key: cord-0700093-a4hqdb24
authors: Grasselli, Giacomo; Scaravilli, Vittorio; Mangioni, Davide; Scudeller, Luigia; Alagna, Laura; Bartoletti, Michele; Bellani, Giacomo; Biagioni, Emanuela; Bonfanti, Paolo; Bottino, Nicola; Coloretti, Irene; Cutuli, Salvatore Lucio; De Pascale, Gennaro; Ferlicca, Daniela; Fior, Gabriele; Forastieri, Andrea; Franzetti, Marco; Greco, Massimiliano; Guzzardella, Amedeo; Linguadoca, Sara; Meschiari, Marianna; Messina, Antonio; Monti, Gianpaola; Morelli, Paola; Muscatello, Antonio; Redaelli, Simone; Stefanini, Flavia; Tonetti, Tommaso; Antonelli, Massimo; Cecconi, Maurizio; Foti, Giuseppe; Fumagalli, Roberto; Girardis, Massimo; Ranieri, Marco; Viale, Pierluigi; Raviglione, Mario; Pesenti, Antonio; Gori, Andrea; Bandera, Alessandra
title: Hospital-acquired infections in critically-ill COVID-19 patients.
date: 2021-04-20
journal: Chest
DOI: 10.1016/j.chest.2021.04.002
sha: 74682ba3cc51b276bd54a8087f71d710684eea48
doc_id: 700093
cord_uid: a4hqdb24

Background Few small studies have described hospital-acquired infections (HAIs) during COVID-19. Research Question What patient characteristics in critically ill patients with COVID-19 are associated with HAIs and how do HAIs associate with outcomes in these patients? Study Design and Methods Multicenter retrospective analysis of prospectively collected data including adult patients with severe COVID-19, admitted to 8 Italian hub hospitals from February 20, 2020, to May 20, 2020. Descriptive statistics, univariable and multivariable Weibull regression models were used to assess incidence, microbial etiology, resistance patterns, risk factors (i.e., demographics, comorbidities, exposure to medication), and impact on outcomes (i.e., ICU survival, length of ICU and hospital stay and duration of mechanical ventilation) of microbiologically-confirmed HAIs. Results Of the 774 included patients, 359 (46%) patients developed 759 HAIs (44.7 infections/1000 ICU patient-days, 35% multi-drug resistant (MDR) bacteria). Ventilator-associated pneumonia (VAP) (389, 50%), bloodstream infections (183, 34%), and catheter related blood stream infections (74, 10%) were the most frequent HAIs, with 26.0 (23.6-28.8) VAPs/1000 patient intubation-days, 11.7(10.1-13.5) BSIs/1000 ICU patient-days, and 4.7 (3.8-5.9) CRBSIs/1000 patient-days. Gram-negative bacteria (especially Enterobacterales) and Staphylococcus aureus caused 64% and 28% of VAPs. Variables independently associated with infection were age, PEEP and treatment with broad-spectrum antibiotic at admission. 234 patients (30%) died in ICU (15.3 deaths/1000 ICU patient-days). Patients with HAIs complicated by septic shock had almost doubled mortality (52% vs. 29%), while non-complicated infections did not affect mortality. HAIs prolonged mechanical ventilation (24(14-39) vs. 9(5-13) days; p<0.001), ICU and hospital stay (24(16-41) vs. 9(6-14) days, p=0.003; and (42(25-59) vs. 23(13-34) days, p<0.001). Interpretation Critically-ill COVID-19 patients are at high risk for HAIs, especially VAPs and BSIs due to MDR organisms. HAIs prolong mechanical ventilation and hospitalization, and HAIs complicated by septic-shock almost doubled mortality.

Patients affected by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) 2 infection can develop coronavirus disease 2019 , which is associated with a high 3 rate of hospitalization, admission to Intensive Care Unit (ICU) 1 , and death 2 . 4

For several weeks, starting February 20, 2020, Italy was the epicenter of the first COVID-19 5 outbreak in the western world 3 . In two previous studies, we reported 1,4 that critically ill 6 COVID-19 patients were primarily relatively old males with several comorbidities (e.g., 7

hypertension, diabetes, chronic obstructive pulmonary disease), who suffered from severe 8 respiratory failure and needed invasive mechanical ventilation (IMV) in almost 90% of the 9 cases.. Comorbidities, immune-suppression associated with SARS-CoV-2 infection 5,6 and 10 with the critical illness per se, use of immunomodulators (particularly steroids), and the 11 frequent need of invasive life support procedures, predispose COVID-19 patients to a high 12 risk of hospital-acquired infections (HAIs). To date, few studies have described HAIs during 13 COVID-19, mostly without a specific focus on ICU patients [7] [8] [9] [10] [11] [12] [13] [14] . 14 This retrospective, single nation, multicenter study aimed to evaluate the incidence, microbial 15 etiology, resistance patterns, risk factors, and impact on outcome of HAI in a large cohort of 16 COVID-19 patients admitted to ICU. 17 18

