Impact of behavioral economics to improve antihypertensive therapy adherence, a pilot randomized controlled trial in Los Angeles

Abstract

Routinized pill-taking can enhance medication adherence but is difficult to achieve. In this pilot randomized controlled trial we assess the feasibility, acceptability, and preliminary efficacy of a novel behavioral economics-based approach to medication-adherence. We enrolled 60 hypertensive adults, who all received information on pill-routinization and selected an existing behavioral routine (‘anchor’) to assist with routinization of pill-taking. Participants were randomized into 3 groups: 1)‘Control’ receiving usual care (n = 20); 2)‘Messages’ receiving daily text messages (n = 20); and 3)‘Incentives’ receiving both text messages and rewards for medication adherence (n = 20). Interventions lasted 3 months, followed by a 6-month post-intervention period during which we assessed medication-adherence and conducted standardized assessments of acceptability. The study demonstrated high feasibility and acceptability, with 90% of participants willing to refer others to the study. Mean adherence during the extended follow-up period (months 7–9) was numerically higher in both intervention arms than the control arm (Control 75%, Messages 84%, and Incentives 77%) though this pilot study was not powered to detect a statistically significant difference (P = 0.73). While this pilot study was under-powered to detect between-group differences, the novel behavioral economics habit formation approach was feasible, acceptable, and yielded promising results, warranting completion of a fully-powered trial.

Registration: URL: https://www.clinicaltrials.gov; Unique identifier: NCT04029883; registered 23/07/2019.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-94805-5.

Introduction

Nearly one-third of the adult global population suffers from hypertension, the most common modifiable risk factor for cardiovascular disease including myocardial infarction, heart failure, stroke and renal failure, contributing to excess morbidity and mortality worldwide^1,2. Despite widely and readily available medications to lower blood pressure (BP) and reduce the risk of adverse health events¹, data indicate that patients possess prescribed antihypertensive medications on only 50% of days during the first year of antihypertensive therapy, and only 20% take their medications frequently enough to obtain cardiovascular benefit^3,4. Numerous studies have examined interventions, including text messages and financial incentives^5–7, aimed at improving medication adherence, but meta-analyses have shown inconsistent results of these existing interventions. For chronic conditions, it will be crucial to increase adherence not only while an intervention is provided, but in the long run, for which habits offer unique potential. However, participants in interventions using the predominant approach in the literature called ‘anchoring’ often experience high attrition due to the lengthy habit formation period during which many people drop put. The role of interventions designed to counter such high attrition and habituate medication taking remains under-explored.

Habits, i.e. the integration of pill taking into patients’ daily routine, are a commonly reported strategy for sustaining high medication adherence among those successfully managing chronic conditions^8,9. According to the Habit Formation model (Fig. 1A)¹⁰, habits shift the cognitive pathways to the subconscious system, thus allowing for the behavior to persist without the need for high motivation^11–14. Unfortunately, most patients have difficulty forming new habits on their own, as doing so requires sustained daily repetition of the new behavior in response to the same contextual cue for approximately 3 months^15–18. The most common strategy to initiate habits involves tying the targeted behavior to an existing routine (‘anchoring’) such as brushing one’s teeth or eating breakfast that acts as the contextual cue (Fig. 1B). Anchoring has been used to improve several health behaviors such as physical activity¹⁹, smoking cessation²⁰, and improved dietary patterns²¹, however the existing interventions rely on participants’ high intrinsic motivation for successfully establishing the habit^22–25. Consequently, typically fewer than half of the participants benefit from these interventions²⁶.

Fig. 1.

Habit formation model and the proposed intervention mechanism.

