The photo arrived blurry and overexposed, taken under a single, flickering bulb. A nurse typed: “Fever, 3 days. No transport today.” The clinic’s generator had enough fuel for two hours, maybe three if they didn’t run the refrigerator too hard. A video call was never going to happen. But a decision still had to.
That’s the version of telehealth I think about most often. Not the glossy demo of a doctor smiling into a webcam, but the unglamorous work of moving clinical decisions across distance when everything around it is fragile.
It’s tempting to tell a simple story: telehealth was niche, then COVID arrived, then everyone adopted it. The truth is longer and less flattering. Telehealth has been tried for decades.
It succeeded only when four conditions lined up:
Bandwidth - can the connection carry anything usable?
Billing - will the system pay for the work in a predictable way?
Belief - do clinicians and patients trust it enough to use it?
Back office - can it fit the real workflow (documentation, follow-up, labs, referrals) without doubling the burden?
I’ll use two terms carefully:
Telemedicine is clinical care at a distance. Diagnosis and treatment delivered remotely.
Telehealth is the broader system. Telemedicine plus triage, remote monitoring, patient communication, and operational back office.
So how did we get here?
Telehealth didn’t become critical infrastructure because video got better. It became infrastructure when it started to fit the systems that had to carry it. And when those systems, under stress, stopped treating distance as a special case.
I. The Prehistory (1960s–1990s): Telemedicine Before the Internet
Telemedicine began as a serious experiment funded by organizations that could afford serious experiments.
In the 1960s, Massachusetts General Hospital linked clinicians to Boston’s Logan Airport using telecommunications and later an interactive TV microwave link with tools like stethoscope and electrocardiograph capabilities. This was an early attempt to move urgent evaluation across a short (but consequential) distance. Around the same time, NASA pushed remote biomedical monitoring forward for spaceflight, and then brought that mindset back to Earth through programs like STARPAHC (Space Technology Applied to Rural Papago Advanced Health Care), developed with the Indian Health Service and the (now) Tohono O’odham Nation to extend care into a remote reservation using communications technology and a mobile health unit. The military also explored telemedicine as operational support. U.S. Army telemedicine efforts in the early 1990s helped normalize remote consultation inside constrained, high-stakes environments.
On paper, it looked like the future. In practice, it was often structurally doomed.
The technology worked (sometimes) but bandwidth was expensive and brittle. Systems were specialized enough that one broken part could cancel a clinical day. In STARPAHC, providers reported major problems like unreliable equipment and the time burden of TV consultations. Cost was not just equipment; it was maintenance, training, and the hidden price of asking clinicians to work differently without reshaping the rest of the process.
The deeper mismatch was workflow. Telemedicine could transmit a conversation, but it rarely carried the surrounding system: medication access, diagnostic follow-up, documentation, scheduling, and accountability. Even in the mid‑1990s, reviewers noted that most early programs failed to survive once grant funding ended and that expensive broadband video often wasn’t justified when cheaper, more reliable channels could do the job.
A counterexample matters here: teleradiology gained traction earlier than many other telemedicine forms because it fit existing professional norms and payment realities better than live video consults did. It didn’t require a shared room at the same time, and reimbursement for radiology interpretation already had pathways that didn’t demand face-to-face contact.
The lesson from this era is blunt: telemedicine solved narrow problems, but it didn’t fix the systems that made those gaps routine. Bandwidth existed in pockets, but billing, belief, and back office were still missing.
II. The First Internet Wave (2000–2010): Telehealth as a Feature
The commercial internet made a basic form of telehealth possible for ordinary clinics: web portals, email follow-ups, and early video consults. It did not make telehealth easy to adopt.
In the U.S., billing was the decisive limiter. Medicare telehealth coverage under the Physician Fee Schedule began in 2001 and was largely framed as a rural access exception, which often required patients to be at an approved originating site in a rural area rather than at home. Payment rules emphasized live, interactive telecommunications, with limited allowances for asynchronous (“store and forward”) use in specific demonstration contexts like Alaska and Hawaii. The signal was unmistakable: telehealth was permissible on the margins, but not yet a default delivery channel.
Physician behavior also slowed adoption. Most clinicians didn’t wake up wanting to add a new visit type. They were already overloaded, and telehealth often meant extra, burdensome steps. In many cases it meant separate scheduling, awkward documentation, uncertainty about malpractice exposure, and licensure constraints when patients crossed state lines.
Patients had their own “belief” barriers. In 2003, a choppy webcam call from a desktop computer felt less like care and more like a tech support session. Trust is a clinical ingredient. Then, it was in short supply.
Regulation added friction where it mattered most. The Ryan Haight Act of 2008 generally required at least one in‑person medical evaluation before prescribing controlled substances via the internet, with specified telemedicine exceptions. That law was responding to real harms, but it also reinforced a broader reality that certain high-risk clinical actions would remain tightly tethered to in-person norms.
Conversely, large integrated systems and public institutions could make telehealth work internally because they could subsidize it and embed it inside existing care pathways. The average independent practice and early telehealth startup could not.
