Evaluation of the Validity of International Classification of Diseases Codes for Dermatologic Conditions

Importance: Accurate classification of dermatologic conditions using International Classification of Diseases (ICD) codes is essential for research that uses large administrative datasets. Misclassification can be associated with biased epidemiologic estimates and misleading conclusions in population-based studies. Objective: To systematically identify and evaluate validated classification approaches for dermatologic conditions using ICD codes in US-based administrative, claims, or electronic health record data. Evidence Review: A systematic review was conducted that was registered with PROSPERO (CRD420250654233) and reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. A comprehensive search of Ovid MEDLINE, Embase, Web of Science, and CINAHL was conducted for studies published from January 1, 2000, to October 21, 2025. The data were analyzed in October 2025. Eligible studies evaluated International Classification of Diseases, Ninth Revision (ICD-9) or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes used to identify dermatologic conditions in US-based datasets and reported at least 1 classification metric (eg, positive predictive value). To minimize selection and extraction bias, all screening and data extraction were performed independently by 2 reviewers, with discrepancies resolved by consensus. Findings: A total of 59 studies met inclusion criteria. Most reported positive predictive value, with few reporting sensitivity or specificity. Classification accuracy varied widely by condition and coding strategy. Studies included inflammatory and autoimmune conditions (eg, acne vulgaris, perioral dermatitis, psoriasis, palmoplantar pustulosis, hidradenitis suppurativa, atopic dermatitis, prurigo nodularis, dermatomyositis, cutaneous lupus erythematosus, pyoderma gangrenosum, cutaneous sarcoidosis, pemphigus, pemphigoid, granuloma annulare, alopecia areata, and vitiligo), actinic keratosis and skin cancer, pigmentary and hair disorders (eg, androgenic alopecia, cicatricial alopecia, lichen planopilaris, and melasma), drug reactions (eg, Stevens-Johnson syndrome, toxic epidermal necrolysis), and infections (eg, herpes zoster, herpes simplex virus, and cellulitis or abscess). Classification algorithms that incorporated 2 or more codes, dermatologist attribution, or treatment/procedural data often achieved the highest accuracy. Conditions lacking validated algorithms included seborrheic dermatitis, rosacea, fungal infections, and specific alopecia subtypes. Conclusions and Relevance: This systematic review provides a summary of the most accurate classification approaches to identify various dermatologic conditions in large administrative datasets. These results may inform study designs when using these datasets. In addition, some common conditions lack validated classification approaches, highlighting important areas for future research. As administrative and electronic health record data increasingly support dermatology research, use of rigorously validated algorithms will be essential for generating trustworthy findings.