A new study on the implementation of AI in healthcare

It is widely known that European healthcare systems are facing complex challenges, including the increasing demand for healthcare due to an ageing population, the growing prevalence of chronic and complex diseases, rising costs, and a shortage of healthcare personnel. Artificial intelligence ("AI")-based solutions can undoubtedly be valuable tools to address these challenges, offering the potential to improve operational efficiency, reduce administrative burdens, and advance diagnostic and therapeutic pathways.

Despite the potential and availability of AI-based tools, their implementation in clinical practice remains slow. A study commissioned by the Directorate-General for Health and recently published by the European Commission demonstrates how the European Union is uniquely positioned to support the safe, effective, ethical, and equitable expansion of AI in healthcare, balancing innovation with the safeguarding of patients' fundamental rights.

The document, titled "Study on the deployment of AI in healthcare", was developed following a literature review and consultation with five working groups composed of patients and patient associations, healthcare professionals, hospital representatives, AI developers, and regulatory experts. All of this culminated in an interesting final report of 241 pages, available on the European Union's website, which shows that interest in the use of AI in healthcare is steadily and rapidly growing in the European Union, (from 33 research projects funded in 2015 to 85 in 2022; from 6 patents registered in 2016 to 122 in 2024; and from the exponential increase in clinical studies based on AI and machine learning from 2014 to 2024).

However, the study highlights numerous challenges to be addressed on technological, infrastructural, regulatory, organisational, social, and cultural levels. In particular, the European Union faces delays in the standardisation and interoperability of data, as well as a regulatory framework that is not always coordinated and consistent in the interaction between the various applicable disciplines, (consider, for example, the AI regulation, the regulation on personal data protection, and the regulation establishing the European Health Data Space), not to mention regulatory gaps, related to the lack of rules on liability for damages caused to third parties by the use of AI.

Furthermore, from an economic perspective, challenges are identified in relation to the high implementation costs, the lack of clear reimbursement mechanisms, and the absence of comprehensive evaluations to determine the added value of using AI compared to existing solutions. Lastly, from a social and cultural perspective, doubts and concerns about the accuracy of diagnostic or therapeutic decisions made based on AI systems remain widespread among both doctors and patients. These concerns are partly generated by an inadequate level of digital literacy among healthcare professionals, who are also worried that the use of AI could compromise the doctor-patient relationship.

As for proposals, the study calls for the adoption of initiatives aimed at:

1. Establishing common standards for data governance and interoperability across all European healthcare systems;
2. Creating centres of excellence for AI in healthcare that act as hubs for talent and resources, offering advanced training programmes, promoting public digital literacy, and facilitating collaboration;
3. Consolidating funding and introducing financing mechanisms to support specific strategic priorities;
4. Introducing local performance tests, real-world value-added assessments, and post-implementation monitoring of AI tools, with the possible creation of "assurance laboratories"; and
5. Developing a catalogue of AI solutions in healthcare that serves as a centralised repository with detailed performance metrics, user reviews, and informational resources.

This article, originally authored by Vincenzo Salvatore, was originally published (in Italian) in AboutPharma and is reproduced here with permission.