Introduction
Among the factors that could contribute to COVID-19 mortality and morbidity, the role of bacterial superinfections remains unclear. The latter were common during pandemics such the Spanish flu pandemic in 1918 [Reference Morens, Taubenberger and Fauci1] and H1N1 influenza A in 2009 [Reference Lee2], in which invasive species such as Staphylococcus aureus, Streptococcus pyogenes or S. pneumoniae were identified. Comprehensive data regarding superinfections which complicate COVID-19 would allow rationalisation of the prescription of antimicrobial agents. In the literature, one meta-analysis covering 30 studies reported rates of bacterial superinfections of 7% in hospitalised patients and 14% of ICU patients; figures that appeared to be lower than those with influenza [Reference Lansbury3]. Surprisingly, in the latter study, Mycoplasma pneumoniae was the most frequent pathogen detected in COVID-19 patients followed by Gram-negative bacterial species. Moreover, a retrospective of patients hospitalised with COVID-19 reported acquisitions of superinfections to be both community-acquired and healthcare associated [Reference Garcia-Vidal4]. Interestingly, S. pneumoniae and S. aureus were the most frequent aetiological agents of community-acquired bacterial pneumonia, while enterobacteria and Pseudomonas aeruginosa were mostly found in healthcare-associated infections. Although a number of published studies have reported cases of co-infections with COVID-19 [Reference Nori5–Reference Gerver10], they were conducted with a limited number of patients, often without a control group, thereby rendering it difficult to deduce whether such infections are actually associated with the acquisition of COVID-19 [Reference Santos11].
Since the beginning of the pandemic, we have diagnosed over 40 000 patients with COVID-19 and have access to microbiological results from patients admitted to public hospitals in Marseille. In this retrospective study, we aimed to provide a snapshot of bacterial superinfections in all COVID-19 patients for which microbiological investigations were conducted. To assess whether SARS-CoV-2 infection is associated with a specific epidemiology, we compared these results with those obtained from controls during the same period, as well as to a control group of influenza-positive patients between 2017 and 2019.
Methods
Study design and patients
This retrospective study was conducted at the IHU Méditerranée Infection in Marseille, France. We first extracted records of all patients with a positive diagnosis of COVID-19 (cases) between 26 February 2020 and 1 May 2021. A COVID-19 diagnosis was confirmed by real-time reverse transcription PCR (RT-PCR) testing performed on nasopharyngeal throat swab specimens, as previously described [Reference Lagier12]. Firstly, to compare infections among the cases with controls, we extracted the list of patients who had had at least one COVID-19 PCR test, and for whom all tests were negative (controls). Finally, we aimed to compare COVID-19 superinfections with influenza superinfections through extraction of all positive PCR results for influenza A, B or H1N1 [Reference Bal13] between 1 January 2017 and 31 December 2019. All microbiological tests results from cases, controls and influenza patients were analysed.
Data collection and outcomes
For all cases, controls, and influenza patients for whom a bacterial infection had been identified, information was collected from electronic health records concerning demographics (age, gender), length of hospital stay, outcomes including death, and admission to the intensive care unit (ICU).
Microbiological procedures
Microbiological investigations included standard culture for urinary samples and blood cultures [Reference Dubourg14]. Respiratory specimens (i.e. sputa, pleural fluids, bronchoalveolar lavages (BALs), bronchial aspirates and pulmonary biopsies) were examined by standard cultures [Reference Dubourg15]; urine specimens were tested for pneumococcal antigen [Reference Edelstein, Jørgensen and Wolf16], and BAL by S. aureus (PCR) [Reference Oh17].
Definitions
We defined a superinfection by a positive sample to any bacteria and collected between two days before and 30 days after the diagnosis of the viral infection among COVID-19 and influenza patients. We did not consider the detection of Candida albicans and Enterococcus spp. from respiratory samples in this study. Bacteraemia was defined as the growth of a commensal non-skin flora in at least one blood culture vial, and involving a common skin coloniser (coagulase-negative staphylococci or Corynebacterium species) when at least two vials sampled within 48 h were positive for the same species. When the timescale between sampling the positive specimen exceeded 48 h after initial admission, these were defined as healthcare-associated infections. We removed duplicate data when a patient was positive to the same bacterial pathogen from the same anatomical site. The positivity rate for each category (i.e. respiratory infections and bacteraemia) was calculated by dividing the number of positive patients by the number of patients for which at least one blood or respiratory specimen was taken following their admission.
