A Composite Index Predicts Disease Progression in Early Stages of COVID-19: a Propensity Score-matched Cohort Study

Background: Thus far, studies on COVID-19 have focused on the epidemiology of the disease and clinical characteristics of patients (14-19), as well as on the risk factors associated with mortality during hospitalization in critical COVID-19 cases. However, no research has been performed on the prediction of progression in patients in the early stages of the disease. The aim of this work was to identify the early predictors of COVID-19 progression. Methods: The study included 338 patients with COVID-19 treated at two hospitals in Wuhan, Chian, from December, 2019 to March, 2020. Predictors of the progression of COVID-19 from mild to severe stages were selected by the logistic regression analysis. The predictive accuracy was evaluated further in the propensity score-matched cohort. Results: COVID-19 progression to severe and critical stages was conrmed in 78(23.1%) patients. The average value of the neutrophil-to-lymphocyte ratio (NLR) was higher in patients in the disease progression group than in the improvement group. Multivariable logistic regression analysis revealed that elevated NLR, LDH, and IL-10, were independent predictors of disease progression. The optimal cut-off value of NLR for predicting the progression of COVID-19 was 3.75. In the propensity score-matched cohort, NLR ≥ 3.75 was still an independent predictor of COVID-19 progression after multivariate analysis. Conclusions: The performed analysis demonstrates that NLR quali ﬁ es as an independent predictor of disease progression in COVID-19 patients at the early stage of the disease. The combined evaluation of NLR and LDH improved the accuracy of the prediction of COVID-19 progression. Assessment of predictors might facilitate early identication of COVID-19 patients at high risk for disease progression and ensure timely administration of appropriate treatment to prevent mild cases from becoming severe. propensity urea erythrocyte TNFα: tumor PPV:

Wuhan, patients with severe or critical COVID-19 were admitted to a hospital, while patients with mild or general COVID-19 were advised to be isolated at home. With increased awareness of the spread and progression of COVID-19, efforts have been made throughout China to admit and treat all con rmed or suspected COVID-19 patients. It has been reported that patients in Guangzhou, China, had a better prognosis than in Wuhan. Patients in Wuhan were more likely to be admitted to an ICU and had higher mortality (5). This difference highlights the importance of appropriate care and treatment in early mild cases, especially in patients at high risk of disease progression. Early detection of risk factors for progression to severe illness can help to provide adequate supportive care and treatment, decreasing the number of patients that develop the severe condition, reducing mortality, and alleviating the shortages of medical resources.
The neutrophil-to-lymphocyte ratio (NLR) and systemic immune-in ammation index (SII), both easily calculated from a routine blood test, are indicators of in ammation and immune response (2,6). Elevated NLR represents a risk factor for mortality not only from infectious diseases, but also from malignancies, intracerebral hemorrhage, and dermatomyositis (7)(8)(9). Similarly, increased SII has a prognostic value in a variety of malignancies, including hepatocellular carcinoma, breast cancer, and esophageal squamous cell carcinoma (10)(11)(12). It has been documented that severe and critical COVID-19 cases tend to have higher neutrophil counts and lower lymphocyte counts (1,13). Whether NLR or SII could be an independent predictor of COVID-19 progression in the early (mild or moderate) stages of the disease remains to be determined.
Thus far, studies on COVID-19 have focused on the epidemiology of the disease and clinical characteristics of patients (14)(15)(16)(17)(18)(19), as well as on the risk factors associated with mortality during hospitalization in critical COVID-19 cases. However, no research has been performed on the prediction of progression in patients in the early stages of the disease. To address this lack of data, the present research was designed to determine whether baseline values of NLR and SII in patients with mild or moderate COVID-19 can be used as reliable predictors of the progression of the disease in its early stages.

