Characterization of antimicrobial use and co-infections among hospitalized patients with COVID-19: a prospective observational cohort study

Purpose To investigate antimicrobial use and primary and nosocomial infections in hospitalized COVID-19 patients to provide data for guidance of antimicrobial therapy. Methods Prospective observational cohort study conducted at Charité–Universitätsmedizin Berlin, including patients hospitalized with SARS-CoV-2-infection between March and November 2020. Results 309 patients were included, 231 directly admitted and 78 transferred from other centres. Antimicrobial therapy was initiated in 62/231 (26.8%) of directly admitted and in 44/78 (56.4%) of transferred patients. The rate of microbiologically confirmed primary co-infections was 4.8% (11/231). Although elevated in most COVID-19 patients, C-reactive protein and procalcitonin levels were higher in patients with primary co-infections than in those without (median CRP 110 mg/l, IQR 51–222 vs. 36, IQR 11–101, respectively; p < 0.0001). Nosocomial bloodstream and respiratory infections occurred in 47/309 (15.2%) and 91/309 (29.4%) of patients, respectively, and were associated with need for invasive mechanical ventilation (OR 45.6 95%CI 13.7–151.8 and 104.6 95%CI 41.5–263.5, respectively), extracorporeal membrane oxygenation (OR 14.3 95%CI 6.5–31.5 and 16.5 95%CI 6.5–41.6, respectively), and haemodialysis (OR 31.4 95%CI 13.9–71.2 and OR 22.3 95%CI 11.2–44.2, respectively). The event of any nosocomial infection was significantly associated with in-hospital death (33/99 (33.3%) with nosocomial infection vs. 23/210 (10.9%) without, OR 4.1 95%CI 2.2–7.3). Conclusions Primary co-infections are rare, yet antimicrobial use was frequent, mostly based on clinical worsening and elevated inflammation markers without clear evidence for co-infection. More reliable diagnostic prospects may help to reduce overtreatment. Rates of nosocomial infections are substantial in severely ill patients on organ support and associated with worse patient outcome. Supplementary Information The online version contains supplementary material available at 10.1007/s15010-022-01796-w.


Introduction
The Coronavirus Disease 2019 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), has major implications for health care worldwide. Bacterial co-infections are common in viral pulmonary infections, such as influenza, significantly contributing to mortality [1,2]. While co-infections are reportedly low among COVID-19 patients with rates of primary bacterial co-infections of 1-4% [3][4][5][6], use of antimicrobial therapy upon admission is reported in up to 74% [7]. The rate of nosocomial co-infections varies between 4 Tilman Lingscheid and Lena J Lippert contributed equally. Likewise, Miriam S Stegemenn and Florian Kurth should be noted as senior authors with equal contribution.

3
and 50%, depending on disease severity and level of care (non-intensive care unit (non-ICU) vs. ICU) [8]. Data on microbial workup in COVID-19 are sparse with only few studies reporting the most common pathogens or antimicrobial use, and there is hardly any data differentiating between primary and nosocomial co-infections [8,9]. There is clear research need to define the exact incidence of co-infections at the different phases of COVID-19, antimicrobial susceptibility profiles and risk factors to establish evidence-based antibiotic stewardship interventions for COVID-19 patients.
We performed a prospective observational cohort study investigating antimicrobial use and primary and nosocomial bacterial and fungal pathogens in respiratory and blood samples from hospitalized COVID-19 patients between March and November 2020 in a German tertiary care centre. We analysed patient characteristics, microbiological, clinical and laboratory data, aiming to identify determinants for co-infections, provide evidence for guidance of optimized antimicrobial therapy, and prioritize antimicrobial stewardship interventions.

