Non-ventilator-associated ICU-acquired pneumonia (NV-ICU-AP) in patients with acute exacerbation of COPD: From the French OUTCOMEREA cohort

All adults admitted to one of the 32 ICUs participating in the prospective national OutcomeRea™ database between 1^st January, 1997 and 31^st December, 2018, with a diagnosis of COPD exacerbation or acute respiratory failure with a medical history of COPD were included in the analysis. The exclusion criteria are summarised in Fig. 1.

The French Advisory Committee on Data Processing in Health Research, the French Commission on Informatics and Liberty, and the Ethics Committee of the University of Clermont-Ferrand, France, (IRB No. 5891) approved this data collection (ref. 2007–16). Patients were informed of the inclusion of their de-identified data in the database and could oppose this if they so wished.

OutcomeRea™ is an ongoing, prospective, observational, collaborative, and multicentre database. All codes and definitions used to describe diseases, comorbidities, and outcomes were established before the study began and have been described previously [25]. Clinical and outcome data, treatments, and prescribed medications are prospectively entered into the database daily for a random sample of patients admitted to 32 French ICUs. Of these 32 participating ICUs (including 18 university hospitals), 16 were polyvalent or surgical ICUs and 16 were primarily medical ICUs.

Patient characteristics, prescriptions, and use and type of ventilatory support were extracted from the database. The diagnosis of “very severe COPD” was defined as previous long-term treatment with home oxygen therapy, home NIV, or documented forced expiratory volume in 1 s < 30% predicted value (GOLD classification, stage 4) [1]. Patients who did not meet these criteria, with limited follow-up data, without recent spirometry or without available data were classified as having “not very severe COPD or unknown COPD severity.”

The risk of NV-ICU-AP was considered from the end of the first 48 h in ICU (without IMV) until ICU discharge, death, or need for at least 48 h of IMV later during their ICU stay. The period considered as risk of VAP ranged from 48 h after intubation to weaning from the invasive ventilation. Pulmonary infection was suspected in patients with a new or persistent infiltrate on the chest radiograph that was associated with any of the following criteria: (1) purulent tracheal secretions, (2) fever ≥ 38.5 °C or hypothermia ≤ 36.5 °C, and (3) leucocytosis > 10⁹ G/L or leucopenia < 4.10⁸ G/L. The diagnosis of NV-ICU-AP was confirmed by bacteriological tests or by a senior investigator in cases of strong clinical suspicion and impossibility of sampling /and or bacteriological tests. The diagnosis of VAP was confirmed by bacteriological tests.

The bacteriological samples considered were a sputum culture (threshold, > 10⁵ cfu/mL of a good-quality sample), bronchoalveolar lavage fluid (threshold, ≥ 10⁴ cfu/mL), plugged telescopic catheter (threshold, ≥ 10³ cfu/mL), or a quantitative endotracheal aspirate (threshold, ≥ 10⁵ cfu/mL), received within the first 24 h after suspicion of NV-ICU-AP or VAP. When bacteriological examinations yielded only coagulase-negative staphylococci or Enterococcus species, a nosocomial pneumonia diagnosis was confirmed only after checking the patient’s data by a senior investigator.

Appropriate antibiotic therapy (AT) required a positive culture result of bacteriological tests, received within the first 24 h after suspicion of NV-ICU-AP. The time to administration of appropriate AT was the number of days before the patient received appropriate AT (in case of inadequate AT or delay to AT prescription) after suspicion of NV-ICU-AP. The presence of multidrug-resistant (MDR) bacteria required documented evidence of at least one of the four classes of MDR: methicillin-resistant Staphylococcus aureus, extended-spectrum β-lactamase–producing Enterobacteriaceae, AmpC-producing Enterobacteriaceae, and Pseudomonas aeruginosa resistant to ticarcillin and/or imipenem and/or ceftazidime in the bacteriological samples.

Categorical parameters are expressed as numbers and percentages, and continuous parameters are given as median and interquartile range [IQR]. For missing data concerning body mass index (BMI), multiple imputations were performed using a fully conditional specification method with linear regression and twenty imputed datasets were created to consider this variable with less than 20% of missing values [26]. Factors associated with the occurrence of NV-ICU-AP were identified using the Fine and Gray competing risk model [27], with 2 days of IMV, discharge from the ICU, and death as competing risks. An adjusted analysis was performed including variables collected at ICU admission, with p < 0.2 in the univariate analysis and considered potentially collinear. The association between NV-ICU-AP and survival on Day-28 was assessed by a Cox model with NV-ICU-AP as a time-dependent variable. To assess the association between NV-ICU-AP and the length of stay in the ICU (using the daily instantaneous risk of alive discharge from the ICU), and association between NV-ICU-AP and intubation, we used a cause-specific regression model controlling for the competing risks of death or of discharge from the ICU with NV-ICU-AP as a time-dependent variable. The association between inadequate antibiotic treatment, time to appropriate antibiotic treatment, and documented presence of MDR bacteria on mortality and intubation, were assessed using the same models. For all the multivariate regression models, variables with p < 0.2 in the univariate models were introduced using a stepwise selection method. In cases of multiple episodes of NV-ICU-AP, only the first episode was considered. All models were stratified by centre, and a 2-sided alpha threshold of 0.05 was considered statistically significant. Statistical analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC, USA).

Among the 23,249 patients registered in the OutcomeRea™ database between 1^st January 1997 and 31st December 2018, 1,816 patients had a diagnosis of acute exacerbation of COPD or acute respiratory failure with a medical history of COPD. Of these 844 were not initially given IMV and were thus were exposed to the risk of NV-ICU-AP. Fifty episodes of NV-ICU-APs occurred in 42 patients (six patients presented two episodes of NV-ICU-AP, and one patient had 3 episodes) with an incidence density of 10.8 per 1,000 patients-days exposed to the risk of NV-ICU-AP (Figs. 1 and 2). The first episode of NV-ICU-AP occurred after a median delay of 6 [4; 11] days after ICU admission, and 32 patients with NV-ICU-AP (76.2%) were intubated within 48 h after diagnosis of NV-ICU-AP (Table 1). The baseline characteristics and outcomes of patients are presented in Table 2. The main microorganisms found were S. aureus, P. aeruginosa, and Escherichia Coli. Six (14.29%) NV-ICU-AP cases were diagnosed without bacteriological documentation (Table 1). MDR bacteria were found in 13 of 36 patients (36%) with a first episode of NV-ICU-AP (with bacteriological documentation) and mainly concerned S. aureus and P. aeruginosa (Table 1). 219 patients were intubated and 59 episodes of VAP occurred in 39 patients with an incidence density of 23.1 per 1,000 patients-days at risk of VAP (Fig. 2). The first episode of VAP occurred after a median delay of 11 [8; 14] days after ICU admission and after 9 [5; 11] days of invasive mechanical ventilation. Baseline characteristics and outcomes of patients according to the occurrence of VAP are shown in Additional file 1: Tables S1 and S2.