INTRODUCTION

The role of evidence-based practice in enhancing intervention outcomes for individuals with communication challenges is widely acknowledged and highly valued within the literature (Bernstein Ratner, 2006). Although Sackett et al. (1996) identified multiple forms of evidence for clinicians to integrate into their practice, the term evidence-based practice has become most synonymous with research or efficacy evidence i.e. the result of efficacy research such as systematic reviews and randomized controlled trials (McCurtin et al., 2019). Less attention has been paid to other forms of knowledge such as patient-based evidence (knowledge from patient experience, e.g., a client's intervention preferences) and practice-based evidence (knowledge from clinical experience, e.g., a clinician's opinion) (Buetow & Kenealy, 2000; Connery et al., 2020a; Greenhalgh et al., 2014; McCurtin et al., 2019). More recent frameworks of evidence-based practice have highlighted the need for inclusivity of all evidence forms to support clinical decision-making and treatment evaluation (Buetow & Kenealy, 2000; McCurtin et al., 2019). The total evidence and knowledge approach (TEKA) proposed by McCurtin et al. (2019) is one such framework. It acknowledges and values multiple forms of evidence or knowledge such as the three pillars of evidence (research-, practice- and patient-based evidence), and expands the definition to include other forms of evidence such as contextual evidence (e.g., intervention cost, resources and policies) and judico-ethics based evidence (e.g., a country's legal system and legal principles such as do no harm). This framework provides guidelines for researchers and clinicians to collect an array of information to enhance programme design, evaluation and intervention outcomes for clients with communication difficulties. While efficacy evidence has significantly advanced the field of stuttering, integrating multiple evidence forms is advocated to ensure best clinical practice.

A synthesis by Johnson et al. (2016) exploring clients’ and clinicians’ experiences of non-pharmacological stuttering interventions highlighted the reduced range of knowledge forms in the field of stuttering, particularly patient and practice evidence. The findings from this review identified no studies investigating the experiences of healthcare professionals delivering interventions and only three articles examining adult experiences of intervention over a 25-year period. In comparison, systematic reviews of efficacy evidence point to significantly more studies examining stuttering intervention efficacy (Baxter et al., 2015; Connery et al., 2020b). Some recent studies have demonstrated the value in collecting other forms of knowledge such as patient- and practice-based evidence to inform clinical decision-making and ensure intervention efficacy for adults who stutter (Connery et al., 2020a; Sønsterud et al., 2020). In their qualitative examination of the perspectives of expert academics and clinicians on the components of effective intervention, Connery et al. (2020a) demonstrated the need for individualized intervention that considers multiple factors beyond therapeutic technique. These included person-related variables such as client motivation or clinician's knowledge, and therapeutic alliance to enhance treatment outcomes for adults who stutter. Additionally, Sønsterud et al. (2020) demonstrated the effectiveness of a multidimensional individualized stuttering intervention that incorporated the therapy preferences of adults who stutter (i.e., feedback from participants on how useful they found the therapy elements and how frequently they had implemented them outside of the clinic), resulting in a significant reduction in the impact of stuttering for participants. Many efficacy studies, although often stating they have incorporated the goals and preferences of adults who stutter, are less explicit about how this has been achieved. Best practice would suggest that the therapeutic process is a complex one involving multiple factors that influence empirically–driven intervention approaches. Multiple evidence forms must therefore be considered to inform intervention design, support best practice, maintain clinical relevance and guide future stuttering research.

