The research, published today in BMC Medicine, is the first systematic large-scale evaluation of the UK National Early Warning Risk Score (NEWS) 2, a scoring system that is currently recommended for predicting severe Covid-19 outcomes in patients.
The research team at King’s College London – which included applied informatics researchers at NIHR ARC South London – found that NEWS2 had poor-to-moderate accuracy for identifying Covid-19 patients at risk of being transferred to intensive care units (ICUs) or dying after 14 days of hospitalisation. However, it was shown to be moderately successful at predicting short term severe outcomes (ICU or death at three days).
For people who are hospitalised with severe Covid-19, it is vital to quickly identify which patients may deteriorate and require transfer to an intensive care unit (ICU) for organ support or who may die. NEWS2 is an early warning score that combines physiological parameters such as respiration rate, oxygen saturation, blood pressure and temperature. NEWS2 is currently used almost universally in UK NHS Trusts to identify which patients are at risk of deteriorating early.
Researchers analysed data from 1,276 Covid-19 patients admitted to King’s College Hospital NHS Foundation Trust during the first wave of the pandemic in March-April 2020. The team then validated their models using data for more than 6,000 patients across eight other hospitals globally (five in the UK, one in Norway, and two in China). At all UK sites, around one third of patients with Covid-19 were transferred to ICU or died within 14 days of hospital admission.
The researchers evaluated how well patients’ NEWS2 scores measured at hospital admission predicted who would have severe Covid-19 outcomes, which means either being transferred to ICU or dying. In all UK sites, combining NEWS2 and age to predict outcomes showed moderate success in the short term (three days), but for poor-to-moderate success for medium-term (14 days) outcomes.
We have conducted the largest study to date evaluating the accuracy of NEWS2 for predicting medium-term Covid-19 outcomes. NEWS2 is widely used in UK NHS Trusts, but little is known about how well it can predict severe Covid-19 outcomes and so evaluating its accuracy is important as we look to improve patient care now and in future.
Dr Carr added: "By collecting data from nine hospitals globally, our results have robust external validation. We found consistency across sites both in the performance of NEWS2 alone, as well as for the supplemented model. In short, NEWS2 tended to have poor-to-moderate performance but was improved by adding common blood and physiological measures."
The research team used CogStack, an existing health information platform developed by NIHR Maudsley BRC and in use at King’s College Hospital NHS Foundation Trust, which allows for rapid extraction and processing of data from a patient’s electronic health records.
The CogStack platform allows us to extract information from deep within hospital records at King's College Hospital NHS Foundation Trust in real time in order to really explore complex questions such as this. This includes being able to extract information relating to co-morbidities, for example, that may only be mentioned in passing in the physician narrative.
Researchers found that accuracy in predicting severe outcomes was improved by considering routinely-collected blood and physiological parameters from patients including age, oxygen saturation and neutrophil count, which is an important infection-fighting immune system cell. In models that supplemented NEWS2 with these parameters the ability to predict severe outcomes was improved.
Dr James Teo, a consultant neurologist at King’s College Hospital, clinical director of data science, and lead of the CogStack platform, said: “Our results for the first time validates NEWS2, and shows how it could be improved by adding common blood and physiological parameters. Thankfully, this NHS scoring system is easy to adapt and implement in real-world clinical practice, compared to other complex risk-scoring models.”
The study was a collaboration between the National Institute for Health Research (NIHR) Maudsley Biomedical Research Centre (BRC) and the Medical Research Council (MRC). The research team included several applied informatics researchers at NIHR ARC South London.
Data analysed comes from patients admitted to: King’s College Hospital National Health Service (NHS) Foundation Trust, Guy’s and St Thomas’ NHS Foundation Trust Hospitals, University Hospitals Southampton, University Hospitals Bristol and Weston NHS Foundation Trust, University College London Hospitals, University Hospitals Birmingham, one hospital in Norway (Oslo University Hospital), and two hospitals in Wuhan, China (Wuhan Sixth Hospital and Taikang Tongji Hospital).