The Past and Future of Fungal Diagnoses in Health Records: A Comparative Analysis and Harmonization of Mycoses ICD Codes

Background: International Classification of Diseases (ICD) codes are used in healthcare for documenting disease and billing, but also for research in epidemiological and translational studies for cohort identification and specimen selection. Periodic code revisions, while necessary, significantly differ in structure and descriptive content, preventing direct cross–version mapping while also complicating use in longitudinal research and coding accuracy during the transition to newer versions. Methods: Mycoses codes within ICD–9, ICD–10, and ICD–11 were manually identified within fungal and non–fungal disease sections using systematic searches for fungal–associated text and eponymous diseases with confirmed fungal etiology. Code descriptions were analyzed to classify etiologic and clinical characteristics for harmonized match scoring. Code similarity was assessed using a heuristically weighted match–score algorithm incorporating classification variables with description–based text mining for clinically relevant semantic harmonization. Codes were categorized into twelve fungal groups (Aspergillus, Blastomyces, Candida, Coccidioides, Cryptococcus, Histoplasma, Mucorales, Paracoccidioides, Sporothrix, Other Fungus–specific Mycoses, Dermatomycoses, and General Mycoses). Bidirectional best–match mapping was performed within each fungal group and manually verified. Results: Fungal codes were organized within and outside the mycoses section in all ICD versions, with several requiring expertise for identification. ICD–9, ICD–10, and ICD–11 contained 94, 117, and 148 mycoses codes accounting for 0.7%, 0.2%, and 0.4% of all codes, respectively. Most codes were taxon-specific (56%, 69%, 61%), but code proportion varied by fungal group and was version-specific. Cross–version bidirectional mapping revealed distinct patterns of code continuity, expansion, and consolidation. Conclusions: A description–based matching approach integrating disease classification and text mining enabled reliable cross–version best–match mapping and produced a resource to improve coding accuracy, support ICD–11 adoption in the United States, and simplify code use in ICD–based studies. Findings emphasize the importance of nomenclature consistency in science and medicine as molecular studies reclassify fungal pathogens.