The field of behavioral economics (BE) merges economics and psychology, providing a coherent framework to examine decision-making, to predict how biases may affect those decisions, and develop interventions to drive one decision over another. BE suggests that failure to successfully form habits, and therefore improve medication adherence, is related to lack of disease salience and present bias. Lack of disease salience results in reduced adherence particularly for asymptomatic conditions such as hypertension, as benefits of pill taking are not readily evident during the habit formation period. Relatedly, present bias (the tendency to give in to current temptations at the cost of potential future rewards) contributes to reduced adherence as other, more pleasurable activities compete with the onerous task of taking medications (that can reap benefits over the longer term)^27,28. A novel approach synthesizing the anchoring approach with BE insights proposes that sending reminders together with incentives for pill taking in accordance with a person’s anchoring plan, can effectively support habit formation for all participants, including those with low motivation²⁹. In this study, we sought to evaluate the feasibility, acceptability, and preliminary efficacy of a BE-based intervention (called Behavioral Economics to improve Antihypertensive Therapy Adherence [BETA]) that uses text messages and small incentives to improve antihypertensive medication adherence through habit formation.

Methods

This three-armed pilot randomized controlled trial (RCT) was composed of an active intervention period for the first three months, and a post-intervention follow-up for six months.

Study population

Participants were enrolled from the Cedars-Sinai Hypertension Center for Excellence in Los Angeles. Eligible patients were adults (age ≥ 18 years) with a diagnosis of hypertension, currently or newly prescribed at least one antihypertensive (AHT) medication (both incident and prevalent users). Participants had to own or have access to a phone throughout the duration of the intervention and be willing to receive study text messages. Individuals who were unable to provide informed consent, were unwilling to use a study provided pill bottle, and those enrolled in another hypertension-specific clinical trial were excluded. A sample size of N = 60 (n = 20 per arm) was prespecified prior to study initiation and identified through the electronic health record (Supplemental Methods).

Randomization

The online Clinical Trial Randomization Tool from the National Cancer Institute (https://ctrandomization.cancer.gov/) was then used to randomize participants in a 1:1:1 ratio (using asymptotic maximal randomization) to the three study arms. Results from the randomization tool were placed in sealed, opaque envelopes by a study team member not involved in recruitment and stored separately in a secured cabinet, separate from other study materials. Multiple staff members enrolled patients and all were unaware of total arm allocation numbers when enrolling patients. Envelops were opened by the study coordinator in front of the participant at the time of enrollment, revealing the arm assignment to both the participant and the study coordinator, who therefore were not aware of the respondent’s treatment assignment during the preceding survey. The nature of the intervention did not allow for blinding of the participant or the coordinator, however, the data analyst who conducted the analysis was blinded to the treatment assignment.

Study design

Prior to randomization, all participants completed a baseline survey and were provided a one-page educational leaflet on routinization and habit formation as it relates to medication adherence. Participants were then asked to select an existing routine behavior (‘anchor’) to cue their medication-taking, and to state the time at which their anchor typically occurs. If participants were taking more than one AHT medication or multiday dosing regimens, they were asked to select a single medication and single dose to be monitored for the study, as has been shown to be appropriate in other studies³⁰. All participants were provided a MEMS (medication event monitoring system) Cap, a pill bottle and cap which electronically captures the date and time of each bottle opening. They were instructed to place their identified medication in this bottle and informed that the bottle would be used to assess adherence.

Participants were randomized (Supplemental Methods) into 1 of 3 arms³¹: standard of care (Control), text messages (‘Messages’), or text messages with conditional incentives (‘Incentives’). In the Control arm, participants received no additional interventions beyond standard of care under the supervision of their treating provider. In addition to usual care, participants in the Messages arm received daily text messages during the 3-month intervention period at fixed times during the day that did not correspond with the patients’ anchor time. Finally, in addition to standard of care and the aforementioned text messages, participants randomized to the Incentives arm could qualify for small monetary incentives conditional on medication adherence. Specifically, if participants took their designated AHT medication within 1-hour of their chosen anchor time (cued adherence) on at least 80% of the days in the month preceding each of the 3 monthly visits during the intervention period, they were allowed to participate in a prize drawing. Prizes were of values $0, $25, or $50 (Fig. 1C).

