EHR (Electronic Health Record): What it is, How the Market Works, and What Actually Changed

An emergency physician makes roughly 4,000 mouse clicks over a ten-hour shift. Almost none of them involve touching a patient.

That finding, from a study of ED workflow published more than a decade ago, has become the shorthand for a larger truth about the electronic health record: it is simultaneously the most consequential piece of software in American healthcare and the most resented. It made prescribing safer, records portable, and population health measurable. It also became the primary reason clinicians describe their work as data entry with intermittent patient contact.

Both things are true. Most writing on EHRs picks one and ignores the other.

This guide covers what an EHR is, how it differs from an EMR, how the United States ended up with near-universal adoption in about a decade, who the vendors are and why market concentration matters, what certification and information blocking rules require, what implementation actually costs, why documentation burden has proven so stubborn, and what ambient AI is and isn’t changing.

What is an EHR?

An electronic health record is a digital, longitudinal record of a patient’s health information, maintained by clinicians and designed to be shared across the organizations involved in that patient’s care.

That last clause is the whole point, and it’s the part that distinguishes an EHR from its predecessor.

An EHR is not just a chart. In practice it’s the operational system of a healthcare organization: it holds clinical documentation, but it also drives ordering, results routing, medication administration, scheduling, billing, quality reporting, and increasingly patient communication.

EHR vs EMR: Is There Actually a Difference?

Yes, though the ehr vs emr distinction is more useful conceptually than in daily conversation.

An EMR (electronic medical record) is the digital version of a single practice’s paper chart. It lives inside one organization and was built to serve that organization’s clinicians.
An EHR (electronic health record) is designed to travel. It assumes the patient will be seen elsewhere and that the record needs to move with them.

The practical test is whether the record was built to leave the building. An EMR is a filing cabinet that happens to be digital. An EHR is meant to participate in exchange.

In everyday use, most people say EHR for everything, most vendors market EHR regardless of architecture, and federal policy uses EHR. If someone insists on the distinction mid-meeting, they’re usually making an interoperability point worth hearing.

What an EHR Actually Does

The functional scope traces back to the core capabilities the Institute of Medicine outlined in the early 2000s, and they’ve held up well:

Health information and data. Problems, medications, allergies, demographics, notes, vitals.
Results management. Labs, imaging, pathology, delivered and tracked rather than faxed.
Order entry. Computerized provider order entry, or CPOE, including e-prescribing.
Clinical decision support. Drug interaction checks, allergy alerts, order sets, best practice advisories.
Electronic communication and connectivity. Messaging between clinicians, exchange with outside organizations, patient portal.
Patient support. Portal access, education materials, self-scheduling, results release.
Administrative processes. Scheduling, registration, eligibility, coding, revenue cycle.
Reporting and population health. Quality measures, registries, public health reporting.

CPOE and e-prescribing are where the EHR’s clearest wins live. Illegible handwriting, transcription errors, and unchecked drug interactions were killing people, and computerized ordering with decision support measurably reduced that. This is worth remembering when the conversation turns critical, because it usually skips over the part that worked.

How the United States Got Near-Universal Adoption

In 2008, fewer than one in ten non-federal acute care hospitals had even a basic electronic health record. Adoption had been creeping along for two decades.

Then Congress spent about $35 billion on it.

The HITECH Act, enacted in 2009 as part of the American Recovery and Reinvestment Act, created incentive payments for eligible hospitals and clinicians who adopted certified EHR technology and demonstrated Meaningful Use of it. Later, incentives turned into penalties for those who didn’t.

It worked, in the narrow sense. By the early 2020s, federal data showed roughly 96 percent of non-federal acute care hospitals and about 78 percent of office-based physicians using certified health IT. Few federal technology programs have moved an industry that fast.

It also produced consequences that shaped everything after:

Certification requirements defined the product. Vendors built to the criteria, which meant features arrived because they were required rather than because clinicians wanted them.
Documentation requirements multiplied. Meaningful Use attestation, and the billing documentation rules layered on top, made the note into a compliance artifact rather than a clinical communication.
The rush was real. Organizations implemented on incentive deadlines rather than readiness timelines, and some of those decisions are still being lived with.
Interoperability was assumed rather than required. Adoption was incentivized before exchange was, which is why the following fifteen years were spent trying to make these systems talk.

Meaningful Use evolved into Advancing Care Information and then the Promoting Interoperability programs that continue today under MIPS and the hospital program.

The EHR Market: Who the Vendors Are

Market structure matters more than most buyers appreciate, because it determines your negotiating position, your integration options, and how much of your roadmap is really yours.

