Multimodal prediction models add an extra data source—such as imaging or an assay—to standard clinical information. Health systems must then decide whether that added test is worth the cost and effort. The new study examines whether the research literature justifies those decisions.
The audit covered 177 studies that directly compared multimodal models with unimodal alternatives. The central finding is that claims of clinical value in this literature run ahead of the evidence. In other words, researchers are asserting benefits that the studies do not yet firmly establish.
The authors suggest this gap matters for practical adoption. Without robust comparative evidence, health systems cannot reliably judge whether the added data source improves predictions enough to justify its acquisition. The study does not say the models lack value, only that the current evidence base is not keeping pace with the claims.