development, implementation and sustainability).

Data

For this analysis, we conducted 51 interviews and 24 non-participant observations (approximately 60 hours’ worth), and collected > 100 documents (see Table 2 and Chapter 2, Understanding development, implementation and sustainability). We interviewed a range of stakeholders, including system leaders, commissioners, stroke service staff and service managers; we observed such events as Stroke Clinical Leadership Group (SCLG) meetings and stroke service reviews, and documents collected included meeting papers (e.g. minutes, project updates), service review summaries and local system documents (e.g. updates to the London stroke service standards).

Analysis

To analyse sustainability of the system we were in part guided by the findings presented in Chapter 7, which indicated that the central factors in achieving a significant impact on quality of care and outcomes were: (1) that a high proportion of patients were treated in a HASU and (2) that HASUs had sufficient capacity (in terms of appropriate staff and equipment) to provide evidence-based care. Our analysis was also informed by factors established in the literature as potentially influencing sustainability of complex organisational change, including contextual factors, leadership, and processes to sustain the change. 47–49,126,215,221–223

Flow of patients through the hub and spoke model

Hub and spoke systems (as in London and GM) rely on the timely transfer of patients through the system. In this section we discuss how the flow of patients through the system was sustained, in terms of (1) direct transfer to a HASU, (2) repatriation from a HASU to a SU and (3) discharge from the acute system to the community.

Care provision in Hyperacute Stroke Unit and Stroke Unit services

As described in Chapter 11, the provision of evidence-based clinical interventions either improved or was sustained over our period of study. However, this was achieved in spite of significant pressures faced by the system, in terms of capacity, staff turnover and a national shortage of stroke specialists.

Governance of the London system

As described in Chapters 6 and 7, planning and implementation of the changes in London were influenced significantly by how they were governed. System leadership, the presence of a central source of authority, was important to ensure that a wide range of stakeholders remained engaged in the changes, while the operational support provided by local Stroke Clinical Networks was instrumental in ensuring standards were met across the system.

Principal findings

The London acute stroke system – like the rest of the English NHS – faced significant obstacles over the period studied. However, this complex system was sustained through a range of factors relating to its design, leadership and management.

Important contextual factors included: turbulence prompted by the 2013 reforms to the NHS in England; national targets (especially the 4-hour A&E target); national staffing shortages (in terms of medical, nursing and allied health professionals); and significant pressures on social care services. These factors contributed to significant pressures on service provision, for example through delays in the transfer of patients through the system and resultant difficulties in finding beds for stroke patients in HASU and SU wards.

The key promoters of sustainability of the system included the characteristics of the model itself (in particular the service standards linked to the London stroke tariff), processes directly aimed at sustaining the model (including regular service reviews and use of national audit data), leadership of the model (achieving a degree of continuity, but also adapting the style of leadership to be more persuasive, for example using evidence to persuade nascent CCGs to engage with the system) and, finally, the prioritisation of generating and sharing independent evidence was key to ensuring ongoing ownership and support of the London model by clinicians, managers and commissioners.

Strengths and weaknesses

This was a rare opportunity to analyse the sustainability of a MSC over several years, drawing on a range of quantitative (see Chapter 11) and qualitative data, assessing sustainability from multiple stakeholder perspectives. This approach reflects recommendations in the literature for optimal assessment of sustainability. 222 The fact that the changes were sustained, despite the loss of several important facilitators (as the result of the NHS reforms implemented in 2013), means that we provide a number of lessons on sustaining change during times of significant turbulence, which should be valuable to planners, managers and clinical leaders in the English NHS and elsewhere.

Our analysis had a number of limitations. First, we sampled only a limited number of HASUs and SUs. However, the governance meetings we observed had representation from across all parts of London, and many governance-level interviewees took a system-wide view of HASUs and SUs; also, several interviewees were based in HASUs or SUs that were not our selected study sites.

Comparison with other studies

This chapter demonstrated clearly the influence of factors identified in the wider organisational change literature on sustainability, suggesting that many of the lessons relating to sustainability or complex change conducted within a single organisation can be applied in relation to MSC.

