During the HIMSS AI and Cybersecurity Virtual Forum on Tuesday, Dr. Xiaoqian Jiang, associate vice president for Medical AI at the University of Texas Health Science Center at Houston, explored the evolving landscape of healthcare data sharing. He emphasized that healthcare data sharing now operates along a broad spectrum, ranging from isolated institutional databases to large-scale, multi-institutional consortia.

"We are moving more and more toward a collective large, consortium data sharing," Jiang explained. However, he cautioned that as the scale of data sharing grows, so too do the associated risks and complexities. The healthcare sector must balance the promise of collaboration with the responsibility of protecting patient privacy.

The Complexity of Protecting Patient Privacy

Jiang highlighted a common misconception: simply removing direct identifiers, such as names or Social Security numbers, does not guarantee anonymity. "If you take the name away, the data looks anonymous but indeed it is still not protected," he said, pointing out that a single attack could re-identify a patient.

Even seemingly harmless data points can become identifiers when combined. For instance, the combination of a zip code, birth date, and gender can uniquely identify an individual in many cases. Jiang also noted that demographic data for specific patient cohorts or information about disease distribution can increase privacy risks.

"Demographics combined with phenotypes provide strong clues to reveal individuals' information," Jiang said. He referenced studies showing that genome data is particularly vulnerable to re-identification. Techniques exist that can recover surnames from personal genomes or use DNA to reconstruct a three-dimensional human face. Another approach involves linking publicly available genotype-phenotype correlations to identify individuals statistically.

Technology as a Shield Against Privacy Breaches

Despite the potential risks, Jiang emphasized that technology offers solutions to safeguard sensitive healthcare information. One approach is secure collaboration, where multiple parties can work together on data analysis without exchanging raw data. This method reduces exposure while still enabling meaningful research.

A key innovation in this space is federated learning. This AI-driven technique trains models across multiple datasets without centralizing the data. By keeping patient information at its source, federated learning minimizes the risk of privacy breaches while allowing institutions to benefit from AI insights.

"AI excels at merging different data modalities," Jiang said, "but privacy and practical constraints often prevent centralizing raw data." Federated learning addresses this challenge, enabling collaborative research without compromising patient confidentiality.

UTHealth Houston Pioneering Privacy-Preserving AI

At UTHealth Houston, researchers are actively developing tools to implement federated learning on a broad scale. Jiang described a federated learning workflow manager designed to allow AI models to train across distributed clinical datasets securely.

This system enables institutions to collaboratively build AI models without directly sharing patient data. The approach not only preserves privacy but also accelerates research by leveraging data from multiple sources simultaneously. "At UTHealth Houston, we are pioneering this kind of development," Jiang said, underscoring the institution’s commitment to innovation in medical AI

Readiness and Future Outlook

When asked whether the healthcare industry is fully prepared to implement these advanced approaches at scale, Jiang acknowledged a gap. "Some institutions have the capacity; not all institutions are ready because few of them move really fast," he said.

Looking ahead, Jiang foresees a more connected and cloud-driven infrastructure. He anticipates the emergence of a common technological layer where institutions increasingly move their data to the cloud. This shift would enable secure, scalable collaboration while simplifying data governance and privacy management.

"I think in the future we would expect there will be a common layer and more and more institutions moving to the cloud so we can leverage the cloud layer and manage the cloud layer to conduct this kind of collaboration," Jiang concluded.

This vision points to a future where healthcare institutions can work together more effectively, harnessing AI and federated learning while maintaining rigorous standards for patient privacy. As data sharing becomes more sophisticated, the combination of advanced technology and careful governance will be critical to unlocking its full potential.