Machine Learning for Clinical Decision Support in the Neonatal Intensive Care Unit

The neonatal intensive care unit (NICU) is a data-rich environment that is an ideal setting for the implementation of machine learning (ML) and artificial intelligence (AI) in clinical decision support (CDS). Despite their potential, ML and AI applications are rarely used in clinical practice because of infrastructure and technical limitations. In this article, we review the technical requirements for data acquisition solutions, storage, and processing needed to handle the varied sources of data generated by hospitalized newborns. In addition, we describe the challenges for integrating structured and unstructured data from electronic health records, bedside monitors, imaging, and other sources and we consider the ethical and legal implications of using ML and AI for CDS. Finally, we emphasize that the study and application of ML and AI models in CDS requires rigorous research and quality improvement methodology. The NICUs that realize the potential of ML and AI in quality improvement and clinical research applications will be uniquely positioned to apply their findings to improve neonatal outcomes.