Secondary bacterial infection in COVID-19 patients is a stronger predictor for death compared to influenza patients

Secondary bacterial infections are a potentially fatal complication of influenza infection. We aimed to define the impact of secondary bacterial infections on the clinical course and mortality in coronavirus disease 2019 (COVID-19) patients by comparison with influenza patients. COVID-19 (n = 642) and influenza (n = 742) patients, admitted to a large tertiary center in Israel and for whom blood or sputum culture had been taken were selected for this study. Bacterial culture results, clinical parameters, and death rates were compared. COVID-19 patients had higher rates of bacterial infections than influenza patients (12.6% vs. 8.7%). Notably, the time from admission to bacterial growth was longer in COVID-19 compared to influenza patients (4 (1–8) vs. 1 (1–3) days). Late infections (> 48 h after admission) with gram-positive bacteria were more common in COVID-19 patients (28% vs. 9.5%). Secondary infection was associated with a higher risk of death in both patient groups 2.7-fold (1.22–5.83) for COVID-19, and 3.09-fold (1.11–7.38) for Influenza). The association with death remained significant upon adjustment to age and clinical parameters in COVID-19 but not in influenza infection. Secondary bacterial infection is a notable complication associated with worse outcomes in COVID-19 than influenza patients. Careful surveillance and prompt antibiotic treatment may benefit selected patients.


Characterization of the bacterial pathogens among influenza and COVID-19 patients. COVID-
19 patients had higher rates of secondary bacterial infections than influenza patients (12.6% vs. 8.7%, p = 0.006). To test whether there was also a difference in the nature of the infecting pathogens between the two groups, we further compared the different types of bacteria in each group (Table 2). For this purpose, we defined infections occurring < 48 h after admission as early infections, while infections occurring between 2-14 days postadmission were defined as late infections. Infections occurring after 14 days were disregarded. Interestingly, the types of isolated bacteria were generally similar between COVID-19 and influenza patients. While infections with gram-negative bacteria represented most infections (75%) in both groups, there were no significant differences between the groups neither in the early nor late infections. In contrast, late infections with gram-positive bacteria were more common in COVID-19 than influenza patients (28% vs. 9.5%, p = 0.01).
Most secondary bacterial infections in both groups were derived from blood cultures (supplementary table 1). A notable difference was observed in the late gram-positive infections. In COVID-19 patients, 85% of these infections were isolated from blood cultures, while only 14.2% were isolated from respiratory samples. In contrast, among influenza patients, the isolated bacteria divided equally between blood and respiratory-derived cultures.
Last, we calculated the percentage of deaths with respect to the presence or absence of secondary bacterial infections in the COVID-19 and Influenza groups (Fig. 2). Bacterial infections resulted in significantly decreased survival rates in both groups. While 13.2% of patients without infections died, 33% of patients with one infection and 61% of patients with two or more infections died (Chi-Square Test, no infections p < 0.001, one infection p = 0.001, more than one p = 0.002).
Comparison between the influenza and COVID-19 groups showed that patients infected with COVID-19 have a higher risk of death. In the influenza group 17.6% of patients with one infection, and 25% of the patients with more than one infection died. While in the COVID-19 group, 48.1% of patients in the group with one infection, and 75.9% of patients with more than one infection died (Chi-Square Test, influenza p < 0.001, COVID-19 p < 0.001). A similar analysis was performed regarding the time of the infection in both COVID-19 and influenza Table 3. Univariate analysis of variables associated with death. OR, odds ratio; CI, confidence interval; min, minimal levels; max, maximal levels.

