University of Groningen Lifestyle change in adults with intellectual disabilities Willems, Mariël

(1)

Lifestyle change in adults with intellectual disabilities

Willems, Mariël

DOI:

10.33612/diss.102031521

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from

it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date:

2019

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Willems, M. (2019). Lifestyle change in adults with intellectual disabilities: use and effectiveness of

behaviour change techniques. Rijksuniversiteit Groningen. https://doi.org/10.33612/diss.102031521

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

(2)

Chapter 4

Exploration of suitable behaviour change techniques for lifestyle change in

individuals with mild intellectual disabilities: A Delphi study

Mariël Willems Aly Waninge Johan de Jong Thessa I.M. Hilgenkamp Cees P. Van der Schans

(3)

96

ABSTRACT

Background: Promotion of a healthy lifestyle for individuals with mild intellectual disabilities is important. However, the suitability of behaviour change techniques (BCTs) for these individuals is still unclear.

Methods: A Delphi study was performed using the Coventry, Aberdeen & LOndon – REfined (CALO‐RE) taxonomy of BCTs (n = 40). Health professionals (professional caregivers, behavioural scientists, health professionals, intellectual disability physicians) participated in an online survey to determine whether BCTs were suitable or unsuitable. Comments from participants were analysed qualitatively.

Results: Consensus was reached for 25 BCTs out of 40.The most suitable BCTs were barrier identification (97%), set graded tasks (97%) and reward effort towards behaviour (95%). No significant differences were found for intergroup effects.

Conclusion: Regardless of their position and education level, health professionals reached consensus about the suitability of BCTs for individuals with mild intellectual disabilities. Increased use of these BCTs could result in more effective promotion of a healthy lifestyle.

97

1 | INTRODUCTION

Adults with intellectual disabilities are more at risk of being overweight, obese and inactive compared to adults without intellectual disabilities (Hilgenkamp, Reis, Van Wijck, & Evenhuis, 2012; Melville, Hamilton, Hankey, Miller, & Boyle, 2007; Waninge et al., 2013). The promotion of a healthy lifestyle, therefore, is important for individuals with intellectual disabilities (Alesi & Pepi, 2015). Many interventions are aimed at supporting a healthy lifestyle for this population by targeting nutrition, physical activity or both (Willems, Hilgenkamp, Havik, Waninge, & Melville, 2017).

To promote a healthy lifestyle, several reviews confirmed the effectiveness of techniques to change behaviour, called behaviour change techniques (BCTs), for the general population (Bird et al., 2013, Greaves et al., 2011; Michie, Abraham, Whittington, McAteer, & Gupta, 2009; Olander et al., 2013; Williams & French, 2011). Bartholomew and Mullen (2011) recommended improving the effectiveness of health‐related behavioural change and increasing the possibilities of replicating a positive effect by increasing our understanding of the underlying action mechanisms. One recommendation was to determine techniques to change lifestyle behaviour (Abraham & Michie, 2008; Rothman, 2004). To achieve this, a 26‐item taxonomy of behaviour change techniques was developed by Abraham and Michie (2008) which was later refined by Michie et al. (2011) resulting in the 40‐ item Coventry, Aberdeen & LOndon – REfined (CALO‐RE) taxonomy. Another taxonomy was developed later, named the Behavior Change Technique Taxonomy v1 which consisted of 93 techniques (Michie et al., 2013). However, there was heterogeneity between results with respect to which particular BCTs were effective for the general population. Also, associations between the number of used BCTs and the effectiveness of the intervention as well as the combination of particular BCTs are ambiguous (Bird et al., 2013; Olander et al., 2013). Some BCTs also do not seem to be interchangeable between specific health behaviours (Michie et al., 2011). In addition, the suitability of BCTs for specific populations is understudied, and they may need to be adapted or extended (Michie et al., 2011; Van Schijndel‐Speet, 2015).

A review regarding the use of BCTs in lifestyle interventions for individuals with intellectual disabilities determined that the BCT’s “Plan social support/social change” and “Provide information on consequences of behaviour in general” were mostly used (Willems et al., 2017). However, it is unclear whether these BCTs are effective for individuals with intellectual disabilities and whether the use of these BCTs is sufficient for changing the lifestyles of these individuals. Also, the BCTs that are used in the previous studies are often not explicitly labelled as such, which may indicate that these BCTs are used unintentionally. Without intended use of BCTs and evaluation of this use, effectiveness of the used BCTs could not be determined. BCTs targeting an increase in knowledge and understanding as well as those primarily using executive functioning may be less suitable for individuals with mild intellectual disabilities because translation of health knowledge into behaviour might be difficult for these individuals (Kuijken, Naaldenberg, Nijhuis‐van der Sanden, & Schrojenstein‐Lantman de Valk, 2016). Until now, the suitability of BCTs in lifestyle interventions for individuals with intellectual disabilities remains unclear.

(4)

ABSTRACT

Background: Promotion of a healthy lifestyle for individuals with mild intellectual disabilities is important. However, the suitability of behaviour change techniques (BCTs) for these individuals is still unclear.

Methods: A Delphi study was performed using the Coventry, Aberdeen & LOndon – REfined (CALO‐RE) taxonomy of BCTs (n = 40). Health professionals (professional caregivers, behavioural scientists, health professionals, intellectual disability physicians) participated in an online survey to determine whether BCTs were suitable or unsuitable. Comments from participants were analysed qualitatively.

Results: Consensus was reached for 25 BCTs out of 40.The most suitable BCTs were barrier identification (97%), set graded tasks (97%) and reward effort towards behaviour (95%). No significant differences were found for intergroup effects.

