A study, published in The BMJ, looked at predicting the risk of death in hospitalised COVID-19 patients.
Prof Duncan Young, Professor of Intensive Care Medicine, University of Oxford, said:
“This is a major piece of high-quality work from the ISARIC collaboration.
“They have used both conventional statistical and machine-learning approaches to calculate the probability of a patient dying after being hospitalised for a COVID-19 infection using information available at the beginning of their hospitalisation. The final product is a simple scoring system that can be used at the bedside to put a patient into one of several risk categories using only a few commonly available pieces of medical information about the patient. It is essentially an objective, numerical way to prognosticate.
“Scoring systems such as this can be used to calculate an expected mortality for all the patients a hospital admits with COVID and compare this with the actual mortality. Hospitals with actual mortalities far lower than expected can then be looked at in more detail to try and find which aspects of care might be causing the reduced mortality. A similar exercise has already been undertaken by ICNARC looking only at patients admitted to intensive care units with COVID.
“The unanswered question is how knowledge of the probability of death at the start of a patient’s hospital care will be used for the benefit of that individual patient. As an example, how do you translate knowledge of a mortality risk of 5% (1 in 20 risk of dying) for an individual patient into an action that will significantly reduce that patient’s risk below 5%? There are treatments that are limited to sub-groups of patients with COVID (ventilation, dexamethasone etc) but the indications for these treatments are already established and not dependent on an illness severity score.
“This was a COVID-specific score, and so was always likely to outperform (predict death more accurately) scores devised for other diseases or patient groups such as the SOFA, CURB-65, NEWS, Surgisphere etc when they were used on patients with COVID. It is reassuring it does so, but not surprising. The 4C score is only modestly better than the E-CURB65, a score derived for patients with pneumonia. This is likely because it uses very similar data for the prognostication.
“This score was derived using only data from patients treated in hospital. The BMJ press release says “……… patients with a 4C Mortality Score falling within the low risk groups might be suitable for management in the community”. The patients in this study had full hospital care with a mortality that resulted from this level of care. Changing the care model may well change the mortality, so switching care to the community might well result in a different, and likely higher, mortality than predicted if the same patients were treated in hospital.”
‘Risk stratification of patients admitted to hospital with covid-19 using the ISARIC WHO Clinical Characterisation Protocol: development and validation of the 4C Mortality Score’ by Knight et al was published in The BMJ on Wednesday 9th September at 23:30 UK time.
All our previous output on this subject can be seen at this weblink: