Telemedicine statistics shape product plans, investor decks and capacity models, yet many figures in circulation have no traceable source. A number copied from a vendor survey can end up driving a staffing plan or a revenue forecast.
Federal agencies publish usage data with documented methods, and those sources answer different questions. This guide explains what the National Center for Health Statistics, the HHS Office of the Assistant Secretary for Planning and Evaluation and CMS each measure, why their figures do not match, and how a product team should use them alongside its own platform data.
Where do reliable telemedicine statistics come from?
Planning a telemedicine product?
Agnotic builds telemedicine platforms for the client, with the event data needed to measure real usage from the first visit.
Reliable telemedicine statistics come from federal sources with published methods: NCHS surveys of the US population, ASPE analyses of national survey data, and CMS reports built from Medicare claims. Each one measures a different population, time window and definition of telemedicine, so a figure is only useful alongside those details.
Surveys ask people what they did. Claims record what was billed. Neither one tells a product team how its own patients will behave, but together they set a defensible starting range for planning.
What does each federal source measure?
The three federal sources cover different ground, and the right one depends on the question. A consumer app needs population survey data; a Medicare-focused service needs claims data.
NCHS runs the National Health Interview Survey. NCHS Data Brief No. 445 (October 2022) reported that 37.0% of adults used telemedicine in the past 12 months in 2021, counting an appointment with a doctor, nurse or other health professional by video or by phone.
ASPE analyzed the Census Bureau's Household Pulse Survey. Its issue brief HP-2023-09 (April 2023) found that 22.5% of adults who answered the telehealth question reported using telehealth, audio-only or video, within the last four weeks, across responses from 14 April 2021 to 8 August 2022.
CMS publishes Medicare Telehealth Trends data on data.cms.gov, built from Medicare Fee-for-Service Part B claims and enrollment information. CMS updates it several times a year and labels the data preliminary, and it does not cover Medicare Advantage enrollees.
Other sources fill gaps the federal ones leave. State Medicaid agencies, state all-payer claims databases and health system research teams publish telehealth analyses for their own populations. Treat them like the federal sources: find the method note, the population and the window before using any figure.
Peer-reviewed studies are useful for specific questions, such as no-show rates in one specialty or follow-up visits after a video consult. Read the setting carefully, because a result from one academic center may not carry over to a rural clinic or a direct-to-consumer service.
Why do two reputable figures disagree?
The NCHS and ASPE figures differ because they measure different things, not because one is wrong. Reading the method notes explains the gap within a few minutes.
- Time window: NCHS asked about the past 12 months, ASPE about the last four weeks. A longer window catches more occasional users.
- Survey design: The two figures come from different surveys with different sampling, question wording and response modes.
- Period: The NCHS figure covers 2021, while the ASPE window runs into 2022, so the two describe overlapping but different phases of telehealth use.
- Definition: Both count phone and video, but other sources may count only video, or include patient portal messages, which changes results again.
- Population: Adults only in both cases. Figures for children, Medicare enrollees or a single state need their own source.
The lesson for planning is to quote a figure with its window, source, year and definition. A slide that says only a percentage invites the wrong comparison.
How should a product team use telemedicine data?
Use federal data to bound assumptions, then replace those assumptions with your own measurements as soon as patients use the service. The steps below keep the two kinds of data separate.
- 1. Write the question: State exactly what the plan needs, such as the share of a clinic's Medicare patients likely to book a video follow-up.
- 2. Match the population: Choose the source whose population matches the question: Medicare claims for a Medicare service, national surveys for consumer products.
- 3. Match the window: Convert the question to the source's time window, or pick a source whose window fits.
- 4. Segment: Use the source's breakdowns by age, region or rural and urban location when the target group differs from the national average.
- 5. Check recency: Note the data year and whether the release is preliminary, and plan to refresh the figure when a new release appears.
- 6. Express a range: Turn the figure into low, middle and high cases, since one survey figure cannot predict one clinic's uptake.
- 7. Log the assumption: Record the source, link, table and date in an assumptions log so anyone can trace the number later.