Conclusion: Regardless of their position and education level, health professionals reached consensus about the suitability of BCTs for individuals with mild intellectual disabilities. Increased use of these BCTs could result in more effective promotion of a healthy lifestyle.

1 | INTRODUCTION

Adults with intellectual disabilities are more at risk of being overweight, obese and inactive compared to adults without intellectual disabilities (Hilgenkamp, Reis, Van Wijck, & Evenhuis, 2012; Melville, Hamilton, Hankey, Miller, & Boyle, 2007; Waninge et al., 2013). The promotion of a healthy lifestyle, therefore, is important for individuals with intellectual disabilities (Alesi & Pepi, 2015). Many interventions are aimed at supporting a healthy lifestyle for this population by targeting nutrition, physical activity or both (Willems, Hilgenkamp, Havik, Waninge, & Melville, 2017).

To promote a healthy lifestyle, several reviews confirmed the effectiveness of techniques to change behaviour, called behaviour change techniques (BCTs), for the general population (Bird et al., 2013, Greaves et al., 2011; Michie, Abraham, Whittington, McAteer, & Gupta, 2009; Olander et al., 2013; Williams & French, 2011). Bartholomew and Mullen (2011) recommended improving the effectiveness of health‐related behavioural change and increasing the possibilities of replicating a positive effect by increasing our understanding of the underlying action mechanisms. One recommendation was to determine techniques to change lifestyle behaviour (Abraham & Michie, 2008; Rothman, 2004). To achieve this, a 26‐item taxonomy of behaviour change techniques was developed by Abraham and Michie (2008) which was later refined by Michie et al. (2011) resulting in the 40‐ item Coventry, Aberdeen & LOndon – REfined (CALO‐RE) taxonomy. Another taxonomy was developed later, named the Behavior Change Technique Taxonomy v1 which consisted of 93 techniques (Michie et al., 2013). However, there was heterogeneity between results with respect to which particular BCTs were effective for the general population. Also, associations between the number of used BCTs and the effectiveness of the intervention as well as the combination of particular BCTs are ambiguous (Bird et al., 2013; Olander et al., 2013). Some BCTs also do not seem to be interchangeable between specific health behaviours (Michie et al., 2011). In addition, the suitability of BCTs for specific populations is understudied, and they may need to be adapted or extended (Michie et al., 2011; Van Schijndel‐Speet, 2015).

A review regarding the use of BCTs in lifestyle interventions for individuals with intellectual disabilities determined that the BCT’s “Plan social support/social change” and “Provide information on consequences of behaviour in general” were mostly used (Willems et al., 2017). However, it is unclear whether these BCTs are effective for individuals with intellectual disabilities and whether the use of these BCTs is sufficient for changing the lifestyles of these individuals. Also, the BCTs that are used in the previous studies are often not explicitly labelled as such, which may indicate that these BCTs are used unintentionally. Without intended use of BCTs and evaluation of this use, effectiveness of the used BCTs could not be determined. BCTs targeting an increase in knowledge and understanding as well as those primarily using executive functioning may be less suitable for individuals with mild intellectual disabilities because translation of health knowledge into behaviour might be difficult for these individuals (Kuijken, Naaldenberg, Nijhuis‐van der Sanden, & Schrojenstein‐Lantman de Valk, 2016). Until now, the suitability of BCTs in lifestyle interventions for individuals with intellectual disabilities remains unclear.

(5)

98

Specific subgroups need specific BCTs that are tailored to the subgroup or the individuals’ preferences and living situation (Van Schijndel‐Speet, 2015). To tailor BCTs to people with mild intellectual disabilities, it is necessary to have a clear understanding of the characteristics of this subgroup. According to the Diagnostic and statistical manual of mental disorders (DSM 5), they suffer from (a) deficits in intellectual functions (e.g., reasoning, problem‐solving, planning, abstract thinking), (b) deficits in adaptive functions and (c) onset of deficits during the developmental period (American Psychiatric Association, 2013). Specifically, people with mild intellectual disabilities need support with abstract thinking, executive functioning, academic skills and additional support within daily living (American Psychiatric Association, 2013). According to the International statistical classification of diseases and related health problems (ICD‐10), it is also important to consider the emotional and social development because of the influence on daily functioning (WHO, 1992). In sum, there is only minimal insight into the suitability of BCTs for individuals with intellectual disabilities, and additional insight is needed to determine which BCTs could be suitable for supporting a change in their lifestyles (Willems et al., 2017). This study aimed to query experts from the field with the Delphi method in order to investigate the suitability of BCTs for supporting a change in the lifestyles of people with mild intellectual disabilities.

2 | METHODS

2.1 | Design

In this Delphi study, the consensus of experts was achieved in several rounds with intermediate feedback regarding the suitability of BCTs in individuals with mild intellectual disabilities (De Meyrick, 2003). This method is considered appropriate for health‐promotion research in “new” areas with only a small knowledge‐base (De Meyrick, 2003; Milat, King, Bauman, & Redman, 2013; Whitehead, 2008). The Delphi study was performed between April 2017 and July 2017. The CALO‐RE taxonomy of BCTs (Michie et al., 2011) was used instead of the Behavior Change Technique Taxonomy (Michie et al., 2013) since this last taxonomy was too extensive for the goal in this Delphi study, exploring the suitability of BCTs specifically for people with mild intellectual disabilities. The panel was completed when 66.67% consensus was reached for the suitability of all BCTs or after a maximum of three rounds since the literature stated that three rounds are usually enough and considering the (time) investment of professionals participating in the panel (Boulkedid, Abdoul, Loustau, Sibony, & Alberti, 2011).

