Development and evaluation of interventions and tools to improve patient and carer centred outcomes in longer-term stroke care (LoTS care) and exploration of adjustment post stroke

Creation of problem-specific reference guides

Having identified the key problem areas and operationalised them through appropriate assessment questions, a range of literature, both within and beyond stroke, was reviewed to identify effective evidenced-based service interventions. The 16 problem-specific reference guides (relating to the 15 patient questions plus one carer-specific question) contain educational text with supporting assessment/treatment algorithms and checklists. The algorithms guide problem-solving as many apparently straightforward problems become multilayered on closer examination. For example, the activity of shopping might be impeded by physical barriers (mobility, lack of suitable transport), cognitive problems (poor comprehension, short-term memory loss) or psychological problems (fear, embarrassment). The final stage in this development phase was to obtain feedback from a range of local community and primary care professionals. This involved presenting and discussing the model during two workshops. Feedback from the workshops enabled us to frame our model into a product that was considered acceptable to primary care professionals. The resulting system of care is presented in a manual.

Manual presentation

Evidence on the implementation of clinical guidelines suggests that presentation can influence uptake into practice. The National Institute for Health and Care Excellence (NICE)29 details five key principles – language and style, bulleted lists, tables and figures, abbreviations and algorithms – and these were used to structure the components of the system of care. The aim of these principles is to make the information clear, accessible to non-specialists, unambiguous, succinct and guiding, but not prescriptive.

The resulting manual comprises patient- and carer-structured assessments representing the identified problem areas linked to, a reference guide and treatment algorithm; patient and carer goal and action plans; a directory of service information; and a selection of validated assessment scales for specific areas such as depression30 and cognitive impairment,31 included as appendices. Thus, we have created a manualised system for longer-term stroke care that is comprehensive (encompasses all areas of potential concern to patient and carer) but individualised (patient-specific action plans constructed).

Clinical implementation

The next stage in the development of the system for longer-term stroke care was to identify means of clinical implementation. Our previous work has identified a need for a service embedded in the community, where staff would have a greater awareness of locally available services. 22 Other work has identified that nurse-delivered interventions have been effective. 32 A survey of local district nurses demonstrated that nurses had relevant insights and skills but would appreciate additional training. 33

The National Service Framework for Older People16 identified a new role of ‘stroke care co-ordinator’ (SCC) to provide advice, arrange reassessment when necessary and co-ordinate long-term support. But who should be involved and how this role might be fulfilled are ill-defined. Through national presentations of our work, through links with the Stroke Association and through the National Stroke Nursing Forum, we were aware that a number of centres were developing SCC roles. To explore this further we undertook a national survey of all UK stroke services. For the purposes of this survey, a SCC was defined as ‘a qualified health professional in regular contact with stroke patients in the community, co-ordinating care inputs on their behalf’.

Pilot work: feasibility of the system for longer-term stroke care

Pilot work was conducted to investigate the practical implementation and feasibility of the proposed system of care. 23 A training programme was provided to community-based staff recruited to use the system of care in clinical practice. In total, 47 stroke patients and 21 carers were assessed using the new system of care. Analysis of care plans, 3 months after the initial assessment, indicated that the process was successful in picking up patient and carer problems. Of 219 problems, 75% had been resolved 3 months after assessment. Patient and carer participants thought that the review process would be more valuable if conducted sooner after hospital discharge (undertaken 8 months after stroke in the pilot study). We were able to demonstrate that a systematic assessment approach incorporated in a disease-specific manual is feasible to implement and was successful in identifying problems and triggering interventions.

Objectives

Projects 1 and 2 of the programme grant described here continue this work:

• project 1 – an update of the evidence-based treatment algorithms using a structured protocol for searching and assessing the available literature (see Chapter 2)

• project 2 – a cluster randomised controlled trial (RCT) of the system of care as delivered by SCCs, evaluating its impact on patients’ and carers’ psychological and physical outcomes and its cost-effectiveness (see Chapter 3).

Since this programme grant was awarded in 2007, there have been considerable changes in stroke service provision across the UK. Our system of care is very much in keeping with current developments. Similar approaches have been developed, for example the Greater Manchester Stroke Assessment Tool (GM-SAT) for 6-month reviews. 34 This indicates that our thinking is feasible, practical and appropriate for ongoing service development.

Objectives

Project 3 in this programme continues this work:

• further development and psychometric testing of a measure to assess and monitor patient longer-term met/unmet stroke-related needs (the LUNS questionnaire) (see Chapter 4).

Identification of needs after stroke has become increasingly topical during the implementation of this programme grant. The Stroke Association commissioned a survey of post-stroke unmet needs and co-applicants of this grant (Forster and Young) contributed to the development of this survey,4 which included some questions from the LUNS tool.

Objectives

Project 4 undertook further exploratory work:

• identify factors contributing to patients’ poor adjustment after stroke through a qualitative and quantitative case–control study nested in the randomised trial (see Chapter 5) [it was clarified during the review of the grant application that the design is not in fact a nested case–control study but rather a qualitative substudy (with purposive sampling based on the quantitative trial data)].

Clinical engagement

Clinical engagement has been sustained throughout the programme through:

• Yorkshire Stroke Research Network (YSRN) therapy meetings – held three times a year and open to all allied health professionals in Yorkshire

• the attendance of the chief investigator at the quarterly Clinical Stroke Network meetings, where research progress is reported as a regular agenda item

• the ongoing contact of the chief investigator, clinical lead for the YSRN, with all local stroke services and robust links with national colleagues.

Patient and public involvement

Terry Brady (co-applicant and stroke survivor) has been involved at all stages of programme development. He is a member of the CRAG, which has met bimonthly throughout the duration of the programme. The CRAG has been provided with regular updates on the progress of the programme and has contributed at all stages.

Terry Brady is also on the organising committee for the annual consumer conference, which is hosted by the YSRN and attracts over 70 stroke survivors and their carers each year. Updates on ongoing research are provided at the conference and this includes progress reports on this programme of research. In addition, advice and guidance were sought on specific components of the research through ‘round table’ discussions, for example outcome assessment for the cluster trial and layout of the outcome assessment booklet (project 2); considerable input into the development of the LUNS tool, both content and layout (project 3); and contribution to the adjustment after stroke study and consideration of the results (project 4).

Colleagues in the CRAG have become experienced in the review of stroke research. One of our members, Mick Speed, is now a member of the Stroke Association research panel.

Collaborations

Professors Anne Forster and John Young were members of the Stroke Association UK Stroke Survivor Needs Survey study team and this survey included some of the questions from the LUNS questionnaire.

Background

Our purposely developed system of care was framed around patient- and carer-identified problems. To support treatment delivery a manual was created that included algorithms to assist SCCs in clarifying problems and directing patients to treatment options. To ensure that all of the treatment algorithms were up to date and evidence based, they were rigorously reviewed and updated.

Methods

In collaboration with information specialists at the University of Leeds, a hierarchical, comprehensive structured protocol for identifying evidence in each of the 16 problem areas was developed. This protocol included identifying relevant stroke- and problem-specific guidelines; meta-analyses and systematic reviews; and, if necessary, individual RCTs. Two researchers independently reviewed all outputs. Guidelines, reviews and papers identified for inclusion were assessed for quality using standard tools. Drafted treatment algorithms were peer reviewed by external experts.

Results

Over 71,000 articles were identified by the initial searches. Robust guidelines were identified for three problem areas; for one problem area, information provision, we are the authors of the Cochrane review, which we updated. For the 12 remaining problem areas detailed searches were implemented. Following the review procedure algorithms were updated in accordance with the identified evidence and, following external peer review, they were incorporated into the system of care manual. Presentation of the manual was informed by expert opinion and feedback from clinical colleagues.

Conclusion

The system of care manual was created using robust methods to ensure that advice and guidance provided were up to date and evidence based.

The care plan

The care plan included questions addressing all problem areas supported by additional prompts. It also included goal and action planners for the patient/carer, which aimed to encourage and support shared decision-making about goals and the prioritisation of goals according to patient (carer) preference and joint problem-solving approaches. The layout and presentation of the assessment questions for patients and their carers were initially reviewed by the study management group (SMG) and clinical colleagues. The care plan was further refined by SCCs taking part in the trial, before the start of the trial (see Chapter 3, Intervention arm: system of care).

Part of the care plan is shown in Appendix 3.

The manual

The manual associated with our system of care was created in 2003. The manual included introductory text, reference guides with associated detailed treatment algorithms, which are centred around the patient- and carer-identified problems, and general advice on patient assessment. On commencement of the programme grant the manual needed updating to ensure that it remained a relevant and reliable resource.

Updating the manual

To ensure that all of the treatment algorithms were up to date and evidence based, in collaboration with information specialists at the University of Leeds, a hierarchical, comprehensive, structured protocol for identifying evidence in each problem area was developed (Figure 1).

FIGURE 1.

Protocol for identifying evidence-based interventions for stroke problem areas. AGREE, Appraisal of Guidelines Research and Evaluation; AMED, Applied and Complementary Medicine Database; CINAHL, Cumulative Index to Nursing and Allied Health Literature; HTA, Health Technology Assessment; SIGN, Scottish Intercollegiate Guidelines Network.

The reference guides linked to treatment algorithms addressed the following problem areas:

• transfer of care

• communication and information

• medicines and general health

• pain

• mobility/falls

• personal hygiene and dressing

• shopping and meal preparation

• house and home

• cognition

• driving and general transport

• finances and benefits

• continence

• sexual functioning

• patient mood

• patient social needs (and employability)

• carer social and emotional needs.

The protocol started (stage 1) with searches to identify all potentially relevant evidence-based stroke guidelines (e.g. the National Clinical Guidelines for Stroke2,41). This progressed to the identification of guidelines for problem areas of relevance to stroke but that were not necessarily stroke specific (stage 2) and the identification of systematic reviews and meta-analyses (stage 3). If it was considered that the problem area had not been adequately addressed (e.g. through the availability of national guidelines) then a comprehensive electronic search strategy was developed and implemented to identify all RCTs of effective interventions evaluated for stroke and for other diseases with similar experiences to those of stroke (stage 4).

The search strategy was developed iteratively. Each problem area was discussed between the researchers (A Forster, J Murray and R Breen) and the information specialists and consensus agreement was reached about scope (e.g. age limits, key words, databases to be searched). When available, guidance on search terms was taken from existing reviews (e.g. the Cochrane review on the prevention and treatment of urinary incontinence after stroke in adults42). Example search strategies for pain for MEDLINE and EMBASE are provided in Appendix 4 and other search strategies are available on request from the authors. All searches were restricted to the English language only, with a date restriction from 1995 onwards, and included appropriate methodological filters for guidelines, consensus statements, systematic and other reviews and RCTs. The information specialists then conducted initial searches with the results reviewed by J Murray and A Forster to verify that appropriate titles were being identified. Following further amendments as necessary, the full search was undertaken, with the results downloaded into EndNote (version 5; Thomson Reuters, CA, USA). All searches were undertaken in MEDLINE, the Cumulative Index to Nursing and Allied Health Literature (CINAHL), Evidence-based Medicine Reviews, PsycINFO and EMBASE. A member of the research team was designated as lead reviewer for each problem area.