Methods
Data collection was performed within the Pa-COVID-19 study, a prospective observational cohort study conducted at a tertiary care university hospital of Charité-Universitätsmedizin Berlin, as described before [10]. The study was approved by the ethics committee of Charité-Universitätsmedizin Berlin (EA2/066/20), conducted according to the Declaration of Helsinki and Good Clinical Practice principles (ICH 1996), and registered in the German and WHO international clinical trials registry (DRKS00021688).
All adult patients with PCR-confirmed SARS-CoV-2 infection admitted between March and November 2020 were eligible for inclusion. Diagnosis and treatment followed national guidelines and was independent of study participation. All samples were taken within standard of care, and all results of microbiological specimen from blood cultures (BCs) and lower respiratory tract samples (RSs) and antimicrobial prescription data are included in this analysis. Comorbidities were assessed using Charlson Comorbidity Index (CCI) [11]. COVID-19 severity was stratified by WHO ordinal scale of clinical improvement [12].
The primary objective was characterizing antimicrobial use and microbiologically confirmed blood stream (BSI) and respiratory co-infections in hospitalized patients with COVID-19, stratified by time of occurrence (≤ 48 h after admission: primary co-infections; > 48 h after admission: nosocomial co-infections). For primary co-infections, only primarily admitted patients (directly admitted to Charité hospital or transferred ≤ 48 h after admission) were considered.
BSIs were classified as follows: in primary co-infections, isolates of typical skin commensal and coagulase-negative staphylococci (ConS) were excluded due to the high probability of contamination among patients just admitted to the hospital. In nosocomial BSIs, ConS were documented but not included in the analysis of risk factors and outcome because of uncertainty regarding true infection or contamination. For RSs, all isolated pathogens except Candida spp. and non-specified yeast were considered possible infectious agents. Among patients on invasive mechanical ventilation (IMV), RS were collected via tracheal aspiration (either via conventional aspiration or bronchoscopy); patients not on IMV who reported productive cough were instructed on submitting a (morning) sputum sample.
Differentiation between primary respiratory co-infection and colonization was made by treating physicians with review by an infectious disease specialist (based on radiological findings, laboratory parameters, and clinical worsening without antimicrobial therapy). Antimicrobial treatment without clear focus mentioned in medical charts was defined "empirical for suspected co-infection". Courses of C-reactive protein (CRP) and procalcitonin (PCT) were recorded over the hospital stay, and values at the day of onset of every BSI or respiratory tract infection (± 2d) compared to the lowest corresponding value during the 7 days prior.
Regarding microbiological workup, blood culture bottles were incubated in the BACTEC FX blood culture system (BD, Switzerland). Positive bottles were subcultured on conventional solid media and incubated in aerobic and anaerobic atmospheres at 37 °C. Respiratory samples were cultured on conventional solid media and incubated under aerobic conditions. Aerobic microorganisms (bacteria and yeast) were routinely identified using Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF) using the VitekMS system or by biochemical means using the VITEK 2 system (bioMérieux, France).
Antibiotic susceptibility testing (AST) was conducted in the VITEK 2 System. If necessary, additional commercial methods were applied, such as the disk diffusion method, E-tests or broth microdilution. Anaerobes were identified by MALDI-TOF and tested for susceptibility using ATB ANA (bioMérieux, France) or E tests. If Aspergillus spp. were cultured from respiratory samples, phenotypic identification of the species complex was done by macroscopic features and microscopic examination of cellotape flag preparations mounted in lactophenol cotton blue. AST was conducted according to recommendations of the European Committee on Antimicrobial Susceptibility Testing (EUCAST) and results were interpreted according to EUCAST Breakpoint tables for interpretation of MICs and zone diameters, version 10.0, 2020 [13].
Distribution of continuous variables was summarized by median and interquartile range (IQR). Differences of continuous variables between groups were examined by Mann-Whitney U test. Categorical variables were compared using chi-square and Fisher's exact tests. For all analyses, complete cases were used for the respective evaluation. A p value < 0.05 was considered statistically significant. Analyses were conducted with JMP (version 15 pro, SAS Institute Inc., Cary, United States) and GraphPad Prism (version 9, GraphPad Software Inc., San Diego, United States).