Statistical analysis
Statistical analyses were performed using Graphpad 6 (La Jolla, USA) and XLSTAT 2019 (Addinsoft, New York, USA). Continuous and categorical variables were presented as median (interquartile range) and absolute number (percentage) respectively. The Mann–Whitney U test, chi-square test and Fisher exact test were used when appropriate. Significance was set at P < 0.05.
Results
Patients included
In total, 42 468 subjects were positive according to a COVID-19 RT-PCR test during the study period and were considered as cases. Of these, 631 had at least one superinfection, accounting for 1321 episodes. During the same period, 266 261 subjects tested negative for COVID-19 amplification and were used as controls. A total of 4731 infections were identified among 2884 controls. Finally, from 2017 to 2019 we identified 4059 patients with at least one positive PCR for influenza virus; of these, 124 had at least one superinfection, as previously defined.
Patients’ characteristics
The cases were older (68 years old vs. 58.9 years old and 56.6 years old; P < 0.0001) and were significantly more frequently male (sex ratio M/F: 2.9) compared to controls and influenza patients (sex ratio M/F: 1.6 and 1, respectively) (P < 0.0001) (Table 1). Among cases, the sex ratio of patients admitted to the ICU did not differ from that of patients hospitalised in other units. The proportion of ICU-acquired infections was significantly higher among cases (83.9%) compared to control and influenza groups (46.9% and 34.9%, respectively) (P < 0.0001) as well as to the proportion of hospital-acquired infections (82% vs. 61.1% and 44.1%, respectively) (P < 0.0001). The median length of stay in the ICU preceding the infection was longer among cases (9.63 days) compared to controls (6.71 days) and influenza patients (5.17 days).
Characteristics of infections occurring in COVID-19 patients and those occurring in uninfected subjects between 26 February 2020 and 1 May 2021, and those occurring among influenza-positive patients between 1 January 2017 and 31 December 2019s (***P < 0.001)
Values in bold indicate the group for which the difference is the most significant.
Respiratory infections
We identified a total of 1069, 2697 and 87 respiratory infections among 535 cases, 1492 controls, and 68 influenza patients, respectively (Table 1). When compared to control and influenza groups, respiratory infections were more frequent among controls (41.1%) than among cases and influenza patients (35.6% and 10.7%, respectively) (P < 0.001). However, respiratory infections among cases were more frequently healthcare-associated (84.3% vs. 70.7% and 48.2%, respectively) (P < 0.0001) and more frequently ICU-acquired (88.7% vs. 60.9% and 40.6%, respectively) (P < 0.0001) compared to the control and influenza groups, respectively. Cases with respiratory infections were significantly more frequently male (sex ratio M/F: 3.7 vs. 2.1 and 1.1, respectively), and were older (mean age: 65.3 years old) than the controls only (55.4 years old) (P < 0.0001). The median length of stay in the ICU preceding the infection was higher among cases (9.2 days) when compared to that of controls (6.8 days) and influenza patients (3.9 days), while the mean timescale for acquiring respiratory superinfections was higher among cases (9.8 days) than influenza patients (5.95 days). Regarding the species identified, S. epidermidis (P < 0.0001), K. aerogenes (P = 0.004) and Hafnia alvei (P = 0.001) were more frequently recovered from respiratory specimens of cases, while Haemophilus influenzae (P < 0.0001), Enterobacter cloacae complex (P = 0.003) and Escherichia coli (P = 0.001) were more frequent among while S. pneumoniae (P = <0.0001) was more frequently detected from influenza patients (Fig. 1a). Regarding ICU-acquired infections, H. alvei (P = 0.03) and S. epidermidis (P = 0.03) were more predominant in cases, while E. cloacae complex (P = 0.0001), E. coli (P = 0.0001) S. pneumoniae (P = 0.001) and H. influenzae (P < 0.0001) were more frequent among controls (Fig. 1b). The only observed difference in non-ICU settings was the significantly higher detection of S. pneumoniae among influenza patients (P < 0.0001) (data not shown). The latter species was also more common in community settings (i.e. 62.5%) out of the 10 most frequent species associated with COVID-19 respiratory superinfections.
