The Importance and Challenges of Interoperability in the Age of AI Healthcare
'Nú concreet aan de slag met AI in de zorg' | ICT&health
Summary
Dr. Carina Vorisek, a medical informatics expert, warns that data interoperability is essential for AI implementation in healthcare. She points out that current AI applications rely on siloed data from single institutions, making safe generalization to diverse patient populations difficult.
Details
Carina Vorisek, an expert in digital health, raises concerns about the status of AI adoption in medical systems. She notes that while many hospitals aim to use AI, significant challenges remain regarding interoperability at the basic data level. Many current AI applications are limited to imaging data (such as DICOM format) obtained from a single institution, leading to models trained in such restricted environments failing to generalize safely across diverse patient populations. The problem is particularly acute with text-based or structured clinical data—like diagnoses or lab results—where many systems still rely on proprietary codes or free text, despite international standards like SNOMED existing. This severely complicates data exchange and AI integration. Vorisek highlights her own experience of needing to use seven different IT systems in one department (obstetrics/gynecology), emphasizing that true value from AI requires access to interoperable data from various systems. The article underscores the critical importance of standards such as FHIR and SNOMED, highlighting the ongoing need for systemic integration.
Original content copyright by respective publishers