All participants completed in-person study visits at enrollment (month 0), monthly during the intervention period (months 1, 2, and 3) and at study completion (month 9) (Supplemental Fig. 1). At each visit, BP was assessed per protocol (Supplemental Methods)³², and adherence data were downloaded from the MEMS Cap devices. Participants also completed surveys on their anchoring and adherence behaviors since their last visit; participants who reported degradation of their selected behavioral anchor were asked to identify a new existing routine for the anchoring strategy which was used for future cued adherence assessments. Conditional prize drawings were completed at months 1, 2, and 3 for participants in the Incentives arm if cued adherence was ≥ 80% in the preceding month. Participants were provided with parking vouchers and an honorarium ($50) if they attended all visits.

Once the pilot RCT was completed, it was followed by an adaptation phase to improve the intervention in preparation for a fully powered intervention to test efficacy. During this phase, a sub-sample of participants who completed the RCT (n = 11) and providers who participated and supported in the recruitment of patients for the RCT (n = 5) were interviewed to document their experiences with the study and identify potential improvements to the study design. Interviews were conducted virtually by two qualitatively trained study staff, lasted 30–45 min and were recorded.

Surveys

In addition to the short surveys at each monthly intervention visit, more in-depth surveys were completed at baseline (month 0), end of intervention (month 3), and end of study (month 9). The baseline survey asked about participant demographics, structural barriers^33–35, existing habits associated with regular medication-taking^36,37, attitudes and beliefs pertaining to medication-taking behaviors³⁸, and adherence and regimen-related information. The month 3 survey included questions pertaining to acceptability of the MEMS Caps and intervention components. The month 9 survey included questions regarding overall study acceptability. Surveys were collected using REDCap administered on a study provided tablet device.

Full summaries of the study timeline as well as the survey measures are available in Supplemental Fig. 1.

Outcomes

Primary outcome: extended post intervention daily adherence (preliminary efficacy)

Our primary efficacy outcome was between-group difference in medication adherence during the end of the follow up period (months 7–9) to capture behavioral persistence at the latest possible period. Adherence was defined as the number of actual once-a-day bottle openings during the intervention period out the number of days assessed. Only 1 bottle opening per day was counted, capping adherence at 100%. We additionally assessed for between-group differences in medication adherence during the intervention period (months 0–3), immediate post-intervention period (months 4–6), and entire post-intervention follow-up period (months 4–9). We secondarily evaluated between-group differences in hypertension control at the end of the study, defined by BP of < 130/80 mmHg³².

In sensitivity analyses, we repeated these analyses using cued adherence, defined as the fraction of bottle openings within 1-hour of the participant’s stated anchoring time. Assessed time intervals were the same as in the primary outcome. Finally, we repeated above analyses examining between-group differences between the Control arm and a combined intervention arm consisting of patients in both the Messages and Incentives arms.

Feasibility and acceptability

We assessed feasibility and acceptability as additional outcomes for the trial. Feasibility was assessed by screening success (proportion of participants deemed eligible out of total number screened) and recruitment rates (proportion of enrolled patients out of all patients approached for enrollment). Acceptability was assessed using three types of data: (1) participant retention through study completion, (2) survey responses regarding comfort with different components of the intervention, overall satisfaction with the study, and willingness to recommend BETA to a friend or family member³⁹, and (3) qualitative assessment of responses during interviews conducted during the adaptation phase.

All outcomes were prespecified, as defined in our published protocol⁴⁰. Outcome definitions and time periods are further delineated in Supplemental Table 1.

Statistical analysis

We used an intent-to-treat approach to evaluate group-level differences in the primary and secondary outcomes. Adherence and cued adherence are presented as means with 95% confidence intervals. Given our sample size, we assessed our primary outcome using the Student’s t-test, comparing differences between each treatment arm with the Control arm, and two-sample test of proportions for our secondary outcome. We summarized response statistics for each item assessed for feasibility and acceptability. The significance level for all analyses was set at 0.05. Missing data were imputed using the last survey wave’s data point if a participant remained enrolled in the study but did not complete a particular survey.

Qualitative analysis

We undertook a rapid analysis of the interview data to identify within a short period of time dimensions relevant to the adaptation phase for the planned study at scale^41,42. Two qualitative researchers [IG, AP] iterated on the rapid coding approach, which defined the categories of interest for which we extracted summaries from each interview.

