Theoretical frameworks informing this process evaluation

Two theoretical frameworks with emphasis on translating evidence in the real-world setting were used to inform our methods.

The first was the RE-AIM (Reach, Effectiveness, Adoption, Implementation and Maintenance) framework. It has been used by researchers in the development, evaluation and dissemination of research, using qualitative and quantitative methods.18 The framework's five domains highlight the need to look into the proportion and representativeness of the participants involved in the intervention, the impact of the intervention, the fidelity of trial implementation and at how a programme may be institutionalised into organisational practice after the conclusion of the trial. Thus, pending ATTEND's results, and resources permitting, there may be scope for a post-trial evaluation to explore how and why there were changes in practice and policy.19

The second framework that informed our methods was the Realist framework (Context, Mechanism, Outcomes).20 It has been successfully used in process evaluations as a theoretical basis for identifying potential causal mechanisms for how an intervention works for whom, under what contexts and therefore fosters uptake of research-based knowledge into practice.21 ,22 Given the complexity of the ATTEND trial from contextual macrolevel factors, such as different socioeconomic demographics, cultural differences, health system funding structures at each state, to microlevel factors such as literacy of patients and carers; the realist framework would be valuable in framing our analysis in understanding the mechanism within the ‘black box’ of the intervention.23

Our process evaluation framework and the hypothesised causal mechanisms of ATTEND

The framework (figure 2) highlights the key questions for the process evaluation, incorporating the above theories and our work plan. We have briefly outlined our hypothesised causal mechanisms of ATTEND's complex intervention within the framework as per the MRC's guidance on process evaluations.19 The framework is divided into sections of context, implementation of the trial, mechanisms of the trial and outcomes of the trial with the final objective of reducing the burden of stroke through a sustainable model of care.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

The ATTEND process evaluation framework. The process evaluation components are highlighted in blue boxes—exploring contextual factors, the implementation of the ATTEND trial, mechanisms of impact from the intervention. Questions informed by the RE-AIM and Realist framework fit within these components. These components are informed by the causal assumptions of ATTEND intervention and will inform the interpretation of the primary and secondary outcomes. ATTEND, fAmily led rehabiliTaTion aftEr stroke in INDia; RE-AIM, Reach, Effectiveness, Adoption, Implementation and Maintenance.

Mixed methods used to address the aims of the process evaluation

1. To explore if the trial was delivered as intended (eg, fidelity and dose)

This will be achieved through an analysis of the administratively collected process data (eg, competency measures, frequency of training), electronic case record file, quantitative data and activity logs (eg, the measurement of the total time spent per patient in both arms of the trial with the usual care physiotherapist, the number of meetings and time spent with each family caregiver by the stroke coordinator during their inpatient stay and after discharge). If the analysis provides any evidence of variation between the active and control groups in terms of amount of time spent on different tasks, we will investigate and provide further training if required.

1. To understand whether, how and why the intervention had an impact through exploring the providers', patients' and carers' perspectives of usual care and of the intervention

   • Non-participant observations of all the sites to understand the contextual factors at each site, which may help explain variations in outcomes. Our observation template has been adapted from a process evaluation of a similar intervention24 and includes issues on trial set up, trial implementation and impact. See online supplementary additional file 1 for a copy of the template.

   • Semistructured in-depth interviews will be conducted with patients, carers, stroke coordinators and the stroke unit staff of hospitals to understand their perspectives of the causal mechanisms of the intervention and to seek their suggestions on how the intervention could be scaled up or rolled out. Methods for our qualitative interviews have been described as per the consolidated criteria for qualitative research:25

     1. Research team and reflexivity: The research team is multidisciplinary and comprises of the trial's chief investigators, the clinical trial coordinating staff and research fellows from India, UK and Australia. The team has background in medicine, physiotherapy, health economics, pharmaceutical trials and public health; with varied experience in qualitative research. In particular, three of the team members (HL, MA and SJ) have training in qualitative methods. HL who coordinated the team, organised face-to-face qualitative workshops to train the rest of the team prior to the conduct of interviews and will take a leading role in analyses. A team of six people will conduct the interviews. The team members, who are part of the clinical trial management team, have a good understanding of the trial and are known to the principal investigators, stroke coordinators and blinded assessors but will not have established relationships with the patients and their carers.

