When the pieces leave the room, who owns the patient?
We teach the eye exam as a litany. Vision, pupils, pressure, motility, fields, slit lamp, fundus. We sort it into tiers—intermediate and comprehensive—and we bill it that way: 92002 and 92004 for new patients, 92012 and 92014 for established ones, each defined by a checklist of “elements” a payer can audit. To a trainee, the comprehensive exam looks like a clinical unit, a single indivisible act of medicine.
It was never only that. The comprehensive exam is also a bundle — a set of independent measurements held together as much by reimbursement codes and scope-of-practice law as by clinical necessity. For most of our careers, that distinction didn’t matter because there was no practical way to pull a single thread out of the bundle and run it somewhere else. Now there is. And technology isn’t breaking the exam so much as exposing its seams.
The components are walking out the door
Start with the one I’ve written about before. The retina is becoming a window into systemic disease, and the image that opens that window no longer has to be captured in our chairs. A fundus photo taken at a primary care office, a pharmacy, a lab—or eventually a grocery-store kiosk—and read by an algorithm can flag cardiovascular, renal, neurologic and metabolic risk.
Is that an eye exam? Sort of. One component of one. Is it a “full” exam? No. Is it valuable? Absolutely. And here’s the uncomfortable part: it can surface findings a comprehensive exam without AI would have missed entirely. The decomposed piece isn’t a weaker version of the whole. It’s doing a different, arguably bigger, job.
And it isn’t only the static image. As I argued in my column on functional oculomics, the next layer is behavioral—saccadic velocity, fixation stability, pupillary dynamics, captured on an ordinary webcam. That layer unbundles even more easily, because it needs no specialized hardware at all.
Refraction is unbundling now
Eyebot, a Boston startup founded in 2021, has placed self-service refraction kiosks in malls, on campuses and in grocery chains and big-box stores across several states, and in optometry offices. A user steps up to a contactless terminal, follows on-screen prompts for about ninety seconds, and a licensed doctor reviews the captured refraction, lensometry and acuity before a prescription is released. It is fast, inexpensive and for someone in a vision desert with no eye doctor within an hour, it solves a genuine problem. Notice, too, that the optometrist hasn’t disappeared. The role has simply been relocated to the other end of a data link.
What makes Eyebot the clearest case in this whole discussion is the response it drew. When the AOA issued patient-safety alerts on kiosk refraction earlier this year, a key objection was that a refraction delivered without an eye-health examination can leave patients falsely reassured about the condition of their eyes, and can quietly steer them away from the comprehensive exam that catches glaucoma, a suspicious disc or a retinal lesion before it becomes symptomatic.
That is not a reflexive objection to technology. It is a sound warning about what gets orphaned when a single component is pulled out of the bundle and sold on its own — and it is exactly the right thing to worry about. (This tension isn’t brand new. The first skirmish played out a decade ago over online refraction tests. What’s new is that the test is now a physical kiosk, standing in the store where the glasses are sold.)
Structural imaging is unbundling too
Notal Vision’s SCANLY Home OCT—granted FDA de novo authorization in 2024—lets a wet-AMD patient self-image at home in under a minute, with AI quantifying fluid and routing alerts to the retina specialist between visits. Notice what came with it: a set of new CPT codes for remote OCT review, billable on a 30-day cycle. The bundle didn’t just spring a leak. It spawned a new economic apparatus around the piece that left. That is what unbundling looks like in medicine. Not a component disappearing, but a component acquiring its own billing code, its own monitoring center, its own workflow.
The receiving specialty often isn’t eye care
The pattern repeats across the rest of the litany, and the striking part is how often the test ends up somewhere other than our chair.
- Oculogica’s EyeBOX. FDA-cleared since 2018, uses eye tracking to aid concussion diagnosis. It lives in emergency departments, neurology, sports medicine and orthopedics – not eye clinics. Saccade and pursuit abnormalities are being chased as early markers of Parkinson’s and cognitive decline. “Motility” has quietly become a neurologic vital sign.
- Visual fields. VR perimetry headsets already reproduce the function of a tabletop perimeter. It is not a stretch to expect neurologists managing MS, stroke or tumor patients to monitor fields themselves, on a headset, without routing through us.
- Pupils. Quantitative pupillometry—the handheld NPi devices—is standard kit in the neuro-ICU, where the afferent pathway is read as a brain monitor. The pupil exam left our lane years ago, and we barely noticed.
- Pressure. Home tonometry and contact-lens pressure sensors are pulling IOP measurement out of the slit-lamp bay and into the patient’s day.
And the exam isn’t only decomposing—it’s growing limbs that never existed on the classic checklist. Tear osmolarity, point-of-care inflammatory markers and molecular and genetic readouts have no line item in 92004. The framework we trained on is simultaneously being pulled apart and outgrown.
“Is it an eye exam?” is the wrong question
When a retinal photo gets taken next to the pharmacy counter, the instinct is to ask whether that counts as an eye exam. It’s the wrong question. It’s nostalgia for the bundle. The honest questions are sharper. Who reads the piece and who owns the finding? Who is accountable for the patient once the components scatter across a dozen settings that don’t talk to each other?
This is where unbundling stops being a technology story and becomes a systems problem—which is where the real failures in healthcare AI consistently live. Unbundling is genuinely good for access: it moves each test to where the patient already is. But it orphans accountability. When organized optometry sounds the alarm over kiosk refraction, this is what it is really pointing at. Not the device, but the absent owner of everything the device doesn’t measure.
A retinal biomarker flagged at a grocery store, a positive home-OCT alert, a field defect caught on a neurologist’s headset—each one needs a responsible clinician and a reimbursement path. Frequently, there is neither. The pharmacy that took the photo isn’t going to manage the cardiovascular risk it surfaced. The neurologist watching fields isn’t going to work up the incidental disc.
The quiet product of the comprehensive exam was never any single measurement. It was a single accountable owner. One clinician integrating the pieces and holding responsibility for what they added up to. That is the part that doesn’t survive unbundling on its own, and the part no kiosk, headset or algorithm currently replaces.
The exam isn’t dying. It’s dissolving.
The temptation is to defend the bundle—to insist the pieces belong together and treat every kiosk as a threat. That is a fight the regulation will lose, slowly, because the access argument is too strong and the seams are already open. The more useful posture is to design for the unbundled world: decide who reads each piece, who carries the duty to act, and who gets paid to close the loop.
The eye is becoming a sensor distributed across the entire health system—at the pharmacy, in the ICU, on a neurologist’s headset, in a patient’s kitchen. The open question for our profession is whether we architect that distributed system or get routed around it.
The comprehensive exam taught us to put the pieces together. The work ahead is making sure someone still does — after the pieces have left the room.
Author