The overall purpose was not to develop guidelines but to identify and present the evidence in a format that will aid the clinical application of the structured assessment system, triggering appropriate, evidenced-based interventions. These interventions had to be relevant to SCC practices; therefore, for example, in medicines management the focus was on knowledge rather than the SCC being an active prescriber of drugs. Similarly, when considering the problem area of mobility, interventions that require specialist equipment (such as a treadmill) were not considered as the SCC would refer to a physiotherapist for such specialist treatment.

The outputs from stages 1–3 were initially reviewed by one researcher (J Murray) and presented and discussed at regular SMG meetings (monthly then moving to fortnightly). For stage 4 (Figure 2), initial screening of the titles was conducted in two phases: the first reviewer carried out an initial screen to eliminate any obviously irrelevant papers and then the second reviewer screened all remaining papers. The abstracts/full papers were then considered by two reviewers independently using a standard proforma. A third reviewer was required when there was a conflicting opinion or both reviewers were unsure about study inclusion. Following discussions a consensus of studies for inclusion was reached.

FIGURE 2.

Written search strategy used in stage 4.

The Cochrane review of information provision for stroke patients and carers

The Cochrane review of information provision for patients and their carers after stroke has been updated twice during the programme. The systematic review was conducted using The Cochrane Collaboration-recommended methodology for undertaking systematic reviews of RCTs and full details of the methods and outcomes have been published. 43–45

Studies were included if there was random assignment of participants (patients with a clinical diagnosis of stroke and/or their identified carers) to intervention groups (one of which involved the provision of information and/or education) and to a control group. In addition, studies using a matched pairs design and studies in which strict randomisation procedures were not adopted were also considered for inclusion.

We excluded trials in which information giving was only one component of a more complex rehabilitation intervention, for example family support worker trials (e.g. Forster and Young18 and Dennis et al. 20), which are the subject of a separate Cochrane review. 46

The primary outcomes used to assess the effectiveness of information provision were patient and carer knowledge about stroke and stroke services, and impact on health, especially mood. Secondary outcomes were activities of daily living; handicap; social activities; perceived health status; quality of life; satisfaction with information; hospital admissions, service contacts or health professional contacts; compliance with treatment/rehabilitation; death and/or institutionalisation; and cost to health and social services.

Relevant trials were identified in the Cochrane Stroke Group Specialised Trials Register. Additional intervention-based search strategies were developed for the Cochrane Central Register of Controlled Trials (CENTRAL/CCTR, formerly Cochrane Controlled Trials Register), MEDLINE, EMBASE, CINAHL, ISI Citation Index, ISI Web of Science Service, ASLIB Index to Theses, Dissertation Abstracts International, Applied Social Sciences Index and Abstracts and PsychLIT/PsycINFO. We also searched registers of trials in progress and bibliographies of retrieved papers, articles and books.

Updating the manual

Following a review of all problem areas it was agreed that each category of problem should be subjected to a separate search, with the approach and range of the search varying according to the problem. Broadly three approaches to the search strategy were undertaken:

1. an inclusive general search of the problem area

2. the search was restricted to ‘stroke’

3. the search was restricted to stroke and other similar long-term conditions, which we termed a ‘chronic illness filter’.

A summary of the search methods used is presented in Table 1.

Over 71,000 titles were indentified by the individual searches, details are provided in Table 2. In total, 12 problem types required a review of evidence-based interventions from stage 3 (see Figure 1). Two problem areas, shopping and meal preparation and house and home, were addressed by a combined search and one problem area, patient social needs (and employability), was addressed by two separate searches.

Identified guidelines and papers were subject to appropriate quality assessment. Relevant evidence-based interventions from guidelines and studies found to be of high quality were then used to update treatment algorithms. Each was drafted by a lead reviewer and then reviewed by the SMG before external review. Final versions were further revised and formatted for inclusion in the manual.

The completed reference guides and treatment algorithms with supporting text are included in the manual; an example (for pain) is provided in Appendix 5.

Review of the evidence

Updating these treatment algorithms provided an opportunity to review evidence gaps. Available evidence was reviewed for each problem area. The results were tabulated and summarised to include available evidence and recommendations (if available) made in the Evidence-Based Review of Stroke Rehabilitation (see www.ebrsr.com), a relevant Cochrane review, Royal College of Physicians (RCP) guidelines,2 Australian guidelines52 or Canadian guidelines53 or identified through searches undertaken for the manual update. We also highlighted when a relevant RCT was ongoing. The resulting output is 85 pages long. Although the work was completed, we did not promote it as we were mindful that NICE was developing stroke rehabilitation guidelines (issued June 2013)54 that could replicate the work and the RCP was similarly updating its guidelines (issued September 2012). 55

The Cochrane review of information provision for stroke patients and their carers

For the first update44 we identified 228 abstracts, of which 63 studies were potentially relevant to the review. In total, 17 studies, involving 1773 patients and 1058 carers, were included (published 2009).

For the most recent update45 we reviewed 28,110 titles including 134 full papers, resulting in the inclusion of four additional studies. The current review includes 21 trials from seven countries involving 2289 patients and 1290 carer participants.

Meta-analyses of reported outcomes showed a significant effect in favour of the intervention on patient knowledge, carer knowledge and patient satisfaction with information provision. There was also an effect on patient depression; however, the reduction was small and may not be clinically relevant. There was no effect on the number of cases of anxiety or depression or on patient mortality. Qualitative analyses found no strong evidence in favour of the intervention for other outcomes relevant to stroke recovery, including independence, participation in social activities, service use or modification of health-related behaviours.

Post hoc subgroup analyses demonstrated that, when information was provided in a format that more actively involved patients and carers, for example by offering repeated opportunities to ask questions, it had more effect on patient mood than information provided on one occasion only. However, as we saw no effect on the dichotomous end points of anxiety or depression (number of cases), the effects may be small. The specific component of the active information provision that may provide beneficial effects requires further investigation. There was no evidence that active information provision strategies were effective for other outcomes.

Trial design

A pragmatic, multicentre, cluster RCT compared a new system of care with usual practice delivered by SCCs.

Methods

Stroke services were randomised to the intervention or the control. Patients with new stroke living at home post stroke and referred to a SCC, and their informal carers, were eligible for recruitment before their first SCC assessment. The primary objective was to determine whether the intervention improved patient psychological outcomes (General Health Questionnaire-12, GHQ-12) at 6 months; secondary objectives included further functional outcomes for patients, carer outcomes (if registered) and cost-effectiveness. Follow-up was through self-completed postal questionnaires at 6 and 12 months after registration.

Results

In total, 32 randomised stroke services (29 participated), including 800 patients (399 control, 401 intervention) and 208 carers (100 control, 108 intervention), were recruited. In intention-to-treat (ITT) analysis the adjusted difference in patient GHQ-12 mean scores at 6 months was −0.6 points [95% confidence interval (CI) −1.8 to 0.7 points; p = 0.394], indicating no evidence of a statistically significant difference between treatment arms; analyses of secondary end points and of the per-protocol population resulted in the same conclusion. The intervention involved no additional SCC time. Quality-adjusted life-years and total health and social care costs at 6 months, 12 months and over 1 year were similar between arms. Societal costs were higher for the intervention arm (+£1163 at 6 months, 95% CI £56 to £3271).

Conclusions

This robust trial demonstrated no benefit for clinical effectiveness or cost-effectiveness outcomes for the system of care compared with usual SCC practice.

Trial registration

This trial was registered as ISRCTN67932305.

Background

The diversity of longer-term problems experienced by stroke patients and their carers has long been recognised,56 but they remain poorly addressed by existing services. 57 There are persuasive arguments that a community-based orientation to post-stroke care delivered by primary care in the patient’s local community, to assess, support and co-ordinate relevant services, might be more helpful in minimising longer-term stroke morbidity. 22,58

To address this we have developed an evidence-based system of care that aims to meet the longer-term needs of stroke survivors and their carers living at home in the community. The system of care incorporates structured assessment focused on patient- and carer-centred problems and associated evidenced-based treatment algorithms.

We aimed to evaluate the clinical effectiveness and cost effectiveness of the system of care when delivered by SCCs compared with usual practice in a cluster RCT [Longer-Term Stroke care (LoTS care) trial]. The methodology for the trial was informed by our survey of 39 SCCs. 59 The survey, undertaken in 2006, revealed that approximately 40% of SCCs reviewed the patient within 1 week of discharge from hospital. Most SCCs (n = 27) saw patients more than twice after discharge from hospital, but the number of contact visits was often dependent on patient need (as opposed to defined time points). The majority of SCCs appeared to refer on to other services for carer assessments.

Objectives

Trial design

The trial has been designed as a pragmatic, multicentre, cluster RCT. The unit of randomisation was at the level of the stroke service, with SCCs (single practitioners or part of a community-based team) randomised to deliver the new system of care to all patients (and their carers, if appropriate) or to continue to deliver current practice as determined through local SCC policy and practices. Follow-up was through postal questionnaires at 6 and 12 months after recruitment.

A prospective cluster RCT design was used to reduce between-group contamination. The training component of the intervention was designed to impact on the skills, knowledge and clinical practice of SCCs and therefore the risk of contamination would be high if randomisation was at the individual patient level. If a service had more than one SCC it was likely that the SCCs would meet on a regular basis and work as a team, sharing caseloads. For these reasons the unit of randomisation was at the level of the service.

The system of care was evaluated as a whole to avoid a type III error,61 in which deconstruction of the component parts of a complex intervention may result in the loss of benefits of the interactions between components.

We recognised that the intervention may be effective on a number of domains. We therefore chose a range of outcome measures. The primary outcome measure was the General Health Questionnaire-12 (GHQ-12),62 to reflect patient mood, which would be influenced by a range of service inputs (including overall co-ordination of care). We aimed to capture psychosocial outcomes rather than just physical abilities as reflected in activities of daily living. The GHQ-12 has been used in previous stroke research and is sensitive to change.

Eligibility

Randomisation

Randomisation was undertaken by the University of Leeds Clinical Trials Research Unit (CTRU). Stroke services were randomised on a 1 : 1 basis to either the intervention group or the control group. A detailed overview of the stroke services was obtained before randomisation to inform stratification.