2.2 | Sample

2.2.1 | Criteria for participants

The Delphi panel consisted of three groups of professionals working with individuals in the Netherlands with mild intellectual disabilities, and the desired number of participants was at least ten per group. These groups were selected since they primarily work with individuals with mild intellectual disabilities and provide support in their daily lives. The first group consisted of professional caregivers; the second group consisted of behavioural scientists; and the third group consisted of allied health professionals including physiotherapists, speech therapists, dieticians, psychomotor therapists, movement scientists, occupational therapists, family therapists and nurse practitioners.

99

2.2.2 | Recruitment of participants

Participants were approached through multiple channels for both Delphi rounds in order to achieve the broadest representation possible:

- Through expert networks in the field of lifestyle promotion research for individuals with mild intellectual disabilities. These researchers were asked to share the questionnaire with professional caregivers, behavioural scientists and healthcare professionals in their own network.

- Through a consortium of Dutch intellectual disability care provider organizations. These organizations invited their employees to participate in the panel. Employees were asked to invite colleagues to join the research.

- Through a conference about lifestyle and persons with mild intellectual disabilities that was held on 5 April 2017. Visitors to the conference were invited to participate in the panel. They could participate by emailing the first author.

2.3 | Measurements

The Coventry, Aberdeen & LOndon – REfined (CALO‐RE) taxonomy was used for designing the questionnaires and defining BCTs (Michie et al., 2011). This taxonomy consists of a 40‐item list of theory‐based descriptions of BCTs that are potentially suitable for improving a lifestyle. The CALO‐RE taxonomy was adapted to the CALO‐RE‐ NL in four steps. First, the BCTs were translated into Dutch using the definitions from the CALO‐RE taxonomy. Second, the wording of the BCTs was simplified in order to be understandable for all of the participants with different education levels. Third, examples were drafted for all of the BCTs. Fourth, the CALO‐RE‐NL was converted into a questionnaire. The CALO‐RE‐NL taxonomy is provided in the Appendix.

The questionnaire was pilot‐tested within three groups of intended participants and a group of research experts. First, it was submitted digitally to experts in research on lifestyle changes in individuals with mild intellectual disabilities (n = 5). They provided suggestions for further improvements of the formulations and the examples that were implemented in the questionnaire. Second, the questionnaire was pilot‐tested in a paper form with professional caregivers (n = 15) who were working with individuals with mild intellectual disabilities in a Dutch care provider organization. The caregivers completed the questionnaire and offered feedback regarding the usability (e.g., duration of the questionnaire, the examples used and advised specification of the degree of intellectual disabilities) with which the questionnaire was further improved. Third, the adapted questionnaire was pilot‐tested by the other two groups of participants including behavioural scientists (n = 4) and allied health professionals (n = 4) working with individuals with mild intellectual disabilities. The questionnaire was delivered using online survey software (www.qualtrics.com), and questions were offered at random. Their feedback (e.g., regarding usability of the online version of the questionnaire and used examples) was used to further improve the questionnaire and the online procedure, for example, the visibility of the questions.

Delphi Round 1: The final questionnaire including the CALO‐RE‐NL taxonomy was used in the first Delphi round. The 40 translated BCTs and related examples were presented one‐by‐one in random sequence in order to

(6)

Specific subgroups need specific BCTs that are tailored to the subgroup or the individuals’ preferences and living situation (Van Schijndel‐Speet, 2015). To tailor BCTs to people with mild intellectual disabilities, it is necessary to have a clear understanding of the characteristics of this subgroup. According to the Diagnostic and statistical manual of mental disorders (DSM 5), they suffer from (a) deficits in intellectual functions (e.g., reasoning, problem‐solving, planning, abstract thinking), (b) deficits in adaptive functions and (c) onset of deficits during the developmental period (American Psychiatric Association, 2013). Specifically, people with mild intellectual disabilities need support with abstract thinking, executive functioning, academic skills and additional support within daily living (American Psychiatric Association, 2013). According to the International statistical classification of diseases and related health problems (ICD‐10), it is also important to consider the emotional and social development because of the influence on daily functioning (WHO, 1992). In sum, there is only minimal insight into the suitability of BCTs for individuals with intellectual disabilities, and additional insight is needed to determine which BCTs could be suitable for supporting a change in their lifestyles (Willems et al., 2017). This study aimed to query experts from the field with the Delphi method in order to investigate the suitability of BCTs for supporting a change in the lifestyles of people with mild intellectual disabilities.

2 | METHODS

2.1 | Design

In this Delphi study, the consensus of experts was achieved in several rounds with intermediate feedback regarding the suitability of BCTs in individuals with mild intellectual disabilities (De Meyrick, 2003). This method is considered appropriate for health‐promotion research in “new” areas with only a small knowledge‐base (De Meyrick, 2003; Milat, King, Bauman, & Redman, 2013; Whitehead, 2008). The Delphi study was performed between April 2017 and July 2017. The CALO‐RE taxonomy of BCTs (Michie et al., 2011) was used instead of the Behavior Change Technique Taxonomy (Michie et al., 2013) since this last taxonomy was too extensive for the goal in this Delphi study, exploring the suitability of BCTs specifically for people with mild intellectual disabilities. The panel was completed when 66.67% consensus was reached for the suitability of all BCTs or after a maximum of three rounds since the literature stated that three rounds are usually enough and considering the (time) investment of professionals participating in the panel (Boulkedid, Abdoul, Loustau, Sibony, & Alberti, 2011).

2.2 | Sample

2.2.1 | Criteria for participants

The Delphi panel consisted of three groups of professionals working with individuals in the Netherlands with mild intellectual disabilities, and the desired number of participants was at least ten per group. These groups were selected since they primarily work with individuals with mild intellectual disabilities and provide support in their daily lives. The first group consisted of professional caregivers; the second group consisted of behavioural scientists; and the third group consisted of allied health professionals including physiotherapists, speech therapists, dieticians, psychomotor therapists, movement scientists, occupational therapists, family therapists and nurse practitioners.

