Development and Assessment of an Artificial Intelligence-Based Tool for Skin Condition Diagnosis by Primary Care Physicians and Nurse Practitioners in Teledermatology Practices

The document discusses a study evaluating an Artificial Intelligence (AI) tool's effectiveness in diagnosing skin conditions by primary care physicians and nurse practitioners. The study involved 40 clinicians (20 primary care physicians and 20 nurse practitioners) who used the AI tool to diagnose 1,048 teledermatology cases. The AI tool was trained on 17,777 de-identified patient cases, covering 2,602 unique diagnoses. The results showed that the AI tool significantly improved diagnostic accuracy, increasing agreement with dermatologist diagnoses from 48% to 58% for primary care physicians and from 46% to 58% for nurse practitioners. However, the study noted limitations, including underrepresentation of certain skin types and lower AI-associated improvements for malignant neoplasms. The study concluded that the AI tool could potentially improve the accuracy and efficiency of skin condition diagnosis in primary care and teledermatology settings, but further work is needed to improve the tool for malignant neoplasms and additional skin types.