Developing primary care services for stroke survivors: the Improving Primary Care After Stroke (IPCAS) research programme

Changes to the programme’s original design

In the initial programme application, it was envisaged to run two trials: one cluster RCT to evaluate the IPCAS primary care model and one individual RCT to evaluate the MLAS self-management programme. We did explore nesting the evaluation of MLAS within the cluster RCT, whereby participants in the intervention arm would be individually randomised to be offered MLAS at the time of their review. After discussion with our programme steering committee, we agreed that this would be complex to prosecute and interpret. Furthermore, as the IPCAS research programme evolved during the development stages, it became evident that the two interventions were complementary to each other. Therefore, we decided to run a single trial with both components, IPCAS and MLAS, included in the intervention. We would carry out a non-randomised analysis to explore the effect of MLAS itself by comparing outcomes in stroke survivors in the intervention arm of the trial who did and did not take part in MLAS. A second change to the original programme, made in response to feedback from the funding panel, was to enlarge the process evaluation of the trial.

Multidisciplinary Intervention Development Group

The multidisciplinary IDG was convened at the outset of the IPCAS programme. The aim was to provide researcher, clinician and stroke survivor input to guide the development process and to agree the components of the primary care model. This involved reviewing results from the earlier studies and using these to inform both the content of the later studies and ultimately the design of the intervention.

The members (n = 25) included specialist stroke consultants, stroke nurses, general practitioners (GPs), practice nurses, a stroke survivor, a Clinical Commissioning Group (CCG) representative, representatives from the Stroke Association, community neurorehabilitation team members, clinical psychologists, a rehabilitation specialist, a return-to-work specialist and a behavioural science specialist.

We conducted five meetings, between September 2015 and November 2016, plus an additional meeting with representatives from primary care only (three GPs, two practice nurses) to discuss the implementation of model components in general practice.

During the meetings, we discussed the proposed components of the model and their iterations; results from qualitative work (literature reviews, interviews and focus groups) to inform model development and plans for model implementation in general practice. In the final meeting, the components of the primary care model were agreed.

1. Understand perspectives of stroke survivors and carers on experiences of health and care services since discharge from specialist services

Two studies informed our understanding of stroke survivor and carer perspectives: a systematic review and meta-ethnography of the literature and a qualitative interview study with stroke survivors and their carers.

2. Explore roles of specialist and primary care services for stroke aftercare and avenues of communication between specialist and primary care

As the programme developed, we realised that the roles of specialist and primary care services for stroke aftercare would in fact be better met by a consensus study, which is described under objective 4 of this workstream. We performed a focus group study to explore communication between specialist and primary care in the context of stroke.

3. Consider poststroke checklists for primary care

Traditionally, primary care responds to issues raised by patients rather than by using checklists. However, checklists have underpinned other initiatives aiming to address longer-term issues facing stroke survivors, including LoTS care6 and the Greater Manchester Stroke Assessment tool (GM-SAT, updated in 2018 to GM-SAT2). 26,27 Furthermore, the use of a standardised tool is recommended in implementation guidance for poststroke reviews. 13 While there is evidence that the GM-SAT is feasible to administer in the community using third-sector co-ordinators and that it is acceptable to patients and carers,26 the number of items on the tool (35) and the length of time it takes to complete (a mean of 74 minutes) do not make it a practical tool for primary care facing mounting workload pressures. 28 Recognising the need for a parsimonious checklist, an international expert panel developed an 11-item checklist for the identification of long-term problems after stroke in a primary care setting. 29 However, primary care clinicians, stroke survivors and carers were not represented in the panel that produced the checklist. Therefore, we ran focus groups to explore the potential role of a checklist in the setting of UK primary care.

4. Establish criteria for rereferral from primary care to specialist care

Guidelines in England and Wales for rehabilitation after stroke do not specify where these interventions should happen or when referrals to specialist services should occur. 5 Critical to understanding the potential for a primary care stroke model is whether unmet needs of stroke survivors are best met by developing primary care services or by improving access to specialist services. To inform this debate, a consensus study was performed.

5. Agree the components of the primary care model

In addition to the studies described above that related to specific objectives in workstream 1, two other pieces of work were carried out to inform the primary care model, a scoping review of the literature and PPI workshops.

6. Pilot the primary care model

Aspects of the pilot study, which is more correctly described as a feasibility study,35 were published in our paper on using a checklist to facilitate management of long-term care needs after stroke. 30

This was a 6-month non-randomised, uncontrolled study in a single practice in Cambridgeshire. The aim was to assess feasibility of delivering the new model of primary care for stroke. Specific objectives were to assess the uptake of the new service by patients, the workload implications for the practice and acceptability of the model. We also used the results to inform the subsequent trial design (recruitment rate and completion rate of questionnaires intended as outcome measures).

A training session for practice staff to deliver the IPCAS intervention was set up, and a one-off meeting was organised for general practice staff, representatives from hospital stroke services and the community neuro-rehabilitation team. The study population was adult stroke survivors identified from the general practice stroke register. Patients were excluded if they were on the palliative care register or resident in a nursing home. The practice generated a list of potentially eligible people from their clinical computer system which was reviewed by the GP. Potential participants were sent an invitation letter to take part in the study, with a consent form to return to the researchers. A single reminder was sent if there was no response. On receipt of the consent form, the stroke survivor was invited for a review and sent the checklist to complete. At the end of the 6-month period, a debrief meeting was held with practice staff. Patients completed a feedback questionnaire immediately after the review and after 3 months.

Forty-eight stroke survivors were sent an invite from the general practice, and 13 participants were recruited. All attended a stroke review with a practice nurse. Nine had completed the checklist prior to the review. The four who had not did not feel they had any stroke-related problems. While participants did identify their needs from the checklist, none of them ranked what were their top three. Clear action plans were agreed in 10 reviews. Twelve participants completed the first feedback questionnaire, and 11 the second one. Most found the checklist easy to complete and useful, and most were quite or very satisfied with the review. The mean review length was 44 (range 25–55) minutes. Part of this time was taken dealing with questions about the research. The other components of the model (DPoC; service mapping resource) were found to be acceptable to practice staff.

The results from this feasibility study were used to set realistic target recruitment for practices that would be engaged in the main trial. We also modified the intervention by adapting the staff training, simplifying the checklist and reducing the paperwork needed to be completed by the practice nurse.

Literature review

A literature search was carried out (see publication for search strategy)38 to identify pilot studies, RCTs and systematic reviews that had been published between 2005 and 2015 on self-management interventions for stroke.

The literature review identified 25 RCTs and 6 systematic reviews relevant to the aims of this study. Interventions were delivered individually face to face (n = 11), group-based (n = 7), via telephone (n = 3), online (n = 2) or as a workbook (n = 2). Topics across the identified interventions included: understanding stroke (recovery, recurrence, medication, prevention); practical advice; managing emotions and behaviours; health behaviour change (promotion of health lifestyle); dealing with stress and fatigue and future focus. Resources used within the interventions included personalised manual/workbook, written tip sheets, card tasks (to identify goals), prevention package, risk factor profiles, keeping well plan, fatigue diary and problem rating sheets.

Consultation phase

Building on the findings from the literature review, we undertook consultation with a multidisciplinary healthcare professional group and stroke survivor groups. Workshops were held with HCPs and stroke survivors separately to identify the different perspectives on stroke survivor needs, the timing and format of a stroke self-management programme as well as priorities for content. HCPs included physiotherapists, occupational therapists, nurses, speech and language therapists and stroke consultants. Three separate stroke survivor groups (including carers) were consulted across Leicester and Cambridge.

Key areas that were highlighted by HCPs included emotional needs and practical strategies to manage low mood; prevention strategies to reduce the chance of further strokes; follow-on support and education to aid stroke survivors’ recovery, adjustment and adaptation; and an introductory individual appointment prior to attending group sessions.

It was acknowledged that challenges may exist in offering a programme suitable for all stroke survivors at different stages of recovery with varying disabilities and effects (e.g. catering for physical disabilities as well as cognitive and speech problems).

