Telehealth after the surge: what stayed and what reverted

What is telehealth?

Telehealth is the use of electronic information and telecommunications technology to deliver clinical care, health education and health administration across distance. It is an umbrella term rather than a single product. The federal definition maintained by the Health Resources and Services Administration covers four modalities: live video, store and forward (recorded images, notes or scans sent for later review), remote patient monitoring, and mobile health.

The definition has not moved since 2020. The distribution has. Telehealth stopped working as a general substitute for the doctor’s office and settled into a narrow set of care types where it beats the alternative. Reading the current numbers as one trend is the most common mistake in this market, and it produces the wrong roadmap.

Telehealth vs telemedicine: what is the difference?

Telemedicine is a subset of telehealth. It describes remote clinical services specifically: a live, real time diagnostic or treatment encounter between a patient and a licensed clinician, held by phone or video in place of an in-person visit. Telehealth covers that plus the surrounding activity that is not itself a consultation, including clinician education, administrative meetings, patient portals and asynchronous data exchange.

The distinction is not academic. Telemedicine is the part that gets coded, billed, audited and constrained by state law. Telehealth is the part that gets funded and marketed. Products that blur the two tend to promise coverage the billing rules will not support.

What is telehealth actually used for now?

Mental health, more than anything else. It is the top telehealth diagnostic category nationally and in every region of the country, and it is not close.

Total volume is steady rather than growing. Telehealth accounted for 5.119 percent of all medical claim lines nationally in January 2025 and 5.122 percent in February, before a 3.1 percent national decrease in March 2025.¹ That is a mature channel finding its floor, not a channel in collapse.

On the supply side, adoption is wide but shallow. In 2024, 71.4 percent of physicians reported using telehealth weekly, up from 25.1 percent in 2018 and down from a 2020 peak of 79 percent.² Nearly three times the pre-pandemic share of physicians now touch telehealth in a normal week. How much of their practice runs through it is a separate question, and a claims-based read of 2024 shows the answer splitting hard by specialty.

Psychiatry sits at 85.9 percent weekly use and ophthalmology at 1.8 percent.² Neurology is next at 32.2 percent, and the middle of the table clusters tightly: endocrinology at 24.2 percent, gastroenterology at 20.4 percent, and family and general medicine at 20.1 percent.² The spread between the top and the bottom of that list is not a technology adoption curve. It is four or five different markets wearing one name.

Telehealth statistics: reading the specialty split

Patient-side data tells the same story from the other end. As of December 2025, telehealth accounted for 28.2 percent of mental health visits, 11.4 percent of endocrinology, 9.4 percent of obstetrics and 2.3 percent of urgent care.³

Primary care is the clearest case of reversion. Telehealth use there fell roughly 30 percent, from 8 percent of visits in July 2022 to under 6 percent by October 2025.³ Patients moved the same way earlier: adult telemedicine use fell from 37.0 percent in 2021 to 30.1 percent in 2022, with the steepest drop among adults aged 65 and older, from 43 percent to 31 percent.⁴

So “telehealth is declining” and “telehealth is now standard” are both true, of different things. Routine primary care went back toward its pre-2020 pattern. Outpatient mental health did not go back at all; it changed channel permanently. Any forecast, product roadmap or investment case that averages the two will be wrong in both directions at once.

Why the specialty gap is not only clinical

The convenient explanation is that some care needs hands on a patient. That is plainly true for ophthalmology and for anything procedural. It is also incomplete, because reimbursement explains a large part of the spread.

Coverage and price are separate promises. Service parity stops an insurer refusing to cover care because it was delivered remotely. Payment parity forces the same rate.⁶ Where only the first exists, the specialties carrying the highest fixed cost per encounter drop telehealth first, because a discounted virtual visit does not pay for the room, the staff and the equipment standing behind it. A telehealth product’s specialty mix is partly a map of state insurance law, and that map is redrawn every legislative session.

Which telehealth benefits hold up in the data?

The durable telehealth benefits are the ones that remove friction, not the ones that claim to improve medicine. Three hold up.

Access where geography and logistics block care. Metropolitan telehealth use ran above 6 percent in October 2025 against under 4 percent in small towns and rural areas, roughly double the rate.³ That gap runs backwards from the original rural telehealth promise, and it is a compound failure rather than a single one: broadband availability and clinician supply thin out at the same edge of the map. Widening eligibility without the connectivity underneath it does not close the gap on its own.

Removing the cost of showing up. The same research finds that patients who do not speak English over-index on telehealth adoption.³ Read that as evidence about friction rather than preference. Transport, interpreter scheduling and unpaid time off work are the real barriers, and a video visit removes all three at once. Language access belongs in the growth plan, not only in the compliance checklist.

Satisfaction, which has quietly become a payer story. Direct-to-consumer telehealth scored 730 out of 1,000 in 2024, down 1 point year on year, while payer-provided telehealth scored 708, up 18 points.⁵ The standalone apps have stopped improving. The benefits embedded inside health plans are closing the gap fast, which is where distribution is heading too.

Why telehealth policy is an operating risk, not a settled question

Federal telehealth authority still runs on temporary extensions. The DEA and HHS extended the flexibility that lets clinicians prescribe Schedule II through V controlled substances without a prior in-person exam through 31-Dec-2026, the fourth temporary extension since the pandemic-era rules were first due to expire.⁷ Medicare’s telehealth flexibilities have followed the same rhythm of lapse, short-term extension and restoration since 2023.

Any product that assumes telehealth authority is permanent has placed a policy bet on behalf of its users.

The response is architectural, not legal. Modality belongs in configuration, not in assumptions baked through scheduling, consent, prescribing and billing logic. If a flexibility lapses on a Tuesday, the affected visit types should be switchable to in-person or hybrid the same day, without a release and without a data migration. Teams that hard-coded permanence in 2021 spent the following years shipping emergency patches instead of features. Building that flexibility in early is cheap. Retrofitting it after a lapse is not.

What to build now

Capital has returned to this market with a clear preference about where it goes.

Three priorities follow from the numbers above.

Build for the mental health case first, then widen. When roughly six in ten telehealth patients carry a mental health diagnosis,¹ the workflows that matter are recurring appointment series, no-show recovery, measurement-based care and continuity with one clinician. A generic video-visit product designed for one-off acute encounters is built for the segment that reverted.

Treat compliance as an architecture decision, not a certification. Consent capture, audit logging, data residency and access control need to hold under a modality that can change by statute. Our guide to building a HIPAA compliant app covers the technical baseline, and our engineering teams treat that baseline as the starting point for regulated builds rather than a late-stage audit.

Point AI at the administrative load before the clinical decision. The funding data shows where investor conviction sits.⁸ The defensible near-term applications are documentation, intake summaries, coding support, triage routing and language access, all of which reduce the cost per encounter without moving clinical liability. That is the frame we bring to AI consulting engagements in regulated settings.

Telehealth is no longer one market with one growth curve. It is a set of channels with different economics, different legal exposure and different patients. If you are scoping a build in this space, talk to us about which of those channels your product is actually in.

Written by the group's editorial team with the practice leads who run these builds. Reviewed before publish. Spotted an error? Tell us and we will fix it.