Healthcare Technographic Data: Where to Source It, 9 Providers, and 5 Free Datasets

Here is a failure we see often enough to name. A data team buys a healthcare technographic data feed, wires it into their product, and ships it. Four months later a customer points at a row and says their own hospital is listed on the wrong electronic health record. The vendor column was not broken. It was answering a different question from the one the team thought they had asked.

We all know stale records still happen, and that is not really what goes wrong here. What goes wrong is earlier. Healthcare technographics behave differently from the horizontal kind, and teams buy them using horizontal instincts. A hospital is not one company running one system. It is a corporate parent above facilities that each carry their own federal identifier, an inpatient stack, a separate outpatient stack, and an imaging archive older than both. None of it is on a public website.

There are three things every data team should do before they spend anything on this category. Name which of the three altitudes they are actually shopping at. Build the free federal join first, because for US hospitals it answers more than most people expect. Then price the refresh rather than the record. Do those in order and the shortlist gets short quickly.

What follows is that method, the nine providers worth knowing, and the five public datasets underneath them. We pulled the federal file ourselves while writing this, all 68,446 rows of it, so the numbers below are measured rather than repeated.

What healthcare technographic data actually is, and why the obvious field is dead

Healthcare technographic data records which clinical and administrative technology a provider organisation runs, at which facility, in which version. That is the healthcare case of the same signal business teams buy horizontally, and the definition is the only part that carries over cleanly.

If you have read our explainer on firmographic versus technographic data, the split holds. Firmographics say a hospital has 400 beds inside a three-hospital system. Technographics say those beds are charted in MEDITECH Expanse while the system's two clinics run athenahealth. One sizes the opportunity. The other tells you whether you have anything to sell.

In practice, six layers carry almost all of the commercial signal, and teams routinely buy coverage of one while needing another.

Now the part that decides whether any of this is worth buying. Federal reporting shows more than 99 percent of US non-federal acute care hospitals had adopted a certified electronic health record by 2024, up from under 10 percent in 2008. On the ambulatory side the figure is 91 percent of office-based physicians. The variation between hospital types that was visible in 2010 and 2018 had disappeared entirely by 2024.

A field that is true of 99 percent of your market is a constant, not a signal. Every segment built on "has an EHR" returns everybody. The questions that still discriminate are narrower, and they are the ones worth paying for. Which vendor. Which product line. Which version. Which modules they did not buy. And when the contract comes up. That last one is the most valuable and the hardest to get, and we will come back to why.

The four places healthcare technographic data comes from

Every record in this category was produced one of four ways, and the method decides everything downstream. Coverage shape, refresh rate, which fields exist at all, and price are consequences of collection, not vendor choices. Once you can name the method behind a dataset, you can predict its failure domain without running a trial.

Regulatory disclosure is the class most teams skip. Hospitals in federal payment programmes have to declare the certified health IT they use, and the declaration is published. That produces a hospital-to-vendor mapping with an authority no commercial product can match, because it is the hospital's own filing rather than an estimate. Its limits are equally hard, and we quantify them further down.

Survey research is the only class that captures intent. A filing tells you what a hospital runs today. An analyst who spent an hour with the CIO knows the contract is up in eighteen months and the board has lost patience. That cannot be crawled, filed or inferred. Somebody has to ask.

Web detection is where healthcare technographic projects fail most confidently. Point a crawler at a health system domain and it returns a real, accurate, well-formed answer about the content management system, the analytics tags and the chat widget. It returns nothing about the electronic health record, because the record is not served over the public web. The output validates cleanly and answers a question you did not ask, which is the worst kind of wrong.

The three altitudes, and the denominator that changes every answer

This category is sold as one thing and bought for three different jobs. Naming the three is the fastest way to stop overpaying, because the most common purchasing error here is buying a market-research subscription to answer an account-level question, then discovering nine months in that no report will ever name your prospect.

Market altitude is the cheapest and the least operational. It earns its place in board decks, territory planning and roadmap decisions. It is no use to someone who needs to know what is in front of them on Thursday.

Account altitude is what most people mean when they say technographic data, and it is where free federal data competes far better than its reputation suggests.