2.2.2 | Recruitment of participants

Participants were approached through multiple channels for both Delphi rounds in order to achieve the broadest representation possible:

- Through expert networks in the field of lifestyle promotion research for individuals with mild intellectual disabilities. These researchers were asked to share the questionnaire with professional caregivers, behavioural scientists and healthcare professionals in their own network.

- Through a consortium of Dutch intellectual disability care provider organizations. These organizations invited their employees to participate in the panel. Employees were asked to invite colleagues to join the research.

- Through a conference about lifestyle and persons with mild intellectual disabilities that was held on 5 April 2017. Visitors to the conference were invited to participate in the panel. They could participate by emailing the first author.

2.3 | Measurements

The Coventry, Aberdeen & LOndon – REfined (CALO‐RE) taxonomy was used for designing the questionnaires and defining BCTs (Michie et al., 2011). This taxonomy consists of a 40‐item list of theory‐based descriptions of BCTs that are potentially suitable for improving a lifestyle. The CALO‐RE taxonomy was adapted to the CALO‐RE‐ NL in four steps. First, the BCTs were translated into Dutch using the definitions from the CALO‐RE taxonomy. Second, the wording of the BCTs was simplified in order to be understandable for all of the participants with different education levels. Third, examples were drafted for all of the BCTs. Fourth, the CALO‐RE‐NL was converted into a questionnaire. The CALO‐RE‐NL taxonomy is provided in the Appendix.

The questionnaire was pilot‐tested within three groups of intended participants and a group of research experts. First, it was submitted digitally to experts in research on lifestyle changes in individuals with mild intellectual disabilities (n = 5). They provided suggestions for further improvements of the formulations and the examples that were implemented in the questionnaire. Second, the questionnaire was pilot‐tested in a paper form with professional caregivers (n = 15) who were working with individuals with mild intellectual disabilities in a Dutch care provider organization. The caregivers completed the questionnaire and offered feedback regarding the usability (e.g., duration of the questionnaire, the examples used and advised specification of the degree of intellectual disabilities) with which the questionnaire was further improved. Third, the adapted questionnaire was pilot‐tested by the other two groups of participants including behavioural scientists (n = 4) and allied health professionals (n = 4) working with individuals with mild intellectual disabilities. The questionnaire was delivered using online survey software (www.qualtrics.com), and questions were offered at random. Their feedback (e.g., regarding usability of the online version of the questionnaire and used examples) was used to further improve the questionnaire and the online procedure, for example, the visibility of the questions.

Delphi Round 1: The final questionnaire including the CALO‐RE‐NL taxonomy was used in the first Delphi round. The 40 translated BCTs and related examples were presented one‐by‐one in random sequence in order to

(7)

100

prevent question order bias. For every BCT, participants were asked if the BCT was suitable for influencing the lifestyles of individuals with mild intellectual disabilities. Dichotomous answers included “yes, this technique is suitable” or “no, this technique is not suitable.” With each dichotomous answer, there was also an open field to write down comments. The first questionnaire was delivered using the web‐based survey software, Qualtrics.com, that invited participants by email and which included an anonymous hyperlink to the questionnaire. Participants were asked to respond within 16 days. Personal information was asked about job position, years of work experience and email address (not obligatory). Answers were collected and anonymously analysed. Participants provided informed consent before participating in any of the Delphi rounds.

Delphi Round 2: After the data from the first Delphi round were analysed, a second questionnaire was designed based on the outcomes from it. The purpose of the second round was to determine consensus about the suitability of the BCTs in which no consensus was reached in the first round. The BCTs and corresponding examples were similar to the second questionnaire. The only exception concerned some of the examples that were not completely clear to the participants; this was expressed in the comments provided by the participants in the open fields in the first round. Only minor changes were made to the examples (e.g., the time period in which a behaviour goal had to be completed).

Participants were asked whether or not these BCTs were suitable. Again, there was an open field to write down comments. The group decisions from Round 1 were shown in percentages per technique using only one decimal in case the percentage decimal was exactly 0.5. Comments from participants in Round 1 were translated into statements that were more common and comprehensive by the first author (MW) and checked by the second author (AW). Disagreements were resolved by a consensus discussion. These resumptive comments (e.g., “according to participants of the panel, suitability depends on the client characteristics”) were shown per technique next to the group decisions from Round 1. The questionnaire was tailored to the groups of participants, and questions were offered in a random order. The second questionnaire was also delivered using the web‐based survey software, Qualtrics.com, by which participants were invited through an email containing an anonymous hyperlink to the questionnaire. Participants were asked to respond within 16 days. Personal information was asked about job position, years of work experience and email address (not obligatory).

2.4 | Data analyses

Round 1: After collecting the data from the first questionnaire, consensus was determined. There is no generally accepted cut‐off score for the consensus that was reached (Becker, & Roberts, 2009; Boulkedid et al., 2011). The present authors used a cut‐off score of 66.67% consensus both for evaluating a BCT as suitable or not (in changing lifestyle) in the first round. The 66.67% consensus was determined by dividing the number of corresponding answers of the participants (yes, suitable/no, not suitable) by the number of participants for that group. The present authors calculated the total consensus (in %) of each group and the total consensus (in %) of all of the groups together. Since the questions were offered at random, the answers from incomplete questionnaires were also used in the analyses. The comments of participants were analysed by theme and categorized. These

101

categories or themes were used in the second questionnaire in order to give participants’ feedback on the results of the first round.

