Toggle Nav

MelaTools-ML preferences study

We performed a discrete choice experiment (DCE) using online surveys with patients (425), members of the public (1577), and GPs (300). DCEs present participants with a choice of hypothetical options and ask them to chosse which they would prefer. In our DCE we asked participants to choose between two hypothetical AI technologies designed to help detect skin cancer. We described the technologies in terms of 6 attributes:

  1. How many skin cancers they miss
  2. The number of people who are told they have skin cancer when they don’t have a skin cancer
  3. The cost of using the technology
  4. Whether the technology is used by the patient at home or a GP or nurse in the GP surgery
  5. Whether the technology has been developed to work in all skin tones
  6. Whether using the technology has been recommended in guidelines

All attributes significantly influenced which AI technology the respondent’s chose. How many skin cancers the AI technologies missed was the most important attribute, followed by skin tones on which the technology had been developed to work in. The number of skin cancers missed was more important for GPs than for other groups.. Participants preferred AI technologies used by healthcare professionals in a GP surgery over AI technologies that patients could use at home. Ranking the attributes by their importance to each group showed that the order of importance was the same for all groups.

When developing AI technologies and bringing them into clinical practice to help detect skin cancer, we should focus on missing as few skin cancers as possible. Developing AI technologies that work in patients of all skin tones is also essential. The preferences of stakeholder groups are important to consider and will be important in how successful the implementation of AI technologies will be.

Read the full publication: Jones et al. Preferences for the use of Artificial Intelligence (AI) technologies to help detect skin cancer in primary care settings: a UK-wide discrete choice experiment (DCE) (currently in submission)

MelaTools-ML preferences study