Transferability of ICD-10 Machine Learning Models Is Limited in ATTR-CM Screening

The cross-border transfer of machine learning models based on the International Classification of Diseases, 10th revision (ICD-10) for the identification of rare diseases like transthyretin-mediated amyloid cardiomyopathy (ATTR-CM) is very limited, according to a recent study by researchers in Germany and Denmark.

This is the case even under enriched conditions, the authors noted.

Based on these results, the research team concluded that the semantic harmonization of routine data needs to be improved, and ontologies specific to rare diseases, like the Human Phenotype Ontology and ORPHAcodes, need to be integrated to improve cross-border assistance based on machine learning for the timely diagnosis of patients with rare diseases like ATTR-CM.