Recent systematic reviews of the efficacy evidence for stuttering interventions for adults have identified a wide range of efficacious therapies (Baxter et al., 2015; Connery et al., 2020b). Nevertheless, challenges with stuttering treatment continue to be reported by both adults who stutter and speech and language therapists (SLTs). Adults who stutter have highlighted high unsuccessful treatment and relapse rates, as well as a failure of SLTs (the primary intervention providers) to explore psychosocial needs in treatment (Johnson et al., 2016; Yaruss et al., 2002). This highlights the need for efficacy studies to measure the long-term outcomes of interventions and to integrate the psychosocial aspects of stuttering into treatment and outcome measurement, something which remains a limitation of many intervention trials (Connery et al., 2020b). Additionally, the principles of person-centred care mean it is imperative for SLTs to identify each individual client's values and treatment preferences to guide the intervention process and to enhance treatment outcomes. Challenges with stuttering treatment have also been voiced by SLTs, who have reported discomfort with working with individuals who stutter, and reduced knowledge of stuttering and its origins (Tellis et al., 2008, Byrd et al., 2020). These challenges identified by both stakeholder groups suggest that a timely revisioning of stuttering intervention is required. One way to achieve this is to utilize the expertise of key stakeholders to identify and achieve consensus on core stuttering intervention components. This will inform the design of new interventions for adults who stutter and guide the development of a checklist for SLTs to use when designing intervention plans for individual clients.

Core intervention components are elements of an intervention that are deemed essential and indispensable. They include theory-based and empirically derived principles, intervention practices, contextual factors (e.g., where does the intervention take place) and structural elements (e.g., the format of the intervention) that are required to produce desired outcomes (Blase & Fixsen, 2013; Damschroder et al., 2009). Various benefits of defining core treatment components have been outlined by Blase and Fixsen (2013) and include directing often limited implementation resources to targeted areas (i.e., the core components), supporting better treatment outcomes, increasing the ability to replicate an intervention in different settings and informing standards of practice. Involving key stakeholders including researchers, clinicians and clients in this process facilitates the co-production and co-learning of knowledge between partners (Wallerstein et al., 2017; Wight et al., 2016). Additionally, co-production improves the fit of the intervention with the client group's needs and enhances the uptake by clinicians and policymakers (Wight et al., 2016). Involvement of both adults who stutter and SLTs allows the retrieval of practice- and patient-based evidence for core stuttering intervention components and this provides the opportunity for collective agreement on key components.

This study aims to build consensus on the core components of stuttering intervention for adults using an expert panel of adults who stutter and SLTs who have experience of providing stuttering intervention. It employs an e-Delphi technique (Keeney, 2009) to obtain consensus among stakeholders on the core components of stuttering intervention for adults. Consensus-building is a process that produces ideas and collaboratively solves problems about a specific issue (Keeney et al., 2011). The Delphi technique is a widely used consensus-building method that aims to achieve group consensus on a specific topic in an iterative and structured manner (Diamond et al., 2014; Keeney et al., 2011; Trevelyan & Robinson, 2015). It is an appropriate methodology to use when there is incomplete knowledge or a lack of agreement on a specific topic because it provides the opportunity to represent the views of multiple stakeholders (Powell, 2003). Given the limited patient- and practice-based knowledge available in the stuttering literature, this methodology was appropriate for the current study's aim. It represents the first employment of e-Delphi technique with an expert panel of adults who stutter that is known to the authors. The results of this study will contribute to the development of a guiding framework for the design and development of evidence-based stuttering interventions for adults and support clinicians working with this population.

METHODS Research design

This e-Delphi study consisted of three rounds: Round 1 aimed to identify key stuttering intervention components through engagement with a group of stakeholders; Rounds 2 and 3 aimed to reach group consensus on the components that are essential to stuttering intervention. Ethics approval was obtained for this study through the University of Limerick Faculty of Education and Health Sciences Research Ethics Committee (2020_06_05_EHS).

Participants

The Delphi technique uses non-probability sampling to recruit an ‘expert’ panel (Trevelyan & Robinson, 2015). In this study, adults who stutter (aged 18 years and over) with past experience of stuttering intervention, in either childhood or adulthood, were recruited for the panel. This facilitated participants’ reflections on past treatment experiences to inform their responses to the e-Delphi study. SLTs with experience working with adults who stutter (minimum of 2 years), and adults who stutter who were qualified SLTs and had a history of intervention were also recruited. This study was conducted in Ireland, with adults who stutter recruited via purposive, convenience and snowball sampling using national and international stuttering associations (see Table S1 in the additional supporting information), a North American stuttering treatment and research institute, social media, and personal contacts of the first author (A.C.). SLTs were recruited via purposive, convenience and snowball sampling using national SLT associations (see Table S1 online), international treatment centres for stuttering, national and international special interest groups (SIGs) for stuttering, social media and personal contacts of the authors (A.C., A.M.).