The analysis process began with summary field notes (including selected supporting quotes as available) prepared during the interview and refined by listening to the interview audio recordings. These were then transferred to an Excel extraction sheet⁴³where the sub-summaries were coded across the dimensions of interest, pre-identified from the interview protocols: patient profiles (e.g., changes to AHT regimen); barriers to habit formation (e.g., AHT medication side-effects); facilitators of habit formation (e.g., motivation for habit building); and feedback on the intervention components. The qualitative researchers worked together, cross-checking each other’s summaries, iterating, and refining them. Summaries were consolidated by participant type (patient, provider) to facilitate comparisons across groups⁴⁴.

The Cedars-Sinai Institutional Review Board approved this study and all experiments were performed in accordance with relevant guidelines and regulations, including obtaining informed consent from all participants.

Results

Feasibility

A total of n = 448 patients were screened for eligibility, of whom n = 245 (54.6%) were excluded based on identifiable exclusion criteria during chart review or discussion with the treating provider, resulting in a screen success rate of 45.4%. Of the remaining n = 203, another n = 139 were excluded, resulting in a recruitment rate of 31.5% (Fig. 2). During rolling recruitment, a total of n = 4 participants withdrew consent after randomization but before their month 1 visit (n = 2 in the Control arm and n = 2 in the Messages arm). All 4 indicated that they were unable to commit to monthly in-person visits for the first 3 months of the study. Given the pilot nature of this study, these participants were replaced, bringing the total recruited to n = 64, and total enrolled to n = 60. Of the 60 participants, one withdrew consent prior to month 9, 16 were lost to follow up (did not show up for a study visit for > 6 months) or dropped out before the study ended, and 3 participants did not complete or only partially completed their month 9 survey. Additionally, 2 participants stopped using their MEMS Cap following the intervention period, resulting in a final sample of 39 with survey and MEMS data, and 42 with BP data.

Fig. 2.

CONSORT Flow Diagram.

There were no significant baseline differences between study arms (Table 1). Overall, the average age was 63 ± 16 years and n = 25 (49%) were female, with n = 35 (59%) self-reporting as White. Participants typically had a Bachelor’s degree or higher (63%), were either employed (44%) or retired (35%), and had a household income ≥$50,000 (61%). Polypharmacy was common, with patients taking on average 3 ± 2 AHT medications and 10 ± 8.5 medications for other conditions. At baseline, mean systolic BP was 140 ± 22 mmHg, with 28% of participants having a BP < 130/80 mmHg.

Table 1.

Characteristics across study arms at baseline. BP; blood pressure. SD; standard deviation.

