MICA: A partnership to extend the research utility of a source of real-world health data, the UK National Neonatal Research Database

Aim: Our aim is to improve the usefulness of an established UK resource, the National Neonatal Research Database (NNRD) for parents and for researchers so that they can view data and conduct studies to improve the care of preterm and sick newborn babies more quickly, efficiently and at lower cost than presently. About 1 in 7 (100,000 each year) newborn babies is admitted to a NHS neonatal unit. Neonatal problems and the care received, affect life-long health and well-being.

Background: We established the NNRD, a unique, award-winning resource, in collaboration with parents, doctors, nurses, other healthcare professionals, and researchers to improve care, treatments and outcomes for preterm and sick babies admitted to NHS neonatal units. The NNRD contains comprehensive data, updated quarterly, from the electronic medical notes of all babies admitted to NHS neonatal units in England, Scotland and Wales. Imperial College London hosts the NNRD securely on a computer server. No data that can identify any individual baby are included. The data include details of diseases, daily treatments and outcomes. To-date the NNRD has information on about one million babies; around 25,000 new babies are added each quarter.

Why this work is needed: We established the NNRD because a key challenge in newborn care is the need for up-to-date, timely and accurate data for research to improve, evaluate and develop new treatments. Data are required for all of the many types of studies needed, such as improving understanding of diseases, their causes and the care provided, and to develop new medicines. Different types of studies often need similar data (e.g. age, sex, weight, disease) but traditionally, researchers collect these again and again for each new purpose. This is expensive, wastes time and increases the risk of errors. Studies can fail because data availability or quality are poor. Data also need to be up-to-date otherwise information may be misleading. For example, information on health outcomes of very preterm babies that are widely used in the UK to counsel parents and guide clinical practice was derived from research conducted over 20 years ago and no longer reflects circumstances today. The NNRD provides a single source of up-to-date data for research and other purposes. This work is needed to improve the NNRD and make it more useful.

Our objectives, and what we will do: We will automate processes to check accuracy and add new data into the NNRD that we currently perform manually. At present anyone who wants to use the NNRD must request our assistance, which inevitably incurs a delay. We will identify common types of information that researchers, parents and clinicians would find useful to obtain from the NNRD. This might be to determine the number of patients with particular conditions that are admitted to neonatal units. We will also obtain views on the way in which they would like to see the results (e.g. tables or graphs). This will help us develop web-based tools to enable parents and researchers to answer common questions themselves. We will make these tools available on our website. We will also develop ways to process NNRD data so that we can apply new techniques that can help identify patterns such as where particular types of disease occur and provide clues to their causes. Additionally, we will train young scientists in handling complex health data.

Why this partnership is needed: We have formed a partnership because our objectives require skills across different organisations and disciplines. Our partnership brings clinical neonatologists, academic researchers and data scientists together with the national information technology lead for the health and care system in England (NHS Digital), expertise in data tools (Strategic Intelligence Alliance for Health; SIA), the national charity for preterm and sick newborn babies (Bliss) and the national institute for health data, Health Data Research (UK HDR-UK).

The National Neonatal Research Database (NNRD) is a dynamic relational database of extracts from the electronic patient records (EPR) of NHS neonatal unit admissions. We receive quarterly extracts over secure NHS N3 environment as a password protected .bak file of multiple tables reflecting groups of EPR entry screens, and identifier codes reflecting each table and episode of care. Neonatal care is delivered in a network model hence babies usually have multiple episodes across different hospitals. We remove identifiers and assign pseudonymised codes. We identify duplicate, missing, internally inconsistent and out-of-range entries. We check episode temporal consistency and hospital, provider and network codes, notifying neonatal units of potentially erroneous or missing entries. As each extract contains data on babies born within the last 5 years, corrections are incorporated progressively into the NNRD. With NHS Digital, we will automate the majority of these processes and publish detailed technical specifications. Supported by Bliss, we will identify and prioritise areas that researchers from different backgrounds, and parents, might wish to interrogate the NNRD (e.g. event rates, patient numbers, outcomes). The NNRD is managed in a MS SQL Server environment, hence can be used for automated reporting directly from data subsets. With SIA we will develop on-line analytical processing tools to enable standard queries to be run rapidly by researchers and parents. We will use applications such as IBM Cognos, MicroStrategy and Board MIT, basing initial development upon simulated data. We will also develop approaches to curate NNRD for future application of data-driven techniques such as selection of normalisation and clustering techniques (e.g. K-means, Dirichlet Process Gaussian Mixture Modelling), development of "smart" queries based on commonly utilised derived measures (e.g. birth weight Z-scores) and modelling time-dependent concepts (e.g. feeding patterns).

The work we propose will have wide, substantial and immediate impact on academic research groups and industry sponsors in the UK and around the world, patients and families, clinicians, research funders, policy-makers and the global standing and competitiveness of the UK in relation to use of real-world health data i.e. data derived from routine sources.

The partnership will transform and scale-up technical infrastructure and improve the capabilities of researchers to access and interrogate the NNRD. This will reduce burden of data collection and the costs of research and improve the speed and efficiency of studies. The evidence-base for much of present-day neonatal care is limited, and over 90% of neonatal medications remain off-label or off-license. There would be considerable positive impact on patient care and outcomes from ability to generate new evidence more rapidly.

There is considerable global interest in the use of real-world data, The NNRD is a source of real-world data and already supports many types of studies such as observational research to identify disease determinants, health services research to improve the organisation and delivery of care, pragmatic clinical trials and epidemiological research. The outcome of this partnership will strengthen the use of routine real-world health data to improve newborn care and advance the world-leading position of the UK in this field.

There will be further impact through added value to current parallel work by the applicants. We are part of a newly formed initiative, the ALPHA Collaboration, led by the University of Sydney Clinical Trials Unit to deliver large international randomised controlled trials in perinatal medicine using routine data. We are also collaborating with Shanghai Fudan University to develop a China Neonatal Research Database modelled on the NNRD. Modi is co-director of the eNewborn network of European neonatal units sharing electronic health data for bench-marking and research.

Industry is increasingly interested in utilising real-world data. The lead applicant, NM, is a member of the International Neonatal Consortium, a US Food and Drug Administration funded initiative to accelerate the pace of development of medicines for newborn babies that includes developing the potential of real-world data through creation of standards for data definitions, measurements and methods. NM is also a member of C4C (connect4children) expert groups. C4C is a newly formed European network with 33 academic and 10 industry partners to facilitate the development of medicines for infants and children. C4C aims to bring innovative approaches, including use of real-world data to deliver fast and efficient clinical trials. The NNRD is already being used for an industry sponsored study to determine the optimum dose of a neonatal medication. Hence, the UK is poised to achieve high impact through the NNRD.

The NNRD is already impacting UK health policy; e.g. as sole data source for the UK National Neonatal Audit Programme, through use in the Department of Health Maternal and Neonatal Health and Care Policy Research Unit in which Knight, Modi and Gale are Co-I, the National Maternity and Perinatal Audit, and Maternity Transformation Programme. The outcomes of this partnership will accelerate opportunities for wider policy impact through more rapid availability of quality assured, timely data, and new tools enabling direct interrogation by researchers.

Health informatics is a multi-disciplinary field that aims to improve the safety, delivery and effectiveness of healthcare through use of technology and analysis of health data. However, delivering insights is bottle-necked due to the shortage of relevant skills. This partnership will grow capacity in health informatics and data science. We will publish our approach in high-impact peer-reviewed journals which will facilitate adoption by other specialities in the UK, and research groups world-wide.