Introduction
Hospital-acquired pneumonia (HAP) is a frequent event during an intensive care unit (ICU) stay and is characterized by a pneumonia acquired during hospitalization, in patients without invasive mechanical ventilation [
1‐
3]. Despite improved prevention measures, antimicrobial therapy, and supportive care, it remains an important cause of morbidity and mortality [
1‐
3]. HAP is the leading cause of death among hospital-acquired infections, with estimates of its associated mortality ranging from 20 to 50% [
2,
4‐
6].
Microbiological diagnosis is fundamental for guiding HAP treatment, allowing a targeted, effective antibiotic choice, and reducing the associated impact of ineffective empiric antibiotic regimens or the unnecessary use of broad-spectrum antibiotics [
1]. Yet the current understanding of HAP pathogens is based mainly on data derived from ventilator-associated pneumonia (VAP) [
7‐
15]. Although some studies have reported the pathogens in HAP that occur outside the ICU [
16‐
18], there is no systematic description of the diagnostic approaches that should be used to efficiently obtain an microbiological diagnosis in HAP, mainly when critically ill [
1].
In comparison with VAP, HAP poses additional challenges for respiratory sampling and, thus, for obtaining microbiological diagnosis. The utility of sputum cultures or distal sampling for HAP has not been comprehensively evaluated [
1]. The recent guidelines for HAP/VAP recognized that for some patients, whom non-invasively sampling is not possible, invasive sampling is an option [
1,
3]; however, the literature is scarce to support one method over the other in HAP. In this study, we aimed to describe the diagnostic approaches used in a cohort of HAP acquired during an ICU stay and their associated clinical impact.
Discussion
We could achieve microbiological diagnosis in 50% of 200 patients with HAP occurring during ICU stay using an intensive diagnostic approach. Upon HAP clinical diagnosis, around 40% of patients underwent fiberoptic-bronchoscopy while not receiving iMV. Finally, invasive respiratory sampling was associated with a higher rate of microbiological diagnosis.
Recent recommendations from the FDA recognized that there are three different types of nosocomial pneumonia with different all-cause mortality rates: non-ventilated HAP, ventilated HAP, and VAP [
28,
29]. Interestingly, the highest mortality has been observed in patients with HAP who subsequently required iMV. In a recent summary of these recommendations, Talbot highlighted the necessity to have information about sampling and causative pathogens in the non-VAP population [
28]. Our study is the first one to provide this information in a detailed way, which can be very useful for empirical treatment adequacy and for future RCT studying new antibiotics.
Being able to achieve a microbiological diagnosis in HAP has important consequences for patient care. First, it can support the suspicion of infection in a new lung infiltrate appearing concomitantly with fever in a critically ill patient, a frequent challenge for the attending physician [
30]. Second, it makes possible to target the empiric antibiotic scheme more accurately, thus increasing the likelihood of clinical cure, preventing the selection pressure to further resistances, and reducing costs and unnecessary side effects [
1]. Our findings corroborated two important phenomena reported elsewhere: (1) patients with microbiological diagnosis more commonly had an adaptation in their empiric antibiotic regimen and (2) patients without microbiological diagnosis received shorter total antibiotic treatment [
31,
32]. Although a microbiological diagnosis is central in all infections both for epidemiological studies and for bedside care by clinicians, it becomes fundamental for hospital-acquired infections, because of higher probability of resistant pathogens, greater amount of antibiotic use and side effects, and associated costs.
Interestingly, one third of patients underwent sputum collection, which was positive in 34% of cases after ensuring sample quality and performing quantitative cultures. Very few data are available on the applicability of sputum in HAP [
1,
16,
17]. In our experience, this non-invasive diagnostic method should be encouraged, as it already is for community-acquired pneumonia [
33]. Indeed, when only one diagnostic method was positive, 15% of microbiologic confirmations were due to sputum, and in patients who were not subsequently intubated, this proportion was even higher (33%). Despite the limited numbers of patients allowing for pair-wise comparisons between methods, we observed a good agreement on retrieving the same pathogen (80% on average). As expected it was higher for invasive methods (FBAS vs. BAL, 86% of agreement). In our protocol, we tried to obtain as much as possible respiratory samples to increase the likelihood of identifying a causative pathogen, and the good agreement observed is reassuring. When two methods were discordant, respecting the sample quality check and cutoff values, clinicians interpreted the episode as polymicrobial and treated both pathogens, which is sound in critically ill patients. Taking different respiratory samples also increases the risk of false positives (i.e., colonization). We could not evaluate the actual impact that discordance between methods would have in clinician’s decision in a scenario where there would be a hierarchy between methods, for instance.
