It is a novel mHealth application designed to address medication non-adherence using established behavioral methods. The application can be recommended to patients by healthcare professionals to encourage and monitor adherence. My A:Care associates adherence with self-care and helps patients take small, manageable steps while rewarding positive actions to facilitate lasting behavioral changes. The application provides motivational messages/challenges, health insights, and pill reminders to encourage adherence (24). Further details on the application are described in Supplementary Methods S1.1 and Supplementary Table 1.

Smart Coach is an advanced version of the My A:Care application running on the same core engine with the same app interface. It differs substantially from the latter by delivering SPUR (Social, Psychological, Usage, Rational) framework-enabled patient profiling in combination with clinical and sociodemographic information using the d.Tells^TM algorithm to deliver tailored treatment plans (25, 26). Further details on the application and SPUR questionnaire are described in Supplementary Methods S1.2.

Medication adherence was measured using the Medication Adherence Report with Visual Analog Scale (MARS-5VA) questionnaire. This questionnaire comprised 2 parts: a 5-item Medication Adherence Report Scale (MARS-5) as part 1 and two Visual Analogue Scales (VAS) as part 2. While MARS-5VA part 1 assesses non-adherence to any medication, MARS-5VA part 2 specifically evaluates adherence to lipid-lowering therapy. MARS-5 is a clinically validated and reliable five-item self-report questionnaire that captures both intentional and unintentional non-adherence with a minimal social desirability bias (27). The score in the MARS-5VA part 1 ranges from 5 to 25 for lower to higher adherence, respectively. Responses to the scale are non-dichotomous and provide a nuanced differentiation of the adherence behavior to allow categorizing respondents into multiple groups as per their position along the non-adherence continuum (28). Alternately, VAS is a single-item, simple-to-administer adherence scale utilizing the patient's recall to estimate the number of doses taken in a specified period. It has a 10-point response scale with values ranging from 0% (none used) to 100% (all used) (29).

Patients' general perception towards medications was evaluated using a combination of the twelve-item Beliefs about Medicines Questionnaire-General 12 (BMQ-General-12) (30) and a five-item Perceived Sensitivity to Medicines Scale-5 (PSM-5) questionnaire (31). The BMQ-General-12 score ranges from 4 to 20 for each subscale (overuse, harm, and benefit), with a high score indicating a more positive perception of medicines. Alternatively, PSM-5 score ranges from 5 to 25, with high scores correlating with higher perceived sensitivity to potential adverse effects of medicines.

To appraise patients' perceptions specific to lipid-lowering therapy, the Beliefs about Medicines Questionnaire—Specific (BMQ-S11-Plural) was employed (30). The BMQ-S11 Plural questionnaire has two subsets—BMQ-Specific-Necessity and BMQ-Specific-Concern. The Medicines Adherence and Information Needs (MAIN) Screen was used to determine the barriers and support needed for the participant's medicine adherence. Various barriers and facilitators of medication adherence were examined, for example, difficulty in using the medicine, support required for using the medicine, information, and advice available about the medicine, ease in availability of medicine, and necessity-concern beliefs.

Participants aged 18–75 years with newly diagnosed dyslipidemia, initiated on stable lipid-lowering therapy between 1 and 9 months before the study start date, were included. A total of 150 patients from this primary prevention cohort were enrolled. Screening was conducted at Visit 1 using the MARS-5VA (Part 1), with a score <22 indicating suboptimal adherence and eligibility for inclusion. Key inclusion criteria included ownership of an Android or iOS smartphone and the ability and willingness to use it. Major exclusion criteria were a recent history (within 2 months) of myocardial infarction, stroke, or unstable angina; recent hospitalization for cardiovascular conditions; need for changes in lipid-lowering therapy; adverse effects requiring therapy modification or discontinuation; use of injectable lipid-lowering agents; or concurrent use of other adherence-related mHealth applications.

The study was approved by the Human Research Ethics Board of the Faculty of Medicine, Khon Kaen University (Reference no. HE651047, 1 April 2022). Written informed consent was obtained from all participants prior to their enrollment in the study.

