Current Issue
This Month's Print Issue

Follow Fast Company

We’ll come to you.

1 minute read

When Diagnostic Health Apps Fail: Skin Cancer Edition

A test of apps that claim to identify melanoma just from a smartphone picture were wrong a shocking amount of the time. What makes a difference? Real doctors.

Smartphone health apps are exciting for everyone involved—the consumers, who can take control of their health data; the investors, who see the promise of big business; and the creators, who are making lots of money. All that excitement should be tempered with at least a little bit of caution, however. In a recent study, researchers found that four different smartphone apps all claiming to be able to identify skin cancer came up with variable results. Most disturbingly, three out of the four apps categorized melanomas (i.e. skin cancer) as normal 30% of the time.

In the study, published in JAMA Dermatology, researchers showed 188 images of skin lesions to each of the four apps, measuring the sensitivity, specificity, and positive and negative predictive value of each of them. All the measurements varied widely among the apps—sensitivity ranged from 6.8% to 98.1%, for example, while specificity ranged from 30.4% to 93.7%.

Here’s the important thing: The three apps that failed to spot melanomas 30% of the time all use digital image analysis techniques. Computers decide whether a mole is cancerous or not. The fourth app actually sends images to a dermatologist, who replies with an evaluation in 24 hours. That fourth app correctly identified 52 out of 53 melanomas that researchers sent along. Such accuracy doesn’t come cheap—the app charges $5 per image, which is enough to drive many people towards the cheaper digital image analysis apps (they ranged from free to $4.99, with no individual charge per lesion).

This isn’t to say that all diagnostic apps can’t be trusted. AliveCor’s ECG heart monitor, which doubles as an iPhone case, was recently approved by the FDA, so the company’s claims that its monitor and app provide accurate results are almost certainly true (disclosure: I participated in a clinical trial for the monitor). But it’s worth taking unregulated apps with a grain of salt. Even though the skin cancer app designers warn that their products are for educational purposes only, it’s hard not to breathe a small sigh of relief upon getting the news, however unscientific, that a suspicious mole is benign.