Study design and participants
The present investigation was designed as a retrospective study involving a total of 476 COVID-19 patients admitted consecutively to the Union Hospital of Huazhong University of Science and Technology and the Wuhan Asia General Hospital, from December 2019 to March 2020. The clinical outcomes, discharge from hospital, or death in hospital, were recorded up to March 31, 2020. Both hospitals are located in Wuhan, Hubei Province, and were designated by the Chinese government as hospitals responsible for the treatment of COVID-19 patients. The disease was diagnosed according to the WHO guidelines.
Exclusion criteria were as follows: (1) patients that were pregnant or under 18 years old; (2) patients with severe or critical COVID-19 at admission; (3) patients receiving irregular treatment before the admission (Fig. 1). Based on the exclusion criteria, 338 COVID-19 patients were included in the nal analyses.
All patients admitted to the hospital with mild or moderate COVID-19 were divided into the improvement group and the progression group. The improvement group included patients who recovered after admission and were discharged from the hospital. The progression group included patients who after the admission progressed to severe or critical condition, or died in the hospital, regardless whether the nal outcome was discharged or death, The protocol of the study was approved by the Research Ethics Committee of the Tongji Medical College. The data used in the study were anonymized, and the requirement for informed consent was waived by the Committee.

Data collection
Complete clinical data for all COVID-19 patients were collected from the medical records of the patients. They included demographic, clinical, laboratory, imaging, treatment, and outcome information. The laboratory data and CT scan of the lung were collected corresponded to the results of the rst test performed upon admission. Prior to the analysis, the patient information was de-identi ed and anonymized. The NLR was calculated from the results of neutrophil and lymphocyte counts. SII was calculated according to the results of a routine blood test, according to the formula: SII=N×P/L, where N, P, and L represents the count of neutrophils, platelets, and lymphocytes, respectively (6).

Laboratory examination
Laboratory con rmation of the infection of SARS-CoV-2 was conducted by local CDC in accordance with the Chinese CDC protocol. Pharyngeal swab samples were collected from all patients and the samples were stored in a viral-transport medium for laboratory testing. Real-time RT-PCR was performed to exclude infection with other respiratory viruses including in uenza A virus, coxsackie virus, in uenza B virus,respiratory syncytial virus, enterovirus and parain uenza virus. Automatic biochemical analyser, AU5800(Beckman Coulter, USA), was used for measuring the concentrations of ALT, AST, albumin, and creatinine. Liquid Assayed Multiqual was performed in QC procedures. Automated blood analyser, XE-2100 (Sysmex, Japan), was used for measuring the count of white blood cells, neutrophils, lymphocytes, and platelets. The lymphocyte test kit (Beckman Coulter Inc., FL, USA) was used for lymphocyte subset analysis. Plasma cytokines (IL10, IL6, IL4, IL2, IFN -γ and TNF -α) were detected using the human Th1/2 cytokine kit II (BD Ltd., Franklin lakes, NJ, USA). All laboratory tests were conducted in accordance with the product manual.

Study de nitions
The classi cation of the severity of COVID-19 was based on the "Diagnosis and Treatment Protocol for COVID-19 (Trial Version 7)" (4). The clinical classi cations of illness severity of COVID-19 were as follows: (1) mild type, with mild clinical symptoms and the absence of signs of pneumonia on imaging; (2) moderate type, with fever, respiratory tract symptoms, and signs of pneumonia were identi ed on imaging; (3) severe type, characterized by one of the following: a) respiratory distress, respiratory rate ≥ 30 breaths/min; b) mean oxygen saturation ≤ 93% in the resting state; c) oxygenation index ≤ 300 mmHg; and (4) critical type, characterized by one of the following: a) shock; b) respiratory failure requiring mechanical ventilation; c) organ failure requiring ICU admission. Fever was de ned as an axillary temperature of at least 37.3°C (20). The duration of viral shedding was de ned as the time from the onset of illness to the second negative nucleic acid test (4).

