Location
Moultrie, GA
Start Date
7-5-2025 1:00 PM
End Date
7-5-2025 4:00 PM
Description
INTRODUCTION: Teledermatology has become a vital tool in dermatologic care since its introduction, improving accessibility and efficiency, particularly in underserved communities. However, concerns persist regarding its diagnostic accuracy across diverse skin tones. Dermatologic conditions often present differently in patients with darker skin, and teledermatology’s reliance on image-based diagnosis may amplify disparities in detection and management. Additionally, artificial intelligence (AI) systems used in teledermatology may inherit biases due to underrepresentation of darker skin tones in training datasets. This systematic review examines current evidence on the diagnostic accuracy of teledermatology across diverse skin tones, identifies gaps in research, and explores the potential of AI to mitigate diagnostic disparities.
METHODS: This systematic review conducted a comprehensive search across PubMed, Google Scholar, and OneSearch+. Keywords included teledermatology, diagnostic accuracy, skin of color, race, and artificial intelligence in dermatology. Inclusion criteria encompassed studies published between 2012 and 2025 that assessed the diagnostic performance of teledermatology compared to in-person dermatology across varying skin tones. Articles were screened for quality assessment, and data were extracted regarding diagnostic concordance rates, differences in performance across skin types, and AI applications in teledermatology.
RESULTS: A recent meta-analysis of 44 studies found a pooled teledermatology vs. in-person diagnostic concordance of ~69% (κ ≈0.67). Specialist involvement significantly boosted accuracy: when a dermatologist evaluated patients both via telemedicine and in-person, diagnostic agreement was about 71% compared to only 44% when non-specialists were involved. Individual studies likewise report substantial agreement; a large clinic study observed 76.4% concordance between remote and in-person diagnosis (κ=0.636). Teledermatology’s diagnostic accuracy was slightly lower than in-person examination. Qualitatively, common factors influencing performance included image quality and the availability of clinical context. Incorporating high-resolution photography, patient history, or teledermoscopy was noted to further improve diagnostic accuracy, particularly for skin cancer detection (9% higher than traditional methods). AI-assisted diagnostic tools exhibited up to 15% lower accuracy on darker skin tones compared to lighter skin.
DISCUSSION: Teledermatology is a useful tool with diagnostic accuracy comparable to in-person dermatology. This indicates teledermatology is viable for timely triage and diagnosis in many settings, including areas limited in resources. However, its effectiveness across diverse skin tones remains understudied. This review found no study that explicitly stratified diagnostic accuracy by skin type, highlighting a critical gap. This absence of data on darker skin tones leaves uncertainty about teledermatology’s efficacy across diverse populations. AI integration shows potential to enhance accuracy but risks perpetuating bias if datasets lack proper representation. Future research should focus on ensuring diverse representation in teledermatology studies, optimizing AI training methods, and standardizing imaging protocols for all skin tones. Addressing these disparities is crucial to ensuring equitable access to high-quality dermatologic care as teledermatology continues to expand.
Embargo Period
6-3-2025
Included in
The effectiveness of teledermatology and artificial intelligence across diverse skin tones: a systematic review
Moultrie, GA
INTRODUCTION: Teledermatology has become a vital tool in dermatologic care since its introduction, improving accessibility and efficiency, particularly in underserved communities. However, concerns persist regarding its diagnostic accuracy across diverse skin tones. Dermatologic conditions often present differently in patients with darker skin, and teledermatology’s reliance on image-based diagnosis may amplify disparities in detection and management. Additionally, artificial intelligence (AI) systems used in teledermatology may inherit biases due to underrepresentation of darker skin tones in training datasets. This systematic review examines current evidence on the diagnostic accuracy of teledermatology across diverse skin tones, identifies gaps in research, and explores the potential of AI to mitigate diagnostic disparities.
METHODS: This systematic review conducted a comprehensive search across PubMed, Google Scholar, and OneSearch+. Keywords included teledermatology, diagnostic accuracy, skin of color, race, and artificial intelligence in dermatology. Inclusion criteria encompassed studies published between 2012 and 2025 that assessed the diagnostic performance of teledermatology compared to in-person dermatology across varying skin tones. Articles were screened for quality assessment, and data were extracted regarding diagnostic concordance rates, differences in performance across skin types, and AI applications in teledermatology.
RESULTS: A recent meta-analysis of 44 studies found a pooled teledermatology vs. in-person diagnostic concordance of ~69% (κ ≈0.67). Specialist involvement significantly boosted accuracy: when a dermatologist evaluated patients both via telemedicine and in-person, diagnostic agreement was about 71% compared to only 44% when non-specialists were involved. Individual studies likewise report substantial agreement; a large clinic study observed 76.4% concordance between remote and in-person diagnosis (κ=0.636). Teledermatology’s diagnostic accuracy was slightly lower than in-person examination. Qualitatively, common factors influencing performance included image quality and the availability of clinical context. Incorporating high-resolution photography, patient history, or teledermoscopy was noted to further improve diagnostic accuracy, particularly for skin cancer detection (9% higher than traditional methods). AI-assisted diagnostic tools exhibited up to 15% lower accuracy on darker skin tones compared to lighter skin.
DISCUSSION: Teledermatology is a useful tool with diagnostic accuracy comparable to in-person dermatology. This indicates teledermatology is viable for timely triage and diagnosis in many settings, including areas limited in resources. However, its effectiveness across diverse skin tones remains understudied. This review found no study that explicitly stratified diagnostic accuracy by skin type, highlighting a critical gap. This absence of data on darker skin tones leaves uncertainty about teledermatology’s efficacy across diverse populations. AI integration shows potential to enhance accuracy but risks perpetuating bias if datasets lack proper representation. Future research should focus on ensuring diverse representation in teledermatology studies, optimizing AI training methods, and standardizing imaging protocols for all skin tones. Addressing these disparities is crucial to ensuring equitable access to high-quality dermatologic care as teledermatology continues to expand.