Experiences of using the DiabeText mobile text messaging intervention to support diabetes self-management for people living with type 2 diabetes: Post-hoc process evaluation

Abstract

Objective

Poor adherence to prescribed treatment for people with type 2 diabetes increases the risk of complications, mortality, and associated healthcare costs. Mobile phone text-messaging interventions such as DiabeText can support diabetes self-management and improve health outcomes. This study aimed to explore patients’ experiences with the DiabeText intervention and the contexts and mechanisms through which behavioral change might occur.

Methods

A post-hoc process evaluation using a mixed-methods design was conducted within primary care settings in Spain. The analysis combined a questionnaire-based study with 371 participants, aimed at rating the behavior change techniques embedded in the text messages, and a qualitative interview study with 19 participants. Quantitative data were analyzed descriptively, while qualitative data underwent thematic analysis.

Results

Participants reported predominantly positive experiences with the intervention, highlighting increased support, motivation, and self-care capacity. They valued the reliable and practical content as well as the user-friendly format, which were perceived as facilitating healthier lifestyle changes and supporting medication adherence. However, participants also suggested incorporating greater personalization to enhance acceptability. The mechanisms through which the intervention may foster behavioral change likely relate to the inclusion of behavior change techniques (BCTs) previously associated with positive outcomes—and which participants rated as highly helpful in supporting change. Nevertheless, further research is needed to determine the specific active ingredients that would maximize DiabeText’s overall impact.

Conclusion

DiabeText was well received by individuals with T2DM to support self-care. Introducing personalized features and messages that utilize BCTs previously associated with better outcomes could be a good starting point for better outcomes.

Background

Diabetes mellitus results from insufficient insulin production due to various factors, including genetics and the environment.^1–3 Globally, approximately 537 million people are affected, with type 2 diabetes mellitus (T2DM) being predominant. T2DM is associated with numerous complications, ranging from microvascular to macrovascular issues, leading to significant healthcare costs. ⁴ Therefore, prevention of complications and enhancing quality of life are paramount in diabetes management. ⁵

Traditionally, diabetes management involves in-person follow-up by primary care healthcare professionals, who prescribe lifestyle changes and medication. Like other chronic conditions, diabetes requires significant time and effort from practitioners to educate, support and help patients maintain optimal self-management. Owing to its high global prevalence, mobile health (mHealth) initiatives offer an innovative approach to enhance diabetes care, providing ongoing support and motivation. These initiatives pose challenges in achieving sustained behavioral change, ⁶ but they also hold promise for improving health outcomes among individuals living with diabetes. ⁷

Systematic reviews and meta-analyses^8–12 indicate that mobile phone text messaging interventions can positively influence diabetes management, particularly by improving medication adherence and supporting glycemic control. However, their effectiveness in promoting behavioral changes remains inconsistent, and other health outcomes–such as quality of life, self-efficacy and psychological well-being–are often underreported.

In this context, we conducted a large-scale randomized controlled trial (RCT) ¹³ to evaluate the effectiveness of a text message-based intervention. The trial demonstrated improvements in self-reported medication adherence, health-related quality of life, and diabetes self-efficacy, while other outcomes, such as glycemic control and lifestyle behaviors, remained unchanged. These mixed results underscore the need to understand the mechanisms underlying these effects and the contextual factors influencing intervention success. The added value of this post-hoc process evaluation lies in its ability to go beyond the initial RCT outcomes by exploring participants’ experiences and perceptions, identifying which components were perceived as helpful, and uncovering barriers and facilitators to behavior change. Such insights are essential for refining intervention content, improving theoretical alignment, and optimizing the design of future digital tools to maximize their impact on health outcomes.

The aim of this process evaluation was to explore patients’ experiences and perceptions after 12 months of receiving the DiabeText intervention, with a focus on understanding the contexts and mechanisms through which behavior change might occur. Specifically, the study sought to i) identify key features of the intervention perceived as helpful; ii) assess the acceptability and usability of the text messaging approach, and; iii) gather insights to inform potential improvements that could better meet the needs of the target population and enhance health outcomes such as medication adherence and glycemic control.

Methods

DiabeText is a digital tool that sends personalized text messages via mobile phones to enhance treatment adherence in patients with T2DM. ¹⁴ The design of the intervention was informed by patients ¹⁵ and health care professionals ¹⁶ and then piloted and refined ¹⁷ before its effectiveness was evaluated. ¹³ Here we present a post-hoc process evaluation^18,19 of the DiabeText RCT (NCT05006872), integrating quantitative survey data and qualitative interviews to explore the mechanisms of impact and the contextual factors that may influence the implementation, functioning, and outcomes of the intervention. To this end, we conducted a mixed-methods study within six months after the RCT’s conclusion, carried out within the primary care centres of the Balearic Islands, Spain, a setting characterized by a robust public health infrastructure and a specific bilingual (Spanish and Catalan) cultural context. The implementation evaluations regarding fidelity, dose, adaptations, reach, participant satisfaction, and adverse effects have been published previously. ¹³

