Artificial Intelligence Research as a Continuous Clinical Service

Advances in artificial intelligence (AI) have demonstrated potential in augmenting clinical decision-making and expanding access to care, particularly in resource-limited settings.1 Nevertheless, a persistent disconnect remains between reported research performance of AI systems and their effectiveness in real-world clinical practice. For example, while the COVID-19 pandemic precipitated a rapid increase in scientific publications, including numerous AI-based predictive models, systematic reviews have highlighted that many models, despite reporting high predictive performance, were inaccurate or had a high risk of bias.