Per profession group, the inter‐rater reliability of the binary suitability scores was estimated by the alpha coefficient from the ordinal factor analysis performed on the tetrachoric correlation coefficients (Revelle & Zinbarg, 2009). Reliability was considered poor or moderate (ICC < 0.75), good (ICC 0.75–0.90) or very good (ICC > 0.90) (Portney & Watkins, 2000). To investigate agreement between profession groups, the present authors computed the two‐way intraclass correlation coefficient (ICC) (McGraw & Wong, 1996) by using the average BCT score per profession group. Also for the ICC, reliability was considered poor or moderate (ICC < 0.75), good (ICC 0.75–0.90) or very good (ICC > 0.90) (Portney & Watkins, 2000). To investigate whether there was any systematic bias of a profession group with respect to any other profession group, a mixed model analysis (Pinheiro, & Bates, 2000) was undertaken on the average BCT scores per profession. In the mixed model, random intercepts were specified for the BCTs and fixed effects for the profession groups. A pairwise comparison (Hothorn, Bretz, & Westfall, 2008) was used to analyse the intergroup effects. For all statistical analyses, raters with more than 30% missing BCT suitability evaluations were discarded from further analysis. R v3.3.0 was used (R Core Team, 2017). Round 2: After data collection, consensus for the BCTs in Round 2 was determined within each group. The total number of BCTs with consensus and the number of suitable BCTs for this round was not calculated for the total group since the BCTs in the second questionnaire differed per group depending on the consensus found in the first round per group. This made it also impossible to compute the agreement between profession groups and the systematic bias of a profession group with respect to another profession group.

The present authors calculated the total consensus (in %) for all BCTs of each group and that over all groups. Suitable BCTs were ranked by mean consensus using the percentages of consensus on suitability for all of the groups. Comments from participants were analysed regarding theme and then categorized. The answers of behavioural scientists were compared separately with the other groups since this profession group has the most extensive education and working experience with BCTs. The present authors compared the consensus of the behavioural scientists with the consensus of the other groups to identify possible differences in consensus.

3 | RESULTS

3.1 | First Delphi round

A total of 97 professionals responded to the questionnaire in the first round. Ten respondents did not answer any of the questions. Four participants declared that they did not want to participate in the research and, therefore, did not respond to further questions. Answers from two participants were excluded from the data analysis as they did not satisfy the inclusion criteria for participants. Answers of 81 participants were included in the data analysis of which 58 completed the entire questionnaire. No patterns of missing data were found, and

(8)

prevent question order bias. For every BCT, participants were asked if the BCT was suitable for influencing the lifestyles of individuals with mild intellectual disabilities. Dichotomous answers included “yes, this technique is suitable” or “no, this technique is not suitable.” With each dichotomous answer, there was also an open field to write down comments. The first questionnaire was delivered using the web‐based survey software, Qualtrics.com, that invited participants by email and which included an anonymous hyperlink to the questionnaire. Participants were asked to respond within 16 days. Personal information was asked about job position, years of work experience and email address (not obligatory). Answers were collected and anonymously analysed. Participants provided informed consent before participating in any of the Delphi rounds.

Delphi Round 2: After the data from the first Delphi round were analysed, a second questionnaire was designed based on the outcomes from it. The purpose of the second round was to determine consensus about the suitability of the BCTs in which no consensus was reached in the first round. The BCTs and corresponding examples were similar to the second questionnaire. The only exception concerned some of the examples that were not completely clear to the participants; this was expressed in the comments provided by the participants in the open fields in the first round. Only minor changes were made to the examples (e.g., the time period in which a behaviour goal had to be completed).

Participants were asked whether or not these BCTs were suitable. Again, there was an open field to write down comments. The group decisions from Round 1 were shown in percentages per technique using only one decimal in case the percentage decimal was exactly 0.5. Comments from participants in Round 1 were translated into statements that were more common and comprehensive by the first author (MW) and checked by the second author (AW). Disagreements were resolved by a consensus discussion. These resumptive comments (e.g., “according to participants of the panel, suitability depends on the client characteristics”) were shown per technique next to the group decisions from Round 1. The questionnaire was tailored to the groups of participants, and questions were offered in a random order. The second questionnaire was also delivered using the web‐based survey software, Qualtrics.com, by which participants were invited through an email containing an anonymous hyperlink to the questionnaire. Participants were asked to respond within 16 days. Personal information was asked about job position, years of work experience and email address (not obligatory).

2.4 | Data analyses

Round 1: After collecting the data from the first questionnaire, consensus was determined. There is no generally accepted cut‐off score for the consensus that was reached (Becker, & Roberts, 2009; Boulkedid et al., 2011). The present authors used a cut‐off score of 66.67% consensus both for evaluating a BCT as suitable or not (in changing lifestyle) in the first round. The 66.67% consensus was determined by dividing the number of corresponding answers of the participants (yes, suitable/no, not suitable) by the number of participants for that group. The present authors calculated the total consensus (in %) of each group and the total consensus (in %) of all of the groups together. Since the questions were offered at random, the answers from incomplete questionnaires were also used in the analyses. The comments of participants were analysed by theme and categorized. These

categories or themes were used in the second questionnaire in order to give participants’ feedback on the results of the first round.

