Parts of this chapter are adapted with permission from Fisher et al.1 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: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.
Introduction
The SSNAP is the national stroke register of England, Wales and Northern Ireland in which all acute admitting hospitals and post-acute stroke teams are mandated to participate.61 SSNAP has played a key role in monitoring the performance and improving the provision of acute stroke care. The collection of SSNAP data from community stroke services now offers a unique opportunity to investigate the large-scale impact of ESD.
Clinical guidelines recommend that ESD services should provide responsive and intensive rehabilitation (with treatment at home beginning within 24 hours of hospital discharge), with the aim to promote stroke survivor recovery.5,6,62–65 By investigating if and how these aspects of an effective ESD service can be realised in practice, this study aims to inform the provision of evidence-based care for stroke survivors.
Our previous research has hypothesised that the active ingredients of ESD can be defined with evidence-based core components,13 and that these core components are essential characteristics that need to be implemented for the ESD intervention to be effective in practice.12 The aim of this study was to determine if such core components had been adopted by ESD teams in real-world settings in England and whether or not these related to realised benefits of ESD.
Methods
Study design
We present results from this observational cohort study (), which was conducted as part of the larger mixed-methods study.37 We determined a priori a sample size of 4750 patients for a study power of 80% to detect standardised effect sizes of 0.25 for each outcome.
The ESD study design flow chart. WISE, What is the Impact of Stroke Early supported discharge?
Setting
Early supported discharge services were sampled across a large geographical area of England. The sampling strategy was devised in accordance with the overall mixed-methods study design and included all ESD services in specific regions of England.37 Here we report the findings from the quantitative investigation of ESD effectiveness across the West and East Midlands and East of England (across which a specific initiative to promote ESD was initiated in 2010), and the North of England, a region with a defined lack of ESD.7,28
Data sources and participants
The aim of this study was to examine the association between ESD service models and process and patient outcome measures of ESD effectiveness. Information about ESD service models included in the study was obtained from SSNAP post-acute organisational audit data, which were published freely in the public domain in 2015.7 ESD teams had participated in the 2015 post-acute organisational audit by completing questionnaires that investigated the organisational characteristics of their service in relation to evidence-based standards, which were distributed and collated by SSNAP.7
Patient-level SSNAP data are entered by clinical teams onto a secure webtool with real-time data validations to ensure data quality.61 Historical, prospective, clinical (patient-level) SSNAP data from all SSNAP-participating ESD teams in the geographical area of interest (n = 31) were obtained with permission from HQIP.
Key predictor: early supported discharge consensus score
We hypothesised that the adoption of evidence-based core components of ESD was important for the ESD intervention to be effective in practice. An ESD consensus score was developed using defined evidence-based core components of ESD, as outlined in an international consensus document and evidence-based post-acute organisational audit criteria utilised by SSNAP in the post-acute audit ().7,13 Statements defining core components of ESD from the consensus document (derived using an international panel and modified Delphi process) were compared with items from the post-acute organisational audit questionnaire that was used previously by SSNAP. Using this process, a 17-item ESD consensus score was designed by the study team to measure the adoption of core components of an ESD service model, for example team composition (core team and others), staff training, team meetings and service specificity (). This 17-item ESD consensus scoring system was then applied to organisational audit questionnaire data (categorical data previously collected by SSNAP) for each of the 31 ESD teams involved in the study. The adoption of evidence-based core components was measured by calculating an ESD consensus score for each of the 31 teams.
Configuration of ESD consensus score measure
The ESD consensus score components across 31 ESD teams
To evaluate the level of service provided by the ESD teams, we proposed a scoring system, as set out in , that relates ESD organisational audit data directly to evidence-based core components. Based on our proposed scoring system, an ESD team can score a maximum of 17 points: a maximum of 5 points for core team members meeting or exceeding the recommended WTE level per 100 stroke patients, and a maximum of 3 points each for access to other team members, training opportunities, MDT meetings and level of service provided.
Process and patient outcome measures
The measures of the effectiveness of ESD were based on clinical guidelines and ESD systematic review recommendations, and were dependent on what patient-level SSNAP data variables were collected routinely.5,10 Using historical prospective SSNAP clinical data (1 January 2016 to 31 December 2016), measures of ESD effectiveness were ‘days to ESD’ (number of days from hospital discharge to first face-to-face contact; number of patients, n = 6222), ‘rehabilitation intensity’ (total number of treatment days/total days with ESD; n = 5891) and stroke survivor outcome (modified Rankin Scale score at discharge from ESD; n = 6222). The measure of rehabilitation intensity was based on established approaches used by SSNAP.66 The modified Rankin Scale score, routinely collected at discharge from the ESD service, was used as the stroke survivor outcome and in analysis was controlled for by modified Rankin Scale score at discharge from hospital.