Statistical analysis
Summary statistics of the demographic data, clinical characteristics, laboratory results, and radiographic ndings were expressed as median and the interquartile range (IQR) for continuous variables and as frequencies and proportions for categorical variables. If the variance in the improvement group and the progression group was the same, continuous variables were compared by the Student's t-test; otherwise, the Welch's t-test was used. Categorical variables were compared by the χ2 test or Fisher's exact test.
The relationship between the NLR treated as a continuous variable, and the progression of COVID-19 was examined rst; subsequently, the relationship was evaluated considering NLR as a categorical variable according to the best threshold value. These relationships were examined using univariate and multivariate logistic regression analyses, and odds ratio (OR) and 95% con dence interval (CI) were calculated. After univariate logistic regression analysis, only the variables with a P-value of less than 0.1 were considered for multivariate analysis to identify predictors of progression of COVID-19. To avoid over tting in multivariate analysis, six variables were selected on the basis of clinical constraints and previous studies. Previous investigations have documented that older age was associated with poor prognosis in COVID-19 patients (20). Recent studies indicated that patients with severe COVID-19 had higher C-reactive protein (CRP) and lactate dehydrogenase (LDH) levels than patients with non-severe COVID-19 (13). Moreover, in comparison with patients with mild COVID-19, the levels of IL-10 and IFNγ in patients with severe COVID-19 were signi cantly increased in the early stage of the disease, and most detected cytokines peaked in the serum 3-6 days after the onset (21,22). Therefore, NLR, CRP, LDH, interleukin-10(IL-10), and interferon γ(IFNγ) were selected as the six variables for the multivariable logistic regression analysis.
To further validate the association between elevated NLR and disease progression, propensity score matching (PSM) was used to eliminate confounding bias (23). All COVID-19 patients were divided into two groups, the high NLR (≥3.75) group and the low NLR (<3.75) group, based on the best threshold value of NLR predicting disease progression. The propensity score was calculated using the logistic regression model in which age, gender, hypertension, coronary heart disease, diabetes, chronic obstructive pulmonary disease, fever, and cough were considered. The matching was performed using a 1:1 ratio. By this approach, patients in the high NLR group were matched with patients in the low NLR group having the closest propensity score. Patients selected by PSM were enrolled in a new cohort and subjected to further analysis of the association between NLR elevation and disease progression. The best threshold value of NLR was calculated according to the Youden index. The two-sided P-value of less than 0.05 was considered to indicate a statistically signi cant difference. SPSS 24.0 software was used to perform PSM and for statistical analysis.

Relationship between clinical characteristics and COVID-19 progression
Seventy-eight (23.1%) COVID-19 patients developed disease progression after the admission, and 6 (7.7%) of them died while hospitalized (Table 1). Table 1 also indicates that patients in the progression group were 6.5 years older than in the improvement group (62.5 vs. 56.0 years, P = 0.024). There was no signi cant difference in the ratio of males to females between the two groups (P = 0.548). Compared with patients in the improvement group, patients in the progression group had a higher rate of developing a fever, chest distress, myalgia, nausea or vomiting, and palpitation (all P 0.05). They also had a higher count of neutrophils (3.34 vs. 2.92 ×109/L, P = 0.014), lower count of lymphocytes (0.93 vs. 1.40 ×109/ L, P 0.001), and higher NLR (3.90 vs. 2.08, P 0.001) and SII (807.86 vs. 453.16 ×109/ L, P 0.001) values. In addition, they had higher levels of creatine kinase, LDH, CRP, ESR, IL-10, and IFNγ (all P 0.05) on admission. The incidence of lung consolidation detected by imaging was higher in the progression group (74.4% vs. 43.8%, P 0.001).

Risk factors associated with COVID-19 progression
To identify the predictors of COVID-19 progression, multiple clinical parameters were evaluated by logistic regression analysis (Table 2). In univariable analysis, the odds of disease progression were higher in patients with fever or cough (Table 2). Age, elevated neutrophil count, reduced lymphocyte count, elevated NLR, SII, creatine kinase, LDH, CRP, ESR, IL-10, and IFNγ were also associated with disease progression. To avoid over tting in multivariate analysis, age, NLR, CRP, LDH, IL-10, and IFNγ were selected for multivariate regression analysis, as justi ed in the Methods section. Multivariable logistic regression analysis documented that increased NLR, LDH, and IL-10 were independent predictors of disease progression in COVID-19 patients.
The receiver operating curve (ROC) analysis indicated that the optimal cut-off value of NLR for predicting the progression of COVID-19 was 3.75. The values of the area under the curve, re ecting the accuracy of predicting COVID-19 progression by NLR and SII, were 0.739 (95%CI: 0.605-0.804) and 0.674 (95%CI: 0.603-0.745) (Fig. 2). Therefore, NLR, but not SII, was included in the multivariate regression analysis.
In this analysis, we demonstrate that a NLR of 3.75 is the best cutoff point for predicting disease progression in patients with COVID-19 by ROC curve analysis. The ROC curve analysis revealed that a LDH cutoff point of 213.5 U/L showed the biggest Youden index (Fig 2). The AUC (95%CI) was 0.727(0.670-0.785). Furthermore, we estimated whether a NLR of 3.75 combined with a LDH cutoff of 213.5 U/L would improve the accuracy of prediction of disease progression in patients with COVID-19. Table 3 presents sensitivity, speci city, positive predictive values (PPV), and negative predictive values (NPV)(95%CI) of these variables in combination or either alone.