Stroke survivors and carers also supported the idea of an individual appointment at the outset. They stressed that the programme should be held in an accessible venue and should focus on including management of emotions and confidence building. Stroke survivors felt the programme should consist of four to five group sessions, preferably held in the morning due to fatigue experienced later in the daytime.

Iterative phase

Based on the literature review and consultation, a prototype self-management programme was developed. The philosophy underpinning MLAS was that it should be patient-centred, enable empowerment and assist people who have had a stroke to maximise their well-being and QoL. The aim is to achieve this across four areas: social well-being and integration; acquiring a level of understanding of, and capacity to manage, emotional responses to living with a stroke; maximise own physical potential (including cognitive and sensory abilities) and minimise risk of future stroke. MLAS was grounded in a narrative approach (i.e. participants would become more aware of their journey with stroke) and utilised Social Learning Theory, in particular self-efficacy. 40 The course also utilised aspects of cognitive behavioural theory and the capability, opportunity, motivation and behaviour (COM-B) model. 41 In terms of format, MLAS consists of one 30- to 45-minute one-to-one individual appointment, four weekly group sessions of 2.5 hours each delivered by two facilitators and a final 30-minute one-to-one individual appointment 4 weeks after the final group session.

MLAS was tested in two iterative cycles. The first involved volunteers (10 stroke survivors and 4 carers) from an existing stroke support group in the community. The second was delivered to six stroke survivors and three carers who volunteered from contacts within the research development team or wider support groups. After each iterative session, feedback and observation notes were discussed within the development team, and further amendments were made to the draft curriculum plans. A feedback PPI group was held by a researcher not involved in the delivery at the end of each iteration programme, for participants to share their thoughts and insights on the programme.

Overall, MLAS was well received. Minor amendments were made to content, resources and delivery. Some written activities were changed to a verbal format to accommodate communication needs and physical abilities of participants. Written summaries of key points were added to aid those with memory problems. The narrative journey metaphor was incorporated and referred to more often throughout MLAS to enhance participant engagement. Feedback led to changes to the structure and content of some sessions. For example, some group sessions were made interchangeable based on participants’ preferences. An additional section about relationships was also introduced.

Following feedback from the iterative phase, we developed a final curriculum detailing key outcomes, content, facilitator behaviours, participant activities and resources for each session. We also provided a session outline, example open questions and guidance. We provided resources, including a road map, action plan and photographic images of ways to manage health and well-being. Handbooks for participants for each group session were developed. These included worksheets and opportunities for participants to write down their own notes. A stroke directory of useful local and national services, developed as part of the IPCAS primary care model, was made available.

Feasibility study

The aim of this study was to test the acceptability and feasibility of MLAS as well as to explore what outcome measures to include as part of further testing.

First, six facilitators were trained on a 3-day training course incorporating both the underlying principles of MLAS and its content and stroke-specific knowledge. Second, invitation letters were sent to people on the stroke registers of four general practices (three in Leicestershire; one in Cambridgeshire). People who expressed interest were offered an appointment at a community venue. At this visit, consent was obtained, and a number of baseline questionnaires were completed. The stroke survivors, some with their carers, then attended the MLAS courses. A feedback session was held at the end of the last session, facilitated by an independent researcher, and a separate feedback session was held with facilitators. Further questionnaires were sent out to participants after the final session.

Seventeen stroke survivors and seven carers participated in three MLAS courses. Fifteen (88%) patients and five (71%) carers completed the course. Fourteen out of 15 (93%) participants who completed a follow-up questionnaire said they would recommend the course to someone else. Participants approved of using a community venue (rather than a hospital), felt that the one-to-one with a facilitator prior to the group sessions was important and valued the sense of not being alone that came from the groups. Some participants would have liked to have attended the course earlier after their stroke – the average time lapse was 7 years. With regard to the questionnaires that were used as possible outcome measures, change was observed in the Southampton Stroke Self-Management Questionnaire (SSSMQ),42 but not in the others. Facilitators wanted more time in training and a chance to practice their skills. Suggestions were made for how to improve the training (e.g. showing videos of facilitators delivering sessions; including delivery ‘tips’). In parallel to the participant feedback, facilitators speculated as to whether the course should be offered earlier in the stroke pathway, as some participants had already largely adapted to their stroke.

This study has shown that My Life After Stroke was an acceptable and feasible self-management programme to deliver to stroke survivors in the community. Feedback from MLAS was positive, although completion of all the questionnaires was difficult for some. Attendance rates and retention to MLAS itself were high, suggesting it is a feasible programme to deliver. Retention and feedback from those who attended MLAS were positive, suggesting that attenders found it beneficial.

Cluster randomised controlled trial

The protocol for the trial has been published (http://doi.org/10.1136/bmjopen-2019-030285). 34 Since the intervention was to be delivered at the level of a general practice, a cluster design was adopted. See Figure 2 for the trial profile.

FIGURE 2.

Trial profile.

The aim was to evaluate the effectiveness of the IPCAS model in improving emotional health and increasing participation of people with stroke living in the community compared to standard care. The design was a two-arm cluster RCT, with the unit of randomisation being a general practice. The coprimary outcomes for the trial were two subscales (emotion and participation) of the Stroke Impact Scale (SIS) v3.0 12 months after randomisation (adjusted for baseline). 43 Analysis set significance at 2.5% [and confidence intervals (CIs) correspondingly at 97.5%] to take account of the two coprimaries in the analysis.

We recruited practices with at least 100 patients on their stroke register from the East of England and East Midlands. We aimed to recruit an average of 20 patients per cluster from 46 general practices (i.e. 920 participants). We estimated this would give us 90% power to detect an effect size of 0.33. We recruited patients by mail. Once we had sent all invitation letters and reminders for a practice, it was randomised centrally using a stratified random permuted block design. Practices in the intervention arm would then send out invitations for a stroke review and our 15-item checklist. 30 At the review, the patient was given information about the MLAS programme, including information about how to access it. It was intended that the research team would arrange meetings between the primary care staff of intervention practices and specialist staff from the hospital and the community in order to enhance primary/secondary care communication. In the event, this proved logistically difficult, so in most cases, the primary care staff were provided videos of the specialist staff explaining their service and how the practice could contact them. Training was provided for all intervention practices which included discussion of the mapping of local services that was provided and how the single point of contact would work. In the control arm, we asked general practices to continue to provide ‘usual care’. Outcome data were collected by postal questionnaire at baseline, 6 and 12 months after randomisation. If data were missing for the coprimary outcomes, then patients were followed up by telephone. The general practice records were reviewed after 12 months to ascertain health service use of trial participants.

More details on the results are provided in Appendices 3–5.

Recruitment took place between May 2018 and April 2019. We sent invitations to 4432 patients, of whom 23.5%, 1040 patients (113% of target), were recruited from 46 practices. The number of patients recruited per practice varied from 8 to 36. The mean age of participants was 70.6 years, and 63.1% were male. Twenty-three participants (2.2%) were of non-white ethnicity. The median time since last stroke of the trial population was 5 years. There were no important differences in baseline characteristics between intervention and control participants (see Table 1). Over the course of a 12-month follow-up, there were 21 deaths (9 in the intervention arm; 12 in the control arm), and 124 patients withdrew (78 in the intervention arm; 46 in the control arm).

Primary outcome data at 12 months were available for 778 (emotion) and 776 (participation) patients (75% of those randomised, 76% of those still alive). The primary analysis found that at 12 months, the intervention was not associated with any significant change in either of the coprimary outcomes: 0.08 (97.5% CI −2.3 to +2.5) in the emotion outcome and 1.3 (97.5% CI −2.3 to + 4.9) in the participation outcome (see Table 2). An effect size of 0.33 for which the trial was powered is the equivalent of a 6-point change in the emotion outcome and a 9.1-point change in the participation outcome. The 95% CIs of the observed differences between the change in outcome in intervention and control do not include these prior assessments of a clinically important effect. Therefore, it can be concluded that the intervention did not have a clinically important impact on emotion or participation of participants.

