Are the European reference networks for rare diseases ready to embrace machine learning? A mixed-methods study

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Rare Disease Diagnostics

Background:

• Rare disease (RD) patients face diagnostic delays compared to common diseases.
• Machine learning (ML) technologies present opportunities to improve diagnostic speed and accuracy for RDs.
• European Reference Networks (ERNs) are key stakeholders in advancing RD care.

Purpose of the Study:

• To explore the expectations and experiences of ERN members regarding ML technologies for RD diagnosis.
• To assess the perceived benefits and barriers to ML adoption in rare disease healthcare.

Main Methods:

• A mixed-methods approach combining an online survey and focus group discussions.
• Targeted medical professionals and other individuals affiliated with 24 ERNs.
• Collected 423 survey responses and conducted qualitative focus group discussions.

Main Results:

• Participants reported limited ML knowledge but recognized improved diagnostic accuracy as a key benefit.
• Lack of training and need for collaboration with developers were identified as barriers.
• Most supported optional, recommended ML use in diagnostics, primarily in specialized units within 5 years.

Conclusions:

• ERN members are enthusiastic about implementing ML for RD diagnostics, despite experience gaps.
• Collaboration among healthcare professionals, developers, policymakers, and patient groups is vital for trust and adoption.
• Further research is needed to understand diverse stakeholder perspectives on ML in rare diseases.