National Early Warning Score 2 and laboratory predictors correlate with clinical deterioration in hospitalized patients with COVID-19

Aim: We aimed to determine the prognostic values of the National Early Warning Score 2 (NEWS2) and laboratory parameters during the first week of COVID-19. Materials & methods: All adult patients who were hospitalized for confirmed COVID-19 between 11 March and 11 May 2020 were retrospectively included. Results: Overall, 611 patients were included. Our results showed that NEWS2, procalcitonin, neutrophil/lymphocyte ratio and albumin at D0, D3, D5 and D7 were the best predictors for clinical deterioration defined as a composite of ICU admission during hospitalization or in-hospital death. Procalcitonin had the highest odds ratio for clinical deterioration on all days. Conclusion: This study provides a list of several laboratory parameters correlated with NEWS2 and potential predictors for clinical deterioration in patients with COVID-19.


Statistical analysis
Quantitative variables are expressed as mean and standard deviation when they contain continuous and normal distributed data. When the data were not distributed normally, median and interquartile range (IQR) were used. When they contained categorical data, they were expressed as a percentage (%) and frequency (n). Comparison of qualitative variables was performed by Pearson's Chi-square test. The normal distribution questioning the necessity of using the parametric test was examined by Kolmogrov-Smirnov, Shapiro-Wilk, Kurtosis-Skewness tests and box plot distribution. When normally distributed data could not be determined, nonparametric tests and spearman correlation were used. Kruskal-Wallis test was used for the analysis of continuous and more than two independent nonparametric groups (Bonferroni correction was used when necessary) and Mann-Whitney test was used for post hoc analysis. To evaluate the factors in prognosis which were admission ICU and in-hospital death, univariate logistic regression analysis was performed. Afterward, these dependent groups were handled one by one, receiver operating characteristic (ROC) curves were drawn and cut-off values, sensitivity and specificity and area under the curve (AUC) were demonstrated. To predict clinical deterioration, the prognostic accuracy of NEWS2 and laboratory parameters at D0, D3, D5 and D7 was evaluated by ROC analyses.
• NEWS2 and laboratory parameters at D0; • NEWS2 and laboratory parameters at D3 after excluding from the analysis patients with clinical deterioration within the first 3 days of hospitalization; • NEWS2 and laboratory parameters at D5 after excluding from the analysis patients with clinical deterioration within the first 5 days of hospitalization; • NEWS2 and laboratory parameters at D7 after excluding from the analysis patients with clinical deterioration within the first 7 days of hospitalization.
Additionally, the association of the parameters at admission with the clinical deterioration was evaluated to predict the 3-day, 5-day and 7-day end points by ROC analyses.
The results were evaluated in 95% CI and statistical significance level was defined as p < 0.05. The analyzes were performed using IBM SPSS-21 (Statistical Package for Social Sciences, IL, USA).

General characteristics
Overall, 611 patients were included. Of whom, 329 (53.8%) were male, the mean age was 52.53 ± 15.07 years. Seventy-three patients (11.9%) were admitted to the ICU. In-hospital death occurred in 46 (7.5%) patients. Among 73 patients (11.9%) admitted to the ICU, 40 patients (54.8%) died during hospitalization. Clinical deterioration was observed in 79 patients (12.9%) during hospitalization, 36 (5.9%), during the first 3 days, 54 (8.8%) during the first 5 days and 62 (10.1%) during the first week of hospitalization. NEWS2 was calculated at D0, D3, D5 and D7 of hospitalization. Patients were stratified into three risk groups: low risk from zero to four; medium risk from five to six and high risk above seven. Of 611 patients, 505 (82.7%) at D0, 411 (91.9%) at D3, 375 (92.2%) at D5, 284 (93.8%) at D7 had a NEWS2 score <7. The median length of hospital stay was 8.9 days and 332 patients (54.3%) who did not have fever and did not need oxygen in the last 48-72 h and meet the criteria for home monitoring were discharged within the first 7 days. Demographic characteristics of hospitalized patients with COVID-19 are available in Supplementary Table 1.

ICU admission or in-hospital mortality
The parameters associated with admission ICU and in-hospital death at D0, D3, D5 and D7 were NEWS2, lymphocyte count, neutrophil count, platelet count, NLR, PLR, CRP, procalcitonin, D-dimer, troponin, AST, urea, lactate dehydrogenase (LDH) and albumin. The median and IQR values of the laboratory parameters and NEWS2 are represented in Table 1.
Additionally, we analyzed the association of NEWS2, procalcitonin, albumin and NLR at admission with the end point events at D3, D5 and D7. Albumin was the best predictor for the 3-day end-point. AUC was 0.849 (95% CI: 0.755-0.944), sensitivity of 84.1% and specificity of 72.2% (p < 0.001). Procalcitonin was the best predictor for the 5-day and 7-day end points. AUC was 0.835 (95% CI: 0.748-0.821), with a sensitivity of 79.2 and specificity of 77.1% (p < 0.001) for the 5