It was noted that there were ceiling effects, in that some participants had maximum scores on the SIS at baseline and so could not undergo a clinically important improvement. Therefore, in a secondary analysis, the study population was split into quintiles by baseline scores for emotion and participation (see Figure 3). For patients in the lowest quintile of emotional health at baseline, there was a non-significant 4-point increase in score in the intervention arm relative to the control arm. A similar effect is seen for participation, where people in the lowest quintile in the intervention arm gained a non-significant 3.4 points more than the control arm. However, even in these quintiles, the effect size was not what had been deemed clinically important.

FIGURE 3.

Change in emotional health (SIS3) and participation (SIS8) in intervention and control from baseline to 1 year by baseline SIS score.

With regard to secondary outcomes, there was also no evidence of effect of the intervention on short form SIS, QoL [EuroQol-5 Dimension, five level questionnaire (EQ-5D-5L)], well-being ICEpop CAPability measure for Adults (ICECAP-A), Southampton Stroke Self-Management questionnaire or Health Literacy Questionnaire (HLQ).

Four hundred and twenty-one (81%) of 522 patients in the intervention arm received a structured stroke review. Three hundred and ninety-three patients completed the checklist. One hundred and seventy patients (43%) identified fatigue as a need. Mobility (153, 39%), mood (130, 33%) and cognition (128, 33%) were the next most commonly identified needs. This high prevalence of fatigue, mood and cognition problems was also noted in the feasibility testing of the Greater Manchester checklist. 26 Two hundred and thirty-seven (56%) of the 421 who had a review had an action plan produced. The principal actions included practical advice given (n = 189), follow-up appointments arranged in primary care (n = 127) and referral to specialist care (n = 104). With regard to both the emotion and participation subscales, patients who attended the review had higher (i.e. better) scores at baseline compared to those who did not, but over time the difference narrowed. Among those who attended a review, people with an action plan had worse emotional health at baseline than people who did not have an action plan, and this difference persisted over time. Thus, there was no evidence that attending a review or having an action plan had any impact on emotional health or participation. Four hundred and twenty participants were invited to an MLAS course, but only 139 participants took part, of whom 102 completed it (73%). Given this variable uptake of the intervention components in the trial, a non-randomised analysis was performed to compare the outcomes in people in the intervention arm who had different levels of intervention. People who attended the MLAS course had lower scores on the emotion and participation scales at baseline, and these differences also persisted. Thus, there was no evidence of an effect of attending the MLAS course on emotional health or participation. We also looked at the effect on the Southampton Stroke Self-Management questionnaire, since this had been specifically selected for its relevance to MLAS, but again, there was no evidence of effect.

In conclusion, the trial showed no evidence of any effect of the IPCAS intervention or the MLAS component on any measured aspect of health status of participants. The study was of sufficient size to rule out a positive effect, and secondary analysis suggested that it was not an artefact due to ceiling effects. There was no association in the intervention arm of receipt of intervention with outcome providing further strength to the conclusion that the intervention and its components do not have an important effect on stroke survivors in the population studied. It is possible that the lack of effect relates to intensity of intervention. We successfully developed an approach to a review of longer-term stroke needs that could be delivered in primary care (i.e. delivered within a 30-minute practice nurse appointment). This contrasts with the 74 minutes to administer the GM-SAT, the impact of which has not been subject to trial evaluation. 26 The EXTRAS trial involved five reviews over 18 months, as opposed to the single review in this study. 10 The study was not powered to explore effect in different subgroups. Thus, we cannot exclude the possibility that the intervention or some of its components might be effective, for example, if offered to stroke survivors earlier in their stroke journey. Nevertheless, this trial strongly suggests that a primary care model targeted at all people on stroke registers is unlikely to be effective.

Process evaluation

The approach taken for the process evaluation has been published (http://bmjopen.bmj.com/cgi/content/full/bmjopen-2020-036879). 44 More detail is provided in Appendix 6.

The specific objectives of the evaluation were to describe how the intervention was delivered, how participants engaged with it and the impact of context on intervention delivery.

A mixed-methods design was used. We were interested in fidelity of design, fidelity of training, fidelity of delivery and fidelity of engagement. Methods included questionnaires sent to practices (to ascertain ‘usual care’); video-recording and audio-recording of training sessions of staff; direct observations; and interviews with staff and participants.

Both IPCAS and MLAS components were coded by two raters against the underpinning theoretical framework (Normalisation Process Theory for IPCAS; Social Learning Theory and Narrative Approach for MLAS). Both interventions were found to align with the appropriate framework(s). As noted above, 81% of intervention patients in the intervention arm received a stroke review. On average, this review lasted 28 minutes (within the 30-minute target). Nineteen of the training sessions of intervention practices were recorded. Each session lasted an average of 90 minutes with three members of staff present (typically a GP, a practice nurse and a practice administrator). The fidelity of training was high, with 96% of planned items covered. Two MLAS training sessions were observed, each involving 10 trainees. Again, the fidelity was high, with 87.5% of planned components covered. With regard to delivery of information at the stroke review, this was assessed both by self-report (high fidelity reported) and by audio-recording and review by a researcher (moderate fidelity observed). The biggest discrepancy was for explanation about the DPoC, which the audio-recording found was only done with moderate fidelity. In terms of the structured review, three-quarters of patients completed the 15-item checklist. Actions (n = 431) reported by staff included follow-up appointments (29%), referrals (25%) and provision of advice (45%). Only 139 patients attended an MLAS course. The main reasons for not attending were that patients felt that they had made a good recovery from their stroke, that they had other health issues, that it was not relevant to them or that they would find it difficult to attend. Six MLAS sessions were observed. Overall fidelity of delivery was high, ranging from 78% to 92%. A sample of patients completing post-MLAS questionnaires gave positive feedback that the objectives of the course had been achieved.

Twenty-seven stroke survivors in the intervention arm were interviewed. Some valued the stroke review, while others felt it would have been useful earlier post stroke. Some felt that their healthcare problems were not related to the stroke. Several participants did not remember being told about the DPoC but said that such a service would be useful. Those that did remember it did not use it but said that was because they did not need to. People liked the checklist, and they valued the service directory (given to them in MLAS). People who had attended MLAS reported benefits to participation, including greater self-awareness, more positive mindset, improving their knowledge and providing them with social support. Control patients that were interviewed did not discuss any stroke-specific needs with their practice during the study.

Eleven HCPs were interviewed. Generally, they found the checklist useful, though time-consuming, as often they needed to help the patient complete it during the appointment (i.e. it had not been filled in advance). Those that used the service mapping found it helpful, but others thought it had too much information. Only one healthcare professional reported being contacted as a result of the DPoC service, but they felt that the practice had gone to considerable effort to set this up. The option of the MLAS course was inconsistently conveyed to patients. The reviews were delivered either by practice nurses or by research nurses. The latter found the reviews more difficult as they were not an integral part of the practice team and did not know the patients.

In summary, with regard to the different components of IPCAS, delivery of the structured review had the most uptake, with over 80% of participants receiving a review, with the checklist being used in three quarters, and action plans being generated. Only a minority of patients attended the MLAS course, but those that did appeared to value it. Practice staff perceived that a lot of effort was put into setting up the DPoC service, but this was hardly used. The local directory of services had variable take-up by HCPs. It did not prove possible to support improved communication between primary and secondary care as originally intended. Instead, we needed to rely on videos of specialist staff explaining what they did and how to access their service. Training fidelity was high.

Health economic evaluation

Though the trial did not demonstrate that the intervention had an effect, it is still important to consider the cost consequences of the intervention. For example, in the EXTRAS trial of extended contact with the early discharge support team after stroke, although the intervention had no impact on the primary outcome measure (extended activities of daily living), it did reduce costs and so was found to be a cost-effective intervention. 10

Costs were measured from the perspective of the UK NHS. Resources used to implement the intervention were costed using a micro-costing approach which took into account the time spent by healthcare staff on training and delivery of the intervention. At 12 months, we sent patients a questionnaire on their contacts with secondary care services and with social services over the course of the trial. We obtained data on contact with primary care from the general practice electronic health records. Quality-adjusted life-years (QALYs) over the course of the trial were estimated from the EQ-5D-5L questionnaire at baseline, 6 and 12 months. Missing data were assumed to be missing at random and were imputed using multiple imputations. We calculated incremental cost-effectiveness ratios (ICERs) to represent the incremental cost required to obtain one additional QALY when moving from usual care to the new intervention. Prespecified analyses included use of multiple imputed data and unadjusted estimates. We also performed ad hoc analyses to determine the impact of baseline QALY scores and performed a complete case analysis.

