Incidence of bloodstream infections in patients with COVID-19: a retrospective cohort study of risk factors and outcomes.

Introduction
Prior evidence found that bloodstream infections (BSIs) are common in viral respiratory infections and can lead to heightened morbidity and mortality. We described the incidence, risk factors, and outcomes of BSIs in patients with COVID-19.


Methods
This was a single-center retrospective cohort study of adults consecutively admitted from March to June 2020 for COVID-19 with BSIs. Data were collected by electronic medical record review. BSIs were defined as positive blood cultures (BCs) with a known pathogen in one or more BCs or the same commensal organism in two or more BCs.


Results
We evaluated 290 patients with BCs done; 39 (13.4%) had a positive result. In univariable analysis, male sex, black/African American race, admission from a facility, hemiplegia, altered mental status, and a higher Charlson Comorbidity Index were positively associated with positive BCs, whereas obesity and systolic blood pressure (SBP) were negatively associated. Patients with positive BCs were more likely to have severe COVID-19, be admitted to the intensive care unit (ICU), require mechanical ventilation, have septic shock, and higher mortality. In multivariable logistic regression, factors that were independent predictors of positive BCs were male sex (OR=2.8, p=0.030), hypoalbuminemia (OR=3.3, p=0.013), ICU admission (OR=5.3, p<0.001), SBP<100 mmHg (OR=3.7, p=0.021) and having a procedure (OR=10.5, p=0.019). Patients with an abnormal chest X-ray on admission were less likely to have positive BCs (OR=0.3, p=0.007).


Conclusions
We found that male sex, abnormal chest X-ray, low SBP, and hypoalbuminemia upon hospital admission, admission to ICU, and having a procedure during hospitalization were independent predictors of BSIs in patients with COVID-19.


Background
Coronavirus disease-19  is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV2). By May 16, it had affected more than 168 million people worldwide and had caused more than 3.4 million deaths. [1] The disease can have a wide range of clinical presentations, commonly affecting the lungs, heart, kidneys, neurological, and gastrointestinal systems. [2] Severe presentations occur in up to 40% of hospitalized cases, commonly due to respiratory failure requiring mechanical ventilation and admission to an intensive care unit (ICU). [3] Bacterial coinfections are common in respiratory viral infections. A previous study in Spanish adults admitted to the ICU during the 2009 H1N1 in uenza pandemic found that bacterial coinfection led to increased length of stay and resource utilization. [4] Similarly, bloodstream infections (BSIs) can increase morbidity and mortality when associated with viral infections. In hospitalized adults with COVID-19 from the US, [5,6] UK, [7] Italy, [8] and Sweden, [9] the frequency of BSIs ranged from 2.5-7.9%. In subjects with severe and critical illness, estimates of BSIs from the US [10], China, [11][12][13][14] and Italy [15] ranged from 0.9-49%. Moreover, BSIs are considered a signi cant cause of morbidity and mortality in the US and Europe. [16] The current evidence highlights a wide range of frequencies of BSIs and the importance of identifying risk factors that can be targeted to decreased morbidity and mortality. Furthermore, minority patients such as black/African Americans are disproportionally affected by a higher incidence and greater disease severity of COVID-19. [2,17] We evaluated the incidence of BSIs in COVID-19 patients and described the potential risk factors associated with these infections in predominantly black/African American patients. We sought to identify risk factors that can help predict which patients will develop BSIs, to decrease morbidity and mortality.

