AI in healthcare: How electronic patient records are driving digitization

Why digitization in healthcare took so long

Digitalization in healthcare is an evergreen that hasn't really taken off for a long time. Sandra Heuer compares the process with a rocket that had to go through several preliminary stages: first the electronic health record, then EPA 1.0, EPA 2.0 — all precursors that provided important insights but did not yet bring the breakthrough.

The problem: Lack of connection between service providers

The problem was less the technology itself, but the lack of connectivity between service providers. This term includes not only doctors in various functions — from registered general practitioners to emergency doctors in hospitals — but also pharmacies and other players in the healthcare system.

So there was a product that worked for patients, but the core — the exchange of data with the treatment provider — was not yet there. There was no interoperability between the various systems.

The change from opt-in to opt-out

A decisive difference from the past is the switch from opt-in to opt-out procedure. In the past, insured persons had to actively apply for a file. Not many people did that — at Techniker Krankenkasse, there were around 600,000 out of around one million users across Germany, i.e. a disproportionately large proportion.

Now everyone automatically gets an EPA, unless you object. Techniker Krankenkasse has carried out the largest letter campaign in its history to inform all insured persons. The result: The feared 20-30 percent contradictions from preliminary surveys remained far below expectations.

‍

Better patient care: How the EPA helps in everyday life

The benefits of electronic health records are particularly evident in critical situations. Sandra Heuer talks about a conversation with an emergency room doctor who had to treat a dementia patient with a femoral neck fracture. The patient was unable to provide any information about previous illnesses, allergies or current medications. Without this information, any treatment becomes a risk.

Emergency health data

With the EPA, doctors now have access to this patient data — under strict data protection requirements and with different access rights depending on the situation. Emergency doctors in hospitals, for example, have different rules than for registered doctors, because rapid action has priority in emergency situations.

E-prescription and medication list: Identify interactions

Another specific advantage concerns medication. Since January 15, all e-prescriptions have been automatically stored in the EPA. This enables automatic interaction checks: When pharmacists see the medication list in parallel with the current e-prescription, the system can warn that a specific antibiotic with a painkiller, for example, could be problematic.

This can prevent complications, especially for older patients, who often no longer have all their medications in their heads themselves. The combination of EPA and e-prescription shows how digitization brings concrete improvements in healthcare — not as a theoretical concept, but as a tangible step forward in everyday life.

‍

The biggest challenge: Unstructured data

Despite all progress, there are still significant challenges. The biggest: Most of the data that currently flows into the EPA is unstructured. They are available as PDF documents or image files — readable by humans, but not machine-processable.

Why AI systems need structured health data

In concrete terms, this means that it is not yet possible to search for all documents that are related to headaches. Instead, manual search work is necessary — and doctors simply have no time for that in everyday practice.

Techniker Krankenkasse has tried to start here with AI applications. The idea: have unstructured documents tagged automatically. But the project ran into two hurdles:

First, training data was missing because doctor's letters do not have a uniform format and health data cannot simply be used for AI training for data protection reasons. Second, for security reasons, the AI model should have run directly on the device — the EPA's encryption technology does not allow data transmission to third-party systems.

She explains in detail how Techniker Krankenkasse is tackling these hurdles and what solutions Sandra Heuer sees for the future in our podcast episode — including specific insights into ongoing projects.

Structured data as a prerequisite

This is the central finding from the podcast: The EPA is the first step, but it is only structured health data that enables the meaningful use of AI in healthcare. Without uniform standards and machine-readable formats, options remain limited.

‍

AI in medicine: Specific areas of application

Despite these hurdles, there are already concrete examples of how AI systems can support healthcare.

Diagnostics and treatment: How AI supports doctors

In future, an AI agent could compile a brief medical history from findings and documents from the last few months. This saves time and gives doctors a quick overview of their medical history.

Another example: speech-to-text solutions that accept documentation. A dentist has already built his own solution that automatically converts his findings into text. Such applications show how artificial intelligence can make everyday life in practices and hospitals more efficient.

From diagnostics to decision support

However, the real breakthrough only comes when the data is available in structured form. This opens up further areas of application: automatic drug interaction checks, personalized treatment recommendations, prediction of disease progression, decision support for treatment providers during diagnosis.

The development of the technology — smaller, more efficient AI models, better on-device processing — is playing into the hands of the EPA.

‍

Shortage of skilled workers as a driver

One aspect that underlines the urgency: The population is aging, while at the same time there are fewer and fewer doctors. Older patients need more medical care, but staff is becoming scarcer.

The example of Scandinavia

In Scandinavian countries, many medical cases are therefore no longer being treated by doctors, but by nurses — supported by decision-supporting AI systems. These systems enable healthcare professionals to make the right treatment decisions, paired with their individual perspective and experience.

Artificial intelligence in healthcare is not replacing people here, but expanding their capabilities. It is not about automation at any price, but about support where it provides the greatest benefit for patient care.

‍

What Germany can learn from other countries

Germany traditionally ranks lower when it comes to digitization in healthcare. But Sandra Heuer also sees an advantage in this: Anyone who starts later can learn from the mistakes of others.

Opt-out vs. opt-in: The comparison with Switzerland and Austria

Switzerland, for example, has an opt-in process and is still struggling to motivate enough people to use it today. Austria has relied on opt-out from the outset — with significantly higher acceptance.

The interoperability challenge

What all countries have in common is that they are struggling with data interoperability. With around 200 different IT systems in Germany alone — from market-dominating solutions to niche products for individual practices and hospitals — interacting between the systems is a permanent challenge. But standards are slowly becoming established, even among start-ups in the area of digital health applications.

That data:unplugged 2026 festival offers an exchange with decision makers who face similar challenges in data integration. In master classes, providers present concrete solutions — from data management to implementation.

‍

The future: Connected apps and AI-powered healthcare

For Sandra Heuer and her team, the next big step is to connect various digital offerings. In addition to the EPA, Techniker Krankenkasse operates other apps: TK Doc for telemedicine, a symptom checker, a laboratory value checker.

How connected applications improve treatment

Networking these services opens up new opportunities: The telemedicine service could provide direct access to the EPA in the future. Medical reports could be explained in understandable language. When health data from various sources flows together and AI systems can analyze them, new opportunities arise for diagnostics, treatment and prevention.

‍

Conclusion: The foundation is in place

Electronic health records will continue to evolve — technologies and requirements will change. The foundation for structured health data is now in place.

Sandra Heuer's appeal at the end of the conversation: more pragmatism. The questioner mentality often prevents the opportunities offered by digital solutions from being used. There are risks and mistakes will happen. But you can learn from mistakes.

The EPA is creating the basis for all other digital initiatives in healthcare. Anyone who understands this and consistently implements it — from data structuring to AI strategy — can take advantage of the opportunities for better patient care, medical research and a more efficient healthcare system.

This shows how companies and decision makers are shaping the transition to a data-driven organization data:unplugged 2026 festival on March 26 & 27 in Münster. On five stages, experts share their specific experiences — from data strategy to AI implementation. For effective implementation, it is recommended to bring key people from the company along and bring them into contact. Tickets for the festival are available here.