Questionnaire study

Psychological mechanism evaluation

Text messages from DiabeText were designed according to many different behavioral change techniques (BCTs)^13,14,17 following Michie et al.’s taxonomy ²⁰ and BCT candidates to support medication adherence by Long et al. ²¹ However, 50% of the messages focused on four BCTs: “how to perform behavior”, “health consequences”, “habit formation” and “verbal persuasion to increase self-efficacy”. Researchers of the DiabeText RCT subsequently developed a questionnaire on the degree to which these four BCTs helped participants to take their medication well or follow diet and exercise guidelines in a similar way as Bartlett et al. did it previously. ²² The response categories followed a 5-point Likert scale ranging from 1=not helping me at all to 5=helping me a lot (Supplemental material 1). Additionally, the internal consistency of the questionnaire was explored by calculating Cronbach’s alpha coefficient for the four items assessing the perceived helpfulness of the most frequently used BCTs. The questionnaire was administered via telephone to participants in the intervention group by a team of trained evaluators between the months of November and December 2022 as part of the final trial interview. Eligibility required adults with type 2 diabetes registered in the Balearic Islands public health system who had an HbA1c >7.5% within the preceding six months, at least one prescription for a non-insulin antihyperglycemic agent, and the capacity to receive and comprehend Spanish-language SMS messages (or a caregiver able to do so). Individuals with severe mental health conditions or concurrent participation in other research studies were excluded. The participants’ informed consent was recorded telephonically by trained research assistants when they agreed to be part of the trial ¹³ approved by the Ethical Research Committee of the Balearic Islands under the identifier IB 4320/20 PI. The results are presented as frequencies and the total number of responses per value across all participants for each question.

Qualitative interview study

Participants

We preselected 52 participants from the DiabeText RCT intervention group to invite them to individual qualitative interviews, seeking a diverse sample in terms of demographic and clinical characteristics (age, gender, duration of diabetes, number of prescribed medications, self-reported medication adherence, illness experiences, and familiarity with digital devices, particularly regarding Internet use on their mobile phones) relevant to diabetes self-management,^23–27 to ensure the generation of rich information ²⁸ and representativeness of the sample. Two interviewers (EMME and RZC), both trained in qualitative interviewing, contacted potential participants by phone between January and April 2023. However, 25 individuals did not respond to phone calls after multiple attempts, meaning direct contact was successfully established with only 27 potential participants. Of those reached, 19 consented to participate, representing a 70.4% acceptance rate among contacted individuals. The remaining eight participants who declined did so primarily due to logistical constraints and scheduling conflicts between their work commitments and the researchers’ availability during the specified interview period. Informed consent was audio-recorded according to previously approved protocol IB 4320/20 PI.

Data collection

Interviews were conducted between February and June 2023, in Spanish or Catalan, either in a private room at the primary care centre or via secure telephone call, depending on participant preference. All sessions took place in quiet, uninterrupted settings. The topic guide (Supplemental material 2), which underwent several modifications as data collection progressed, included a global evaluation of the intervention’s utility and satisfaction, as well as key topics related to the development of digital solutions for people with chronic diseases. ²⁹ It was designed with open-ended questions explicitly intended to elicit critical and neutral feedback without leading the participants. To ensure a balanced representation, specific prompts were included to encourage participants to share aspects they ‘liked the least,’ suggest ‘what could be improved,’ or identify any content they ‘didn’t like or found upsetting’. These prompts were specifically included to ensure that participants felt comfortable sharing dissenting perspectives alongside positive experiences. Interviews, which lasted 15-40 minutes, were audio-recorded and transcribed verbatim by EMME and cross-checked by RZC. No prior relationship was established before the study; participants only knew the interviewer as a researcher involved in the study, and no specific personal characteristics, biases, assumptions, or interests of the interviewer were disclosed or emphasized. Confidentiality and anonymity were ensured by assigning study codes and removing identifiers from recordings and transcripts, which were stored on encrypted institutional servers. Data saturation was determined through an iterative and reflexive process conducted concurrently with data collection. The researchers (RZC and XCA) and the interviewer (EMME) met regularly to review transcriptions and field notes, assessing the information gathered in relation to the study’s objective of exploring the 12-month experience with DiabeText. Saturation was considered to have been reached when three consecutive interviews yielded no new themes, conceptual changes, or additional information on perceived impact mechanisms or contextual barriers, confirming that the sample of 19 participants provided sufficient explanatory depth to address the research objectives, regardless of the diversity criteria sought in the characteristics of the participants.