Results
309 patients hospitalized for COVID-19 were included in this study. Of those, 231 (74.8%) patients were primarily admitted to Charité hospital, whereas 78 patients (25.2%) were transferred from another centre. Patient characteristics are summarized in Table 1.
Six BCs were positive (positivity rate 4.3%, 6/137), three in patients with a defined infection focus (pyelonephritis, cholangitis, perforated peptic ulcer) and three without defined focus. In all three primary BSIs, penicillin-sensitive streptococci were isolated. RSs were positive in 12 patients (positivity rate 32.4%, 12/37), of which four were classified as colonization and eight (3.5%, 8/231) as bacterial superinfection. The isolated pathogens are shown in Supplementary  Table 2. Age ≥ 65 years and number of comorbidities were risk factors for primary co-infection in univariate analysis, and patients with co-infections had significantly higher levels of CRP and PCT (Table 1).
Aspergillus (A.) spp. was isolated in RSs of 16 patients (16/309, 5.1%, 13 A. fumigatus, three A. niger) plus one positive galactomannan test without cultural growth of Aspergillus spp.. All patients with Aspergillus spp. isolates were on IMV. Median time from admission and intubation to isolation of Aspergillus spp. was 19d (IQR 4-37) and 15d (IQR 3.5-34), respectively. Computed tomography (CT) substantiated the diagnosis of invasive pulmonary aspergillosis (IPA) in two cases, eight patients showed consolidations consistent with possible IPA, and five patients had no radiological signs of IPA (CT scans missing for two patients). Case fatality in patients with detection of Aspergillus spp. was 47% (8/17) and not significantly higher than in patients without Aspergillus spp. (34%, 31/91; OR 1.7 95%CI 0.6-4.9, p = 0.3). Use of dexamethasone was significantly associated with detection of Aspergillus spp. (

AMR and antimicrobial use over hospital stay
Overall, half of the patients (152/309, 49.2%) in the studied cohort received antimicrobial therapy at least once during their hospital stay, 131 of those (86.2%) were treated on ICU at some point. Among those treated, a median of 3.5 different antimicrobials was used (IQR 2-7). AMR pathogens were isolated in 37/309 (12%) patients, all treated on ICU, and patients with detection of AMR pathogens received a median of seven antimicrobials (IQR 4-10). Among BSIs (Table 2), a small proportion of pathogens exhibited AMR, except A. baumannii isolates. Among RSs, 15% (18/120) of first isolated pathogens displayed AMR (five CR and 13 3GCR isolates), which increased slightly to 22.8% (13/57) in isolates of the second RS (five CR and eight 3GCR isolates).
Bold values indicate p < 0.05

Laboratory markers
Patients with confirmed co-infections had higher levels and more spikes of CRP and PCT than patients without (Fig. 3A-D). Onsets of BSIs were accompanied with a median PCT increase of 233% (IQR 67.4-854.5%) and a median CRP increase of 162% (IQR 104-439%). Laboratory changes were less pronounced in respiratory infections, with a median PCT increase of 18.9% (-14.1-224.2%) and CRP increase of 114.9% (IQR 14.61-290.5%). There was no difference in the levels of CRP, PCT or leucocyte count regarding the isolation of gram-positive, gram-negative, or fungal pathogens. Supplementary Table 3 summarizes inflammatory parameters in relation to nosocomial infections.

