Summary

Advice

The technology described in this briefing is SYNE‑COV. It is a cloud-based software designed to help manage COVID‑19 in hospitals. It does this by predicting the chance of people with COVID‑19 being admitted for intensive care, having mechanical ventilation, or dying while in hospital.
The innovative aspect is that it is a prediction tool using machine-learning algorithms that analyse data from electronic health records to help clinicians' decision making in real time.
The intended place in therapy would be alongside existing early warning scores in hospitals for monitoring deterioration in people with COVID‑19.
The main points from the evidence summarised in this briefing are from 1 published study and 1 unpublished study. The published study retrospectively analysed data from a cohort of people with COVID‑19 who were admitted to an NHS hospital trust (n=879). Machine-learning algorithms were used to predict the chance of 3 clinical outcomes. Results suggest that the best model using one of 3 algorithms showing area under the receiver operating characteristic scores were 0.76 for hospital mortality, 0.84 for intensive care admission and 0.87 for having mechanical ventilation. The unpublished validation study assessed the performance of the SYNE‑COV algorithm and reported the accuracy of risk prediction compared with existing clinical risk scores as a reference standard.
The key uncertainty around the evidence is that it is from only 1 peer reviewed study, which included anonymised patient data from 1 NHS hospital trust during a 5‑month period. Further large studies at multiple sites would be useful to validate the machine-learning algorithms, and to provide further evidence for predicting key clinical outcomes when managing COVID‑19. The experts considered there to be uncertainty around the potential benefits to clinical practice.
The cost of SYNE‑COV used in addition to standard care ranges from £18,000 to £30,000 per year, depending on the number of beds and patients in each NHS trust.