Semistructured interview guidelines based on our study objectives have been developed and pilot tested. The key areas covered include overall views of the health and socioeconomic impact of stroke on patients and families, stroke management in India, acceptability of the family-led rehabilitation intervention, the general healthcare experience of patients and their families and translation of the intervention into practice and policy. Early findings will be discussed with the study team, and minor changes made to the interview guidelines if needed. (See online supplementary additional files 2–4 for the interview guidelines for the health provider, carer and patient, respectively.)

1. Study design: Sampling: Participants for the interviews will be recruited using maximum variation purposive sampling. Variables to be considered for patients and carers are: usual care versus intervention arm, primary outcome (as measured by the mRS at 6 months), gender and the region of India they are from. Variables to be considered for stroke unit providers: healthcare roles (ie, neurologists, physiotherapists, nurses and the stroke coordinators), private versus public government hospitals and the region of India the hospitals are located.

A list of patients and caregivers who match our sampling criteria will be generated. They will be contacted by the local site staff (either the stroke coordinator or the ‘blinded assessor’) and where relevant, reasons for not participating will be elicited. They will be formally consented by the interviewer face-to-face. Interviews with carers and patients will be conducted either individually or with both together either at the participant's home or at the hospital. These interviews will be conducted in local languages and the services of an interpreter may be required. The benefits of interviewing the patient and carer separately would be to gather perspectives which otherwise may not be shared should the other be present. Healthcare providers will be invited to participate in interviews either by a letter or in person by the clinical coordinating team during their site visits, and conducted in English. Written informed consent (see online supplementary additional file 5 for a copy of the form) will be obtained from all interviewed participants. As per qualitative research methods,26 analysis is iterative and thus the interviewer will carry out preliminary thematic data analysis at the end of each interview and discuss any highlights with the rest of the interview team. For example, the findings from the pilot interviews were discussed during the qualitative workshop, in order for the team of interviewers to explore emerging themes in subsequent interviews. It is estimated that at least five sites would be sampled, with 3–4 health providers and about 4–6 patient/carer dyads interviewed from each site. According to the sampling matrix, we will interview equal numbers from both usual care versus intervention arm, and also include sampling for gender. For example, at each site, two usual care dyads and two intervention group dyads will be invited to participate.

In addition, some of the stroke coordinators and the independent assessors at the other sites will be interviewed. That is, an estimated 80–100 interviews will be conducted though the final numbers will be determined by saturation of themes and resources permitting. Interviews will be conducted face-to-face, audio recorded and professionally translated and transcribed verbatim. These will be uploaded into a software program NVivo V.9 (QSR International, Melbourne, Victoria, Australia) to assist with data management.

1. To explore if the results are generalisable, scalable and sustainable through exploring health providers', patients' and carers' experiences of the intervention and its perceived impact. This will include an evaluation of costs from the societal perspective including the health system, and for patients and their families. (note this is separate to the formal cost-effectiveness study which has been set out in the original trial protocol)6

   • One of the key domains in the semistructured interviews will be exploring participants' perspectives of what they expect post-trial in relation to this intervention. The findings related to these questions will address this aim.

   • Cost components of the intervention which will be relevant to implementation outside a trial setting will be considered. This will be extracted from the programme budget and trial contracts and includes, for example, costs of employment of a physiotherapist to implement the intervention, travel costs of the home visits and costs of any educational material required as part of the intervention. Relevant questions asked by the blinded assessor at the 3-month and 6-month follow-up visits include loss of family income (eg, number of hours of work taken off) due to carer's additional responsibility and medical costs. (Participants are encouraged to keep receipts of related medical costs.)

4 To explore implementation barriers and facilitators of a complex intervention by an international collaboration.

A research fellow not involved in the implementation of the trial will conduct focus group interviews and semistructured interviews with members of the clinical trial coordinating team and also with the trial investigators in regards to perceived barriers and facilitators to trial implementation. Findings from these interviews in addition to relevant findings from the interviews with the health providers, patients and carers will inform lessons learnt from this trial for future research collaborations.

Analysis plan for the process evaluation

Thematic analysis will be used for the qualitative analysis to code closely to the data and establish themes within the subheadings of the process evaluation framework (figure 2).26 Constant iterative comparison between sources, for example, patient, carer and health provider will be carried out in order to identify common, as well as distinctive, themes.26 Contextual information from observations, other process data and costs to patients and families will be used to triangulate the emerging themes.27 The quantitative process data will help inform fidelity and the time logs of the stroke coordinator will provide descriptive data on dose. Fidelity data will be reviewed at six monthly intervals.