	Overall (n = 59)	Study Arms					P-value
		Control (n = 19)		Messages (n = 20)		Incentives (n = 20)
Age, mean (SD)	62.58 (16.0)	64.11 (20.8)		63.15 (13.5)		60.55 (13.7)	0.4766
Female, n (%)	25 (0.42)	10 (0.53)		10 (0.50)		9 (0.45)	0.8906
Race, n (%)
White	35 (0.59)	10 (0.53)		11 (0.55)		14 (0.70)	0.4898
African American or Black	9 (0.15)	4 (0.21)		3 (0.15)		2 (0.10)	0.6355
Asian	7 (0.12)	4 (0.21)		2 (0.10)		1 (0.05)	0.2924
Hispanic, n (%)	5 (0.08)	1 (0.05)		2 (0.10)		2 (0.10)	0.8326
Average household income >$50,000, n (%)	36 (0.61)	12 (0.63)		11 (0.55)		13 (0.65)	0.7920
Employment Status, n (%)
Full Time (≥ 35 h/week)	19 (0.32)	6 (0.32)		6 (0.30)		7 (0.35)	0.9429
Part Time (< 35 h/week)	7 (0.12)	2 (0.11)		3 (0.15)		2 (0.10)	0.8684
Retired	21 (0.36)	8 (0.44)		6 (0.30)		7 (0.35)	0.7082
Unemployed	10 (0.17)	2 (0.11)		4 (0.20)		4 (0.20)	0.6679
≥ 1 year of higher education, n (%)	56 (0.94)	19 (1.00)		16 (0.80)		18 (0.90)	0.1228
Insurance, n (%)
Medicare	28 (0.47)	11 (0.58)		11 (0.55)		6 (0.30)	0.1345
Medicaid	7 (0.12)	2 (0.11)		3 (0.15)		2 (0.10)	0.8333
Private	34 (0.58)	10 (0.53)		11 (0.55)		13 (0.65)	0.7371
Other	4 (0.07)	2 (0.11)		1 (0.05)		1 (0.05)	0.7517
Travel time to clinic [minutes], mean (SD)	35.29 (26.0)	28.58 (18.5)		42.05 (35.0)		35.25 (21.3)	0.5240
# of Antihypertensive Medications, mean (SD)	2.69 (1.6)	2.63 (1.6)		3.00 (1.9)		2.45 (1.4)	0.6993
# of non-hypertensive medications, mean (SD)	10.00 (8.5)	10.05 (8.6)		9.40 (7.4)		10.55 (9.8)	0.9930
Systolic BP, mean (SD)	139.98 (22.1)	139.78 (19.5)		142.72 (27.7)		137.43 (18.8)	0.9475
Diastolic BP, mean (SD)	78.80 (16.7)	72.88 (12.7)		80.50 (19.5)		82.72 (16.1)	0.1222
Controlled BP (< 130/80 mmHg), n (%)	17 (0.29)	6 (0.32)		5 (0.25)		6 (0.30)	0.8947
Existing techniques to help remember to take medications, n (%)	29 (0.49)	13 (0.68)		9 (0.45)		7 (0.35)	0.1173
Automaticity total (SRBAI sum; range = 4–20), mean (SD)	15.03 (4.2)	16.05 (3.1)		14.89 (4.4)		14.20 (4.8)	0.4236
Habit score (EHS average; range = 1–5), mean (SD)	4.52 (0.8)	4.40 (1.0)		4.53 (0.8)		4.63 (0.7)	0.7940

P-values derived using the Kruskal-Wallis test.

^These participants noted that they were interested, but requested follow up or said they would follow up later but did not.

*Includes 2 participants who gave both reasons, and while they are counted twice (i.e., once in each category), they are only counted once in total approached.

Overall trial retention rate at the end of the trial period was 67.18% (Control: 68.18%, Text Messages: 59.2%, Intervention: 75%).

Preliminary efficacy

Mean daily adherence during the extended follow-up period (months 7–9) was numerically higher in both intervention arms than the control arm (mean adherence and 95% CI: Control arm 75.27% [59.1%, 91.5%]; Messages arm: 83.97% [67.1%, 100.9%]; Incentives arm 76.74% [59.23%, 94.3%]) though this pilot study was not powered to detect a statistically significant difference (P = 0.73). We found similar results when daily adherence was evaluated during the intervention (months 0–3), immediate post-intervention period (months 4–6), and full post-intervention period (months 4–9). (Fig. 3A). We next examined differences in cued adherence. Overall, cued adherence was lower than mean daily adherence, but similar trends emerged, i.e. greater cued adherence in the Messages and Incentives arms than the control arm (mean cued adherence and 95% CI for months 7–9: Control arm 39.59% [23.6%, 55.6%]; Messages am 53.11% [32.4%, 73.9%]; Incentives arm 49.52% [29.1%, 69.9%]), though we were underpowered to detect a statistically significant difference (P = 0.50) (Fig. 3B). No between-group differences in BP control were observed at month 9 (Fig. 3C).

Fig. 3.

Preliminary Efficacy Outcomes. (A) mean daily adherence, (B) mean cued adherence, and (C) blood pressure control, stratified by intervention arm across prespecified study periods.

Similar findings emerged for the pooled intervention arms (Messages and Incentives arms) (Supplemental Fig. 2).