Risk factors associated with COVID-19 progression analyzed after propensity score matching
To further validate the association between elevated NLR and disease progression, PSM was used to reduce confounding bias. The patients were divided into two groups based on the best threshold value of NLR predicting disease progression: the high NLR (≥3.75) group and the low NLR (<3.75) group. The high NLR group had a higher incidence of COVID-19 progression than the low NLR group (P 0.001). Seventyfour propensity score-matched pairs of COVID-19 patients were selected and enrolled in a new cohort. However, a statistically signi cant difference in COVID-19 progression continued to be present between the high and low NLR groups (P 0.001) ( Table 4). After univariate analysis, the variables with P<0.1, including age, NLR, LDH, CRP, IL-10, and IFNγ, were selected for multivariate regression analysis. This analysis con rmed that NLR ≥3.75 was still an independent predictor of disease progression in COVID-19 patients (Table 5).

Discussion
The present study has demonstrated that NLR, measured at the early stage of the disease, is a signi cant independent predictor for the progression of COVID-19 to severe and critical stages. Moreover, the nding that NLR ≥3.75 is an independent predictor of disease progression was also con rmed in the propensity score-matched cohort. The PPV of NLR is actually quite poor. The combined evaluation of NLR and LDH improved the accuracy of the prediction of COVID-19 progression. The relevance of early adequate treatment to prevent mild or moderate cases from developing into severe ones is well-recognized since the treatment of critical COVID-19 patients requires not only signi cant medical resources but, most importantly, results in a high mortality rate (20). This current analysis provides a new evidence-based strategy for early identi cation of patients at high risk of COVID-19 progression.
Previous studies have documented that older age is a signi cant independent predictor of mortality in MERS and SARS (24,25). The current investigation has found that older age is also associated with increased odds of in-hospital death in COVID-19 patients (26). However, age is not an independent predictor of the progression of COVID-19 to severe or critical stages. COVID-19 is commonly susceptible in the population, and there is a risk of progression to severe disease (4). A possible explanation of this apparent discrepancy might be that older patients with severe or critical COVID-19 have a higher risk of death. Since patients with severe or critical COVID-19 on admission were not included in the study, the fatality rate among the analyzed population was lower than that documented in previous reports (26).
In agreement with previous research (4), the present analysis has shown that an increased level of LDH at the early stage of COVID-19 was also a signi cant independent predictor for disease progression. Moreover, COVID-19 patients in the progression group had higher levels of IL-10 and IL-6 at the early stage of the disease, although differences in IL-6 between the groups were not statistically signi cant. However, IL-10 at the early stage was also a signi cant independent predictor for COVID-19 progression.
The identi cation of elevated levels of in ammatory factors in COVID-19 patients is consistent with previous studies (27). The current work documented that, except for IL-6, the serum concentration of cytokines peaked at 3-6 days after the onset of the disease in critically ill COVID-19 patients; these cytokines included IL-10, IL-2, IL-4, TNFα, and IFNγ (27). In addition, levels of IL-10 and IL-6 continued to increase in patients with severe COVID-19, and with the levels of IL-6 beginning to decrease after 16 days (27).
It is currently well-recognized that COVID-19 is highly contagious in the general population, and caused a global pandemic in a short period of time. The treatment of severe and critically ill patients is imperative, but it requires the consumption of a great deal of valuable medical resources, especially the ventilators. The lack of adequate medical resources became already evident in some developed countries and is more dramatic in the developing ones. It has been reported that in Guangzhou, China, where all patients with mild and moderate COVID-19 receive appropriate care and treatment at an early stage of the illness, their prognosis is better than that for patients in Wuhan, where they are more likely to be admitted to ICU and succumb to the disease (5). The difference between these two cities underscores the importance of appropriate care and treatment for early mild cases, particularly for patients at high risk of disease progression. Timely detection of risk factors for disease progression might help to provide necessary supportive care and treatment to prevent mild cases from becoming severe, reduce mortality, and alleviate medical resource shortages.