This is a retrospective analysis of prospectively collected data of all consecutive COVID-19 20 patients admitted to the ICUs of 8 Italian hub hospitals (see eMethods, Online Supplement) 21 from February 20, 2020 to May 20, 2020. Follow-up ended on July 23, 2020. 22

The participating centers shared the following management approaches for HAIs: 1) routine 23 antibiotic prophylaxis was not recommended; 2) stress ulcer and deep vein thrombosis 24 prophylaxes were provided; 3) ventilator-associated pneumonia (VAP) bundles 15 were 25 J o u r n a l P r e -p r o o f applied; 4) no selective digestive decontamination was employed. The participating centers 1 shared the same policy for microbiological surveillance: routine surveillance cultures for 2 bacterial and fungal infections (perineal and nasal swabs for multidrug-resistant (MDR) 3 bacteria, tracheal aspirate, and urine cultures) were obtained at ICU admission and then at 4 least once a week, while further microbiological examinations were performed in the 5 presence of a clinical/laboratory suspicion of infection. 6

The study was approved by the Ethical Committee of the promoting center (Comitato Etico 7

Milano Area 2; protocol: 0008489) and by the local Ethical Committees and pre-registered at 8 clinicaltrials.gov (NCT04388670). Written informed consent was waived due to the 9 retrospective nature of the analysis. 10

All consecutive patients with laboratory-confirmed SARS-CoV-2 infection (positive 11

Reverse-Transcription-Polymerase Chain Reaction) 16 admitted to the participating ICUs were 12 considered for inclusion. Exclusion criteria were: 1) age < 18 years old; 2) ICU length of stay 13 (LOS) < 24 hours; 3) nosocomial COVID-19; 4) bacterial co-infections at ICU admission; 5) 14 reason for ICU admission different from COVID-19. 15

The following patient data were collected at admission: demographics; weight, height, body 16 mass index (BMI); comorbidities stratified according to Charlson Comorbidity Index; 17 immunocompromised status (i.e., chronic immunosuppressive therapies, active hematological 18 or solid malignancies, autoimmune diseases); hypertension; diabetes; Sequential Organ 19

Failure Assessment -SOFA-score; PaO 2 /FiO 2 ; arterial blood gas analyses; ventilatory 20 settings (i.e., Positive End Expiratory Pressure; respiratory rate; Tidal Volume, Plateau 21 Pressure); blood exams (i.e., complete blood count, bilirubin, creatinine, lactate 22 dehydrogenase, C-reactive protein, D-dimers, ferritin, interleukin-6). We recorded the use of: 23 remdesivir, hydroxychloroquine, lopinavir/ritonavir, corticosteroids (low-dosage: < 2 24 mg/kg/die; high-dosage: > 2 mg/kg/die of methylprednisolone or equivalents), anakinra, 25 J o u r n a l P r e -p r o o f tocilizumab and assessed the use of RRT, extracorporeal membrane oxygenation (ECMO), 1 and pronation. Finally, we assessed the antibiotics administered before ICU admission and 2 then before and during each infectious episode. For the analysis, fluoroquinolones, ≥ III 3 generation cephalosporins, and carbapenem were defined as "broad-spectrum" antibiotics. 4

The following outcomes were recorded: survival at ICU and hospital discharge, ICU and 5 hospital length of stay (LOS), duration of IMV. 6

Infections were identified and recorded considering all microbiological isolates obtained 7 during the ICU course, independently reviewed and classified in light of the available 8 clinical, laboratory, and radiographic data by dedicated intensivists (one for each center) and 9

infectious disease specialists (one for each center), following international guidelines 17, 18 . 10

The timeframe for diagnosing HAIs was limited to the ICU stay, without follow-up after ICU 11 discharge. Infections were considered as ICU acquired infections whether they occurred ≥ 48 12 hours from ICU admission. Furthermore, a specialized intensivist (VS) and an infectious 13 disease specialist (AB) from the promoting center were available to support other 14 onset (i.e., occurring < 5 days from intubation), or late-onset (i.e., ≥ 5 days from intubation). 20 Viral infections were not included in the analysis. Finally, for every infectious episode, the 21 presence of sepsis or septic shock 21 was recorded. 22

