Evaluating AI in Healthcare and Biomedicine

This book represents a comprehensive review of the use of artificial intelligence (AI) to enhance the safety, quality and efficiency of healthcare. It discusses the evidence supporting the effectiveness of AI, which remains limited, and examines the understanding of unintended consequences (e.g. algorithmic bias and impact on equity) and the consequences of contextual variations. While numerous early-stage AI applications have emerged, relatively little attention has been given to the challenges of adoption, implementation, scaling and ensuring long-term sustainability. Strategic decisions about AI implementation, whether at national or local levels, are often not based on robust evidence. This gap is further exacerbated by a lack of evaluation and evidence assessment skills among key decision-makers and implementers. The field of AI in healthcare and biomedicine is evolving rapidly. Traditional methods, such as randomized controlled trials, are often too slow to keep pace with the speed of technological innovation. As a result, observational studies have become a common method of assessment. Many existing evaluations, however, are of low quality and lack integrated assessments of value, impact and processes. They also do not routinely establish robust baselines against which progress can be measured. There is therefore a pressing need for developing new agile and adaptive approaches to evaluation. These must draw on innovative evaluation frameworks that address the increasing tension between the long timelines associated with robust evidence generation and the need for evidence-informed decision making in real-world settings. Evaluating AI in Healthcare and Biomedicine: Foundations for Evidence-Based Decision Making in the Digital Age is an interdisciplinary approach integrating various perspectives and the pragmatic application of best practice in evaluating AI applications, ensuring that evidence emerging from evaluations will inform evidence-based decision making and harness the potential of AI to accelerate the safe adoption of AI in healthcare. This book addresses this gap and helps to educate the next generation of decision makers and evaluators.