Detailed results are provided in Appendix 7. The average cost of the intervention was £68 per person. In addition, there was significantly greater use of practice nurses, healthcare assistants and respite care services. There was no significant difference in use of secondary care services. At 1 year, the mean EQ-5D-5L was 0.734 in the intervention arm and 0.721 in the control arm. There was a non-significant mean QALY difference of 0.013 (95% CI −0.024 to 0.048) in favour of the intervention. The intervention was associated with an incremental cost to the NHS of £267.07 per person. This equated to an incremental cost per QALY of £20,863. Using a threshold of £20,000 per QALY, there is a 48.4% chance that the intervention is cost-effective. If we adjust for differences in baseline QALYs, then the incremental cost per QALY rises to £45,489. Using a complete case analysis, the incremental cost per QALY is £39,532.

In conclusion, it is unlikely that the primary care model that we tested is cost-effective at a threshold of £20,000 per QALY. The intervention increased primary care workload, and while it was associated with a non-significant, higher QALY at 12 months, the size of this difference diminished when adjusted for baseline QALY scores.

The independent Programme Steering Committee

The independent Programme Steering Committee which includes Marney Williams as a lay member, retains oversight of the research programme. This group met eight times between May 2015 and September 2020, through both the developmental and evaluation stages.

The Intervention Development Group

The IDG met on six occasions between September 2015 and November 2016. The group included two representatives from The Stroke Association and Bundy Mackintosh (a research grant co-applicant and stroke survivor).

Consultation with a stroke survivors support group in Cambridge facilitated by The Stroke Association

Around 20 stroke survivors and carers attended on three separate occasions between April 2015 and March 2017. Participants were informed about the programme of work and invited to comment and provide feedback on specific aspects of our research plan. We gained valuable feedback on study materials and outcome measures for both developmental streams of the programme.

Consultation with a stroke survivors support group in Bedford facilitated by The Stroke Association

Two stroke survivors and one carer attended in April 2016. They were informed about the programme of work and invited to comment and provide feedback on specific aspects of our research plan. Specifically, they provided feedback on the findings from the focus group and materials planned to be used as part of the primary care model for stroke (checklist).

Meeting with a stroke survivor and caregiver consumer research advisory group in Leeds

A research associate from the programme presented the aims and objectives of the research to a group of stroke survivors and caregivers at one of their regular meetings (September 2016). We received patient and caregiver feedback on the results of the qualitative studies, validation of the findings, and informed patients of the progress with the research they had advised on previously. This stimulated a group discussion, and multiple suggestions were incorporated into the study materials following the group’s feedback.

Stroke survivors and carers have been involved throughout the development of the Managing Life After Stroke course

Three consultation meetings with stroke survivor groups took place in April 2016 (2 in Leicester, 1 in Cambridge – 20 people per group) to seek their experience of the impact of stroke, support they received, opinions on content of a self-management intervention, its format, as well as on initial drafts of the programme. Once we had a draft intervention, we tried it out with interested patients and carers and refined the intervention based on feedback from this PPI, as well as the facilitators who ran the sessions.

Secondment of a staff member from The Stroke Association

Between April 2016 and April 2017, a staff member from the charity organisation was seconded to support development of the new primary care model. Specifically, they worked on the service mapping exercise relating to the provision of information about health, social and community care availability within the local area.

Piloting of the stroke survivor topic guides used for intervention fidelity assessment of the randomised controlled trial

The topic guides were initially developed from semistructured interviews completed during intervention development work. The topic guides were then pilot tested, involving six e-mails, three telephone calls and a mock interview with a member of our PPI group, and refinements were made following feedback.

Key findings

Impact of patient and public involvement

As summarised previously, PPI had a major impact on the design of the interventions and implicitly provided validation that the interventions would be helpful. We obtained a different perspective from some stroke survivors from our qualitative studies, with feedback that some aspects were not relevant or might have been more helpful earlier in the stroke pathway. Our PPI was composed of longer term stroke survivors, so why the discrepancy? The key point is that it is not the role of PPI members to provide representative views but their own perspectives. It is likely that people who are still engaged with stroke charities and stroke clubs (our sources of PPI members) continue to identify stroke as a major part of their lives and do not reflect the attitudes of people who feel that stroke is no longer something that defines them. This is not to diminish the value of our PPI input but simply to emphasise that other approaches to gaining the patient's voice (such as qualitative research) remain important complementary methods.

Successes

With over 1000 participants, this is the largest completed RCT of an intervention to address longer-term needs of people with stroke. We exceeded our recruitment targets and achieved acceptable follow-up rates (76% of those still alive at 12 months) for our primary outcome measures. In parallel with the trial was an in-depth process evaluation that enabled us to conclude that the lack of impact of the intervention was not due to flawed implementation.

Despite the start of the coronavirus disease 2019 (COVID-19) pandemic during the final year of the trial, we were able to rapidly adapt our process and complete the programme within the original budget and with only a 6-month extension (which was related to the complexities of earlier phases of the programme, rather than to COVID-19).

While the overall result of the trial was to demonstrate that the intervention was not effective, we were able to draw important conclusions for clinical practice and research, as discussed below.

Challenges and lessons learnt

We underestimated the difficulties of incorporating improved primary–secondary care communication into our model. Our original plan, which was prosecuted in the feasibility study, was to set up face-to-face meetings between the specialist stroke team and general practice staff. As we moved into the trial phase, this proved very difficult to arrange due to availability of personnel to attend such meetings in both settings. Therefore, we moved to a more pragmatic, but perhaps less effective approach, of obtaining videos to circulate to the practices of the relevant secondary care staff, which covered what their roles were and how to contact them.

A second challenge was achieving our objective of establishing criteria for re-referral to specialist care through a consensus study. When we originally conceived this study, we anticipated that its output would be a set of consensus-based recommendations to guide referral decisions. However, as we developed this study, we recognised that, in the absence of evidence and in the face of subtle nuances that would influence decisions, such guidelines would have limited face validity. Therefore, we used this study to better understand the types of issues that come into play when rereferral is being considered. Thus, rather than have a set of guidelines to distribute to primary care, we had case scenarios as an outcome that we explored in the training of the primary care professionals.

A third challenge was our choice of primary outcome measure for the trial. We deferred specifying this in advance of developing the primary care model, as we wanted the primary outcome to reflect what the model was specifically aiming to address. We wanted the primary outcome to be sensitive to change, yet not too burdensome to complete. Trials of related interventions addressing long-term issues facing people with stroke had used the Nottingham Extended Activities of Daily Living Scale10 and the General-Health Questionnaire-12. 6 Both these trials had not demonstrated any effects on these outcomes. We selected two outcomes of the SIS – participation and emotion. 43 In the event, we did not demonstrate any difference in outcome between intervention and control in these measures, and ceiling effects were observed. Unlike the EXTRAS trial,10 we did not observe significant differences in any secondary outcomes either, so we have no evidence that had we chosen different primary outcomes, we would have obtained a different result.

Our intervention was delivered within the trial before the COVID-19 pandemic struck. The impact of this pandemic on longer-term support for people with chronic conditions such as stroke has yet to be fully understood. Patterns of care are likely to change permanently with greater use of remote support. This creates some challenges with regard to considering the implications of our findings. It may be, for example, that remote delivery might address some of the practical difficulties of access for some patients and therefore make the intervention more attractive. Similarly, the needs of people with stroke might have changed as a result of the increased isolation imposed by the pandemic.

