Ventilator-Associated Pneumonia in Critically Ill Patients With COVID-19

Mailis Maes; Ellen Higginson; Joana Pereira-Dias; Martin D. Curran; Surendra Parmar; Fahad Khokhar; Delphine Cuchet-Lourenco; Janine Lux; Sapna Sharma-Hajela; Benjamin Ravenhill; Islam Hamed; Laura Heales; Razeen Mahroof; Amelia Solderholm; Sally Forrest; Sushmita Sridhar; Nicholas M. Brown; Stephen Baker; Vilas Navapurkar; Gordon Dougan; Josfin Bartholdson Scott; Andrew Conway Morris


Crit Care. 2021;25(25) 

In This Article


Overall, we managed 94 patients with COVID-19, of whom 81 were ventilated for more than 48 h. From the period 15th March to 30th August we also managed 144 patients without COVID-19 in the liver/general unit who required ventilation for more than 48 h. The demographic and clinical features of these two groups are shown in Table 1 and details of non-COVID admission diagnoses in Additional file 1: Table S1. Ventilator bundle audit data demonstrated high compliance (compliance with the full bundle ranged from 85 to 100%, with 99–100% for the period April–May when most COVID-19 patients were admitted).

Patients with COVID-19 were significantly more likely to be investigated for VAP (Table 1), and had a higher incidence of microbiologically confirmed VAP (39 (48%) COVID-19 patients compared to 19 (13%) patients without COVID-19). Further details of the comparison of the investigation for VAP are shown in Additional file 1: Tables S2 and S3. Patients who were investigated for VAP demonstrated a significant deterioration in oxygenation relative to the period immediately prior to the diagnosis (Additional file 1: Figure S1).

Survival analysis (Figure 2) demonstrated that the increased risk of developing VAP in patients with COVID-19 was not simply a function of longer duration of ventilation. The hazard of early VAP was similar in both groups of patients, however the greater number of later VAPs in COVID-19 led to the increased median duration of ventilation before VAP developed seen in Additional file 1: Table S2. The effect of COVID status on VAP-free survival remained significant when adjusted for age and immunocompromised status (adjusted p value 0.045 by Cox proportional hazards model, Additional file 1: Table S5). Sensitivity analysis of patients with > 72 h mechanical ventilation and > 144 h of mechanical ventilation produced similar survival curves and hazard ratios (Additional file 1: Figure S2A and B). A similar finding was apparent when comparing crude incident density, patients with COVID-19 developed VAP at a rate of 28/1000 ventilator days, whilst those without COVID-19 experienced VAP at a rate of 13/1000 ventilator days (p = 0.009 by mid-P exact test). Incident density censoring for post-VAP duration ventilation, which is confounded by VAP itself prolonging ventilation, shows a similar pattern (40/1000 ventilator days for COVID-19, 19/1000 ventilator days for non-COVID p = 0.004 by mid-P exact test). Further details on timing of VAP are available in the supplemental section (Additional file 1: Table S2). Antibiotic use on admission (Table 1) and in the period leading up to investigation for suspected VAP (Additional file 1: Tables S3 and S4) was similar in frequency and spectrum of agents used.

Figure 2.

Time to development of VAP in patients with and without COVID-19 censored for death or extubation. P value and hazard ratio by Cox proportional hazards. Numbers at risk at each time point indicated below x-axis

The organisms identified on endotracheal aspirate culture and both culture and molecular testing of bronchoalveolar lavage fluid are show in Table 2. The concordance between culture and molecular testing was high, although molecular testing identified a number of additional organisms.

The distribution of organisms in COVID-19 and non-COVID-19 associated VAP is shown in Figure 3, and is broadly similar between both groups.

Figure 3.

Causative organisms of VAP in patients with and without COVID-19. Non-pathogenic organisms detected above threshold levels shown in grey

Lung Microbiota

To investigate changes in the lung microbiota in the COVID-19 positive and negative patients we performed 16S rRNA sequencing on a subset of BAL samples from 24 patients. In general, bacteria detected by TAC or conventional microbiology were abundantly identified in samples by 16S sequencing (Figure 4). Samples with confirmed VAP or colonization with low pathogenic organisms generally yielded higher overall read numbers. When comparing COVID-19 positive to COVID-19 negative patients, there was no specific taxon that was more prevalent in either group. Additionally, in this relatively small subset of samples, the bacterial composition of BALs from COVID-19 positive patients were not significantly different in either the species richness (alpha diversity) or the microbial composition (beta diversity).

Figure 4.

Microbial composition of BAL samples from SARS-CoV-2 positive and negative patients. Bacterial 16S genes were sequenced and classified to the genus level using Kraken2. The number and percent of reads mapping to each genus is shown for individual samples from each patient (A), with kit controls in the first two columns, and longitudinal samples (1, 2 or 3) from individual patients (B). Individuals were classified as either COVID-19 negative, COVID-19 positive, or recovering (previously diagnosed with COVID-19 but SARS-CoV-2 negative at time of sample)

To investigate changes in the microbiota over the course of infection, we next looked at the microbial composition of BAL samples in some individual patients over time. Two patients diagnosed with VAP (patients 1 and 24) showed decreasing species richness over time, as the bacterial pathogen implicated in the illness became the predominant microbe present. For patient 6, the microbial composition shifted significantly over time, as Enterococcus took over from Staphylococcus as the most predominant organism. The microbiome composition of patient 24, who was both VAP and COVID-19 negative, was largely stable over time. In general, the microbial composition of BAL samples from patients who did not have VAP at the time of sampling (sample 1 from patient 14 and both samples from patient 24) were more diverse than samples from patients who had been diagnosed with VAP.

Invasive Aspergillosis

43 patients were investigated for possible pulmonary aspergillosis by PCR and lavage galactomannan, on the basis of senior clinician suspicion of fungal infection. 23 patients with COVID-19 and 20 without. Of these 3 COVID-19 patients met the criteria for IPA outlined in the methods above (one positive by PCR with borderline galactomannan 0.7 optical density index (ODI), and 2 PCR negative but with galactomannan > 1.0 ODI), and all were treated with liposomal amphotericin, 2 of these patients survived to hospital discharge whilst one died. One patient without COVID-19 had a borderline positive galactomannan (0.8 ODI), and met clinical criteria but was not treated as care was withdrawn for other reasons. We estimate the prevalence of COVID-19-associated aspergillosis (CAPA) to be 13%, although with small numbers the confidence intervals are wide (95% CI 5–32%). None of the three patients with CAPA had received steroids prior to the diagnosis.

Reactivation of Herpesvirade

49 patients had lavage tested for herpesvirade, 24 with COVID-19 and 25 without. Although five patients (two with VAP from other organisms, and three without VAP) had detection of herpes simplex virus (HSV) below the Ct cut-off of 32, in viral reactivation the role of viral load is uncertain. We therefore examined the frequency of herpesvirade detection at any level in lavage of patients investigated for suspected VAP. In total 10 patients with COVID-19 had detection of herpesvirade (4 HSV, 5 Epstein barr virus (EBV) and 1 patient with both), whilst 5 patients without COVID-19 had detection (2 HSV, 1 cytomegalovirus, 1 EBV and 1 patient with both HSV and EBV). As only lavage was tested for herpesvirade, the prevalence of herpesvirade detection amongst the tested population was 42% (95% CI 24–61%) in patients with COVID-19 and 20% (95% CI 9–39%) in patients without COVID-19 (distribution of Ct values for herpesvirade are shown in Additional file 1: Figure S3). Only one patient with herpesvirade activation had received steroids prior to detection.