The research team, led by investigators at Rush University System for Health in Chicago along with data scientists at Nuna, Inc., conducted a retrospective observational study using electronic health record data from primary care practices. Adults aged 18 and older with a documented hypertension diagnosis who activated the app between April 2024 and September 2025 were matched one-to-one with hypertensive patients receiving standard care, using a greedy nearest-neighbor algorithm with a caliper of 0.25 standard deviations of the logit of the propensity score. The matching model incorporated a broad set of covariates: age, sex, race and ethnicity, insurance type, comorbidities such as diabetes and hyperlipidemia, the Charlson Comorbidity Index, the Social Vulnerability Index, distance from the primary care provider, visit frequency, prior hospitalizations, and baseline blood pressure values. After matching, 425 app users were compared with 425 controls, each followed for six months before and six months after an index date—app onboarding for users, and a randomly assigned pseudo-engagement date for controls to align calendar time and secular trends.

The resulting cohort was notable for its demographic profile: a mean age of 59 years, 74 percent women, roughly two-thirds identifying as Black, and 11 percent Hispanic. At baseline, 40 percent of participants in each group met criteria for uncontrolled stage 2 hypertension, 33 percent had stage 1 hypertension, and 27 percent had controlled blood pressure. The final analysis relied exclusively on clinic-measured blood pressure values from outpatient and telemedicine encounters, deliberately excluding readings taken during hospitalizations or emergency department visits to avoid confounding by acute illness. When multiple readings occurred on the same day, they were averaged into a single daily value, and participants needed at least one qualifying measurement in both the pre- and post-periods to be included.

The headline results were consistent across outcomes. Mean systolic blood pressure fell from 136.3 to 131.7 mm Hg among app users, a decline of 4.6 mm Hg, compared with a drop of only 1.6 mm Hg among matched controls, whose mean fell from 136.0 to 134.4 mm Hg—a statistically significant difference (p = 0.003). Diastolic pressure followed the same pattern, declining 2.5 mm Hg in the app group versus 0.9 mm Hg in controls (p = 0.012). The effect was even more pronounced among the 40 percent of participants who began with uncontrolled stage 2 hypertension: app users in this subgroup saw systolic pressure fall by 13.6 mm Hg, compared with 9.0 mm Hg among controls (p = 0.009). More than half of these high-risk app users—56.7 percent—transitioned to an improved blood pressure stage during follow-up, versus 46.2 percent of controls, and a greater share of app users achieved a systolic reduction of at least 10 mm Hg, a threshold long associated with meaningful reductions in cardiovascular risk.

That 10 mm Hg benchmark carries substantial clinical weight. A widely cited meta-analysis of blood pressure lowering trials found that every 10 mm Hg reduction in systolic pressure corresponds to a 28 percent lower risk of heart failure, a 27 percent lower risk of stroke, a 20 percent lower risk of major cardiovascular events, and a 13 percent lower risk of all-cause mortality. The 13.6 mm Hg decline observed among app users with uncontrolled stage 2 hypertension therefore exceeds what would be expected from many pharmacologic additions, and it occurred without any protocolized medication changes or mandated clinician actions—suggesting that behavioral support and self-management, rather than intensified prescribing, drove much of the improvement.

Understanding how the app works helps explain why. Participants received an FDA-approved home blood pressure cuff that synchronized readings directly with the application, which was available in both English and Spanish. An AI-driven health coach checked in through voice and chat interactions, gathered context about habits and symptoms, and answered health questions. The app also offered medication reminders with self-reported adherence tracking, daily step goals linked to smartphone activity sensors, nutrition logging with personalized AI-generated feedback, and educational modules on cardiovascular health. Users earned points for completing tasks—recording blood pressure, logging medications, hitting step goals, logging meals—and accumulated points advanced them through engagement levels and entered them into sweepstakes for modest financial incentives averaging five dollars per month, redeemable as gift cards. Critically, the program was not a standalone consumer gadget: a Rush nurse monitored alerts for critically high or low readings and for sustained stage 2 hypertension, triaging patients according to established nurse protocols and documenting every outreach in the electronic health record with the primary care provider copied. Community health workers employed by Nuna and contracted to Rush helped onboard patients, provided education on the cuff and app, administered a social needs screener, and continued periodic outreach to address technical barriers.

Perhaps the most striking finding was retention. Digital health interventions routinely collapse under the weight of user attrition; cross-study analyses of 100,000 participants in remote digital health research have reported median retention of less than 30 days. In this study, 74.7 percent of app users accessed at least one app feature every week during the six months after onboarding, and 87.3 percent were still active in the sixth month. Median retention was 180 days—the entire follow-up period. Engagement also tracked with outcomes: each week of active app use was associated with 2.56-fold higher odds of achieving blood pressure control in the post-period (95 percent confidence interval 1.28–5.10), and users whose blood pressure was controlled at follow-up showed consistently higher weekly activity and retention than those who remained uncontrolled. Even among patients whose blood pressure was already controlled at baseline, the app group showed smaller increases in systolic and diastolic pressure over time than controls, hinting that digital tools may help prevent the clinical backsliding that commonly afflicts hypertension management.

The study’s pragmatic design strengthens the case that these are real-world effectiveness signals rather than idealized trial results. Recruitment itself revealed important lessons about implementation: cold-calling 4,207 eligible patients converted only 12.8 percent of those contacted into enrollees, whereas switching to clinician referral produced a 70.9 percent conversion rate—an observation the authors say underscores how central physician endorsement is to digital health uptake. Of those who enrolled, more than 90 percent ordered a blood pressure cuff, and 553 participants successfully synced their devices. The intervention augmented rather than replaced clinician-directed care, aligning with current recommendations that digital tools support longitudinal primary care rather than operate as parallel systems.

The authors are careful to enumerate limitations. The single-site Chicago population, with high proportions of female and Black participants, may limit generalizability, and as an observational study it cannot fully exclude residual confounding—participants who chose to engage with the app may differ in motivation or health literacy in ways matching could not capture. Blood pressures came from routine clinical care rather than standardized research protocols, the analysis did not isolate which app components drove benefit, and the control threshold of 140/90 mm Hg may need revisiting as quality measures shift toward 130/80. Still, the magnitude and consistency of the blood pressure improvements, the exceptional retention, and the strongest effects precisely in the highest-risk patients make a compelling case that AI-enabled, gamified, clinically supervised digital interventions can meaningfully complement traditional hypertension care—and potentially do so at scale, without proportionally increasing clinician workload.

Subject of Research: Association of an AI-enabled gamified mobile health application with blood pressure outcomes in patients with hypertension

Article Title: Association of an AI-Enabled Gamified Mobile Health App on Blood Pressure Outcomes in a Matched Cohort

Article References: Gottlieb, M., Pallok, K., Zimmermann, L., Thompson, D., Ansell, D., Weaver, A., Kersemakers, D., Niehaus, K., & Walker, G. (2026). Association of an AI-Enabled Gamified Mobile Health App on Blood Pressure Outcomes in a Matched Cohort. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10684-0

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10684-0

Keywords: hypertension, mobile health, artificial intelligence, gamification, blood pressure, digital health, telehealth, home blood pressure monitoring, health informatics, cardiovascular risk, patient engagement, primary care