Data analysis

A thematic approach based on Braun and Clark’s methodology^30,31 was used to identify recurring patterns and generate initial themes. An interpretive approach was subsequently applied to explore the underlying meanings and contextual nuances of the emerging themes. Although internal validity cannot be guaranteed in interpretive qualitative research, efforts have been made to ensure the consistency and trustworthiness of the analysis.^28,30 Specifically, 8 members of the research team independently analyzed four interviews each, conducting line-by-line reading and identifying meaningful units that informed the initial inductive coding scheme. These independent coding were then discussed in pairs during two collaborative sessions to compare interpretations, reconcile discrepancies, and refine preliminary themes. Following this, the full research team participated in three analytic workshops where themes were critically reviewed, adjusted, and validated through group consensus. This iterative and reflexive process–combining independent coding, pairwise comparison, and team-level consensus–served as our approach to achieving inter-rater reliability within an interpretive qualitative paradigm. Also, we employed an iterative analysis where we specifically sought out divergent interpretations and ‘dissenting voices’ during our analytic workshops to ensure a balanced representation of the data. Consistency was ensured through analyst triangulation, repeated discussions, and shared interpretive decision-making. Following the initial inductive coding, a deductive approach was applied to organize and refine the themes using the COM-B model.^20,32–35 This framework—which posits that behavior change occurs through the interaction of Capability (psychological and physical), Opportunity (social and physical), and Motivation (reflective and automatic)—served as our overarching methodological reference. By mapping the emergent qualitative categories onto these three components, we were able to systematically explore the mechanisms of action through which DiabeText may influence participant behavior. As part of discrepancy resolution, particular attention was given to divergent interpretations related to the perceived motivational impact of the text messages. To address this, the team jointly examined verbatim excerpts to ensure that even minority perspectives were accurately captured and interpreted. This multi-layered approach enhanced the credibility, coherence, and dependability of the analysis. No qualitative analysis software (e.g., NVivo, Atlas. ti) was used. All coding and theme refinement were conducted manually by the research team.

Researcher reflexivity

The research team comprised professionals with diverse backgrounds including pharmacy (EMME), nutrition (MCR, RZC), primary care & public health (IRC, JRA, JMTA), social science (XCA), psychology (MJSR), and digital health (IRC, MJSR, RZC), the majority of whom had prior experience in diabetes management, qualitative research and behavioral interventions. Several members of the team were also involved in the design and development of the DiabeText intervention, which may have influenced data collection, coding, and interpretation through expectations regarding its potential benefits or mechanisms of action. To mitigate this, the research team included both interviewers and analysts with no prior involvement in the intervention alongside those with in-depth knowledge of the tool and the RCT. Data collection followed a structured topic guide designed to elicit both positive and negative experiences. During analysis, multiple researchers independently coded transcripts and engaged in iterative discussions to challenge interpretations and identify divergent perspectives. The team also actively sought data that did not align with initial assumptions, ensuring that both supportive and critical views were adequately represented.

Statistical analysis

To assess the representativeness of the samples, the baseline characteristics of the interview study participants (n=19) were compared with those of the total intervention arm of the DiabeText trial (n=371). Since continuous variables followed a non-normal distribution, they are expressed as Median and Interquartile Range (IQR) and were compared using the Wilcoxon rank-sum test. Meanwhile, categorical variables are presented as frequencies and percentages (n,%) and were analyzed using the Chi-square test or Fisher’s exact test (when expected frequencies were less than 5). Results were considered statistically significant when the p-value was < 0.05. It is important to note that statistical comparisons were not necessary for the questionnaire-based study, as it included the entirety of the participants (n=371) from the intervention group.

Integration of quantitative and qualitative data

To ensure a meaningful integration of quantitative and qualitative strands, we adopted a convergent mixed-methods approach. Both datasets were analysed independently and subsequently integrated through a process of joint interpretation. Integration was guided by the identification of convergence, divergence, and complementarity between findings. Quantitative results were compared with qualitative themes to assess whether they confirmed, expanded, or contradicted each other.

Results

Participants characteristics

Participants had a median age of 66[59-73] years, and 42% were women. They had been living with T2DM for an average of 11[5-15] years and were prescribed between 1 and 11 medications including antidiabetic drugs, with 55-58% reporting being adherent to them (Table 1). Overall, both participants in the questionnaire (n=371) and interview studies (n=19) were representative of the total intervention arm (n=371) in almost all sociodemographic and clinical variables, including age, gender, and baseline adherence. The only notable differences in the interview sample were a trend toward lower obesity (37% vs 56%), higher depression (37% vs 20%) and a statistically significant difference in the number of prescribed medications (p=0.017), with the interview group reporting a lower median of 4 medications [IQR 3–8] compared to 7 [IQR 4–10] in the larger intervention arm. However, the qualitative interviews included individuals with obesity and normal weight, with and without depression, and taking both lower and higher numbers of medications, so their diverse potential perspectives were represented and reflected in the collected data.

Table 1.

Characteristics of participants in their respective baseline for the RCT DiabeText ¹³ who participated in the post hoc process evaluation.

​	Participants from the intervention arm (n=371)	Questionnaire study (n=371)*	Interview study (n=19)	Statistical comparison (Interview vs. Intervention arm) (p)**
Sociodemographic characteristics
Women (n, %)	154 (42)	154 (42)	9 (42)	0.789
Age (years) (median, IQR)	66 [59-72]	66 [59-72]	65 [61-73]	0.895
Current smoker (n, %)	94 (25)	94 (25)	5 (26)	0.873
Use of internet in the mobile phone (n, %)	268 (73)	268 (73)	15 (74)	0.707
Clinical characteristics
Time since T2DM diagnosis (years) (median, IQR)	11 [6-15]	11 [6-15]	10 [5-15]	0.648
Number of drugs prescribed at the moment (median, IQR)	7 [4-10]	7 [4-10]	4 [3-8]	0.017
Adherent to antidiabetic medication (n, %) – based on self-reported outcomes	205 (55)	205 (55)	11 (58)	1
Hypertension (n, %)	263 (71)	263 (71)	13 (68)	1
Obesity (n, %)	209 (56)	209 (56)	7 (37)	0.153
Depression (n, %)	74 (20)	74 (20)	7 (37)	0.086

Note. IQR: Interquartile Range (Q₁–Q₃).