We defined as multidrug-resistant (MDR) all microorganisms resistant to at least one agent in 23 three or more antimicrobial classes of agents 22 or the microorganisms with specific antibiotic 24 resistance mechanisms (e.g., methicillin-resistant Staphylococcus aureus -MRSA). Each 25 J o u r n a l P r e -p r o o f antibiotic susceptibility testing was analyzed, and resistance patterns for each antimicrobial 1 agent were classified. Of note, the presence of extended-spectrum beta-lactamases and 2 carbapenemases in surveillance specimens was directly tested either with phenotypic assays 3 (growth-based assays or immunochromatographic methods) or with molecular tests 4 depending on the clinical workflow of the microbiology laboratory. Conversely, the 5 antimicrobial resistance profile of bacterial strains isolated from other clinical specimens was 6 either tested (immunochromatographic or molecular tests) or deduced as per clinical practice. 7

Descriptive statistics were produced for demographic, clinical, and laboratory characteristics 9 of cases. Mean and standard deviation (or, in case of skewed distribution, median and 10 interquartile range (IQR)) are reported for continuous variables, and number and percentages 11 for categorical variables. 12

Groups were compared with parametric or nonparametric tests, according to data distribution, 13 for continuous variables, and with Pearson's Chi 2 test (Fisher exact test when appropriate) for 14 categorical variables. The crude incidence rate per 1000 patient-days (pd) of ICU (IR/ 1000 15 ICU pd) and relative 95% confidence intervals (CIs) was calculated. All the infectious 16 episodes, including multiple infectious episodes for each patient, were considered. The 17 analysis time scale was the time since ICU admission until the date of ICU discharge. 18

Time at risk of ICU HAIs was from ICU admission to HAIs, death, or discharge from ICU. 19

Risk factors for HAI were explored through multilevel Weibull regression models, with 20 random intercepts for hospital and patient, with drugs as time-dependent variables. 21

Univariable and multivariable models were fitted; variable selection strategy for 22 multivariable models was: clinically relevant variables, not colinear (i.e., rho <0.30), <10% 23 missing data, with no further selection. 24

Competing risk analysis was used to estimate the cumulative incidence of HAIs, with death 1 as a competing event; patients were censored at discharge from ICU; in this analysis, only the 2 first HAI was included. Fine and Gray competing risk regression models were used to assess 3 independent risk factors for HAIs; SubHazard Ratios (SHRs) and their corresponding 95% CI 4 are reported. Death was considered a competing event for developing an infection. 5

Univariable and multivariable models were fitted, with the same analysis strategy as above. 6

As exploratory subgroup analyses, the distributions of microorganisms identified in HAIs 7

were calculated by the type of infection; also, mortality risk in VAP, BSI, and other relevant 8 subgroups was assessed. Other secondary analyses included risk factors for HAI due to MDR 9 pathogens. 10

All tests were two-sided, and p<0.05 was chosen to indicate statistical significance. JMP 11 11 statistical (SAS, Cary, NC, USA) and Stata computer software version 16. J o u r n a l P r e -p r o o f Three hundred fifty-nine patients (47%) developed a total of 759 microbiologically-1 confirmed HAIs during the ICU course, corresponding to a median of 1 (0-3) episode per 2 patient (range 0 to 9) and an incidence rate of the first HAI of 44.7 infections/1000 ICU-pd. 3

The first HAI occurred after a median of 12 (8-18) days from hospital admission, 8 (5-12) 4 days from ICU admission, and 7 (5-12) days from intubation. Of note, none of the 82 non-5 intubated patients developed a HAI during their ICU stay. 6

The probability of developing an infection increased rapidly with the days since admission 7 after considering the competing event of death ( Figure 1) . 8

Univariable analysis comparing clinical characteristics of infected and non-infected patients 9 is presented in Table 1 . At the multivariable analysis, variables independently associated with 10 infection were age, PEEP, and broad-spectrum antibiotic treatment at admission (Table 2) . 11

Notably, the multilevel Weibull regression model showed that higher age was associated with 12 a lower risk of infection. This association was absent at Fine and Gray analysis, which 13 accounts for the competing risk of death (see eTable 2, Supplementary Material). 14 No association was observed between the use of immunomodulating drugs and infections. HAIs associated with septic shock (i.e., 54% of the overall septic shock events). Gram-24 J o u r n a l P r e -p r o o f negative and gram-positive agents caused septic shock in 23.6% and 18.0% of the HAIs, 1 respectively. 2