There is currently no agreement on the number of participants required to comprise a representative sample in a Delphi Survey; however, eight to 12 participants have been recommended for a homogenous sample recruited using specific criteria (Keeney et al., 2011). As this study comprised of two potentially separate homogenous samples, that is, adults who stutter and SLTs, recruitment continued until a minimum of eight participants for each group had agreed to participate. Participants made contact with the first author by email to express their interest in participating in the study and they were subsequently emailed a participant information sheet that included details of the study aims, methodology and benefits of participating.

Questionnaire development

The Round 1 questionnaire was developed from findings of three previous research phases which included a meta-ethnography on the lived experience of stuttering, a systematic review and meta-analysis on stuttering intervention effectiveness, and semi-structured interviews with expert academics and clinicians in the field of stuttering (Connery et al., 2019, 2020a, 2020b). The questionnaire consisted of 25 statements generated by two authors (A.C., A.M.) to represent the themes, subthemes and categories identified in the results of these studies (see Table S2 in the additional supporting information). These statements were plotted onto Blase and Fixsen's (2013) definition of core components of evidence-based interventions, which are theory-based and empirically derived principles, contextual factors, structural elements and specific intervention practices. There were 14 statements listed as principles and 11 listed as intervention practices. These initial statements provided a foundation for participants to understand the meaning of core components and to support their identification of additional components that were not represented. As no statements were plotted onto contextual factors or structural elements, a definition was provided to support participants in identifying any essential components within these categories (see Table S2 online). The questionnaire was piloted with one adult who stutters and one SLT with experience working with individuals who stutter, both of whom were recruited using convenience sampling. Feedback from this exercise resulted in the wording of two statements being modified to clarify meaning, as well as an additional written reminder being added to advise participants that their level of agreement for statements was not required for the current round (Round 1) but would be requested during the next round (Round 2).

Data collection and analysis

This e-Delphi Survey was conducted using three rounds of online questionnaires developed and distributed using Qualtrics (Qualtrics, 2020). Three questionnaire links representing each round were emailed to participants over a period of 3 months. Each questionnaire was open for a period of 2 weeks, with a reminder email for completing it sent 1 week after the initial email. Only the participants who had completed Round 1 were emailed the link for Round 2, and only those who had completed Rounds 1 and 2 were sent the link for Round 3. Demographic data, including age, gender, country, history of stuttering intervention received, and years working as an SLT with adults who stutter, were collected and inputted into IBM SPSS Statistics, version 26, which produced descriptive statistics (Table 1).

TABLE 1. Demographic data of the participants in the Round 1 questionnaire Demographic variable Adults who stutter (n = 35) SLTs (n = 13) Total (n = 48) Age: mean (SD); range 44.31 (17.80); 19–89 39.45 (11.64); 25–58 43.07 (16.46); 19–89 Gender: n (%) Female 9 (25.7) 9 (69.2) 18 (37.5) Male 26 (74.3) 4 (30.8) 30 (62.5) Continent residing in …: n (%) Oceania 1 (2.9) 2 (15.4) 3 (6.3) North America 24 (68.5) 3 (23.1) 27 (56.2) Europe 9 (25.7) 8 (61.5) 17 (35.4) Asia 1 (2.9) 0 1 (2.1) Years working as an SLT with adults who stutter: mean (SD); range n.a. 13.46 (10.48); 2–37 Self-help support group involvement: n (%) Yes 11 (31.4) 4 (100) 15 (38.5) No 24 (68.6) 0 24 (61.5) Received SLT for stuttering as child: n (%) Yes 22 (62.9) 4 (100) 26 (66.7) No 13 (37.1) 0 13 (33.3) Received SLT for stuttering as adult: n (%) Yes 25 (71.4) 1 (25) 26 (66.7) No 10 (28.6) 3 (75) 13 (33.3) Round 1: Exploring core components of intervention for adults who stutter