‘Days to ESD’ was a binary variable (0 = ESD team sees the patient within 1 day; 1 = ESD team sees patient after ≥ 1 day) and ‘rehabilitation intensity’ was a natural log-transformed continuous measure (the results presented in the text have been back transformed to give the per cent change per unit). The stroke survivor outcome measure of modified Rankin Scale (at ESD discharge) was treated as an ordinal categorical variable with the following categories of increasing dependency: 0, 1, 2, 3 and 4–5 (combined owing to small patient numbers).
Other variables
To investigate the effect of ESD consensus score on the process and patient outcomes, we controlled for a number of covariates, which were measured at the ESD team level (level 2 in our multivariate model described in Statistical analyses) or the patient level (level 1).
We identified a need to control for the effect of preceding hospital care and geographical context of delivery of rehabilitation. At the site (or ESD team) level, we included two confounding variables: a rurality score and a hospital SSNAP rating score. The rurality score was based on the rural–urban classification reported for the geographical area associated with the NHS CCG who had procured each ESD team.67 Each CCG in England has a geographical area over which it operates to procure NHS services. Where an ESD team included in this study was managed by multiple commissioning groups, the weighted average level of rurality was calculated based on the prevalence of stroke and transient ischaemic attack in that commissioning area (figures obtained from NHS Quality and Outcomes Framework68).
The hospital rating scores used in this study were an overall quality rating for each hospital obtained from SSNAP (total key indicator score derived across 10 domains of stroke care, with adjustments made for case ascertainment levels and the quality of data submitted to SSNAP). The score for each referring hospital (associated with each ESD team of interest) was used as an indication of the overall standard of inpatient care prior to ESD referral.69 For ESD teams with multiple discharging hospitals, a weighted average SSNAP rating score was calculated based on the number of patients being discharged to those ESD teams.
To account for differing patient characteristics between ESD teams, we also included variables at the patient level. These were stroke patient characteristics, reflecting validated stroke case-mix models and collected as part of the SSNAP data set, and included age, sex, pre-stroke independence, comorbidities, NIHSS score on admission, type of stroke and modified Rankin Scale score at discharge from hospital.31,32
Statistical analyses
Multilevel modelling was used to investigate the relationships between ESD model and process and patient outcomes in an approach consistent with previous observational studies of this type.31,32,70,71 By combining SSNAP post-acute organisational audit data at the site (ESD team) level with SSNAP clinical audit data at the patient level, we fitted generalised linear mixed models on two levels, ESD team (level 2) and patient nested within an ESD team (level 1), to process and patient outcome variables. Covariate adjustments were made for site (ESD team) (level 2) and patient (level 1) variables. Models were fitted for ‘days to ESD’, ‘rehabilitation intensity’ and modified Rankin Scale score at ESD discharge using multilevel logistic, linear logistic and ordinal logistic models, respectively.
The ESD consensus score was used in three different ways: total score, disaggregated by component and, where appropriate, as an individual item. We began by assessing the significance of the total score in relation to our outcomes of interest (both unadjusted and adjusted). If a significant association was found, further analyses by components and then by individual items were conducted to uncover the key driver(s) behind the significant association(s). Any statistically significant components were tested for linearity (using likelihood ratio tests) to assist with substantive inference. Where possible, variables were interpreted in a continuous fashion, otherwise they were treated as categorical if any variable could not be interpreted in a linear way.
We chose multilevel modelling to evaluate the effectiveness of ESD service provision because it can accommodate and appreciate the variation that may exist within and between different ESD teams. Furthermore, the intraclass correlation coefficient was calculated as a measure of the proportion of the total variance in outcomes that is attributable to variance within ESD services as opposed to between services.
The adequacy of different statistical models was compared using the log-likelihood, Akaike information criterion and Bayesian information criterion values from single-level and multilevel regression models for each outcome variable, with multilevel preferable on each occasion. Multicollinearity was investigated by examining variance inflation factor scores of all predictor variable sets and was found not to be an issue. Covariate linearity was examined by checking the consistency of a linear trend in relation to each outcome variable. To explore the impact of missing data, we conducted a sensitivity analysis excluding any teams that had missing outcome data; no substantial differences were found. A two-tailed significance level of 0.05 was used in all hypothesis tests. We carried out all analyses using Stata/SE® 15.1 (StataCorp LP, College Station, TX, USA).
Results
shows the variation of ESD consensus scores across the 31 ESD teams and four English regions. The total ESD consensus scores across the 31 teams varied between 5 and 15 {mean 10.6 [standard deviation (SD) 2.4]}, with no team achieving 100% adherence, reflecting that a range of ESD models had been adopted. In terms of the English regions, adherence to the core components of ESD service delivery was greatest in the East of England.
