A new study from the Netherlands found that many physicians are hesitant to support mobile health apps that use artificial intelligence to assess skin lesions for cancer.
While some saw potential benefits, others warned that the apps could misdiagnose cases, exclude vulnerable patients, and disrupt clinical care if not properly developed and validated.
Researchers included 6 focus groups totaling 33 general practitioners and dermatologists. They discussed the risks, benefits, and conditions under which they would feel comfortable recommending skin cancer apps to patients. The most serious risk identified was incorrect diagnosis. False negatives and false positives may occur—flagging harmless lesions as suspicious. Such errors could delay necessary care or cause undue anxiety and unnecessary medical visits.
The focus groups also raised concerns about digital exclusion. Patients with lower digital literacy or darker skin tones might not benefit equally, as AI tools often perform better on lighter skin and require some technical skill to use. “I think what you’re going to get more and more in the future is inequity in access to care,” said one GP.
GPs were further concerned about losing clinical autonomy. If an app indicates cancer, patients may insist on a referral—even if the GP disagrees. This dynamic could increase unnecessary referrals and burden the health care system. Additionally, widespread reliance on apps might reduce GPs' hands-on diagnostic experience over time.
Despite these concerns, participants acknowledged potential benefits if the apps are properly designed. Some believed they could raise public awareness and support earlier detection. “We’ll see people with possible skin cancer sooner,” another GP noted.
However, several conditions would need to be met for endorsement. First, apps must undergo independent accuracy testing using representative populations. “The study population must also correspond to the population that will be using the product,” emphasized a dermatologist.
Second, integration into clinical practice must be seamless. Doctors stressed the need for clear, non-alarming language, the ability to share app results with clinicians, robust privacy protections, and well-defined roles in the event of diagnostic errors. Liability issues were a major concern, especially if patients used the apps independently, noted researchers led by Marjolein Lugtenberg, PhD, of the University Medical Center Rotterdam, The Netherlands.
Third, the apps must be accessible to a diverse range of users. Participants emphasized the importance of clear instructions, simple interfaces, quality checks for image submissions, and inclusive design tested across different ages, skin tones, and physical abilities. One GP cited the Dutch COVID-19 tracing app as a successful model, having been tested by various user groups before release.
Overall, the physicians agreed that skin cancer apps could be valuable tools—but only if they are accurate, equitable, and responsibly integrated into healthcare systems. Until these criteria are met, most remain cautious, reported the researchers.
Full disclosures are available in the published study.
Source: BMC