We will use focus groups to fulfill the objectives of Aim 1. Each focus group will have 6-8 participants, which is a size that allows for easy exchanges of ideas and a diversity of perspectives to be represented. We will recruit a convenience sample of 24-32 participants for four focus groups. Aims 1 and 2 will include the same participants who will attend a total of four focus group sessions (Figure 3).

The participants from San Francisco Veterans Affairs Medical Center (SFVAMC) are likely to be predominantly male; therefore, we will likely recruit one group of female Veterans from SFVAMC and one focus group of female non-Veterans from John Muir Medical Center (JMMC) in the East Bay. The four groups will account for variability in mobile phone use (low/high) and sex. Inclusion criteria for Aims 1 and 2 include: (1) >21 years of age, (2) history of ACS or PCI within one year, and (3) current/former antiplatelet (thienopyridine) prescription. Exclusion criteria are: (1) cognitive impairment, and (2) lack of English proficiency/literacy.

Thematic analysis will include reading and coding the completed field notes and interview transcripts. We will create an initial set of codes that account for the study aims and other topics that emerge from the data. Some codes will be generated deductively, while others will be generated inductively based on themes and issues that emerge from ongoing coding. We will compare coding, resolve differences, and revise the coding scheme. We will assess intercoder reliability by having 10% of the data cross-coded. The key outcomes will be content and frequency of text messaging, particularly focusing on how preferences vary across groups. We will use ATLAS.ti (Thousand Oaks, CA) for all qualitative analyses to determine the preferred app for Aim 3.

Aim 2 consists of Aim 2a (content analysis) and Aim 2b (usability testing). The usability study will include the same individuals from the focus groups in Aim 1 who are stratified by low/high mobile phone use and sex, to accommodate easy exchanges of ideas and a diversity of perspectives. In contrast to Aim 1, there will be three focus group sessions to fulfill Aim 2. Recruitment, group assignment, and data collection will be the same for Aim 2a and Aim 2b.

The research design for Aim 3 is an RCT that lasts 12 months. In contrast to Aims 1 and 2 that included participants who were post-ACS/PCI within one year and recruited from a cardiac rehabilitation program, participants for Aim 3 will be recruited from the hospital after ACS/PCI.

We plan to enroll 225 total participants (75 participants per group). Allowing for a 15% attrition rate over the 12-month study, we conservatively estimate that this recruitment target will result in a total sample size of n=192, which will give 86% power to detect a medium effect size (half-SD difference) between at least one treatment group and the control group using Dunnett’s method to adjust for multiple comparisons. This sample size would give more than 95% power to detect an effect size of d=0.69, as found in our previous clinical trial [33]. Analysis of the longitudinal data points with mixed-effects models will confer additional power.

The convenience sample will consist of 159 male and female Veterans from SFVAMC and 66 female participants (29% of sample) from JMMC. A sample size of 225 is considered to be feasible considering the number of post-ACS/PCI patients who are treated at both institutions. We have incorporated a 29% oversample of women to ensure a sufficient sample size to examine subgroup differences in intervention effects, given the low enrollment of female Veterans in Veterans Health Administration (6%) [38]. Inclusion criteria include: (1) >21 years of age, (2) recent ACS or PCI within one week, and (3) new antiplatelet (thienopyridine) prescription. Exclusion criteria are: (1) cognitive impairment, and (2) lack of English proficiency/literacy. For the process evaluation, we will recruit a subset of 10 participants from each experimental group (text messaging and mobile app; total n=20) for in-depth individual interviews.

We will recruit from SFVAMC and JMMC inpatient units using two recruitment strategies. First, we will inform the cardiologists and nurses in the cardiac units at the hospital sites about the study so they can identify patients who meet eligibility criteria. A patient-friendly handout will be made available for the health care team to give to interested participants. Second, we will obtain a daily list of patients who had PCI and their electronic medical records will be screened for eligibility in the study using a checklist to examine the inclusion and exclusion criteria. Participants who agree to participate will provide written informed consent.

Group assignment will be generated by random allocation sequence using random blocks stratified within recruitment site (SFVAMC or JMMC) of either three or six, which will be prepared by the study biostatistician. The research team will assign patients to their groups by distributing presealed envelopes in consecutive, numbered order. Due to the nature of the study design, the research team and participants cannot be blinded to the intervention once group assignment is determined.

After written consent is obtained, we will collect clinical data by reviewing electronic medical records. Participants will complete the following baseline questionnaires: (1) sociodemographic information and medical history, (2) mobile phone use, and (3) Mini-Cog Test [39]. All instruments and frequency of administration are described in Table 2.

We will conduct an RCT with random assignment to one of three groups: (1) text messaging, (2) mobile app, or (3) website-control, for a total of 12 months. Both intervention groups will also be offered the patient education website that the control group will be given. The text messaging group will receive text messages with content and frequency informed by Aim 1. We will use the Veterans Affairs (VA) Annie text messaging program. For non-Veteran females, we will use a text messaging program from CareSpeak Communications. The app group will use the preferred app that is selected in Aim 2. The major purpose of the app is to remind participants to take medications; however, the app will likely include other features such as providing educational messages. The website group will be offered the American Heart Association patient education website (My Life Check - 7 Steps To Healthy Living) and will serve as an active control group [43]. The website will be offered to participants in all three groups. The 7 small steps to big changes are to manage blood pressure, control cholesterol, reduce blood sugar, get active, eat better, lose weight, and stop smoking.