Methods
Study subjects. We performed a retrospective, single-center historical cohort study of patients admitted for COVID-19 from March 8 to June 14 who had BSIs. BSIs were de ned as a positive blood culture (BC) with a known pathogen in one or more BCs or the same commensal organism in two or more BCs.
Inclusion criteria were adult patients (18 years and older) admitted for COVID-19 diagnosed with a positive reverse-transcriptase-polymerase-chain-reaction (RT-PCR) assay of a nasopharyngeal swab and had a BC done during their hospitalization. Exclusion criteria were adults who did not have BCs drawn during hospitalization or if the positive BC was due to an organism thought to be a contaminant. Clinical signi cance and source of BSIs were collected from infectious diseases (ID) consultation or progress notes. The study protocol was approved by the Institutional Review Board (IRB) of Ascension St. John Hospital.
Data collection. We collected data from the electronic medical record (EMR) via retrospective chart review.
Demographic data collected included age, sex, race, nine-digit zip code, type of insurance, and admission source. Anthropometric data included weight, height, and body mass index (BMI). Obesity and morbid obesity were de ned using the Centers for Disease Control and Prevention (CDC) guidelines. [18] We collected relevant medical comorbidities and calculated the Charlson Comorbidity Index, a validated weighted score established as a predictor of mortality. [19] COVID-19 severity was based on the World Health Organization (WHO) disease severity classi cation. [20] We collected data on vital signs and pulse oximetry on admission along with chest x-ray and chest computed tomography (CT) ndings based on the o cial radiology report. Laboratory data included total white blood cell, lymphocyte, and platelet counts, C-reactive protein (CRP), serum creatinine, alanine aminotransferase (ALT), and aspartate aminotransferase (AST). We classi ed elevated and low laboratory values based on the hospital's cut points and categorized them as presented in Table 1. Elevated creatinine on admission was de ned as an increase in serum creatinine ≥ 1.5 times the baseline within seven days prior or an increase ≥ 0.3mg/dL within 48 hours of admission. [21] We calculated the quick Sequential Organ Failure Assessment (qSOFA) on admission and used it to predict in-hospital sepsis-related mortality. [22] Septic shock was de ned as hypotension requiring vasopressors. We de ned central line as placement of any catheter (tunneled or non-tunneled) in any central vein during hospitalization. For procedures, we included any major emergent surgeries and any emergent endoscopic or percutaneous procedures done during the hospital stay. The most common symptoms were shortness of breath, cough, and fever. Patients with positive BCs were signi cantly more likely to present with altered mental status and lower systolic blood pressure (SBP) but less likely to present with fatigue, shortness of breath, and cough, compared to those with negative BCs. (Table 1). Notably, 94.0% of the population had elevated CRP without signi cant difference between BC groups. Compared to those with negative BCs, those with positive BCs were more likely to present with leucopenia and hypoalbuminemia (Table 1).
In the positive BC group, patients had a lower incidence of abnormal chest x-ray on admission compared to those with negative results (66.7% vs. 81.3%, p = 0.04). Contrastingly, CT chest on admission was more likely to be abnormal in the positive BC group compared to the negative group (16.2 vs. 6.6%, p = 0.04).
Twenty percent (n = 58) and 3.1% (n = 9) of the analyzed population had at least one central line placed and one procedure done during hospitalization. Both were signi cantly higher in the positive BC group (Table 1). Additionally, more than half of the population was treated with steroids during hospitalization (55.5%), and those with positive BCs were more likely to receive steroids compared to those with negative BCs, approaching signi cance (69.2%. vs. 53.4%, p = 0.06) ( Table 1).
The prevalence of severe disease was 21.7%, and around a third of patients were admitted to ICU (30%) ( Table 2). Compared to those with negative BCs, those with positive BCs were more likely to have severe disease, be admitted to ICU, require mechanical ventilation, have septic shock, and die (64.1% vs. 32.7%, p < 0.0001). In multivariable analysis, the variables initially introduced in the stepwise forward logistic regression included sex, Charlson Comorbidity Index, obesity, altered mental status, abnormal chest x-ray on admission, leucopenia, thrombocytopenia, hypoalbuminemia, CRP level, steroid treatment, admission from a facility, central line placement, procedures, severity of disease, admission to ICU, and SBP ≤ 100 mmHg. The nal model included sex, abnormal chest x-ray on admission, hypoalbuminemia, admission to ICU, SBP ≤ 100 mmHg, and procedures. Notably, compared to those with negative BC, those with positive BCs during hospital stay were ten times more likely to have undergone a procedure, ve times more likely to be admitted to ICU, twice as likely to be male, and three times more likely to have hypoalbuminemia and low SBP, and three times less likely to have an abnormal chest x-ray on admission (Table 3).

