Principles for building confidence in artificial intelligence systems for healthcare

Abstract

The introduction of artificial intelligence technologies is a key area for the digital transformation of healthcare. It is believed that the use of AI-based software will revolutionize healthcare by improving clinical decision-making, digital image analysis, risk management, predicting and monitoring the correctness of prescribed treatment. At the same time, many publications state that there is currently a serious problem of insufficient data on the clinical efficacy and safety of AI products brought to the market. Manufacturers make claims about product efficacy, but few provide credible evidence. In this regard, over the past few years, regulators, supervisory authorities, experts and professional communities have intensified the discussion around the need for thoughtful and strict regulation of issues of responsible attitude to the development and application of AI technologies in healthcare and clinical medicine. Based on the analysis of various documents and recommendations, we defined 4 key principles for a responsible attitude to the development, application and supervision of the use of AI technologies in Russian healthcare: “Confidence in the manufacturer”, “Data confidence”, “Model confidence” and “Product confidence”. What is more, we proposed evaluation criteria and suggestions for the development of legal and technical regulation.