Moment altitude is where the money is and where supply is thinnest. Knowing a system runs a given vendor is worth something. Knowing they are eleven months from a renewal decision is worth an order of magnitude more, and no catalogue sells it.

One number shows why altitude confusion is expensive. In the 2026 KLAS Research acute care report, Epic holds 43.7 percent of hospitals and 56.9 percent of beds. Oracle Health holds 21.9 percent and 20.4 percent. MEDITECH holds 14.7 percent and 12.5 percent. Epic's two figures sit 13 points apart because it is concentrated in large systems. Both are correct. Which one is right for you depends entirely on whether you sell per site or per bed.

It gets sharper. A widely cited vendor analysis puts the same three at 43.9 percent, about 19 percent and 10.7 percent. Two reputable sources, the same market, MEDITECH four points apart. Neither is lying. They count different hospitals and different moments. We are not going to pick a winner here, because the useful lesson is the one you can apply to any vendor: a coverage or share claim without a stated denominator is not a claim you can evaluate.

Worth knowing if you are timing a market entry: KLAS also found the number of hospitals affected by an EHR purchase decision fell roughly 40 percent against 2024 and nearly 50 percent against 2023, as systems deferred transitions amid policy questions and a pivot toward AI. Fewer moments to catch makes catching them worth more.

Healthcare technographic data sources at a glance

Nine providers and five public datasets, with the altitude each one genuinely serves. If you read one block on this page, read this one, then jump to the two or three rows that match your altitude.

The nine healthcare technographic data providers, compared

Grouped by the altitude each genuinely serves rather than the one it markets. Several of these are good products that will disappoint badly if bought for the wrong job.

Built to spec: when the field you need is not sold

The category has a structural hole, and it is the same hole for everyone. Renewal timing, practice-level ambulatory coverage and commercial licensing terms are three fields that often decide a deal, and no catalogue carries them. When a requirement lands in that hole, the choice is to commission the dataset or drop the requirement.

Forage AI

Forage AI is the commission-it option rather than another subscription. That distinction matters more in healthcare than in horizontal business data, because the gap between what is sold and what is needed is wider. A seller of imaging AI does not need a general hospital database. They need every facility running a named archive at a named version above a volume threshold, and that combination is not a product anybody lists.

What that looks like in delivery: on one healthcare provider engagement the team delivered more than a million provider profiles across 350,000 practices at 99.7% field-level accuracy on the golden set, cutting the client's data-collection time by 90 percent. Every delivery passes a 3x QA team, and the 200% QA approach means automated checks and human verification both run before anything lands. Person-level and facility-level healthcare records are handled under contractual controls and client-specific handling rules agreed up front.

Who it is for: teams whose target definition includes at least one field they have already failed to find on a price list, and teams needing a cadence faster than an annual regulatory file. Watch-out: a build has a lead time a download does not. If the question is simply which EHR a hospital runs, start free and escalate only when the free data fails you.

Account level: what this facility runs

These answer the question a product team asks before it ships a record. The first is healthcare-native. The rest are horizontal products with healthcare coverage of varying honesty.

Definitive Healthcare

One acquisition explains why this is the default name in the category. In January 2019 Definitive Healthcare acquired the data services business of HIMSS Analytics, including the Logic, Predict and Analyze products and the custom research practice. The HIMSS Analytics dataset was the industry reference for healthcare technology install data and IT purchasing contracts, and it was merged into Definitive Healthcare's provider data.

The practical consequence is that a dataset many buyers still request by name has not existed independently since 2019. When somebody on a team says "let us just get the HIMSS Analytics install file", they are describing something that is now a capability inside another company's platform. That is the most useful structural fact in this market, and it explains why a healthcare-native account-level file is harder to shop for than the vendor count suggests.

Who it is for: teams that need firmographics, technology installs and contacts on US providers in one place. Watch-out: ask whether the technology record sits at the facility or the parent, which is the failure mode described earlier and applies to any account-level source.

HG Insights

This is the strongest horizontal option precisely because it does not rely on web detection. It models installs and spend rather than reading page source, so it can say something about systems that never touch the public internet. For the enterprise IT layer inside a health system, meaning security, infrastructure, communications and general enterprise software, that is a real capability.