Acceptability

Most participants were satisfied with the BETA intervention and found many of its components acceptable: 82% of survey respondents said they were moderately satisfied, and 90% would recommend BETA to a friend. 80% of participants found the anchoring strategy to be effective in integrating medications into existing routines. Most survey respondents were comfortable (76%) or very comfortable (15%) with using MEMS Caps, however, in qualitative interviews some expressed mixed perceptions about MEMS Caps and their influence on habit formation. For instance, while some noted that MEMS Caps reminded them to take pills because they felt that it was like a “trigger mechanism” since they were being “watched,” others found it disruptive to already established habits, such as when they were already using a pill box for their medications. More than two thirds of survey participants agreed that text messages helped them stay motivated to continue taking their medication. By comparison, some participants found these texts to be superfluous due to already existing routines, while others felt the messages were repetitive and too frequent. One, for example, noted that they “kept coming all the time.” Most survey respondents (80%) reported it was easy to understand how to qualify for prizes, and 60% agreed that the prizes helped them stay motivated. In interviews, some participants suggested that prize draws should occur more often. Finally, almost 70% of those surveyed agreed it was easy to come to the clinic for study visits, though 60% noted they would prefer remote visits. Most respondents in the qualitative interviews also preferred remote visits, with reasons ranging from living far away from the medical center or experiencing transportation issues. Based on attendance data, 63% participants returned for all three intervention visits.

Discussion

This pilot RCT demonstrated the preliminary efficacy, feasibility, and acceptability of a behavioral economics-based intervention aimed at improving healthy AHT medication adherence habits. While this pilot study was not powered to detect between-group differences, results indicate the potential for efficacy of the intervention that should be evaluated in a fully-powered study. Specifically, higher medication adherence levels were observed in both intervention arms than in the control arm for both daily and cued adherence. Additionally, BP control increased numerically more in the intervention arms than the control arm across study periods.

BETA was adapted from formative work examining the use of BE-based incentives to improve antiretroviral therapy medication habits in sub-Saharan Africa²⁹. This novel approach to habit formation combines anchoring (a traditional approach) with incentives based on insights from BE to counter the frequently observed high attrition during the relatively long period it takes to form a habit. BETA consequently sought to overcome the barriers of present bias and lack of disease salience through the introduction of conditional rewards and text message reminders to support habit formation and avoid attrition^28,45,46. Text messages increase the salience of the habit formation behavior while continuing to reinforce information about the importance of habits provided at recruitment, while small incentives conditional on cued adherence (i.e. adherence in line with a participant’s anchoring plan) target present bias^47–50. The key methodological innovation is that the incentives are explicitly conditioned on cued adherence rather than on mean adherence as in most existing incentive-based interventions, with the goal to improve timeliness as a key dimension of habitual adherence.

Results demonstrate the potential feasibility and acceptability of the intervention, evidenced by both the recruitment (on average, and over a 30-day period) and retention rates. We did however observe a relatively high exclusion rate that we found to be due to the requirement of in-person study visits, rather than concerns regarding the intervention itself. Additional changes to the design of the protocol have been noted through the results from the adaptation phase of the trial, including those concerning in-person visits, and the use of pill bottles for storing the medication.

Study participants also reported high acceptability of the intervention including comfort with using the electronic pill caps, likelihood to refer a friend to the program, and a belief that both the messages and incentives helped with habit formation and maintaining motivation. We do, however, recognize that the results were driven by those participants that stayed in the trial at least until the end of the intervention period, i.e. 73.4% of the sample. We also identified points of improvement from the adaptation phase, including the frequency and content of the text messages, as well as the frequency of prize drawings. We discuss these in more detail below.

While our pilot intervention was not powered to detect statistically significant group differences, we found promising preliminary evidence that it improved medication adherence habit formation, as evidenced by increased mean adherence as well as timeliness of AHT adherence in both intervention groups compared to the control. Mean adherence dissipated somewhat in the post-intervention observation period, indicating that there is room for improvement based on the promising preliminary results of this pilot study.

