A Qualitative Study of Barriers to Medication-Taking Among People With Type 2 Diabetes Using the Theoretical Domains Framework

Thirty individuals were recruited for 45-minute one-on-one interviews. Participants were recruited from a general sample of Canadians who have previously consented to be contacted to complete interviews for research. These people were initially recruited through referrals, social media, advertising, and patient advocacy groups. To participate in this study, individuals were required to have been living with type 2 diabetes for at least 1 year (i.e., since 2020) and to be currently on medication(s) to manage their type 2 diabetes. Quota sampling was used to ensure that the distribution of individuals included representation from various geographical regions in Canada, different medication types, and different levels of medication-taking.

An interview discussion guide was developed, informed by the results from our previous scoping review (21), inputs from a national steering committee of clinical experts in diabetes, prior TDF medication-taking research (19), and the TDF (18) (Supplementary Material). The discussion guide included background questions on participants’ history with type 2 diabetes and medications and questions exploring issues related to medication-taking for each of the 14 TDF domains. One-on-one, 45-minute telephone interviews were conducted from 8 July to 12 August 2021 in English or French by one of three interviewers. After receiving consent from participants, the interviews were digitally recorded, and themes (barriers) were identified for analysis. Participants were classified by sex, geographical region (Atlantic Provinces, Quebec, Ontario, or Western Provinces), years since type 2 diabetes diagnosis, types of medication(s) (metformin only, oral medications other than or in addition to metformin, insulin with or without metformin or other oral medications, or noninsulin injectable medications with or without metformin, other oral medications, or insulin), and levels of medication-taking (low, medium, or high).

Levels of medication-taking were determined based on answers to the Morisky, Green, Levine Medication Adherence Questionnaire (MGL-MAQ) (22). The MGL-MAQ is a concise, reliable, and validated self-reported scale to measure medication-taking behaviors in people with chronic diseases (23,– 25). This four-item scale includes yes/no questions, which are framed in reverse wording to overcome yes-saying bias. These items are 1) Do you ever forget to take your medicine? 2) Are you careless at times about taking your medicine? 3) When you feel better, do you sometimes stop taking your medicine? and 4) Sometimes if you feel worse when you take the medicine, do you stop taking it? Individuals who answered “No” to all questions were considered to have high engagement, those who answered “Yes” to one or two questions were considered to have medium engagement, and those who answered “Yes” to three or four questions were considered to have low engagement.

An inductive approach was used to analyze the qualitative data collected from the interviews (18). Transcripts from all interviews were reviewed and condensed into brief summaries and further summarized into overarching belief statements (i.e., themes) by two reviewers (G.N. and Andrean Bunko of IQVIA Solutions Canada, Inc.) (26). The themes were then mapped to the best fitting of the 14 TDF domains listed in Table 1 by a single reviewer (G.N.). The mapping of themes into domains was interrogated and confirmed by a second reviewer (Andrean Bunko) (18) and subsequently reviewed by all authors. Domains were considered relevant if they had a high frequency (i.e., including several belief statements mentioned by multiple respondents) or included belief statements that contrasted between respondents who were considered high versus low engagement.

The demographic characteristics of the 30 people with type 2 diabetes interviewed for this study are presented in Table 2. Of the included individuals, 14 (47%) were female, with the largest proportion from Ontario (43%) and smallest proportion from the Atlantic Provinces (7%). The average time since diagnosis of type 2 diabetes for all individuals was 12.9 ± 7.9 years (range 1–30 years), and half (50%) of the individuals reported currently taking more than one medication to manage their type 2 diabetes. Among all participants, 9 (30%) were taking metformin only, 9 (30%) were taking oral medication(s) other than or in addition to metformin, 3 (10%) were taking insulin with or without metformin or another oral medications, and 9 (30%) were taking a noninsulin injectable medication with or without metformin, other oral medication(s), or insulin. In all, 10 individuals each were classified into the low, medium, and high medication-taking categories. Table 2 presents the participant characteristics stratified by level of medication-taking.

Forty-two themes were identified and mapped to 12 of the 14 theme domains, with the knowledge and skills domains being the exceptions. The most prominent domains, as determined by high-frequency themes or themes for which people with low and high medication-taking had contrasting perspectives were emotion (2 themes); memory, attention, and decision processes (6 themes); behavioral regulation (5 themes); beliefs about consequences (8 themes); goals (3 themes); and environmental context and resources (4 themes). The key themes from the most prominent domains are discussed below, along with illustrative quotes captured during the interviews. A list of all of the identified themes is provided in Supplementary Table S1.

Two identified themes mapped to the emotion domain and illustrate a contrast in mindset between individuals who reported low versus high medication-taking engagement. People with type 2 diabetes who reported low engagement described feeling overwhelmed by their medication plans and diabetes self-management (Supplementary Table S1, theme 13.1).

Similarly, it was found that people with type 2 diabetes who reported low engagement often associated taking medication and living with diabetes with negative emotions (e.g., feelings of annoyance, frustration, shame, and regret), which contrasted with those reporting higher engagement, who spoke about taking their medication with little emotional connotation, but rather as something that just needed to be done (Supplementary Table S1, theme 13.2).

