Artificial intelligence (AI) models could help detect patients who may have undiagnosed systemic mastocytosis (SM), a rare disease caused by abnormal mast cell growth, according to a study published recently in The Journal of Allergy and Clinical Immunology.
By identifying patients with symptoms and medication patterns linked to SM, researchers developed a predictive model that could improve early detection and screening recommendations.
“Interpretable AI models can use EHR [electronic health record] data to identify patients who should be screened for SM. Further work is needed to refine, evaluate generalizability, and prospectively evaluate models,” the study’s authors explained.
This study examined data from 46,543 patients within the Penn Medicine health system who had at least two clinical features commonly associated with SM. Researchers trained machine learning models, including a logistic regression model with LASSO regularization, to predict elevated ambulatory blood tryptase levels, a key marker for SM. The model identified 26 predictive factors, including symptoms such as flushing, itching and low blood pressure, and the use of H2 blockers.
Read more about SM testing and diagnosis
In testing, the model demonstrated a strong predictive performance, and correctly identified 40% of patients with SM. The model had a number needed to test of 10.3, indicating that screening 10 patients identified by the model could lead to one diagnosis of SM.
The AI system successfully identified all known SM cases confirmed through medical chart reviews, including a patient whose tryptase levels were not elevated.
Early identification of SM is crucial because its symptoms vary widely and often resemble other conditions, making diagnosis difficult. Many cases go undetected for years, delaying treatment that could prevent complications. By analyzing patterns in electronic health records, AI can flag high-risk patients and prompt further testing, potentially leading to earlier and more accurate diagnoses.
The researchers emphasized that while the AI model performed well in this study, further refinement and validation are necessary before it can be widely implemented in clinical practice. Additional testing in diverse patient populations would help ensure the model is generalizable and effective across different healthcare settings.
For patients, AI-assisted screening could mean earlier detection and better management of systemic mastocytosis. Physicians may use these models as decision-support tools to identify at-risk individuals and recommend appropriate diagnostic testing. As AI technology advances, it has the potential to transform rare disease detection and improve patient outcomes.
Sign up here to get the latest news, perspectives and information about SM sent directly to your inbox. Registration is free and only takes a minute.