Finally, we faced the challenge of how to evaluate a multicomponent intervention where there is value in knowing the impact of the individual components as well as the impact of the whole package. While initially we had considered evaluating one of the components (the self-management component, MLAS) using an individually randomised trial nested within the cluster-randomised trial, we recognised that this was going to be impractical. Instead, we decided to do a cluster RCT of the whole model with an embedded observational analysis of the impact of MLAS and a process and qualitative evaluation to explore what was likely to have had the greatest effect. In the event, the overall trial did not detect any benefit, and the observational analysis of MLAS did not suggest any effect. However, given the relatively small numbers of people (n = 139) who attended an MLAS course, we cannot exclude an effect of that component, particularly in an appropriate subgroup (e.g. patients in the first 6 months after their stroke).

Limitations

Although we exceeded our recruitment target per practice, the age, sex and ethnicity distribution of our trial population was not typical of people on a primary care stroke register, being younger with a higher proportion of men and low representation of minority ethnic groups. 45 Given our qualitative findings that some older people identified their health problems as related to ageing rather than to their stroke, the age distribution of our trial population does not have a major impact on the applicability of our findings, since such people would be less likely to benefit from the intervention.

Some components of the intervention had only limited take-up. There was over 80% uptake of the structured stroke review, but only a third of people who were invited to take part in the self-management training programme took up the offer. While available, the single point of contact at the practice was hardly used. We had not been able to implement our improved communication between primary and secondary care in the way that we had planned.

Our primary outcome measures did have ceiling effects in that some participants had maximum scores on the SIS at baseline. Nevertheless, clinically important differences between intervention and control groups were not observed in the subgroups where changes in scores were possible.

A key component of the intervention was a structured review of needs. An implicit assumption in this is that once a need has been identified it can be addressed. This gives two possible explanations for why the intervention did not work. Firstly, there is a lack of evidence as to how to address the most commonly identified needs. Secondly, even if there is an evidence base, there may be a lack of service availability. As we showed, psychological needs post stroke are common, but access to specialist psychological services is limited, even when patients are still under specialist care. 46

Our follow-up was limited to 1 year. It is conceivable that the effect of the intervention might have increased over time. While the observed differences in the coprimary outcomes were larger at 12 months rather than 6 months, they were still well short of what we had prespecified as a clinically important threshold. The intervention was not repeated at 1 year, so it is unlikely that the differences would have crossed this threshold had we followed up for a longer period of time.

Implications for practice

1. We found no evidence that structured stroke reviews offered in primary care to all stroke survivors in the community are effective at addressing stroke-related needs. Such reviews should perhaps therefore be time-limited after stroke and only continued in selected patients.

2. The 15-item checklist that we developed is practicable to be used in primary care for patients where it is perceived to be useful.

3. Given that stroke survivors in this study report high prevalence of fatigue, low mood and cognitive issues, greater attention needs to be paid to services that can address these needs, informed by evidence on interventions, described below under research recommendations.

4. From the evidence in the wider literature, self-management programmes have a role in improving QoL in stroke survivors in the community. 37 Our findings suggest that such programmes are likely to be of greatest value if offered early after discharge from hospital.

Implementation issues

The process evaluation found that the intervention was largely implemented as intended, though we did not implement improved communication between primary and secondary care in the way we intended. The new Integrated Care Systems that came into operation in the NHS in July 2022 are an explicit acknowledgement of the need to better integrate health and social care. The Integrated Care Partnerships that will be set up under these may facilitate better integration between primary and secondary care. New efforts to integrate primary and secondary postacute stroke care might fall within the remit of the Integrated Community Stroke Service (ICSS). 47 It is envisaged that ICSS provision will be for 6 months post discharge from hospital or 6 months for community referrals. However, the model does not specify how the ICSS will engage with primary care, which will be particularly relevant for referrals from the community.

In the post-July 2022 NHS, it would be appropriate for the ICSS to consider referral of patients to self-management programmes, since they are supporting people in the time period when our qualitative feedback suggested they might be most beneficial. The checklist that we developed could be appropriate for use by the ICSS, since the intention is that it will be a needs-led service.

The ability of primary care to undertake routine reviews is likely to be more limited than it was prepandemic. 48 We found no benefit of needs-based reviews applied to the whole population of stroke survivors, so any implementation of needs-based reviews in primary care should be focused only on people where there might be benefit.

Full publications

Aziz NA, Pindus DM, Muillis R, Walter FM, Mant J. Understanding stroke survivors’ and informal carers’ experiences of and need for primary care and community health services – a systematic review of the qualitative literature: protocol. BMJ Open 2016:e009244. http://doi.org/10.1136/bmjopen-2015-009244

Pindus DM, Lim L, Rundell AV, Hobbs V, Aziz NA, Mullis R, Mant J. Primary care interventions and current service innovations in modifying long term outcomes after stroke: a protocol for a scoping review. BMJ Open 2016;6:e012840. http://doi.org/10.1136/bmjopen-2016-012840

Pindus DM, Mullis R, Lim L, Wellwood I, Rundell AV, Aziz NAA, Mant J. Stroke survivors’ and informal caregivers’ experiences of primary care and community health services – a systematic review and meta-ethnography. PLOS ONE 2018;13(2):e0192533. https://doi.org/10.1371/journal.pone.0192533

Mullis R, Aquino MRJ, Dawson SN, Johnson V, Jowett S, Kreit E, Mant J; on behalf of the IPCAS Investigator Team. Improving Primary Care After Stroke (IPCAS) trial: protocol of a randomised controlled trial to evaluated a novel model of care for stroke survivors living in the community. BMJ Open 2019;9:e030285. http://doi.org/10.1136/bmjopen-2019-030285

Turner GM, Mullis R, Lim L, Kreit L, Mant J. Using a checklist to facilitate management of long-term care needs after stroke: insights from focus groups and a feasibility study. BMC Family Practice 2019;20:2. https://doi.org/10.1186/s12875-018-0894-3

Aquino MR, Mullis R, Kreit E, Johnson V, Grant J, Lim L, et al. Improving Primary Care After Stroke (IPCAS) randomised controlled trial: protocol for a multidimensional process evaluation. BMJ Open 2020;10:e036879. URL: http://bmjopen.bmj.com/cgi/content/full/bmjopen-2020-036879

Aquino MRJ, Mullis R, Moore C, Kreit E, Lim L, McKevitt C, et al. ‘It’s difficult, there’s no formula’: qualitative study of stroke related communication between primary and secondary healthcare professionals. Int J Integr Care 2020;20:11:1–10. http://doi.org/10.5334/ijic.5465

Lim L, Mant J, Mullis R, Roland M. When is referral from primary care to specialist services appropriate for survivors of stroke? A modified RAND-appropriateness consensus study. Family Practice 2020;21:66. https://doi.org/10.1186/s12875-020-01139-4

Johnson VL, Apps L, Kreit E, Mullis R, Mant J, Davies MJ; on behalf of the MLAS Development Group. The feasibility of a self-management programme (My Life After Stroke; MLAS) for stroke survivors. Disability and Rehabilitation, published online 1 February 2022. URL: https://www.tandfonline.com/doi/full/10.1080/09638288.2022.2029960

Johnson V, Apps L, Carey ME, Kreit E, Mullis R, Hadjiconstantinou M, et al.; on behalf of the MLAS Development Group. The development of a self-management interventions for stroke survivors – My Life After Stroke (MLAS). Disability and Rehabilitation, published online 3 February 2022. https://www.tandfonline.com/doi/full/10.1080/09638288.2022.2029959

Scott Reid P, Neville E, Cater C, Mullis R, Mant J, Duschinsky R. Accounts of preventative coping: an interview study of stroke survivors on General Practice registers. BMJ Open 2022;12:e058441. https://doi.org/10.1136/bmjopen-2021-058441

Blatchford EG, Aquino MRJ, Grant J, Johnson V, Mullis R, Lim L, Mant J. Patients’ experience of and participation in a stroke self-management programme, My Life After Stroke (MLAS): a multimethod study. BMJ Open 2022;12(11):e062700. https://doi.org/10.1136/bmjopen-2022-062700

Conference abstracts

Pindus DM, Mullis R, Lim L, Wellwood I, Rundell AV, Aziz NA, et al. The experiences of and need for primary care and community health services in informal carers of stroke survivors – a systematic qualitative review. Eur Stroke J 2016;1(1_suppl):570.