Discussion
This study found a 13.4% incidence of BSIs in an urban population from Southeast Michigan, which is slightly higher than previously reported rates in patients from New York, [5,6] Sweden, [9] London, [7] and Italy, [8] but lower than those from New Jersey. [10] The difference in estimates is likely related to variations in case de nition and severity of disease. About a third of our population was admitted to the ICU, and the patients with positive BCs were more likely to be admitted to the ICU, have more severe disease, and require vasopressor support. Notably, admission to ICU and systolic hypotension were independent predictors of BSIs. Our ndings are consistent with a study conducted in adults with severe COVID-19 from several centers from New Jersey [10] that found that patients with positive BCs were more likely to be admitted to ICU, have septic shock, and present with altered mental status than those with negative BCs. Notably, in univariable analysis, patients with positive BCs were more likely to present with altered mental status when compared to those with negative BCs.
Male sex and an increased number of comorbidities were associated with increased BSIs, but only male sex remained an independent predictor. A previous meta-analysis by Wu et al. [23] described an almost two-fold increased odds of severe COVID-19 in males compared to females. One of the proposed mechanisms for males having increased disease severity comes from the hypothesis that androgens facilitate SARS-CoV2 entry via the transmembrane protease serine 2 (TMPRSS2)-pathway. SARS-CoV2 spike proteins need trimming by TMPSS2 to bind to angiotensin-converting enzyme 2 (ACE2) receptors for entry to host cells. Androgens activate its receptor (AR), and the AR acts as a promoter of TMPRSS2 transcription resulting in facilitated host cell entry of SARS-CoV2 [24]. We believe males have increased BSIs related to the increased severity of COVID-19.
Several laboratory abnormalities have been described in patients with COVID-19 [2]. In our study, patients with positive BCs were more likely to present with leucopenia and hypoalbuminemia when compared to those with negative BCs. On multivariable analysis, only hypoalbuminemia was an independent predictor of BSIs. This is consistent with a previous metanalysis by Aziz et al. [25] that found hypoalbuminemia was associated with increased odds of severe COVID-19 vs. non-severe disease. Additionally, hypoalbuminemia has been associated with increased severity of illness in the context of viral, fungal, and bacterial infections, including COVID-19 [26]. This evidence supports the role of hypoalbuminemia as a marker of more severe disease and can be considered a valuable predictor of BSIs in patients with COVID-19.
We also described that abnormal chest x-ray on admission was less likely in the positive BC group when compared to the negative BC group, and this association remained signi cant in multivariable analysis. Contrastingly, CT chest on admission was more likely to be abnormal in the positive BC group compared to the negative group, but it was not a signi cant predictor in multivariable analysis. Similar to our results, a previous retrospective study of Chinese adults with COVID-19 reported abnormal ndings in 59% of patients with a chest x-ray [27]. We hypothesize that the abnormal CT chest did not remain a signi cant predictor in the multivariable analysis because of a low number of patients who had a CT chest done on admission compared to those who had a chest x-ray done (6 vs. 29). Additionally, an observational study conducted among adults from the UK described a lower sensitivity of the chest x-ray when compared to CT chest for COVID-19 diagnosis using RT-PCR as gold standard [28], and we hypothesize that changes in CT chest occur earlier compared to xray, which could explain the discrepancy in our results.
Finally, we found that having at least one central line placement or a surgical procedure done during admission was associated with a positive BC, but in multivariable analysis, only having a procedure was associated with higher odds of positive BCs, despite the small number of patients (n = 9). Surgical procedures are associated with a higher risk of BSIs but, to the best of our knowledge, no other studies have described an increased risk of BSIs associated with procedures in patients with COVID-19.
Our study is not without limitations. First, because of the small sample size, we were not able to stratify our results based on BC collection timing, and therefore our results also re ect the occurrence of hospitalrelated BSIs. Thus, we cannot know if BSIs resulted from greater disease severity or if having a BSI led to a more severe presentation. Moreover, due to the observational nature of our data, causal inference cannot be made. Because our study population is predominantly black/African American, and race can affect the association between COVID-19 and disease severity, our results may not be generalizable to a broader patient group. We did not nd any association of steroid therapy with bloodstream infections, likely because of the low sample size. Finally, we did not have information on immunomodulatory therapies such as tocilizumab, which has been associated with increased risk of BSIs in critically ill patients with COVID-19. [29] Our study has several strengths. We were able to identify predictors of bloodstream infections in patients hospitalized for COVID-19. Our results are consistent with previous studies, as stated above. We described potential predictors that can be used for early detection and to identify patients at risk, speci cally, patients with COVID-19 admitted to ICU. Our population was predominantly black/African American, previously reported as disproportionally affected by a higher number of COVID-19 cases, and presenting with more severe disease [17]. We were also able to adjust for important risk factors such as placement of central lines and emergent procedures and described the role of procedures as a risk factor, despite the small sample size.
This historical cohort study found that independent predictors of bloodstream infections included male sex, hypoalbuminemia, admission to ICU, SBP ≤ 100 mmHg, abnormal chest x-ray, and procedures. We recommend further investigations to con rm these potential predictors that can lead to early identi cation and treatment of BSIs and reduce morbidity and mortality in adults with COVID-19, speci cally identifying patients requiring emergency procedures as a high-risk group.

Declarations
Ethics approval and consent to participate The study protocol was approved by the Institutional Review Board (IRB) of Ascension St. John Hospital

Consent for publication
Not applicable Availability of data and materials The datasets generated and/or analyzed during the current study are not publicly available due to HIPAA regulations, but are available from the corresponding author on reasonable request.
SS: Data analysis, collection, and nal manuscript editing and revision ML: Study design and nal manuscript editing and revision LS: Study design and nal manuscript editing and revision AB: Study design and nal manuscript editing and revision