EUNIWELL ASTUTENESS - A Visual Guideline for Trustworthy AI-driven Tools in Healthcare
Is this AI-driven tool for healthcare trustworthy?
Which questions are important to ask yourself before using new technologies using artificial intelligence (AI) in healthcare?
ASTUTENESS - A VISUAL GUIDELINE FOR TRUSTWORTHY ARTIFICIAL INTELLIGENCE IN HEALTHCARE
ASTUTENESS project is part of the EUniWell Research Incubator .
Guidelines alone are unlikely to persuade healthcare professionals to modify their practice by using AI-driven tools in healthcare, we often lack the proper indicators and metrics to measure relevance of a tool. We need more actionable tools assessing new technologies. The decision to adopt, accept, and utilize an innovation is not an instantaneous act, but more often an iterative process. Thus we worked to provide a visual guideline to assure the artificial intelligence decision support systems tools healthcare professionals use are trustworthy. It is a hybrid model that combines elements from the Pasteur Quadrant[1] and the Gartner (“Magic”) Quadrant[2] for assessing measure of trust in AI-driven tools in healthcare.
ASTUTENESS is a Euniwell Research Incubator project that brings together researchers and students from Med AI Lab University of Murcia - Spain, Centre for Data Intensive Sciences and Applications University of Linnaeus - Sweden, Digital Health Solutions in Medicine University of Semmelweiss - Germany, DEep Learning Proposal for Healthcare and Innovation University of Nantes - France, and University Hospital Morales-Meseguer - Spain. By studying real use case of decision support tools for drug related problems and antibiotic resistance using two types of clinical datasets (proprietary and open data), we analyzed two main Machine Learning paradigms: supervised deep learning methods (e.g., RNN, LSTM) and unsupervised interpretable methods (e.g., pattern mining, subgroup discovery algorithms) and discussed key factors that helps qualify the trustworthiness of an AI-driven tools for healthcare. Thus, in our discussion at Linnaeus University, we highlighted the importance of education to improve adoption of new technologies of AI-driven tools in healthcare. The need of curricula that portray multifaceted issues necessary to evaluate AI-driven tools in healthcare has become a priority in many European countries.
Also, adaptation is frequently a crucial aspect of adoption when building a discovery for a population subgroups other than those from which it originated. We confirmed our strong believe that translating innovation of AI-driven tools to practice can be stimulated or enhanced by interdisciplinary work flow to collaborate to the disciplines of one another in informatics, mathematics, education science, epidemiology, etiology, behavioral and social sciences, biostatistics, and healthcare specialties.
The hype cycle of Gartner in 2023[3] showing the interest for Healthcare Data, Analytics and AI clearly points that Population Health Management solutions and advanced analytics architecture for payers and providers such as clinical decision support system have passed the disillusionment and are now on the slope of enlightenment. It means more instances of how the technology can benefit the enterprise start to crystallize and become more widely understood. Second- and third-generation products appear from technology providers. We will soon meet the plateau of productivity and mainstream adoption so it is important to provide tools to assess trustworthiness in AI-driven decision support system in healthcare.
Adoption of a new technology depends on specific incentives such as the degree to which planners consider the innovation better than the practice it precedes and the extent to which a potentially effective innovation will improve outcomes. The Gartner Quadrant looks at the competitive ability of four types of tech providers and rates them on a scale of the ability to execute vs completeness of vision to attribute them a position of niche players, challengers, visionaries or leaders. We decided to adapt this concept to a radar plot and mix it with Pasteur’s Quadrant concept in the result interpretation to verify if the AI-driven tools is Pure Basic research, Use-inspired basic research or Pure applied research so it can be classified by wether it advances human knowledge by seeking a fundamental understanding of nature, or wether it is primarly motivated to solve a practical problem.
To translate the “Effective”, “Safe” and “People centered” paradigm of quality in healthcare from the WHO, in ASTUTENESS consortium we strongly believe AI-driven tools are of good quality and trustworthy if they are “FR: Utile, Utilisable, Utilisé” or “ES: Útil, Utilizable, Usado” or “SW: Användbar, Annan, Använd” or “DE: Nützlich, Nutzbar, Benutz”!
“To him who devotes his life to science, nothing can give more happiness than increasing the number of discoveries, but his cup of joy is full when the results of his studies immediately find practical applications.”
—Louis Pasteur
It takes an average of 17 years for 14% of original research to be integrated into physician practice[4], we must consider a more effective way to bring discovery to practice for the better goods of patients. If dissemination of clinical guidelines using passive methods (publication of consensus statements, mailings, …) or single-source prevention messages has been ineffectives, scoring grid is a an alternative way to engage and promote adoption of new technologies despite the lack of confidence, insufficient technological knowledge and low literacy on responsibilities awareness towards AI-driven tools.
With this visual guideline in an eight dimensions scoring system to download, healthcare professionals, buyers, and innovation managers will be able to provide evidence-based of the trustworthiness of an AI-driven decision support system for healthcare assuring that it is developed in ways that match adopters’ needs, resources and infrastructure.
To go further, in order to build trust with the ultimate beneficiaries (patients) of AI-driven tools in healthcare as well at the intermediate beneficiaries (doctors), Semmelweiss Universities researchers developed a pilot study to measure trust in AI-driven tool:
This project result from a single day workshop. It highlighted the importance of collaborative research in different European countries with different National Health systems and confirmed the relevance of the Euniwell program. Discussion are taking place in order to identify knowledge gaps and opportunities leading to further research projects together. Feel free to use & modify it openly (CC by NC 4.0).
Authors: Martin-Gauthier, J.1, Fadeenkov, I.1, Paris, J. 1, Löwe, W.2, Hammar, T. 2, Lincke, A. 2, Unander, D. 2, Juarez, J. M.3, Canovas, B. 3, Kim, D. 3, Mora, F. J. 3, Girasek, E.4, Döbrössy, B. 4, Belando, S. A.5, Juarez, J. M.3, & Gourraud, P.-A. 1.
Affiliations:
1 Nantes Université, CHU Nantes, INSERM, Center for Research in Transplantation and Translational Immunology, UMR 1064, ITUN
2 Linnaeus University
3 University of Murcia
4 Semmelweis University
5 Hospital University Morales Meseguer
[1] Stokes, D. E. (1997). Pasteur’s quadrant: Basic science and technological innovation. Washington, DC: Brookings Institution Press.
This project has received funding from Euniwell Research Incubator.
This project has received support from the DELPHI cluster I-Site NExT through a France grant managed by ANR PIA (reference ANR-16-IDEX-0007), DELPHI also receives financial support from Région Pays de la Loire and Nantes Métropole.
This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 101035821. With the support of the Erasmus+ Programme of the European Union.
Documents à télécharger
- Euniwell ASTUTENESS Quadrant XLS, 347 Ko