Lorraine L. Rosamilia, MD
Dr. Rosamilia is from the Department of Dermatology, Geisinger Health System, State College, Pennsylvania.
Dr. Rosamilia reports no conflicts of interest in relation to this post.
JAMA Dermatology recently published a case-control study from the University of Pittsburgh wherein 4 smartphone applications (apps) designed to aid consumers in determining if a skin lesion is benign or malignant were put to the test. Digital images taken by dermatologists of 60 melanomas and 128 benign control lesions (prior to skin biopsy and dermatopathologist-determined diagnosis) were uploaded into these consumer apps, and there was a very broad range of diagnostic sensitivity and specificity. Notably, the app that included a board-certified dermatologist who reviews the image in 24 hours demonstrated the highest sensitivity (98.1%). The other apps utilized an automated algorithm, and even the best-performing nonphysician-inclusive app displayed only 70% sensitivity while the worst-performing app exhibited 6.8% sensitivity, meaning that 18 of 60 and 56 of 60 melanomas, respectively, were classified as benign.
What’s the issue?
As concerns about patient medical coverage and cost escalate, it is not surprising that the smartphone universe has created avenues for patients to bypass portions of the health care system to manage some medical problems from home. Although certain apps may have their place in patient-centered care, such as those that catalogue patient nevi or educate them on salient clinical features and sun protection, others appear to wrongly substitute for medical consultation, even if their disclaimers suggest otherwise. As the aforementioned study demonstrates, the accuracy of automated analysis of photographs does not substitute for physician evaluation, whether it be telemedicine or otherwise. The US Food and Drug Administration is aware of certain misleading medical apps, but in a broad cyber universe, how do we ensure that dermatology stays on the map for app regulation and truth in advertising?
c9fadce4-b3fa-44c8-9693-d47c9acfaabd|1|5.0