Pulmonary bacterial infections in patients hospitalized for COVID-19: a retrospective observational study

Backround During the COVID-19 pandemic, antibiotics use was very common. However, bacterial co/secondary infections with coronaviruses remain largely unknown, especially outside of intensive care. The aim of this study was to investigate the pulmonary bacterial infections characteristics associated with COVID-19 in hospitalized patients. A retrospective monocentric observational study was conducted in Bichat hospital in France, between February 26 and April 22, 2020. All patients hospitalized in standard wards with COVID-19 (positive nasopharyngeal PCR and/or typical aspect on CT scan) and diagnosed with a pulmonary bacterial infection (positive bacteriological samples) were included. Bacteriological and clinical data were collected from the microbiology laboratories and the patient's medical records.

deaths [3,4]. Pulmonary bacterial infections, a key factor in the potential severity of viral infections [5], are still poorly known with Human coronavirus. Few studies investigated the co-infection and secondary infection rates. One reported up to 10% rate with Mycoplasma, Legionella and Streptococcus pneumonia in Severe Acute Respiratory Syndrome (SARS) [6] and 1% with Mycoplasma, Legionella and Chlamydia spp. in Middle East Respiratory Syndrome (MERS) [7]. Recent studies regarding COVID-19 reported highly variable rates of bacterial infection, up to 45% [8,9]. However, most studies focused on intensive care patients and ventilator associated pneumonia [10,11].
Antibiotics have been extensively used during the rst outbreaks of SARS-CoV-2 [8 12]. The main reasons were symptom similarities between COVID-19 and pulmonary bacterial infections, the severity of SARS-CoV-2 pneumonia, and the lack of knowledge of the virus pathogenicity. Based on early data reported on COVID-19, it is time to question whether the broad use of antibacterials is warranted, especially in the context of rising antibiotic resistance.
Thus, a retrospective study of documented pulmonary bacterial co/secondary infections in COVID-19 hospitalized patients in standard wards was conducted. The aim of this study was to investigate the characteristics of pulmonary bacterial infections associated with COVID-19 in hospitalized patients.

Study design
A monocentric retrospective observational study was conducted in Bichat University Hospital in Paris, France, during the SARS-CoV-2 outbreak. All departments in charge of SARS-CoV-2 infected patients, except intensive care units, participated in this study. The microbiological database of the hospital's bacteriological department was screened to identify positive respiratory tract secretions sample (sputum samples, bronchial aspirations and bronchoalveolar lavages), urinary antigen (pneumococcal and Legionella) tests and blood cultures of hospitalised SARS-CoV-2 infected patients, between February 22 and April 22, 2020. The hospital activity based payment registry was assessed for all patients diagnosed with COVID-19, to recover positive bacterial nasopharyngeal PCR (QIAstat-Dx® or BIOFIRE® RP2.1+) for hospitalised patients during the same period. Microbiological investigations and biological samples were obtained as part of the routine patient care, at the discretion of the treating physician.
Sputum cultures were considered positive regardless of the culture threshold if one or several bacteria were isolated from quality sample, de ned by at least > 10 leukocytes/ eld and < 25 epithelial cells/ eld. Film arrays® on sputum samples, bronchial aspirations and bronchoalveolar lavages were included if one or several bacteria were found, regardless of the threshold. Data collection was further completed using the hospital electronic medical records of patients identi ed. de ned as either co-infection at the symptoms onset, or secondary infection occurring during the course of illness or hospitalisation [8].

Patient inclusion
All patients with the following criteria were included: Patients over 18 years old hospitalised in a ward in charge of SARS-CoV-2 infection.
SARS-CoV-2 infection diagnosed on microbiological criteria (positive PCR on nasopharyngeal swab) and/or imaging criteria (typical aspect on CT scan).
Positive bacterial exam with either positive respiratory tract secretions sample (PCR, sputum samples, bronchial aspirations and bronchoalveolar lavages), urinary antigen (pneumococcal and Legionella) tests or blood culture.
Patients were excluded if they met the following criteria: Invasive ventilation prior to the documented bacterial exam.
Positive blood culture with presumed extra-pulmonary origin or considered as a contamination.
Positive respiratory tract sample considered as a colonisation.