*Includes all RCT intervention group participants (n=371).

**Continuous: Median [IQR] (Wilcoxon test); Categorical: n (%) (Chi-square/Fisher’s). Bold: p < 0.05.

Quantitative results from the psychological mechanism evaluation

All the participants in the intervention group that completed the trial (n=334) scored the four BCTs of the questionnaire on the basis of their experiences after receiving the intervention for 12 months (Table 2).

Table 2.

Ratings of the behavior change techniques most frequently used in DiabeText messages

Likert scale	1=not helpful at all n (%)	2=slightly helpful n (%)	3=moderately helpful n (%)	4=very helpful n (%)	5=extremely helpful n (%)	Cronbach’s alpha
Instruction on how to perform a behavior	19 (6)	18 (5)	48 (14)	95 (28)	154 (46)	0.92
Information about health consequences	21 (6)	15 (4)	45 (13)	87 (26)	166 (50)	0.92
Habit formation	24 (7)	16 (5)	53 (16)	84 (25)	157 (47)	0.93
Verbal persuasion to boost self-efficacy	20 (6)	12 (4)	46 (14)	86 (26)	170 (51)	0.92

Data are distribution of frequencies as percentages (%) and total number of responses (n) from the total of respondents (N=334).

The four techniques were scored very similarly: approximately 75% of the participants considered that these techniques helped them (Likert scale=4-5) to take their medication as prescribed and follow lifestyle recommendations; 14% scored each technique with a neutral score (Likert scale=3), so they were not very sure if BCTs helped them in some way or not. Approximately 11% thought that any of these techniques helped them to make changes to improve diabetes self-management (Likert scale=1-2). The questionnaire showed high internal consistency across the four items (Cronbach’s alpha = 0.94).

Interview study results

We conducted 19 qualitative individual interviews with participants from the intervention group of the DiabeText RCT. Two overarching themes emerged: first, “DiabeText is a helpful digital tool for supporting diabetes care”; second, “People with T2DM suggest new personalized features to improve DiabeText uptake and engagement.” Participants’ accounts revealed both contextual factors and potential mechanisms of change, offering insight into how the intervention was experienced and how it might be optimized.

DiabeText is a helpful digital tool for supporting diabetes care

Participants felt practically and emotionally cared for

The majority of participants expressed satisfaction with the intervention. As they received text messages throughout the year, the participants reported feeling supported and cared for, both in a practical and emotional way. This support was also irrespective of specific self-care information contained within texts (such as diet, physical activity, medication, etc.). Consequently, the messages appeared to fulfill participants’ needs for support related to T2DM, surpassing the mere content of the messages. The participants viewed the messages as an additional aspect of their healthcare, and reported that the messages were perceived to contribute to a sense of being better cared for by their doctors.

“Knowing that someone was on the other side of the message made me feel very good because I needed someone to be there for me.” Woman, 64 years old, 17 years since T2DM diagnosis

“ I felt the closeness a lot. It gives you the impression that someone is interested in your health… it stimulates you and helps you overcome bumps sometimes.” Woman, 74 years old, 6 years since T2DM diagnosis

DiabeText provides trusted expertise and easy access to reliable information to support diabetes self-care

The participants expressed trust in the information they received. They found the messages to be an easy and convenient ways to access this information, noting that the content was clear and easily understood.

“It appears to be they are written by professionals, because they are well-written and contain good information.” Woman, 71 years old, 18 years since T2DM diagnosis

“These messages allow me not to be tied to specific appointments with healthcare professionals. This flexibility is helpful as it allows for agility and comfortable expansion of knowledge.” Man, 73 years old, 20 years since T2DM diagnosis

The identification of the sender as the local public health care system led participants to associate the messages with their trusted general practitioners. However, one participant suggested including a phone number for addressing doubts and providing further clarification on specific topics.

“Of course, it helps a lot. Any information, especially when it comes from the health service, is beneficial because it helps improve the quality of life you have.” Woman, 74 years old, 6 years since T2DM diagnosis

“It would be good if at the time of receiving the message, if there is something you do not understand or need to explain better, to be able to call a phone number. I wish it were that way.” Woman, 54 years old, 7 years since T2DM diagnosis

Most participants expressed that they had improved their knowledge about diabetes, which participants perceived as helpful for improving their diabetes self-management. However, some declared that they already knew all the information received.

“Yes, there are new things. There are things that I learn now with you.” Woman, 74 years old, 6 years since T2DM diagnosis

“The advice is very good for people who do not know.” Man, 61 years old, 10 years since T2DM diagnosis

DiabeText raises awareness and supports type 2 diabetes self-management

Participants reported that receiving the messages made them more conscious of their diabetes and highlighted the importance of self-care in avoiding related complications. Some participants stressed that since T2DM often develops silently, these messages were perceived as tools that could increase their awareness of their health condition.

“I used to think, 'Well, I'm diabetic, I take some pills, and that is it.' Now I realize it is not just about the pills, but also about what you think, eat, and do. I have become more aware.” Woman, 60 years old, 8 years since T2DM diagnosis

For some participants, the messages acted as daily reminders to be mindful of their dietary choices and medication adherence. Additionally, the messages may have prompted them to monitor their blood test results more carefully and to be proactive in managing their condition if their control was poor.