Randomisation was stratified by the quality of the stroke unit [National Stroke Audit (NSA) score based on 2006 data64], the annual number of referrals, whether SCCs work alone or within a community-based MDT and the Strategic Health Authority (SHA). When a SCC received referrals from more than one stroke unit, a weighted average of the NSA scores reflecting referral rates was used.

A method of obtaining a balanced design from the covariates available at the start of the trial, as proposed by Carter and Hood,65 was used. This method uses an imbalance measure to compare the extent to which all possible designs balance important covariates between the arms of the trial and then randomly selects one member from the class of designs that achieves maximum balance. The advantage of the method is that it allows randomisation in multiple phases, conditioning on previous allocations. Because of the complexities of the current research approvals process, which created variable time delays (see Appendix 6), the services were randomised in two phases.

Blinding

As this was a cluster trial, all patients within the services randomised to the intervention received the new system of care and all patients within the services randomised to the control received usual care.

Recruitment of trial participants was by members of the independent reseach team and the SCCs were unaware which of their patients had consented to participate. Members of the research team were blinded to whether they were recruiting within an intervention service or a control service. This trial design has the advantage of reducing the potential for selection bias from differential recruitment, inherent if the SCCs were responsible for patient identification, and it minimises altering of SCCs’ clinical activity for trial participants.

Interventions

All patients received the services of a SCC in a community stroke service. SCCs within the intervention arm provided care according to the new system of care. Control SCCs continued with usual practice.

Assessments

Outcomes

In common with other stroke rehabilitation trials78,79 the primary outcome point was at 6 months, with the final follow-up at 12 months to assess whether any intervention effect was sustained.

Sample size

Initial sample size calculations, based on the primary outcome measure, GHQ-12 at 6 months, indicated that recruitment of 800 patients from 40 services would provide 90% power at the 5% significance level to detect a clinically relevant difference of 2.5 GHQ-12 points [standard deviation (SD) 7 points], as defined in the Trial of Occupational Therapy and Leisure (TOTAL). 78 This sample size accounted for loss to follow-up and clustering; the inflation factor of 1.95 was derived from a maximum cluster size of 20 and an intracluster correlation coefficient (ICC) of no greater than 0.05. 80,81 The TOTAL study achieved an 80% response rate to postal questionnaires at 6 months and a 71% response rate at 12 months. We therefore estimated the maximum loss to follow-up at 6 months as 25%. Losses to follow-up are likely to increase over time78 and the interpretation and credibility of results are difficult if losses exceed 30%. Therefore, follow-up was limited to 12 months post recruitment and the primary outcome was defined as 6 months, when there will be a higher follow-up rate.

The SCCs indicated that they were likely to receive two or three new referrals per week, of which one or two were likely to consent to inclusion in the trial. The initial aim was to recruit 40 SCC services; if 40 services recruited an average of 20 patients, this would result in the required number of patients (800) and the trial having 90% power to detect the clinically relevant difference (2.5 GHQ-12 points). Overall, 32 services were randomised, each with the aim of recruiting 25 patients, providing a power of 88%. To minimise unequal recruitment, the maximum number of patients per service was capped at 45, maintaining 85% power.

Statistical methods

Statistical analyses were based on the ITT population. All statistical testing was performed at a two-sided 5% significance level.

The ITT population was defined as all patients registered for active follow-up regardless of non-compliance with the intervention. All patients within a stroke service were analysed according to the intervention that that stroke service was randomised to. The ITT population was used for both primary and secondary analyses.

Per-protocol analysis was also undertaken in which major protocol violators or patients not receiving care from a SCC were excluded from the ITT population. The following groups of patients were excluded from the per-protocol population:

• non-eligible patients

• patients registered but then not referred to the SCC

• patients referred to the SCC at baseline but who were not known to the SCC as a referred patient

• patients not receiving any care from the SCC for the following reasons: patient declined the SCC service, the SCC was unable to contact the patient and the patient was registered but died before the first SCC assessment.

As the trial was a cluster RCT, outcome measures at follow-up were compared between the intervention group and the control group using a two-level multilevel model, with patients nested within stroke services.

Both primary analysis and secondary analysis involved a two-level linear model. The models were adjusted for the patient-level covariates (level 1) [baseline Barthel Index score (pre and post stroke); gender; age; living circumstances (living alone vs. living with carer); stroke severity, reflected by speech and language impairment (normal/impaired) and baseline 6CIT score (normal/impaired), and patient baseline score for the outcome measure] and the following stroke unit-level covariates (level 2): quality of stroke unit (NSA score), referral rate and SCCs working alone compared with working within a community MDT.

Two-level linear models for analyses of carer outcomes were adjusted for carer covariates (level 1) (carer baseline score for the outcome measure, gender, carer education, patient baseline Barthel Index score) and stroke unit covariates (level 2) [quality of stroke unit (NSA score), referral rate and SCCs working alone vs. working within a community MDT].

No interim analyses were undertaken.

All losses to follow-up because of death, dropouts and loss of contact are fully reported.

Details of patient deaths and hospital readmissions, carer deaths, any serious adverse events and related and unexpected serious adverse events are reported for each treatment group.

Sensitivity analyses to examine the robustness of the conclusion of the primary analysis were employed. Each of these analyses was compared with the primary analysis separately. The following sensitivity analyses were performed:

• Including patients who died – patients who died were assumed to have a GHQ-12 score of 36 (worst possible outcome).

• Only including patients returning postal questionnaires at 6 months; patients who provided primary outcome data by telephone were excluded.

• Repeating analysis without proxy responses (the proxy response was when the whole questionnaire was completed on the patient’s behalf without consulting him or her) to assess the impact of proxy responses on the analysis of the primary end point.

• Using data collected at 12 months for patients who did not send their questionnaires back at 6 months.

• Accounting for all participants in the ITT population assuming data missing at random using multiple imputation. Information based on Edinburgh stroke case mix adjuster, clinical classification of stroke, patient status in terms of response compared with non-response (alive, died, too poorly) and prespecified model covariates was used to impute the missing outcomes. The analysis with imputed values was repeated using the same model as the model in the primary analysis.

The relationship between the adjusted primary outcome and compliance with care plans and completion of time logs was explored graphically.

Economic evaluation

Participant flow

The Consolidated Standards of Reporting Trials (CONSORT) flow diagram depicting the progress of participants and clusters through the phases of the trial, including assessment of eligibility of clusters, cluster randomisation, participant screening, recruitment and follow-up and the ITT populations, is provided in Figure 3, with additional detail given in Appendix 8 (see Tables 76–79).

FIGURE 3.

CONSORT flow diagram. PI, principal investigator.

In total, 32 clusters were randomised and, of these, 29 participated (14 control clusters and 15 intervention).

During screening, 2488 participants were screened for eligibility, 1347 in the control group and 1141 in the intervention group.

The numbers of registered patients were balanced between the two arms (399 control, 401 intervention). At all time points the number of clusters remained the same (14 control, 15 intervention). The numbers of registered carers were also balanced (100 carers were recruited in the control group and 108 in the intervention group).

Response rates for patient-reported outcomes at 6 months were 75.2% (300 patients) in the control group and 77.3% (310 patients) in the intervention group; these dropped to 67.2% (268 patients) and 70% (281 patients), respectively, at 12 months’ follow-up.

Response rates for carer-completed outcomes at 6 months were 88% (88 carers) in the control group and 82.4% (89 carers) in the intervention group; these dropped to 71% (71 carers) and 67.3% (73 carers), respectively, at 12 months’ follow-up.

Recruitment

Service eligibility and randomisation were undertaken in two phases. Fourteen services were randomised in phase 1 in January 2009. In phase 2, in September 2009, an additional 18 services were randomised. Following the training and implementation period (intervention services), participant recruitment began in July 2009 in phase 1 and in March 2010 in phase 2, which is reflected in the steep increase in monthly accrual (Figure 4). Overall, 800 patients were registered from July 2009 to March 2011, 399 patients from 14 control services and 401 patients from 15 intervention services. Follow-up was completed in May 2012.

FIGURE 4.

Accrual into the trial.

Baseline data

Numbers analysed

All analyses and data summaries were carried out using the ITT population. Per-protocol analysis was undertaken for analyses of primary and secondary end points. Numbers of patients and carers in each population are shown in Table 16.

Outcomes and estimation

Patient-reported outcomes

Patient-reported outcomes were collected at three time points: baseline, 6 months and 12 months.

Unadjusted scores

Summaries of unadjusted scores for primary and secondary patient-reported outcomes at baseline and 6 and 12 months are shown in Table 17. Baseline unadjusted scores are balanced between treatment arms.

TABLE 17 - Patient unadjusted questionnaire scores at different time points by arm

Questionnaire	Baseline, [mean (SD)]		6 months, [mean (SD)]		12 months, [mean (SD)]
	Control	Intervention	Control	Intervention	Control	Intervention
Barthel Index n	15.2 (4.48) 398	14.4 (5.06) 401	16.2 (4.24) 296	15.5 (4.48) 307	16.3 (3.91) 266	15.7 (4.26) 282
GHQ-12 n	15.2 (7.26) 396	15.9 (7.03) 397	15.3 (7.32) 305	16.2 (7.17) 318	14.5 (6.78) 268	14.3 (6.87) 281
EQ-5D n	0.56 (0.340) 382	0.51 (0.378) 381	0.61 (0.339) 288	0.56 (0.322) 301	0.61 (0.316) 259	0.54 (0.318) 270
FAI n	28.1 (9.71) 398	27.8 (9.70) 399	20.3 (11.41) 293	18.4 (11.27) 304	21.5 (11.30) 266	18.8 (11.53) 281

Adjusted scores

Two-level hierarchical modelling was used to compare the intervention and control groups, with patient-level covariates (baseline Barthel Index score, baseline GHQ-12 score, gender, age, living circumstances, stroke severity and baseline 6CIT score) being level 1 and service-level covariates (quality of stroke unit, referral rate, SCC working alone or in a team) being level 2. In all models we allowed for different variances between the treatment arms.

A summary of the adjusted scores, differences in scores with 95% CIs, p-values and adjusted ICCs for each treatment arm for the primary and secondary end points at 6 and 12 months is provided in Table 18.

TABLE 18 - Patient adjusted questionnaire scores at 6 and 12 months

SE, standard error.

