The NNRD-AI is a version of the National Neonatal Research Database (NNRD) curated for machine learning and artificial intelligence applications. The NNRD is an award-winning resource, a dynamic relational database containing information extracted from the electronic patient records of babies admitted to NHS neonatal units in England, Wales and Scotland (Northern Ireland is currently addressing regulatory requirements for participation).

The NDAU team established the NNRD in 2007 as a resource to support clinical teams, managers, professional organisations, policy makers, and researchers who wish to evaluate and improve neonatal care and services. Recently, supported by an award from the Medical Research Council, and with the support of the National Institute for Health Research Imperial Biomedical Research Centre, the neonatal team and collaborating data scientists at the Institute for Translational Medicine and Therapeutics, Data Science Group at Imperial College London, created NNRD-AI.

The NNRD-AI is a subset of the full NNRD with around 200 baby variables, 100 daily variables and 450 additional aggregate variables. The guiding principle underpinning the creation of the NNRD-AI is to make available data that requires minimal input from domain experts. Raw electronic patient record data are heavily influenced by the collection process. Additional processing is required to construct higher-order data representations suitable for modelling and application of machine learning/artificial intelligence techniques. In NNRD-AI, we encode data as readily usable numeric and string variables, utilise imputation methods, derived from domain knowledge, to reduce missingness, identify and remove out of range values, and apply clinical consistency algorithms. A wide range of definitions of complex major neonatal morbidities (e.g. necrotising enterocolitis, bronchopulmonary dysplasia, retinopathy of prematurity), aggregations of daily data and clinically meaningful representations of anthropometric variables and treatments are also available.


If you are thinking of using the NNRD-AI, please follow the process shown and complete the HDR UK data access request on https://web.www.healthdatagateway.org/dataset/cf4c4419-7c10-4376-a246-77b0411f9928