Mobile Health Apps Versus Dermatologists: Comparing Preferences for Skin Cancer Screening

Patients with generally higher mistrust of medical professionals are more likely to choose a mobile health (mHealth) smartphone application as opposed to a dermatologist for the purposes of skin cancer screening, according to new findings.¹

These findings and others were the result of a recent study—led by Susanne Gaube of UCL Global Business School for Health at the University College London—in which provider and user characteristics were evaluated in their effectiveness of influencing patient determination to seek screening for skin cancer by a mHealth app or by a dermatologist.

While mHealth apps have been shown to be somewhat helpful, their skillset is known to be below that of specialists. As such, use of mHealth apps is acknowledged by Gaube and colleagues as not being a substitute for a professional screening for skin cancer.²

“In ideal conditions, these algorithms have shown high sensitivity and reasonable specificity,” Gaube and colleagues wrote. “However, in natural settings, their performance is still below specialists’; therefore, using mHealth apps is not a substitute for a professional skin cancer. Nevertheless, since laypeople and GPs show even worse accuracy, mHealth apps might be valuable tools for supporting self-examination and indicating whether to follow up with a dermatologist.”

Background and Methods

The investigators enrolled participants from the US through the implementation of the online research platform 'Prolific,’ ensuring their sample was representative by setting quotas for subjects’ gender, age, ethnicity, and status of health insurance status based upon available US 2020 census data. The team’s approach to sampling did not have any emphasis on specific geographical regions.

Over the course of the main study, the research team determined that a minimum sample size of 1442 would achieve a power of 0.80 and an effect size of 0.09 for the average marginal component effects (AMCEs). At first, there were 1702 participants involved in the survey, though a variety of reasons led to a final sample of 1591 individuals included for the team’s analysis, with those who finished out the main study not being eligible for the replication study.

The investigators’ goal in their replication study was to recruit around 20% of the main study's sample size, and they determined a minimum of 283 subjects would be necessary to detect the significant effects with a power of 0.80 and an AMCE effect size of 0.20. Among the 317 subjects who initiated the replication survey, the team made several exclusions and ended up with a final sample of 308.

Within both of the research team’s samples, it was noted that over half of those included held a university degree (main: 57.5%, replication: 55.8%). Many had also never been given a skin cancer assessment (main: 69.5%, replication: 75.6%).

Additionally, a minority noted that they did not have health insurance (main: 10.0%, replication: 7.5%) and most reported having no history of any kind of skin cancer (main: 93.4%, replication: 93.5%). The team’s collection of data for the pre-registered research occurred in the period between April - May 2022 within the main study, and between December 2022 - March 2023 for the replication study.

The research design would also involve a forced-choice conjoint experiment that was then followed by a survey. The attributes of providers that were manipulated over the course of the investigators’ experiment included costs, type (mHealth app versus dermatologist), expertise, details of results, and privacy policies.

The team implemented a subsequent questionnaire for subjects which would assess many different user characteristics, some of which included attitudes towards AI technology, and levels of medical mistrust, and demographic data.

Findings

Overall, the investigators found their results had been comparable in both of the studies. They reported that subjects’ choice when it came to providers had played a major role, noting that dermatologists were generally preferred over the use of an mHealth app.

Other heavily influential elements to decisions were factors such as expertise, cost, and privacy policy. The research team’s evaluation of demographic subgroups indicated that subject preferences were shown to be somewhat stable across different genders, ages, and ethnic identities.

Individuals that were shown to have increased levels of medical mistrust were noted by the team as tending to favor the mHealth app option. Dermatologists were found by the investigators to have been given higher ratings for accuracy, trust, and quality, regardless of whether or not they were chosen.

“This research provides a comprehensive understanding of the factors influencing people's preferences and decisions between AI-enabled mHealth apps and dermatologists for skin cancer screenings,” they wrote. “Our findings provide valuable insights for technology developers, healthcare providers, and policymakers.”

References

1. ^{Gaube S, Biebl I, Engelmann MKM, Kleine AK, Lermer E. Comparing preferences for skin cancer screening: AI-enabled app vs dermatologist. Soc Sci Med. 2024 May;349:116871. doi: 10.1016/j.socscimed.2024.116871. Epub 2024 Apr 15. PMID: 38640741.}
2. ^{Freeman K, Dinnes J, Deeks JJ, et al. Algorithm based smartphone apps to assess risk of skin cancer in adults: systematic review of diagnostic accuracy studies. BMJ, 368 (2020), p. m127, 10.1136/bmj.m127.}