Acute Care Hospitals

Epic Systems is the clear leader and has been gaining share for years, with a share of the US acute care market that industry analysts have placed at roughly 40 percent or higher and climbing. Privately held, famously opinionated, and known for a single integrated codebase rather than acquisitions.
Oracle Health, formerly Cerner, is the second-largest and has been losing share. Oracle acquired Cerner in 2022 for approximately $28 billion and has since been rebuilding the product on its cloud infrastructure with AI embedded.
MEDITECH holds a meaningful share, particularly among community and rural hospitals, with its cloud-based Expanse platform.
Altera Digital Health (the former Allscripts hospital business, now under Harris), TruBridge (formerly CPSI), and others serve smaller and critical access facilities.

Ambulatory and Specialty

The ambulatory market is far more fragmented: Epic and Oracle Health at the enterprise end, then athenahealth, eClinicalWorks, Veradigm, NextGen, and a growing set of newer cloud-native platforms including Elation and Canvas aimed at independent and value-based practices.

Specialty-specific systems remain viable in ophthalmology, dermatology, oncology, behavioral health, and dentistry, where general-purpose workflows fit badly.

Why Consolidation Matters to You

Concentration changes the buyer’s position. A few practical consequences worth naming:

Your data model is the vendor’s. Analytics built directly on vendor table structures couple your reporting to their release cycle.
Integration options are gated. Third-party apps reach your clinicians through the vendor’s app program, on the vendor’s terms.
Switching costs are enormous. Migration is a multi-year, nine-figure undertaking for a large system, which limits how much leverage you actually have at renewal.
Network effects favor the leader. When most referral partners in a region run the same system, exchange gets easier inside that network, which is both genuinely useful and a competitive moat.

None of this argues against picking the market leader. It argues for going in clear-eyed about what you’re buying.

How EHRs Are Built

The architecture underneath these systems explains a surprising amount about how they behave.

Epic runs on Chronicles, a hierarchical database built on MUMPS, a language designed for medical record keeping in the 1960s. It is fast, extremely reliable at transactional workloads, and unlike anything a modern developer expects. Epic exposes analytics through Clarity, a normalized relational extract, and Caboodle, a dimensional warehouse.

Oracle Health’s Millennium platform is more conventionally relational, and Oracle’s strategy has centered on moving it to cloud infrastructure with AI capabilities built in.

Two architectural realities that affect everyone:

The transactional system is not the analytics system. Every major EHR ships separate reporting structures because the database optimized for a clinician saving a note is the wrong shape for population queries. This is why your healthcare data model work exists as a separate discipline.
Modularity arrived late and partially. FHIR and SMART on FHIR made it possible for third-party applications to run inside the EHR with the right patient in context. That’s a real change from the monolithic era, but write access remains limited and vendor app programs control distribution.

EHR Regulation, Certification, and Information Blocking

This is where the fresh material lives, and where most EHR explainers are years out of date.

Certification

Certified EHR technology must meet criteria set by the Assistant Secretary for Technology Policy and the Office of the National Coordinator for Health IT, now referred to as ASTP/ONC following a 2024 reorganization. Certification matters because participation in Medicare quality programs generally requires certified health IT.

Recent requirements worth knowing:

The standardized API criterion requires certified systems to expose a FHIR R4 API conforming to the US Core Implementation Guide, with SMART App Launch authorization and Bulk Data Access. This is what makes third-party app connectivity possible at all.
USCDI v3 became the required data element baseline for certified health IT as of January 1, 2026, replacing earlier versions.
The Decision Support Interventions criterion, introduced in the HTI-1 final rule, replaced the older clinical decision support requirement and added transparency obligations for predictive decision support. Developers must disclose defined “source attributes” describing how a predictive model was developed, validated, and maintained. This is the first meaningful federal transparency requirement for AI embedded in clinical software, and it applies whether the model came from the vendor or was built in-house on certified technology.

Information Blocking

The 21st Century Cures Act prohibits information blocking, meaning practices that unreasonably interfere with the access, exchange, or use of electronic health information. It applies to providers, developers of certified health IT, and health information networks and exchanges.

Two enforcement realities:

Developers, HIEs, and HINs face civil monetary penalties of up to $1 million per violation.
Healthcare providers face disincentives rather than penalties, established through a 2024 HHS rule and applied through Medicare programs including the Promoting Interoperability programs, MIPS, and Shared Savings Program participation.

The rules include defined exceptions, including one covering fulfillment through TEFCA. Read the actual exception conditions before relying on any of them, because the summaries circulating in the industry are consistently broader than the regulation.

One consequence patients notice: clinical notes and most test results are released to patients without delay. “Open notes” changed how clinicians write, for better and worse.