Who it is for: vendors selling enterprise technology into health systems where the buyer is the CIO rather than the chief medical information officer. Watch-out: the closer your product sits to the chart, the thinner this gets. Test it on fifty accounts where you already know the clinical answer before relying on it.

ZoomInfo

It earns a place here on breadth rather than depth. If a motion needs named contacts at thousands of healthcare organisations plus a rough technology picture, one subscription does that, and the contact data is the reason to buy it. We cover its horizontal technographic layer in more detail in our guide to B2B data enrichment tools.

Who it is for: teams whose primary need is reach, with technographics as a filter rather than the thesis. Watch-out: do not build a clinical segmentation on it. The attributes populate, they look plausible, and their provenance is not clinical.

BuiltWith

It is genuinely useful for exactly one of the six layers. If you sell patient engagement, scheduling, digital intake or healthcare marketing technology, what you sell against is on the public website, and the history lets you see what a system already tried and abandoned.

Who it is for: digital health and healthcare marketing vendors, and anyone selling to health tech companies rather than to providers. Watch-out: the failure mode is not inaccuracy, it is category error. A crawl of a hospital domain returns a clean answer about a content management system and nothing about the chart.

Wappalyzer

This is the practitioner tier of web detection. It began as a browser extension and still behaves like one, which makes it the cheapest way to put a web-layer field into your own pipeline through an API rather than a platform.

Who it is for: small teams wanting a web-stack signal inside an existing workflow. Watch-out: no coverage number is published, so there is nothing to test the claim against. Run your own account list through it and count how many resolve before building on it.

Market and moment: the shape of the market, and who is replacing what

None of these three will build a target list, and none pretends otherwise. What they carry is the context record-level files cannot: satisfaction, replacement intent, and direction of travel.

KLAS Research

This is where the numbers everyone quotes come from. The 2026 acute care report puts Epic at 43.7 percent of hospitals and 56.9 percent of beds, Oracle Health at 21.9 and 20.4 percent, MEDITECH at 14.7 and 12.5 percent. In 2025 Epic added 77 hospitals and 18,679 beds while Oracle Health lost 56 hospitals and 14,676 beds, its third consecutive year of major net losses, and Epic was the only vendor selected by health systems with more than ten hospitals.

Those movement figures are the most useful decay estimate available. Roughly 130 hospitals changed acute care EHR in a single year, before counting version migrations. On the forward view, KLAS researchers wrote that "2026 will be a critical year for Oracle Health customers, as the new AI-enabled EHR becomes a reality", noting its "success will be imperative to restoring confidence among both existing customers and the broader market." The same report found 84 percent of MEDITECH legacy customers migrating to Expanse, a reminder that running MEDITECH and running MEDITECH Expanse are different conversations.

Who it is for: product, strategy and territory planning. Watch-out: buying it to answer an account question is the altitude error described above, and it is an expensive one.

Signify Research

It answers a question the EHR-centric sources cannot. If you sell into imaging, then archives, visualisation and pathology are your market, and that layer is barely visible in EHR market-share reporting. Signify covers it as a first-class subject, and covers it globally rather than US-only.

Who it is for: imaging and diagnostics vendors, device makers, and anyone whose market the EHR analysts treat as adjacent. Watch-out: the deliverable is understanding. Budget the included analyst time deliberately, because that is where anything account-shaped comes from.

Black Book Research

It measures the dimension the others leave out. Market share tells you who won the deal. Satisfaction data tells you whether the people living with that decision would make it again, which is the leading indicator of the replacement activity that shows up in market-share reporting two years later.

Who it is for: vendors positioning against an incumbent, and provider organisations doing diligence before a selection. Watch-out: read the sample size on a category before quoting it, because survey depth varies a lot by segment and a thin segment is a thin signal.

The five free datasets, and the one that does most of the work

Before buying anything, spend an afternoon here. For a US hospital use case these five sources, joined, produce a target list with the vendor name attached, at no cost, from official filings. We ran this while writing the article, so the numbers below are measured rather than quoted. The limits are stated just as plainly.

