JPCB Journal

Introduction

Around 40% of patients with bipolar disorder (BD) do not adhere to their prescribed medication regimen, resulting in relapse, hospitalization, increased suicidality, and higher healthcare costs [1-3]. One of the key challenges to successfully addressing nonadherence is identifying an individual’s modifiable barriers and facilitators (determinants) of adherence which is critical for providing personalised support. In the scope of this study, we define modifiable determinants as ‘any determinants of medication adherence that can be modified by the patient, carer, or prescriber within a short timeframe (days or weeks) to improve adherence’[1]. There is a notable absence of questionnaires to identify individual’s adherence determinants in mental health, particularly in BD. Currently available validated adherence questionnaires such as  BARS and MARS-5 [4,5] focus on the nature of adherence, e.g., frequency and magnitude, or explore very limited adherence determinants (e.g., medication beliefs, forgetfulness). Additionally, there is an absence of comprehensive behavioural theory underpinning the development of most adherence questionnaires in mental health despite this being critical for subsequent development of any behaviour change intervention [6,7].

Questionnaire development is a systematic process that involves creating a structured set of questions to elicit specific information from respondents. It requires careful attention to questionnaire design principles, such as crafting clear and unbiased questions and choosing appropriate response formats [8,9]. Boateng et al. provides a best practice guideline for developing and validating a questionnaire [10]. It involves nine steps: 1) identification of the domain(s) and item generation, 2) consideration of content validity, 3) pre-testing questions, 4) sampling and survey administration, 5) item reduction, 6) extraction of latent factors, 7) tests of dimensionality, 8) tests of reliability, and 9) tests of validity.

In our prior research, as a part of domain identification, we conducted: I) a systematic review identifying modifiable adherence determinants in bipolar disorder [1] and II) a qualitative study prioritizing these determinants and identifying new ones [11]. This study focussed on item generation, content validity and pre-testing questions.

Generating items requires careful attention to clarity, tone, structure, and readability, as these elements directly influence response quality and the questionnaire’s validity. The BRUSO model (Brief, Relevant, Unambiguous, Specific, Objective) [9] provides a useful framework for item design. Visual aspects like font and question order also affect engagement [12,13].

Creating clear questionnaire items from modifiable adherence determinants requires input from various stakeholders to ensure consistent interpretation and understanding. Review by subject matter experts such as healthcare professionals is recommended to assess face and content validity of questionnaires. Face validity refers to whether, on the surface, a questionnaire appears to measure what it is intended to measure. Content validity assesses whether the questionnaire comprehensively covers the full range of concepts it aims to measure [14]. Questionnaire review by 5-7 experts is recommended [10]. Pre-testing questionnaires is a crucial part of questionnaire development. Cognitive interviews assess understanding, information retrieval, and response accuracy. Thus, the goal of stakeholders engagement is to assess face and content validity and to ensure questionnaire’s relevance, clarity, and ease of use.

This article describes the development of ‘Collaborative Medication Adherence in Bipolar disorder Questionnaire (C-MABQ)’, a 50-items adherence questionnaire to identify an individual’s adherence determinants. The C-MABQ is grounded in the Theoretical Domains Framework (TDF), a comprehensive behavioural science framework encompassing 14 domains: Knowledge, Skills, Social/ Professional Role and Identity, Beliefs about Capabilities, Optimism, Beliefs about Consequences, Reinforcement, Intentions, Goals, Memory, Attention and Decision Processes, Environmental Context and Resources, Social Influences, Emotion and Behavioural Regulation  [15]. Each C-MABQ question is mapped to relevant TDF domain. And each TDF domain is linked to one or more behaviour change techniques. Thus, the C-MABQ aims to facilitate personalised adherence support by identifying each person’s specific adherence determinant, allowing interventions to be tailored to their needs.

Methods

Ethical Approval

Ethical and governance approvals were obtained from Cambridgeshire and Hertfordshire Research Ethics Committee (Reference:19/EE/0288) and United Kingdom Health Research Authority (IRAS project ID:261687), respectively.

Informed consent was obtained from all the participants.

C-MABQ Development

The development of C-MABQ involved three steps:

1. initial item formulation and refinement
2. Review by healthcare professionals
3. Cognitive interviews with patients to finalise the questionnaire.

Formulating preliminary items and their refinement

Preliminary items for the C-MABQ were developed by AP, based on previously reported modifiable adherence determinants [11]. To ensure clarity, simplicity and to reduce response bias, item construction followed the BRUSO model emphasising brevity, relevance, unambiguity, specificity, and objectivity [9]. Items avoided jargon, acronyms, complex phrasing, and double-barrelled or negatively worded statements. Where adherence determinants involved multiple concepts (e.g., forgetfulness in administering or collecting medication), separate statements were created. A mix of positively and negatively worded items was included to minimise acquiescence bias.

Design elements such as font style, font size, and item order were carefully considered to enhance readability and engagement. The questionnaire began with a broadly applicable, easy-to-answer item to encourage participation. Items followed a logical sequence, with more sensitive questions placed mid-way to maintain respondent comfort.

These preliminary items were then discussed with behavioural science experts, AD (Psychologist with experience of developing adherence scales), CG (Psychologist with expertise in patient related outcome measures) and DB (behavioural scientist with expertise in medication adherence) in group meetings. We held three separate one to two hours meetings to review each item and response options. Items were refined and consolidated resulting in C-MABQv1 (See Supplementary File I).