“It has been beneficial for me to receive comparisons between analytics. Previously, I did not pay much attention to the results, but with the messages, I began to take notice.” Man, 65 years old, 14 years since T2DM diagnosis

“It has been difficult to take good care of myself, mainly because of my job. Therefore, when you sent me the messages, it was a reminder.” Woman, 51 years old, 4 months since T2DM diagnosis

Although a few participants acknowledged rereading the messages to review or refresh the content, the majority struggled to recall specific details. However, intervention as a whole succeeded in significantly increasing awareness of diabetes and the importance of self-care.

“I do not remember any specific messages, but the messages made me think more about diabetes than I had before.” Man, 65 years old, 14 years since T2DM diagnosis

Additionally, the messages encouraged participants not only to be more attentive and receptive to self-care, but also to adhere to recommendations. For example, prompts to refill prescriptions at the pharmacy assisted some individuals in remembering to do so before running out of medication. Nevertheless, some participants did not require assistance in pill-taking, as they were already diligent in their medication adherence.

“The messages you sent me truly supported me and helped me a lot with my illness. If I ever had any doubts, I would search for more information about it.” Man, 75 years old, 13 years since T2DM diagnosis

“I am quite careful about taking medication. I already had it under control.” Woman, 63 years old, 4 years since T2DM diagnosis

Participants felt motivated to increase physical activity upon receiving messages promoting walking and reducing sedentary behaviors, prompting adherence to recommendations either on the same day or shortly thereafter.

“It helps to plan your day-to-day when they advise you about exercising and all those things that sometimes you do not give importance to. At least they remind you, and I do not know, it is like putting yourself in order.” Man, 59 years old, 6 years since T2DM diagnosis

The participants greatly welcomed messages offering nutritional counseling, especially in areas where guidance was lacking, emphasizing the challenge of dietary adjustments in diabetes self-management.

“Food is where I needed the most advice. The messages helped me behave well […] they told me what I should do, and that made me more attentive to following the advice.” Man, 73 years old, 11 years since T2DM diagnosis

DiabeText helps people with T2DM maintain healthy habits

The participants exhibited varying degrees of behavioral change. Some already adhered to certain healthy habits and continued to do so throughout the intervention, whereas others acknowledged inappropriate behaviors and made improvements in response to the messages they received.

Among participants who were already adhering to healthy habits, such as exercising and taking their medication correctly, the messages appeared to reinforce their existing routines and were perceived as helpful for maintaining these habits. However, for others, the messages served as a reminder of the importance of adopting self-care behaviors.

“The habits were already established. Therefore, they have been reaffirmed […] in terms of being more conscious about maintaining medication and exercising.” Man, 73 years old, 20 years since T2DM diagnosis

“Many times it motivated me […] For instance, if I had been eating a lot of fruit, and [a message] recommended what I should do with the fruit, I followed through with it diligently.” Man, 73 years old, 11 years since T2DM diagnosis

Participants who reported improvements attributed to the intervention described adopting healthier dietary choices and engaging in more frequent physical activities. Additionally, some participants started taking medication as prescribed because of the reminders received from DiabeText.

“Yes, I have made some changes. I avoid sweets, pasta, and things like that more than before. Now I pay more attention to the products; if I buy something, I look at the amount of sugar it contains.” Woman, 60 years old, 8 years since T2DM diagnosis

“In my case, the messages have encouraged me to walk and do gymnastics, which I have put into practice, and continue to do so.” Man, 67 years old, 19 years since T2DM diagnosis

Individuals with T2DM suggest new personalized features to improve DiabeText uptake and engagement

The participants had varied preferences regarding message content, frequency, and timing. Most patients requested more personalized dietary advice for diabetes.

“I suggest that messages should provide more diet plans and menus because honestly, I struggle to find ways to incorporate vegetables into my meals when I don't feel like having a salad.” Man, 62 years old, 4 years since T2DM diagnosis

Medication details were less valued due to confidence in adherence. The frequency and duration were generally acceptable, although one participant found it too lengthy, whereas others wanted more weekend messages.

The timing of message delivery, particularly in the morning, was well received by users, who appreciated the flexibility to read messages immediately or at a later, more convenient time. The majority of participants agreed with the possibility of including monitoring for new clinical variables such as blood cholesterol and blood pressure.

With respect to the web links included in some messages, the participants admitted that they rarely clicked on them to view the recommended content, despite having internet access and using their mobile phones regularly.

“Regarding web links included in some messages, I didn’t check them.” Man, 59 years old, 6 years since T2DM diagnosis

Integration of quantitative and qualitative findings

The joint interpretation of quantitative and qualitative findings was conducted through an explicit integrative analysis to identify convergence, divergence, and complementarity between data sources, providing a more comprehensive understanding of how and why DiabeText influenced self-management behaviors (Table 3). Quantitatively, participants rated the four most frequently used BCTs—instruction on how to perform a behavior, information about health consequences, habit formation, and verbal persuasion to increase self-efficacy—as highly helpful for supporting medication adherence and lifestyle changes, with approximately 75% rating them as “very” or “extremely” helpful. The qualitative findings reinforce and contextualize these results: participants described feeling practically and emotionally supported, perceiving the messages as trustworthy, and experiencing increased awareness and motivation toward healthier habits. These accounts illustrate potential mechanisms through which the highly scored BCTs may have operated in practice. Overall, the findings showed a high degree of convergence between both data sources, with qualitative accounts consistently supporting and contextualizing the quantitative ratings, while also revealing complementary insights regarding underlying mechanisms and user experience.

