Electronic Health Records: What Was Promised and What Was Delivered

In 2009, Congress passed the Health Information Technology for Economic and Clinical Health Act, known as HITECH, and set in motion the largest coordinated push toward electronic health records the country had ever seen. The idea was simple: pay hospitals and physicians to adopt certified EHR systems, set meaningful use criteria they had to meet to keep the payments, and eventually penalize those who did not comply. By the mid-2010s, EHR adoption rates across hospitals had climbed sharply. On the adoption metric alone, the program worked.

The question I keep coming back to is whether adoption is the right metric.

When policy makers designed Meaningful Use, they described a future where patient records followed the patient, where care teams had complete information at the point of care, and where the data collected in daily clinical work would drive quality improvement and lower costs. Those were real promises, attached to billions of dollars in federal incentive payments. The adoption numbers came in. The outcomes are a different story.

Start with what happened to physicians. There is a substantial body of medical literature, and it has grown larger every year since the mid-2010s, on the relationship between EHR adoption and physician burnout. I am not going to cite a specific burnout percentage because the numbers vary by study and I do not want to overstate what any single survey shows. But the direction of the finding is consistent enough to take seriously: many physicians report spending more time on documentation and data entry than on direct patient interaction. The EHR created administrative work that did not exist before, and that work does not feel like clinical care. It feels like paperwork that happens to appear on a screen.

Part of this is structural. EHR systems were built to serve multiple masters at once. Billing and compliance drove a lot of the design decisions. Insurance reimbursement requires specific coded fields to be filled. Regulatory requirements create mandatory documentation templates. The result is a system that is very good at capturing what payers and regulators want to see, and sometimes awkward for capturing what a clinician actually needs to record. When a doctor spends twenty minutes clicking through required fields to document a fifteen-minute patient visit, the EHR is doing exactly what it was designed to do. It is just not designed primarily for clinical care.

Epic dominates the hospital EHR market, and Cerner (now part of Oracle Health) is a major competitor. Both companies built systems that handle billing, compliance, scheduling, ordering, and documentation in one platform. That integration has real value. But it also means the system has to serve all those functions simultaneously, and the interfaces reflect that complexity. Clinicians who spend their careers learning workarounds and documentation shortcuts to navigate these systems are not being irrational. They are adapting to tools that were not built with clinical efficiency as the primary objective.

The interoperability problem is where the original promise fell apart most visibly. When HITECH passed, one of the central arguments for EHR adoption was that digital records would enable data sharing. A patient who visits an emergency department should not have to describe their medications from memory. Their primary care record should be available. Their specialist notes should follow them. The vision was a connected health information infrastructure, and Meaningful Use was supposed to help build it.

What actually happened is that hospitals and health systems adopted different EHR systems, and those systems do not talk to each other well. A patient who receives care at a hospital that runs Epic and then visits a specialist at a clinic on Cerner may have records in both systems that never synchronize. The specialist may not know about a medication the hospital prescribed. The primary care physician may not receive the discharge summary from the hospitalization in a format that loads into their system automatically. In 2026, fax machines remain common in US healthcare not because the industry is backward but because a fax works across organizational boundaries regardless of which EHR either party runs. Fax is the lowest-common-denominator protocol that everyone has and everyone understands.

There were technical and regulatory attempts to address this. HL7 FHIR (Fast Healthcare Interoperability Resources) is a standard designed to enable data exchange between EHR systems using web APIs, and it has gained real traction. The 21st Century Cures Act, passed in 2016, included provisions requiring EHR vendors to support interoperability and prohibiting practices that blocked information sharing. That the law needed to prohibit information blocking tells you something about how the market was behaving before the prohibition existed. I wrote a longer piece on this specific problem, why your doctor still sends faxes, if you want to go deeper on the interoperability side.

The billing incentive structure also shaped what kind of data gets collected. A well-documented visit in an EHR looks good for reimbursement purposes. It may or may not reflect what actually happened in the room. Researchers who try to use EHR data for clinical studies regularly encounter this problem: the data is abundant, but its completeness reflects documentation incentives as much as clinical reality. Fields that matter for billing are filled carefully. Fields that matter for research or care coordination may be incomplete.

None of this makes HITECH a failure by every measure. EHR adoption did create the infrastructure for digital health. There are genuine improvements in medication tracking, allergy alerts, and clinical decision support that were much harder to deliver with paper records. The question is whether the investment and the disruption were proportional to the outcomes achieved. Meaningful Use set very specific adoption metrics and hit them. It set less specific outcomes metrics and the evidence on those is mixed, sometimes quite mixed.

I think the clearest lesson is one that IS researchers talk about a lot in other contexts: adoption is not use, and use is not effective use. Having an EHR system in a hospital is adoption. Physicians logging encounters in that system is use. Having the right information available at the right moment to improve a clinical decision is effective use. The policy program optimized hard for adoption. It did not and maybe could not easily optimize for effective use. The gap between those levels is where most of the promised value was supposed to live.