Questionnaire	Control, mean (SE) n	Intervention, mean (SE) n	Difference (SE)	95% CI of the difference	p-value	Adjusted ICC
						Control	Intervention
6 months
GHQ-12	14.9 (0.60) 300	15.5 (0.60) 310	−0.6 (0.65)	−1.8 to 0.7	0.394	0.013	0.025
Barthel Index	15.8 (0.33) 296	15.3 (0.28) 307	0.5 (0.33)	−0.2 to 1.1	0.133	0.022	0.000
EQ-5D	0.58 (0.025) 288	0.55 (0.022) 301	0.03 (0.025)	−0.02 to 0.08	0.252	0.014	0.059
FAI	19.0 (0.76) 293	18.0 (0.76) 304	1.0 (0.80)	−0.6 to 2.5	0.229	0.000	0.014
12 months
GHQ-12	14.4 (0.58) 268	13.9 (0.72) 281	0.5 (0.73)	−0.9 to 2.0	0.454	0.000	0.063
Barthel Index	15.6 (0.36) 266	15.4 (0.30) 282	0.2 (0.37)	−0.5 to 0.9	0.585	0.049	0.023
EQ-5D	0.56 (0.030) 259	0.51 (0.028) 270	0.05 (0.033)	−0.02 to 0.11	0.167	0.050	0.044
FAI	20.2 (0.78) 266	18.7 (0.84) 281	1.5 (0.87)	−0.2 to 3.2	0.078	0.000	0.025

The primary and secondary analyses were based on a complete case analysis with no substitution for missing patient-level outcome data; numbers of patients analysed for each end point are shown in Table 18.

In the analysis of the primary end point, the adjusted GHQ-12 mean score at 6 months was 14.9 [standard error (SE) 0.6] in the control group and 15.5 (SE 0.6) in the intervention group, a difference of −0.6 points (95% CI −1.8 to 0.7 points), with a p-value of 0.394 and an adjusted ICC of 0.013 for the control group and 0.025 for the intervention group, indicating that there is no statistical evidence of a significant difference between the treatment groups in psychological outcomes for stroke patients living at home at 6 months, as measured by the GHQ-12.

Results for secondary end points (Barthel Index, EQ-5D and FAI at 6 and 12 months and GHQ-12 at 12 months; see Table 18) also indicate that there are no statistically significant differences between the treatment groups for any of the secondary end points.

Longer-term Unmet Needs after Stroke questionnaire

Summaries of the patient-completed LUNS questionnaire at 6 and 12 months, showing the number of unmet needs out of a possible 22, are provided in Tables 19 and 20 respectively. The numbers and types of unmet needs reported were similar for the control group and the intervention group at 6 and 12 months. A median of 2.5 unmet needs were reported, and the prevalence of individual unmet needs ranged from 2% to 51% of those completing the questionnaire (Figure 5). The most commonly reported unmet need was for more information; this was followed by unmet needs in the areas of falls, memory/concentration, accessible holidays and pain, all reported by over one-fifth of responders.

TABLE 19 - Longer-term Unmet Needs after Stroke questionnaire: number of unmet needs for patients at 6 months

No. of unmet needs at 6 months	Control (N = 399), n (%)	Intervention (N = 401), n (%)	Total (N = 800), n (%)
0	57 (14.3)	43 (10.7)	100 (12.5)
1	46 (11.5)	46 (11.5)	92 (11.5)
2	44 (11.0)	41 (10.2)	85 (10.6)
3	25 (6.3)	39 (9.7)	64 (8.0)
4	28 (7.0)	31 (7.7)	59 (7.4)
5	15 (3.8)	16 (4.0)	31 (3.9)
6	18 (4.5)	17 (4.2)	35 (4.4)
7	15 (3.8)	18 (4.5)	33 (4.1)
8	13 (3.3)	17 (4.2)	30 (3.8)
9	8 (2.0)	8 (2.0)	16 (2.0)
10	3 (0.8)	10 (2.5)	13 (1.6)
11	3 (0.8)	5 (1.2)	8 (1.0)
12	7 (1.8)	1 (0.2)	8 (1.0)
13	4 (1.0)	7 (1.7)	11 (1.4)
14	2 (0.5)	1 (0.2)	3 (0.4)
15	1 (0.3)	1 (0.2)	2 (0.3)
16	0 (0.0)	1 (0.2)	1 (0.1)
17	1 (0.3)	0 (0.0)	1 (0.1)
18	0 (0.0)	1 (0.2)	1 (0.1)
Missing	109 (27.3)	98 (24.4)	207 (25.9)

TABLE 20 - Longer-term Unmet Needs after Stroke questionnaire: number of unmet needs for patients at 12 months

No. of unmet needs at 12 months	Control (N = 399), n (%)	Intervention (N = 401), n (%)	Total (N = 800), n (%)
0	62 (15.5)	62 (15.5)	124 (15.5)
1	35 (8.8)	36 (9.0)	71 (8.9)
2	34 (8.5)	43 (10.7)	77 (9.6)
3	29 (7.3)	31 (7.7)	60 (7.5)
4	20 (5.0)	28 (7.0)	48 (6.0)
5	21 (5.3)	11 (2.7)	32 (4.0)
6	11 (2.8)	12 (3.0)	23 (2.9)
7	12 (3.0)	19 (4.7)	31 (3.9)
8	10 (2.5)	8 (2.0)	18 (2.3)
9	10 (2.5)	7 (1.7)	17 (2.1)
10	6 (1.5)	7 (1.7)	13 (1.6)
11	4 (1.0)	3 (0.7)	7 (0.9)
12	2 (0.5)	6 (1.5)	8 (1.0)
13	3 (0.8)	2 (0.5)	5 (0.6)
14	1 (0.3)	1 (0.2)	2 (0.3)
15	1 (0.3)	0 (0.0)	1 (0.1)
18	1 (0.3)	0 (0.0)	1 (0.1)
21	0 (0.0)	1 (0.2)	1 (0.1)
Missing	137 (34.3)	124 (30.9)	261 (32.6)

FIGURE 5.

Prevalence of individual unmet needs: (a) 6 months; and (b) 12 months.

Ancillary analyses

Adverse events

Deaths, hospital readmissions, institutionalisation and treatment on an emergency outpatient basis were reported at 6 and 12 months’ follow-up by the clinical research team using health and social care records.

Deaths

Throughout the trial, no safety issues were highlighted. Overall, 67 patient deaths were reported, 32 in the control group and 35 in the intervention group; most of these occurred in hospital and within the first 6 months of registration, as shown in Table 26 and Figure 6.

TABLE 26 - Total patient deaths and place of death

Summary of all deaths up to 9 months from registration (end of 6-month follow-up).

These deaths are not included in the total death count or place of death as they were reported after the follow-up period.

Time/Location of death	Control (N = 399), n (%)	Intervention (N = 401), n (%)	Total (N = 800), n (%)
Deaths
6 monthsa	27 (6.8)	31 (7.7)	58 (7.3)
12 months	5 (1.3)	4 (1.0)	9 (1.1)
After 12-month follow-upb	2 (0.5)	2 (0.5)	4 (0.5)
Total	32 (8.0)	35 (8.7)	67 (8.4)
Place of death
Home	2 (0.5)	4 (1.0)	6 (0.8)
Hospital	23 (5.8)	19 (4.7)	42 (5.3)
Nursing home	3 (0.8)	3 (0.7)	6 (0.8)
Unknown	3 (0.8)	7 (1.7)	10 (1.3)
Other	1 (0.3)	2 (0.5)	3 (0.4)

FIGURE 6.

Time between registration and patient death.

One carer death was reported in the control group, which occurred 7 days after the carer was registered to the trial.

Withdrawals

Withdrawals of patients and carers from the trial are shown in Tables 29 and 30 respectively. Overall, there were 49 withdrawals among the patients and more patients and carers withdrew in the control group than in the intervention group (35 patients and 8 carers in the control group and 14 patients and two carers in the intervention group).

Process data: implementation of control and intervention services

Completion of process data

Control services were asked to complete a time log for all patients summarising the contacts. Intervention services were asked to complete the care plan for all contacts with all patients. Table 31 shows the completion of the time logs and Table 32 shows the completion of the care plans. Of patients registered to the trial, 314 (78.7%) and 318 (79.3%) received the SCC service in the control and intervention arms respectively. Reasons for not receiving the service include the SCC not receiving a referral, the patient declining to receive the service, patient death or patient not contactable. SCCs completed time logs for 207 (51.9%) patients in the control arm and completed care plans for 283 (70.6%) patients in the intervention arm. Of the 283 completed care plans, three had all contacts completed > 12 months after patient registration to the trial.

TABLE 31 - Completion of time logs

Service	Registered, n (%)	Time log completed		Reasons not completed
		Yes, n (%)	No, n (%)	Referral not received by SCC, n (%)	Patient declined SCC service, n (%)	Patient died before receiving SCC service, n (%)	SCC unable to contact patient, n (%)	Patient withdrew from all follow-up, n (%)	Received SCC service, no time log completed, n (%)
Total	399 (100.0)	207 (51.9)	192 (48.1)	67 (34.9)	9 (4.7)	3 (1.6)	2 (1.0)	4 (2.1)	107 (55.7)
S001	15 (3.8)	8 (53.3)	7 (46.7)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	7 (100.0)
S003	18 (4.5)	8 (44.4)	10 (55.6)	4 (40.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (10.0)	5 (50.0)
S004	32 (8.0)	20 (62.5)	12 (37.5)	1 (8.3)	1 (8.3)	1 (8.3)	0 (0.0)	1 (8.3)	8 (66.7)
S008	25 (6.3)	22 (88.0)	3 (12.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	3 (100.0)
S010	17 (4.3)	13 (76.5)	4 (23.5)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (25.0)	3 (75.0)
S011	45 (11.3)	18 (40.0)	27 (60.0)	13 (48.1)	7 (25.9)	0 (0.0)	0 (0.0)	0 (0.0)	7 (25.9)
S014	29 (7.3)	5 (17.2)	24 (82.8)	7 (29.2)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	17 (70.8)
S015	46 (11.5)	35 (76.1)	11 (23.9)	1 (9.1)	0 (0.0)	1 (9.1)	1 (9.1)	0 (0.0)	8 (72.7)
S018	35 (8.8)	4 (11.4)	31 (88.6)	14 (45.2)	1 (3.2)	0 (0.0)	1 (3.2)	0 (0.0)	15 (48.4)
S019	45 (11.3)	33 (73.3)	12 (26.7)	4 (33.3)	0 (0.0)	1 (8.3)	0 (0.0)	0 (0.0)	7 (58.3)
S022	19 (4.8)	3 (15.8)	16 (84.2)	13 (81.3)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	3 (18.8)
S025	18 (4.5)	7 (38.9)	11 (61.1)	4 (36.4)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	7 (63.6)
S029	41 (10.3)	25 (61.0)	16 (39.0)	6 (37.5)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	10 (62.5)
S032	14 (3.5)	6 (42.9)	8 (57.1)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (12.5)	7 (87.5)

TABLE 32 - Completion of care plans

N/A, not applicable.

One care plan with all contacts completed > 12 months after registration.

Two care plans with all contacts completed > 12 months after registration.