The characteristics of the innovation (i.e. the service standards linked to the London stroke tariff, measured regularly) were a key facilitator of sustained staffing levels within services across the system. 41,47,48,221,223 However, reported difficulties in adapting standards (e.g. to meet requirements for 7-day working) was in line with the observation of Greenhalgh et al. 49 that ongoing development of a MSC can be challenging. The context in which change is implemented is highly influential, and in the case of MSC the complexity of this relationship grows. Sustainability of a MSC requires ongoing responsiveness to such changes. 47,49,215,221 We found that continuity in leadership was vital to ensure ‘organisational memory’ for the purpose and processes of the London system, but that the ways in which leadership is enacted must adapt to changes in the organisational context. One issue emerging from the pan-London approach, emphasised from the beginning of this change process, is that local adaptations to the system tended to be brought to system leadership for discussion before being put into action, and then used as a basis for consideration of system-wide adaptations. This suggests that following a MSC, once the system reaches a certain stage of maturity, with certain principles established (e.g. the service standards and ‘hub and spoke’ structure), a degree of flexibility may be introduced to permit localised piloting.

Implications

A key implication of this chapter is that, when planning a MSC, there would be clear value in looking beyond its implementation to its long-term sustainability. A range of factors, relating to innovation design, the context in which it is to be introduced, how it is led, and how it is managed and supported, should be considered. Especially important was prioritising the generation of meaningful evidence about the implementation and impact of the MSC: this addressed the question of whether or not the innovation was having the intended effect, but was also fundamental to ongoing dialogue with clinicians, managers, commissioners, patients and the wider public, to ensure ongoing engagement and support of the system.

Implications relating to the ‘what works’ question

Our study provides evidence to support the centralisation of acute stroke services in urban areas, and that this service model should include all stroke patients, not just those suitable for thrombolysis:

• This is supported by our analysis of the original changes in London and GMA, which found that the London model, in which the majority of patients were treated in a HASU, resulted in reductions in mortality, whereas the GMA model in which most patients were not treated in a CSC/PSC, did not.

• The finding above was further strengthened by our analysis of the GMB model, in which the majority of patients were treated in a CSC/PSC that showed reductions in mortality.

• These findings are supported by our analysis of the provision of evidence-based clinical interventions (i.e. that these were more likely to be provided in the London model compared to GMA, and that the provision significantly increased with GMB and was maintained in London). The findings in relation to improvements in the provision of clinical interventions provide some explanation for our findings in relation to improvements in clinical outcomes.

• Our study suggests that centralisation of acute stroke services can be cost-effective (London and GMA at 10 years, and GMA and GMB at 90 days), predominantly through improved outcomes or reductions in LOS.

Our findings suggest that the centralisation of care provision can offer patients and carers a good experience of care:

• The disadvantages of travelling further were perceived to be outweighed by the opportunity to receive the best-quality care. The major contributions of this study are highlighting the necessity for clear and understandable information to patients and their families about every stage of the centralised care pathways, in order to maximise their experience of care and highlighting the importance of wide stakeholder engagement in the development and provision of centralised services, so that relevant components of the wider system have sufficient ownership and understanding of system delivery (e.g. the importance of the role of the ambulance service in explaining the centralised pathway on initial contact with stroke patients).

Implications relating to the ‘how’ question

Our findings provide implications for other urban areas considering centralising acute stroke services:

• In relation to what issues to consider when deciding on the model –

  • Service models should be simple and inclusive to ensure that health-care staff and the public understand the new service and to ensure that all patients who might benefit from specialist care are eligible for treatment in a specialist unit, not just a selection.

• in relation to the processes of implementation –

  • Service standards should be used to ensure that all services are able to provide the best possible care; providers should receive support to achieve standards and ongoing achievement of standards should be linked to financial incentives.

  • The reconfiguration should be implemented ideally as ‘big bang’ (for instance, the London centralisation was launched on a single day) or at least minimise the phases. The timing of these stages should be clearly communicated so that the change is clear, and a smooth transition is achieved.

