Improving Test Interpretation through Personal Reference Ranges, AI and Statistical Literacy

In this episode of the PLUGS Podcast, Dr. Mike Astion and Dr. Geoff Baird interview Dr. Brody Foy, a mathematician and computer scientist who researches practical improvements in interpreting clinical lab data. Dr. Foy discusses the practical use of personal reference ranges, the potential overemphasis on omics vs common testing, why Mike should not hate the future, AI in lab medicine, AI safety, the
problem of statistical illiteracy in test interpretation, and, of course, how “my mother medicine” applies to AI to keep moms safe. Overall, the episode provides valuable insights into the intersection of reference ranges, AI, and statistics in lab medicine.

For more information on Dr. Foy's work on personal reference ranges see:  Foy BH, et al. Haematological setpoints are a stable and patient-specific deep phenotype. Nature. 2025 Jan;637(8045):430-438. doi: 10.1038/s41586-024-08264-5. Epub 2024 Dec 11. PMID: 39663453.

PLUGS Website:
https://www.schplugs.org/

MTS Website
https://medtraining.org/

Chapters

• 0:00 Introduction: Mike invented AI. Dr. Foy shares his background in mathematics and computer science, his education at Oxford, and his transition into lab medicine and pathology.
• 1:13 Conclusion: The podcast concludes with a "My Mother Medicine" discussion on the future of AI and personal reference ranges in healthcare, addressing concerns about job displacement and the continuous need for human expertise in the medical field
• 4:11 Personal Reference Ranges: Dr. Foy explains the concept of personal reference ranges, how they differ from population intervals, and their potential benefits in providing more accurate health assessments.
• 8:00 Sports Doping Testing: Dr. Geoff Baird discusses the use of personal reference ranges in sports doping testing, highlighting the biological passport and its application in elite sports.
• 12:00 Data Requirements: The discussion covers the amount of data needed to develop personal reference ranges, the role of genetic predisposition, and the use of Bayesian inference in refining these ranges.
• 16:00 Use Cases and Challenges: Dr. Foy talks about the use cases for personal reference ranges, such as in diagnosing anemia and PSA levels, and the challenges of overdiagnosis and underdiagnosis.
• 21:00 Implementation and Future: The conversation shifts to the implementation of personal reference ranges in clinical practice, the need for defined use cases, and the potential future developments in this area.
• 30:00 AI in Healthcare: Dr. Foy discusses the current state of AI in healthcare, comparing its advancement in radiology and cardiology to its lag in lab medicine and pathology, and the reasons behind this disparity.
• 39:00 Safety and Validation: The discussion covers the safety concerns of using AI in clinical practice, the importance of rigorous validation, and the need for better statistical education in medical training.
• 1:00:00 Statistical Literacy: Dr. Foy emphasizes the importance of statistical literacy among clinicians, common statistical errors, and the need for better education to improve the use of data in healthcare.