Distribution of ESD consensus score by (a) study team; and (b) region.
All 31 ESD teams included in this study reported that they provided a stroke-specific service (see ). Only three teams reported having at least the recommended level of input from doctors and only four teams said that they had access to social workers. For the range of ESD models, there was a mixture of urban and rural settings [mean level of rurality 35.6 (SD 21.8)], as well as varying performance of associated referring hospitals [mean SSNAP hospital rating score 72.2 (SD 12.1)].
Data from 6260 patients with a completed NIHSS score were included in the primary analysis, and their characteristics are shown in . The majority of patients (91.9%) had a mild or moderate stroke (NIHSS score of < 15). The most common age group was 70–79 years (30.8%) and 4151 (66.3%) patients were functionally independent prior to their stroke (modified Rankin Scale score of 0).
In terms of the outcomes, 69% of the sampled patients were seen after ≥ 1 day, with 31% of patients seen within 1 day for the ‘days to ESD’ variable. The median rehabilitation intensity value of the sampled patients was 0.38 treatment days for every day with the ESD team, with the 25th percentile being 0.19 and the 75th percentile being 0.59. For the stroke survivor outcome measure, 9% of sampled patients were classified as moderate to severe at ESD discharge (modified Rankin Scale score of 4–5), with the percentages of patients with a modified Rankin Scale score of 0, 1, 2 and 3 being 9%, 31%, 31% and 20%, respectively.
Results of the multilevel modelling are presented below. The degree of clustering was greater for the process measures ‘days to ESD’ and ‘rehabilitation intensity’ than that for the patient outcome measure of modified Rankin Scale score (adjusted intraclass correlation coefficients: 0.56, 0.26 and 0.08, respectively). shows the amount of variation among the 31 ESD teams in relation to the outcomes of this study.
Distribution of study outcome variables across study teams. (a) Proportion of EDS teams that make patient contact within 24 hours; (b) mean rehabilitation intensity; and (c) mean modified Rankin Scale score.
Results for the association between the total ESD consensus score and the ‘days to ESD’ variable are shown in , which are unadjusted and adjusted for all patient characteristics, the level of rurality and weighted average SSNAP hospital score. Odds ratios are also presented in , with percentage odds reported here. From the adjusted results, a 1-unit increase in the ESD score was associated with an odds ratio of 0.71 [95% confidence interval (CI) 0.51 to 0.99] or, in other words, with a reduced odds (by 29%) of the ESD team seeing the patient after ≥ 1 day following hospital discharge. Hence, an increase in ESD consensus score was associated with a more responsive ESD service. Exploring the effect of components, this association appeared to be driven by having more core team members meeting or exceeding the recommended WTE level per 100 stroke patients [a 1-unit increase was significantly associated with a 47% reduction in the odds of the ESD team seeing the patient after ≥ 1 day (95% CI 14% to 67%)]. There was some evidence at borderline significance of an effect of access to other team members (reduced odds of 70%, 95% CI –8% to 92%). Further investigation at an individual item level showed that having access to a social worker was associated with more responsive ESD service, with a 97% reduced odds of the ESD team seeing the patient after ≥ 1 day (95% CI 61% to 99%).
Association between ESD consensus score/components and days to ESD
presents the linear multilevel model results for the rehabilitation intensity outcome measure, which are unadjusted and adjusted for all patient characteristics, the weighted level of rurality, the average SSNAP hospital score and the total ESD consensus score. Focusing on the adjusted results and coefficients (presented as percentages here), the ESD consensus score was significantly associated with treatment intensity, such that a 1-unit increase in ESD consensus score increased treatment intensity (total number of treatment days/total days with ESD) by 2% (95% CI 0.3% to 4%). With respect to this significant association, holding weekly MDT meetings with the core team attending (see ) and a member of the ESD team attending the acute meetings were all positively associated with increased rehabilitation intensity: specifically an average of 8% (95% CI 0.9% to 16%) improvement in rehabilitation intensity.
Association between ESD consensus score/components and rehabilitation intensity
presents the ordinal logistic multilevel model results for the patient outcome measure, which are unadjusted and adjusted for all patient characteristics, weighted average SSNAP hospital score and level of rurality. There was no significant association between the ESD consensus score and the stroke survivor outcome measured by the modified Rankin Scale score at ESD discharge. Site-level control variables, namely percentage rurality and hospital SSNAP rating score, had no statistically significant relationship with any of the outcomes.