In this observational study, patients assessed with an invasive diagnostic method had higher rates of microbiological diagnosis. Although there is evidence that invasive and non-invasive approaches have a comparable impact on patient-centered outcomes in VAP [
1], no evidence is available for HAP in immunocompetent patients [
1]. In fact, the 2016 IDSA/ATS guidelines propose non-invasive respiratory sampling in HAP, although the panel agreed that there may be factors that prompt clinicians to consider invasive sampling [
1]. In a small single-center randomized trial aiming to compare invasive and non-invasive approaches in patients with HAP outside the ICU, Herer et al. found that clinical cure rates at 28 days were similar between groups; however, the study was rather exploratory in nature, with several limitations and a small sample size [
18].
Because of the barriers to obtaining lower respiratory tract samples in HAP, we cannot straightforwardly extrapolate the evidence from VAP to HAP. Indeed, an invasive approach might have higher clinical utility in HAP, particularly in those patients who will not require iMV. A key point when discussing invasive vs. non-invasive tactics in HAP is the feasibility and safety of performing a fiberoptic-bronchoscopy. Several reports show that fiberoptic-bronchoscopy, followed by BAL or mini-BAL, can be conducted in patients with acute respiratory failure and community- and healthcare-acquired pneumonia and is even safer when non-invasive ventilation and high-flow oxygen therapy are applied [
34‐
39]. In a landmark trial, Azoulay et al. showed that an invasive approach had a similar rate of intubation to a non-invasive approach in non-ventilated, immunosuppressed patients with acute respiratory failure [
40].
Invasive mechanical ventilation after HAP diagnosis was commonly needed in our population of critically ill patients, being applied 60% of the time within 24 h. Despite its clear implications for prognosis, having an endotracheal tube vastly facilitates access to a lower respiratory tract sample using either invasive or non-invasive approaches. The ability to predict which patients will need iMV in the next hours can help guide clinicians faced with the decision of performing a prompt fiberoptic-bronchoscopy or postponing it until after the intubation. The development of a prediction tool is beyond the scope of this study, but we observed that severity, hypoxemia, and chest X-ray patterns were associated with intubation after performing a fiberoptic-bronchoscopy.
Our study has some strengths. We included prospective cases of HAP acquired during an ICU stay from six ICUs. Our center also has a comprehensive clinical decision-making protocol for achieving microbiological diagnosis in lung infections, which means that our data are relevant for the description of microbiological diagnosis in HAP. Moreover, the causative pathogens responsible for HAP in our cohort are similar to those reported elsewhere, where Gram-negative bacteria have been implicated in 55% to 85% of HAP cases and Gram-positive cocci (particularly
Staphylococcus aureus) account for 20% to 30% [
7,
9‐
11,
13,
14,
41], thus increasing the generalizability of our results. In addition, the results of this study cover an unmet need of knowledge (microbiological diagnosis of HAP) highlighted by the recent IDSA ATS and International guidelines for HAP and VAP [
1,
3].
However, there are several limitations that must be highlighted. First, our study is retrospective and single-center and, although we collected data from six ICUs with different profiles (from general medical to respiratory and liver units), the single-center characteristic decreases the generalizability of our findings. Second, our study is observational and allowed us for a reliable description of real-life diagnostic methods approaches for achieving microbiological diagnosis in HAP, our primary objective. However, the crude associations found for the potential benefit of invasive methods are exploratory and not causal; a well-designed, controlled randomized trial is now warranted to define the management of HAP regarding the use of invasive or non-invasive methods. Third, we could recruit 200 patients, which limited our ability to explore subgroups and pair-wise comparisons between different methods, but to the best of our knowledge, this is one of the first and largest studies reporting all these different diagnostic methods in critically ill nonventilated HAP [
42]. Third, we could not achieve 100% of respiratory samples in the cohort; however, we believe that 93% represents a very high proportion of patients, considering the daily care in an ICU. Fourth, our population comprised critically ill patients, who commonly require iMV, and our results may not be applicable to patients outside the ICU. Fifth, we did not have a “gold standard” to confirm that the pathogen identified was responsible for the infection and not only colonizing the airways, a potential limitation particularly for sputum cultures. To limit the number of false positives, we used the most standard quality assessment to accept only lower airway representative samples. Sixth, at the time the current study was conducted, our center did not have routine access to rapid diagnostic methods because they were not standard of care, but these methods have been shown to be promising tools for pathogen identification in HAP [
43]. The performance of rapid diagnostic methods in nonventilated HAP, utilizing different sampling strategies, must be evaluated and could produce different results compared to our findings. Particularly, rapid diagnostic methods could increase the sensitivity for pathogen identification in those patients already receiving a new antibiotic upon sample collection, a fact that might explain the reason we achieved only 50% of microbiological diagnosis using traditional culture methods [
44]. Finally, we did not conduct a cost-effective analysis [
1,
18], which is a key element when comparing different respiratory sampling methods.