The API Mandates Pushing EHRs Forward

Several converging requirements are driving EHR capability whether organizations planned for it or not:

CMS payer API requirements, including the Patient Access API and the Prior Authorization API that impacted payers must implement by January 1, 2027, built on FHIR implementation guides. If you’re following prior authorization reform, this is the EHR-side counterpart.
TEFCA’s move toward FHIR-based exchange
Bulk FHIR export as the emerging path for population data extraction

What EHRs Genuinely Improved

Worth stating plainly, because criticism of EHRs has become so reflexive that the wins get lost.

Prescribing safety. E-prescribing with interaction and allergy checking eliminated an entire class of error.
Availability. The chart is in the room, at home, at 2 a.m., and in the specialist’s office. Records no longer go missing.
Results follow-up. Tracking and routing beat paper for closing the loop on abnormal findings.
Measurement became possible. Quality reporting, registries, and population health work all require structured data. You cannot manage what only exists in a filing cabinet.
Research at scale. EHR data underpins real-world evidence and pragmatic trials in ways paper charts never could.
Patient access. Portals, results release, and messaging shifted the balance of information toward patients meaningfully.

The Documentation Burden Problem

And now the part that hasn’t been solved.

The Evidence

Multiple time-motion and audit-log studies over the past decade have converged on similar findings: ambulatory physicians spend roughly as much or more time on EHR and desk work as on direct patient care, and a meaningful share of that work happens after hours. The phenomenon has its own name, “pajama time,” which tells you how normalized it became.

International comparisons are the most revealing. Research comparing US clinicians to non-US clinicians using the same EHR found US clinicians spending substantially more time in the system per day, and writing notes several times longer. Same software, very different burden.

Why US Notes Are Longer

Because the note stopped being a clinical communication and became a multi-purpose legal and financial document.

Billing documentation requirements drove note length for years, rewarding volume of documented elements over clarity.
Malpractice defensiveness encourages documenting everything.
Quality program attestation added structured field requirements.
Copy-forward tooling made bloat frictionless. A note assembled from prior notes is fast to produce and nearly useless to read.
Prior authorization and payer documentation demands generate their own documentation load.

Notice that almost none of these causes are the software. The EHR made a document that was already becoming bloated much easier to bloat further. Blaming the vendor for note length is like blaming the printer.

Evaluation and management coding changes in recent years removed some of the element-counting incentive, which was real progress. Note length has not fallen accordingly, because habits and templates outlive the rules that created them.

Ambient AI Documentation: What It Is Changing

The most significant shift in clinician EHR experience in a decade is ambient documentation: an application listens to the patient encounter and drafts the note.

The category moved from pilot to broad deployment quickly, with several vendors competing and large health systems rolling out to thousands of clinicians. What the reported evidence generally shows:

Consistent improvement in clinician-reported burnout and cognitive load, which is the finding that matters most and shows up most reliably
Reduced time spent documenting, though the magnitude varies widely across studies and settings
Mixed effects on total time in the EHR, since documentation is only part of the burden and inbox volume keeps growing
Note quality that clinicians generally rate favorably, with the caveat that review is still required

Three honest cautions:

Ambient AI addresses note writing, not the reason notes are long. If billing, legal, and quality requirements still drive content, an AI writes a long note faster.
The clinician remains responsible for the note. Attestation and review obligations don’t transfer.
The inbox is the next frontier and a harder one. Message volume grew substantially after the pandemic, and drafting replies is a more consequential task than transcribing a conversation.

Under the DSI transparency requirements described earlier, predictive models embedded in certified EHRs now come with disclosure obligations. Ask for the source attributes. For any model influencing clinical decisions, you want to know the training population, the validation approach, and the maintenance plan.

EHR Safety Risks Worth Managing

EHRs prevent errors and create new ones. The recurring categories:

Alert fatigue. When decision support fires constantly, clinicians dismiss reflexively, including the alerts that mattered. Override rates above roughly 90 percent are common and should be treated as a configuration failure rather than a clinician failure.
Copy-paste propagation. An error entered once and copied forward becomes permanent and hard to trace.
Wrong-patient errors. Multiple charts open, similar names, interruptions.
Default and dropdown errors. A wrong default dose or unit selected from a picklist.
Interface failures that fail silently. A results feed that stops delivering without alerting anyone.
Downtime. Planned and unplanned, including ransomware. Paper downtime procedures that nobody has practiced are not procedures.

The practical governance step: treat EHR configuration as a patient safety domain with its own review process, not as an IT ticket queue.