The internet improved bandwidth and lowered hardware costs, but telehealth remained mostly an add-on. It was useful one the edge, but rarely central. In this decade, telehealth wasn’t blocked by a lack of video. It was blocked by reimbursement, workflow fit, and trust.
III. Smartphones Change the Equation (2010–2015)
Smartphones did something the first internet wave couldn’t. They made the endpoint universal.
By the end of 2010, the ITU estimated 5.3 billion mobile cellular subscriptions globally, including 940 million 3G subscriptions, with mobile network access available to about 90% of the world population (and 80% of rural populations). By 2015, the ITU reported more than 7 billion mobile subscriptions and 3.2 billion internet users worldwide, with roughly 2 billion in developing countries.
That shift changed “bandwidth” from a clinic problem to a population feature. It also changed user expectations: people stopped needing to learn “telehealth.” They already knew how to use a camera, send a message, or share a photo.
Cloud infrastructure mattered too. It became cheaper to run scheduling, messaging, and data storage without installing bespoke systems at every site. In high-income settings, telehealth could now augment mature healthcare infrastructure which included labs, pharmacies, payer systems, and credentialing. In emerging markets, the order was often reversed.
I witnissed firsthand how connectivity typically appeared before healthcare infrastructure. A village could have mobile signal and cheap Android devices long before it had reliable clinic hours, stable medication supply, or enough clinicians. Telehealth suddenly looked like a shortcut to care. But the shortcut still had to connect to something real.
However, one counterexample is instructive: many early mobile health efforts assumed constant connectivity, consistent device storage, and stable patient identity. They failed not because the idea was wrong, but because reliability (power, data, staffing) was still missing.
The key shift from 2010–2015 was that access became possible at scale. The hard truth was that possible is not the same as dependable. Smartphones solved the endpoint, but they didn’t solve the system.
IV. Building in the Real World (2013–2019): Telehealth Outside Ideal Conditions
For years, I worked on telehealth deployments in places where the “ideal conditions” assumed by many designs simply did not exist. Doctors were scarce. Clinics were unreliable. Paper systems dominated. And the ordinary failures of infrastructure were part of the care environment.
This is where typical assumptions quietly break.
Many telehealth models assume that care is delivered through predictable appointments, stable patient IDs, electronic records, and a pharmacy and lab network that reliably completes the plan. In the field, phone numbers change. People share devices. Names are spelled three different ways in three notebooks. Transport collapses the schedule. Backorders turn prescriptions into suggestions.
So telehealth had to become more boring and more operational.
We treated uptime like a clinical metric. If a system is down during the two hours a clinic has power, it may as well not exist. That forced an obsession with unglamorous engineering, such as offline queues, retries, local caching, and fallbacks to voice calls or SMS when data disappeared.
We also learned quickly that live video was a luxury. Asynchronous “store and forward” flows, which included structured history, photos, and vitals collected locally and reviewed later were the workhorse. For this same reason, programs like Alaska’s AFHCAN leaned on asynchronous consultation. It fits unreliable data links.
Many used to think video quality would decide whether telehealth worked. In practice, the decisive variable was clinician time. Anything that added even two extra steps (re-entering notes, struggling with logins, chasing a missing identifier, etc…) quietly killed adoption. The best telehealth systems didn’t feel like “remote care.” They felt like less friction around the same care.
Trust was not automatic either. In many settings, a local nurse’s presence carried more legitimacy than a remote physician on a screen. We kept local caregivers central and used telehealth to extend them, not replace them.
I watched more than one program fail after being designed like a policy presentation. Some implementations assumed stable workflows, constant internet, and a trained workforce ready to change. Telehealth didn’t fail because it was impossible. It failed because it didn’t match the environment.
By 2019, the lesson was clear. Telehealth works when it adapts to reality, not the reality of a funding proposal, but the reality of power cuts, staffing gaps, and human trust.
V. COVID as an Accelerant (2020–2021)
Many people first encountered telehealth during COVID. However, COVID didn’t invent telehealth. It simply removed the option to keep treating it as optional.
On March 17, 2020, CMS announced it was expanding Medicare’s telehealth benefits using emergency authority (including 1135 waiver authority), waiving key limitations and allowing many beneficiaries to receive telehealth services in their homes. Around the same time, HHS’s Office for Civil Rights issued enforcement discretion so clinicians could use certain non-public-facing audio/video tools in good faith without facing HIPAA penalties during the public health emergency.
Those were structural moves. They changed billing and reduced compliance friction quickly.
The utilization jump was enormous. One HHS/ASPE analysis described a 63-fold increase in Medicare fee-for-service telehealth visits from ~840,000 in 2019 to 52.7 million in 2020. The same analysis highlighted how telehealth concentrated differently across specialties. In 2020, telehealth visits made up roughly a third of behavioral health specialist visits, versus 8% for primary care and 3% for other specialists.
But it’s important to be honest about what changed and what didn’t.
Permission, payment, and cultural legitimacy changed. Clinicians and patients suddenly shared a reason to avoid waiting rooms. A telehealth visit stopped feeling like a novelty and started feeling like a responsible substitute.
