This multicentre study recruited from 12 English sites from December 2006 to January 2010. Sites were NHS hospitals that admitted adults with acute stroke . All had a speech and language therapy service with expertise in aphasia and dysarthria, and they agreed to reorganise their service to provide continuity of care from inpatient to community setting and to work to a consensus based intervention.
Adults admitted to hospital with a stroke were eligible for inclusion if they had impaired communication due to aphasia or dysarthria, were considered by a speech and language therapist as likely to benefit from this intervention, and gave informed consent (or proxy consent by carers). The intention was to be as inclusive as possible since the intervention was a flexible, enhanced version of current best practice and intensity could be modified to individual need.
Exclusion criteria were based on practical resource limitations such as the participant living outside the area served or being unable to communicate in the English language (therapists believed that translation or interpretation services were inappropriate for aphasia). Therapists also excluded people for whom intervention was deemed unsuitable—for example, those receiving end of life care or with pre-existing learning disabilities or dementia likely to prevent benefits from therapy, subarachnoid haemorrhage, or serious medical conditions (such as terminal disease); patients unable to complete the eligibility screening after three attempts; and patients whose communication problems resolved or were likely to resolve without intervention.
Initially, all stroke admissions to hospital were screened by speech and language therapists to determine eligibility. However, this proved unnecessary, and (from April 2008) speech and language therapists screened only those with suspected communication problems to confirm the presence and persistence of communication problems, provide a differential diagnosis (aphasia, dysarthria, or both), determine the severity of the communication problem, and rule out non-obvious exclusion criteria. They used the Frenchay Aphasia Screening Test (FAST),20
or informal assessment if the communication problems were too severe, and the impairment and activity scales of the Therapy Outcome Measure (TOM).21
Dysarthria was diagnosed and its severity determined by a TOM rating based on the speech sample from the FAST picture description task and conversation. The presence or absence of dysphagia (swallowing) was also diagnosed. Apraxia of speech was outside the remit of this study.
Before randomisation, research assistants also rated patients on the 10 item Modified Barthel Index22
with the help of the multidisciplinary team. The Barthel Index gives a score out of 20 and indicates severity of overall disability (beyond communication). Since the communication impairment was specifically assessed (TOM), no other stroke severity measures were required.
Before randomisation of eligible patients, the speech and language therapists provided them with an “aphasia friendly” leaflet about their diagnosis and asked if they would like to meet a research assistant to discuss the study. The multidisciplinary team was also alerted to the diagnosis. Speech and language therapists did not provide further communication support to participants, their families, or the multidisciplinary team until after randomisation. Research assistants immediately met potential participants to provide study information. Participants were given 24 hours to allow discussion with their family. We used aphasia friendly, accessible information materials developed in collaboration with our Research User Group partners.18
When necessary, proxy consent was requested, and research assistants later gave regular opportunities for participants to directly provide or withdraw their consent.
To ensure concealment of allocation, randomisation was by an external, independent, web based randomisation service from a trials unit activated by research staff. Randomisation was stratified by the severity of communication impairment and recruiting site. Stratification by diagnosis was also requested (aphasia only, dysarthria only, or both), but during data checks after study completion it became clear this had not occurred. Participants were randomised using a 1:1 allocation ratio and randomly permuted blocks.
The speech and language therapy was an agreed, best practice, flexible intervention developed by speech and language therapists for delivery early after stroke in usual care settings but better resourced than standard practice at most sites.18
In this study “early” refers to the acute to post-acute period of the stroke pathway (the first four months) and is used to distinguish this clinical population from those with chronic aphasia or dysathria. This allowed therapy to start as soon as clinically indicated and, if deemed appropriate, up to three contacts per week for up to 16 weeks, following participants along their stroke pathway. Start date, duration, and frequency of therapy varied within and between participants, as determined by each therapist’s clinical judgment and agreement with the participant about what was appropriate. This was not a “one size fits all” intervention, and these levels of therapy were not prescriptions but were fixed upper limits. Although well resourced, this study was not intended as a trial of different levels of intensity.