Service	Registered, n (%)	Care plan completed		Reasons not completed
		Yes, n (%)	No, n (%)	Referral not received by SCC, n (%)	Patient declined SCC service, n (%)	Patient died before receiving SCC service, n (%)	SCC unable to contact patient, n (%)	Patient withdrew from all follow-up, n (%)	Received SCC service, no care plan completed, n (%)
Total	401 (100.0)	283 (70.6)	118 (29.4)	45 (38.1)	21 (17.8)	14 (11.9)	2 (1.7)	1 (0.8)	35 (29.7)
S002	27 (6.7)	20 (74.1)	7 (25.9)	0 (0.0)	0 (0.0)	2 (28.6)	0 (0.0)	1 (14.3)	4 (57.1)
S005	19 (4.7)	15 (78.9)	4 (21.1)	0 (0.0)	1 (25.0)	0 (0.0)	0 (0.0)	0 (0.0)	3 (75.0)
S006	30 (7.5)	26 (86.7)	4 (13.3)	3 (75.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (25.0)
S007	31 (7.7)	28 (90.3)	3 (9.7)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	3 (100.0)
S009	17 (4.2)	16 (94.1)	1 (5.9)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (100.0)
S012	45 (11.2)	25a (55.6)	20 (44.4)	8 (40.0)	4 (20.0)	2 (10.0)	0 (0.0)	0 (0.0)	6 (30.0)
S013	45 (11.2)	33b (73.3)	12 (26.7)	8 (66.7)	1 (8.3)	2 (16.7)	0 (0.0)	0 (0.0)	1 (8.3)
S016	27 (6.7)	19 (70.4)	8 (29.6)	2 (25.0)	1 (12.5)	4 (50.0)	1 (12.5)	0 (0.0)	0 (0.0)
S020	2 (0.5)	1 (50.0)	1 (50.0)	0 (0.0)	1 (100.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
S023	13 (3.2)	11 (84.6)	2 (15.4)	0 (0.0)	1 (50.0)	1 (50.0)	0 (0.0)	0 (0.0)	0 (0.0)
S024	27 (6.7)	10 (37.0)	17 (63.0)	4 (23.5)	1 (5.9)	0 (0.0)	0 (0.0)	0 (0.0)	12 (70.6)
S026	45 (11.2)	29 (64.4)	16 (35.6)	8 (50.0)	5 (31.3)	0 (0.0)	1 (6.3)	0 (0.0)	2 (12.5)
S027	26 (6.5)	26 (100.0)	0 (0.0)	N/A	N/A	N/A	N/A	N/A	N/A
S030	29 (7.2)	10 (34.5)	19 (65.5)	10 (52.6)	5 (26.3)	3 (15.8)	0 (0.0)	0 (0.0)	1 (5.3)
S031	18 (4.5)	14 (77.8)	4 (22.2)	2 (50.0)	1 (25.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (25.0)

For control services, the relationship between the adjusted mean primary outcome and the mean percentage of completed time logs (of registered patients) by service is shown in Figure 7. There is no linear trend between the percentage of completed time logs observed and services’ adjusted means.

FIGURE 7.

Adjusted differences of each service from adjusted control mean in GHQ-12 scores ordered by service completion of time logs: control.

Compliance with the intervention

Compliance with the intervention assessed from completed care plans is shown by service in Table 33. Compliance with the intervention was defined as at least 12 of 16 (75%) assessment areas (each linked to a reference guide in the manual) asked on patient contact 1 of the care plan (with at least one question in the assessment area being asked). In total, 10 care plans were not compliant and 269 (96.1%) met the definition of compliance.

TABLE 33 - Compliance with completion of care plans

Only the front page of the care plan was completed; no assessment contacts were used.

Service	Compliant record, n (%)	Non-compliant record, n (%)	Missing, n (%)	Care plan completed, n (%)
S002	19 (95.0)	0 (0.0)	1a (5.0)	20 (7.1)
S005	15 (100.0)	0 (0.0)	0 (0.0)	15 (5.4)
S006	25 (96.2)	1 (3.8)	0 (0.0)	26 (9.3)
S007	28 (100.0)	0 (0.0)	0 (0.0)	28 (10.0)
S009	16 (100.0)	0 (0.0)	0 (0.0)	16 (5.7)
S012	24 (100.0)	0 (0.0)	0 (0.0)	24 (8.6)
S013	31 (100.0)	0 (0.0)	0 (0.0)	31 (11.1)
S016	18 (94.7)	1 (5.3)	0 (0.0)	19 (6.8)
S020	1 (100.0)	0 (0.0)	0 (0.0)	1 (0.4)
S023	11 (100.0)	0 (0.0)	0 (0.0)	11 (3.9)
S024	10 (100.0)	0 (0.0)	0 (0.0)	10 (3.6)
S026	28 (96.6)	1 (3.4)	0 (0.0)	29 (10.4)
S027	20 (76.9)	6 (23.1)	0 (0.0)	26 (9.3)
S030	9 (90.0)	1 (10.0)	0 (0.0)	10 (3.6)
S031	14 (100.0)	0 (0.0)	0 (0.0)	14 (5.0)
Total	269 (96.1)	10 (3.6)	1 (0.4)	280 (100.0)

For intervention services, the relationship between the adjusted mean primary outcome and the mean percentage of compliant care plans (of registered patients) by service is shown in Figure 8. There is larger variation in adjusted scores for services with higher compliance, but no linear trend is observed between the percentage of compliant care plans and services’ adjusted mean scores.

FIGURE 8.

Adjusted differences of each service from adjusted intervention mean in GHQ-12 scores ordered by service compliance with care plans: intervention.

Economic evaluation

Cost-effectiveness and cost–utility

Of the 12 cost–outcome combinations examined for the cost-effectiveness and cost–utility analyses, none suggested that there were statistically significant between-group differences for both cost and outcome elements. Therefore, we have not calculated ICERs.

Cost-effectiveness planes (Figures 9 and 10) show that differences in health and social care costs between the groups at 6 months are fairly strongly centred around zero, that is, no difference, whereas outcome differences suggest a disadvantage for the intervention group in terms of less improvement on the GHQ-12 and QALY losses rather than gains. This conclusion is further represented in Figures 11 and 12, which show the probabilities that the intervention group is cost-effective compared with the control group. The probabilities of cost-effectiveness were low based on both outcome measures, not exceeding 30% for the threshold ranges examined. Probabilities based on the societal perspective were particularly low as these costs are dominated by informal care costs, which were significantly higher in the intervention group. Therefore, the intervention is unlikely to be cost-effective over a 6-month period at current policy thresholds of £20,000–30,000 per QALY gain. It is unclear what the willingness to pay for a GHQ-12 point improvement would be in practice.

FIGURE 9.

Cost-effectiveness plane of incremental total health and social care costs and point changes on the GHQ-12 at 6 months. Note: GHQ-12 scores have been reversed so that a positive difference is an improvement.

FIGURE 10.

Cost-effectiveness plane of incremental total health and social care costs and QALYs at 6 months.

FIGURE 11.

Probability that the intervention is cost-effective compared with the control at 6 months, from each cost perspective, for a range of willingness-to-pay values for an additional point improvement on the GHQ-12.

FIGURE 12.

Probability that the intervention is cost-effective compared with the control at 6 months, from each cost perspective, for a range of willingness-to-pay values for an additional QALY.

Key findings

This trial was a pragmatic multicentre cluster RCT of a complex intervention. The trial was designed to evaluate a newly developed system of care compared with usual practice when implemented by stroke services, specifically SCCs, across the UK. The system of care was predicated on patients’ and carers’ needs and was developed through systematic reviews and the views of patients and carers, with input from primary health-care providers. The intervention was manualised and the treatment algorithms were developed through robust, evidence-based methodologies.

Services randomised to the intervention were trained in using the system of care whereas control services continued with usual practice. Patients (and optionally carers) were recruited on referral to the service and patient and carer outcomes were assessed using postal questionnaires at 6 and 12 months post recruitment. The primary outcome was patient psychological health (GHQ-12) at 6 months and secondary outcomes included patient functional health, carer psychological and functional health, and cost-effectiveness.

In total, 800 patients and 208 carers were recruited from 29 stroke services in England, Scotland, Northern Ireland and Wales. SCCs in 14 services provided the intervention whereas those in 15 services continued with usual practice. Patient groups were well matched between the two arms in demographic and clinical characteristics. There was no evidence of statistically significant differences between the arms in primary or secondary outcomes using either ITT analysis or per-protocol analysis. There was no difference in adverse events between the two groups.

The costs of SCC inputs (mean difference £42; 95% CI – £30 to £116) and total health and social care costs at 6 months, 12 months and over 1 year were similar between the groups. Societal costs were higher in the intervention group (mean difference at 6 months £1163; 95% CI £56 to £3271) because of greater use of informal care. There was no difference in QALYs between the groups and the probability of the intervention being cost-effective at 6 months was low at the current policy threshold.

Strengths and limitations of the study

This trial is one of the largest stroke rehabilitation trials completed to date (worldwide). It successfully recruited to time and target and involved centres from a wide geographical area. Implementation was supported by the National Institute for Health Research (NIHR) SRN and demonstrates the importance of the network in facilitating set-up and recruitment.

Generalisability

Including a wide range of disparate geographical regions ensured a good representation of different health-care settings. The eligibility criteria were kept to a minimum, in keeping with the pragmatic trial design, to ensure that a stroke patient population representative of referrals to SCCs was recruited, including patients with language and cognitive impairments. Of those patients eligible for the trial, 56% were recruited. In total, 29% of patients recruited had a cognitive impairment and 9% were unable to complete the cognitive test as a result of communication difficulties.

Context

Although the role of the SCC had been recommended in the National Service Framework for Older People,16 our pre-trial survey revealed that relatively few services had adopted this model and even fewer used a dedicated SCC role. It might be suggested therefore that the services that were recruited to the trial were particularly forward thinking and slightly atypical compared with other services across the UK. Since this programme grant was awarded, in 2007, there have been considerable changes in stroke service provision across the UK. Our system of care is very much in keeping with current developments. Similar approaches have been developed (e.g. GM-SAT for 6-month reviews34). This indicates that our thinking is feasible, practical and appropriate for ongoing service development. However, because of the increased awareness of the longer-term needs of stroke patients and carers, as promoted through the work of the Stroke Improvement Programme, the SCCs in the control arm may have adapted their practice accordingly. However, none of the control SCCs reported using structured review tools in the survey and interview that they completed after the end of the trial.

It would seem that longer-term outcomes for patients after stroke do remain quite poor. A Stroke Association unmet needs survey4 demonstrated the range and depth of unmet needs. Our own work reported similar outcomes (see Chapter 4).

