Medical Informed Consent, Artificial Intelligence and Law - MED-ICAiL

  Scientific evidence shows how the use of AI in healthcare and the integration of such technology into medical devices - from diagnostics to personalised care to high-precision surgery - will have a beneficial effect both on the standard of care for patients and on the overall efficiency of the healthcare system (with benefits in terms of cost savings that would allow more resources to be allocated to prevention strategies and biomedical research).

  On the other hand, the use of AI also raises a number of legal issues. First, there is the opacity of such technology and the so-called "black box" problem, which makes it not always possible to foresee ex ante and understand ex post the operating mechanisms of AI and thus to explain the reasons why the latter has produced a given output.

The second issue relates to the much-discussed inability of intelligent devices to make ethical decisions, which is certainly not uncommon in the medical field. Finally, the use of AI-enabled medical devices reduces, but does not eliminate, the risk of adverse events affecting the health of patients (e.g., exceptional device malfunctions).

  In the light of the framework outlined, the issues relating firstly to the area of informed consent, and in particular:
-    should patients be informed about the use of AI in health care? If so, is it sufficient to provide a general description of the technology and its benefits, or is it also necessary to disclose the associated risks, including those of system malfunction (with any consequent risk that the patient might be induced to refuse treatment, and all the benefits that the use of AI entails)?
-    in order to ensure the effectiveness of the consent given by the patient, is it necessary to have a full explanation of the functioning of the AI system used in the medical service, extending to the hypothesis of autonomous learning of the system itself, which may also be beyond the control not only of the user but also of the developer?
-    if the doctor deliberately fails to disclose to the patient the use of AI in diagnosis or treatment, is he or she liable for violating the patient's right to informed consent?

  These issues - which entail serious risks for the doctor-patient relationship, and whose specificities have received little attention from Italian scholarship - will be examined through a multi-level analysis, starting with the scope of art. 32 of the Italian Constitution, also in relation to articles 2 and 13 of the same Constitution, and to the jurisprudence that, with regard to this norm, has reconstructed the relationship of care as a "relationship of trust", without forgetting that art. 32 of the Constitution qualifies the right to health not only as a "fundamental right of the individual" but also as a "collective interest".

  The subject of the research will also be examined taking into account other normative references, such as Article 3 of the Charter of Fundamental Rights of the European Union - which, on the one hand, enshrines the principle of individual self-determination and, on the other hand, limits it in relation to the protection of human dignity - and Law No. 219/2017 on informed consent and provisions for prior treatment, as well as the positions of the Italian National Bioethics Committee on this issue and the deontological rules governing the exercise of the medical profession.

  There are obvious overlaps with the issue of consent to the processing of personal data in the medical field, where the widespread use of AI, including in health research, is putting the strictness of data protection regulations to the test, especially with regard to EU Regulation 2016/679 (GDPR). On the one hand, AI-based medical research inevitably requires as much (health) data as possible; on the other hand, machine learning techniques facilitate the re-identification of originally anonymised data and the creation of derived personal data.

  Thus, while AI represents an 'antagonistic' element with respect to the rights of individuals over their own data, the collection of health data and the use of AI in research could bring significant benefits in terms of protecting individual and public health. Even in the area of data protection, therefore, AI raises potential conflicts between individual self-determination and other fundamental interests that are only partially addressed by the current data protection legal framework, while the recent proposal for an EU Regulation on the European Health Data Space (EHDS) appears to be more aware of the multiple interests at stake, but does not provide definitive solutions.

 The analysis will therefore start from the issues related to the consent of the data subject as a legal basis for the processing of health data in computational research, also in relation to principles such as the prior establishment of the purpose of the processing. In this context, the MED-ICAiL project will assess, among other things, whether the so-called "progressive consent", already envisaged by recital 33 of the GDPR and by a recent opinion of the Italian Data Protection Authority (ref. 9791886), is adequate to meet the needs of rapid medical research while respecting legal certainty and the protection of individuals.

  On the other hand, it should also be considered that even legal bases other than the consent of the data subject are not always sufficient on their own to strike an appropriate balance between the individual's right to control personal data and the interest in public and private health. These criticisms do not seem to be sufficiently taken into account by either the EU Regulation for an Artificial Intelligence Act or the EHDS, where the latter makes the conformity of AI systems (and thus the authorisation to place them on the market) subject to full compliance with European data protection law (see Rec. 49 and Art. 44 EHDS).

The MED-ICAiL project is based on a threefold method of analysis:

-    theoretical, with the aim of analysing the current legal framework - national and supranational, and in particular European - relevant to the issues to be researched, also by mapping and analysing the legal sources relating to the derogation of the principle of self-determination in the medical field (informed consent and consent to the processing of personal data);
-    technical, based on a deep understanding of relevant healthcare AI applications from both a medical and engineering perspective;
-    empirical, based on the analysis of specific practical cases, also with a view to identifying solutions that are immediately useful to health professionals and researchers.

  The research will thus be developed in two directions: a theoretical assessment of the regulatory framework and an empirical analysis of its practical application, also with a view to developing policy recommendations and operational proposals. All this will be done through a multidisciplinary comparison, which will entail:

i)    workshops with experts in computer science and biomedical engineering to identify specific issues at the intersection of AI applications in medicine and the principle of self-determination;
ii)    interviews with experts and researchers in the health field to identify current and foreseeable critical issues;
iii)    surveys to measure people's awareness and trust in AI technologies.
 

The MED-ICAiL project aims to achieve four outcomes:

a.    Improving knowledge: we will classify AI-based systems used in healthcare and identify the specific legal issues associated with each type of such system. It will also provide empirical findings (in particular, the opinions of experts and researchers working in the field of healthcare will be collected through interviews, as well as patients' perceptions and doctors' opinions through surveys) that could form a cornerstone for further research on the same issues. This will contribute to both legal theory and substantive law by showing how the concepts of self-determination and informed consent need to be shaped (at least) to keep pace with technological developments in healthcare. In addition, the project will outline a realistic approach that takes into account the complexity of situations and subjective pluralism in adapting the framework.

b.    New tools (guidelines): the project aims to have both a direct and indirect impact on the development of health technologies by removing some of the barriers to the implementation of AI-based systems in the health sector, and by providing researchers and practitioners with guidelines and criteria that will enable them to use AI techniques in a way that avoids violations of patients' right to self-determination in clinical treatment and health data protection.

c.    Policy recommendations: the project will provide policy recommendations to national and EU legislators for the adoption of legislation that facilitates the digitisation of health and the implementation of AI in healthcare, without weakening the protection of individual rights.

d.    Aims of social, economic and legal progress: the research results will have a major impact on (i) the collective interest in health promotion and disease prevention; (ii) the efficiency of health and health care systems; (iii) the competitiveness, sustainability and innovation of the EU health industry; (iv) the decision-making process of legislative bodies.