Digital App Successfully Detects Skin Cancers

As the use of artificial intelligence expands across the medical spectrum, researchers sought to determine the effect of a mobile phone app (mHealth) for identifying suspicious skin lesions on healthcare systems. Study authors found that use of an mHealth app for early detection of skin cancers was followed by an increase in the number of claims for premalignancies when matched with a control group.

peopleimages.com/AdobeStock

In 2019, more than 2.2 million customers of a large Dutch insurance company were invited to use the mHealth (SkinVision) app free of charge. All insureds >18 who completed at least 1 app-based skin assessment of a suspicious lesion were included in the study. App users were matched to controls on a 1 to 3 ratio. Study authors were primarily interested in the frequency and type of dermatological claims between app users and controls.

Secondary objectives were to examine differences between therapeutic and diagnostic interventions for suspicious skin lesions, to determine the differences in direct healthcare costs, and to conduct a short-term, cross-sectional cost-effectiveness analysis.

The mHealth app used in the study uses a convolutional neural network (CNN) to identify skin lesions as high or low risk and send the information to the user. The recently validated CNN showed a sensitivity of 87–95% and a specificity of 70–78%.

Claims for benign skin tumors and nevi were almost 4 times higher among mHealth users (5.9%) compared to controls (1.7%) (OR 3.7 (95% CI 3.4–4.1)). App users also had a higher average annual cost for dermatological care (€64.97 vs. €43.09).

Study authors suggest that while use of an mHealth app for full body examinations is not recommended, an app could be helpful for examining high risk lesions. Users of the app had an increase in claims for (pre)malignancies so the authors suggest that use of the app could be a step toward identifying skin cancers. However, because the app detects all cutaneous (pre)malignancies, including melanomas, keratinocytic carcinomas, and actinic keratosis, which have different morbidities and mortalities, overdiagnosis is a risk of the app.

The incidence of false positives is another issue of the app. A diagnosis of skin cancer will likely cause anxiety for the user and lead to additional costs.

Limitations of the study included that mHealth app users tended to be younger, resulting in a large part of the older population for which the app is intended not being included. Also, study results were based on claims data, which most likely resulted in an underestimation of the impact of the app. In addition, users of the app are likely people who are more concerned about their skin, which could result in overdiagnosis.

Reference

1. Smak Gregoor AM, Sangers TE, Bakker LJ, et al. An artificial intelligence based app for skin cancer detection evaluated in a population based setting. npj Digit. Med. 6,90.(2023). https://doi.org/10.1038/s41746-023-00831-w