The trial was not designed to evaluate whether the SCC service was effective but whether the structured patient-centred system of care improved patient outcomes. Our hypothesis was that the system of care configured around patient needs would be more holistic and more responsive in addressing the well-recognised unmet needs of patients after stroke.

Further analysis is ongoing to determine whether some subsets of patients, possibly over selected time periods, did have an enhanced outcome from the intervention approach, or whether service/process factors contributed to differences in outcomes. A realist evaluation of the trial has also been undertaken,88 which may shed further light on how the SCCs have implemented the system of care.

The trial results apply to a heterogeneous population. It might be that the SCC service should be targeted more towards patients (and carers) with specific needs, leading to a more specialised bespoke service.

Aim

The study aim was to finalise the development of the LUNS questionnaire and evaluate the acceptability, test–retest reliability and validity of the resulting 22-item questionnaire for measuring longer-term unmet needs of patients living at home post stroke.

Methods

Patients with a primary diagnosis of stroke being discharged to their own home or that of a carer, after a minimum 3-day hospital stay, were recruited from 40 stroke units across England. A questionnaire pack including the LUNS questionnaire, GHQ-12, FAI and Short Form questionnaire-12 item (SF-12) was posted to participants 3 or 6 months after stroke to assess acceptability and validity. The LUNS questionnaire was re-sent 1 week after receiving back the first pack to assess test–retest reliability. Semistructured interviews were conducted with participants who reported low unmet need on the LUNS questionnaire to explore whether they genuinely considered this to be the case.

Results

In total, 850 patients, including 199 (23%) with impaired communication and/or cognition, were recruited. A total of 529 pack 1s were returned (69% of those sent), with 3.5% missing LUNS items, comparable with the concurrent measures. The median number of unmet needs was four (range 0–19) and participants reporting low unmet need verified this to be the case and that they did not have other unmet needs not captured by the LUNS questionnaire. Identification of an unmet need was consistently associated with poorer outcomes on concurrent measures. The test–retest reliability of the LUNS questionnaire (n = 326) was moderate to good (77.9–98.8% item agreement, κ = 0.45–0.67).

Conclusions

The LUNS questionnaire shows suitable acceptability, test–retest reliability and validity for identifying longer-term unmet needs after stroke.

Background

Stroke is the leading cause of complex disability in adults. 89 Research into the longer-term experiences of stroke survivors has found that they experience a range of difficulties, including, for example, incontinence, depression, lack of information, pain, and restrictions in social and household activities. 5,90–93 Many of these problems are highly prevalent and exist even a decade after stroke,94 suggesting that they are not appropriately identified or addressed by service providers. To address the diverse and complex difficulties that stroke survivors face, effective methods are required to identify their unmet needs. Regular patient reviews, as part of an integrated longer-term stroke pathway, are recommended2,15 and could incorporate needs assessment. Measuring unmet need is also a way of identifying stroke service deficiencies95 and this may enable effective commissioning and provision of services. Assessment of needs can be defined as the ability to benefit from health (or social care)96 and unmet needs may be defined as ‘expressed needs that are not satisfied by current service provision’ (p. 1052). 97

Existing instruments to measure problems after stroke, such as the FAI,98 tend to deal with limited aspects of the stroke experience and therefore do not capture the range of domains pertinent to longer-term life after stroke. A recent survey in the UK identified the unmet needs of stroke survivors but the survey instrument was not evaluated for reliability. 4 Tools designed for measuring unmet needs in people recovering from stroke have been developed for research purposes but are long and have poor response rates. 99–101 There is a need for a short, easy-to-complete tool for identifying unmet needs after stroke.

Development of the Longer-term Unmet Needs after Stroke questionnaire

We developed a 22-item questionnaire (termed ‘LUNS’) to provide a method for identifying the longer-term unmet needs of stroke survivors. A multistage iterative process was used to develop, refine and carry out preliminary testing of the LUNS questionnaire, summarised in Figure 13 and described in detail below.

FIGURE 13.

Tool development summary.

Items for the LUNS tool were identified through systematic reviewing of qualitative and quantitative literature reporting the longer-term stroke experience24,25 and, subsequently, semistructured interviews with stroke patients and carers. 26 The data were represented in a framework comprising 92 longer-term stroke problem types organised into six domains (transfer of care, information, services, social and emotional consequences, caring role, and health problems and related areas).

This framework was used as a basis for constructing an initial series of 48 statements representing patient unmet needs (items relating to the domain of caring role were excluded). Through internal peer review, similar items were consolidated (such as bathing and dental into personal care) and statements were removed, leaving 28 items. The 28-item LUNS tool underwent small-scale reliability testing with 29 stroke patients, revealing that 12 items had poor test–retest reliability. Further modification to each item within the tool was facilitated by discussion with members of two stroke clubs, a patient research advisory group and internal peer review. This resulted in a 21-item LUNS tool.

Objectives

The primary objective of the LUNS study was to evaluate the acceptability, test–retest reliability and validity of the LUNS questionnaire for identifying longer-term unmet needs of patients living at home post stroke.

Secondary objectives were to:

• evaluate whether the LUNS questionnaire has suitable properties to be used to give a score for measuring the level of unmet need (internal consistency, dimensionality)

• explore perceptions of unmet need using qualitative methods.

Study design

The LUNS questionnaire was evaluated in a large multicentre study. It was administered by post at 3–6 months post stroke in a self-completion questionnaire pack of the same style as that used in the pilot study. The pack contained the LUNS questionnaire and four validated questionnaires on health status, which were used as concurrent measures to assess the validity of the LUNS questionnaire (see Data collection). There are no measures of unmet need capable of capturing broad life areas in this population with which we could evaluate the concurrent validity of the LUNS tool; therefore, the LUNS tool was compared with health measures that could reflect the level of need. The LUNS questionnaire was also administered at a second time point to assess test–retest reliability.

The study was conducted in two phases. Factors such as cognitive impairment and unfamiliarity with the language may affect ability to accurately complete a questionnaire. To assess properties of the LUNS tool without interference from these factors, in phase 1 a selected population of English-speaking patients without communication or cognitive impairment was selected. Interim analysis of phase 1 data was conducted to ensure adequate acceptability, test–retest reliability and validity of the tool in this optimised group before proceeding to phase 2. To ensure inclusion of a sample with a broad range of stroke impairments in phase 2, we recruited patients who had spent longer in hospital (likely to be more disabled) and set a target of 50% of patients with cognitive or communication impairment; patients who did not speak English were also included in this group because it was expected that they would have difficulty completing an English-language questionnaire. This target was reduced to 40% during recruitment to meet target timelines.

Participants

Patients were recruited by SRN staff from stroke units across England. Eligibility criteria were purposefully broad: a diagnosed new stroke, age ≥ 16 years and patient being discharged to his or her own home or that of his or her carer. Patients were not eligible if concurrent illness required palliative care or if permanent discharge to a nursing or residential home was planned.

In addition to the above, patients were eligible for phase 1 only if they were English speaking, had no cognitive or communication impairment and had undergone a minimum hospital stay of 72 hours. In the second phase, patients who were unable to read and understand English were recruited if they had an English-speaking carer, patients with a communication or cognitive impairment were included and the minimum hospital stay was increased to 14 days.

With verbal consent, cognition and communication were assessed in both phases using 6CIT65 or the Frenchay Aphasia Screening Test (FAST)105 respectively. Patients who scored ≥ 8 on 6CIT were considered to have cognitive impairment and patients who scored ≤ 6 for FAST comprehension or ≤ 3 for FAST reading were considered to have communication impairment. The group assessed as having no communication or cognitive impairment and who are English speaking are described as the ‘no impairment group’ whereas those assessed as having a cognitive or communication impairment or who are not English speaking are described as the ‘impairment group’ (not speaking English could be considered an impairment in relation to completing an English-language questionnaire).

Written informed consent was obtained from patients before collection of baseline data or, when applicable in the second phase, a relative or friend of a patient lacking capacity was asked to consider the patient’s wishes and provide a consultee declaration.

Recruited patients who became ineligible (e.g. moved to a nursing home) were withdrawn from follow-up and analysis.

Data collection

Baseline data were collected on the ward by SRN staff. Demographic information, stroke details, pre- and post-stroke Barthel Index scores38 and comorbidities106 were recorded.

After confirming patients’ contact details and survival status through GP or NHS database records, the LUNS tool was posted to the participants’ home address in a self-completion questionnaire pack. The first pack (pack 1) contained the LUNS tool and the following validated and reliable measures:

• Impairment Manikin107 – This asks whether the respondent has completely recovered from stroke and includes a diagram of a body on which patients indicate ongoing problems, for example difficulties with their leg or vision.

• GHQ-1262 – A measure of emotional health and well-being. A higher score indicates a poorer level of emotional well-being.

• FAI98 – A measure of extended activities of daily living. The FAI is scored from 0 to 45 with a higher score indicating a higher activity level.

• SF-12108 – A measure of health-related quality of life. The two component scores of the SF-12 [mental component score (MCS) and physical component score (PCS)] were used. Scores are standardised to a 0–100 scale, with 50 representing the norm value in the general population, < 50 indicating health worse than the norm and > 50 indicating health better than the norm.

The second pack (pack 2) included the LUNS and SF-12 tools. Both packs included a question asking whether any help was needed with completion of the questionnaires and pack 2 included a question asking whether the patient had experienced any health changes since completing pack 1. Proxy responses were allowed.

In phase 1, pack 1 was posted to study participants at 3 or 6 months post stroke (dependent on their date of recruitment into the study). This was to enable timely completion of data collection and the interim analysis. In phase 2 all pack 1s were posted at 6 months post stroke when patients were more likely to be home from hospital (considering the increased length of stay and the possibility of cognitive or communication impairment). Using these time points allowed patients a period of time living at home so that they could understand their longer-term needs. Patients were asked to complete the pack and return it to the research team using a prepaid addressed envelope. Patients who had not returned their pack to the study team within 2 weeks were telephoned to check receipt and if necessary the pack was re-sent. If questionnaires were missing < 50% of the data, patients were telephoned to complete the missing data. Questionnaires missing ≥ 50% of the data were re-sent. To assess test–retest reliability, pack 2 was sent approximately 1 week after the fully completed pack 1 was received, or if missing data in pack 1 had not been returned within 2 weeks.

Sample size

A sample size of 350 was required in phase 1 for the interim analysis evaluating whether the LUNS tool was suitable for this optimised group. The sample size in phase 2 was 500 to give a total sample size of 850 to assess the full psychometric properties of the LUNS tool. For the psychometric Rasch analysis,109 a sample size of 243 is required to generate 99% confidence that item calibrations are stable within half a logit, even with poor scale targeting. 106 In phase 1, with an estimated response rate of 70%, a sample size of 350 would generate 245 cases, which is adequate. Phase 2 targeted a more impaired sample and so a lower response rate was anticipated. A response rate of 50% was estimated within this target population, meaning that a sample size of 500 would generate 250 cases. Phase 1 and phase 2 data can also be pooled to create a more powerful analysis.

