Evaluation of the clinical applicability of artificial intelligence in determining systemic therapy indications and treatment selection in psoriasis.

OBJECTIVE: To evaluate the concordance between artificial intelligence (AI)-recommended and dermatologist-selected treatments in real-world psoriasis care. METHODS: Eighty-two patients with psoriasis were included in this study. An AI model (ChatGPT) was provided along with treatment guidelines and clinical trial data to generate treatment recommendations. The concordance was assessed for systemic therapy indications (n = 73) and drug selection (n = 58). Clinical outcomes were evaluated using Psoriasis Area and Severity Index (PASI) 75 and 90 responses. RESULTS: Among 43 patients treated with biologics or deucravacitinib, the PASI 75 and 90 response rates at week 48 exceeded 90% and 80%, respectively. Concordance for systemic therapy indication was 79.5% (95% CI, 68.7-87.6%) overall and 89.4% (95% CI, 76.9-96.5%) among patients with PASI ≥5. For drug selection, concordance was 46.6% (95% CI, 33.7-59.8%) within the top three and 62.1% (95% CI, 48.5-74.5%) within the top five recommendations. Binary κ values were 0.41 and 0.58, and weighted κ values were 0.52 and 0.68 for the top three and top five recommendations, respectively, indicating moderate-to-good agreement. CONCLUSION: AI showed moderate-to-good concordance in determining systemic therapy indications and drug selection. These findings highlight the potential of AI in supporting treatment decisions for psoriasis.