Interobserver variability of teledermoscopy: an international study

Abstract

Teledermoscopy is a rapidly developing field of dermatology with studies demonstrating excellent agreement with face-to-face diagnosis. However, we are unaware of studies evaluating interobserver variability in diagnosis between dermatologists from different continents. This evaluation is important to determine the robustness of teledermoscopy and allow comparisons to be made between different studies. To assess the interobserver diagnostic variability between five independent experienced dermatologists (A-E) in New Zealand, Australia and the U.S.A. Images from 979 lesions from 206 patients were distributed to five dermatologists. The lesions were viewed and diagnoses recorded using MoleMap Diagnose (MoleMap, Auckland, New Zealand) software. The diagnoses were analysed for interobserver variability. There was excellent agreement between four of five dermatologists (A-D) for lesions that were agreed upon as melanoma (κ = 0·81-0·97) and benign naevus (κ = 0·77-0·82).The fifth dermatologist (E) made a more frequent diagnosis of atypical naevus and melanoma than the others. For nonmelanocytic lesions, there was moderate to very good agreement for seborrhoeic keratosis (κ = 0·64-0·80) and basal cell carcinoma (κ = 0·55-0·67), but poor agreement for invasive squamous cell carcinoma (SCC) (κ = 0·05-0·15). Agreement for actinic keratosis (κ = 0·32-0·67) and SCC in situ (κ = 0·15-0·32) was only moderate. When atypical and benign naevi were grouped together and actinic keratosis and SCC in situ grouped together, there was better agreement among all dermatologists. There was good ability to distinguish malignant from benign lesions (κ = 0·57-0·93). There was good agreement among dermatologists A-D but dermatologist E varied from the group with more frequent diagnosis of melanoma and atypical naevus. This difference could be due to different definition of terms with lack of consensus guidelines in definition of atypical naevus, lack of familiarity with the specific patient population and/or diagnostic drift.