Comparison of Blood Counts and Markers of Inflammation and Coagulation in Patients With and Without COVID-19 Presenting to the Emergency Department in Seattle, WA

Christopher M. Chandler, MD; Molly C. Reid, MPH; Sindhu Cherian, MD; Daniel E. Sabath, MD, PhD; Kerstin L. Edlefsen, MD


Am J Clin Pathol. 2021;156(2):185-197. 

In This Article


We present a description of changes in laboratory tests associated with SARS-CoV-2 infection in a cohort of patients in the United States, with a COVID-19–negative comparison group stratified by clinical acuity. Concordant with what has been previously published, our findings suggest that laboratory alterations seen in COVID-19 are related to disease severity but highlight that these alterations may not be specific to pathogenic mechanisms distinct to COVID-19.[6,15,16]

Regarding WBC counts, ED patients with COVID-19 were overall more likely to be leukopenic (16% vs 5%, P <.0001) and lymphopenic (45% vs 27%, P <.0001) compared to those without COVID-19 in the ED. Among patients with COVID-19, leukocyte counts increased significantly with illness severity, driven largely by an increase in neutrophils. Rates of absolute neutrophilia also increased significantly with disease severity (3% up to 34%, ED to critical courses; P <.0001; OR, 15). Meanwhile, lymphocyte counts and illness severity were inversely related; the ED COVID-19–positive group had significantly higher lymphocyte counts than either the critical COVID-19–positive or admitted COVID-19–positive groups. This finding supports the use of lymphocyte count as a marker that may flag patients with COVID-19 more likely to require admission to the hospital. However, a low lymphocyte count alone is not sufficiently sensitive or specific to predict COVID-19 status (sensitivity of 45% and specificity of 73% in our dataset). Our findings are similar to those recently published by Pozdnyakova et al,[17] who found critically ill patients with COVID-19 had lower leukocyte counts compared to ICU patients without COVID-19, but higher rates of neutrophilia and lymphopenia compared to patients with COVID-19 who did not require critical care.

SARS-CoV-2 is similar to the original SARS-CoV in its tendency to cause lymphopenia, and the higher neutrophil counts and higher rates of neutrophilia in patients with severe COVID-19 were also observed in SARS-CoV and MERS-CoV infections.[18,19] Lieberman and colleagues[20] have shown an inverse correlation between SARS-CoV-2 viral transcripts and neutrophil and B-cell related transcripts in nasopharyngeal specimens. This finding, in combination with the data of Liu et al[21] showing increasing levels of circulating neutrophils over time in both severe and nonsevere COVID-19 infections, suggests neutrophilia and lymphopenia may lag behind the periods of highest viral replication and presumed infectivity.

Several prior studies have suggested the NLR is a useful marker of disease severity in COVID-19.[7–9,11] Our data support the utility of NLR to stratify patients but, importantly, our data show the association of NLR with disease severity is not unique to patients with COVID-19 (Figure 2). Overall, the NLR may be useful as a proxy for disease severity as a measure of a patient's overall inflammatory state and may help stratify those at risk of needing a higher level of hospital care.[12]

Other hematologic laboratory values responsive to inflammation include hemoglobin and RDW. Interestingly, anemia was less common in patients with COVID-19 (P = .075) and did not vary with illness severity (Figure 2), fitting the muddled picture in the literature; most analyses similar to our study show no significant change in hemoglobin with worsening disease.[5,22] However, a small meta-analysis found a significant association between decreased hemoglobin concentrations and severity of COVID-19.[23] The lack of consistent impact on RBC levels is somewhat surprising, as inflammatory cytokines blunt erythropoiesis (anemia of inflammation).[24] The hematologic data in our study represent early time points in the disease course for COVID-19 (all coming from the ED) and may be too early to detect any changes in erythropoiesis or erythrophagocytosis due to the disease.

Higher values of RDW are thought to reflect a proinflammatory state and have been associated with more severe illness in the critical care literature.[25,26] RDW increased steadily, and significantly, with disease severity in both patients with and without COVID-19 in our dataset. This finding is concordant with the findings of Foy et al[27] showing increased mortality risk with increasing RDW in patients with COVID-19. Somewhat surprisingly, we found patients with COVID-19 tended to have lower RDWs compared to patients without COVID-19 overall (P = .035), and this held true for each level of acuity examined (Figure 2). Also, patients with COVID-19 were no more likely than patients without COVID-19 overall to have an abnormally high RDW (35% vs 36%, P = .9278). As such, RDW holds utility as an indicator of general illness severity but is not specific to COVID-19.

