Availability: Data consortiums for sensitive health data should continue to maximize the volume of usable data that have already been collected but are sitting in servers across the globe. For example, the World Economic Forum’s Breaking Barriers to Health Data initiative advocated for a data consortium approach to help solve the shortage of genomic data available to researchers for diagnosing rare diseases.35 Providers and governments should accelerate data sharing through the Forum’s framework36 of outlining the problem, aligning on incentives, deploying a governance model and planning for long-term viability. Providers and payers, meanwhile, should make increased use of untapped data hiding in undigitized medical records. Mayo Clinic, a leading US-based research institution, has collaborated with the digital pathology company Pramana to digitize 5 million glass pathology slides.37 Interoperability: In its January 2023 report, Global Health and Healthcare Strategic Outlook: Shaping the Future of Health and Healthcare, the World Economic Forum identifies data interoperability as one of the major barriers to improving global health and healthcare outcomes.38 If data is to be collected responsibly to make it fit for AI, it must be harmonized across disparate geographies and organizations. Government must define data ownership policies and reinforce data security to encourage the sharing of valuable insights in an ecosystem where data security is a shared responsibility and data ownership varies by country and situation. This report recommends taking inspiration from the aviation industry’s global participation standards and practices or the payment industry’s data security standards, which codify data security standards and serve as a repository for sharing best practices. These standards have been implemented globally, sometimes in deep partnership with the private sector. Responsibility: AI has the power to solve an incredible variety of challenges. Technology companies, however, must be cautious and conscious about choosing the “right” algorithms – those with the potential to do the greatest good while limiting harm. From pulse oximeters misreading skin tones featuring different melanin concentrations to algorithms misdiagnosing non-white patients, there are countless examples of algorithmic misuse. Moreover, when data science teams aren’t representative of the populations their products aspire to serve, they can fail to spot potential sources of bias and their real-world implications. In 2021, the World Health Organization provided guidance on the Ethics and Governance of Artificial Intelligence for Health. The report highlights responsibility and accountability as critical ethical principles for using AI in healthcare.39 The World Economic Forum has also strongly advocated for an ethical and responsible approach to AI in prior reports across sectors such as Earning Digital Trust: Decision-Making for Trustworthy Technologies (2022) and The AI Governance Journey: Development and Opportunities (2021). Governments, payers, life sciences and technology companies must encourage design team diversity internally and with key partners and hold stakeholders accountable when algorithms create or perpetuate bias. Providers should acquire a basic level of AI fluency, starting with updating medical and nursing school curricula. If data is to be collected responsibly to make it fit for AI, it must be harmonized across disparate geographies and organizations. 5.2 Designing AI with adoption in mind Scaling Smart Solutions with AI in Health: Unlocking Impact on High-Potential Use Cases 30
RkJQdWJsaXNoZXIy NzQwMjQ=