Moore C, Pindus DM, Kreit E, Lim L, Mullis R, Mant J. Changing care needs of stroke survivors in primary care. Eur Stroke J 2017;2(1_suppl):369.

Pindus DM, Mullis R, Moore C, Kreit E, Lim L, Mant J. Improving communication between generalists and specialists for patient benefit: insights from focus groups on long-term care after stroke. Eur Stroke J 2017;2(1_suppl):369.

Lim L, Pindus DM, Mullis R, Moore C, Kreit E, Mant J. Factors influencing re-referral decisions to specialist care in stroke survivors: insights from focus groups on long-term care after stroke. Eur Stroke J 2017;2(1_suppl):370.

Moore C, Pindus DM, Kreit E, Lim L, Mullis R, Mant JM. Identifying the long-term needs of carers of stroke survivors living in the community to inform a new primary care model. Eur Stroke J 2017;2(1_suppl):371.

Kreit E, Pindus DM, Mullis R, Chowienczyk S, Mant J. Clinical practice guidelines for stroke rehabilitation and long-term management: similarities and differences in recommendations for mood, aphasia and cognitive deficits. Eur Stroke J 2017;2(1_suppl):372.

Chowienczyk S, Nguyen Q, Pindus DM, Kreit E, Mant J, Mullis R. A comparison of national guidelines for secondary prevention of ischaemic stroke. European Stroke Journal 2017;2(1_suppl):347.

Pindus DM, Mullis R, Lim L, Wellwood I, Kreit E, Lim L, Mant J. Primary care interventions for long-term outcomes after stroke: a scoping review of reviews and recent trials. Eur Stroke J 2017;2(1_suppl):453.

Moore C, Pindus DM, Kreit E, Lim L, Mullis R, Mant J. Exploring stroke survivor and informal carer need: informing a new primary care model. Int J Stroke 2017;12(S3):14.

Johnson V, Schreder S, Kreit E, Mullis R, Mant J. A feasibility study of My Life After Stroke (MLAS) – a structured self-management programme for stroke survivors. Int J Stroke 2018;13(S3):25.

Lim L, Mullis R, Roland M, Mant J. Care of post-stroke patients: identifying circumstances in which specialist referral may be beneficial. Int J Stroke 2018;13(S3):30.

Changing needs of stroke survivors in primary care

Needs were categorised into four key areas: physical, psychological, information and health services. Examples of specific needs within these areas are shown in Table 4. The needs changed in importance over the time course since the stroke (see Figure 4). Psychological needs, particularly anxiety and feelings of abandonment, were most apparent in the postdischarge phase. Abandonment is illustrated by the following quote:

Well I thought someone would come in and see if I was alright or if there was anything they could do but we got nothing at all. Did I need this, did I need that ….

FIGURE 4.

Three phases of poststroke recovery.

Physical needs were identified in the longer-term phase by some stroke survivors. It was notable that those survivors who talked about their longer-term physical needs referred to them in the context of their other comorbidities and in comparison to other stroke survivors. Need for information and for health services was apparent at all stages in the recovery process. Illustrative quotes include:

To say I’m getting better, that’s what you want to hear, you know, or why can’t I lift my leg? Well it has affected so and so, so and so, and you can think: oh well, that’s alright, you know.

If you want to see him you’ll have to ring up first thing of a morning … after you’ve done that for about a week and there’s no appointments you say oh forget it. It’s just really difficult to get an appointment.

Needs of carers

These related both to those of the cared-for stroke survivor and to the carer’s own needs, but sometimes the boundaries were blurred between what was a carer need and what was a patient need. An example of a need that related to the stroke survivor:

I had to fight to get my mum physio. I went to the doctor’s to fight for physio because my mums’ right side, literally it was nothing. She would literally just slump to one side.

An example of a carer need, slightly blurred:

He’s never been my husband anymore. I’ve never had any help since then … I was 54 (crying) and I’ve never had a husband anymore. So we just muddle through really.

The key issues from a carer perspective in terms of the stroke survivor were fear of a further stroke, lack of stroke-related information and lack of follow-up. The consequences for the carers themselves were loss of physical and emotional relationships, loss of confidence, anxiety and depression and feelings of abandonment and being trapped in the caregiver role. Indeed, psychological needs were the ones that were most forcibly expressed. In terms of seeking support, the general practice was regarded as the first point of contact.

Key findings

• Thirteen (62%) reviews and 5 (50%) trials reported positive small- to medium-sized effects of interventions on long-term outcomes after stroke.

• Only 19% (n = 4) of reviews and 10% (n = 1) of trials included caregivers.

• Forty-three per cent (n = 9) of reviews and 60% (n = 6) of trials focused on physical function (e.g. mobility, balance, physical fitness).

• Increased interest in psycho-social outcomes was observed in recent trials: 33% of trials compared to 19% of reviews addressed caregiver burden or QoL in survivors.

• Two reviews reported positive effects of information provision on depression in survivors (Smith et al. , 2010) and education and support interventions on QoL in caregivers (Lee et al. , 2007).

• Sixty per cent of positive review and trial findings related to physical (e.g. mobility, fitness, balance) and functional (activities of daily living, ADLs or disability) outcomes.

Intervention process data

This section reports on GP practice, participant characteristics and data on participation at the structured stroke review. Each will be discussed in turn.

Fidelity of design

1. Fidelity of IPCAS training

Checklist

Overall, healthcare professionals found the 15-item checklist of needs useful, albeit with some reservations that there were a lot of items listed. Some HCPs reported that this tool facilitated a focused conversation with stroke survivors and carers, thereby optimising available consultation time:

… with them coming in with that pre- filled in it really did focus one’s mind on areas that were obviously important to the patient. So yeah, it was good.

[GP1_20]

If the patient had ticked items of concern on the checklist, the HCP would decide with the patient whether a referral was required. Depending on the nature of the referral, concerns could be actioned by a clinician within the practice. For example, clinical referrals were made in the usual practice way (see also theme 2), and self-referrals were suggested using the SMT.

Some HCPs also reported that using the checklist encouraged stroke survivors to openly discuss their needs and allowed them to identify the issues that were important to them, making it a patient-centred consultation:

… I found that in a lot of cases it was the first time in ages that they’d actually had time to actually talk about how their stroke had affected them and what it does affect. […] I think the checklist was really good because it opened up the conversation.

[PN1_29]

Differences were observed between stroke survivors’ use of the checklist, with some ticking most items and others hardly any at all, demonstrating the varying levels of patient needs.

I found it good, yeah, how it was planned out in different sections it was quite easy and people understood which bit they had to tick and, if not, they used to ask when they get here and we used to sort them out and that was fine. Yeah, I found it really easy and good.

[HCA1_13]

And that was really interesting because some patients just ticked everything …

[PN1_29]

Those HCPs who found that the checklist was lengthy recalled having to support some stroke survivors with completing the checklist during the enhanced stroke reviews, making time management challenging.

And at the end of the day, you’ve got a patient in front of you. You want to do what’s right for the patient. I understand you have to tick these things. I do understand why. I do get that. But at the same time you think, hang on, there’s a … if you squeeze your time too much you can’t actually give what patients often really, really need.

[RN1_46]

It’s really difficult to steer a consultation in that kind of forum with a list that long, because even the well … it’s like I always say, in practice, when you have the ‘worried well’, they’ve still got a story to tell you …

[PN1_11]

And a lot of patients, they would tick a lot, because they’re obviously issues but they weren’t always prioritised, they’d go through the checklist and they’d say, yes, I’m having trouble here, I’m having … so it was quite difficult to get … to keep pertinent for that top three to four when some patients found that they had a lot of problems, so that made it more difficult for people to prioritise what their main problem was …

[PN1_04]

… if they didn’t bring their checklist, we would go through it at the appointment.

[PN1_31]

HCPs also suggested improvements for the future. These included reminding patients to complete the checklist before the review appointment, reducing the checklist to one page and making clear on the checklist instructions to select patients’ three main concerns, in line with the design of the structured review.

the way that we did that is that [name] sent out the checklist and things with the letter and everything for them to complete, then she also rang them all before they came in, one, to make sure they were going to attend because obviously I was given a half an hour appointment so to make sure that they were actually attending, but also to remind them to fill that in before they came. So the fact that she did that, I think that’s what really helped.

