The Healthcare Data Innovation Council hosted the “Advancing Healthcare with Generative AI and Data in Italy” webinar, with a panel of experts from both industry and academia. Elena Bonfiglioli and Andrea Pescino talked about the potential of Generative AI in healthcare, including some cutting-edge use cases, Paolo Parini covered the importance of accessibility to data, with examples where AI is already improving processes in hospitals and, to conclude, Andrea Puligheddu gave an introduction about the AI Act and European Health Data Spaces.

The new advancements in AI are benefiting the healthcare sector at all levels, from providers to payors, to researchers. Generative AI, in particular, was proven to improve several metrics, for example reducing patient waiting time by 50%, speeding up collaboration by 17.7%, and reducing burnout by up to 70%. The most recent models had a leap at zero-shot medical understanding and can also take into account patients’ feelings, a capability that previous models did not have.

However, most of the Generative AI users are still in the phase of “exploration” of such technology, they are aware of it, they adopt it and, in some cases, they integrate it with some Retrieval Augmented Generation. But the unexploited potential is still very high. Models can be adapted and fine-tuned to perform specific tasks efficiently and cost-effectively. They can be combined in multi-agent systems where several agents operate different tasks, communicate with each other, and could even be embedded into devices and androids. The impact we foresee is significant. However, for a deep transformation to happen, the shift to these systems must be easy and safe, with a clear improvement in the results and the collaborations. Timing will be crucial; adoption shouldn’t be too slow or too fast to ensure acceptance. Moreover, to ease the process, there is a need to invest in AI literacy, one of the two main barriers to the adoption of these technologies.

There is a need for a mindset shift towards the creation of collaborative environments between industry and research that can foster early innovation and even better economic returns for the developers. The second barrier might be the legal and regulatory aspects, which in Italy are often disregarded, while they should be taken into account from the design phase of the AI systems. The recently formalized AI Act was written exactly to ensure that transparency, privacy, and accountability are considered. It defines the dos and don’ts for providers of General Purpose Artificial Intelligence, as well as the sanctions for those who do not abide by the rules.

Instead, the European Health Data Space regulation establishes standards and practices regarding access to electronic health data for primary and secondary use. This includes the rights of individuals, rules for EHR systems, rules for cross-country data-sharing, governance, data access, secondary use of data, measures for prevention, security, etc. These new regulations aim to ease data-sharing for the greater good, not to discriminate against patients but rather to inform them of the benefits deriving from the use of AI. Unfortunately, Italy is lagging in the matter of data-sharing, compared to some other countries in Europe. The example of Sweden is striking, in fact, despite the data being regionalized, there are national databases that doctors across the country can consult and allow better support and treatments for the patients. Some national registries contain disease data and can be accessed and used for research too. Work is constantly done to harmonize the data and tackle the biases, to make the databases fairer and more interoperable.

In conclusion, the advancements in Generative AI and data accessibility are offering many benefits from patient care to research. While the potential is immense, success hinges on careful adoption, regulatory compliance, and fostering collaborative environments.