Per profession group, the inter‐rater reliability of the binary suitability scores was estimated by the alpha coefficient from the ordinal factor analysis performed on the tetrachoric correlation coefficients (Revelle & Zinbarg, 2009). Reliability was considered poor or moderate (ICC < 0.75), good (ICC 0.75–0.90) or very good (ICC > 0.90) (Portney & Watkins, 2000). To investigate agreement between profession groups, the present authors computed the two‐way intraclass correlation coefficient (ICC) (McGraw & Wong, 1996) by using the average BCT score per profession group. Also for the ICC, reliability was considered poor or moderate (ICC < 0.75), good (ICC 0.75–0.90) or very good (ICC > 0.90) (Portney & Watkins, 2000). To investigate whether there was any systematic bias of a profession group with respect to any other profession group, a mixed model analysis (Pinheiro, & Bates, 2000) was undertaken on the average BCT scores per profession. In the mixed model, random intercepts were specified for the BCTs and fixed effects for the profession groups. A pairwise comparison (Hothorn, Bretz, & Westfall, 2008) was used to analyse the intergroup effects. For all statistical analyses, raters with more than 30% missing BCT suitability evaluations were discarded from further analysis. R v3.3.0 was used (R Core Team, 2017). Round 2: After data collection, consensus for the BCTs in Round 2 was determined within each group. The total number of BCTs with consensus and the number of suitable BCTs for this round was not calculated for the total group since the BCTs in the second questionnaire differed per group depending on the consensus found in the first round per group. This made it also impossible to compute the agreement between profession groups and the systematic bias of a profession group with respect to another profession group.

The present authors calculated the total consensus (in %) for all BCTs of each group and that over all groups. Suitable BCTs were ranked by mean consensus using the percentages of consensus on suitability for all of the groups. Comments from participants were analysed regarding theme and then categorized. The answers of behavioural scientists were compared separately with the other groups since this profession group has the most extensive education and working experience with BCTs. The present authors compared the consensus of the behavioural scientists with the consensus of the other groups to identify possible differences in consensus.

3 | RESULTS

3.1 | First Delphi round

A total of 97 professionals responded to the questionnaire in the first round. Ten respondents did not answer any of the questions. Four participants declared that they did not want to participate in the research and, therefore, did not respond to further questions. Answers from two participants were excluded from the data analysis as they did not satisfy the inclusion criteria for participants. Answers of 81 participants were included in the data analysis of which 58 completed the entire questionnaire. No patterns of missing data were found, and

(9)

102

answers ranged from 60 to 69 per BCT (Mean 64.2; SD 1.98). Figure 1 illustrates a flow chart of the Delphi rounds for all of the participant groups.

Figure 1 Flow chart number of BCTs per Delphi round per group of participants

Three groups of participants were initially included. In addition, a fourth group was added after the first Delphi round that consisted of physicians specialized in care for individuals with intellectual disabilities, henceforth referred to as intellectual disability physicians. This group was separated from the other groups because of their different levels of education, expertise and work experiences. From the participants who described their job function in the “other” category, five were added to the group with the professional caregivers, 12 were added to the group with allied health professionals, and 15 were added to the group with the intellectual disability physicians. This results in the distribution amongst the four groups as described in Table 1.

Table 1 Overview of consensus after Delphi round 1

Professional caregivers Behavioural scientists Allied health professionals ID Physicians Total group participants No. of participants 18 10 38 15 81

No. of completed questionnaires 13 (72%) 6 (60%) 28 (74%) 11 (73%) 58 (72%) No. of BCTs with consensus 28 (70%) 31 (77.5%) 31 (77.5%) 28 (70%) 18 (45%) Suitable BCTs 25 (62.5%) 25 (62.5%) 29 (72.5%) 25 (62.5%) 17 (42.5%)

The total group of participants reached consensus for 18 BCTs: 17 BCTs were considered suitable, and one was considered unsuitable, that is, “Prompt anticipated regret”. For two BCTs, no consensus was found in all of the four groups: “Facilitate social comparison” and “Stress management/emotional control training”. For 20 BCTs, consensus about suitability was found for some but not for all groups of participants. See Table 1 for additional details regarding group consensus.

An overview of the ordinal alpha for inter‐rater reliability regarding the agreement over raters for each profession group is shown in Table 2. Alpha was moderate for behavioural scientists (α 0.72, CI 0.562–0.853) and good for

Round 2 Round 1 Behavioural scientists N=40BCTs Suitable N=25 Unsuitable N=6 No consensus N=9 Suitable N=5 Unsuitable N=1 No consensus N=3 Round 2 Round 1 Professional caregivers N=40BCTs Suitable N=25 Unsuitable N=3 No consensus N=12 Suitable N=5 Unsuitable N=4 No consensus N=3 Round 2 Round 1 Physicians ID BCTs N=40 Suitable N=25 Unsuitable N=3 No consensus N=12 Suitable N=5 Unsuitable N=0 No consensus N=7 Round 2 Round 1 Allied health professionals N=40BCTs Suitable N=29 Unsuitable N=2 No consensus N=9 Suitable N=5 Unsuitable N=2 No consensus N=2 103

professional caregivers (α 0.81, CI 0.715– 0.876), allied health professionals (α 0.88, CI 0.820–0.915) and intellectual disability physicians (α 0.83, CI 0.627–0.915). Among the excluded raters with missing evaluations, there was one with 38% missing, and the others had more than 50% missing. ICC for agreement between profession groups was good (ICC 0.876, CI 0.8; 0.929) which indicates a high degree of agreement between the groups. Analysis of the intergroup effects showed no evidence for significant differences in mean scores between the four profession groups whereas none of the pairwise comparisons were significant (see also Table 3).