Association between ESD consensus score/components and stroke survivor outcome
Discussion
This study was designed to inform the large-scale implementation of ESD by evaluating its effectiveness in real-world conditions, at scale, using recommended methodology.71 This addresses recent recommendations for investment in stroke rehabilitation made in NHS England’s Long Term Plan4 and the lack of large-scale development of ESD worldwide.15 The study found that a variety of ESD service models have been adopted in regions of interest, as reflected by the variability in the ESD consensus score. The ESD consensus score was significantly associated with a more responsive ESD service (reduced odds of patient being seen after ≥ 1 day) and increased rehabilitation intensity when controlling for patient characteristics and other confounding variables, but no effect on stroke survivor outcome, as measured by the modified Rankin Scale, was demonstrated. We conclude that adopting defined core components of ESD was associated with providing a more responsive and intensive ESD service, suggesting that adherence to evidence-based criteria is likely to result in more effective services in practice. This builds on methods used to investigate the organisation of stroke unit care, bringing a much needed focus on community-based stroke care.70
There are limitations inherent to observational data that we aimed to address with the study design. Although the study used a large sample of stroke patient data, it must be acknowledged that data from a relatively small sample of ESD services were used in this study; further research would be required to confirm wider transferability, particularly beyond England. A key feature of this study was the development of the ESD consensus score. Although we acknowledge that more in-depth investigation of ESD model features is required to make definitive conclusions, this approach offered a useful way to quantify the adoption of core components for quantitative analytical purposes.37 It provided a simple means by which to evaluate services based on international consensus and clinical guidelines relating to ESD.5,6,13 We attempted to control for a number of confounders, however we cannot rule out the possible influence of unobserved variables. Outcomes of interest were reliant on a relatively small SSNAP data set, which was entered by community stroke service staff. Findings are reliant on accurate reporting and the possibility of bias cannot be excluded. It should also be noted that previous studies have suggested that ESD reduces length of hospital stay; investigation of this, using hospital SSNAP data, is reported in Chapter 5.37
Clinical guidelines emphasise the importance of seamless transfers of care and previous studies have reported the negative impact of delayed or un-coordinated transfers on patients.23,38 In addition to teams with higher total ESD consensus scores being more likely to see patients sooner, findings highlighted the importance of the ratio of staff to patients. Hence, teams that met (or exceeded) consensus-recommended WTE levels of staff per 100 stroke patients were more likely to be responsive, emphasising the need for ESD services to be appropriately resourced.13 Previous studies have also highlighted the transfer problems relating to lack of joint working between health care and social care.23,38,72 This study adds to this debate by highlighting the importance of access to a social worker as part of the ESD team.
Early supported discharge has been recommended as a high-intensity rehabilitation intervention, with guidelines and systematic reviews referring to daily visits or four or five visits per week.5,6,10 In this study, the intensity of rehabilitation delivery was measured by calculating the percentage of treatment days in relation to the patients’ total time with the ESD service. In addition to the total ESD score, the MDT working component was associated with increased intensity of rehabilitation delivery. This resonates with previous studies emphasising the importance of MDT working in the delivery of stroke care and, in particular, MDT meetings.73–75
Routine collection of patient outcomes in SSNAP is currently limited to the use of the modified Rankin Scale. Findings could be interpreted such that the model of ESD adopted did not influence patient outcomes, as measured by the modified Rankin Scale; however, caution is required. Robust modified Rankin Scale data were available only at discharge from the ESD service (instead of at a later follow-up stage) and, therefore, it is possible that there was not sufficient time to investigate ESD effects. There was also a lack of variability of this outcome measure in the study, possibly reflecting a focus of ESD services on treatment of mild to moderate stroke survivors. There have also been concerns from teams about the reliability of the use of this score across the stroke care pathway.76 We suggest that routine collection of additional validated patient outcome measures (e.g. measuring activities of daily living, general health/mood and quality of life) at longer follow-up periods in national stroke audits or registries is required.10,12,13
Finally, at the site level, the lack of effect of rurality was surprising. It is encouraging that we found examples of evidence-based ESD models in rural regions; however, reported challenges with health-care provision in these settings cannot be overlooked.77,78 Further investigation of the impact of geographical location on implementation of ESD is required.
Conclusion
Original clinical trials of ESD were conducted across the world and implementation of ESD is recommended in many countries’ stroke guidelines.5,6,62–65 This study supports the use of an international ESD consensus document as a means to guide the implementation of effective, evidence-based ESD in practice.13 We suggest that the extension of national stroke registries with the inclusion of community stroke data could offer important opportunities to evaluate stroke service delivery beyond the hospital setting.16 This could go some way towards addressing the current gaps in the provision of stroke rehabilitation that exist globally, moving towards the goal of ensuring that stroke survivors receive the evidence-based care that they deserve.