Step 6 matters most for capacity planning. Clinician schedules built on the middle case can be adjusted, while schedules built on a single optimistic figure leave patients waiting or clinicians idle.
Which breakdowns change design decisions?
Breakdowns by group can be more useful to a build team than the headline figure. They show which patients need which features, and they turn statistics into design requirements.
- Modality by group: The ASPE brief reported differences between groups in audio-only versus video use. For a product team, that makes a reliable audio path and a clean switch from video to audio core features, not extras.
- Age: Survey breakdowns by age inform onboarding design, such as larger controls, fewer steps to join and support for a family member who helps.
- Urban and rural location: Differences by location point to bandwidth limits, so test the visit at low upstream speeds and keep the join page light.
- Insurance: Medicare, Medicaid and commercial coverage differ in what is paid for, which affects which visit types the product offers to which patients.
- Language: Survey and state data on language use help decide which languages the booking flow, reminders and interpreter requests support first.
Record which breakdown supported each design choice. When a later release of the data changes the picture, the team can see which features to revisit.
Questions to ask of any telemedicine statistic
Some telemedicine figures in market reports and articles cannot be traced to a source. Run each one through these questions before it enters a plan.
- Who published it: A federal agency or peer-reviewed study, or a vendor with a product to sell?
- What was the method: Survey, claims or platform logs, with a published sample size and question wording?
- What counts as telemedicine: Video only, phone, messaging, remote monitoring, or all of them?
- Which population: Adults, children, Medicare enrollees, one health system or one country?
- Which period: The emergency years, a later year, or a projection?
- What is the denominator: All adults, adults with a visit, or visits rather than people?
- Can you find the table: If the original table cannot be located, leave the figure out.
Market size forecasts deserve extra care. They can combine many inputs with undisclosed assumptions, so they fit an investor narrative better than an operating plan.
What should your own platform measure?
Your own platform data is the most relevant telemedicine statistic you will ever have, because it describes your patients and your workflow. Design the event model before launch so the numbers are trustworthy from day one.
- Booking events: Visit requested, scheduled, rescheduled and cancelled, each with channel and lead time.
- Join events: Patient and clinician join times, waiting room time and failed join attempts with error codes.
- Modality events: Video start, switch to audio-only and reconnects, which also feed claim coding.
- Completion: Visit ended, note signed and follow-up booked, so completed care is counted rather than started calls.
- Outcome links: Whether an in-person visit, an emergency visit or a repeat telemedicine visit followed within a defined period.
- Segments: Age band, location type and language, stored without names so analytics stays separate from clinical data.
Define each metric in writing, with its formula and owner, before anyone builds a dashboard. A no-show rate that one team calculates from bookings and another from confirmed visits will never agree.
Keep identifiers out of the analytics store. Event data that names patients turns a reporting tool into another system holding protected health information.
How should figures appear in a plan or pitch?
Every figure in a plan or pitch should be traceable in one step. Readers who check a number and cannot find it stop trusting the rest of the document.
- Cite beside the number: Put the source name, year and window next to the figure, not in an appendix.
- Separate measured and assumed: Mark which numbers come from published data, which from your platform and which are assumptions.
- Show the range: Present low, middle and high cases for any forecast that depends on uptake.
- Date the deck: State when each figure was last checked, and refresh it before each new round of review.
Once the service is live, put your own measured results beside the federal benchmark, with the difference in population and window explained. That comparison is more persuasive than any national figure on its own.
How Agnotic can help
Good telemedicine numbers start with instrumentation built into the visit flow. Agnotic Technologies builds that for the client: booking, join, modality and completion events, a de-identified analytics store, and dashboards with written metric definitions, so leadership can compare its own results with federal benchmarks.
Agnotic can also review an existing business case and trace each telemedicine figure back to its source table before it reaches an investor or board.
Measure your telemedicine service
Review telemedicine platform development, or find more interoperability articles in the HL7 hub.
Related reading on Agnotic
Continue with telemedicine requirements, the policy outlook, telemedicine benefits and healthcare data visualization.