Table 2: Interrater reliability per group of professionals

Profession group Number of raters (total no. of raters) α (ordinal Alpha) 95% CI*

Professional caregivers 13 (18) 0.81 0.715- 0.876 Behavioural scientists 6 (10) 0.72 0.562-0.853 Allied health professionals 28 (38) 0.88 0.820- 0.915

ID Physicians 11 (15) 0.83 0.627-0.915

* CI, Confidence Intervals

Table 3: Group comparisons of mean rating per BCT for each group of professionals

Profession groups Estimate Standard Error p-value

Professional caregivers - Behavioural scientists 0.04872 0.02894 0.3324 Allied health professionals - Behavioural scientists 0.07173 0.02894 0.0632 ID Physicians - Behavioural scientists 0.03333 0.02894 0.6573 Allied health professionals - Professional caregivers 0.02301 0.02894 0.8567 ID Physicians - Professional caregivers -0.01538 0.02894 0.9514 ID Physicians - Allied health professionals -0.03839 0.02894 0.5459

3.2 | Second Delphi round

A total of 123 professionals responded to the questionnaire in the second round. Nine respondents did not answer any of the questions. Three respondents stated that they did not want to participate further in the research and, therefore, had not responded to further questions. Two participants stated that they wanted to participate, however, did not answer any of the further questions. The answers of 112 participants were included in data analysis, and 87 participants completed the questionnaire. Figure 1 provides a flow chart of the Delphi rounds.

For the group of professional caregivers, Rounds 1 and 2 resulted in consensus about 30 suitable and seven unsuitable BCTs; see also Figure 1. For three BCTs, consensus was not reached, that is, “Provide information on consequences of behaviour to the individual,” “Goal setting (outcome)” and “Agree behavioural contract”. An overview of consensus after Delphi round 2 is shown in Table 4.

For the group of behavioural scientists, Rounds 1 and 2 resulted in consensus about 30 suitable and seven unsuitable BCTs (see Figure 1). For three BCTs, consensus was not reached the following: “Prompt review of outcome goals”, “Environmental restructuring” and “Facilitate social comparison”.

(10)

answers ranged from 60 to 69 per BCT (Mean 64.2; SD 1.98). Figure 1 illustrates a flow chart of the Delphi rounds for all of the participant groups.

Figure 1 Flow chart number of BCTs per Delphi round per group of participants

Three groups of participants were initially included. In addition, a fourth group was added after the first Delphi round that consisted of physicians specialized in care for individuals with intellectual disabilities, henceforth referred to as intellectual disability physicians. This group was separated from the other groups because of their different levels of education, expertise and work experiences. From the participants who described their job function in the “other” category, five were added to the group with the professional caregivers, 12 were added to the group with allied health professionals, and 15 were added to the group with the intellectual disability physicians. This results in the distribution amongst the four groups as described in Table 1.

Professional caregivers Behavioural scientists Allied health professionals ID Physicians Total group participants No. of participants 18 10 38 15 81

No. of completed questionnaires 13 (72%) 6 (60%) 28 (74%) 11 (73%) 58 (72%) No. of BCTs with consensus 28 (70%) 31 (77.5%) 31 (77.5%) 28 (70%) 18 (45%) Suitable BCTs 25 (62.5%) 25 (62.5%) 29 (72.5%) 25 (62.5%) 17 (42.5%)

The total group of participants reached consensus for 18 BCTs: 17 BCTs were considered suitable, and one was considered unsuitable, that is, “Prompt anticipated regret”. For two BCTs, no consensus was found in all of the four groups: “Facilitate social comparison” and “Stress management/emotional control training”. For 20 BCTs, consensus about suitability was found for some but not for all groups of participants. See Table 1 for additional details regarding group consensus.

An overview of the ordinal alpha for inter‐rater reliability regarding the agreement over raters for each profession group is shown in Table 2. Alpha was moderate for behavioural scientists (α 0.72, CI 0.562–0.853) and good for

Round 2 Round 1 Behavioural scientists N=40BCTs Suitable N=25 Unsuitable N=6 No consensus N=9 Suitable N=5 Unsuitable N=1 No consensus N=3 Round 2 Round 1 Professional caregivers N=40BCTs Suitable N=25 Unsuitable N=3 No consensus N=12 Suitable N=5 Unsuitable N=4 No consensus N=3 Round 2 Round 1 Physicians ID BCTs N=40 Suitable N=25 Unsuitable N=3 No consensus N=12 Suitable N=5 Unsuitable N=0 No consensus N=7 Round 2 Round 1 Allied health professionals BCTsN=40 Suitable N=29 Unsuitable N=2 No consensus N=9 Suitable N=5 Unsuitable N=2 No consensus N=2

professional caregivers (α 0.81, CI 0.715– 0.876), allied health professionals (α 0.88, CI 0.820–0.915) and intellectual disability physicians (α 0.83, CI 0.627–0.915). Among the excluded raters with missing evaluations, there was one with 38% missing, and the others had more than 50% missing. ICC for agreement between profession groups was good (ICC 0.876, CI 0.8; 0.929) which indicates a high degree of agreement between the groups. Analysis of the intergroup effects showed no evidence for significant differences in mean scores between the four profession groups whereas none of the pairwise comparisons were significant (see also Table 3).

Table 2: Interrater reliability per group of professionals

Profession group Number of raters (total no. of raters) α (ordinal Alpha) 95% CI*

Professional caregivers 13 (18) 0.81 0.715- 0.876 Behavioural scientists 6 (10) 0.72 0.562-0.853 Allied health professionals 28 (38) 0.88 0.820- 0.915

ID Physicians 11 (15) 0.83 0.627-0.915

* CI, Confidence Intervals

Table 3: Group comparisons of mean rating per BCT for each group of professionals

Profession groups Estimate Standard Error p-value

Professional caregivers - Behavioural scientists 0.04872 0.02894 0.3324 Allied health professionals - Behavioural scientists 0.07173 0.02894 0.0632 ID Physicians - Behavioural scientists 0.03333 0.02894 0.6573 Allied health professionals - Professional caregivers 0.02301 0.02894 0.8567 ID Physicians - Professional caregivers -0.01538 0.02894 0.9514 ID Physicians - Allied health professionals -0.03839 0.02894 0.5459

3.2 | Second Delphi round

A total of 123 professionals responded to the questionnaire in the second round. Nine respondents did not answer any of the questions. Three respondents stated that they did not want to participate further in the research and, therefore, had not responded to further questions. Two participants stated that they wanted to participate, however, did not answer any of the further questions. The answers of 112 participants were included in data analysis, and 87 participants completed the questionnaire. Figure 1 provides a flow chart of the Delphi rounds.