Several changes in the study design may increase the impact of our approach for supporting habit formation in a future, fully-powered trial. First, given the high baseline AHT adherence rates observed in the current study, the intervention should be targeted to those in need of adherence support, such as those failing to show drug refills or missing clinic appointments. Second, coming to the clinic in person for study purposes was a commonly expressed drawback of the study; a fully remote intervention approach could reduce this barrier to study participation and increase completion rates. Third, because mean adherence is observed to decrease over time, it would be prudent to revise the intervention components based on feedback received from the extensive qualitative data collected as part of this pilot study, including the exit interviews. Improvements could include making the text messages less monotonous or increasing the frequency of the prize drawings. A fully powered follow-up study should also include a longer post-intervention observation period to both fully evaluate the persistence of habits after the intervention is withdrawn, as well as study when and why habits dissipated over time, to what extent individuals manage to re-establish them on their own, and what support they need to re-establish a previously formed habit. A future intervention at scale should also include a cost effectiveness analysis, as the novel approach of using a short-term intervention to bring about long-lasting improvements in AH medication adherence is likely to be cost effective.

Strengths of our study include the testing of a novel approach to habit formation based on insights from behavioral economics, a robust study design based on a RCT, and the use of an objective measure of cued medication adherence. Several limitations also merit consideration. First, we are unable to determine medication adherence prior to enrollment. Additionally, participants demonstrated relatively high daily adherence compared with reported rates from prior studies^51–53. Despite this, we were still able to achieve the objectives of demonstrating feasibility and acceptability and noted trends indicating potential efficacy of the approach. Future studies should focus on recruiting patients with known low adherence or new drug initiators who do not yet have established habits to assist them with medication adherence. Our enrolled cohort was also relatively well educated and retired. These two factors may be associated with both improved medical literacy and fewer competing factors which otherwise may contribute to low adherence. Further, we used MEMS bottle opening as a surrogate for medication adherence compared with more direct techniques such as blood level monitoring or directly observed therapy. This measure of adherence has been used successfully in numerous other studies and been found to closely correlate with biomarkers of adherence^54–56. Answers in the surveys may have been impacted by acquiescence or social desirability bias; we attempted to reduce this concern by mostly using previously tested survey questions, used balanced question wording (i.e. a mix of positively and negatively framed questions), and performed analyses of response patterns. We also thoroughly trained the study coordinators to make sure they were appropriately able to convey to participants that they could answer in a truthful manner. We also stressed this point in the consent forms, i.e. that the participant would receive the same level of care no matter whether they left the study or did not answer (parts of) the survey. Finally, we observed relatively high study refusal, with the primary reason being the need for in-person visits. Reluctance to come to the clinic in person also was the major reason responsible for the dropout of study participants (n = 4, or 6.25% of the total number enrolled). This insight may inform future trials to consider use of virtual visits and select monitoring devices that can transmit data remotely, rather than requiring in-person data extractions. We attempted to contact several subjects who dropped out and offer the opportunity to share their thoughts on the study as part of exit interviews, however, these subjects declined further participation.

Conclusion

This pilot randomized controlled trial demonstrated the feasibility, efficacy, and preliminary efficacy of a BE-based intervention to form healthy habits for AHT medication adherence. Future fully powered trials should focus on engagement of patients with lower baseline medication adherence and leverage remote data transmission to allow for virtual visits to boost retention.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

SL and JE designed the study and led all aspects of developing the protocol. JE, SL, and AP advised on the implementation of the interventions. DB, IG, RV, CB, MM, SJ, and NG contributed substantially to the design of the instruments and data collection. SL drafted the initial manuscript. SL, JE, and IG contributed to the quantitative analyses, while AP and IG contributed to the qualitative analyses. SL, JE, AP, and IG edited and refined the manuscript. All authors read and approved the final manuscript.

Data availability

Due to the sensitive nature of the data collected for this study, requests to access the dataset from qualified researchers trained in protocols on the protection of human subjects may be sent to Cedars-Sinai Medical Center at cda-research@cshs.org.

Declarations

Competing interests

The authors declare no competing interests.