Discussion
In this study, we compared two large cohorts of influenza, and COVID-19 patients admitted to a large tertiary center and from which bacterial cultures from blood, sputum, and/or BAL were obtained. Most secondary bacterial infections in both groups were derived from blood cultures. Patients infected with these viruses suffer from acute respiratory distress; therefore, clinicians mostly avoid taking sputum and BAL tests. This might affect the number of infections identified in blood cultures vs. cultures from respiratory origins. However, it should be also taken into account that respiratory infections might become systemic with time as well. In general, COVID-19 patients were more severely ill, as reflected by various disease severity markers, and had worse outcomes, reflected by a higher percentage of intubation or death than influenza patients. Importantly, COVID-19 patients had more documented secondary bacterial infections than influenza patients, and these infections were independently associated with death in COVID-19 but not in influenza patients. These findings suggest that secondary bacterial infections might be a significant, potentially treatable, contributing factor for disease severity among COVID-19 patients.
Our results are in line with previous studies reporting that COVID-19 patients suffer from a more severe disease and ~ 3 times higher death rates) than influenza patients. In addition, in-hospital death of patients with pulmonary secondary bacterial infection was two-times higher in COVID-19 patients 8 .
In this study, the most common bacterial infections in patients with either influenza or COVID-19 were Pseudomonas aeruginosa and Staphylococcus aureus, generally in agreement with previous studies 8 . Staphylococcus aureus is a known pathogen associated with secondary pneumonia during influenza infection 9 . Its dissemination to the lungs is attributed to a combination of environmental changes and immune responses that create suitable conditions for Staphylococcus aureus infection 9 . Pseudomonas aeruginosa is also associated with chronic predisposing respiratory conditions, including upper respiratory tract infections such as influenza 10,11 . While Pseudomonas aeruginosa is a common respiratory opportunistic pathogen, it is also known as the most common gram-negative bacterial species associated with severe hospital-acquired infections in some hospitals 12 .
Interestingly, a significant difference was observed in the percentage of Enterococcus infections between influenza and COVID-19 (0% vs. 8.6%, p = 0.018). Furthermore, late infections with gram-positive bacteria were more common in COVID-19 compared to influenza patients. Of note, a large multi-center study from Sweden comparing bacterial growth in 15,103 blood cultures from COVID-19 patients with non-infected controls also found that the rate of infections with gram-positive bacteria was significantly higher in patients with COVID-19 (66% vs. 50%, p < 0.0001) 13 .
Generally, the overall time from admission to bacterial growth was longer for COVID-19 compared to influenza patients. This possibly reflects the disease's natural history, characterized by a late deterioration, typically 7-10 days after symptoms onset, that might be accompanied by a secondary bacterial infection. Alternatively, COVID-19 patients are admitted in isolated dedicated wards under strict isolation protocol, limiting free medical and nursing personnel access. This might affect the quality of medical treatment and increase the likelihood of complications such as nosocomial infections, including line, device-related, skin, and soft tissue infections associated with gram-positive bacteria 14 . This hypothesis is in line with the increased hospitalization length, the longer overall time from admission to bacterial growth, and the higher rates of late gram-positive bacterial infections observed in COVID-19 than influenza patients. In addition, the higher fraction of late gram-positive infections isolated from blood cultures in COVID-19 compared to influenza, further supports this notion.
Although several studies characterized the secondary bacterial infections in COVID-19 patients, our study is the first to compare the infecting bacterial pathogens with those observed in influenza patients from the same center, and to correlate secondary bacterial infections in both groups with disease severity and outcome.
Our study, however, has several limitations. First, it is retrospective in nature and relies on proper documentation of cultures and clinical parameters in the medical records. The study might also suffer from selection bias since the decision to obtain bacterial cultures was done by the treating clinician and most probably was affected by the severity of the disease. Second, given the local screening policy of every admitted patient for COVID-19 during the pandemic, we also cannot rule out that a fraction of patients (especially among the COVID-19 group) was hospitalized due to other medical conditions implicated in bacteremia that were completely unrelated to their viral infection. Furthermore, since there were no influenza cases in Israel during the COVID-19 pandemic, the two groups of patients might reflect different time periods. Last, the study's data was from a single center, and thus the infecting bacteria might reflect a site-specific microbiological profile.Taken together, our results show that secondary bacterial infections, in particular late gram-positive infections, are a clinically important complication with a significant correlation to poor outcomes in hospitalized COVID-19 patients. This calls for increased awareness of the treating physicians to the possibility of secondary bacterial infection as an etiology to late deterioration, suggesting that antibiotic treatment may be an essential component of the therapeutic armamentarium in selected patients with severe COVID-19. As gram-positive bacteria are increasingly becoming resistant to antibiotics 15 , our results highlight the importance of implementing infection control measures specific for COVID-19 hospitalized patients in addition to modification of antibiotic treatment protocols. Blood or sputum cultures were taken routinely at admission or according to the discretion of the treating clinician. The samples were transferred to the microbiology laboratory and processed according to the current standard practice procedures 16 . Sputum or BAL specimens were cultured, some of the samples were also tested using BIOFIRE® FILMARRAY® Pneumonia plus Panel according to the discretion of the treating clinician. Blood cultures were analyzed using the BACTEC™ blood culture system /BACTEC™ FX system (Becton Dickinson, Inc). Positive cultures were further analyzed according to the current standard practice procedures 16 . Identification of isolates was performed using matrix-assisted laser desorption ionization-time-of-flight mass spectrometry. GeneXpert assay was performed directly on positive blood cultures suspected of containing staphylococci after the gram staining.
Bacterial growth during the first 48 h from admission was regarded as early or coinfection, and bacterial growth up to 2 weeks from admission was defined as a late infection. The study was approved by Rabin Medical Center Review Board (#RMC-20-0142) and the Tel Aviv University Ethics Committee (0001269-3) and was performed according to the Helsinki declaration. Statistical analysis. The patient's cohort was divided into two groups based on the viral infection diagnosis (influenza and SARS-CoV-2). Normality tests were conducted for all variables. Due to the non-normal distribution of our variables, we used the non-parametric Wilcoxon and Kruskal Wallis tests when appropriate, and median and interquartile range (IQR) are presented. Chi-square or Fisher's exact tests were used to compare categorical variables between study groups accordingly, p < 0.05 (two-tailed test) was considered statistically significant. Univariate and multivariate regression models were constructed to estimate predictors for the primary outcome. ORs with 95% confidence intervals (CI) were calculated. Logistic regression was used to calculate ORs. Statistical analysis was performed using SPSS 25, R 4.04, and GraphPad PRISM 8.