Data collection
A standardized form on Excel was used for data collection. Data were collected on demographics, bacterial infection risk factors, clinical parameters, in ammatory biomarkers within the 24 hours following the day of the suspected infection and the previous and following values (at least 48 hours from the suspected day of infection), bacteriological samples (respiratory tract secretions samples, urinary antigen (pneumococcal and Legionella) and blood culture), thoracic CT scan on admission and in the 48 hours following the day of the suspected infection, speci c COVID-19 antiviral and immunomodulatory treatment, empirical antibiotic therapy, duration of treatment, and nal outcome at discharge from the unit.
Depending on the severity of the viral infection, patients were treated according to the local guidelines at the time of the screening and research protocols, using high dose of glucocorticosteroids (Dexamethasone) and immunomodulators (Anakinra or Tocilizumab). The imaging classi cation of the present study was based on the French Society of Radiology's SARS-CoV-2 classi cation system: mild (< 10%), moderate (10-25%), extended (25-50%), severe (50-75%) and critical (> 75%) [13]. Patients were treated by the attending physicians from the participating wards, who decided whether it was an actual infection requiring antibacterial treatment.

Statistical analysis
All continuous variables were either expressed as median and interquartile range or means and con dence intervals. Categorical variables were expressed in number and percentage. Paired samples t-test was used to compare continuous variables among patients at different time points. All statistical analyses were performed using Excel sheets.

Results
A total of 784 patients were hospitalized with COVID-19 during the inclusion period. Forty-six among 2075 (2.2%) samples were positive for bacteria of which 23 samples were excluded. Finally, 23 samples (1.1%) from 22 patients (2.8%) were identi ed as secondary pulmonary infections and included in the nal selection (Fig. 1). Patients' characteristics are presented in Table 1 Table 2.  Total number of pathogens exceeds total number of samples since more than one isolate have been identi ed in a single culture (with a maximum of three bacteria). BAL: broncho alveolar lavage The mean level of neutrophil count before infection was 7905/mm3, at the day of infection 9546/mm3 and after 8056/mm3 ( Fig. 2A). The differences were non-signi cant, respectively p = 0.15 and p = 0.07. Similarly, there was no difference for the mean CRP levels before (124 mg/dl), during (98 mg/dl) or after bacterial infection (78 mg/dl), respectively p = 0.37 and p = 0.26 (Fig. 2B). Leukocyte count data are not shown, and the differences were also non-signi cant. Lack of data prevented procalcitonin (PCT) analyses.
At admission, 11 patients (50%) had mild to moderate parenchymal involvement and 17 patients (77.3%) had pulmonary consolidation. Only ten patients (45.5%) had a CT-scan in the 48 hours following the suspicion of secondary infection, preventing comparison between the diagnosis day and the suspected bacterial infection day. Among them, the CT scan control showed a lung consolidation appearance in ve patients (50%) and the worsening of at least one previous lung consolidation in three patients (30%).
Finally, among the 22 patients with documented bacterial infections, nine (40.9%) were discharged from hospital, ve (22.7%) were transferred to an intensive care unit and eight (36.4%) died in the standard ward. The transfer to intensive care unit occurred mostly the day of the bacterial infection suspicion. Out of these ve patients, four survived after the intensive care, and one died.
A total of 496 multiplex PCRs on nasopharyngeal swab were carried out. Seven were positive including three samples later excluded due to the absence of associated SARS-CoV-2 infection (Fig. 3). Four patients (0.8%) had co-infections: two Mycoplasma pneumonia and two Bordetella pertussis. The median age was 56 years, 3 were men. The patients with whooping cough, including one with an underlying sarcoidosis under corticoids, were more severe, requiring oxygen support. Mycoplasma pneumonia infections were from a family cluster and mild, without any speci c symptoms.

