Tailoring the Determinants of Implementation Behaviour Questionnaire (DIBQ) to best-practice low back pain primary care program implementation: a mixed-methods validity-testing in Sweden and Denmark

Background: Best-practice low back pain (LBP) primary care programs have been developed based on evidence-based clinical guidelines and are implemented in Sweden and Denmark. The Theoretical Domains Framework and its linkage to The Behavioural Change Wheel has been utilised in the design of the implementation object and its implementation strategy. Based on the Theoretical Domains Framework domains, the Determinants of Implementation Behaviour Questionnaire (DIBQ) has been developed to evaluate implementation determinants but its feasibility and validity needs to be tested and adapted to study specific contexts while maintaining its linkage to the Behavioural Change Wheel. The aim of this study was to tailor the DIBQ for evaluation of the implementation processes for best-practice LBP primary care programs in Sweden and Denmark. More specifically, the objectives were to i) Translate the DIBQ into Swedish and Danish, ii) Adapt the DIBQ into DIBQ-tailored (DIBQ-t) to study content validity, iii) Test the DIBQ-t for feasibility and iv) Perform initial validity testing of DIBQ-t. Methods: A mixed methods design with a four-step process was used. First, forward translation of the DIBQ, then adaptation into DIBQ-t using qualitative face validity assessed by the project group followed by quantitative content validity assessment by an expert group. Finally, primary care clinicians completed the DIBQ-t directly after participation in a 2-day educational course prior to the implementation of the program to determine feasibility and construct validity using confirmatory factor analyses. Results: In total 598 clinicians out of 609 responded, with only 2% of the items missing. The final DIBQ-t included 28 items describing 10 of the original 18 TDF domains and was considered feasible. The confirmatory factor analyses showed good fit after removing items 4 and 13 with the lowest domain loading. The DIBQ-t maintained linkage to all domains of the Capability–Opportunity-Motivation-Behaviour model within the Behavioural

Change Wheel. The clinicians´ expectations according to the DIBQ-t indicate facilitating determinants outweighing barriers at the initiation of implementation processes.
Conclusions: The study resulted in a tailored version of the DIBQ, DIBQ-t, which is feasible and valid for evaluating clinicians' expectations regarding implementation determinants of best-practice LBP primary care programs.

Background
Evidence-based guidelines provide recommendations for clinical practice and have been developed in several countries to assist clinicians in managing patients presenting with low back pain (LBP) [1][2][3]. However, there is often a mismatch between recommendations and clinical practice [4][5][6] because barriers at the level of stakeholders such as patients, providers, organisations and health care systems can complicate the implementation of guidelines [7-10]. Although clinicians consider evidence-based guidelines important, they may not adopt and adhere to them in routine practice for several reasons [11]. Some clinicians consider guidelines as a threat to professional autonomy and inconsistency with clinical reasoning [12], or they may have beliefs and traditions that are not in line with the guideline recommendations [12,13]. Therefore, they may choose to rely upon experience and well-established habits using an intuitive, experiential approach [12,14].
In an effort to rectify these barriers of guideline implementation identified in previous literature, best-practice LBP primary care programs have been developed in Sweden and Denmark aiming to facilitate the adoption of guideline consistent care in the management of people seeking care for LBP [15,16].
The Theoretical Domains Framework (TDF) [17] and the Behavioural Change Wheel [18] are useful when developing implementation strategies. The TDF is a comprehensive behaviour change framework based on 128 constructs from 33 psychological theories categorized into 18 domains considered relevant to behaviours and cognitions involved in evidence-based practice implementation [17,19]. The TDF has also previously been linked to the Behavioural Change Wheel, which aids in interpreting how potential determinants of behavioural change could influence the effects of behavioural change interventions (i.e. the implementation strategy) on the central source of behaviour (i.e. use of a bestpractice LBP primary care program) [20]. The Behavioural Change Wheel incorporates the Capability -Opportunity -Motivation -Behaviour (COM-B) model [21] to describe the central source of behavior (Fig. 1).
The TDF has also been utilised in the development of a survey instrument, the Determinants of Implementation Behaviour Questionnaire (DIBQ) [17,19,22], which quantitatively evaluates the 18 TDF domains role in implementation processes through 93 items. From these domains, researchers can identify the most relevant ones in relation to the aims and target population of a specific research and implementation context. However, feasibility and validity of tailoring the DIBQ to a specific context, here bestpractice low back pain primary care program implementation, needs to be tested while maintaining its linkage to the Behavioural Change Wheel.

Methods
The aim of this study was thus to tailor the DIBQ for valid and feasible evaluation of the clinician expectations regarding implementation processes of best-practice LBP primary care programs in Sweden and Denmark. More specifically, the objectives were to i) Translate the DIBQ into Swedish and Danish, ii) Adapt the version into DIBQ tailored (DIBQ-t) to study best-practice low back pain primary care program implementation for content validity, iii) Test the DIBQ-t for feasibility and iv) Perform initial validity testing of DIBQ-t.

