The usefulness of NEWS2 at day 7 of hospitalization in predicting COVID-19 evolution and as an early endpoint in therapeutic trials.

Dear Editor,

We read with interest in your journal the paper by Sze et al. who reports that National Early Warning Score 2 (NEWS2) in elderly patient is not useful to predict outcome during COVID-19.

However, NEWS2 has been described as a robust tool to predict outcome during sepsis and seems to predict evolution of COVID-19 upon admission to the hospital in other recently studies.3, 4, 5

Considering the evolution of the epidemic, there is a need for reliable tools to predict some evolution in particular with regard to hospitalized patients, in order to discharge more quickly the patients from medical wards to home.

Moreover, more than 2000 clinical trials have been registered regarding COVID-19. Most of them use an endpoint defined 28 days after inclusion or hospitalization, as recommended by the World Health Organization. This relatively long time needed to conduct these trials may lead to uncontrolled prescription of drugs with unproved efficacy. Thus, early endpoints predictive of long-term evolution may not only reduce the risk of lost to follow-up and protocol deviations, but could also allow knowing the results of the trials faster.

We evaluate if NEWS2 could predict outcome of COVID-19 at day 28 and look for the earliest time point of assessment which could be used as a strong surrogate marker of day 28 evolution.

All adult patients who were directly hospitalized in our department for a confirmed COVID-19 between the 27th of February and the 30th of April 2020 (during the first peak of the epidemic in France) were retrospectively included. For each patient, epidemiological, demographic, clinical, biological, radiological, therapeutic data, and outcomes were collected from medical records. NEWS2 was calculated at admission, at day (D) 7 and D14 of hospitalization. With the total score, patients were classified into 3 groups of risk: low risk from 0 to 4; medium risk from 5 to 6 and high risk above 7.

The occurrence of an unfavourable event (defined as death or transfer to ICU) during hospitalization was the primary outcome.

The diagnostic accuracy of the NEWS2 to predict unfavourable outcome at D28 was evaluated thanks to the area under receiver operating characteristic curve (AUROC) for:

NEWS2 at admission,

NEWS2 at D7 after excluding from the analysis patients with unfavourable outcome within the first seven days of hospitalization.

NEWS2 evolution at D7 (delta NEWS) after excluding from the analysis patients with unfavourable outcome within the first seven days of hospitalization.

Sensitivity and specificity, negative predictive value (NPV), positive predictive value (PPV) and positive and negative likelihood ratio (LR+ and LR-) were calculated for the best corresponding cut-off.

Overall, 222 patients were included. Characteristic of patient, NEWS2 value and NEWS2 class at admission, D7 and D14, are represented in Table 1

Table 1

Characteristics, NEWS2 value and NEWS2 class of patients hospitalized for COVID-19.

	Overall (n = 222)
Age, mean ±SD	70.1 (±17.1)
SexMale, n (%)	123 (55)
Serious COVID-19 risk factors, n (%)Age > 75 y.o, n (%)Diabetes mellitus, n (%)Cardiovascular diseases, n (%)Chronic respiratory diseases, n (%)BMI < 16 kg/m², n (%)BMI ≥ 30 kg/m², n (%)Pregnancy, n (%)Immunodepression, n (%)Cirrhosis, n (%)Chronic kidney failure, n (%)	176 (79)90 (41)51 (23)133 (60)30 (14)1 (0)34 (16)2 (1)19 (9)1 (0)18 (8)
Other past medical historyCurrent smoker, n (%)Chronic alcoholism, n (%)Depression, n (%)Dementia, n (%)	8 (4)13 (8)32 (14)48 (22)
Charlson's score, median (IQR)	4.0 (2.0–6.0)
Charlson's score without age, median (IQR)	1.0 (0.0–2.0)
Duration between first symptoms and admission at hospital, median (IQR)	6.0 (3.0–9.0)
Clinical signs before admissionFever, n (%)Cough, n (%)Dyspnea, n (%)Flu syndrome, n (%)Digestive disorders, n (%)Confusion, n (%)Anosmia, n (%)Dygueusia, n (%)	169 (76)146 (66)137 (62)82 (37)66 (30)37 (17)13 (6)17 (8)
Biological data at admissionLymphocytes, G/L, median (IQR)Fibrinogen, g/L, median (IQR)C-reactive protein, mg/L, median (IQR)Albumin, g/L, median (IQR)Prealbumin, g/L, median (IQR)Creatinine, µmol/L, median (IQR)Urea, mmol/L, median (IQR)Alanine aminotransferase, UI/L, median (IQR)	0.9 (0.6–1.4)6.0 (5.3–6.8)83.7 (37.7–127.0)26.0 (23.0–30.0)0.1 (0.1–0.1)76.0 (60.0–100.0)6.6 (4.8–10.4)34.0 (23.0–49.5)
Thoracic CT-scanNon evocative of COVID-19, n (%)Minimal, n (%)Moderate, n (%)Extent, n (%)Severe, n (%)Critical, n (%)	17 (10)19 (12)53 (32)48 (29)22 (13)5 (3)
Specific treatment studied for COVID, n (%)Hydroxychloroquine, n (%)Remdesivir, n (%)Lopinavir/ritonavir/Interferon Beta, n (%)Lopinavir/Ritonavir, n (%)Corticoids, n (%)Antibiotic therapy, n (%)	15 (7)6(40)4(27)3(20)2(13)7 (3)174 (78)
NEWS2 at admissionMedian (min, max)Class 1 of NEWS2, n (%)Class 2 of NEWS2, n (%)Class 3 of NEWS2, n (%)	Overall (n = 222)5.0 (0.0, 17.0)87 (39)50 (23)84 (38)
NEWS2 at D7Median (min, max)Class 1 of NEWS2, n (%)Class 2 of NEWS2, n (%)Class 3 of NEWS2, n (%)	Overall (n = 170)0.0 (0.0, 13.0)134 (79)18 (11)17 (10)
NEWS2 at D14Median (min, max)Class 1 of NEWS2, n (%)Class 2 of NEWS2, n (%)Class 3 of NEWS2, n (%)	Overall (n = 158)0.0 (0.0, 15.0)151 (95)6 (4)1 (1)