Table 3.

Joint display integrating quantitative and qualitative findings of the DiabeText process evaluation.

Quantitative findings (BCT ratings)	Qualitative themes & illustrative mechanisms	Integrated interpretation (meta-inferences)
Instruction on how to perform a behavior rated as very/extremely helpful by 74% of participants.	Participants described that messages provided clear, practical guidance (e.g., diet tips, reminders about physical activity) that helped them “plan their day-to-day” and facilitated actionable self-care steps.	High acceptability of this BCT aligns with participants’ use of messages to structure daily diabetes management, showing how practical guidance translates into increased capability and opportunities for action.
Information about health consequences rated as very/extremely helpful by 76%.	Participants reported that messages increased awareness of diabetes, risks, and the importance of self-care, noting they “paid more attention” to test results and consequences of behaviors.	Awareness-raising messages appear to strengthen reflective motivation, explaining why participants perceived increased vigilance and responsibility for T2DM despite limited quantitative changes in clinical outcomes.
Habit formation rated as very/extremely helpful by 72%.	Messages acted as prompts and reminders to take medication, walk, or make healthier dietary choices. Some said messages “reaffirmed” existing routines, while others credited them with initiating behavior change.	Habit reinforcement through repetition may be particularly influential in sustaining behavioral routines, supporting the idea of DiabeText as a low-intensity yet consistent behavior-shaping tool.
Verbal persuasion to increase self-efficacy rated as very/extremely helpful by 77%.	Participants frequently mentioned feeling supported, cared for, and motivated. The emotional tone of messages helped them feel “someone is interested in your health”.	Emotional support mechanisms appear to amplify self-efficacy, complementing BCT ratings and helping explain why participants felt more capable and motivated despite not recalling specific message content.
Across all BCTs, only ∼11% reported low perceived usefulness (Likert 1–2).	A minority stated they “already knew” most diabetes information or felt confident in medication adherence before DiabeText.	The intervention may be more impactful for participants with lower baseline knowledge or less structured routines, highlighting the need for personalization in future iterations.

For example, participants’ repeated references to feeling “cared for,” “motivated,” and “more conscious” of their condition align with the high acceptability of verbal persuasion and health consequences BCTs. Similarly, descriptions of acting upon reminders or adjusting dietary and physical activity routines complement quantitative ratings of habit formation and instruction on how to perform a behavior. Together, the integrated findings suggest that DiabeText was particularly effective in enhancing reflective and automatic motivation, as well as practical capability, even when quantitative clinical outcomes such as glycemic control did not change. Some divergence was also observed, particularly among participants who reported limited added value of the intervention due to prior knowledge or established self-management routines, highlighting heterogeneity in perceived usefulness.

Discussion

In this process evaluation of the DiabeText intervention, participants reported predominantly positive experiences, largely due to feeling supported and receiving reliable, useful content that was described as facilitating diabetes self-management. They also highlighted the intervention’s practical and user-friendly format, noting perceived increases in motivation to adopt healthier behaviors and potentially improved medication adherence. Importantly, the messages appeared to raise awareness of diabetes-related health consequences and encourage behavioral change.

Consistent with the COM-B framework, ³² our findings suggest that DiabeText addressed all three essential drivers of behavior change. Specifically, the intervention enhanced Capability by providing trusted knowledge and practical self-care instructions; it facilitated Opportunity through a user-friendly, non-intrusive mobile delivery format; and it bolstered both reflective and automatic Motivation by fostering a sense of being ‘cared for’ and providing timely behavioral prompts (Figure 1).

Figure 1.

Mechanisms through which DiabeText may support diabetes self-management, mapped within the COM-B framework.

By mapping these themes onto the COM-B model, we identified a reinforcing cycle where increased support leads to higher self-efficacy and more consistent self-management routines. This theory-driven interpretation allows for a more robust understanding of why the intervention was perceived as effective, even among patients with lower clinical complexity and those less engaged with diabetes management. However, previous research indicates that the strength of associations between theory use and intervention effectiveness is generally limited. ³⁶

Identifying the specific BCTs that drive positive outcomes remains essential for the development and replication of mHealth tools. ²⁰ While “instruction on how to perform behavior” and “information about health consequences”–both central to DiabeText–have been linked to improved adherence in earlier studies,^37,38 other techniques such as “self-monitoring”, “feedback”, “credible sources” and “goal setting” are often more strongly associated with successful behavior change.^39,40 In medication-taking interventions, “prompts and cues” and “credible source” have shown particular effectiveness. ³⁷ Evidence is mixed regarding how these BCTs influence different behavioral domains, ⁴⁰ and variations in target populations and theoretical approaches ⁴¹ make it challenging to define an optimal combination of techniques.

