A sweeping new study from Johns Hopkins University, published July 30 in JMIR Aging, uncovers a fundamental disconnect in the AI health tools being pitched to older adults: the people designing, buying, and deploying these systems don’t agree on what “cost,” “usability,” or “value” even mean—and the mismatch often leaves the intended user with an expensive, inscrutable product.
The Gap Between What Patients Need and What Builders Build
Researchers interviewed 49 stakeholders across six groups: older adults and care partners, clinicians, health-system leaders, payers, developers, and investors. Every group named cost, usability, and value as critical factors. But they meant wildly different things.
For older adults and their families, the immediate worry is out-of-pocket price and whether the tool accommodates physical or sensory limitations. A blood-pressure app that demands a tiny smartphone screen or a medication reminder that relies on high-pitched beeps won’t cut it. Clinicians want AI that doesn’t dump more work on their plates and that patients can actually afford upfront. Health-system leaders talk ROI and interoperability—will this tool prevent enough costly emergency visits to justify integration? Payers want evidence that it lowers claims. Developers and investors, staked on four-to-seven-year regulatory timelines and the need for a viable market, optimize for addressable market size and margins, not the narrow but pressing needs of a 78-year-old with arthritis and a fixed income.
“The result,” the study authors note, “can be a product strategy optimized for scale and margins rather than narrow but important needs in aging care.” In plain terms, the industry keeps building Swiss Army knives for people who need a simple can opener.
Affordability Isn’t Just a Licensing Fee
One of the study’s starkest findings is how ROI calculations at the system level obscure what a patient actually pays. A tool that looks like a bargain to a hospital CFO—cutting readmission rates, for example—may still be unaffordable or unworkable for the individual expected to use it every day. Recurring subscription fees, the need for a late-model tablet, or cellular data bills can lock out the very people who stand to benefit.
Developers told researchers that heavy regulatory costs force them toward high-margin business models. That’s understandable, but it also explains why so many health AI products float toward “healthy wellness” markets instead of sticking with dementia care or fall prevention. For IT leaders, the lesson is straightforward: compare total cost of adoption, not just license fees. That means adding in patient-facing charges, accessibility accommodations, staff training, workflow redesign, help-desk burden, and the productivity hit when clinicians become ad hoc tech support.
Accessibility Can’t Be an Afterthought
The most pointed criticism from the study’s participants: too many tools feel like “solutions in search of a problem.” Designers often adapt an existing AI model to a health-care scenario without starting from the specific tasks, constraints, and day-to-day realities of an older adult’s life.
Late-stage usability testing doesn’t fix this. An AI-powered pill dispenser that performs flawlessly in a clinic pilot can still fail at home if the setup wizard requires a Wi-Fi password typed on a tiny keyboard, or if the only onboarding materials are YouTube videos that assume digital confidence. The researchers urge development teams to bake accessibility into the very first requirement—not as a feature, but as a foundational expectation. For older users, that means large touch targets, clear audio prompts, low-friction authentication, and offline fallback modes for when connectivity stumbles.
How We Got Here: The AI Hype Cycle Meets Geriatric Reality
This tension isn’t new, but it’s accelerating. Over the past five years, venture capital has poured into health AI, often chasing the same “aging in place” trend without the patience to understand it. The assumption has been that a model trained on clinical data will naturally serve older patients. The evidence now says otherwise.
Regulatory timelines stretch from four to seven years for anything classified as a medical device, pushing startups toward wellness-adjacent apps that bypass the FDA but still claim health benefits. That financial pressure skews design toward younger, more affluent users who can absorb subscription costs. Meanwhile, older adults—who are disproportionately affected by chronic conditions—get left out of the design conversation. The Johns Hopkins study quantifies this divide through its stakeholder map, showing that the people who most need these tools are rarely the ones setting the product requirements.
What This Means for You—By Audience
If you’re an older adult or family caregiver
When a provider suggests a health app or monitoring device, don’t assume it’s ready for you. Ask direct questions: What will this cost me out of pocket? Has it been tested with people over 75 who have vision or hearing loss? What happens if I need help at midnight? If the answers are vague, dig deeper. You’re not being difficult—you’re doing the usability testing the developer should have done.
If you’re a clinician
Before endorsing a tool, probe whether it genuinely integrates into your workflow or just adds another layer of alerts. Check if the vendor has published any co-design research with older adults. A pilot that worked with a tech-savvy 65-year-old isn’t the same as one that worked with an 85-year-old who has never owned a computer. Factor in the time your staff will spend untangling login issues.
If you’re a health IT leader or procurement officer
Build a total-cost-of-adoption template that goes beyond the contract price. Include:
- Patient-facing fees (and whether they’re one-time, recurring, or insurance-billable)
- Hardware and connectivity requirements for patients
- Staff training and workflow adjustments
- Estimated support tickets for authentication resets or feature confusion
- Accessibility accommodations (screen-reader compatibility, captioning, etc.)
Push vendors to show evidence of early-stage stakeholder engagement—not just a last-minute focus group. If the tool’s dashboard is beautiful but the patient experience is a black box, you’ll inherit the hidden costs.
If you’re a developer or investor
This study is your roadmap out of expensive rework. Start with older adults and care partners before a single line of code is frozen. That doesn’t mean asking what features they want; it means observing their tasks, mapping their constraints (income, dexterity, hearing, comfort with technology), and designing interactions that work within those constraints. Consider public-private partnerships like the Johns Hopkins Artificial Intelligence and Technology Collaboratory for Aging Research to de-risk early development. Simplify onboarding relentlessly. And if your business model depends on a $19/month subscription that half your target market can’t afford, you don’t have a product—you have a liability.
What to Do Now: Action Steps
- For patients and families: If you’re offered a health tech tool, request a trial period or demo before committing. Ask for written instructions in large print. Check if your local Area Agency on Aging or senior center can help evaluate it.
- For clinicians: Review any AI tool you prescribe with the same rigor you’d apply to a new medication: what’s the evidence, what are the side effects (in this case, user frustration and abandonment), and who shouldn’t use it.
- For health systems: Create an accessibility checklist for AI procurement. Include criteria like “co-designed with target age group,” “plain-language onboarding,” and “no recurring patient fee beyond insurance copay.”
- For developers: Adopt the “nothing about us without us” principle from disability activism. Pay older adults and care partners to participate in your design sprints. Build prototypes that assume slow internet, unsteady hands, and low digital literacy.
Outlook: A Reckoning Ahead
The Johns Hopkins study lands at a moment when health AI is facing a credibility test. Regulators are beginning to ask harder questions about real-world usability, not just algorithmic accuracy. The FDA’s Digital Health Center of Excellence has signaled interest in patient-reported outcomes. And health systems, squeezed by staffing shortages and rising costs, are becoming less tolerant of shiny pilots that don’t deliver.
The researchers’ call for early, inclusive stakeholder engagement and public-private partnerships isn’t just academic. It points to a practical future where AI tools for aging earn their place not by nailing a benchmark test, but by proving they work in the messy, daily lives of the people who rely on them. As lead author Zhang Zhang and senior author Nancy L. Schoenborn make clear, the alternative is a market flooded with technology that’s 99% accurate and 100% unusable for the people who need it most.