For the group of professional caregivers, Rounds 1 and 2 resulted in consensus about 30 suitable and seven unsuitable BCTs; see also Figure 1. For three BCTs, consensus was not reached, that is, “Provide information on consequences of behaviour to the individual,” “Goal setting (outcome)” and “Agree behavioural contract”. An overview of consensus after Delphi round 2 is shown in Table 4.

For the group of behavioural scientists, Rounds 1 and 2 resulted in consensus about 30 suitable and seven unsuitable BCTs (see Figure 1). For three BCTs, consensus was not reached the following: “Prompt review of outcome goals”, “Environmental restructuring” and “Facilitate social comparison”.

(11)

104

For the group of allied health professionals, Rounds 1 and 2 resulted in consensus on 34 suitable and four unsuitable BCTs (see Figure 1). For two BCTs, consensus was not reached the following: “Prompt use of imagery” and “General communication skill training”.

Professional caregivers Behavioural scientists Allied health professionals ID physicians Total group participants No. of participants 34 13 32 33 112

No. of completed questionnaires 30 (88%) 9 (69%) 24 (75%) 24 (73%) 90 (73%)

No. of BCTs evaluated 12 9 9 12 22

No. of BCTs with consensus 9 (75%) 6 (67%) 7 (78%) 5 (42%) - Suitable BCTs 5 (42%) 5 (56%) 5 (56%) 5 (42%) -

For the group of intellectual disability physicians, Rounds 1 and 2 resulted in consensus on 30 suitable and three unsuitable BCTs (see Figure 1). For seven BCTs, consensus was not reached the following: “Normative information others’ behaviour”, “Generalization of behaviour”, “Self‐monitoring of behaviour”, “Behavioural contract”, “Identification as role model”, “Fear arousal” and “Self‐talk”.

3.3 | Results for the total group after two Delphi rounds

A comparison of the answers from the four groups (professional caregivers, behavioural scientists, allied health professionals and intellectual disability physicians) shows only minor differences as depicted in Table 5. Differences were determined for seven BCTs whereby consensus differed between participant groups since one of them reached consensus about suitability whereas another reached consensus about unsuitability. These seven BCTs are as follows: Stress management/ emotional control training”, “Review outcome goals”, “Self‐talk”, “General communication skill training”, “Identification as role model”, “Normative information others’ behaviour” and “Facilitate social comparison”. Three participant groups differed in consensus for two BCTs from the other groups, and only one participant group (allied health professionals) differed from the other group for one BCT (“Normative information about others’ behaviour”). Specifically, the answers of behavioural scientists differed for two BCTs compared to the other participant groups including the following: “General communication skill training” and “Identification as role model.” The behavioural scientists reached consensus for these BCTs being unsuitable whereas the other groups indicated that these BCTs were suitable or did not reach consensus about the BCT. See also Table 5 for the consensus about the BCTs for each group as well as the direction of consensus.

Consensus was reached for 25 BCTs, 24 of which were considered as suitable, and one BCT was considered unsuitable: “Prompt anticipated regret.” Therefore, the total mean consensus for suitability is low for “Prompt anticipated regret” in Table 5 (18% suitability, mean total consensus). According to the mean total consensus of suitability, most participants considered the following BCTs to be suitable: “Barrier identification/problem‐ solving” (97.3% consensus), “Set graded tasks” (96.6% consensus), “Prompt rewards contingent on effort or

105

progress towards behaviour” (95.3% consensus), “Motivational interviewing” (94.4% consensus) and “Action planning” (90.8% consensus). For an overview of mean consensus per BCT, see Table 5.

A total number of 613 comments were given in the two Delphi rounds (Mean 15.3; SD 7.9; range 3–38) including 368 comments given with a response that the BCT was suitable and 245 comments with a response that the BCT was unsuitable. The comments were categorized into five categories, including:

1. Methods/validity: comments about the methods or validity of the study methods or the text in the questionnaire (e.g., “the example is confusing”);

2. Feasibility: comments about the feasibility of the BCT (e.g., “it depends on what the reward entails”);

3. Relative techniques: comments about the suitability of the BCT in relationship to other BCTs (e.g., “encouraging on itself is rarely enough. Pre‐structuring and shaping is a prerequisite”);

4. Client characteristics: comments about the suitability of the BCT in relationship to client characteristics (e.g., “depending on the client and his possible fears”);

5. Support from others: comments about the suitability of the BCT in relation to support from others (e.g., “the BCT is only suitable if the personal caregiver supports the client”).

Consensus seemed to be reached primarily for the BCTs with fewer comments (see, for more details, Table 5). Most comments were offered for the category “client characteristics” (n = 241), and fewer comments were given for the category “methods/validity” (n = 31); see also Table 6. The most suitable BCTs (barrier identification, graded tasks, reward effort towards behaviour, motivational interviewing and action planning) received a lower number of comments than average (mean 8.4; SD 3.0; range 5–13). Table 7 provides an overview of the number of comments per category for every BCT ranked by suitability.

(12)