Our findings also provide implications for other types of MSC. 110

An important contribution of our study was that we were able to combine a qualitative analysis of processes of change with a quantitative analysis of the impact and sustainability of these changes. In doing so in Chapter 6, we extended five lessons for MSC from Best et al. :5

• Interplay between bottom-up and top-down leadership in achieving MSC – system-wide political authority or governance may be needed to align multiple stakeholders over a large scale and encourage clinical commitment to system-wide improvement goals and thus overcome resistance to change. Continuity of leadership is important, both to implementation and sustainability of change. However, the successful implementation of GMB suggested that MSC can be implemented successfully without making use of system-wide authority. However, we cannot discount the possibility that the previous reconfiguration (GMA) familiarised individuals with the issues and thus facilitated implementation.

• Combining feedback with other tools (e.g. use of audit data) was important to build the case for change and to assess its impact. Feedback may need to be combined with other tools to encourage behaviour change (e.g. financial incentives).

• Although learning from previous experiences of change can be a valuable guide to planning and implementing change, contextual factors can be a barrier to implementing these lessons (e.g. local resistance to downgrading or closing services); political authority may be needed to challenge the existing context and enable more radical forms of transformation.

• There is a need to involve a range of stakeholders in planning MSC (including but not limited to physicians) and have a system-wide governance structure to align their interests.

• Awareness that the drivers of MSC (e.g. clinical, political, social, financial) influence how different stakeholders’ views come to count during implementation; potential tension between patients’ and others’ perspectives.

The importance of system-wide leadership was illustrated further in our analysis of changes that were not implemented in the Midlands and East of England, following major reform of the NHS in England in 2013 and the removal of a key source of system leadership (SHAs). When this system-wide leadership does not exist or is less effective, we suggest that consideration should be given to how available sources of leadership can be used to enable the joint planning and implementation of MSC. The approaches employed in London to sustain their services (e.g. consistency of leadership, standards linked to financial incentives) and in GM in implementing further reconfiguration during this period (engaging senior commissioners to lead; recreating network capacity to facilitate the new changes) offered some examples of how the turbulence and threat of major structural reform might be mitigated.

Our study adds new empirical knowledge to the relatively small body of literature on involvement of patients and the public in MSC.

This may have wider implications than just for the reconfiguration of acute stroke services. The value of PPI in this case was found not in its contribution to acute service re-design, but in how involvement practices facilitated its implementation. Involvement was seen to have strategic and intrinsic value. Its strategic value lay in facilitating the implementation of a model of care that aimed to deliver evidence-based care to all; its intrinsic value was in the idea of citizen participation in change processes as an end in its own right. The concept of value, rather than impact, may provide greater traction in both the enactment of involvement in MSC and in analyses of involvement practices.

Impact on service reorganisation

Impact on national policy and recommendations

Our findings on clinical outcomes25 were presented in the Five Year Forward View as evidence of the ‘compelling case for greater concentration of care’ (p. 23),80 associated guidance on transforming urgent and emergency care services across the English NHS (p. 11)226 and the National Clinical Strategy for Scotland 2016 (pp. 70, 72). 208 Our findings on clinical interventions26 have been cited in the 2016 edition of the National Clinical Guidance for Stroke (pp. 13, 16 and 18)209 as evidence of the benefits of stroke service reorganisation.

Impact on our framework for major system change

Our qualitative findings in relation to the decision to further centralise services in GM and the sustainability of change in London suggested the potential influence of evidence of impact on care and outcomes on the decision to sustain or change a service model. We have, therefore, produced an updated figure (Figure 19) to illustrate feedback from implementation outcomes (C4) and intervention outcomes (C5), to decisions to change (C1), selection of model to be implemented (C2), and implementation approaches used (C3). Revised framework of MSC.

FIGURE 19.

Revised framework of MSC. Light blue shading denotes how key stages of change feed into the progress and outcomes of a given change. Darker green shading denotes how evidence on the outcomes of the different stages of change can feed back into future changes.