Discussion
In this study, we presented a detailed analysis of the NEWS2 score and laboratory parameters in hospitalized patients with COVID-19. Our results showed that NEWS2, procalcitonin, NLR and albumin at D0, D3, D5, and D7 were the best predictors for clinical deterioration (ICU admission or in-hospital death  Early and accurate discrimination of need for ICU improves the clinical course of COVID-19 and reduce unnecessary use of ICU beds. There are several published studies on the use of NEWS2 in COVID-19 patients [2][3][4][7][8][9][10][11][12][13]. However, most studies evaluate NEWS2 at admission only [11][12][13]. In the study of Sze et al., they suggested that NEWS2 score was not a valuable tool to predict clinical deterioration in elderly patients with COVID-19 [14]. However, they reported the results of only 17 elderly patients. Kim et al. showed that NEWS2 scores on D0 significantly differed in noncritical and critical patients (2.6 ± 2.6 vs 8.2 ± 3.3; p < 0.001) [9]. In the study of Volff et al. the AUC value of NEWS score to predict ICU admission or death was 0.74, in consistent with our result. However, the AUC values of NEWS2 at D3, D5 and D7 were 0.798, 0.833 and 0.842 whereas NEWS2 at D0 was not accurate (AUC <0.750) [7]. Similarly, Sixt et al. showed that AUC values of NEWS2 was 0.74 at D0, with a best cutoff of six and was 0.98 at D7, with a best cut off of seven. They reported high sensitivity and specificity at D7 (92 and 97%, respectively) [8].
In the study of Shang et al. Spearman's rank correlation analysis revealed that leukocyte, neutrophil, CRP, procalcitonin and LDH were positively correlated and albumin was negatively correlated with mortality in patients with receiving maintenance hemodialysis. Additionally, they showed that CRP had the highest AUC value (0.895) and the values of AUC of neutrophil count, LDH, leukocyte, albumin and procalcitonin were 0.813, 0.758, 0.757, 0.743 and 0.728, respectively [20]. In contrast, we found that procalcitonin was the best predictor for clinical deterioration our study. The optimal cut-off value of procalcitonin at D0, D3, D5 and D7 were 0.065, 0.125, 0.155 and 0.120 ng/ml and the sensitivity and specificity to predict clinical deterioration were 81.1 and 67.9% on D0, 80.6 and 89.3% on D3, 87.5 and 96.7% on D5 and 85.7 and 88.2% on D7, respectively. Procalcitonin is not well studied for COVID-19 cases. However, some studies suggested that increased procalcitonin levels were found to be associated with the disease severity in patients with COVID-19. A meta-analysis future science group www.futuremedicine.com  showed that severe patients with COVID-19 had increased procalcitonin levels [18,19]. Similarly, we found that procalcitonin was the best prognostic parameter for the clinical deterioration in our study. Elevated procalcitonin levels could be associated with acute secondary bacterial pneumonia or systemic secondary bacterial infection in patients with COVID-19 due to the production and release into the circulation from procalcitonin-producing extrathyroidal tissues [21]. In our study, despite elevated procalcitonin levels, this elevation was limited (the cut- off values were less than 0.5 ng/ml). In a previous study by Xu et al., they suggested that a limited increase in procalcitonin levels (cut-off value = 0.1 ng/ml) could be associated with increased IFN-γ [16].
Low-serum albumin levels in studies with COVID-19 patients are suggested to be associated with an increased risk of mortality [22][23][24][25]. In consistent with other studies, our results confirm that albumin is a valuable predictor for ICU admission or in-hospital death. Albumin is a negative acute phase reactant produced in the liver, and causes downregulation of the expression of angiotensin-converting enzyme-2 receptors, which play a role in the cell entry mechanism of SARS-CoV-2. Liu et al. reported that albumin was associated with clinical deterioration and significantly higher in patients with the improvement/stabilization than in those with disease progression (36.62 ± 6.60 vs 41.27 ± 4.55 g/l, p = 0.006) [24]  3.50 g/dl (CI: 3.26-3.74 g/dl) in the severe group and 4.05 g/dl (CI: 3.82-4.27 g/dl) in the nonsevere group (p < 0.001). They reported that hypoalbuminemia was associated with 12.6-fold increased risk of mortality [25]. An increase in neutrophils and a decrease in lymphocytes have been found in various studies. Some studies have shown that NLR may be an important indicator for the severity of COVID-19 patients. Yan et al. showed that NLR was significantly correlated with all-cause in-hospital mortality (OR: 44.351; 95% CI: 4.627-425.088) [26]. The NLR reflects the balance between the innate and adaptive immune systems [26] and increased NLR levels were found to be associated with clinical deterioration in COVID-19 [15][16][17][18]26].
This study has several limitations. First, it was retrospectively conducted in a single center. Second, this study had a small sample size and a control group was not included. The generalizability of our results may be limited. Thus,   we need new large scale studies providing important information to better understand COVID-19 pandemic. Our study has also several strengths. First, we were able to admit all critically ill patients requiring intensive care to the ICU during the first months of pandemic. This prevents a selection bias. Second, longitudinal evaluation of the association between clinical deterioration and the dynamic changes of laboratory parameters was performed, since we regularly monitored laboratory parameters during the clinical course.

Conclusion
This study provides a list of several laboratory parameters correlated with NEWS2 and potential predictors for ICU admission or in-hospital death during the clinical course of COVID-19. NEWS2, procalcitonin, NLR and albumin have a high accuracy to predict clinical outcomes/disease progression in hospitalized patients and should future science group www.futuremedicine.com be considered in the clinical decision of ICU admission. In conclusion, dynamic monitoring of NEWS2 and laboratory parameters is vital for improving clinical outcomes.

Supplementary data
To view the supplementary data that accompany this paper please visit the journal website at: www.futuremedicine.com/doi/suppl/10.2217/bmm-2021-0061 G Tuncer proposed the concept, designed the study, wrote the protocol and managed the study. G Tuncer, S Surme, IY Nakir and M Yazla, performed the statistics, interpreted the data and wrote the manuscript. S Surme, A Buyukyazgan, AK Cinar, B Copur, E Zerdali, G Tuncer, H Balli, IY Nakir, M Yazla, Y Kurekci were involved in collecting the data. MM Sonmez, S S-Yavuz, F Pehlivanoglu, G Sengoz performed a critical review of the manuscript. All authors provided inputs for revision of the manuscript. S Surme communicated with the journal and addressed comments from reviewers. All authors contributed to data acquisition, data analysis or data interpretation and reviewed and approved the final version.