A manual was developed by the therapists setting out specific components of the intervention and service delivery, and can be described as a set of best practice guidelines and a compendium of resources. There were six core components of the intervention (see box). Adherence was ensured by an agreed coding system for all contacts and by employing a part time therapy monitor, an experienced speech and language therapist. Therapists attended regular facilitated peer group meetings where sites presented data on eligibility decisions and descriptive single case therapy. The therapy monitor visited sites to observe delivery and inspect the coded data.
Core components of the intervention
- 1. Assessment—Initial and ongoing, standardised, functional, case history, goal setting
- 2. Information provision—Communication problem, strategies or equipment to assist communication, intervention plan, therapist opinion of progress, available information resources and support networks
- 3. Provision of communication materials—Communication book for recording activities, an “alternative and augmentative communication device” if required
- 4. Carer contact—Discussion and provision of information, observation and participation in therapy, conversation training for patient’s partner, preparation for the end of the research intervention
- 5. Indirect contact—Written descriptions of needs, abilities, and strengths; discussions with clinical teams; goal planning with multidisciplinary team
- 6. Direct contact—Therapy to improve language skills at all levels of the World Health Organisation ICF model: Impairment (improving language skills), Activity (compensatory strategies), Participation (developing confidence, accessible information)
A similar frequency and amount of social contact was offered to those in the control group by employed visitors (not therapists or volunteers). Visitors had excellent social skills and general competency and were trained to deliver social attention absent of any intuitive form of communication therapy or strategy. They followed a short manual allowing everyday activities (such as conversation, television, music), but visits were mostly led by participants. Visitors were monitored to ensure adherence, including visits from the part time visitor monitor, who selected and trained visitors, observed their contact with participants, inspected the coded data from each contact, facilitated group supervisory meetings of visitors, and provided one to one support for visitors.
Objectives and outcomes
The randomised controlled trial aimed to evaluate the clinical effectiveness of the communication intervention compared with the control social contact six months after entry to the study from the perspective of different stakeholders:
- Therapists’ ratings of participants’ functional communication (primary outcome)
- Participants’ self reported functional communication and quality of life
- Carers’ perceptions of participants’ functional communication
- Carers’ own wellbeing
- Carers’ quality of life
- Adverse events.
For the primary outcome, a semi-structured conversation between each participant and an unfamiliar communication partner (study research assistant) trained in communication supportive techniques was videotaped. An independent speech and language therapist, blinded to treatment allocation and not involved in treating the study participants, rated functional communication on the communication activity scale of the Therapy Outcome Measure (TOM). Despite its ecological validity, conversation between a participant and a familiar partner was not used as an outcome because of the difficulties in standardising the content, amount, and support provided. The chosen method (unfamiliar partner plus blinded rating of videotapes) has demonstrated validity and reliability.24
Secondary outcomes were:
- Participants’ perception of their functional communication and quality of life—Based on the validated 20 item Communication Outcomes After Stroke (COAST) scale,25 which covers both understanding and expression in a range of communication situations, including five items on quality of life. The score is converted to a percentage, and higher scores indicate better outcomes.
- Carers’ perceptions of participants’ functional communication—Based on the relevant 15 questions on the validated Carer COAST scale.26 Higher scores indicate better outcomes.
- Carers’ wellbeing—Based on the 15 item Carers of Older People in Europe (COPE) Index.27 There are three subscales: negative impact (a high score is a poor outcome as it indicates stress), positive impact, and quality of support (high scores are good outcomes as they indicate satisfaction and support).
- Carers’ quality of life as affected by participants’ communication problems—Assessed with the relevant five questions from Carer COAST.26 Higher scores indicate better outcomes.
- Adverse events—Death, subsequent stroke, events leading to increased hospital stay or readmission to hospital.