1. The ONC Promoting Interoperability certified health IT file

This is the closest thing to an official hospital-to-vendor map, and it is a download. It exists because hospitals must declare the certified software they use to qualify under federal payment programmes, so the vendor name is the hospital's own filing rather than an analyst's estimate. It is published by ONC at healthit.gov.

The headline finding, and the reason to read this section before a procurement: deduplicated correctly, this free file reproduces the paid market-share numbers to within about two points.

The residual gap is explained by the file covering only hospitals that reported. For a team that needs the shape of the market and a list to work from, that is a subscription's headline output, reproduced in an afternoon.

Pull the file and compute vendor share correctly

import json, urllib.request
from collections import Counter, defaultdict

URL = ("https://healthit.gov/data/open-api"
       "?source=hospital-promoting-interoperability-chpl-linkage.csv")
rows = json.load(urllib.request.urlopen(URL))       # 68,446 rows, 17 columns

latest = [r for r in rows if r["year"] == "2024"]   # 35,438 rows
by_ccn = defaultdict(set)
for r in latest:                                     # Facility.ID IS the CCN
    by_ccn[r["Facility.ID"]].add(r["developer_name"].strip())

hospitals = len(by_ccn)                              # 4,627
share = Counter(d for devs in by_ccn.values() for d in devs if d)
for dev, n in share.most_common(3):
    print(f"{dev:<34} {n:>5}  {100*n/hospitals:.1f}%")
# Epic Systems Corporation        1925   41.6%
# Oracle Health                    936   20.2%
# MEDITECH                         692   15.0%

The set() on line nine is the whole article in one line. Use a list there and you reintroduce the error below.

One more thing this file quietly contains: module-level detail, for free. Certified products are named, not just vendors. Epic ships eight distinct products here, MEDITECH lists 30, Oracle Health 21. The base products separate care settings directly, with EpicCare Inpatient Base listed at 1,893 hospitals and EpicCare Ambulatory Base at 1,171, and 1,139 hospitals carrying both. Module-level targeting is usually described as unbuyable. For certified modules at hospitals, it is already public.

2. CHPL, the Certified Health IT Product List

CHPL is the product dimension the hospital file points at. The hospital file says which product a facility runs and hands you a CHPL product ID; CHPL says what that product is, which version, and what it was certified to do. Together you get vendor, product and version, which is two fields more than many commercial records carry. In the 2024 data, 34,869 of 35,438 rows carry both a CEHRT ID and a CHPL ID, so the join lands on 98 percent of rows.

A practical correction, since most write-ups skip it. The CHPL API is public but not open. We tested it while writing this: calls without credentials return HTTP 400, and calls with an invalid key return 401. Register for a free key first at chpl.healthit.gov, or your pipeline will fail on its first run for a reason the error message does not make obvious.

It is also the cleanest list of who is in the market at all, including the long ambulatory tail that never appears in market-share reporting because individually none of those products clears a reporting threshold.

3. CMS hospital files, and the CMS Certification Number that joins everything

These carry no technology information whatsoever, and they are what makes the technology information usable. Joined on the CCN, they turn "this hospital runs product X" into "this 340-bed non-profit short-term acute care hospital in this state runs product X", which is the difference between a data point and a target list.

Isolate care setting, then prepare the CCN join

inpatient = {ccn for ccn, rs in by_ccn_rows.items()
             if any("Inpatient Base" in r["product_name"] for r in rs)}   # 1,893

# CCN joins straight to CMS Hospital General Information on "Facility ID".
# Zero-pad it. CCNs are strings ("010001"), not integers. Casting to int drops
# the leading zero and silently fails the join for every hospital in a state
# with a low CMS prefix, which looks like a coverage gap rather than a bug.
targets = [{"ccn":     r["Facility.ID"].zfill(6),
            "name":    r["Facility.Name"],
            "state":   r["State"],
            "vendor":  r["developer_name"],
            "product": r["product_name"],
            "chpl_id": r["chpl_id"]}
           for r in latest if r["Facility.ID"] in inpatient]

Facility category matters more than teams expect. Short-term acute care, critical access, psychiatric, rehabilitation and long-term care hospitals all sit in these files and behave nothing like each other commercially. Filter deliberately, and use the same filter when you ask a vendor for a coverage count. We ranked the public provider sources in more detail in our guide to US healthcare provider data sources.