Review by healthcare professionals to generate C-MABQv2

C-MABQv1 was then reviewed by healthcare professionals with experience of consulting patients and prescribing medication for bipolar disorder. Healthcare professionals currently working in mental health settings, had at least one year of experience in such settings, and were involved in the frontline care of patients with bipolar disorder were eligible to review. Each healthcare professional participated in a brief phone call to clarify the purpose of the review, followed by completion of a structured feedback form (See Supplementary File II). The form included C-MABQv1 items and prompted feedback on clarity, relevance, item order, and potential omissions of adherence determinants. Content validity was assumed if no additional determinants were suggested or if suggestions were deemed irrelevant or inapplicable by researchers. Cognitive interviews with patients to generate final C-MABQ

Patients who had a diagnosis of bipolar disorder, prescribed at least one medication for bipolar disorder, and able to provide informed consent were eligible to participate in cognitive interviews. AP conducted cognitive interviews with patients using C-MABQv2.

Consenting patients received C-MABQv2 before the interview. The cognitive interviews aimed to identify difficulties in understanding or responding to C-MABQv2 items. AP explained the purpose of the interview at the start. Participants read each C-MABQv2 item aloud and responded verbally. AP then asked if they understood the item or found any ambiguity and how they decided their response. Feedback on presentation features (e.g., font style, size, colour) was also sought. Based on the findings, C-MABQv2 was refined to create the final version C-MABQ. Face and content validity were measured subjectively based on responses from patients. C-MABQv2 is considered to have face validity if patients understood and interpreted questions correctly, responded without difficulty, and suggested minimal or no changes during cognitive interviews. Similarly, content validity was considered to have established if no new determinants was identified.

The final C-MABQ was evaluated for word frequency and readability. To ensure clarity and ease of understanding, it is recommended to keep items to 15 words on average and not longer than 25 words [16-18]. Specifically, sentences with 8 words or fewer are very easy to read; 11 words are easy; 14 words are fairly easy; 17 words are standard; 21 words are fairly difficult; 25 words are difficult; and 29 words or more are very difficult [16-18]. These parameters were applied to support comprehension and minimise cognitive load during questionnaire completion. A Flesch–Kincaid reading ease score of 60+ or grade level  ≤ 8 (readable by ages 11-12) - recommended for patient materials- indicates good readability and accessibility [19,20].

Results

Development of C-MABQv1 

AP formulated 75 statements representing the 50  previously reported and prioritised modifiable determinants of adherence [11]. Multiple statements represented each determinant to allow for variations such as negatively and positively phrased and differences in the vocabulary used to represent the determinant. For example, four statements were generated for the determinant ‘Not knowing the risks of stopping medication’: I) I am unsure about the risks of stopping my medications. II) I understand very well the risks of stopping my medications. III) I don’t know what would happen if I stop taking my medications. IV) I know that I will become unwell if I stop taking my medications.

Following discussion with experts, 51 out of the 75 statements were selected (See supplementary file I). We opted for a five-point Likert scale response format. The items were divided into three sections with following response options:

1)  Very Easy, Easy, Neutral, Difficult, Very Difficult,

2)  Strongly Agree, Agree, Neither Agree nor Disagree, Disagree, Strongly Disagree

3)Totally Acceptable, Acceptable, Neutral, Unacceptable, Totally

Unacceptable

Review by healthcare professionals to evaluate face and content validity

Six healthcare professionals (two psychiatrists, two mental health pharmacist independent prescribers and two psychiatric nurse independent prescribers) were recruited for review of C-MABQv1.

The review informed the following changes to C-MABQv1 statements:

• Response option ‘Neutral’ in section one and ‘Totally Acceptable’ in section three were changed to ‘Neither easy nor difficult’ and ‘Perfectly Acceptable’ respectively.
• ‘Condition’ was changed to mental health condition in statements 8, 9, 16, 39 and 40.
• ‘Healthcare team’ was changed to ‘mental health team’ in statements 4, 17 and 18.
• ‘prescriber’ was changed to ‘mental health team’ in statement 19.
• ‘illness’ was changed to ‘mental health condition’ in statement 20

Further changes to C-MABQv1 are presented in supplementary file III. Apart from the changes suggested, the healthcare professionals agreed that the statements were clear and represented the determinants of adherence.

Although healthcare professionals considered the list of determinants comprehensive, some suggested additional determinants: not liking the effects of their mood being artificially flattened or controlled, low mood, substance misuse, knowing which medications are for bipolar disorder, being able to talk to the healthcare team about side effects, knowing where to find good quality information about my medications and side effects. However, all of these were either considered to have been addressed (with slightly different wording) or non-modifiable within the context of this study.

Cognitive interviews with patients

Seven patients; four females and three males aged 27 to 69, were recruited for the cognitive interviews. All patients understood and interpreted most of the C-MABQv2 statements and responded as intended suggesting good face validity. The presentation, layout, font size, colour and style were well received, with no change recommended. During cognitive interviews, patients recommended very minimal changes to the C-MABQv2 as shown in supplementary file IV. When there were two different statements for the same determinants, patient preferred statements were kept for final version C-MABQ.

Statements 31 and 32 represented the same concepts but were checked with patients for their preferred statement. Four patients preferred statement 31 over 32 and others did not have any preference, therefore we chose statement 31 for the final C-MABQ which now has a total of 50 statements – referred to as ‘C-MABQ’ from here onward. Two statements (5 and 15) initially phrased as ‘unsure’ were recommended to be changed to ‘don’t know’.

The final C-MABQ statements, their reordering as recommended by healthcare professionals and patients and their corresponding TDF domains are presented in table 1 below. For the full final C-MABQ with instructions, response options and scoring method see supplementary file V. 

Table 1: The Final C-MABQ.