Since there is no speci c drug to treat COVID-19, the treatment consists mostly of symptomatic support therapy. The Chinese management guidelines for COVID-19 (version 7.0) recommend that the general treatment for mild or moderate COVID-19 patients includes 1) bed rest, intensive supportive treatment, adequate supply of energy; maintenance of internal environmental stability; 2) monitoring of blood indicators and chest images; 3) timely delivery of effective oxygen therapy in time; 4) administration of antiviral therapy, such as α-interferon, lopinavir/ritonavir, or abidor; 5) antimicrobial treatment (4). The care for patients with the severe or critical stage of COVID-19 relies on active prevention and treatment of complications, management of comorbidities, prevention of secondary infection, and supporting organ function on the basis of symptomatic treatment (4). Surprisingly, the value of traditional Chinese medicine represented by lotus qingwen capsule in the treatment of COVID-19 has been re ected in the guidelines (4).
It has been reported that a cytokine storm might be associated with the severity of COVID-19 (28). Immunoregulatory therapy downregulating the cytokine storms might be an effective strategy for the treatment of COVID-19. A study of 201 COVID-19 patients documented that treatment with methylprednisolone reduced the risk of death in patients with acute respiratory distress syndrome (29). The Diagnosis and Treatment Protocol for COVID-19 (Version 7) (4) recommends corticosteroids only in certain critical COVID-19 patients, and at low-to-moderate doses for a short duration. It was reported that elevated serum levels of IL-6 might be predictive of mortality in COVID-19 patients (30). The treatment with tocilizumab, a speci c monoclonal antibody that blocks IL-6, led to a reduction in fever and lung lesion opacity and normalized the percentage of lymphocytes in peripheral blood in the COVID-19 patients with upregulated IL-6 (31). Moreover, the use of tocilizumab has been recommended by management guidelines (4) in COVID-19 patients with extensive lesions in both lungs and con rmed rising levels of IL-6.
Some limitations of the present work should be acknowledged. First, this investigation was designed as a retrospective study, and relatively few cases of progressing COVID-19 were included in the analyses. Thus, large-scale multicenter prospective cohort studies are necessary to con rm the results and strengthen the conclusions reached. Second, mild or general cases were treated before they progressed to severe cases. As stated earlier, patients with mild COVID-19 are at risk of progressing to the severe stage, and once they become severe, they have a high mortality rate. Therefore, it would be unreasonable not to treat the patients already admitted to the hospital. Although they were treated at different sites, the two designated hospitals followed the same guidelines for treating mild or moderate COVID-19 speci ed in the Diagnosis and Treatment Protocol for COVID-19 issued by the National Health Commission of China.
Thus, the use of the data from two different centers did not affect the results of the study, since similar treatment strategies were employed.

Conclusions
The performed analysis demonstrates that NLR qualifies as an independent predictor of disease progression in COVID-19 patients at the early stage of the disease. The combined evaluation of NLR and LDH improved the accuracy of the prediction of COVID-19 progression. Assessment of predictors might facilitate early identi cation of COVID-19 patients at high risk for disease progression and ensure timely administration of appropriate treatment to prevent mild cases from becoming severe.

Declarations
Ethics approval and consent to participate: The protocol of the study was approved by the Research Ethics Committee of the Tongji Medical College (NO: IORG0003571). The data used in the study were anonymized, and the requirement for informed consent was waived by the Committee. Consent for publication: All authors familiar with the contents of the nal draft and Agreed to the publication of the article.
Availability of data and materials: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Antimicrob Agents. 2020; doi: 10.1016/j.ijantimicag.2020.105954.      ROC curve: Predicting COVID-19 progression with NLR, SII and LDH