[PN1_19]

But […] even though they don’t need your help now, they want to tell you their story […] they would tick several of the boxes, and then it was a case of saying to them, right prioritise these for me, which ones do you feel are … and trying to just keep them on … you could have spent two hours, sat discussing some of the lists.

[PN1_11]

Using the service mapping tool/use of service mapping tool

Participants had mixed views about the usefulness of the SMT, which was available for healthcare professionals to use during the structured reviews and any other contact with stroke survivors during the course of the study. Some HCPs found it too lengthy and complex:

Yes, that was as clear as mud! That was a very long list of … that definitely could be improved […]Yeah, and it was very small print, and it was very, kind of, you really were having to, you know, it … yeah […]The print was small, there was too much on one … was it a piece of paper or was it a screen?

[PN1_11]

Yeah, I mean there’s quite … it’s quite a big document isn’t it? And it’s unpicking what would be appropriate for that patient. I mean we had the template for different areas.

[RN1_37]

However, on familiarisation with the SMT, other HCPs reported that it was not only helpful but, on familiarisation, became easier to use:

Actually I found them very useful and, off the top of my head, to be honest, I can’t find anything that I’d like to add to it because on the mapping side and information I seemed to have found it [ … ] and if I didn’t then I must have just gone somewhere and got hold of the information and just posted it to the patient anyway.

[HCA1_13]

Direct point of contact

All but one HCP reported that they had not knowingly received any communications from stroke survivors using the DPoC, despite there being a clearly understood process for the practice staff on how to direct and deal with any queries. At some practices, it was expected to generate a heavy workload of patient enquiries:

A GP that was also involved in it didn’t want his point of contact to be put on there, ‘cause, I think, he thought it was more workload. I think, he was … so he said that the nurse would be the point of contact and then I would then go to him with anything that I felt relevant when … but other than once patient, I didn’t have anybody else call.

[PN1_11]

HCPs reported that they had received clear instructions on how to deliver DPoC, for example, using the template and a named contact. Some even provided DPoC information in print to stroke survivors.

None of the patients had made any contact with us on the direct point of contact, because we have a special template on for that, and all the staff were educated as to what to do if a patient who is on the IPCAS trial did phone in and needed to speak to either myself or the nurse who was involved, and we’ve not had one phone call about that. So that was interesting.

[GP1_20]

[…] we did, we printed out our name and our manager put a dedicated line to the stroke … but, yeah, we didn’t get any calls … I always gave them the number and said, if there was anything that you wanted to bring up that you haven’t brought up then just drop us a line, ask for myself or [other HCA], who was the other person, and we’ll be happy to return your call any time [...]. But we never got any so I assume everyone was happy, I don’t know.

[HCA1_13]

R: No, no, I’ve had no messages because I gave my contact and also the practice partner, so if I wasn’t around then she would take the messages and then either deal with them or come back to me, yeah.

I: And she had none to pass on to you..?

R: No, no. Interesting, isn’t it?

[PN1_04]

We had some little cards made up in the practice, that we gave out at the consultation … And, the contact was for me and for the other nurse but then, I think, hers got taken off ‘cause she was off on long term sick.[…]

One HCP felt that IPCAS as a whole intervention was a success because they confirmed they were the DPoC with every study participant, yet received no further patient queries going forward, which to them indicated having met stroke survivors’ needs:

I know it must have been successful because although I religiously told every single person that I was the main point of contact and they knew how they could contact me and request a telephone call back, I haven’t had anybody ring me. […] All I’m saying is they must have got all their relevant needs met, because otherwise they would have been in contact, wouldn’t they? […] I again dutifully told them, every single one of them that all they needed to do, as I was the main point of contact for any queries, if they rang in and left a message, I would ring them back, and I have protected telephone slots every day in the morning and in the afternoon, but not one.

[PN1_19]

Communicating follow-up services (My Life After Stroke and Action Planning)

Healthcare professional role

This theme concerns factors identified by HCPs that made a difference to how the intervention was delivered.

A distinction in IPCAS delivery emerged between the roles of the practice nurse and research nurse. Where a practice nurse conducting reviews was an integral member of staff within the surgery, there was often an already-established rapport with the patient. Having some familiarisation with patients and prior knowledge of the patients’ needs as a whole led to a smoother delivery.

I think it’s good to have a person that you know, most of the patients I would say know that if they’ve got anything to do with their heart circulatory stuff they’ll give me a shout …

[PN_04]

I think, because I’ve seen these patients or I had seen them for the last five, six years, a lot of information that we were trying to find out, we had already discussed in previous appointments, not maybe as much in depth …

[PN_31]

Research nurses, on the other hand, reported experiencing some challenges to delivering IPCAS. These included not knowing the patient beforehand, taking time from the structured review to build a rapport and not having prior familiarisation of IT hardware and administrational processes.

I’m just there for this brief period of time and that I don’t get any of the follow up.

[PN_04]

And this is where probably … this is where I think it’s a … a little shame, if I had been a practice nurse at any of these different surgeries, then maybe I would have, the likelihood of me seeing these patients again would have occurred.

[RN_37]

(Or this quote – same RN)

But I did think it would have been kinder, maybe more direct, if I’d have just done that referral myself to them. So, whether it may have been just a … a patient could just have gone to see that person direct, say if it was a physio or a speech therapist. But then, you know, I … as not working in that practice regularly, I had to ensure that it was an appropriate referral too. So, it kind of made sense to go either via the GP or the practice nurse, yeah.

[RN_37]

It wasn’t all negative though; one HCP felt that it was a good thing to be a research nurse as she had more time for the patients:

… as the research nurse doing the research is our, actually our job and our priority […] I mean, fortunately, as a research nurse, we can just concentrate solely on that research, yeah.

[RN_37]

Having the clinical autonomy to make referrals also appeared to make a difference to how an action plan could be executed. For example, hospital-based referrals appeared to be accepted mainly from clinicians, suggesting other staff were restricted to what they could achieve within the consultation:

… it has got to be through a GP, because a lot of the hospital-based clinics don’t like nurse referrals, so, you know.

[PN_04]

HCPs in the control arm reported a similar role limitation of being reliant upon GPs to refer patients to other services:

I guess I do rely a little bit heavily on a GP, so because of the limitations of my role, if I have a patient that comes to see me, and they have these problems that I cannot fix, then I refer them to the GP, and I then assume that the GP is going to potentially sort it out, or signpost them in the right direction.

[PN1_07]

Healthcare professional experience

HCPs with prior clinical experience said that it helped in delivering the IPCAS review:

This is a new study that they wanted me to take on, previously I worked in the (hospital) for 19 years in cardiology. […] I enjoyed doing the study, I learnt a lot and I found it very interesting and very helpful for the patients that I actually saw.

[HCA_13]

Inevitably you do find whatever experience you’ve had in your personal career, you bring it in anyway don’t you …

[RN_46]

Experience of technology use between colleagues appeared to have influenced the smooth delivery of the intervention. For example, HCPs with limited IT experience reported that they struggled with the templates, whereas others said that they found them straightforward and easy to use.

… what we did is we put it on the computer. One of my colleagues is a computer geek – he sets up templates on our computer system, so we had an IPCAS template, we’ve actually still got it on the system.

[GP1_20]

Descriptions of participating in training

All HCPs who attended the training sessions described how training met their needs adequately, giving them confidence to deliver the structured review.

… after a couple I found that all the stuff that I had been taught and I’ve learnt actually came in useful and it all came flooding back after a couple […] I was impressed with the training that I did get and how informative it was.

[HCA_13]

Those HCPs who missed some or all of the training felt they ‘caught up’ adequately with what was required:

I was a late attender, I did go to the IPCAS meeting, but due to the pressures in primary care, I think, I got there about 45 minutes in, so missed a big chunk and then had patients booked. So, kind of, had the catchup brief and picked up on the, I think, there was a PowerPoint presentation, if I remember correctly.