Statistical methods

Baseline data were summarised using descriptive statistics and compared using Mann–Whitney U-test, chi-squared and independent t-tests for responders and non-responders to pack 1. The number and types of unmet needs reported in pack 1 were summarised and individual unmet needs were compared between the impairment group and the no-impairment group using Pearson’s chi-squared test.

The acceptability of the LUNS tool was determined by the percentage of pack 1s returned and the amount of non-completed items on initial receipt.

The test–retest reliability of the LUNS tool was assessed by comparing the percentage agreement for pack 1 and pack 2 and kappa statistics for individual items. 111 SF-12 test–retest results were calculated to provide context to the LUNS results. Patients who identified a change in their health status between packs 1 and 2 were excluded from the test–retest reliability assessment. The strength of agreement for kappa was defined as ranging from poor (κ = 0–0.20) to fair (κ = 0.21–0.40), moderate (κ = 0.41–0.60), good (κ = 0.61–0.80) and to very good (κ = 0.81–1). 112

The Mann–Whitney U-test was used to test differences in health status (GHQ-12, FAI, SF-12) between the group who identified unmet need and the group who identified no unmet need on each LUNS item. FAI total scores were calculated only for patients who were sent pack 1 at 6 months post stroke because of the inclusion of questions regarding the preceding 3 and 6 months.

Whether the LUNS questionnaire measures a single underlying dimension or multiple separate dimensions was investigated. Factor analysis allows us to produce an internally reliable scale by identifying the items that most correlate with each other. 113 It was first hypothesised that the LUNS tool would fit one dimension and would be tested using confirmatory factor analysis; should this fail, exploratory factor analysis would be used to explore potential dimensions. Factor analysis was carried out using MPlus computer software (version 6; Muthén & Muthén, Los Angeles, CA, USA).

A Cronbach’s coefficient alpha value114 was also calculated for the item set. Cronbach’s coefficient alpha is referred to as a measure of internal consistency reliability and is the proportion of variance in a set of scores that can be attributed to a common influence on the scores of the individual items. 115 However, it has also been argued that alpha is unrelated to the internal structure of the test and should therefore be viewed with caution. 116

Additionally, Rasch methods109 were used to formally test the LUNS tool against a unidimensional measurement model. Rasch analysis provides a unified framework to assess a number of measurement characteristics of individual items, as well as the total scale. Rasch analysis was carried out using RUMM2030 computer software (RUMM Laboratory, Perth, WA, Australia). Full details of Rasch analysis and the various tests of fit available can be found elsewhere. 117,118 Briefly, we examined individual person and item fit, with fit residuals between ± 2.5 displaying adequate fit. Chi-squared and analysis of variance (ANOVA) item fit statistics were also used to identify measurement anomalies. Differential item functioning (DIF) was examined to identify any bias by sample subgroup, considering age, gender, living arrangements (alone/with someone) and cognitive status (impairment or no impairment). Response dependency occurs when the response to one item has a direct influence on the response to another item, indicated by a positive residual correlation > 0.2. Targeting was considered to assess the relative distributions of the item and person locations. The person separation index (PSI) and Cronbach’s alpha statistics were used to assess the internal consistency (reliability) of the LUNS questionnaire.

Qualitative evaluation of unmet needs

During phase 2, a substudy was designed to further understand whether stroke patients who self-reported having no unmet needs on the LUNS questionnaire genuinely feel that they have no unmet needs or whether their unmet needs were missed by the questionnaire. Purposive sampling selected people who self-reported no (0) or low-level (1) unmet needs within a single region (Yorkshire) and semistructured interviews were conducted. A topic guide was written to ask about current problems people may have, how these are managed and how they may be identified as unmet needs. Data analysis used a thematic approach, working through the data until a small number of themes were identified that sufficiently describe the data.

Participant flow

A total of 350 participants were recruited from 29 sites in phase 1. In total, 237 pack 1s and 200 pack 2s were returned and included in the interim analysis. Interim analysis of phase 1 data found that the LUNS tool demonstrated adequate acceptability, test–retest reliability and validity to progress to phase 2 of the study. In phase 2, 500 participants were recruited from 40 sites. All results are reported for the combined study population of 850. Recruitment, questionnaire returns, losses and withdrawals for the combined phase 1 and phase 2 study population are summarised in Figure 14.

FIGURE 14.

Study flow chart. Reproduced with permission from LoTS care LUNS study team. 119

Recruitment

Recruitment to phase 1 was conducted between 11 December 2008 and 1 May 2009. Phase 1 follow-up was completed on 26 September 2012. Recruitment to phase 2 was conducted between 1 February 2010 and 10 January 2011. Follow-up was completed on 29 June 2011.

Baseline data

Of the 850 patients recruited, 199 (23%) had a communication or cognitive impairment or did not speak English. The 6CIT identified cognitive impairments in 138 patients, the FAST identified communication difficulties in 56 patients, three people did not speak English and two people could not provide a 6CIT or FAST score but were assessed as having a cognitive impairment. The median age of the population was 73 years and the median time in hospital was 12 days in phase 1 and 42 days in phase 2. Comorbidities are shown in Table 49 and additional baseline characteristics are shown in Table 50.

Table 50 also shows the baseline demographic and clinical characteristics of patients returning or not returning questionnaire pack 1 (responders and non-responders respectively). Overall, responders had a lower level of cognitive and physical impairment than non-responders. Significant differences (p < 0.01) were observed for number of previous strokes, ethnicity, communication or cognitive impairment, and post-stroke Barthel Index score. No other significant differences were found when comparing responders with non-responders.

The baseline characteristics of the no-impairment and impairment groups were broadly similar, except that the impairment group was slightly older, had a longer hospital stay and was more disabled according to the Barthel Index score; in addition, a greater proportion of the impairment group had suffered a haemorrhage or had right hemiparesis (Table 51).

Numbers analysed

In total, 529 pack 1s were returned. Of these, 20 were excluded because of missing LUNS items and 509 were included to assess acceptability, for comparison with concurrent measures, for calculation of Cronbach’s alpha and for factor analysis (423 in the no-impairment group and 86 in the impairment group). Of the pack 2s, 460 were returned and, of these, 124 were excluded because of a reported change in health status and 336 were included in the test–retest analysis (275 in the no-impairment group and 61 in the impairment group).

Rasch analysis was based on the responses of 449 participants (phases 1 and 2 are included but only the pack 1 data from each phase). All 529 response sets were read into the analysis but one person had no LUNS data (all missing) and 79 persons were removed from the analysis as they were judged to be ‘extreme’. An extreme person is someone who is either at the floor or the ceiling of the scale, meaning that they have scored either maximally or minimally on the scale. Extreme patients are removed from the analysis as they do not offer any information; they are beyond the measurement parameters of the scale.

In this case, all of the extreme patients were at the floor of the scale, meaning that they had no unmet needs.

Response to questionnaire packs

Packs were sent out for completion at 3 or 6 months post stroke, with the median time between stroke and completion 3 months (range 3–6 months) (n = 115) and 6 months (range 5–10 months) (n = 414) respectively.

Of all patients who returned pack 1, 213 (40%) reported that they had received some kind of help to complete the pack. The majority (68–76%) of those receiving help reported help with ticking boxes or reading or discussing the questions; in addition, 14 (6%) reported that someone had completed the pack on their behalf without consulting them (proxy completion) and two (1%) reported help with translating the questions. Proxy completion represented 2.6% of all patients who returned pack 1.

Of the responders with no impairment, 32% reported receiving help to complete pack 1, compared with 78% in the impairment group of responders. In the no-impairment group, 1.3% of the packs were completed on the patients’ behalf by proxy, compared with 10.1% in the impairment group.

The median number of unmet needs as measured by the LUNS questionnaire was four (range 0–19), with the prevalence of individual unmet needs ranging from 3% to 58%. The most commonly reported unmet need was for more information; this was followed by unmet needs in the areas of falls and memory/concentration.

In total, 16% (n = 69) of the no-impairment group and 12% (n = 10) of the impairment group reported no unmet needs (first pack only). The total number of unmet needs reported was higher on average for the impairment group than for those with no impairment (Figure 15). The median number of unmet needs for the no-impairment group was four and the mode was zero (range 0–19). For the impairment group, the median number of unmet needs was five and the mode was five (range 0–17).

FIGURE 15.

Total number of unmet needs reported by individuals.

A significantly higher proportion of patients in the group with impairment than in the group with no impairment had unmet needs in the areas of memory/concentration, incontinence and accessible holidays (Figure 16). The proportion of responders who identified memory/concentration as an unmet need was 26% (n = 114) in the no-impairment group and 45% (n = 41) in the impairment group.

FIGURE 16.

Types of unmet need in the impairment and no-impairment groups. a, p < 0.005; b, p < 0.05.

On the Impairment Manikin, responders reported impairments as listed in Table 52. A lower proportion of the no-impairment group than of the impairment group reported difficulties with speech and thinking.

Primary analysis: evaluation of acceptability, validity and test–retest reliability

Ancillary analyses: evaluation of internal consistency and dimensionality

Rasch analysis

The LUNS-22 PSI was 0.66 and overall chi-squared probability was < 0.0001. Three items (information, driving and daily occupations) were found to be poor fitting because of under- or overdiscrimination (all displayed significant chi-squared and ANOVA item misfit). There were response dependencies between items relating to aids/adaptations inside and aids/adaptations outside (0.224) and between mood and memory/concentration (0.223). DIF analysis showed a bias by age for the item relating to employment and a bias by age and gender for the item relating to physical relationships. Removing the five problematic items (with regard to fit indices and DIF) and accounting for the apparent response dependency within the scale results in a 17-item solution (LUNS-17), which displayed reasonable model fit as none of the items were individually misfitting. However, this 17-item tool remains unsuitable for use as a scale because of poor targeting and reliability/internal consistency issues (PSI = 0.52), The low PSI also reduces the power of the tests of fit, meaning that the observed adequate fit may be misleading. The Cronbach’s alpha value at this point is 0.77, but this does not take into account the targeting, or the response dependency that is present. The targeting plot can be seen in Figure 17, in which it is apparent that there is a skew between the locations of the items and the locations of the sample that is being measured.

FIGURE 17.

The LUNS Rasch targeting plot, displaying the relative distributions of person and item logit locations on the underlying ‘unmet need’ continuum. Grouping set to Interval length of 0.25 making 28 groups. n = 528, mean = –1.711, SD = 1.189.