An unfavourable event was observed in 64 patients (29%), 52 during the first week and, 12 in the second week of hospitalization.

For the 170 patients who did not present with any unfavourable event within the first week of hospitalization, 154 (90%) had a NEWS2<7 at D7.

ROC curves of NEWS2 at admission and at D7 to predict unfavourable events within the first 28 days of hospitalization are shown in Fig. 1 . At D0, AUROC curve was 0.74, and the best cut-off of NEWS2 was 6. At D7, AUROC curve was 0.98, with a best cut-off of 7. For this latter cut-off, sensitivity, specificity, PPV, NPV, LR+ and LR- were respectively of 0.92, 0.97, 0.75, 0.99, 2.5 and 0.08. The AUROC curve of NEWS2 evolution at D7 (Delta NEWS) was 0.82, significantly lower than the area under the curve of NEWS2 at D7 (0.82 vs 0.98, p<0.005).

Fig. 1

Receiver Operating Characteristic (ROC) curve and performance value for the best cut off for: A: NEWS2 at admission using unfavourable event within the 28 first days as the gold standard B: NEWS2 at D7 using unfavourable event within the 28 first days as the gold standard and after excluding from the analysis patients with unfavourable outcomes within the first seven days of hospitalization. C: NEWS2 evolution at D7 (Delta NEWS) using unfavourable event within the 28 first days as the gold standard and after excluding from the analysis patients with unfavourable outcomes within the first seven days of hospitalization.

Herein, we showed in our study that a NEWS2 ≥ 6 at admission predicts unfavourable outcome with 76% sensitivity and 64% specificity, these results being close to those reported in other studies. , Compared to other scores specifically developed for COVID-19, , NEWS2 is an easily and rapidly applicable score. Furthermore, its performance reported in the literature seems to be superior to other usual scores like the quick Sequential Organ Failure Assessment and the Systemic Inflammatory Response Syndrome. ,

However, we observed that these performances of NEWS2 at admission are weaker than those at D7, with high sensitivity and specificity rates by using a threshold of 7. This indicates a close to 0 probability of unfavourable event for patients who did not present with any unfavourable event within the seven first days of hospitalization and who have a NEWS2<7 at this time. In our cohort, this latter threshold concerned 90% of the patients.

The assessment at NEWS2 at D7 could thus be useful in two ways.

First, the patients with a NEWS2<7 at this time could be confidently, safely, and more quickly discharged from medical wards towards follow-up care facilities, thus reducing the burden of hospital occupation and improving the turnover of the patients in a highly epidemic context.

Second, a NEWS2<7 at D7 combined with the lack of unfavourable outcome during the first week could be used as an important judgement criterion in therapeutic trials, as they were observed in many patients, and associated with a nearly non-existent likelihood of subsequent unfavourable evolution. It may thus reduce the time of participation in therapeutic trials, as well as the risk of loss to follow-up. The fact that no specific antiviral therapy with significant efficacy was used in our study strengthened the potential usefulness of such a surrogate marker in further trials on drugs with genuine clinical efficacy.

Last, the main characteristics of our cohort are close to those previously published, making our results likely to be observed in other settings as well. In addition, even though it could be advocated that the delay between contamination and hospitalization could impact on the observed results, the median delay between the onset of symptoms and hospitalization in our study (6 days) also correlates with that reported in other studies, as well as the median time elapsed from admission to unfavourable event (within the first two weeks).

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.