The reason for the slightly higher platelet counts in our study compared to some others is unclear.[28,29] It may represent heterogeneity in our patient subgroups or reflect local differences in disease characteristics, as our median platelet count is similar to those reported in case series from the Seattle, WA, area.[30–32] Lastly, Qu et al[10] found that hospitalized patients with COVID-19 presenting with an elevated platelet count had worse outcomes and that PLR was a risk factor for prolonged hospitalization. Platelet counts may increase due to the inflammatory milieu, and a slight increase in platelet count in our critical group may be confounding the statistical analysis.

The non-CBC analytes in our dataset were not uniformly available for all patients, but several trends of interest will be discussed further. We observed significant increases in CRP and IL-6 in patients with COVID-19 and corresponding changes in acute phase reactants, including ferritin, fibrinogen, and albumin, compared to patients without COVID-19. Serum albumin in particular showed a dramatic decrease with increased illness severity in patients with COVID-19 and was lower than the comparison group for each clinical course (Supplemental Figure 1). Patients in our critical COVID-19–positive group were over 47 times more likely to have a low minimum albumin compared to their peers in the ED COVID-19–positive group (OR, 47.05; 95% CI, 13.95–158.7), suggesting this common laboratory parameter may have utility as a marker of disease severity in COVID-19.

Much has been made of coagulopathy in association with COVID-19 due to the high rates of thrombotic complications in reports from Wuhan with focus shifting to a microthrombotic pathology as described in one of the first autopsy studies.[13,14,33,34] Our dataset was not specifically focused on investigating markers of coagulation and we did not collect information regarding rates of anticoagulation, thus limiting interpretations. However, our data fit the findings in the literature in which small to negligible increases in PT and significantly longer PTTs are noted in patients with COVID-19.[6,13,14,35–37] Aside from the use of prophylactic or therapeutic anticoagulation, especially heparin, one potential explanation for the prolonged PTT we observed is the presence of a lupus anticoagulant, antibodies against phospholipids formed in hyperinflammatory states. Lupus anticoagulant have been reported in 45% to 91% of patients with COVID-19.[37–40]

Interpretation of D-dimer levels poses several challenges. Our study fits with numerous others showing that D-dimer levels increase with disease severity in COVID-19, but the use of this marker and an indicator of thrombotic tendency in COVID-19 is fraught due to its correlation with systemic inflammation.[35,36,41,42] Our data did not show a significant difference in D-dimer levels between patients with and without COVID-19 overall (P = .1972). However, this finding is confounded by differences in test acquisition by clinical course between COVID-19–negative and –positive groups. Over half (65%) of D-dimer tests in our COVID-19–negative group were ordered in the ED course, where the analyte is frequently used to rule out pulmonary thromboembolism (compared to 17% of D-dimer tests in the COVID-19–positive group). In contrast, orders for D-dimer in COVID-19–positive patients were enriched in the critical course (66%). While an elevated D-dimer is an indication of fibrinolysis, this is not an uncommon finding in hospitalized patients, particularly those that are critically ill. The choice of comparison group colors the interpretation of D-dimer elevation observed in COVID-19.

Similar to our data, Helms et al[38] found lower D-dimer levels in patients with COVID-19 in the ICU compared to matched patients with ARDS. Meanwhile, Yu et al[36] compared D-dimer levels in severe cases of COVID-19 to admitted patients with community acquired bacterial pneumonia (CAP) and found higher D-dimers in patients with COVID-19. This was despite the fact that markers of inflammation, including CRP and erythrocyte sedimentation rate, were higher in patients with CAP. In most of the patients with COVID-19 they followed, D-dimer decreased along with CRP following treatment. However, in a subset of patients with COVID-19, D-dimer levels remained elevated, suggesting the presence of thrombi and possibly the need for more aggressive anticoagulation.[36]

Our study has a number of limitations. It is a retrospective, observational analysis of a narrow geographic region during a specific period of time. Pediatric patients (younger than 18 years) and those likely to have a neoplastic diagnosis, including leukemias and lymphomas, were excluded. As we approximated clinical course from origin of laboratory order instead of directly reviewing the medical record for each patient, there is a risk of incomplete or miscategorization of patients. While we chose a significance level of P = .05, many of the laboratory values are likely correlated, thus the statistical significance of those close to this cutoff should be interpreted with caution. Lastly, non-CBC laboratory tests are not uniformly represented in our dataset and thus conclusions are limited by the possibility of ascertainment bias from differences in ordering practices.