A consent form for participation in the study was completed by each participant at the start of the first questionnaire. Participants were requested to identify and add any additional essential intervention components that had not been included in the list of 25 statements presented by the authors. Participants were also given the opportunity to list any other intervention components that were outside of the four categories presented. Deductive content analysis was used to analyse the data collected (Elo & Kyngäs, 2007). New statements provided by participants were categorized into the four core components of the categorization matrix (Blase & Fixsen, 2013). Statements in each category were then examined and similar items were collapsed into one statement (Keeney et al., 2011). This was an iterative process completed by two authors (A.C., A.M.) involving discussing, reflecting and agreeing on the final set of statements to be presented to participants in Round 2.

Round 2: Building consensus

The second questionnaire was developed from the list of statements generated following content analysis of Round 1 data. To maintain the exploratory nature of the study and to reduce bias, the statements were returned to the participants without categorization. Participants were requested to rate their level of disagreement or agreement for each core intervention component using a five-point Likert scale (1 = strongly disagree to 5 = strongly agree). A five-point scale was chosen as poor reliability and discriminating power have been reported for scales of four or fewer points (Preston & Colman, 2000). The presentation of statements was randomized to minimize any order effect. Resulting data were inputted into IBM SPSS Statistics, version 26, which produced descriptive statistics. Median and interquartile range (IQR) were calculated for each statement and those that had a median ≥ 4 (i.e., high level of agreement) with a small IQR (≤ 1) were identified as statements that had reached consensus on their importance. Von der Gracht (2012) recommend that an IQR of ≤ 1 is an appropriate consensus indicator for five-unit scales as it will ensure that ≥ 50% of all opinions fall within 1 point on the scale. Statements with a median score of ≤ 3 (i.e., low level of agreement) and a small IQR (≤ 1), were deemed to have reached consensus on a lack of importance and were therefore omitted from Round 3. The statements that did not reach consensus were used to develop the Round 3 questionnaire.

Round 3: Providing feedback and continued consensus-building

The third round provided participants with a document detailing the results of Round 2 in the form of two tables of randomly listed statements: those that reached consensus and those that did not reach consensus on their importance as components of intervention. Participants also completed the third questionnaire containing the 22 statements that had not reached consensus. Participants were provided with their own responses and the group's responses to the statements, and they were asked to re-examine each response in the context of the group response. They were advised that they could choose whether or not to change their responses and were requested to rate all 22 statements again using the five-point Likert scale used in Round 2. Once final consensus on importance had been reached for the statements from Rounds 2 and 3, they were ranked in order of their level of importance using calculated means for each statement. This ranking exercise offered another way to understand the data, and facilitated the presentation of high priority statements to clinicians who may be working in resource-constrained contexts.

Categorization of statements

Numerous frameworks for intervention development, implementation, evidence-based practice and health were initially reviewed for their suitability in categorizing the statements into an accessible format (Damschroder et al., 2009; WHO, 2013; Wight et al., 2016, O'Cathain et al., 2019). The application of statements to all frameworks was trialled to examine best fit and an example is provided in Table S3 in the additional supporting information, which outlines the categorization of statements into the original ICF framework (WHO, 2013). Following this task, the authors selected Tichenor and Yaruss's (2019a) framework which is structured using the primary domains of the ICF. This framework was chosen because of the ubiquity of the ICF in healthcare in general and because of the framework's adaptation of the ICF to the study of stuttering based on the experiences of adults who stutter. All statements that reached group consensus after Rounds 2 and 3 were then categorized by the first and second authors (A.C., J.S.Y.) using this framework. As some statements referenced more than one domain of the framework, it was agreed that they would be categorized based on the domain to which they were primarily linked. The first author (A.C.) discussed the results of this activity with the other authors (R.G., A.M.) and final consensus was reached on the categorization of statements into their primary domains.