Qualitative evaluation of unmet needs

In total, 10 semistructured interviews were conducted. Eight interviewees were male and the median age was 78 years. Nine interviewees lived with their husband/wife and one lived alone. During the LUNS study, six participants had identified no unmet needs, three participants identified one unmet need in either the first or the second pack and one participant identified one unmet need in both the first and the second packs. The qualitative study found that most people felt that they genuinely had no unmet needs. Despite this, all interviewees were able to identify current problems. However, interviewees often rejected the word ‘problem’ and therefore these are termed ‘identified issues’. Identified issues may relate to stroke, comorbidities and/or ageing. The type, impact and importance of these issues varied. People can be living with significant limitations yet still identify no unmet needs. People may be in the process of seeking help for their issues or may feel that they are managing them. Issues were grouped according to their effect on life (number of people who reported each in brackets):

• leg, arm or hand pain, weakness, spasticity or other limitation (nine)

• walking, getting out, mobility (nine)

• other (nine) (e.g. feeling slow, frustrated, experiencing a changed or strained relationship with partner, weight gain, breathlessness)

• balance or falls (eight)

• hobbies, leisure (seven)

• feeling tired (six)

• mood, anxiety (five)

• domestic activities (four)

• social life (four)

• memory and/or concentration (three)

• vision (two)

• throat and voice (two)

• communication (two)

• bladder and bowel (two)

• personal care (two)

• holiday (one)

• crying (one)

• sleeping (one).

The following themes emerged:

• Level of acceptance – relates to adjusted expectations; high levels of acceptance were common and this links with experience of ageing and may explain how issues can become redefined in the context of current life.

• Support received – people are often receiving support from their partner; family, friends or services also provide support.

• Relative situation – people compare themselves to other people and report feeling fortunate and valuing information from peers.

• Presentation of character – some people identify themselves as proud, being independent and/or determined.

• Experience of frustration – frustration was common and includes limitations in relation to carrying out tasks or things taking longer to do, for example.

• Experience of services – people have experiences and expectations about services, including what treatment is available and acceptable to ask for.

Interpretation

The LUNS questionnaire was developed to identify the longer-term unmet needs of stroke survivors. It consists of 22 items, covering information, services, social and emotional consequences and health problems and related areas. The study aim was to evaluate the acceptability, test–retest reliability and validity of the LUNS questionnaire in a broad sample of patients living at home post stroke. The response rate (69%) and minimal missing data, comparable to those for concurrent measures, suggest that the LUNS questionnaire is acceptable to the target group. Individual-item test–retest reliability was good and compares favourably to that for SF-12. It was found that identification of an individual unmet need was consistently associated with poorer outcomes on the concurrent health measures, demonstrating validity. Interestingly, unmet need was more strongly associated with lower mental health than with lower physical health or activity.

The acceptability of the LUNS questionnaire was lower in stroke survivors with cognitive or communication impairment than in those with no impairment. However, the acceptability in stroke survivors with cognitive or communication impairment was comparable to the acceptability of concurrent measures in this group. The test–retest reliability of LUNS items was slightly lower for the impairment group and comparison with concurrent measures suggests a reduced association between LUNS items and health status. Cognitive and communication impairments were assessed at baseline; however, Impairment Manikin results suggest that cognitive and communication impairment caused by stroke may persist up to 3 or 6 months post stroke for some people.

A secondary aim was to evaluate the internal consistency and dimensionality of the LUNS questionnaire. An initial Cronbach’s alpha value of 0.815 suggests a reasonable level of internal consistency among the 22 LUNS items. However, additional factor and Rasch analysis revealed that the LUNS questionnaire does not function as a scale because of multidimensionality and poor targeting; therefore, items cannot be added up to give a total score for level of unmet need.

A substudy interviewed a sample of patients who identified a low level of unmet need (no or one unmet need) to further understand self-reported unmet need following stroke. People in this study felt that they genuinely had no unmet needs, yet all participants identified current issues or problems, which may relate to stroke, comorbidities or ageing. Six themes demonstrate the importance of adjusting: help received from others; comparisons to other people; characteristics such as pride, independence and determination; frustration when engaging in activities; and experiences and expectations of services. This adds to the understanding of how problems are managed and what shapes the identification and perceptions of unmet needs.

To put the LUNS questionnaire in context, few suitable tools exist to assess an individual’s unmet needs following stroke. The Stroke Association needs survey consists of 44 items and has not been psychometrically evaluated. 4 The Southampton Needs Assessment Questionnaire has not been evaluated for test–retest reliability99 and the postal response rate was low at 49%. 100 This questionnaire was evaluated only in people aged ≤ 65 years, yet older adults make up the majority of the stroke population. 120 The Stroke Impact Scale is long (59 multiple choice items) and low postal response rates of 41%104 and 63%101 have been reported.

Generalisability and limitations

Comparison with data from the UK NSA population120 indicates that the LUNS study sample is reasonably representative of the wider UK stroke population (age, gender, time in hospital, post-stroke Barthel Index score). The ethnicity of the sample is representative of the ethnicity of the population aged 65+ years in England in 2001. 85 By setting a target for the proportion of patients with communication and cognitive impairments, and allowing proxy responses, the LUNS questionnaire has been evaluated on a sample with varying stroke severity. The target recruitment of the impairment group was reduced to meet study deadlines. Inherent barriers exist to the recruitment and inclusion in research of stroke survivors with cognitive and communication impairments. Therefore, stroke survivors with communication or cognitive impairment should be actively included, especially with regard to health or needs assessment and questionnaire evaluation. Lower test–retest reliability and problems relating to validity may be properties of all questionnaires when including stroke survivors with cognitive or communication impairment.

The LUNS questionnaire demonstrated a response rate and median number of unmet needs that are similar to those found for a UK stroke needs survey (response rate 60–78%, median unmet needs four). 4 A median of two unmet needs has been reported during post-stroke reviews using GM-SAT. 34 The LUNS questionnaire is a pragmatic tool which can identify the unmet needs of stroke survivors that relate to either stroke or comorbidities. These include unmet needs that may require inputs from health providers and other providers such as local authorities and third-sector organisations. Results from the LoTS care trial (see Chapter 3, Overview of stroke care co-ordinator practice) indicate that SCCs currently refer/signpost stroke patients to such whole system services. It is possible that the LUNS questionnaire may also have wider relevance beyond stroke care, to capture general issues that affect older people with long-term conditions.

Use of dichotomous responses is considered a strength as it is simple and results in an unambiguous response. 121 The potential complexity of double-barrelled questions was managed by providing instructions to tick ‘no’ if there is no need or if the need has been met. As a result, the LUNS questionnaire is quick for stroke survivors to complete and easy to interpret; items ticked ‘yes’ indicate the type and number of unmet needs that an individual has. The results show that, for the majority of individual items, responders commonly tick ‘no’; as such, this may result in vulnerability to a floor effect.

When assessing validity we were unable to compare the LUNS questionnaire with validated benchmark measures of unmet need as no such tools exist. Therefore, health status questionnaires were chosen that could reflect health-related needs.

It is acknowledged that some topics do not specifically map onto the LUNS questionnaire, for example fatigue, vision and communication. However, tool development incorporated stroke patient- and carer-identified problems and, furthermore, the qualitative exploration of the unmet needs study did not find that the LUNS questionnaire was missing these or any other needs. Iterative development and robust methods were used to capture evidence from the literature and the views of experts and > 150 patients. The subjectivity of unmet needs meant that patient involvement was an important part of the development of the LUNS questionnaire, and patient understanding of items demonstrates provenance.

Recommendations for research and implications for health care

We have demonstrated that the LUNS questionnaire has adequate acceptability, test–retest reliability and validity to provide a simple method for identifying longer-term unmet needs after stroke. Policy recommends that stroke patients are reviewed in the longer term15,55 and that this includes measurement of unmet needs. The LUNS questionnaire is suitable to be used during individual patient consultations to guide service responses. It is also suitable to be used to promote service development, compare different services or inform commissioning, for example evaluating whether services are meeting patients’ needs by measuring the percentage of the local stroke population whose mobility needs are unmet. Further research is needed to evaluate the sensitivity to change of the LUNS questionnaire, to determine whether it can be used to monitor change at a patient or service level over time. Results from factor and Rasch analysis suggest that the LUNS questionnaire should be used at the individual-item level rather than to give a score representing the burden of unmet need.

The observed lower test–retest reliability and validity of the LUNS questionnaire in patients with cognitive or communication impairment suggests that it may be important to research the accuracy of proxy responses and potentially to develop a version of the questionnaire that is suitable for people with communication impairment.

The subtlety and complexity of measuring unmet needs described in the substudy interviews should be considered when planning stroke reviews and using questionnaire tools, for example to ensure that individual circumstances are considered when assessing unmet need.

Aim

The study aimed to investigate longer-term recovery and adjustment post stroke with a particular focus on the processes and mechanisms that contribute to social inactivity and/or social activity.

Study design

A qualitative substudy of the system of care trial was undertaken, drawing on grounded theory techniques.

Methods

A purposive sampling strategy was adopted to identify trial participants who were less or more socially active than anticipated based on social activity and physical disability outcome measures and to ensure variation in key characteristics known to shape adjustment post stroke (age, socioeconomic status, living alone). A combination of qualitative methods was used, including multiple semistructured interviews, limited observations, solicited diaries, and informal and formal support mapping techniques.

Results

In total, 22 stroke survivors and 12 carers/significant others participated in the study. From participants’ accounts, four different post-stroke recovery trajectories have been identified: (1) disruption followed by adjustment and acceptance; (2) cycles of disruption followed by adjustment and acceptance; (3) disruption without adjustment and acceptance; (4) stroke as a continuation of ongoing decline. The different trajectories illustrate how multiple interacting factors shape the process and meaning of recovery and adjustment over time and therefore social participation. The factors that shape the different trajectories were identified.

Conclusions

Knowledge of different recovery trajectories and the factors that shape adjustment and social participation post stroke may enable stroke survivors, caregivers and professionals to identify factors that act as barriers to adjustment and social participation. This may help stroke survivors shape their own recovery trajectory and facilitate professionals in better tailoring services provided.

Background

Many factors are known to affect the process of adjustment following stroke and these factors may impact on social participation. They include:

• physical and functional improvement

• emotional factors

• the understanding given to the stroke, rehabilitation and disability

• stigma

• formal and informal support

• various contextual factors.

Physical and functional recovery shapes longer-term outcomes after stroke. 25,122,123 Stroke-related impairments and the resulting changed relationship that people have with their body can result in people experiencing their body as being unreliable. 123 In addition to the physical and functional implications of stroke, various emotional and psychological consequences have been highlighted in the literature, including depression, low mood, apathy, aggression, irritability and anxiety. 10,124,125