In this cohort study, we found that a high rate of patients with SARS-CoV-2 developed a respiratory coinfection with another microorganism during their ICU stay. The vast majority had late coinfections, and a high proportion had more than one coinfection. Comparing between groups, previous history of corticosteroids, the use of dexamethasone during the hospitalization and days in MV were all independently associated with a significant higher risk of coinfection.
In our cohort, 40.6 % of our patients developed at least one respiratory coinfection during their ICU stay, which is higher than what has been previously reported [9, 18, 19] but similar to the proportion of coinfections reported in post influenza ICU patients [20, 21]. The proportion of early vs late coinfection shows a similar trend to other studies, where late coinfections were higher than early ones [5, 6, 13, 22]. This is to be expected since patients with COVID-19 usually have long hospital stays, making them more susceptible to hospital-acquired infections. Microbiologic analyses in our ICU patients were made only if there was a coinfection suspicion by the attending physician, so we cannot exclude that there could be an under-diagnosis of early infections. Other concerning finding was the high rate of cumulative coinfections, probably associated with longer length of stay. Our bacteria resistance rate is higher than the findings of García- Vidal [5], but lower than the one reported by Li et al (ref 23), which was over 50%. These findings reaffirm the fact that an adequate use of antibiotics is urgently needed.
The most common bacterial agent in late coinfections in our cohort was Pseudomonas aeruginosa. Earlier findings describe the same agent as predominant [6, 7], however other publications show enterobacteriae or Staphylococcus aureus as the most common respiratory coinfection [24, 25]. The latter reflects differences in local microbiology and the importance of knowing this information to make appropriate antimicrobial empirical treatment in every institution. Cases of fungal infections were found in apparently immunocompetent patients. Nevertheless, other risk factors for fungal coinfection such as broad-spectrum antibiotic, corticosteroids use or prolonged ICU length of stay in patients with COVID-19 can explain these coinfections [26]. Invasive fungal infections have been reported in ICU patients with COVID-19 from different countries [6, 7, 27, 28], similar than what was previously found in influenza [29]. We did not find coinfections with other community viruses, even considering that during the recruitment of patients we were on our winter season. In contrast to preceding years, during the first wave of covid in our country, classical viruses were displaced by SARS-CoV-2. The only viral coinfections we found were CMV coinfections, which can be explained because of COVID-19 immunosuppression.
In our center, tocilizumab or ruxolitinib were indicated to patients with high suspicion of having a hyperinflammatory syndrome associated to COVID-19 pneumonia. We did not find an association between these drugs and higher co-infections rates, results that are coherent with other recent studies [5, 7, 24]. Conversely, patients with chronic use of corticosteroids or patients exposed to dexamethasone in the hospital because of COVID-19, developed significantly more coinfections than patients that were not exposed. There are conflicting results on this association; some studies reported a direct one [5, 24, 30] while others did not [7]. Our study was done before the findings of RECOVERY trial [31] so we could compare patients with and without steroids. Systemic corticosteroids for treatment of COVID-19 patients are now broadly used based on studies that showed their efficacy on preventing invasive mechanical ventilation and reducing 28- day mortality for COVID-19 severe cases [31, 32]. However, their impact on immunity, the risk of associated infections and their ultimate effect on outcome are still unclear, furthermore when considering steroids potential side-effects and contraindications [32].
The fact that all coinfected and non-coinfected patients in our cohort used antibiotics could be explained because of institutional guidelines for their prescription for all patients with severe pneumonia. The high frequency of antibiotic exposure previous to a coinfection development has been previously reported [24]. However, since a minority of our patients admitted to the ICU presented an early coinfection, our findings, as well as other studies [5, 22, 33, 34], support a more restrictive use of antibiotic therapy on admission and must be limited to suspected or confirmed bacterial infections.
Most published literature has shown that in COVID-19, the most common chest radiography and CT findings are bilateral ground glass opacities, with or without consolidation [35]. Making diagnosis of bacterial coinfections in COVID-19 with radiological images is challenging because a consolidation on image is not pathognomonic of bacterial infection. In our study we found that only half of the patients with proven bacterial coinfection had a consolidation on X-ray or CT. Fungal infections are difficult to diagnose too, because of the presence of extensive viral pneumonia infiltrates [26]. None of our patients with fungal infection presented a nodular pattern.
Despite the high coinfection rate, ICU mortality in our cohort was similar to other ICU reports [18, 19]. Higher rates of survival than our study in patients with coinfections have been reported, however these studies included non-ICU patients [5, 7]. We did not find a correlation between coinfections and mortality; however, we saw an association with a greater need for tracheostomy, vasoactive drugs, longer ICU and hospital stay as well as prolonged ventilation time in coinfected patients, compared to non-coinfected patients. The nature of this study does not allow us to establish if these findings are a cause or a consequence of coinfection. Some studies suggest higher risk of mortality in coinfected patients [5, 9, 22, 36], findings that are contradicted by other publications [6, 11]. More directed studies are needed to answer this question, which is crucial to determine the real clinical relevance of coinfections in the COVID-19 pandemic.
Our study has some limitations. First, it was conducted at a single center, thus our microbiological results cannot be homologated to other institutions. Additionally, interpretation of our findings might be limited by the sample size. Third, the nature of the study does not allow us to determine causality between coinfection and worst outcomes such as longer hospital or ICU stay. Despite these limitations, our study has important strengths, such as an extensive microbiological study, antimicrobial resistance and the comparison with a group of patients with no-coinfection.