Knowledge Graphs: The Data Architecture Powering Next-Generation Healthcare Interoperability

All it takes is a visit to your primary care doctor and a specialist in the same month to realize

that healthcare data is in a communication crisis.

The medical records your primary care physician has frequently can’t talk to the data at the

specialist’s office you’ve been referred to. Those lab results from your annual physical? They

exist in a completely different universe from your imaging scans.

This isn’t just inconvenient for patients. For many, it’s dangerous. Data fragmentation continues

its damaging effects and must be eliminated-now. According to healthcare IT leaders interviewed by Becker’s Hospital Review, next generation of EHRs won’t look like today’s systems at all.

Instead of rigid, screen-bound interfaces, the EHR of tomorrow will be “headless” architectures

with APIs and services decoupled from the database. This will allow workflows to be uniquely tailored – no more “one-size-fits-all” solutions.

Healthcare Data Lives in an Interoperability Illusion

Healthcare systems and patients have been promised data interoperability for decades. Yet

many healthcare providers still struggle with what Becker’s describes as “forklift upgrades”.

In other words, the traditional approach (forcing EHR systems to speak the same language) is

no longer sufficient. Each system guards its data in proprietary formats, creating digital

“fortresses” that are even harder to breach than paper files.

But here’s where things get interesting. The recent surge in AI adoption has fundamentally

changed how we need to think about healthcare data.

A recent article published in Towards AI, revealed that 79% of online traffic is now agentic

– meaning AI agents are doing the browsing instead of humans.

This same principle applies to healthcare. Soon, AI agents will be navigating patient EHR records

helping to carry out multiple different tasks. The question is: are our healthcare systems ready for this future?

They don’t have a choice. The future is coming quickly – and the answer lies in making

healthcare data not just logically interoperable, but machine-readable and AI-optimized. That requires moving from a traditional data representation architecture to a future built on knowledge graphs.

Knowledge Graphs – the Solution ERPs Were Supposed to Solve

Rather than forcing patient data to live in rigid tables and unstructured documents, knowledge graphs create what researchers call “deterministic relationships”– clear, logical connections between different pieces of medical data that both humans and machines can understand.

Think of traditional EHRs as filing cabinets where each document sits in isolation. Knowledge

graphs are more like living “webs” where every piece of information connects to every other

relevant piece.

For instance, instead of storing "Patient X has diabetes" as an isolated fact, a knowledge graph

“learns” that this diagnosis connects to relationships such as medications, lab values, lifestyle

factors, dietary restrictions, and potential complications.

As one Becker’s Hospital Review CIO predicts, “Eventually, an EHR as a standalone core

system may even cease to exist, with all technology needs of a modern health system being

met in a dynamic and fluid manner by an intelligent platform."

Knowledge graphs are the foundation of that intelligent platform.

Making Data Conversational

The real breakthrough comes when we combine knowledge graphs with large language

models (LLMs). Suddenly, healthcare providers are moving beyond querying databases.

Instead, they're having real, living conversations with the data they have and receive.

Several health system CIOs have reported that they are preparing for a future where patients

walk into care spaces and use voice-driven interactions with the EMR, with contextually

relevant information displayed automatically.

Knowledge graphs help voice-activated systems understand context and meaning. For example, when a physician asks, "Why was this patient’s medication changed last month?" the system can quickly reply: "The change was made due to elevated liver enzymes found in recent lab tests, consistent with the patient's history of similar adverse reactions."

Self-hosted LLMs can now navigate knowledge graphs and grasp clear connections within medical data.

The Path Ahead – An Agent-First Healthcare Future

Healthcare systems are moving toward smarter EHRs by adopting knowledge graph architectures, which will improve patient care and strengthen organizational reputation. Leading medical centers are testing ontology-based knowledge graphs for interoperable records, but key steps are needed to implement these solutions.

These steps include:

1. Implementing comprehensive structured data using standards such as Schema.org and FHIR to ensure all medical

information is clearly labeled and machine-readable.

2. Developing robust APIs that enable AI agents to interact with data and carry out complex tasks programmatically.

3. Building cloud-native platforms to reduce the need for local customizations and support ongoing, accurate updates.

There are existing technologies and established approaches. Advancement in this field requires cooperation to enhance traditional methods of managing fragmented and siloed data sources.

By conceptualizing EHRs as dynamic knowledge graphs rather than static storage systems, interactions can be enhanced for both healthcare professionals and AI applications, thereby expanding the potential benefits within digital healthcare, especially for patients.

The goal isn't to replace doctors with AI, but to equip providers and patients with tools to maximize the value of their healthcare data.

About Recorded Health

Recorded Health turns fragmented medical records into intelligent patient journey knowledge graphs as a part of our MyHealthJourney platform. Our, MyDocSaid application, lets patients record and remember healthcare conversations, tackling the issue that most medical information is quickly forgotten. We help healthcare teams access machine-readable, conversational data to improve outcomes.

Find out more at www.recordedhealth.com.