[PN_11]

One HCP had little recollection of the training and misremembered some aspects. This was reflected in some miscommunication within the consultation:

I knew there were going to be some, like, ‘clinics’, or like an outpatients thing, [...] I wasn’t really very sure who was running them or what it would entail in the groups that they were being invited to … so I don’t know if that just hadn’t been filtered down to my level, or … I can’t remember. They probably did say; they wanted to be doing the presentation just before I had my lunch! [laughs].

[PN_33]

Perceived benefits of training

HCPs reported that having ongoing support and knowing the research team was always on hand helped build and maintain confidence.

But I know that when I started doing them, I felt well-prepared, and I think I’d spoken to somebody beforehand and I definitely got a lot of support during, so if I had any queries, there was always someone there that I could get support from.

[PN_14]

as soon as [researcher’s name] came in and just explained stuff […] I felt a lot more confident. And I felt confident in myself that actually I can go back and I can actually do this and feel like I’ve done it properly.

[HCA_13]

Benefits also included having a clear understanding of the research methodology.

Well, I think for me personally, it was a nice in depth look at the post stroke pathways and really how when patients are discharged from secondary care, they are very much dumped in primary care and there’s not too much happens in between.

[PN_11]

HCPs reported that being part of the review process enhanced their knowledge of the heterogeneity of stroke survivors:

It was quite an eye-opener for me […] to see how patients’, you know, experiences were very different … Yeah, yeah […] their journey post stroke had been very different.

[RN_37]

Did not adopt IPCAS

Some HCPs reported that they did not adopt IPCAS and reverted to their usual practice model of care post study. Those HCPs that did this, however, were able to use their knowledge to enhance their practice:

… going forward we haven’t continued with the template that we were given, we reverted back to just a QOF one […] No, but it’s still good, in that it’s probably nudges me to ask … to delve a little bit deeper to what is just basically on the QOF, kind of, checklist. And, I will now ask about personal relationships and things that I, perhaps, wouldn’t have done prior.

[PN_11]

New (IPCAS) practice normalised

Aspects of the intervention have been kept and are still being used within normal practice.

As I said, we’ve still got the IPCAS template on the computer, so I’ve still got that as an aide memoir, so I will be using that when I next see stroke patients just to keep my mind on what we went through. So yeah, it has helped from that point of view. […] Just an aide memoir to go through everything it was brilliant.

[GP_20]

Measuring costs

A UK NHS perspective was adopted for the costing component of the analyses, with costs expressed using 2018–9 Great British pounds.

Intervention costs

The resources used to implement the intervention were costed using a micro-costing approach that considered the time needed by medical personnel for their training and for reviewing the patients’ needs, alongside the attendance of participants to the MLAS self-management programme.

Costs of health service use

At 12 months, participants were asked to complete a questionnaire to retrospectively record their contacts with the secondary care services of the NHS over the course of the trial. Each patient was asked to provide data on the frequency of visits to outpatient clinics and their hospitalisations due to any reason related to their stroke, including the length of stay. The hospitalisations with a length of stay ≤ 1 day were treated as day cases. The questionnaire collected information on contacts with social services, including home care, private home help, day centre, meals on wheels, laundry service and respite care. Data on the access to primary care were obtained from the medical records provided by the practices involved in the study via SystmOne. These were clinic visits, home visits and phone calls with GPs, practice nurses, healthcare assistants, psychologists and community services. Practices also collected information on prescribed drugs to patients at 12 months, and it was assumed that these medications were prescribed for 6 months and the dosage followed the defined daily dose. Drug use was costed using the unit costs publicly available from the British National Formulary.

Measuring outcomes

Individuals completed the five-level version of the EQ-5D-5L at baseline, 6 and 12 months. The EQ-5D-5L is a generic health-related quality-of-life instrument which contains questions on five domains: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Each domain has five levels: no problems, slight problems, moderate problems, severe problems and extreme problems. Individuals identify which level they are in for each of the five domains, placing them in one of 3127 unique health states (55 plus unconscious or dead), which can then be converted into a single score by using a tariff estimated using data from a sample of the UK population and a crosswalk algorithm. The resulting utility score is anchored at zero and one, which reflect death and perfect health, respectively. Negative scores represent health states considered to be worse than death. A linear trend was assumed between a patient’s scores at each time point, and the area under the resulting curve gave the number of QALYs experienced by that patient to 12 months.

Statistical analysis

It was assumed that the probability of data being missing depended on observed data and not on unobserved data, namely, the data were missing at random and could be predicted by using the information collected during the trial. Missing data were imputed using multiple imputation, a regression-based approach to predict m values for each missing data cell. This accounts for the uncertainty generated in the process and allows both complete and incomplete variables to be used to predict the values of missing data cells. Fifty imputed data sets were used in our analyses, and Rubin’s rules were used to summarise their data. Thus, this approach accounts for the variance between and within imputed data sets.

Resource use, costs, EQ-5D scores and QALYs were summarised using mean, SD and 95% CIs for each study group. Ninety-five per cent CIs in the complete case analysis were obtained using 5000 non-parametric bootstrap replications; in the multiple imputed data, these were obtained evoking Rubin’s rules and assuming normality. Mixed-effect models were used to model each of the dependent variables with a random intercept for each centre and fixed effects for the intervention. In an ad hoc analysis, we controlled for EQ-5D-5L scores at baseline. To account for the skewness of data, the CIs of costs and QALYs were obtained by sampling 5000 non-parametric bootstrap replicates, accounting for the clustered nature of the trial. We therefore calculated 5000 sets of costs and QALYs for the intervention and control and used these to calculate 5000 ICER values, representing the incremental cost required to obtain one additional QALY when moving from usual care to a new intervention. The ICERs were plotted as single points on the cost-effectiveness plane, a two-dimensional figure with four quadrants representing all possible outcome combinations of cost and QALYs differences.

A study’s ICER is commonly compared to the maximum reimbursable ICER (i.e. cost-effectiveness threshold), which for the UK ranges from £20,000 to £30,000 per QALY. Calculating the proportion of the bootstrapped simulations with a corresponding ICER below £20,000 per QALY, we estimated the probability that the new model of care will be cost-effective when compared with usual care (i.e. the cost-effectiveness probability). Varying the cost-effectiveness threshold and re-estimating the cost-effectiveness probability, enabled the construction of a cost-effectiveness acceptability curve (CEAC). This represents the probability that the new model of care will be accepted based on different values of the cost-effectiveness threshold.

The prespecified analyses included the cost-effectiveness assessment using multiple imputed data and unadjusted estimates. A set of ad hoc analyses was performed to determine the impact of (1) adjusting for EQ-5D-5L scores at baseline and (2) using only complete cases.

All statistical analyses were performed using Stata SE v14.0 (StataCorp, College Station, TX, USA).

Resource use – intervention

Resource use – healthcare services

Cost

Quality-adjusted life-years

Resource use

Cost

Quality-adjusted life-years

Complete cases: unadjusted

FIGURE 14.

Cost-effectiveness plane (a) and cost-effectiveness acceptability curve (b): complete cases – unadjusted.

Complete cases: adjusted

TABLE 45 - Incremental analysis: complete cases – adjusted for EQ-5D-5L scores at baseline

	Control	Intervention	Difference	ICER
Cost	1086.246 (831.326 to 1402.219)	1302.617 (918.993 to 1764.421)	216.371 (−272.306 to 741.538)
QALY	0.74 (0.72 to 0.76)	0.746 (0.726 to 0.766)	0.005 (−0.007 to 0.018)	39,532.387

FIGURE 15.

Cost-effectiveness plane (a) and cost-effectiveness acceptability curve (b): adjusted for baseline EQ-5D-5L scores.

Multiple imputation: unadjusted

FIGURE 16.

Cost-effectiveness plane (a) and cost-effectiveness acceptability curve (b), multiple imputation, unadjusted.

Multiple imputation: adjusted

FIGURE 17.

Cost-effectiveness plane (a) and cost-effectiveness acceptability curve (b), multiple imputed data, adjusted for baseline EQ-5D-5L.