Summary of Results
This study included 1,288 COVID-19 positive hospitalised patients in the West-of-England. Patients were more likely to be male and older and these groups were also more likely to die, in line with other COVID-19 studies.9,14-16 Twenty-five percent died during their initial admission and a further 2% died during a subsequent admission. 10% of patients were admitted to ICU. The majority of first NEWS2 values were low (50% NEWS2=0-2, 27% NEWS2=3-4), even though these patients were sick enough to be hospitalised; this is important when considering what threshold to use as an admission trigger in primary and community care. However, many patients went on to have high scores at some point during their hospital stay (17% maximum NEWS2=5-6, 47% maximum NEWS2=7+). Patients with higher first NEWS2 values were more likely to require ICU admission and/or die, in line with findings from non-COVID-19, mostly pre-hospital, populations.17-19 Only 4% of patients with maximum NEWS2=0-2 and 7% with maximum NEWS2=3-6 required ICU or died, compared to 47% of patients with maximum NEWS2=7+. LOS for survivors increased as first and maximum NEWS2 increased. The AUC for 2-day mortality was 0.77, reducing to 0.70 and 0.65 for 7-day and 30-day mortality; these AUCs were considered acceptable at 2- and 7-days according to our predefined criteria13, and support previous research suggesting early warning scores are best at predicting short-term outcomes.20,21 An increase in most first and maximum component scores was associated with an increased risk of hospital mortality, although this relationship was less clear for temperature and systolic blood pressure. For 2-day mortality, only oxygen requirement had an AUC which met the threshold for acceptable (0.71); oxygen saturation and respiratory rate were marginally predictive (AUC 0.65 and 0.64, respectively), but the other components alone were not predictive at all (AUCs 0.53-0.56).
Strengths and limitations
The main strength of this paper is the inclusion of all patients with a positive COVID-19 diagnosis over a period of more than three months admitted to four hospitals encompassing communities of varying ages and deprivation. This enables our findings to be generalisable across the UK. A key limitation was the absence of electronic observations in ICU, three EDs, or the respiratory admission unit in Bath. This meant that 17% of patients were not included in the analysis of first scores, 10% were not included in maximum score analysis, and maximum scores may not have been true maximums. We would have liked to look at pre-hospital NEWS2, however without data linkage this would have meant relying on suspected COVID-19 status rather than confirmed. In addition, remote GP consultation meant that observations, and therefore NEWS2, in primary care data is often incomplete. We therefore focussed on hospital data for confirmed COVID-19 patients. A limitation of this approach was that 45% of patients scored for supplemental oxygen on first score which does not reflect the situation in primary care. However, it is likely that the score for supplemental oxygen would be replaced by a score for hypoxia in the community.
Comparison with other literature
NEWS2 in COVID-19 patients
There are a few small studies looking at NEWS2 in patients with COVID-19, but none in UK populations. A study of 66 patients in Norway22 admitted to ED found that NEWS2 predicted outcomes and performed better than other commonly used clinical risk scores qSOFA, SIRS and CRB-65. 42% of patients had NEWS2≥5 at presentation compared to 23% with NEWS≥5 in our study. This is probably because many of our patients were already receiving treatment when their first score was recorded. A study of 68 admissions to an Italian hospital23 confirmed our finding that ICU admission was more likely for patients with higher NEWS2.
A group in China described how they added 3 points to NEWS for age >65 to predict severe COVID-19 disease, but did not report any outcome measures.24 Similarly, a UK group wrote a letter to the editor of BMJ stating that NEWS2 requires modification for COVID-19 patients, to account for oxygen demand rather than a binary component for oxygen delivery.25 The RCP have released revised guidance making staff aware that an increase in oxygen requirements should be a trigger for re-assessment.
Component scores in COVID-19 patients
Hypoxia (low oxygen saturation) has been shown to predict COVID-19 mortality in other studies, for example, Xie et al found that higher SpO2 levels after oxygen supplementation were associated with reduced mortality, independent of age and sex.16,22,26 High respiratory rate has also been found to predict poor outcomes in this population27, as has low systolic blood pressure.28
Concerns have emerged regarding ‘silent hypoxia’, where a patient has low oxygen saturations but normal respiratory rate.29 We found 57/1,071 first scores had low oxygen saturation (scoring 3 for that NEWS2 component), and of those 28/57 (49%) had a normal respiratory rate (12-20 breath per minute); this was more common for patients without supplemental oxygen (14/23, 61%) that with oxygen (14/34, 41%). Therefore, silent hypoxia did exist in these patients, but numbers were small.
Pyrexia (high temperature) is one of the key COVID-19 symptoms described by the NHS30, but only 19% of our COVID-19 positive patients had a temperature >38 degrees at first score. Other authors9 found pyrexia to occur in only 30% of patients, but it has been reported in up to 79%.22
Implications for research and/or practice
An evidence review of NEWS2 and COVID-1912 raised three research questions for the use of NEWS2 in primary care. The first was whether NEWS2 is valid as a measure of severity in COVID-19 and does it predict who is likely to deteriorate? We have demonstrated that NEWS2 predicts mortality, especially in the short term. The second question was whether a single NEWS2 score is sufficiently sensitive and specific? Again, we have shown that a single score can predict short-term mortality, and based on the AUC, NEWS2=4 is the best value to balance sensitivity and specificity. A low score suggests that mortality is unlikely in the subsequent 2 days but over time scores deteriorate in many patients, so the use of serial scores is likely to be superior to a single score. In either case, NEWS2 should always be used alongside clinical judgement and not as a rule in/rule out test. The final question was whether NEWS is likely to be practical? Some components are measurable at home, but blood pressure and oxygen saturation require equipment. We found blood pressure alone was a poor predictor of mortality in the short term, but that oxygen saturations and respiratory rate are the most predictive, as supported by other studies. This suggests that the proposed role out of pulse oximeters for remote monitoring4,31 is likely to be valuable.
Further research measuring NEWS2 in COVID-19 patients in primary and community care is required. However, given that COVID-19 status may not be confirmed or recorded in the community, a retrospective study is likely to be difficult. NEWS2 is mandated in the UK acute sector, and used widely in primary and community care, so a randomised controlled trial is also unlikely to be possible. Linking ambulance data to hospital data would provide some of the picture, but obtaining linked data is often difficult and slow. Given these difficulties, we believe the evidence presented in this study can be applied in primary/community care settings, despite being collected in hospitalised patients.
This study has demonstrated that NEWS2 is associated with mortality in COVID-19 patients. In particular, NEWS2 is a reasonably good predictor of 2-day mortality in this population, and the respiratory components (respiratory rate, oxygen saturation and supplemental oxygen requirement) are of the most value in the short term. These findings support the RCP’s recommendations to use NEWS2, alongside clinical judgement, in the assessment of COVID-19 patients.