We had a complete follow-up of the patients until ICU discharge. 234/774 patients (30%) 3 died during the ICU stay. The ICU death rate was 15.3 deaths/1000 ICU-pd. Only 8 (9%) of 4 the non-intubated patients died during their ICU stay. On July 23, of the 540 patients 5 discharged alive from the ICU, 474 (87.7%) had been discharged from hospital, 25 (4.6%) 6 had died, and 41 (7.5%) were still hospitalized. Overall in-hospital mortality was 33% 7 (259/774). Infected patients had similar ICU mortality (31.5%) as compared to non-infected 8 patients (29.1%), p=0.483). Patients with at least one infection presenting with septic shock 9 (98/774, 13%) had higher ICU mortality (52%) as compared to non-infected patients 10 (415/774, 54%; mortality 29%), to patients with infection (168/774, 22%; mortality 21%) and 11

to patients with sepsis (93/774, 12%; mortality 28%), p<0.001. The same mortality risk (i.e., In this study, we analyzed the epidemiology, etiology, and impact on outcomes of ICU-19 acquired infections in a large cohort of critically ill COVID-19 patients. 20

The incidence of infectious complications was very high, with almost half of the patients 21 developing at least one infectious episode during their ICU stay. Specifically, 14 days after 22 ICU admission, the probability of having an infection was over 40%. 23

Previous reports documented a frequency of HAIs in COVID-19 patients ranging from 10% 24 to 45% 23-26 but provided limited information about microbial etiology and impact on 25 J o u r n a l P r e -p r o o f outcomes. Giacobbe et al. 8 , in a small cohort (n=78) of critically-ill patients, reported an 1 incidence of BSI of 47 episodes (95% CI 35-63) per 1000 patient-days, mainly due to 2 Staphylococcus aureus, Enterococcus spp, and Coagulase-negative Staphylococci. He at al. 7 , 3 in a single-center study on a mixed patient population (i.e., 33% with critical and 66% with 4 severe disease), described the microbial etiology of 65 HAIs and showed that patients with 5 HAI had a higher mortality rate. 6

Our work provides a detailed description of infectious complications in critically-ill COVID-7 19 patients. Infections occurred relatively early (about one week after intubation), and their 8 frequency increased with extended ICU stays. The most common infections were VAP due to 9

Enterobacterales, while Staphylococcus aureus was the most frequent Gram-positive 10 organism. Bloodstream infections accounted for one-quarter of all HAIs and were caused 11 almost equally by Gram-positive and Gram-negative bacteria. This rate of BSIs is 12 significantly higher than that reported in the largest study published so far in the general 13 population of ICU patients 27 and even than that observed in patients undergoing ECMO for 14 refractory respiratory failure 28 . Another key finding with potentially important clinical 15 implications is that MDR bacteria caused about one-third of the infectious episodes. 16

Although patients were treated in ICUs where MRSA prevalence was previously reported to 17 be low 29 , this organism accounted for more than 50% of all Staphylococcus aureus infections. 18 A high incidence of MRSA infections has been reported also in patients with severe H1N1 19 influenza, possibly favored by excessive mucus production, impaired mucociliary clearance, 20 and epithelial cell breakdown 30 . However, since our ICUs were overwhelmed by an 21 unexpected number of critically ill patients and personnel from different wards had to be 22 recruited to surge the ICU capacity, the high incidence of infections due to MDR germs may 23 be at least in part attributed to suboptimal adherence to the standard infection control policies. 24 J o u r n a l P r e -p r o o f In addition, infections by MDR germs may have been favored by the selective pressure of 1 antibiotic therapies. 2

Regarding the use of antimicrobial prophylaxis, the recent Surviving Sepsis Campaign 3

Guidelines suggest, with a weak level of evidence, the use of empiric antibacterial agents in 4 patients with COVID-19 and respiratory failure. As expected, a high rate of our patients (i.e., 5 68 %) were already receiving a broad-spectrum antibiotic treatment (usually with 6 cephalosporins and/or fluoroquinolones) before ICU admission. All participating ICUs' 7 policy was to interrupt empiric antimicrobial treatments if cultures at admission resulted 8 negative for bacterial coinfections. Notably, only 8 (1%) of our patients had a coinfection at 9 ICU admission, and a low rate of coinfections (8%) was also reported in a recent 10 metanalysis 31 . Even if we observed an apparently protective effect of broad-spectrum 11 antibiotics at ICU admission upon the occurrence of infection, our study was not designed to 12 specifically evaluate this aspect. Thus, this observation should be taken with caution, and,-13

given the high rate of infections due to MDR organisms-recommending routine 14 antimicrobial prophylaxis with broad-spectrum antibiotics does not seem justified in our 15

