A patient receiving cardiology care at a Paris hospital moves to Spain. Their new cardiologist receives a digital medical record, but the local software cannot recognize the format of the imaging reports or the coding used for drug allergies. The doctor starts from scratch: tests, medical history, lab work. This scenario, common in cross-border care, sums up a structural problem that technological innovations alone will not solve.
Health data interoperability: the weak link between hospitals
There is plenty of talk about artificial intelligence in diagnostics, digital twins and bioprinting. There is less discussion of what happens when a patient’s record moves from one facility to another, even within the same country.
The French Digital Health Agency reminds us that cybersecurity plays a direct role in continuity of care. A record that is poorly transmitted or misinterpreted by the receiving system can lead to duplicate tests, delayed care or even a medication error. This is not a theoretical problem.
You can follow the latest news from La Santé de Demain to see just how much this issue now shapes debates around digital medicine. It goes beyond IT: it touches on governance, standards and trust between systems.
The European Health Data Space, which is being rolled out gradually, is specifically designed to make medical information identifiable, documented and usable across borders. Its implementation remains a long-term undertaking: sharing a file is not enough; two different software systems must understand the same information when reading the same fields.
Artificial intelligence in diagnostics: reliability and clinical validation before deployment
AI applied to medical imaging is advancing rapidly. Algorithms are already helping radiologists detect abnormalities in CT scans and mammograms. The promise is real: faster image review, fewer missed findings and better prioritization of urgent cases.
Reliability and explainability are prerequisites for large-scale deployment. G_NIUS, France’s national digital health innovation hub, reports that generative AI is already being used by French healthtech companies. Validating performance across diverse populations using representative data remains a work in progress.
An AI tool trained on data from a university hospital will not necessarily perform as well in a rural clinic with different imaging equipment. This variation between sites raises questions about whether results can be reproduced—and, therefore, about patient safety.
What doctors actually expect
In practice, healthcare professionals are not asking for a tool that replaces their judgment. They want software assistants that flag, sort and suggest, while making their decision-making logic transparent. When an algorithm suggests a diagnosis, doctors need to understand why, not just see a probability score.
- Full traceability of the AI’s reasoning, accessible in the patient record, to document every diagnostic suggestion
- Alerts graded according to the algorithm’s level of certainty, distinguishing a strong suspicion from a preliminary lead
- Compatibility with existing hospital information systems, without requiring a complete infrastructure overhaul
Personalized medicine and biomarkers: tailoring treatment to the patient
Precision medicine is based on a principle that is simple to state but complex to apply: tailoring each treatment to the patient’s biological characteristics. In oncology, biomarkers are already used to select targeted therapies based on a tumor’s profile.
Liquid biopsy is a good example of this shift. Instead of a surgical sample, a blood test detects fragments of circulating tumor DNA. Monitoring becomes less invasive and more regular, and can potentially respond more quickly if the cancer returns.
The limits of personalization at scale
Personalized care requires access to reliable, up-to-date genomic, biological and clinical data that can be shared among care teams. This brings us back to interoperability: a biomarker identified in one laboratory must be readable and usable by the treatment center, even if the two organizations use different software.
Differences in laboratory protocols, detection thresholds and result formats complicate this process. A positive result at one center may be interpreted differently at another if their reference standards are not aligned.
Connected devices and monitoring patients with chronic conditions
Wearable sensors—continuous glucose monitors, connected blood pressure monitors and pulse oximeters—are transforming chronic disease management. Patients transmit data in real time, allowing clinicians to adjust treatment remotely. On paper, the benefits are clear: fewer trips, faster responses and better adherence.
In practice, the quality of monitoring depends on whether the device can communicate with the patient’s medical record. A glucose monitor that sends readings to a proprietary app disconnected from the diabetologist’s software creates yet another data silo.
- Incoming data must be integrated into the shared medical record without requiring manual re-entry by the healthcare professional
- Device-generated alerts must follow clinically validated thresholds, not just the manufacturer’s default settings
- Patients must retain control over how their data is shared, in accordance with the current European regulatory framework
The convergence of connected devices, diagnostic AI and interoperable patient records points to a smoother care journey. Today, however, this convergence exists only in fragments. Each technological component is advancing at its own pace, while common standards are slow to take hold among manufacturers, software vendors and healthcare facilities.
Healthcare innovations are not in short supply. What is missing is the glue that holds them together: shared formats, aligned reference standards and clear data governance. The medicine of tomorrow will not only be more technological; it will also need to be better coordinated.