Although DiabeText improved self-reported medication adherence and diabetes self-management, its effect on glycemic control and lifestyle outcomes was limited. ¹³ This may reflect the interventions’s one-way, automated nature, which restricts the use of key BCTs such as self-monitoring and goal setting. Moreover, the most frequently applied techniques—habit formation and verbal persuasion—are not consistently associated with positive outcomes across studies.^37–40 Refining the BCT strategy may therefore enhance the intervention’s overall impact. ⁴²

The participants generally found the messages helpful, informative, and supportive, as seen in studies by Leon, ⁴³ Bartlett, ⁴⁴ and Moyano, ⁴⁵ where the messages offered comfort and a sense of being cared for. However, some, as reported by Lauffenburger et al., ⁴⁶ felt overwhelmed despite appreciating the practical advice and several participants struggled to recall specific messages, felt they already knew the information, or did not need support with medication adherence. Preferences for content, timing, and frequency varied—one participant found the program too long—and web links were seldom accessed despite regular mobile internet use. Some participants requested more tailored dietary advice and an optional channel for clarifications. These perspectives highlight the need for greater personalization and flexible, user-driven features ⁴² to accommodate diverse engagement patterns. ⁴⁷

More recently, Newhouse et al. conducted a qualitative process evaluation of a text-messaging intervention similar to DiabeText and identified comparable contextual factors and mechanisms of action. ⁴⁸ They found that participants’ engagement was shaped by existing daily routines, whereas disruptions such as heavy workloads or irregular schedules hindered adherence to medication plans. Importantly, sociodemographic characteristics–including time since diagnosis, gender and age–did not appear to influence perceptions of the intervention’s usefulness. Consistent with these findings, our results highlight similar emotional and cognitive mechanisms through which such interventions may support behavior change, including feeling cared for, learning from trusted sources, increased awareness of T2DM, and behavioral reinforcement through repetition.

Broader mHealth literature also shows a consistent preference for accessible, reliable, and user-friendly digital tools,^49,50 although some studies note that these interventions may increase patients’ sense of responsibility. ⁵¹ While certain users find them motivating for goal-setting, others perceive limited added value when their diabetes is already well controlled ⁵² —patterns that closely mirror the range of responses observed in our study. As suggested by Wang et al., prioritizing individuals who are more recently diagnosed and who present lower baseline levels of diabetes knowledge, lower self-efficacy, or higher diabetes distress may enhance the overall effectiveness of the intervention. ⁵³

The contextual setting is also relevant. DiabeText was implemented within the Spanish public primary care system, characterized by high trust in public health services, ⁵⁴ strong continuity of care, ⁵⁵ and widespread mobile phone use, ⁵⁶ all of which likely facilitated engagement. Participants’ emphasis on feeling cared for may also reflect cultural expectations of relational support, which could differ in other healthcare contexts and should be considered when assessing transferability. In terms of future design and implementation, a recent review has summarized the key elements and theoretical foundations needed to develop effective text-message interventions to enhance medication adherence among patients with diabetes. ⁴² The authors highlight that DiabeText already incorporates a comprehensive structure–combining evidence-based content with theory-driven behavior-change components. Building on these foundations, our post-hoc process evaluation indicates several actionable considerations for future iterations of the intervention. First, refining message content to more directly strengthen capability, opportunity and motivation—aligned with the COM-B mechanisms identified—may increase its behavioral impact. Second, greater personalization, including dynamic tailoring informed by users’ ongoing feedback and engagement patterns, could enhance acceptability and retention. Third, from an implementation standpoint, integrating DiabeText seamlessly within existing digital infrastructures, ensuring interoperability with electronic health records, and enabling real-time data capture will be essential to support scalability. Together, these insights offer concrete implications for both the future design of digital adherence interventions and the practical strategies needed to implement them effectively within the routine care pathways. ⁵⁷

Strengths and limitations

The present process evaluation, which explores how a tailored text message intervention for diabetes self-management produces its outcomes within the context of an effectiveness trial, is the first of its kind in Spain. Moreover, the analysis is grounded in participants’ assessments of BCTs and supported by qualitative methods—an approach widely recognized for adding significant value to process evaluations of effectiveness trials.^58–61 In fact, qualitative feedback indicated behavior changes among participants in response to the messages, especially regarding medication and lifestyle habits, even though the quantitative lifestyle outcomes measured after trial completion did not differ between the intervention group and the control group.

The methodology adopted in the DiabeText project embraces the dynamic and interactive nature of context, as revealed through participants’ lived experiences. This approach moves beyond the limitations of relying solely on causal inference to understand how complex interventions “work,” thus addressing the challenges of the so-called “complexity turn”.^62–64