4. NPPES and the NPI registry

NPPES is the identity layer underneath everything else. It carries no technology field at all and is still load-bearing, because it connects a facility to the clinicians practising there and resolves organisations that appear under three different names across three files.

Watch-out: NPPES is self-reported, and deactivated records arrive as sparse rows rather than clean deletions, so a naive load produces silent gaps. We covered the tooling around this in our comparison of NPI data tools.

5. State attestation lists and state HIE connectivity reports

This is the layer almost nobody cites, and it is the only free route into ambulatory territory. Louisiana's Medicaid programme publishes the EHR products used in-state from its incentive attestation system. North Carolina's health information exchange authority publishes more than 80 EHR vendors live in production on NC HealthConnex, exportable as CSV.

Be honest about what these are. Both are usually vendor-level rather than facility-to-vendor maps, and coverage is uneven state to state. They will not replace the federal file. What they do is tell you which products have real regional presence, which is exactly the question the national market-share reports cannot answer, and they surface the small ambulatory vendors that never clear a national reporting threshold.

Every healthcare technographic source compared side by side

All fourteen sources on the six decisive columns. Altitude sits first, because it explains more disappointing purchases than price does.

The four gates, and what building it actually costs

Run these in order. Most teams start at gate four, comparing vendors, which is why the comparison feels impossible. The first three usually eliminate most of the field before anyone takes a call.

Gate 1: which altitude is the question?

Write down the sentence somebody will actually say when they use this data. If it is "the market is moving toward X", that is market altitude and an analyst subscription. If it is "Memorial in Springfield runs X", that is account altitude and a record-level file. If it is "Memorial is replacing X in the third quarter", that is moment altitude, and there are perhaps two routes to it.

A market subscription cannot be made to answer an account question by paying more for it. This is the most expensive mistake in the category and it is entirely avoidable at gate one.

Gate 2: hospitals only, or ambulatory too?

This gate decides whether free is on the table. Hospitals are federally documented. Ambulatory practices are not, at practice level, by anybody, for free. If your market is health systems and hospitals, the public join gets you a long way. If it includes physician groups, urgent care, ambulatory surgery centres or specialty clinics, you are in commercial or commissioned territory from the start, with state sources as partial relief.

In practice most teams discover their market is both, which usually means a free spine for the hospital half and a paid or built layer for the ambulatory half, rather than one product covering everything at uniform quality.

Gate 3: is the free federal join enough?

Build it before deciding, because the answer is yes more often than the market would suggest. If the requirement is the EHR vendor at US hospitals refreshed annually, the public files answer it completely, and the honest recommendation is to stop there. Escalate past this gate only for a reason you can name: you need ambulatory, you need faster than annual, you need version or satisfaction context, or you need non-EHR layers like imaging and revenue cycle.

Gate 4: does anybody actually sell the field you need?

List the required fields and mark each against the four source classes. Fields no class produces are the ones that decide buy or build. Renewal timing, practice-level ambulatory coverage and commercial licensing terms come up most, and all three sit outside the catalogue.

Here is the number worth having before that conversation. A healthcare data provider covering 6,000 hospital sites ran an RPA-class license at around $200,000 a year plus five to seven offshore engineers, roughly $400,000 to $500,000 all-in. It reached 15 to 20 percent coverage, and the crawlers kept breaking. That is not an argument against building. It is an argument for pricing the build honestly, because the license was the small half of the cost and the coverage was the part that never arrived.

When a required field is genuinely unsold, the choice is to commission it or drop the requirement, and dropping it is a legitimate answer. What does not work is buying the nearest available product and hoping the field is in there somewhere. It is not, and the discovery usually happens nine months into a subscription.

Frequently asked questions

Where can I find out which EHR a hospital uses? The free ONC Promoting Interoperability certified health IT file lists the developer and product name for every hospital that reported, keyed on the CMS Certification Number, for programme years 2023 and 2024. It is a CSV download with a JSON API and covers 4,627 hospitals in the 2024 data. For hospitals that did not report, for ambulatory practices, or for anything fresher than annual, you need a commercial source or a commissioned build.

