Back

Are current NEWS2 clinical response thresholds optimised for a general in-patient population?

Pankhurst, T.; Sapey, E.; Gyves, H.; Evison, F.; Gallier, S.; Gkoutos, G.; Ball, S.

2020-09-15 health informatics
10.1101/2020.09.12.20136804 medRxiv
Show abstract

BackgroundThe National Early Warning Score 2 (NEWS2) is mandated in acute hospital trusts in England. Assessment of the implications of this policy across an unselected in-patient population has been limited. ObjectiveEvaluate NEWS2 performance in an acute, in-patient, population by relating potential costs and benefits of specific alerting thresholds 24 hours prior to a composite outcome event (unplanned intensive care admission or death). MethodsAll in-patient spells between Nov 2018 - Jul 2019 in a single acute hospital in the UK were analysed. Standardised Early Warning Score(SEWS) and NEWS2 data acquisition was from the electronic health record (EHR). Existing SEWS alert thresholds were maintained. The performance of NEWS2 and SEWS threshold score against the composite outcome was assessed. A single clinical review cost ({euro}129.50) was used to model the whole system cost of triggered responses at different NEWS2 thresholds. ResultsIn patients [≥]24 hours post-admission, a mean daily rate of progression to the outcome event was 1.95/1000. An increase in alert threshold from NEWS2 [≥]5 to [≥]6 reduced the proportion that would trigger clinical review from 10.0% to 5.3% per day. This was associated with the false negative rate at threshold increasing from 1.13/1000 patients to 1.36/1000. A simple resource model allowing one triggered clinical response every 24 hours defined an incremental cost per patient benefiting ={euro}26,463, equating to 18 additional healthcare professionals per 1000 patients to deliver clinical response to an additional 0.23 patients/day benefitting. ConclusionThe low event rate across the whole in-patient population, moderate performance of a single NEWS2 score and associated resource requirements mean that in any resource limited setting, rules-based unmodified NEWS2 response thresholds may divert clinical resource and focus. Summary BoxO_ST_ABSWhat is already known on this topic?C_ST_ABSNEWS2 is mandated as an early warning score for all NHS acute hospitals in England. There are associated graded clinical response recommendations progressing to urgent clinical review at a NEWS2 score [≥]5. Evaluation of the clinical and health economic implications of applying these recommendations across a whole in-patient population has been limited. What this study addsThis is the first study to report NEWS2 alert thresholds in a longitudinal hospital cohort and to model the cost of implementing a key threshold NEWS2 [≥]5 for all in-patients. For in-patients [≥]24 hours post-admission, approximately 10% trigger at least one NEWS2 score [≥]5 per day (excluding those on intensive care or following a palliative care decision). The corresponding daily adverse event rate was 0.19% (admission to intensive care or death). The real world performance of current NEWS2 thresholds, using the observed classification performance, event rate and a response cost derived from the literature, identifies high incremental costs of increasing NEWS2 sensitivity.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.