One in four people who download a health app deletes it without ever coming back.
That number comes from a scoping review published in JMIR mHealth and uHealth this month, covering 52 studies on user engagement in digital health applications. The authors, Melek Aktas, Linda Cambon, and Olivier Aromatario, didn't set out to produce a damning verdict on the field. But the 25% single-session abandonment rate is sitting right there in the data, and I think it's worth sitting with for a minute before jumping to solutions.
Because the real finding isn't the abandonment rate. It's that most product teams are measuring the wrong thing when they try to understand engagement at all.
What "engagement" actually means, depending on who you ask
The review identifies four distinct perspectives on engagement that show up across the literature. They're worth naming because I don't think most product teams I've worked around have a shared vocabulary for this, and the confusion shows in how they build.
The first perspective is usage metrics: opens, sessions, time in app, retention curves. This is the one most teams default to because it's the one analytics tools give you for free. Open your Mixpanel or Amplitude dashboard and it's all right there. Retention at day 1, day 7, day 30.
The second is subjective experience: does the user actually find this useful, interesting, or motivating? This requires asking people, which means research, which means someone has to schedule it and someone else has to do it. You can see why this one gets skipped.
The third is a hybrid approach: combining both. Usage data plus qualitative input plus validated scales like the Mobile App Rating Scale (MARS) or the User Engagement Scale (UES). This is what the literature increasingly points to as the gold standard.
The fourth is goal-oriented engagement: whether the person is actually moving toward the health outcome they came for. Not just "did they open the app this week" but "are they smoking less, moving more, checking their blood pressure regularly." This is the one that matters most and is measured the least.
The problem with defaulting to usage metrics
Here's what I've seen happen in practice. A product team looks at their day-7 retention and sees 60%. They call that a win. The app stays installed. Sessions are regular. The numbers look good.
But the users aren't actually changing anything. They're logging meals they don't think about. They're tapping through reminder notifications to make them stop. They're doing the activity that keeps the streak alive, not the activity that changes the behavior.
The review makes this point methodologically: usage data tells you that people opened the app, not why they came back or what they got from it. An app can have strong retention numbers and zero behavior change. That's not a theoretical problem. That's a description of a lot of apps that are live right now.
Where engagement actually breaks down
The review organizes what drives engagement into three categories, and I think this framing is genuinely useful.
User-related factors include motivation, self-efficacy, digital literacy, and prior familiarity with the health topic. These are things you can't design away, but you can design for. A first-generation smartphone user managing a chronic condition has a different starting point than a 32-year-old who's already been to therapy and owns a continuous glucose monitor. The app that treats them identically is the app that loses one of them after one session.
Intervention design features include gamification, personalization, and whether the app is grounded in an actual behavioral theory. The review flags gamification specifically as having a positive association with engagement, but only when it's tied to meaningful progress, not points for points' sake. A streak counter that rewards opening the app is not the same as a streak counter that rewards completing a behavior that matters.
Implementation context covers things like whether the app is integrated into a care pathway, whether there's human support alongside it, and whether the organization deploying it actually trained anyone on how to use it. A well-designed app dropped into a clinical workflow with no training gets used badly or not at all.
The finding that actually surprised me
The review looked at how teams involved users in development. The ones that included users throughout the full development cycle, not just in a usability test at the end, produced better engagement outcomes.
This is not news. We've known this in UX for decades. And yet.
The standard pattern in health tech is still: build a thing, run a usability study at version 0.8, launch, watch retention drop off, do a post-launch survey. The involvement is real but it's a single touch, at a single point, after most of the decisions have been made. The review is pretty clear that this approach consistently underperforms compared to teams that have users in the room at every major design decision.
The 25% who delete after one session aren't failing. The product failed to give them a reason to come back on day two.
What this means for how product teams should work
Honestly, the frameworks in this review aren't complicated. What's complicated is that they require changing how teams prioritize.
If you're only measuring usage metrics, you're working with a partial signal. You need at least one qualitative channel, something that tells you what users actually think is happening when they use your product. That doesn't have to be monthly research sprints. It can be in-app surveys, session recordings, support ticket themes. Something that captures the subjective layer.
If your engagement strategy is gamification, check what behavior the gamification actually rewards. If the answer is "opening the app," that's not enough. The habit loop has to close on a meaningful action.
And if your roadmap currently has "user research" scheduled as a phase before launch and a phase after launch, that's probably not going to get you to the outcomes you want. The review is clear on this: ongoing involvement across development stages, not one-time participation.
The 52 studies don't give you a magic metric. But they do give product teams a sharper vocabulary for diagnosing where things went wrong. That's actually a useful thing to have.
Jameson Daines is a senior product designer specializing in virtual healthcare. BehaviorUX is his knowledge base on behavioral science applied to digital health product design. If I've gotten something wrong here, please let me know.