To examine pediatrician diagnostic skill development of dermatology image-based cases via a web-based tool and to determine case-level variables that were associated with diagnostic error.

This was a multicenter, prospective, cross-sectional study. A convenience sample of pediatric trainees and attendees were eligible for participation. Using a web-based tool, physicians practiced 334 pediatric dermatology image-based cases until they achieved a performance standard. Participants identified whether the case was concerning, the morphologic category, and the specific diagnosis. After every case, participants received corrective feedback and their progress toward the performance standard.

Among 185 participants, there was a significant improvement in diagnostic performance in classifying concerning vs nonconcerning (+19.2% [95% CI 17.7, 20.6]), morphologic category (+17.9% [95% CI 16.5, 19.3]), and specific diagnosis (+25.2% [95% CI 23.4, 26.7]). The median number of cases required to achieve the performance standard was 142 (IQR 96, 209; min 58, max 330), with a median time to achievement of 57.3 minutes (IQR 38.7, 84.3). Based on 38 502 case interpretations, children with darker vs lighter skin color had a lower odds of correct identification of “concerning” (OR = 0.87; 95% CI 0.83, 0.93), morphologic category (OR = 0.91; 95% CI 0.85, 0.97), and specific disease (OR = 0.96; 95% CI 0.90; 0.99). Fewer than 60% of participants correctly identified bullous variations of diseases, psoriasis, herpes infections, and nonspecific viral infections.

The deliberate practice of dermatologic presentations in the context of an education intervention significantly and efficiently improved pediatrician diagnostic ability. The specific diagnostic challenges identified also provide opportunity for targeted learning opportunities in these areas.