RESULTS

Figure 1 summarizes the e-Delphi process and the results of each round. A total of 35 adults who stutter, and 13 SLTs, four of whom also stutter, agreed to become members of the expert panel. It was not possible to calculate the overall response rate due to the nature of the recruitment which included social media distribution. Demographic data of the participants, who resided in four different continents, are provided in Table 1. Response rates for each round were > 70%, which is the recommended rate for Delphi rounds in order to maintain rigour (Keeney et al., 2011). All 48 participants (100%) completed the Round 1 questionnaire, 40/48 (83%) completed Round 2 and 36/40 (90%) participated in Round 3. Of the original 48 respondents, 75% completed all three rounds. Table 2 provides a description of the different participant groups and their response rate for each round.

Summary of the e-Delphi process [Colour figure can be viewed at wileyonlinelibrary.com] TABLE 2. Participants’ response information for each round Round 1, N = 48 Round 2, N = 40 Round 3, N = 36 Participant type n % n % n % Adults who stutter 35 72.92% 28 70% 24 66.67% SLTs 13 27.08% 12 30% 12 33.33%

Content analysis of the participants’ responses to the Round 1 questionnaire resulted in 101 statements which were presented to participants in Round 2 (see Table S4 in the additional supporting information). Based on the authors' a priori definition of consensus, 76 of these statements reached consensus in Round 2 on their importance as intervention components (median ≥ 4 and IQR ≤ 1). Three statements reached consensus on their lack of importance as intervention components (median ≤ 3 and IQR ≤ 1) and were omitted from the framework categorization. A total of 22 statements that did not reach group consensus were used to develop the Round 3 questionnaire. Table S5 in the additional supporting information provides the median scores and IQRs for all statements. Of the 22 statements that did not reach consensus in Round 2, 13 reached consensus in Round 3 on their importance as intervention components. Another four statements reached group consensus on their lack of importance as intervention components and five statements did not reach group consensus. These nine statements were omitted from the framework categorization. Medians and IQRs for all these statements are shown in Table S6 in the additional supporting information. The level of importance of each of the final total of 89 statements that achieved consensus, which was calculated using their means, is presented in Table 3.