Microorganisms detected in respiratory specimens, the frequency of which is significantly different among COVID-19 patients, controls and influenza-positive subjects (*P < 0.05; **P < 0.01; ***P < 0.001). (a) All units and (b) ICU only.
Bacteraemia
We identified a total of 252, 2034 and 79 bacteraemia among 209 cases, 1686 controls, and 67 influenza patients, respectively (Table 1). Bacteraemia was significantly more frequent among controls (14.6%) than cases (3.75%) and influenza patients (3.6%) (P < 0.0001). However, bacteraemia among cases were predominantly healthcare-associated (72.2% vs. 48.2% and 39.5%, respectively), more frequently ICU-acquired (63.5% vs. 28.5% and 22.8%, respectively, P < 0.0001), with an overrepresentation of male subjects (sex ratio M/F = 3.3 vs. 1.6 and 1.3, respectively, P < 0.0001) (Table 1). The median length of stay in the ICU preceding the infection was longer among cases (13 days) compared to controls (6.5 days) but not for influenza patients (14.3 days), while the mean timescale between viral diagnosis and acquisition of bacteraemia was longer among cases (9.1 days) than influenza patients (5.1 days). Regarding the species identified, S. epidermidis (P < 0.0001), E. faecalis (P < 0.0001) and P. aeruginosa (P = 0.002), were more frequently isolated from blood cultures of cases while E. coli (P ≪ 0.0001) and K. pneumoniae (P = 0.03) were more frequent among controls, and S. pneumoniae among influenza patients (P = 0.02) (Fig. 2a). Similar findings were apparent with ICU-acquired infections for S. epidermidis (P = 0.005), E. faecalis (P = 0.002), P. aeruginosa (P = 0.016), and E. coli (P < 0.0001), but not for K. pneumoniae and S. pneumoniae (Fig. 2b). In non-ICU settings, S. pneumoniae was, however, more frequent among influenza patients (P = 0.014), as well as E. coli among controls (P < 0.001). Of the 10 species most frequent among cases, only E. coli and S. hominis were more associated with community settings. Among COVID-19 patients, bacteraemia due to E. faecalis generally occurred later when compared to those due to other pathogens (13.7 days vs. 8.5 days, respectively) (P < 0.0001), and more frequently in ICU settings (87% vs. 52.8%, respectively, P < 0.0001), but were not associated with more deaths.
Microorganisms detected from blood cultures, the frequency of which is significantly different among COVID-19 patients, controls, and influenza-positive subjects between 1 January 2017 and 31 December 2019 (*P < 0.05; **P < 0.01; ***P < 0.001). (a) All units and (b) ICU only.
Discussion
In this study, we provide a snapshot of the superinfections (associated with COVID-19 over a 13-month period). Our data rely on microbiologic investigations that were assessed by microbiologists and validated by two study operators. First, we found that bacterial infections in COVID-19 patients were more frequently healthcare-associated, often acquired in ICUs, associated with a longer ICU stay, and occurred in older male patients, compared to controls and influenza subjects (Table 1). The usual main pathogens were less frequent in COVID-19 patients as illustrated by fewer bacteraemias involving E. coli and K. pneumoniae than among controls (Fig. 2a). Most importantly, the respiratory pathogens such as H. influenzae and S. pneumoniae, commonly involved in viral superinfections, were under-represented in COVID-19 patients compared with controls and influenza patients (Fig. 1a and b). These findings are in line with other studies agreeing that community-acquired superinfections are more common in COVID-19 positive subjects [Reference Adler18]. The present work showed that infections in COVID-19 patients are generally dominated by nosocomial-transmitted microorganisms, such as Gram-negative bacteria in respiratory infections, and coagulase negative staphylococci and P. aeruginosa in bacteraemia.
Nevertheless, it was surprising to observe the high frequency of E. faecalis bacteraemia among COVID-19 patients that were mainly ICU-acquired (87%) with a longer timescale of acquisition when compared to the other microorganisms. We wondered whether empirical antimicrobial therapy could have contributed to selecting enterococcal species, but fewer than the half of the patients with a positive blood culture for E. faecalis had received either cephalosporins or carbapenems prior to sampling (data not shown). Nevertheless, gut microbiota enrichment with E. faecalis has been associated with an abnormal immune response in COVID-19 patients [Reference Zhou19].