The study has several limitations. First, patients were not involved in designing the questionnaire or interview topic guide, nor were they involved in the analysis or interpretation of the qualitative findings. Lack of patient involvement in design and analysis risks pro-innovation bias and limited scope, potentially undermining the findings’ interpretive validity. In addition, selection bias likely favoured motivated or health-conscious participants, potentially overestimating impact and limiting generalizability. The qualitative subsample exhibited lower clinical complexity (fewer medications), with a trend toward lower obesity rates and higher levels of depression compared with the overall intervention arm. Although participants with diverse characteristics were included, this imbalance may have influenced the results and should be acknowledged. Nevertheless, the high satisfaction reported among these “lower-need” patients suggests that the intervention’s value is robust, potentially offsetting this bias and indicating that the findings may even be conservative. Furthermore, the transferability of our findings is constrained by the specific geographical and institutional context of the study. We acknowledge that the implementation within the Spanish public primary care system and the specific cultural context of the Balearic Islands may limit the transferability of the results to other settings, languages, or cultures. The high levels of trust in the public health service and the regional cultural expectations of relational support—which likely influenced the reported mechanism of ‘feeling cared for'—may differ significantly in other healthcare systems or cultural backgrounds, potentially affecting the intervention’s reception and impact. Also, interviews were conducted shortly after the COVID-19 pandemic, when recent shifts in healthcare delivery–such as increased telehealth use and fewer in-person visits^65–68— may have influenced participants’ views and increased openness to digital interventions. Recall bias was also possible; however, questionnaires were completed within one month after follow-up and, interviews within six months, helping reduce the risk of substantial memory decay. Additionally, participant characteristics were measured at baseline13 one year before the process evaluation, which may not fully reflect their status during the post-hoc assessment. Also, several members of the research team were involved in the design and development of the intervention, which may have influenced data interpretation despite efforts to mitigate this through reflexive practices and analytic triangulation. Moreover, although an exploratory analysis indicated high internal consistency among questionnaire items, this should be interpreted with caution. The instrument was self-reported, developed ad hoc, and not formally validated, and the items capture conceptually distinct BCTs rather than a single underlying construct. Therefore, the results provide only a partial and potentially biased assessment of the perceived mechanisms of action of the intervention. Future research should incorporate validated instruments or more comprehensive measures to assess the contribution of specific intervention components. Finally, the post-hoc evaluation design carries important methodological constraints that should be considered when interpreting the findings. As data were collected after completion of the intervention, the mechanisms identified in this study reflect participants’ retrospective accounts and perceived experiences rather than processes observed in real time. Consequently, these findings should be interpreted as hypothesis-generating rather than confirmatory. This retrospective approach limits our ability to establish temporal relationships between exposure to specific intervention components, the activation of putative mechanisms, and subsequent behavioral outcomes. Therefore, causal pathways cannot be robustly inferred, and the ability to attribute observed effects to specific behavior change techniques or intervention features is limited. In addition, participants’ accounts may be influenced by recall bias and by the retrospective reconstruction of their experiences, potentially leading to coherent narratives that align with perceived outcomes rather than accurately reflecting dynamic processes during the intervention. The absence of prospectively collected longitudinal process data also prevents examination of how participants’ perceptions, engagement, and behavioral responses evolved over time. Despite these limitations, this post-hoc evaluation provides valuable insights into perceived mechanisms of action and contextual factors that can inform the refinement of the intervention and guide the design of future studies incorporating prospectively embedded process evaluations.

Conclusion

This study suggests that DiabeText, a theory-grounded mHealth intervention could significantly enhance patients’ perceived capability and motivation to manage T2DM by fostering a continuous sense of being ‘cared for’ within the primary care framework. Our findings suggest that while low-intensity text messaging is highly acceptable and effective for reinforcing routines, transitioning from general support to meaningful clinical impact requires a shift in mHealth design. Future iterations should move beyond one-way automated messaging toward dynamic, personalized systems that incorporate ‘active’ behavior change techniques, such as self-monitoring and goal-setting, tailored to the patient’s baseline knowledge and clinical complexity.

Appendix.

List of abbreviations

T2DM

type 2 diabetes mellitus

BCT

behavior change technique

RCT

randomized controlled trial

mHealth

mobile health

COM-B

Capability, Opportunity and Motivation for Behavior Change

MRC

Medical Research Council

Multimedia Appendix

• - Supplementary material 1. Questionnaire to evaluate the main behavior change techniques used in the DiabeText study among people with type 2 diabetes and the behavioral change techniques used in the DiabeText study

• - Supplementary material 2. Interview topic guide

• - Consolidated criteria for reporting qualitative research (COREQ): 32-item checklist

ORCID iDs

Rocío Zamanillo-Campos https://orcid.org/0000-0001-7162-0889

Maria Antonia Fiol-deRoque https://orcid.org/0000-0001-8566-0929

Ethical consideration

This study was approved by the Ethical Research Committee of the Balearic Islands (https://www.caib.es/sites/comiteetic/es/portada-44578/) under the identifier IB 4320/20 PI.

Consent to participate

All participants provided audio-recorded informed consent prior to enrollment. The consent procedure ensured that participants received comprehensive information regarding the study’s objectives, procedures, potential risks, and anticipated benefits. They were also informed of their right to withdraw from the study at any time without penalty or consequence. Confidentiality and anonymity were strictly safeguarded throughout the research process. All personal and health-related data were collected, recorded and stored in accordance with applicable national and international data protection regulations.

Consent for publication

The copyright holders of DiabeText, which is protected through Blockchain-based intellectual property registration (Docusign Envelope ID: 00875219-78AF-46F0-8926-9070C975931B), are members of the project to which the present study belongs.

Data Availability Statement

The data that support the findings of this study, such as the coding tree, are available from the corresponding authors upon reasonable request.*