Themes that mapped to the memory, attention, and decision processes domain were prominent in the interviews, with contrasting accounts from people with low medication-taking compared with their medium- or high-engagement counterparts. For example, those who reported low engagement stated that they often forget or do not take their medication with them when leaving home for errands or social events or when traveling for longer periods of time, an issue not raised by individuals who reported high engagement (Supplementary Table S1, theme 10.1).

Participants reporting low engagement also often stated that they choose to skip doses rather than delay them when they were running late or had conflicting commitments (Supplementary Table S1, theme 10.2).

Finally, there was a difference noted between the way participants reporting low or high engagement talked about taking medication as a priority. Those reporting low engagement more often believed that their diabetes could be managed by diet and/or exercise alone (Supplementary Table S1, theme 10.4).

Additionally, people reporting low engagement often discussed forgetting about their medication, including forgetting to take medication at all and forgetting which medications they have already taken (Supplementary Table S1, theme 10.5).

Five themes were identified in the behavioral regulation domain, highlighting that people with type 2 diabetes and low reported engagement may not have integrated taking their medication into their daily schedules or did not have an established routine. This was in contrast to people with higher levels of medication-taking, who commonly had an established routine for taking their medications (Supplementary Table S1, theme 14.1).

Similarly, participants reporting low engagement less often mentioned using resources such as pill boxes and alarms to stay organized and remind themselves to take their medication. Even when resources were available, these participants did not use them as consistently as their high-engagement counterparts (Supplementary Table S1, theme 14.2).

Finally, it was also noted that disruptions to established daily routines often made it easier to miss taking medications as recommended (Supplementary Table S1, theme 14.5).

Eight themes were identified in the beliefs about consequences domain—the greatest number of themes across the TDF domains. Within these themes, people reporting low engagement were more often unaware of the consequences of not taking their medication or were less likely to take medication as prescribed because of perceived or actual side effects that disrupt their lifestyles and schedules (Supplementary Table S1, theme 6.2).

Three themes from our interviews were mapped to the goals domain. Of these, it was noted that some participants reporting low engagement wanted to lower their doses and minimize the number of medications they take or eliminate the need for medication altogether (Supplementary Table S1, theme 9.2).

Another theme that was commonly discussed was that some participants were not motivated to take medication by setting and focusing on long-term goals. Those reporting high engagement often talked about having more resolute long-term goals that involved loved ones (Supplementary Table S1, theme 9.3).

Four identified themes mapped to the environmental context and resources domain, including some people discussing that they are affected by high costs of and limited insurance coverage for diabetes care, including medications and glucose monitoring, and these problems negatively affect medication-taking (Supplementary Table S1, theme 11.1).

Furthermore, participants who reported low engagement suggested that they may be influenced by unreliable sources of information that negatively affect their medication-taking (Supplementary Table S1, theme 11.3).

Through qualitative interviews, we identified themes and barriers affecting medication-taking among people with type 2 diabetes in Canada using the TDF, with six prominent theme domains identified. These domains were 1) emotion; 2) memory, attention, and decision processes; 3) behavioral regulation; 4) beliefs about consequences; 5) goals; and 6) environmental context and resources. To our knowledge, this is the first study to apply the TDF to qualitative one-on-one interviews with people with type 2 diabetes to provide a greater understanding of the situational context and drivers of medication-taking behavior.

The results from our study extend those of prior studies in several ways, providing further context for the ways in which complex dosing plans, tolerability of medications, and perceptions of risks and benefits of medications (13) contribute to engagement. These findings also present additional context for how both unintentional nonengagement behaviors, such as forgetfulness and beliefs about the need for medication, and intentional nonengagement behaviors contribute to taking medication as recommended (27,28).

Polonsky and Henry (4) emphasized that innovative strategies to improve the attitudes of people with type 2 diabetes toward medication or encourage habit formation regarding medication-taking may help increase engagement. Our study used the TDF to identify specific intentional and unintentional behavioral influences on medication nonengagement that were common among people with low medication-taking. This information was further enriched by contrasting the attitudes and behaviors of people with higher medication-taking to derive recommended areas of change. By leveraging findings from our approach, strategies to increase medication-taking can be tailored to target specific behavioral influences and contextual challenges.

