Seventeen years ago, I was part of a team that led the redesign of a home telemedicine system called IDEATel, a Columbia University initiative that put telemedicine units into the hands of elderly, medically underserved Medicare beneficiaries with diabetes across New York State, and I published the results of that redesign in Telemedicine and e-Health in 2009, showing that our second-generation device cut the average delay between a patient taking a glucose reading and that reading reaching their care team from 22.44 days down to 10.75 days. I consider that finding a genuine engineering success, and it is the one the paper is usually cited for, but a second finding sat inside that same dataset, and I have thought about it far longer and far harder than the first: when I broke the data down by region, patients living in rural, upstate New York uploaded their readings more often and more promptly than patients living in New York City, and that gap held regardless of which generation of device a patient used, which told me that the technology itself was never the real variable.
I have spent more than two decades working at the intersection of health information technology, clinical operations, and the populations who get left behind when new systems are designed around the patient we wish we had rather than the patient who is actually there, and the regional gap I found in IDEATel has stayed with me because it traced back to factors that had nothing to do with hardware and everything to do with the conditions of people's lives. The majority of our downstate population had between one and eleven years of formal education, compared to more than twelve years upstate, and only about a quarter of our downstate participants spoke English, compared to over ninety-five percent upstate, at a time when our case managers were not always able to match that language need. Many of our downstate participants were also grandparents serving as daytime caregivers for young children, and that responsibility competed directly with the time and attention the intervention required, so the upload delay I measured in days was really measuring the accumulated weight of education, language, and caregiving load rather than any property of the device itself.
Gen 1 vs. Gen 2 device (2009)
upstate vs. downstate (2009)
all adults offline (2026)
rural vs. urban seniors
I was not alone in noticing this pattern, and I want to credit my colleagues David Kaufman and Jimmy Pevzner, with whom I had already published a companion paper a few years earlier whose title stated the finding plainly, on redesigning a telehealth diabetes management program for a digital divide seniors population. An earlier usability study some of us worked on had already shown that standard interface conventions like mouse use, scroll bars, and font size created real barriers for older adults regardless of where they lived, and taken together, that body of work taught me a lesson I still lead with today, which is that technology access does not equalize outcomes on its own, and that education, language, caregiving load, and the continuity of human support are what actually determine whether access becomes use.
The regional gap I found upstate and downstate in 2009 never closed, and based on what I am seeing today, I expect it will keep finding new places to hide in every generation of technology that gets built without that discipline in mind.
I would like to be able to say that smartphones, broadband expansion, and wearable devices have closed this gap in the years since, but the evidence I have reviewed suggests the gap has simply relocated rather than closed. Older adults remain disproportionately offline today, and where they are online, they remain disproportionately unsupported, since Pew Research's most recent survey of American technology adoption found that ten percent of adults sixty-five and older do not use the internet at all, more than double the rate across all adults, and a peer-reviewed study examining rural versus non-rural internet access among seniors found that rural older adults have twenty-nine percent lower odds of having any internet access at all than their urban counterparts, with lower use across every category researchers measured, from communication to health technology, and with non-users significantly more likely to describe technology itself as too complicated or too hard to learn. I read that as the same usability and support gap I documented in IDEATel, simply attached to a newer generation of devices.
What has actually changed since 2009, in my view, is not the shape of this problem but the level of public investment aimed at closing it, and the direction of that change concerns me. The Affordable Connectivity Program, which subsidized home internet for lower-income households, ran out of funding in 2024 and was never renewed by Congress, and nearly two in five of the roughly twenty-three million households enrolled in it were headed by someone fifty or older. Around the same time, the federal Digital Equity Act, which had funded the kind of state-level computer skills training and device-lending programs that would have directly served populations like our downstate IDEATel participants, was also ended, so the infrastructure supporting older adults online has continued to improve in raw capability even as the public programs meant to help people actually use it have been allowed to lapse.
If I could leave any health system or payer building a remote monitoring program today with one lesson from the IDEATel redesign, it would be that the device is never the bottleneck for long, because whatever friction a team designs out of the hardware tends to resurface as a gap in language access, education, or caregiving bandwidth if those factors have not been designed for as deliberately as the device itself. In practice, that means staffing care teams with language-concordant clinicians rather than relying on translated materials, building scheduling around a patient's actual day rather than an assumed one, and preserving continuity with the same case manager over time, since my data showed that turnover in that relationship degraded engagement just as much as any hardware limitation did. I was writing about all of this in 2009, when the technology in question was a touchscreen unit that dialed the internet over a telephone line, and although the tools have changed almost beyond recognition since then, I believe the discipline required to make them equitable has not changed at all, and I doubt it ever will.
The regional gap I found upstate and downstate in 2009 never closed, and based on what I am seeing today, I expect it will keep finding new places to hide in every generation of technology that gets built without that discipline in mind.