However, fragmentation, workforce shortages, and inequities in connectivity didn’t change. Many programs were improvised under pressure. Audio-only visits filled gaps where video failed. “Telehealth” often meant “we found a way to talk to you,” not “we redesigned care.”
All that progress aside, much care that required hands-on exams, procedures, labs, or imaging did not magically become virtual. Telehealth substituted for some care and preserved continuity for many patients, but it did not replace the physical system.
COVID made telehealth widespread. The post-COVID era would decide whether it became integrated.
VI. The Post-COVID Sorting (2022–2024): What Survived
After the emergency peak, telehealth entered a sorting period. Temporary adoption is not the same as structural integration.
The U.S. COVID‑19 public health emergency expired on May 11, 2023. Yet many Medicare telehealth flexibilities outlived the PHE because Congress extended them. CMS guidance notes that Section 4113 of the Consolidated Appropriations Act, 2023 extended many Medicare telehealth flexibilities through December 31, 2024, while making some provisions permanent.
Regulators also tightened their focus on risk. The HHS Office of Inspector General reported that in the first pandemic year, more than 28 million Medicare beneficiaries used telehealth, and it flagged billing patterns that raised program integrity concerns. This was an early preview of why “telehealth at scale”, despite the convenience, would bring scrutiny.
Controlled substance prescribing became a flashpoint. DEA (jointly with HHS) extended COVID-era telemedicine flexibilities for prescribing controlled medications through December 31, 2024, citing the need to avoid care disruption while permanent rules were developed.
Meanwhile, the market corrected. Some companies built for pandemic conditions struggled when demand normalized and reimbursement uncertainty returned. Babylon Health, for example, filed for Chapter 7 bankruptcy in the U.S. in August 2023. Teladoc, a major player, recorded large impairment charges in 2022 as expectations reset for digital health assets.
What survived was telling:
Mental health, where conversation is often the core intervention and telehealth reduces travel friction.
Chronic care follow-up and condition management, where outcomes can be maintained without constant in-person visits.
Triage and navigation, directing people to the right level of care instead of defaulting to the emergency department.
Hybrid care, where telehealth is the front door and in-person care handles exams, imaging, and procedures.
Peer-reviewed studies before and after the pandemic support the idea that telemedicine can be noninferior to in-person care for certain conditions (and useful in chronic disease management), when the pathway is designed appropriately.
Here’s the infrastructure test many keep coming back to: a thing becomes infrastructure when downtime is treated as a failure of care, not a product bug. By 2024, the surviving telehealth models were the ones that behaved like that. They were embedded, accountable, and operational.
VII. Telehealth Today: Mature, Boring, and Essential
Telehealth today is not a single product. It’s a set of channels that, when well-integrated, forms an access layer for care: phone, video, asynchronous messaging, and remote monitoring.
In plain terms, telehealth now acts like a load balancer for healthcare systems. It keeps the system from collapsing by moving the right problems to the cheapest safe channel. Some problems need a room, a hand on a patient, or a lab. Many don’t. Telehealth’s power is in handling the “many don’t” without breaking continuity.
It’s also a force multiplier for clinicians, but not in the science-fiction sense. It multiplies clinician reach by reducing wasted motion: unnecessary travel, avoidable in-person check-ins, and administrative dead time between steps of care.
Policy has not fully caught up, but it has moved. CMS has continued to track telehealth use and to specify which flexibilities are time-limited versus permanent, reflecting that telehealth is now part of the operational baseline, not an edge case.
AI is beginning to integrate into telehealth workflows in practical, unglamorous ways. For example, it can assist with drafting notes, summarizing visits, translating language, and sorting patient-submitted information. More “reduce paperwork” than “replace clinicians.”
And plenty still breaks:
Reimbursement remains uneven and time-limited in key areas, creating planning risk.
Incentives are misaligned: some systems still lose revenue when care shifts away from billable in-person encounters.
Fragmentation persists: telehealth can become yet another silo if it isn’t integrated with labs, referrals, and records.
Licensure still complicates cross-state practice, even as tools like the Interstate Medical Licensure Compact offer an expedited pathway for some physicians.
A clear-eyed definition helps.
Telehealth today is: a distribution system for care including triage, follow-up, monitoring, and communication sembedded into real pathways.
Telehealth is not: a replacement for physical exams, a magic-bullet cure for clinician shortages, or a shortcut around broken payment models.
Telehealth is mature now, which mostly means it’s less exciting to talk about, and even more necessary to keep running.
VIII. What the Next Decade May Reward
The next decade won’t reward the most impressive telehealth ideas. It will reward the most reliable ones.
What lasts will be defined by reliability, integration and durability.
The future belongs to operators, not pitch decks. Healthcare is a chain, and telehealth is only one link. The work is making the whole chain holds.
Telehealth didn’t become real because the technology finally arrived. The technology arrived many times. Telehealth became real when systems (pushed by necessity, enabled by networks, and legitimized by policy) started treating distance as an ordinary condition rather than a special exception.
And if there’s a final lesson worth keeping, it’s the one embedded in that grainy photo from a flickering clinic: in healthcare, the systems that last will always outcompete the ideas that merely impress.