New AI Model Enhances Blood Test Interpretation by Integrating Individual and Population Data
Researchers have developed NORMA, an AI framework that combines personal and population data to improve blood test result interpretation, aiming to reduce misclassification and enhance disease prediction.
Blood tests are essential for diagnosing and managing various health conditions. Traditionally, doctors interpret these tests using standard reference intervals derived from large populations. However, these fixed ranges may overlook significant changes in an individual's health, as personal baseline levels can vary and remain stable over time.
To address this limitation, researchers analyzed nearly 2 billion blood test measurements from over 1.6 million individuals across North America, the Middle East, and East Asia. They discovered that while blood test values are highly individual, relying solely on personalized reference intervals often misclassified up to 68% of results as abnormal, even when these did not correlate with adverse health outcomes.
In response, the team developed NORMA, a conditional transformer-based framework that generates reference intervals by considering both a patient's history and population-level data about normal variation. By integrating these two data sources, NORMA provides a more balanced and precise interpretation of blood tests.
Testing demonstrated that NORMA-derived intervals achieved higher precision in predicting outcomes such as mortality, acute kidney injury, and chronic diseases. This suggests that anchoring individual trajectories to population-level priors enhances early disease detection and treatment decisions.
The researchers have publicly released the NORMA model, code, and an interactive user interface to promote transparency and accessibility in individualized laboratory interpretation. This initiative aims to assist healthcare providers and patients in better understanding blood test results, thereby improving patient care.
