The standard assumption in digital health, the one I've heard from product leaders at three different companies and read in roughly half the strategy decks I've ever opened, goes like this. Older adults don't engage with apps. They don't trust them. They drop off fast. So if you're building a behavior-change tool, optimize for the under-50 crowd and treat the 65+ segment as a secondary use case, maybe with a phone-call fallback.
A study published April 17, 2026 in Frontiers in Digital Health makes that assumption hard to keep.
The team analyzed real-world data from 250 adults using the Perx Health app, a gamified medication adherence platform. Mean age was 70.1 years. 67% had three or more chronic conditions, which is roughly the population you'd actually want a medication app to serve.
The median retention was 595 days. Not 595 hours, not 95 days. 595 days. The mean was 779. The longest user in the dataset stayed for 2,465 days, which is just under seven years on the same medication app.
That is not the engagement profile of a population that "doesn't do digital health."
What the app does
Perx is a behavioral-science-informed adherence app. Users log medications by taking a photo (mobile directly observed therapy, mDOT), get reminders, and earn rewards through what the paper describes as "lottery-style" incentives, point accumulation, streaks, and social leaderboards.
The behavior change techniques on display map cleanly to the BCT taxonomy. You've got self-monitoring of behavior (BCT 2.3), feedback on behavior (2.2), prompts and cues (7.1), social support (3.1), social comparison (6.2), and material incentive (10.1). It's a stack of relatively well-evidenced BCTs, applied consistently to a single behavior with a clear daily anchor, taking a pill.
The engagement numbers
Daily engagement was the part I had to read twice.
Average users opened the app 6 times a day. They spent 8 minutes 35 seconds in the app daily. They completed 4.9 tasks per day. The cohort was 61.8% female, mostly Australia-based, mean age just over 70.
For comparison, the persuasive design meta-analysis I wrote about elsewhere reported that across 92 mental-health-app RCTs, only 76% of studies even reported engagement, and when they did, completion rates and login frequencies varied wildly. The Perx cohort isn't a randomized trial, so the comparison isn't fair, but the absolute numbers are still striking. Six logins a day from a 70-year-old population is not what the discourse predicts.
Median medication adherence in the cohort was 95.0%. Mean was 86.2%. To put that in context, meta-analyses of medication adherence in older adults typically report adherence in the 50 to 60 percent range. So either this cohort is wildly self-selected (it almost certainly is to some degree, this is real-world observational data), or the app is doing something useful, or both.
What surprised me
Two things, honestly.
The first is that gamification worked, and the part of gamification that worked was the social part, not the money part. Leaderboard engagement was 62.5%. Streak engagement was 26.5%. The paper explicitly notes that older adults preferred social and intrinsic motivators over monetary rewards.
This contradicts a story I've told in slide decks before. The story goes that older users are skeptical of social features and prefer simple, transactional interactions. The data here says the opposite. Leaderboards beat points beat money, in a population whose median age was 70.
The second is the duration. Mean retention of more than two years. We have got to stop using "30-day retention" as the dominant metric for behavior-change apps. Thirty-day retention is fine for consumer media products where the marginal day is worth roughly the same as every other day. For a medication adherence app, day 600 is worth more than day 6, because you're treating chronic conditions and the value compounds. Measuring 30-day retention on a tool whose entire purpose is multi-year behavior change is a metric mismatch, and we keep doing it because that's what the dashboards default to.
What this changes about how we should design for older users
A few takes I think actually hold up.
The "older adults don't engage" assumption is doing damage. Every product team I've watched scope a digital health tool has under-invested in the over-65 segment because of an unspoken belief about engagement. This paper is one data point against that. There are others. The Perx team's earlier publication is consistent. The assumption has had a long enough run. Time to update.
Build for the use case, not the age. The reason this cohort engaged is because they had three or more chronic conditions on average, the medication regimen was complicated, and the app made the daily ritual easier and more rewarding. If you build for that situation, age stops being a signal. People who actually need the tool, use the tool. The ones who don't need it, won't, and that's true at 30 and at 70.
Lean into social, not monetary. This is the Perx finding I want product teams to actually internalize. If you're building incentives into a behavior-change app for older users, the leaderboard outperforms the gift card. I'd be cautious about generalizing too far (this is one cohort, one platform, one set of design choices), but it's directionally consistent with what self-determination theory would predict, where intrinsic and relational motivators outlast extrinsic ones.
Photo-based adherence verification is underrated. The mDOT mechanism (take a photo of your pill, get credit) does two things at once. It captures objective adherence data, which gives the team a real outcome metric, and it ritualizes the medication moment, which is itself a behavior-change technique (BCT 8.3, habit formation). I think more digital health tools should steal this pattern.
Stop measuring 30-day retention on chronic-condition apps. Use 6-month, 12-month, multi-year retention. The entire reason the tool exists is sustained behavior. Pick a metric that actually rewards what you're trying to do.
The caveats, because I'm not an academic
This is observational, real-world data. People who self-selected into Perx and stayed long enough to be in the analysis are not a random sample of older adults. The 95% adherence rate is in that selected cohort, not in the population. The retention numbers are conditional on already being a user. This is not an RCT and it does not establish causation.
But the engagement numbers are descriptive, and they describe a population that exists. The 250 people in this dataset spent more than two years on average using a medication app, daily, with measurable behavior. That is real. That is not what most product teams plan for when they design for older users.
I think this paper is one of the more useful things I've read in a while, less because of the effect sizes and more because of the assumption it disturbs. Most behavior-change software is being built by people in their 30s for an audience the team imagines as people in their 30s, and the actual best users may be twice that age, doing more with the product than the team built it for.
Read the paper if you build in adherence, geriatrics, or chronic care. Pay attention to the engagement breakdown by feature. The leaderboard number is the one I keep coming back to.
If you've shipped behavior-change software for older users and your data tells a different story, I really want to hear it. The default narrative needs more cases on both sides before any of us should be confident.