A review of the 34th Medical Informatics Europe Conference
Submitted by Dr. Dana Nyholt, recipient of the 2024 CSCC Grant for Leadership and/or Administration The Medical Informatics Europe (MIE) conference occurs annually and consists of peer-reviewed presentations (oral and poster), workshops, and tutorials. All accepted papers are published in the series of Studies in Health Technology and Informatics. The 34th MIE Conference was held in Athens, Greece, from August 25-29, 2024, with the theme “Digital Health and Informatics Innovations for Sustainable Healthcare Systems”. The conference brought together health informatics experts including data scientists, educators, researchers, developers, and users.
There were a broad range of topics covering database development, system interoperability and standardization, privacy and security of health data, artificial intelligence and machine learning, medical imaging, large language models, health literacy, and education, among others.
The depth of topics ranged from beginner, to intermediate, and advanced, with ample opportunities for clinical chemists and laboratory professionals to gain a foothold in this evolving field. Importantly, it was recognized by the planning committee that it is critical to not only develop digital technologies to improve healthcare, but also to equip healthcare professionals with adequate knowledge and skills in health informatics to ensure appropriate implementation and use of the new tools. Therefore, a collaborative approach between the experts in digital health and informatics and the users is required. This mandate was readily apparent through the conference, opening the door to potential collaborations between digital health researchers and laboratory professionals.
Although many topics were covered throughout the conference, a few themes emerged, as relevant to laboratory professionals:
- We must educate and empower patients and citizens to improve health literacy and enable shared clinical decision making:
- Health Literacy refers to the degree to which individuals have the ability to find, understand, and use information and services to inform health-related decisions and actions for themselves and others. Those with low health literacy tend to have low level of education and training, difficulty in communicating and understanding medical information, and a preference to not know about their condition. In a study examining social media preferences for diabetes awareness, participants were interested in using social media, interacted the most with media posts for awareness days, personal stories, and interviews, features that included videos and had emojis, and posts that had social support, such as tangible assistance and network support. Participants engaged the least with posts that had only pictures, research and development information, food recipes, and X (formerly Twitter) posts.
- “Healthcare continues to prioritize doctors over patients in terms of equity of access, delivery, safety, and effectiveness; healthcare is data-rich, but information quality is poor. We are in need of innovative solutions. We need to empower citizens to access information and advice. This is a long-term process that requires vision and leaders to make it reality. We cannot rely on global crisis to motivate action.”
2. The era of artificial intelligence in medicine has arrived:
- 25% of accepted papers were artificial intelligence in medicine
- Machine learning (ML) and deep learning (DL) are branches of AI. Both are opaque AI models, in which the model’s functioning and behaviour are not explained, and causality
cannot be guaranteed. Without being able to explain why and how, such models present a challenge for adoption into clinical practice. A solution to this is explainable artificial intelligence (XAI), which ideally would include users in the development phase, and engage users in determining the context of use, user requirements, design, and evaluation. With our understanding of medicine, laboratory diagnostics, and statistics, laboratory professionals are well positioned to collaborate with data scientists and computer scientists to develop data-driven, XAI algorithms. - Another perceived barrier to AI implementation in clinical applications is low quality data, including non-standardized and unstructured data. Therefore, interoperability through the Data Life Cycle is seen as critical. Low quality data, including semantics, syntax, and ambiguity, requires restoring integrity of the data set, which is time intensive and laboursome, making downstream work of the data scientist challenging. If data would be collected through interoperable means, this would enable processing, analysis, interpretation, and sharing.
- Implementing AI is about reshaping existing practices, shaping the workflow and clinical pathway, training new routines for healthcare professionals.
3. Large Language Models will serve as a tool to improve the efficiency of medical chart review:
- Large Language Models (LLMs) are increasingly applied to improve the efficiency of reviewing medical reports. Various models, including ChatGPT, Med Llama, GPT-4, Mixtral-
8x7B, are available as both open-source and closed source LLMs. One study demonstrated that humans are better than machines in terms of correctness, while LLMs are better in terms of completeness, making human review absolutely necessary. LLMs were able to process 50 oncology reports in 0.25 seconds, greatly reducing the time required for human review. Use of English language currently is an advantage in LLMs.