G amification and Social Incentives to A ugment ME dication Adherence (GAME Adherence) was a randomized clinical trial conducted from September 2022 through September 2023. The trial protocol (Supplemental Methods) was approved by the University of Pennsylvania Institutional Review Board, and all patients provided informed consent for participation and use of their data. Data were not deidentified. The study was conducted using Way to Health, a research technology platform based at the University of Pennsylvania used to implement and test behavior change interventions. Participants were eligible for the trial if they were ≥ 18 years of age, were prescribed 1 or 2 antihypertensive medications and a statin, had a medication possession ratio (MPR; number of days with a supply of medication divided by total number of days) 40-80% for at least one antihypertensive or statin medication over the past 6 months, and had a primary care provider at a participating Penn Medicine clinic. The upper bound for MPR was selected to ensure that we were enrolling patients who were nonadherent to cardiovascular medications; the lower bound was selected to avoid enrolling patients with more substantial barriers to adherence that may not have been addressed by the gamification intervention. MPR was calculated for each patient using data from Surescripts included in the electronic health record.

Potentially eligible patients were identified using data from the health system’s clinical data warehouse and were contacted by email, text message, and phone. Each potentially eligible participant received up to 2 emails and up to 3 phone calls from study staff, with each phone call preceded by a text message. Messaging was designed to maximize response to unsolicited outreach based on our team’s experience from previous studies, and noted that the patient’s primary care provider thought they would be a good fit for the study, indicated that the purpose of the study was to identify ways to help patients increase medication adherence, noted that the entire study would be conducted remotely, and included an offer of compensation for participating (up to $100 and a free blood pressure cuff). During the phone calls, participants could agree to participate or decline to participate; if the potentially eligible participant did not answer 3 phone calls, they were assumed to have declined participation. Participants who agreed to participate were provided with a link to the Way to Health platform, where they provided formal consent, enrolled in the study, and completed baseline questionnaires. Study staff then mailed the patient a blood pressure cuff, stickers to place on pill bottles indicating whether the medication was an antihypertensive or statin, and $50. Five days after this shipment, participants received an automated message from the Way to Health platform asking if they had received their blood pressure cuff and had a supply of their cardiovascular medications available. Once participants had a blood pressure cuff and a supply of their medications, the participant was asked by automated text message to check their blood pressure and send it to the study team by return text message. After reporting a blood pressure, the participant was randomized in a 1:1 ratio to control or gamification, using an electronic number generator in the Way to Health platform. Treatment assignment was necessarily unblinded to participants, but participants were not told anything about the other study arm. Investigators, statisticians, and data analysts remained blinded to arm assignments until the analysis was completed.

Participants in both arms received daily automated text messages asking if they took their antihypertensive medications and statins (“Did you take your blood pressure meds today? Text back yes if you did.”) and twice weekly text messages asking them to take their blood pressure and send in the result (“It’s time for your blood pressure check. Please check and send us your blood pressure.”) Participants in the attention control arm received these text messages and no other intervention for 18 weeks.

In the gamification arm, in addition to receiving the same text messages as the control arm, participants were entered into an 14-week game grounded in behavioral economic theory and adapted from gamification interventions that successfully increased physical activity. ^,, First, each participant signed a pledge to strive to take their medications each day. Second, at the start of each week, each participant received 90 points. Each day the participant did not take all of their medications or did not report their home blood pressure as requested (twice weekly), they lost 10 points. When participants took their medications and reported their blood pressure, they retained their points. We chose this “loss-framed” approach because prospect theory shows that highlighting potential losses is more effective at driving behavior change than emphasizing gains. Third, at the end of each week, participants moved up or down through 5 levels based on their points retained the previous week. Those with 70 points or more advanced one level; those with less than 70 points dropped down a level. Each participant began in the middle level so that they would have immediate motivation to at least maintain this status. Daily text messages noted the number of points each participant had retained for the week, and weekly text messages informed participants if they moved up or down a level. Fourth, each participant picked a family member or friend to receive an email each week summarizing the participant’s performance, and each participant’s primary care physician received a monthly report about the participant’s blood pressure and medication adherence. Involvement of the patient’s primary care physician and support partners was intended to leverage social accountability, as people typically derive additional motivation from not wanting to disappoint others.

At the end of the study, participants in both arms completed an end-of-study questionnaire, for which they were compensated $25. Participants were also offered an additional $50 compensation to undergo a blood draw for measurement of LDL cholesterol at any approved laboratory.

The primary outcome was patient-reported adherence, defined as days taking all medications divided by study days over the entire study duration. This was determined based on patients’ responses to daily text messages asking about adherence; patients who did not respond to the text message were assumed to be nonadherent. Secondary outcomes included change in blood pressure over the study duration, medication possession ratio (MPR) over the study duration and over weeks 14-18, and change in LDL cholesterol from baseline through the end of the study. Blood pressure was measured and reported to study staff by participants twice per week, MPR was captured from the electronic health record using data provided by SureScripts, baseline LDL cholesterol was captured from the electronic health record, and end-of-study LDL was measured by venous blood sample.

A priori power calculations were based on data from the WayToText study of bidirectional text messages to improve medication adherence, which closely approximated the design of the control arm of this study. Consistent with data from this trial, we assumed adherence would be 77% in the control arm with standard deviation of 18%. The trial was designed to include 84 patients (42 in each arm), providing 80% power to detect a 12 percentage point difference in self-reported adherence between groups, with a two-sided alpha equal to 0.05 and accounting for 15% drop-out or loss to follow-up. The enrollment rate was lower than anticipated, and we ultimately enrolled 43 participants. The study had 57% power to detect a 12 percentage point difference between groups, and 80% power to detect a 16 percentage point difference.

All randomly assigned patients were included in the intention-to-treat analysis.

The primary analysis fit generalized logistic regression models for patient-reported adherence to statins and BP medications on a daily level, adjusting for study arm and with participant random effects. Secondary analyses additionally adjusted for age, sex, race, and adherence during the pre-study period. We repeated these analyses for MPR, modeling this variable by determining whether the participant had or did not have a supply of their BP medication or statin on each day. The output of these models was the odds ratio (OR) for adherence in gamification vs. control; OR > 1 indicates greater adherence in gamification arm, and OR < 1 indicates greater adherence in the control arm. For systolic blood pressure, diastolic blood pressure, and LDL cholesterol, we fit generalized linear regression models, adjusting for study arm, baseline value, and with participant random effects. Secondary analyses additionally adjusted for age, sex, and race (Black vs. not Black). In these models, negative numbers indicate greater blood pressure decrease from baseline in the gamification arm, and positive numbers indicate greater blood pressure decrease in the control arm.

Statistical analyses were performed using SAS version 9.4 (SAS Institute) from October through December 2025.