Setting
The BetterBackJ model of care in Sweden [15] and the GLA:D® Back programme in Denmark [16] are best-practice programs for LBP in primary care. They have been developed in collaboration between researchers in the two countries to support the implementation of guideline consistent care. A comparable multifaceted implementation strategy, including a 2-days course with lectures, workshops and access to the supporting material, were used in both countries to enable clinicians to deliver the programs to patients with LBP.

Design
This study applies a mixed-method design in 4 phases. Translation of the DIBQ (phase 1); content validity assessment by an expert group (phase 2); adaptation into DIBQ-t and determining feasibility (phase 3); construct validity of DIBQ-t (phase 4) ( Figure 2). The GRRAS [23] checklist (see Additional file 1) was used to guide our reporting of the study.   Second, quantitative content validity was tested by experts with a professional or methodological research background in the musculoskeletal and/or implementation fields (Table 1). Table 1 The experts were asked to rate each item of the DIBQ on a 1-4 Likert scale from 'not relevant' to 'very relevant' regarding evaluation of the implementation process of the best-practice LBP primary care programs in Sweden and Denmark. The ratings of the experts were indexed using Content Validity Index (CVI) [28]. An item was considered 'relevant' when scoring a CVI of 0.80 or more, i.e. 80% or more of the experts rated the questions 'relevant' or 'very relevant'. Inclusion of items in the DIBQ-t, was based upon three criteria: 1) Swedish and Danish project leaders selected the items related to domains relevant to the project, and experts rated the item with a CVI ≥0.80, or 2) an expert rated relevance score with a CVI = 1.00, regardless of the item being selected by the project leaders, and 3) the project leaders of each country added items included by the project team and with 80% CVI of either the Swedish or Danish experts to anticipate differences in contexts between the countries. The main differences in contexts were that Danish clinicians to a larger degree worked in private clinics and chose to self-fund their participation in the best-practice LBP primary care program. The Swedish clinicians worked in public clinics, and the decision to participate was made by their clinic managers and participation was mandatory and without costs for the participants. The project team considered it of importance that DIBQ-t content validity be relevant for both private and public health care system settings.

Phase 3 and 4 -Feasibility and Construct validity
All clinicians from public physiotherapy clinics in the Östergötland health care region in Sweden (n=110) involved in a Hybrid type 2 implementation-cluster randomised effectiveness trial (BetterBackJ ClinicalTrials.gov NCT03147300) [15] and clinicians from private primary care clinics (physiotherapists and chiropractors) in Denmark (n=488) involved in a Hybrid type 3 implementation-observational clinical intervention cohort (GLA:D Back) [16,29] were asked to complete the DIBQ-t after a 2-day educational course.
During the course, the clinicians were trained in delivering the programs to patients through lectures and workshops [15]. The items (statements about implementation) in DIBQ-t were scored on a 5-point Likert Scale anchored 'strongly agree', agree', 'neither agree nor disagree', 'disagree', and 'strongly disagree' and for items 5-9 similarly anchored 'very easy', 'easy', 'neither easy nor difficult', 'difficult', and 'very difficult'.
Answers for the DIBQ-t were obtained directly after course participation, thereby representing the expectations of the implementation of the programs shortly after the clinicians had gained detailed theoretical knowledge about the content and execution of the program but had not yet delivered it in practice. Data were collected via a digital platform in Denmark (OPEN REDCap, Vanderbilt University) via a link to the questionnaire emailed to the participants within 24 hours after they had attended a course in 2018. In Sweden, the data was collected using paper-based questionnaires completed immediately upon finishing the course in the period 27th of March 2017 to 30th of January 2018.

Data analyses
Descriptive statistics were used to present qualitative content validity results by the project group, as well as the results of the quantitative content validity testing using CVI scores.
The joint data from the Swedish and Danish clinicians on the DIBQ-t was tested for feasibility, missing data, and construct validity, the latter using a confirmatory factor analysis. Domain and item-level data of the DIBQ-t were analysed as categorical data based upon the results of the 5-point Likert scale. The proportion of clinicians responding 'agree' or 'strongly agree' to each domain, as well as the results for the items in each domain is reported. Ratings of 'agree' or 'strongly agree' indicate that the domain is a facilitating determinant of implementation behaviour, whereas 'disagree' or 'strongly disagree' indicate that the domain is a hindering determinant of implementation behaviour. Data was analysed for missing data and reported descriptively with the proportion of missing data used to judge feasibility.

