Home / Technology / AI's Clinical Blind Spot: Hospitals Can't Measure Impact

AI's Clinical Blind Spot: Hospitals Can't Measure Impact

Summary

  • Few hospitals globally can assess AI's true clinical benefit.
  • Healthcare AI trained on Western data risks bias elsewhere.
  • AI's real-world performance often ignores lab accuracy scores.
AI's Clinical Blind Spot: Hospitals Can't Measure Impact

Few hospitals globally possess the capability to accurately assess the real-world clinical benefits of deployed artificial intelligence (AI) systems, according to Emmanuel Bacry, Scientific Director of France's Health Data Hub. This deficiency raises significant questions about the true clinical and economic value of healthcare AI technologies. Hospitals often lack the integrated patient information systems needed to track outcomes post-discharge, making it difficult to establish AI's tangible impact.

Bacry cautioned that AI systems' performance in laboratory settings, measured by accuracy scores against doctors, does not guarantee effectiveness in actual clinical environments. He cited an example where an AI alert system for kidney failure, despite functioning as designed, was linked to increased mortality in some US hospitals. This underscores the critical need to integrate AI into existing clinical workflows rather than treating it as an isolated tool.

Furthermore, Bacry pointed out a significant bias in healthcare AI development, with an estimated 70% of training data originating from Western, male populations. This results in AI models that are essentially 'Occidental AI,' potentially performing poorly or even disastrously when deployed in diverse global populations, particularly in developing countries. He advocated for smaller, locally adapted AI models trained on representative data.

Disclaimer: This story has been auto-aggregated and auto-summarised by a computer program. This story has not been edited or created by the Feedzop team.

Read more news on

Property Code: 5571