Work

Patient perceptions of artificial intelligence integration in dermatology: a cross-sectional study of trust, comfort and equity across multiple care modalities

Year: 2025

Type: article

Abstract: Background Artificial intelligence (AI) and telemedicine are rapidly changing the way dermatological care is delivered. As these tools are increasingly used in tandem, understanding how patients perceive the integration of AI across different care settings is important for responsible implementation. Objectives To assess patient perceptions of AI in dermatology across five care modalities and examine how demographic factors influence acceptance. Methods A cross-sectional survey was conducted among 130 adults at a US academic dermatology clinic between December 2024 and April 2025. Participants rated trust, comfort, perceived quality, privacy and confidence in equitable performance across three AI-involved modalities: standalone AI apps, AI-assisted in-person visits and AI-assisted telemedicine visits. Differences in perception outcomes across the three care modalities were analysed using repeated measures Anova. Logistic and linear regressions analysed predictors of acceptance, including age, race, skin tone, socioeconomic status, rurality and technology experience. Results Patients strongly preferred dermatologist-involved care over standalone AI, with 73.8% trusting dermatologist-guided AI and only 1.5% trusting AI apps alone. Comfort and perceptions of equal performance across skin tones were significantly higher for telemedicine and AI-assisted visits compared with AI apps (P < 0.001). Darker skin tone and Black race predicted lower acceptance of AI-assisted care (P = 0.01 and P = 0.003, respectively), while greater technology familiarity predicted higher acceptance (P = 0.05). Comfort varied by clinical scenario, with in-person visits showing dramatically higher odds of patient comfort compared with AI apps alone [odds ratio (OR) 232.8 for new concerns, OR 137.3 for serious concerns, OR 18.4 for sensitive concerns]. AI-assisted in-person visits also showed significantly higher odds of comfort over AI apps (OR 18.4 for serious concerns, OR 3.6 for ongoing concerns). Conclusions Patients strongly prefer AI as clinical support systems rather than autonomous decision-makers, especially for high-stakes and sensitive concerns. Differences in acceptance by race and skin tone point to the need for better representation in datasets and clearer communication about how these tools perform. Moving forward, development and implementation should emphasize clinician and patient involvement, fairness and patient choice to ensure AI is integrated into dermatology in a way that earns patient trust.

Source: Skin Health and Disease

DOI: 10.1093/skinhd/vzaf086