We did not find increased mortality in patients with bloodstream infections involving E. faecalis although other studies have also noted an unexpected frequency of bacteraemia involving enterococci [Reference DeVoe20], particularly in ICU settings with often equal distribution between E. faecalis and E. faecium [Reference Bonazzetti21, Reference Giacobbe22], One of the hypotheses raised is the possible increased permeability of the gut barrier induced by the virus when replicating in enterocytes. However, patients developed E. faecalis bacteraemia within a mean timescale of 13.7 days in our series, when the virus replicates little, rendering this hypothesis less probable.
The rate of superinfection during COVID-19 has been reported to be usually low, ranging from 6% to 8% depending on the studies [Reference Lansbury3, Reference Garcia-Vidal4, Reference Rawson23, Reference Hughes24]. Our study shows that COVID-19 patients experience more respiratory superinfections than individuals with influenza, due to a high proportion of pathogens with nosocomial transmission, whereas the rate of bloodstream infections is similar. Discrepant results were also reported which found similar rates of superinfection between COVID-19 patients and influenza-positive individuals, but it remains difficult to extract how many infections were ICU-acquired [Reference Hughes24].
We acknowledge that the clinical application of our findings could be limited by the fact that the study was conducted primarily from a microbiological viewpoint, and that patients were not classified according to the severity of their illness. However, stringent inclusion criteria were used for the definition of bacteraemia to avoid inclusion of contaminants when collecting our data, and interpreted the results according to relevant parameters, including the proportion of ICU admissions, and the length of stay in the ICU. It is also acknowledged that only standard cultures and two rapid assays (i.e. pneumococcal antigen detection in urine and molecular detection of MRSA) were utilised and possibly may have missed pathogens detected by other approaches such as serological testing or other specific PCRs. Finally, we did not analyse the antibiotic treatments administered during the patient's stay. Indeed, antimicrobial agents were largely prescribed for COVID-19 patients, which could have thereby reduced the incidence.
Overall, this study shows that the superinfections occurring with COVID-19 are more frequent than with influenza primarily due to prolonged stays in the ICU. Pathogens that are commonly isolated from patients presenting with routine clinical infections are under-represented, while those most often associated with nosocomial transmission were predominant. Notably, there was an unexpectedly high frequency of Enterococcus faecalis bacteraemia among COVID-19 patients, a finding which warrants further investigation.
Acknowledgements
None.
Author contributions
Conceptualisation: G. D., D. R.; Methodology: D. R., G. D., E. L. G.; Validation: G. D., D. R.; Formal analysis: E. L. G., G. D.; Investigation: E. L. G., G. D.; Writing: E. L. G., G. D., D. R.; Visualisation: E. L. G., G. D.; Supervision: G. D., D. R.
Financial support
This work was supported by the French Government under the ‘Investments for the Future’ programme managed by the National Agency for Research (ANR), Méditerranée-Infection 10-IAHU-03 and was also supported by Région Provence Alpes Côte d'Azur and European FEDER PRIMMI funding (European Regional Development Fund–Plateformes de Recherche et d'Innovation Mutualisées Méditerranée Infection), ERDF PA 0000320 PRIMMI.
Conflict of interest
Didier Raoult has been a consultant for Hitachi High-Technologies Corporation, Tokyo, Japan, from 2018 to 2020. He is a scientific board member of Eurofins company and a founder of a microbial culture company (Culture Top). Other authors declare no conflicts of interest.
Ethical standards
All data were generated as part of routine work at the Assistance Publique-Hôpitaux de Marseille (APHM) (Marseille university hospitals), and this study results from routine standard clinical management. Access to the patients’ biological and registry data issued from the hospital information system was approved by the data protection committee of the APHM and was recorded in the European General Data Protection Regulation registry under number RGPD/APHM 2021-131.
Consent to participate
Not applicable.
Consent for publication
All authors have approved the manuscript and gave their consent for submission and publication.
Data availability statement
The data that support the findings of this study are not openly available, as requested by the data protection committee of APHM.
Open access