TABLE 3. Statements that reached consensus listed by mean (highest to lowest level of importance) Statements that reached group consensus on their importance as intervention components Mean Intervention should be affordable for all individuals who stutter 4.97 That there is no quick fix or cure for stuttering should be explained to the individual who stutters when intervention commences 4.83 The individual who stutters needs to be take responsibility for making changes in their life 4.80 An important outcome of intervention is that the individual who stutters sees themselves as valuable and worthy 4.78 Management of communication-related anxiety 4.75 Good listening skills by the SLT/SLP 4.75 Intervention goals should be realistic and achievable 4.75 A positive therapeutic relationship between the individual who stutters and the SLT/SLP is important 4.72 Intervention goals should be meaningful to the individual's real-life situation 4.70 The rationale for choosing different intervention approaches should be explained to the individual who stutters 4.70 ‘Curing’ stuttering should not be a goal of intervention 4.64 Intervention should target the individual needs of the individual who stutters 4.63 Motivation by the individual who stutters to engage with the intervention and associated activities 4.60 Following discharge from intervention, the individual who stutters should be advised to return for intervention in the future if required 4.60 The SLT/SLP should have empathy as a core skill 4.60 Feedback from the individual who stutters regarding their experiences and perspectives of the intervention being provided should be sought 4.60 Stuttering intervention should be freely available (publicly funded) 4.60 Acceptance of being an individual who stutters 4.58 Specialization courses in stuttering need to be available to SLTs/SLPs during which they can gain practical and mentored experience 4.58 Follow-up sessions (e.g., 6 months after intervention) should be offered 4.58 The SLT/SLP and individual who stutters should work together to make decisions and set goals throughout intervention 4.58 Development of a positive self-identity 4.55 The SLT/SLP providing the intervention should have stuttering-specific training 4.55 The SLT/SLP providing intervention should attend regular CPD trainings in stuttering 4.55 The individual who stutters should be supported in having goals other than increased fluency 4.55 SLTs/SLPs should employ best practice guidelines to guide intervention 4.55 Communication self-efficacy 4.53 Identifying and targeting underlying shame which may be associated with stuttering 4.53 The number of sessions provided should be based on the needs and preferences of the individual who stutters 4.53 Avoidance reduction activities such as reducing avoidance of certain speaking situations should be completed in real life situations 4.53 Both the assessment and intervention should be holistic, meaning they target the whole experience of stuttering for the individual 4.53 Reduction of cognitive, behavioural and emotional reactions to stuttering 4.50 Support with identifying and understanding stuttering behaviours 4.50 Reducing avoidance of speaking or social situations 4.47 Mindfulness and acknowledgment by the SLT/SLP of the significant impact stuttering has on an individual's mental health 4.45 The frequency of sessions should be based on the needs and preferences of the individual who stutters 4.45 Targeting effective communication skills such as speaking confidently whether stuttering occurs or not 4.43 Outcomes related to improved mental health are as important as improved speech outcomes 4.43 Resources and information about local and international stuttering organizations should be made available 4.40 The expertise and knowledge of stuttering of the individual who stutters should be acknowledged and respected 4.40 A range of therapeutic approaches may be employed in stuttering intervention as multiple therapies are effective 4.40 The individual who stutters should be allowed time to learn from intervention at their own pace 4.40 The individual who stutters should determine if an intervention has been effective 4.40 Responding to unfavourable reactions from others 4.38 A willingness to restructure their cognitive view of themselves, others and their communication difficulties by the individual who stutters 4.38 Intervention should be available locally for all individuals who stutter in their immediate geographical area 4.38 Management of emotions that affect stuttering including anger or nervousness 4.35 Intervention should include a discussion on the personal lived experience of stuttering for the individual 4.35 Management of employment-related challenges 4.33 Different intervention formats including individual, group, and intensive should be available for the individual who stutters to choose from 4.33 Multiple intervention outcomes including positive changes to speech, participation in daily activities and quality of life should be considered and agreed on with the SLT/SLP 4.33 Long-term outcomes need to be measured and recorded 4.33 The SLT/SLP providing direct speech intervention should be trained in the specific speech programme they are using 4.33 Speech and language therapy/pathology services should be available both publicly and privately 4.33 Knowledge of relevant counselling principles and approaches by the SLT/SLP 4.30 Intervention should be available outside typical working hours for example. evenings or at weekends 4.30 Improvement in quality of life of the individual who stutters 4.28 The individual who stutters should decide whether or not intervention is required 4.28 Intervention should be provided by a registered SLT/SLP 4.28 Targeting the reduction of the experience of stigma 4.25 SLTs/SLPs should have access to supervision with a senior therapist or clinical specialist in stuttering 4.25 Intensive intervention should be available publicly 4.22 Working directly on speech to reduce the amount or severity of stuttering 4.20 Group intervention should be offered to allow individuals who stutter to meet and learn from each other 4.20 Acknowledgment by the SLT/SLP that not wanting help for stuttering is ok 4.20 Online intervention groups should be available for individuals who stutter 4.18 The SLT/SLP should combine therapies targeting speech change and therapies targeting the psychological impact of stuttering 4.15 Development and maintenance of relationships with others 4.14 A choice between face-to-face or telehealth delivery of intervention should be offered to the individual who stutters 4.14 The SLT/SLP and individual who stutters should agree on what and how intervention outcomes will be measured 4.14 Involvement of a partner or a family member in the intervention should be encouraged 4.13 Intervention should support the individual who stutters to embrace their stuttering and find joy, gain and pride in it 4.11 Legalities such as workplace discrimination should be discussed in intervention 4.06 Where appropriate, the SLT/SLP should liaise with other services such as a disability office at a third level educational setting