Implementing artificial intelligence in chest diagnostics for lung disease: A mixed-methods evaluation

Area of Science:

• Radiology and Medical Imaging
• Health Informatics
• Artificial Intelligence in Healthcare

Background:

• Artificial intelligence (AI) tools are being deployed to enhance diagnostic capabilities in radiology.
• The UK's National Health Service (NHS) England invested significantly in AI for chest X-ray and CT diagnostics.
• There is limited understanding of AI implementation, staff experiences, and effectiveness in practice.

Purpose of the Study:

• To evaluate international evidence on AI tools in radiology.
• To assess the implementation of AI for chest diagnostics within England.
• To investigate methods for measuring the effectiveness and cost-effectiveness of AI in chest diagnostics.

Main Methods:

• A 10-month mixed-methods study combining a rapid scoping review and an empirical investigation.
• Empirical work involved staff interviews, observations, and documentary analysis across NHS trusts.
• Analysis employed rapid assessment procedures integrating qualitative, quantitative, and health economic approaches.

Main Results:

• A review of 114 articles revealed evidence gaps regarding real-world AI implementation and its broader impacts.
• AI implementation for chest diagnostics varied significantly, with only 24/66 trusts having deployed tools by November 2024.
• Key barriers included time, resources, and process navigation, while facilitators involved stakeholder engagement; data limitations impacted evaluation capacity.

Conclusions:

• AI tools can potentially support efficient chest diagnostics, but successful implementation hinges on adequate time, resources, stakeholder engagement, and simplified governance.
• Learning from past innovations suggests AI may not provide simple solutions as anticipated by policymakers.
• Addressing implementation, adaptation, sustainability, and impact on care requires further investigation in phase 2.