This study explored the role of machine-learning based clinical decision support (CDS) algorithms to support (rather than replace) human decision-making and the impact on diagnostic performance. This systematic review of 37 studies found limited evidence that the use of machine learning-based CDS systems contributes to improved diagnostic performance among clinicians. Interobserver agreement, user feedback, and clinician override were the most commonly reported outcomes. The authors emphasize the importance of further evaluation of human-computer interaction.