A qualitative study was prospectively nested within the randomised controlled trial to support the interpretation of the trial findings in a way that would help commissioning decisions. Qualitative interviews with participants from both arms of the trial provided rich detail on service users’ perceptions and their views on the impact of early regular contact and are reported elsewhere.18
In most rehabilitation trials it is impractical and unacceptable to blind participants and those delivering the intervention to the allocation group. We endeavoured to blind the research assistants who collected the outcome data (password protected data related to allocation, participants asked not to mention allocation when visited). The risk of bias for the primary outcome was minimised by using expert speech and language therapists who did not know the participants and were blinded to allocation to rate the videotaped structured conversation, although the therapists may have seen communication aids in use. We felt strongly that participants’ and carers’ reported outcome measures should be included as secondary outcomes although they could not be blinded.
The primary analysis used regression methods to estimate group differences in outcomes at six months after adjustment for the intended stratification criteria (site, diagnostic group, and baseline severity of communication impairment on the Therapy Outcome Measure). The adverse event rates were compared without adjustment as they were not anticipated to be sufficiently common to allow multifactorial analysis.
Analyses included all participants assigned to their allocation group regardless of adherence to protocol—a complete case analysis under the “intention to treat” approach. Participants who were lost to follow-up or declined assessment were excluded. Those known to have died were included as having the worst possible outcome (no functional communication on the primary outcome). No other imputation was undertaken.
For outcomes assessed with COAST and Carer COAST, we compared valid assessments (that is, those with at least 90% of applicable items answered). No adjustment took place where responses between participant and carer might be construed as incompatible, as the instruments were designed to respect the individual’s self perception.
Sensitivity analysis was used to re-analyse the primary outcome data in several ways to assess how robust conclusions were to the choice of approach. Non-adjustment for intended stratification criteria, allowance for possible therapist effects, omission of people who had died, and per protocol analyses were all considered. The exact choice of such sensitivity analyses was inevitably data driven to some extent. For example, if primary analysis suggested a group difference the robustness of this conclusion needed to be examined (for example, allowance for possible therapist effect). Conversely, if primary analysis did not suggest a group difference the sensitivity analysis would focus on approaches that might identify possible explanations (such as per protocol analysis). We conducted pre-planned subgroup analysis of type of communication problem (aphasia or dysarthria) and of severity.
All statistical analyses were undertaken in STATA (version 10.1).
Sample size determination
The original protocol proposed a total sample size of 300 participants for 90% power to detect a difference of 0.5 points on the primary outcome of the Therapy Outcome Measure. The target effect size of 0.5 on the Therapy Outcome Measure was chosen as it is the smallest measurable difference on the scale. This calculation allowed for differential clustering between the two arms due to different numbers of therapists and visitors, with intra-cluster correlation coefficient of 0.05 in each arm.
Recruitment was slower than anticipated, leading to revision of the target. The observed standard deviation of the primary outcome for the first 43 recruited participants, adjusted as for primary analysis, was 1.1 points. The initial plan to incorporate therapist effects in the primary analysis was dropped as there was insufficient power to examine these potential effects and the independent Steering Committee considered an intra-cluster correlation coefficient of 0 to be a more reasonable assumption given our confidence in our monitoring system that ensured therapists adhered to the therapy manual. This led to recalculation of a target sample size of 170 participants to give 80% power at the 5% significance level to detect a difference of 0.5 points, allowing for approximately 10% loss to follow-up.
Interim analyses and stopping guidelines
No formal stopping rule was applied to interim analyses reported to the Data Monitoring and Ethics Committee. The videotapes for assessment of the primary outcome data were stored for distribution in batches. This precluded early stopping as the primary outcome remained largely unmeasured at the time of the committee meetings. Consequently no adjustments were made to the significance levels or confidence intervals in presented analyses.
Participant withdrawal criteria
No specific withdrawal criteria were defined for the study. Participants were able to withdraw their consent and stop their allocated intervention prematurely. In these cases they could choose to enter standard services for speech and language therapy at their site. Regardless of this potential protocol deviation, research assistants attempted to collect outcome data for all participants.
Given that participants were assessed for eligibility soon after stroke admission, specific exclusion criteria might come to light after randomisation. In these cases, the independent Data Monitoring and Ethics Committee was consulted as to the appropriateness of the participant continuing in the trial.