Reporting guidelines for clinical trials evaluating artificial intelligence interventions are needed

As of June 2019, more than 30 artificial intelligence (AI) algorithms have been approved by the US Food and Drug Administration (including those for the detection of diabetic retinopathy, stroke, brain hemorrhage and atrial fibrillation)1 and over 300 clinical trials have been registered at ClinicalTrials.gov These algorithms have the potential to transform healthcare, by offering earlier and more accurate diagnoses, providing novel insights for the understanding of diseases, enabling faster and more efficient service delivery and making medical care more available to those who really need it. Optimal reporting is key for evaluating the clinical utility of algorithms, for informing health policy and evidence-based recommendations and for preventing research waste2. Most AI interventions thus far, particularly diagnostic algorithms, have been evaluated only in the context of diagnostic accuracy. Although this initial validation stage is important, a demonstration of good diagnostic accuracy does not necessarily translate to improved patient outcomes.