At the primary Weibull multilevel analysis, we observed older patients to have a lower risk of 17 death. This counterintuitive association is reasonably due to the immortal time bias; that is, 18 older patients have a higher risk of death and die prior to developing a HAI. Accordingly, the 19 secondary Fine and Gray competing risk analysis, which accounts for the competing risks of 20 death and infection, did not confirm such association. 21

Treatment with the interleukin-1 receptor antagonist anakinra, corticosteroids, or tocilizumab, 22

was not associated with an increased risk of secondary infections. However, this finding must 23 be interpreted with caution for the following reasons: a) treatment was not protocolized and 24 J o u r n a l P r e -p r o o f the indications differed among centers; b) anakinra and tocilizumab were administered on a 1 compassionate basis to the most severe patients, possibly introducing a selection bias; c) 2 there was a significant variability in dosing and timing, especially for corticosteroids. Since 3 the study design does not allow to rule out the effect of these confounders, our findings 4 should be considered as merely explorative and no definitive conclusions can be drawn about 5 the association of treatment with immunomodulators and the risk of secondary infections. 6

Infected and non-infected patients had similar mortality. We did not analyze the factors 7 associated with mortality, since a specific study design and causal inferencing to disentangle 8 multiple covariates would have been required. Nevertheless, HAIs complicated by septic 9

shock had almost doubled mortality and -overall-HAIs were associated with significantly 10 increased MV and ICU and hospital LOS duration. However, we cannot exclude a cause-11 effect relationship between the length of IMV or ICU stay and the risk of infection. 12

Our study has several limitations. First, since it is a retrospective analysis of data collected 13 primarily for clinical reasons in one of the regions most severely hit by the pandemics, not all 14 data were available for all patients. Second, there was no standard management approach 15 during the study period across different centers. In particular, the antibiotic strategy and 16 medical treatment were not uniform among the centers. Third, since we included in the 17 analysis exclusively the microbiologically-confirmed infections, we may have 18 underestimated the incidence of infectious episodes, neglecting some difficult-to-diagnose 19

infections (e.g., invasive aspergillosis) or infections characterized by low-yield cultures (e.g., 20

cultures obtained while the patients were receiving an antibiotic treatment). Fourth, a formal 21 "blind revision" of chest radiological imaging was not performed. However, an intensivist 22

and an infectious disease specialist from the promoting center were available to 23 independently adjudicate the diagnosis of HAI. Fifth, the study was conducted in a single 24

Western European country with a high incidence of MDR infection. Hence, it may be improper to generalize and extrapolate the study findings to the worldwide population of 1 COVID-19 patients. Also, sample size was dictated by contingencies and the study was not 2 designed nor powered to detect the effect of HAIs on mortality, thus all results should be 3 viewed as hypothesis-generating only. Finally, the study design does not allow a comparison 4 between the cohort of COVID-19 patients with that of patients with ARDS of different 5 etiology, and we cannot draw any conclusion about a causal association between COVID-19 6 and an increased risk of HAIs. 7

Interpretation 8

Critically ill COVID-19 patients are at high risk for HAIs, especially VAP and BSIs, 9

frequently caused by multidrug-resistant bacteria. Patients with HAIs complicated by shock 10 had almost doubled mortality, and infected patients had prolonged IMV and hospitalization. 11

Clinicians should make every effort to implement protocols for surveillance and prevention and HAIs complicated by septic-shock almost doubled mortality. 

Incidence of coinfections and 11 superinfections in hospitalized patients with COVID-19: a retrospective cohort study

Secondary infections in patients hospitalized with 15 COVID-19: incidence and predictive factors

Risk Factors and Outcomes of Hospitalized 19

COVID-19) and Secondary 20 Bloodstream Infections: A Multicenter Case-Control Study

Ventilator-associated pneumonia in adults: a 1 narrative review

Infectious Diseases Society of America

Guidelines on the Diagnosis of Coronavirus Disease

International ERS/ESICM/ESCMID/ALAT 8 guidelines for the management of hospital-acquired pneumonia and ventilator-9 associated pneumonia

IDSA Guidelines for the Diagnosis and Management of Intravascular 12

Catheter-Related Bloodstream Infection

Clinical practice guidelines for the 16 management of candidiasis: 2009 Update by the Infectious Diseases Society of

Microbiological laboratory testing in the 19 diagnosis of fungal infections in pulmonary and critical care practice: An official 20

American thoracic society clinical practice guideline

Surviving Sepsis Campaign: International 23

Guidelines for Management of Sepsis and Septic Shock