Subsequent in-person follow-up visits at 1, 6, and 12 months will take place in participants’ homes or other mutual meeting places as requested. Follow-up visits will be scheduled at times that are convenient for participants to meet.

Medication adherence will be assessed with five different measures at various intervals to validate patterns of adherence, including: Medication Event Monitoring System (MEMS; key primary endpoint being percentage of prescribed doses taken), Morisky Medication Adherence Scale (MMAS-8), 7-day recall, VA Corporate Data Warehouse (Veterans), and Peoplechart Meds Incontext (non-Veterans; Table 2). In-person follow-up visits will occur at 1, 6, and 12 months, while monitoring of adherence through refill data from computer software will be done monthly to reduce participant burden.

MEMS technology uses integrated microcircuits with bottle caps that record the date and time that an individual opens a pill bottle. Participants will be given one bottle to store their antiplatelet medication, and will be directed to store their other medications as they normally would. For those who use 7-day pill boxes, we will emphasize the need to use the MEMS bottle for the antiplatelet medication. Data will be wirelessly transferred to a MEMSCap reader, which is a small device that connects to a computer software program.

Refill data from computer software will also allow us to monitor prescription adherence passively, without requiring patient participation. We will use administrative data from the national VA repository (VA Corporate Data Warehouse) to provide adherence to antiplatelet medication for Veterans. For non-Veteran females from JMMC, we will use an equivalent data source through Peoplechart Meds Incontext, which is a software program that tracks adherence through sanctioned prescription data sourced directly from retail pharmacies and Prescription Benefits Managers (ie, mail-order companies).

We will examine other variables (health-related quality of life, social support, self-efficacy, depression, anxiety, and post-traumatic stress disorder) and their relationship to medication adherence. Patients will complete the instruments at baseline, 1, 6, and 12 months (Table 3).

Upon completion of the intervention study, we will conduct a process evaluation so that improvements can be made for a multi-site full scale study. The objectives of the process evaluation are to assess participants’ perspectives on feasibility, acceptability, usability, challenges, strengths, and weaknesses of the intervention. In addition, we will ascertain reasons for nonadherence to antiplatelet medications since understanding the reasons for nonadherence is crucial to avoid negative patient outcomes. We will adopt a qualitative research approach using in-depth individual interviews from a diverse sample of participants (total n=20).

Descriptive statistics, means, and SDs for quantitative variables, and frequencies and percentages for categorical variables, will be provided for all demographic and study variables. Random assignment of the patients to the text messaging, mobile app, and website groups should ensure that there are no preexisting differences between the three groups in personal factors (eg, sociodemographic, clinical, and psychosocial characteristics).

The primary outcome will be the percentage of prescribed doses taken based on MEMS data (aggregated over 12 months). Secondary outcomes will include: (1) percentages of days correct number of doses were taken, (2) doses taken on schedule, (3) MMAS-8, (4) 7-day recall, (5) VA Corporate Data Warehouse (Veterans), and (6) Peoplechart Meds Incontext (non-Veterans). The MMAS-8 will be analyzed as a range of 0-8. We will synthesize adherence measures with the key primary endpoint of MEMS data compared with secondary measures of adherence to assess for concurrent and predictive validity. We will also calculate associations between all of the adherence measures but expect that they will track well together. Prior to fitting the mixed-effects linear model, we will examine univariate and bivariate relationships between the outcomes and both demographic and coded qualitative responses.

Our primary analysis will use a mixed-effects linear model. The predictors will be a categorical predictor of group (with control as reference), time, and the group-by-time interactions, as well as the recruitment site (SFVAMC vs JMMC), as randomization is stratified by site. The parameters in the group-by-time interaction are the primary parameters of interest, describing whether adherence changes differentially over baseline in each of the treatment groups compared to the control group. We will also investigate whether other models for change over time are more appropriate, including arbitrary means or nonlinear time functions through comparison of Akaike information criterion/Bayesian information criterion. Secondary analyses will adjust for select covariates that are strongly associated with adherence.

To determine the factors related to medication adherence, we will examine sociodemographic, clinical, and psychosocial variables. The variables that show at least a moderate correlation and have a significant relationship to medication adherence (P<0.10) will be included in a multiple regression analysis in which medication adherence is the dependent variable. The multiple regression analysis will provide the optimum combination of these variables to explain the total percent of variance in medication adherence, and indicate the unique contribution of each variable to the regression model.

The in-depth individual interviews for process evaluation will be transcribed, coded, and analyzed in the same manner as in Aims 1 and 2. We will explore reasons for attrition over the 12-month period. We will complete field notes after the conclusion of each interview, and all interviews will be transcribed verbatim. Thematic analysis will include reading the completed field notes and interview transcripts. The key outcomes will be related to patient satisfaction and process evaluation, particularly focusing on how views vary across groups.