Discussion
In Overall, only a very small fraction of microbiological testing revealed positive samples, leading to the positive diagnosis of secondary infection. Recent meta-analyses showed an overall prevalence of 7% of bacterial infection in hospitalized COVID-19 patients [8,9], with 3,5% of co-infections and 14.3% of secondary infections [8]. It is noteworthy that there was a high heterogeneity including various populations (ICU and non ICU patients), antibiotics policies and microbiological samplings. Furthermore, several recent observational studies showed a bacterial infection rate between 4,7 and 6,1% [14][15][16], including for some of them ICU and non ICU patients and bacteremia from extra-pulmonary origin. This low rate of bacterial infection has to be confronted with the rate of antibiotic prescription in patients with COVID-19, reported as high as 72% [8,16]. Broad spectrum antibiotics were mainly used and varied with the study. The extreme protective and cleaning measures in place for COVID-19 patient treatment can also be highlighted as a potential limiting factor in secondary bacterial infection occurrence.
The present study showed a relatively high proportion of aged men and comorbities among its patients including underlying pulmonary diseases (COPD, asthma), underlying immunosuppression factors (diabetes mellitus, kidney failure …) or immunosuppressive therapy. The lack of control group prevented any comparison to highlight potential risk factors for co or secondary infection.
Microbiological results were varied with a balanced ratio between gram-positive cocci and gram-negative bacilli. Main bacteria were Pseudomonas aeruginosa, Staphylococcus aureus and Streptococcus pneumonia. This distribution contrasts with previous ndings on other viruses describing a possible association between in uenza virus, rhinovirus, hMPV and S. aureus, and Klebsiella spp [17]. Similarly, the same study suggested an association of regular coronavirus, parain uenza virus and RSV infections with Acinetobacter spp. and Klebsiella spp. The previously described meta-analysis showed mainly M. pneumonia (42%), P. aeruginosa (12%) and H. in uenza (12%) infection in SARS-CoV-2 infected patients [9]. The proportion of M. pneumonia was surprisingly high and very likely overestimated while relying on IgM serology [18]. Most included studies were from Asian countries, which might explain the relatively high proportion of Acinetobacter infection. Hugh et al. [15] and Townsend et al. [16] showed more similarities with the present study.
The ndings of the present study highlighted diagnosis di culties. Clinical features leading to pulmonary bacterial infection suspicion can be easily confused with the usual signs of an advancing SARS-CoV-2 infection [19]. Despite sputum production seems to be an interesting sign, it has been reported in one third of SARS-CoV-2 infected patients and remains not speci c of bacterial infection [20]. Additionally, bacterial infection occurrence within the COVID-19 timeframe, around the 10th day after symptoms' onset, corresponds to the classical worsening of COVID-19 symptoms [21], which adds to the di culty of the identi cation of a bacterial infection. Furthermore, in the studied population, biological features did not show any difference between a potential bacterial infection and the underlying evolution of COVID-19.
Imaging bene ts in support of bacterial infection diagnosis should be further investigated. No distinctive marker has been highlighted in COVID-19 yet. Huang and colleagues [22] described PCT level in blood on admission in SARS-CoV-2 patients showing 75% of PCT > 0.5 ng/mL in con rmed secondary infections, whereas 69% of the other patients had a PCT level < 0.1 ng/mL. In contrast to the previous ndings, Wan et al. [23] found that CRP and PCT levels on admission of severe patients were signi cantly higher than in the mild patients group, with no difference in documented bacterial infection rate. PCT levels could not be assessed in this study because of missing data.
The nal outcome was positive in only 62% of the selected patients, with a third requiring intensive care before being discharged. Study design did not allow to claim that poor outcome in this population was the consequence of secondary bacterial infections. However this ascertainment seems consistent with the literature [24,14], and might mainly involve patients with bacteraemia [15].
The design of this study was retrospective. Databases screening were conducted with two different methods for PCR and bacteriological samples data collection to optimise the extraction quality and limit a potential selection bias. Despite authors' endeavour, both extraction methods showed some aws and resulted in additional exclusions during the selection process. Though all patient results were thoroughly described and analysed, there were some missing data, especially for PCT levels and CT scans, preventing further analyses.
In order to catch all potential secondary infections, no threshold was held for microbiological samples in the screening of cases. Thus, a proportion of positive samples could be the result of colonisation and not actual infection leading to an overestimation of bacterial infections. It has to be put in balance with actual non documented infection, with no bacterial investigation or false negatives due to the large presumptive use of antibacterials or poor quality sampling. However for the latter, during a crisis and outside of ICU, only regular non-invasive microbiological samples are easily available and usable.
There is growing evidence suggesting a global overestimation of pulmonary bacterial infection in SARS-CoV-2 infected patients, resulting in an overuse of antibiotics and its consequences, mainly an increase of bacterial resistance [25]. This work supports the low prevalence of co infection and secondary infection in non ICU hospitalised patients with COVID-19. Empirical use of antibiotics are unlikely to provide signi cant bene t to COVID-19 patients. Systematic use of antibiotic, especially uoroquinolones or macrolides on admission, does not seem justi ed. To support these ndings, prospective studies are required to collect further data on the actual prevalence of bacterial infection in COVID-19 patients. Future research needs to focus on the role of antibiotics in SARS-CoV-2 infected patients, to guide its use in daily care and hopefully reduce the adverse consequences of its overuse.

Conclusion
Pulmonary bacterial secondary and co-infections with SARS-CoV2 are uncommon. Their diagnosis is di cult due to similarities with the natural course of the disease. Cough with sputum around day ten might be a sign of bacterial infection. There is a balanced ratio between gram-positive and gram-negative among involved bacteria.