Is there a free hospital EHR dataset, and is it any good? Yes, and better than its reputation. Deduplicated correctly it puts Epic at 41.6 percent of reporting hospitals, Oracle Health at 20.2 and MEDITECH at 15.0, against reported figures of 43.7, 21.9 and 14.7. The catch is the deduplication: the file carries 7.66 rows per hospital because hospitals attest to many certified modules, so counting one row each drops MEDITECH to 1.6 percent and promotes e-prescribing vendors into looking like EHRs.

What is the difference between healthcare technographic and firmographic data? Firmographic data describes the organisation: beds, ownership, system affiliation, location. Technographic data describes what it runs: EHR vendor, imaging platform, revenue cycle system, version and modules. In healthcare the two are unusually dependent on each other, because a technographic record is only actionable once attached to a specific facility identifier. Our explainer on firmographic versus technographic data covers the general distinction, and our guide to healthcare data providers covers the wider provider, claims and clinical data market this sits inside.

Which EHR has the largest market share? Epic, on every measure, and the spread between measures is the point. The 2026 KLAS report puts Epic at 43.7 percent of hospitals and 56.9 percent of beds, Oracle Health at 21.9 and 20.4, MEDITECH at 14.7 and 12.5. Epic's 13-point spread reflects concentration in large systems, so the right figure depends on whether you sell per site or per bed. A separate widely cited analysis puts the three at 43.9, about 19 and 10.7 percent, which is a useful reminder to ask what any share number is a share of.

Can BuiltWith or Wappalyzer tell me a hospital's EHR? No, and it is worth being blunt because several ranking pages imply otherwise. Web detection reads what a public website serves to a browser. An electronic health record is an internal clinical system behind the firewall and never appears in a page scan. These tools return confident, accurate answers about a hospital's content management system, analytics and chat widget, which is genuinely useful if you sell patient engagement or healthcare marketing technology, and is not an answer about the chart.

How fast does healthcare technographic data go stale? Faster than the annual cadence of the free files. Epic added 77 hospitals and 18,679 beds in 2025 while Oracle Health lost 56 hospitals and 14,676 beds, so roughly 130 hospitals changed acute care EHR in one year. Version changes move considerably more, with 84 percent of MEDITECH legacy customers migrating to Expanse. Treat any file older than a year as wrong about a predictable share of your list.

Is there a free source for ambulatory or physician-practice technology? Not at practice level nationally. The federal Promoting Interoperability file is hospitals only, CHPL lists every certified ambulatory product without saying who uses it, and NPPES identifies practices with no technology field. State Medicaid attestation lists and state health information exchange connectivity reports are the partial exception, and they are usually vendor-level rather than facility-level. This remains the largest genuine gap in the category.

Why can I not buy contract renewal dates? Because no single party both knows them and will publish them. Hospitals do not disclose contract terms in federal filings, vendors treat them as confidential, and no regulation forces them into the open. What exists instead is inference: procurement and board records for public institutions, analyst conversations that surface replacement intent, hiring signals and public statements. That is why the moment altitude is served by research subscriptions and commissioned collection rather than by a database, and why it stays the most valuable field in the category.

Can I still get the HIMSS Analytics install file? Not as an independent product. The HIMSS Analytics data services business, including Logic, Predict and Analyze, was acquired in January 2019 and its technology install and IT contract data was merged into Definitive Healthcare's provider data. Buyers still ask for it by name, which is worth knowing when somebody on your team proposes it as an option.

What this looks like a year in

A team that runs this method for a year ends up somewhere specific. They hold their own denominator, rebuilt each time the federal file updates, so every vendor claim arrives pre-scored. They know which accounts the free data answers and buy only the gap, which tends to mean smaller contracts that survive renewal. And they have stopped treating the file as a purchase and started treating it as a pipeline with a refresh cost, which is the shift that makes the budget defensible.

The unsolved part stays unsolved, and we would rather say so. Nobody has a good public answer for renewal timing or for practice-level ambulatory coverage. Those are the two fields that would change how this category works, and right now they are assembled by hand or not at all. If your team has found a route into either that holds up at scale, that is a conversation worth having, and it is the part of this problem we are still working on ourselves.