The process evaluation framework (figure 2) will aid in the analysis by triangulating the process's quantitative data with the relevant qualitative data addressing the questions within its subheadings.27 For example, under the heading ‘implementation—fidelity and dose’, a specific question would be whether usual care is provided equally in both arms of the intervention, and thus the quantitative process data would be the time spent by the usual care physiotherapist and should be almost equivalent in both arms documented in the logs (or not), and the qualitative data would include, for example, the usual care physiotherapists' responses as to whether they did treat all the patients equally, or the neurologist's description of what happens to the study participants.

Other forms of triangulation to increase the reliability of our results include the sampling of different perspectives, that is, patient, carer, neurologists and stroke coordinators and also through the triangulation of different analysts in the team who bring their own cultural backgrounds, academic experience (eg, rehabilitation medicine, pharmacy, physiotherapy) and knowledge about different aspects of the trial into the analysis.26

In line with the MRC guidelines, the process evaluation data should be analysed prior to knowing the trial outcomes, first to remove bias, though there is a role for post hoc exploration of reasons for trial outcomes.2 First, we will analyse our process evaluation data iteratively on an ongoing basis. If there are any process issues which would impact on trial integrity that need to be addressed, these will be fed back through the usual management communication channels.28

The framework serves as a template to consolidate the findings, and will be a dynamic structure with changes to be made if required. This means that our understanding of the causal mechanisms of the intervention may change with the iterative analysis of the process evaluation data.

There will also be a post hoc examination of the process evaluation findings, in light of the main results of the trial.29 For example, in our experience, our assumption is that early supported discharge, as part of the intervention, will decrease costs to the system and family, enable early patient rehabilitation which may improve patient recovery (primary outcome) and result in shorter hospital stays in the intervention arm. However, in piloting our observation template at one site, we discovered that there was shortage of beds such that at that government hospital patients were discharged at the earliest possibility, for example, when they were medically stable. This may perhaps be different to developed country settings, such as UK, and ultimately may explain potential divergences in the findings of this study to that of a recent meta-analyses of rehabilitation trials (which showed positive results of early supported discharge).30 ,31 A major consideration in the process evaluation therefore may be regarding the length of stay in hospital. The examination of such secondary outcomes and contextual findings from the process evaluation is an example of how we could gain a deeper understanding of the assumed causal mechanisms of the ATTEND intervention (as depicted in the logic model in the overall process evaluation framework). Such insights will help inform the final logic model of how the intervention truly impacted the trial effectiveness outcomes, and inform the generalisability of the intervention.

Ethics

Ethical approval for the trial and process evaluation has been obtained from Research Integrity, the Human Research Ethics Committee at the University of Sydney and at each local site. See online supplementary additional file 6 for the details of ethic committees at the local sites.

The ethical implications of this scope of research include the considerations of NHMRC ethical guidelines such as research merit and integrity, beneficence and respect in relation to qualitative methods.32

Dissemination

Our process evaluation aims to complement a robust blinded outcome evaluation of the trial end points and inform our stakeholders how, for whom and why this model of family led rehabilitation could have an impact. This is especially relevant for India which is in ‘epidemiological transition’ with a diverse sociodemographic profile, and an increasing burden from non-communicable diseases which could result in greater health inequity.4 ,16 A stated aim in India's key national strategy for non-communicable diseases is to increase access to healthcare to 80% with costs not being a key barrier. Strategies such as the family-led rehabilitation programme, which marshal family and community resources in the care of patients with chronic conditions will, by financial necessity, play a greater role in the future. In low-income and middle-income countries, the available human capital can form one approach to reducing the evidence practice gap, compared with the high cost alternatives available in established market economies. This process evaluation will provide insights into how such a programme can be implemented in practice and brought to scale.

The Indian Institute of Public Health has been allocated resources for the dissemination of results through the engagement of local policymakers and health practitioners. Apart from stakeholder engagement, dissemination of our findings globally will be accomplished through publishing our results in relevant journals and conferences to build on the literature in providing affordable, holistic and accessible stroke rehabilitation models of care.

We describe our protocol to encourage development in process evaluation methodology, transparency in reporting and to build on this emerging area of health services research which is much needed in addressing the complicated global health needs through sustainable, patient-centred and evidence-based complex interventions.