Trial participants were randomized 1:1:1 (50 each) into three groups, My A:Care and Smart Coach (intervention) and No App (control) using variable block sizes of 3, 6, or 9. The final block size was adjusted to ensure balanced allocation at the target sample size (n = 150). To obscure the sequence, the final assignment was moved to the first position, creating unbalanced blocks at both ends. An independent consultant generated the sequence and prepared sealed, opaque, sequentially numbered envelopes. Investigators assigned interventions by opening envelopes in order. Tamper-Resistant Envelope Allocation Technique (TREAT) and a random number chain check were employed to prevent tampering and ensure sequence integrity (32).

As this was an exploratory study, the sample size was not based on a formal sample size calculation. Participants' baseline characteristics, medical history, details of the lipid-lowering agent, comorbidities, and other concomitant treatments being administered were documented at visit 1. Patients allocated to the intervention groups were assisted in downloading the respective mobile application and received training on application usage. Patients assigned to the Smart Coach group were administered the in-app SPUR questionnaire for patient profiling (Supplementary Methods S1.2).

The study duration was 3 months, with participants followed up at 12 weeks (±7 days) during Visit 2. Lipid panel data were collected at both baseline (Visit 1) and follow-up. Concomitant medications and any adverse drug reactions related to Abbott products were recorded. Mobile application usage data, including individual user summaries, were retrieved and assessed in pseudonymized form to evaluate interaction with the two applications.

Patient's general perception of medications and their concerns about side effects were measured using BMQ-General-12 and PSM-5, respectively. At the beginning of the study, participants had a neutral to slightly positive perception of medication usage in general and a slightly negative perception of potential adverse effects. Patient's perception of medication did not change significantly during the study duration, either within the study groups or in between the control and App intervention arms (Figure 4 and Supplementary Table S4).

Patient's specific perceptions of lipid-lowering agents were measured using the BMQ-S11 Plural questionnaire. At the baseline, patients had a positive perception of the need for the medicines and were relatively neutral toward potential negative effects. During the study, the BMQ-Specific-Necessity scores increased in the My A:Care-All group as compared to the control, but the change was not statistically significant (Mean [SD] 0.6 [3.7] vs. −0.2 [2.7]; p-value = 0.759). The BMQ-Specific-concern uniformly decreased across all study groups with no statistical difference among the groups (Figure 5 and Supplementary Table S5).

Furthermore, the MAIN Screen was used to determine the barriers and support needed for the participants' medicine adherence. At the end of the study, a higher proportion of participants reported no difficulty in using their medicines, higher support in using medicines, and higher availability of information and education across all the study groups. However, no significant difference was found between the groups in week 12 scores of MAIN (Supplementary Table S5).

Participants engaged with motivational messages on at least 50% of the observational days. The proportion of green-yellow days in the cube was similar among the two groups (p-value = 0.866). Cube status can range from green, yellow, or red depending on the level of engagement and is measured using an AUC score on a scale of 1–3, with a higher score ascribed to a green cube. Both My A:Care- Complete and Smart Coach groups had similar AUC scores [mean (SD): 1.518 (1.282) vs. 1.501 (1.071); p = 0.891] (Figure 6).

Overall, engagement with pill reminders was found to be better as compared to motivational messages. Participants had high battery status on 64.3%–70.6% of observational days. The battery status was similar in My A:Care-Complete and Smart Coach groups [mean (SD): 0.706 (0.380) and 0.643 (0.413), respectively] (Figure 6).

The correlation between changes from baseline to week 12 in MARS-5VA scores and engagement with the mobile application was probed to understand the influence of app usage on improving adherence. The My A:Care-Complete group displayed a positive correlation between App usage and the change in MARS 5VA part 1 score (correlation coefficient ranging from 0.19 to 0.37). On the other hand, the usage of Smart Coach showed a weak negative correlation with the MARS 5VA part 1 score modifications (ranging from 0.10 to 0.19). It was perhaps due to outlier values in 7 participants, which negated the values of the whole group. Nevertheless, both these App usages did not yield any statistically significant correlations with changes in MARS-5VA part 1 scores at week 12 (Table 2).

The correlation between App usage and the change in MARS 5VA part 2 score in the My A:Care-Complete and Smart Coach groups was positive, ranging between 0.04 to 0.25 and 0.05 to 0.29. The correlation between high battery status and MARS 5VA part 2 scores in the Smart Coach group was statistically significant (Pearson's r: 0.29, p = 0.042) (Table 2).