Discussion
We present in-depth data on antimicrobial use and primary and nosocomial co-infections within a prospective cohort study of hospitalized COVID-19 patients, identifying risk factors for the primary and nosocomial co-infections and providing evidence of significantly worse outcome in coinfected patients with COVID-19. We included ICU and non-ICU patients, and both primarily admitted patients and patients referred due to clinical worsening and need for acute respiratory distress syndrome (ARDS) therapy including ECMO. Thus, this data is representative of the heterogeneous disease course of patients hospitalized for COVID-19 and reflects a real-life clinical picture.
BC sampling upon admission was performed in nearly 60% of patients, while respiratory sampling was performed in only 16%. Sparse respiratory sampling in COVID-19 has been reported previously and might be partly attributable to safety concerns [5]. Sampling was significantly associated with increased CRP und PCT and clinical worsening. Almost half of the patients with RS taken were already on IMV, facilitating sampling also regarding safety concerns of SARS-CoV-2 aerosolization.
Primary microbial co-infections were rare in our cohort (4.8%) and particularly primary BSIs were negligible at around 1%. These findings are slightly lower than previous studies reporting BSI rates of around 3% [5,14,15]. Primary pulmonary co-infections were more frequent with 3.5% and comprised mostly gram-positive bacteria, which is in accordance with prior studies [8]. Older age and number of comorbidities were identified as risk factors. Compared with bacterial co-infections in influenza, where they add significantly to morbidity and mortality, incidence of primary bacterial co-infections is drastically lower in COVID-19 [2,16]. In contrast to the small numbers of primary co-infections, 26.8% of patients received antimicrobial treatment at primary admission. Upfront antimicrobial therapy was associated with and probably triggered by increased inflammatory parameters, disease severity, and clinical deterioration, which is in line with other reports [9]. Thus, high inflammation parameters trigger anti-infective therapy which actually are elevated in case of a primary co-infection, however, initiated antimicrobial therapies outnumber true co-infections by far. About three-fourths of patients were treated empirically,  PCT and D CRP). Every row represents one patient and each column 1 day. In A and C " + " represents a positive microbiological respiratory sample and "*" a confirmed blood stream infection. CRP c-reactive protein; PCT procalcitonin; d day mainly for bacterial pulmonary co-infection analogous to community acquired pneumonia. Herein, the use of macrolides, whose anti-inflammatory role is known for severe pneumonia yet has no proven efficacy in COVID-19, [17,18] requires further research. That 56.4% received antimicrobial treatment and more broad-spectrum substances at referral is likely reflective of their critical condition and need of ICU-treatment.
Compared to previous studies, we report higher numbers of nosocomial microbial co-infections (15% of all patients), which might be explained by the concise sampling practice and the high proportion of ICU-patients. In addition, we used a longitudinal approach and did not focus on early infections only [3,19]. Nosocomial infection rate on ICU was 28% in our cohort and is in line with what has been reported before [20]. Respiratory co-infections occurred in 29.4% of all patients, most of them on IMV with an overall rate of 52.7%. The event of any nosocomial bloodstream or respiratory infection was significantly associated with longer hospital stay (44 days (IQR 30-73) vs. 10d (7-15), p < 0.0001) and in-hospital death (33.3%, 33/99 vs. 11.2%, 23/205) with an odds ratio of 3.95 (CI95% 2.16-7.22, p < 0.0001) reflecting the seriousness of these co-infections.
We found considerable numbers of enterococci, Candida spp., P. aeruginosa and Enterobacteriaceae in nosocomial BSIs. The frequency of A. baumannii and P. aeruginosa isolates with AMR was caused by a contained outbreak on two ICUs. Taking into consideration that all BSIs occurred in ICU patients, most on more than one organ replacement therapy, the pathogen spectrum reflects invasive ICU therapy rather than patient characteristics. In detail, we show that IMV, ECMO and HD are significantly associated with nosocomial infections. In line with this finding, duration of organ replacement therapies were associated with candidaemia, possibly explained by multiple and longer inserted catheters.
Our study has some limitations. It was performed in an ARDS/ECMO referral centre with a patient collective likely to be severely ill. Urinary samples and viral co-infections were not included, and sampling was done as part of clinical standard procedures, so we cannot exclude co-infections missed due to incorrect sampling.
In summary, we show that primary microbial co-infections in COVID-19 are rare. Nosocomial co-infections are associated with severe disease and organ replacement therapies and significantly contribute to COVID-19 mortality. Given the low numbers of pathogens isolated yet high antimicrobial use, antimicrobial therapy in COVID-19 should be accompanied by antimicrobial stewardship interventions. Our study adds solid data on incidence and causative organisms of co-infections in COVID-19, a basis for informed guidance of empirical antimicrobial therapy to optimize patient outcomes and combat the spread of AMR.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.