Phase 1 -Translation of DIBQ in Swedish and Danish
The Swedish and Danish versions of the DIBQ are presented in Appendix 1 and Appendix 2.
There were no important disagreements between the translators. Experts' comments on the translation were incorporated into the questionnaire. Comments involved specific phrasing (questions 2, 18, 86) that could be misunderstood by respondents and concerned questions that were phrased as if the response options were 'yes/no' instead of rated on a Likert-scale (questions 34, 65).

Phase 2 -Adaptation of the DIBQ into the DIBQ-tailored version and content validity assessment.
The project team selected 53 items in 10 domains. The selection of items from the DIBQ to DIBQ-t by the project team and experts is described in Table 2. A total of 28 items in the final Swedish/Danish version of the DIBQ, DIBQ-t were considered feasible to evaluate the implementation process of the programmes. Eighteen items were included as they met criteria 1: selected by the project leaders plus having 80% or 100% CVI of both Danish and Swedish experts. Additional 10 items were included, based upon criteria 3: being selected by the project leaders and had 80% or 100% CVI by either the Danish (DIBQ-t question 8, 9, 21, 23, 24) or Swedish (DIBQ-t question 13, 14, 20, 27, 28) experts, anticipating the differences in context for the countries. No item met the criteria 2 (expert rated relevance score with a CVI = 1.00, regardless of the item being selected by the project leaders). influences, Behavioural regulation. The items and related domains included in DIBQ-t are described in detail in Table 3. Table 3 The DIBQ-t maintains linked to all domains of the Capability -Opportunity -Motivation -Behaviour (COM-B) model within the Behavioural Change Wheel, which is reported in Table   4. Table 4 Phase 3 -Feasibility of the DIBQ-t Sweden, Denmark combined  Analysing each item independently provided a more detailed presentation of the results ( Figure 4). In 21 items, 75% of clinicians scored 'agree' or 'agree strongly'. In seven items, 25% or more clinicians scored 'neither, nor': item 13 -recognition from work (52%), item 28 -planning of the program when patients are not motivated (47%), item 23 and 24 -opinions and support from colleagues (46%, 37%), item 26, 27 -planning how and when to deliver the program (25%, 28%) and item 16 -adapt the program to clinician's need (25%). Items with highest score on 'disagree' were item 27 and 28 (10% and 14% of clinicians) both related to planning the delivery of the program.  (Table 5). Table 5 The estimated factor loadings of the items in relation to the domains are between 0.365 and 0.819 where 3 items (items 12, 15 and 16) were below 0.4, but all items had p-values < 0.001.
The correlation between the domains are between 0.11 (patient and intention) and 0.74 (innovation and consequences) (  Discussion Key findings DIBQ-t is a feasible and valid version of the DIBQ developed to evaluate the expectations of the implementation process regarding best-practice LBP primary care programs. The Regulation' had the lowest frequency of 'agree' or 'strongly agree' item responses, whereas the TDF domains 'Knowledge' and 'Skills' had over 90% of clinicians that strongly agreed or agreed with the items. This suggests that even though the clinicians expect to be capable by having the skills and knowledge to implement, they are less certain on a clear plan on how to operate the program. When analysing the ´behavioural regulationT DF domain on an item-level, planning management of unmotivated patients had a low expectancy by the clinicians. This highlights the importance to include tools for patient motivation in implementation sustainability strategies.
Regarding the COM-B category of Opportunity, in the TDF domain 'Social Influences' the item 23 'influences of important others' had lowest frequency of 'agree' or 'strongly agree' responses. One can therefore consider that providing opportunity for positive influence of important others such as clinical champions is an important aspect to include in implementation sustainability strategies to improve their facilitation longitudinally.
Regarding the COM-B category Motivation, overall, TDF domains had over 80% scoring 'agree' or 'strongly agree'. This implies that, overall, clinicians' intentions, beliefs of capabilities and beliefs of consequences towards implementing the programs were perceived as a highly facilitative for implementation of the best-practice LBP primary care program. However, the TDF domain 'beliefs of consequences', item 13: 'to receive recognition from the work context' had the lowest frequency (40%) of 'agree' or 'strongly agree' responses. This may imply that recognition from work does not have a high importance for motivation the implement the program.

Strengths and limitations
This study was conducted in two different countries with different contexts. The total item-bank of DIBQ-t therefore covers a wide spectrum to evaluate implementation processes generalisable to private and public LBP primary care internationally.
This current study focuses on clinicians´ expectations on best-practice LBP primary care program implementation and confirms a stable construct of the DIBQ-t for monitoring barrier and facilitator qualities of the determinants. Considering that the construct validity applies to the clinicians´ expectations of the implementation process, further analyses are planned after a longitudinal period of volition. Also, studies on changes in different TDF domains over time and their potential mediational role on clinician confidence, beliefs and intervention behaviour will be performed.

Consent for publication: not applicable
Availability of data and material: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.