What an EHR Actually Costs

Numbers vary enormously by size and situation, so treat any single figure with suspicion. What’s more useful is knowing the components, because buyers consistently underestimate several of them.

Total cost of ownership includes:

Software licensing or subscription, priced per provider, per bed, or as a percentage of revenue
Implementation services, frequently exceeding the software cost itself for large systems
Infrastructure or hosting, declining in relevance as cloud deployment grows
Interfaces and integrations, priced per interface and easy to underestimate by an order of magnitude
Data migration, which scales with how much history you convert
Training, including backfill for clinicians who are in class instead of seeing patients
Lost productivity at go-live. This is the line item most often omitted and frequently the largest short-term cost. Expect a meaningful drop in throughput for weeks to months.
Ongoing maintenance, commonly a substantial annual percentage of license value
Internal staffing, which is permanent. Analysts, builders, report writers, and interface engineers do not go away after go-live.
Optimization, which is where clinical value is actually realized and which almost never gets budgeted

Enterprise implementations at large health systems have run into the hundreds of millions of dollars. Small practice costs are dramatically lower but still typically far above the license quote once training and productivity loss are counted.

The most expensive mistake is budgeting the project and not the operating model. An EHR system is not a purchase. It’s a permanent internal capability.

Selecting and Implementing an EHR System

Selection

Define workflow requirements before scheduling demos. EHR software demos beautifully. A scripted demo against your own workflows and your own edge cases is worth ten polished ones.
Talk to reference sites of your size and type, ideally ones that went live 18 to 36 months ago. Recent go-lives are still in the honeymoon or the crisis; older ones have forgotten.
Evaluate the app and API program, not just core features. Which third-party tools are certified for the platform, and on what terms?
Ask specifically about FHIR and Bulk FHIR maturity if analytics or third-party integration matters to you.
Weigh regional network effects. If your referral partners and the hospital you admit to run one system, that has real value.
Read the contract for the parts that bite later: interface fees, API access terms, data extraction rights on termination, and price escalators.

EHR Implementation

Governance decides success more than configuration does. You need a body that can make decisions and make them stick, with clinical leadership that carries real authority.
Invest in physician and nurse builders. Clinicians who understand both the workflow and the system are the highest-return staffing decision in the entire project.
Decide data migration scope early and defend it. Converting everything is expensive and often unnecessary. Converting too little creates a decade of parallel lookups.
Standardize before you build. Implementation exposes every workflow variation in the organization. Automating existing chaos produces automated chaos.
Plan for the productivity dip explicitly, with reduced schedules and at-the-elbow support. Pretending it won’t happen is how go-lives become crises.
Budget the optimization phase from day one. The first year is about going live. Value comes in years two and three, from the work most organizations defund after go-live.

Where EHRs Are Heading

Four things worth watching.

Ambient and agentic AI moving deeper into the workflow. Documentation was the entry point. The direction of travel is chart summarization, inbox draft replies, order suggestions, and coding assistance. The governance question shifts from “does it save time” to “who is accountable for what it produced,” and the DSI transparency requirements are the first regulatory answer.

Cloud migration and modularity. Both major vendors are moving toward cloud delivery, and FHIR-based app marketplaces make it plausible that specialized functions get bought separately rather than waiting for the core vendor to build them.

External API pressure. CMS payer API requirements, TEFCA participation, and information blocking enforcement all push the EHR from a system of record toward a node in a network. That’s a different product requirement than the one Meaningful Use created.

Consolidation continuing. Market share concentration shows no sign of reversing, which means the practical question for most organizations is not which vendor to pick but how to preserve leverage and data ownership with the one they have.

Frequently Asked Questions

What is an EHR?

An electronic health record is a digital, longitudinal record of a patient’s health information maintained by clinicians and designed to be shared across the organizations involved in that patient’s care. In practice it also runs ordering, results, scheduling, billing, and quality reporting.

What does EHR stand for?

Electronic health record.

What is the difference between an EHR and an EMR?

An EMR is the digital version of a single practice’s chart, built for use inside that organization. An EHR is designed to be shared across organizations and to follow the patient. The practical test is whether the record was built to leave the building.

Who are the largest EHR vendors?

In US acute care, Epic leads by a wide and growing margin, followed by Oracle Health (formerly Cerner) and MEDITECH. The ambulatory market is more fragmented, including athenahealth, eClinicalWorks, Veradigm, NextGen, and newer cloud-native platforms.

What is a certified EHR?

Health IT that meets criteria set by ASTP/ONC, including a FHIR R4 API with US Core and SMART App Launch, the current USCDI data baseline, and decision support requirements. Certification is generally required for participation in Medicare quality programs.