Our results can be applied to the development and application of effective interventions to increase medication-taking, and these should take into account the broader TDF theme domains we identified, as well as target specific themes within each domain (29). For example, people with type 2 diabetes with low engagement experience emotional challenges associated with being overwhelmed by living with diabetes and needing to take medication. Therefore, diabetes self-management support interventions might benefit from incorporating emotion management strategies for people who struggle with diabetes and treatment acceptance (Supplementary Table S1, theme 13.1). Diabetes self-management education that provides concise and reliable diabetes and medication knowledge might mitigate issues associated with individuals’ past experiences of having unreliable information sources (Supplementary Table S1, theme 11.3) and also address feelings about medication-taking (Supplementary Table S1, theme 6.2). The individuals expressed a desire for information that is simple and easily accessible, supports setting expectations regarding side effects, and increases their understanding of the long-term benefits of treatment. Furthermore, we suggest that improving access to resources such as diabetes clinics, pharmacists, and support groups where people with type 2 diabetes can ask questions, receive coaching, and get essential education and training would be useful. These strategies would provide an opportunity to reframe negative feelings about diabetes and medications (Supplementary Table S1, theme 13.2). These approaches could also specifically target identified themes such as the importance of long-term goal setting (Supplementary Table S1, theme 9.3) and understanding the role of medications in type 2 diabetes treatment and the progressive nature of the disease (i.e., that it usually requires long-term medication use) (Supplementary Table S1, themes 9.2 and 10.3).

Likewise, practical support that facilitates medication-taking and increases accessibility may help to increase engagement. It was apparent in this study that disruptions to routines affected individuals who reported low engagement more than those with medium or high engagement, resulting in unintentionally or intentionally forgetting, skipping, or delaying medication doses (Supplementary Table S1, themes 10.1, 10.2, and 14.5). This finding highlights an important consideration when selecting medication plans, which should incorporate individuals’ convenience and lifestyle considerations to encourage engagement.

Our findings also highlight the importance of people with type 2 diabetes establishing a medication-taking routine. A number of participants reporting low engagement did not use reminders, whereas those with high engagement had a clear and defined process, including organization resources such as pill boxes and pharmacy-prepared personalized blister packs and tracking resources such as phone alarms, calendars, and diaries (Supplementary Table S1, theme 14.2). Thus, encouraging or enabling the use of such tools among people with type 2 diabetes may increase engagement, although success in using these tools may be more a result of individuals’ motivations and organizational skills rather than the mere availability of the tools.

There is also a role for pharmacist teams to identify people with or at risk for low engagement and to support these individuals through actions such as sending refill reminders, dispensing medication in personalized blister packs, and accommodating emergency medication needs, in addition to having conversations with and supporting people with type 2 diabetes in individualized appointments.

Finally, financial barriers such as high costs of medications and glucose monitoring supplies (Supplementary Table S1, theme 11.1) can be diminished through health system changes (specifically, covering diabetes-related care products and medications) and through support programs.

There are limitations to our study that should be considered for proper interpretation of the results. First, although quota sampling was conducted to ensure diversity among the participants, little diversity was reflected in recruited participants. Therefore, our participants may not have been representative of all people with type 2 diabetes in Canada. Second, social desirability bias, in which respondents tend to provide answers that overreport desirable and underreport less desirable attributes is a known concern with interviews. However, this problem was mitigated by our comprehensive discussion guide, which approached the topic of medication-taking from many perspectives, as well as by contrasting the responses of people with lower and higher medication-taking engagement. Third, although generating data using the TDF has been shown to be a comprehensive and inclusive approach to exploratory research, it is still a descriptive framework rather than a theory (30). The results generated do not specify relationships between the domains and do not generate testable hypotheses (31). Finally, we recognize that qualitative interviews about behavioral influences represent the perceptions of the individuals interviewed and may not reflect the actual causes of their behaviors or be generalizable to broader populations (30).

Our study identified behavioral influences contributing to lower medication-taking engagement, highlighting key areas for change among a small group of people with type 2 diabetes. Future research is warranted to validate these findings within a larger sample and to explore the suitability of various intervention and implementation options to improve medication-taking (32). Additionally, the results of this study can inform the development of medical education and training programs for health care providers supporting people with type 2 diabetes in Canada.

The authors thank the people with type 2 diabetes who took part in qualitative interviews for this study. They also thank Carole Hamersky, PhD, and Cory Gamble, MD, of Novo Nordisk Canada, Inc., for expert advice on developing the qualitative discussion guide. The authors thank the following individuals from Real World Solutions, IQVIA Solutions Canada, Inc.: Andrean Bunko, MSc, and Pierre-François Meyer, PhD, for their support conducting interviews, coding, and reviewing theme domain mapping; Atif Kukaswadia, PhD, and Calum S. Neish, PhD, for their guidance and oversight throughout the project; and Vibha Dhamija, MSc, and Saurabh Trikha, PhD, for their assistance with medical writing.

This study was funded by Novo Nordisk Canada, Inc.

M.V., S.J., and N.M.I. received advisory fees from IQVIA Solutions Canada, Inc., for contributions to the design and interpretation of the study but received no compensation for development of the manuscript. A.K.-A. is an employee of Novo Nordisk Canada, Inc. G.N. is a consultant employed by IQVIA Solutions Canada, Inc., to support this study sponsored by Novo Nordisk Canada, Inc. No other potential conflicts of interest relevant to this article were reported.

All authors reviewed the TDF mapping results and participated in the interpretation of results, reviewed and revised the manuscript, and approved the final manuscript. G.N. conducted the interviews, coded the themes, and conducted the theme domain mapping. G.N. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Parts of this article were presented in abstract form at the American Diabetes Association’s 82nd Scientific Sessions, 3–7 June 2022.