Community-acquired pneumonia (CAP) remains a major cause of morbidity and mortality cause in Western countries, despite significant advances in antimicrobial and supportive therapy.1 In Portugal, disease burden has been increasing in the last decade, especially among the elderly. According to the latest data, pneumonia is the main cause of respiratory hospitalization, accounting for 3.7% of all hospital admissions in Portugal between 2000 and 2009, with an intra-hospital mortality rate of 20.4%.2

Early introduction of appropriate therapy figures as one of the major prognostic modifying factors.3 The choice of initial antimicrobial treatment is usually empirical and it is essential that it is broad enough to cover all the likely pathogens. Despite international guidelines recommending specific empirical regimens, local differences in etiological pathogens have been found.4 More recently, the emergence of antibiotic-resistant agents has created new concerns for a more judicious use of broad-spectrum antibiotics, which poses important challenges in initiating early therapy.5 For this reason, and in order to propose the most appropriate treatment in a specific context, it is crucial to establish microbiological epidemiology data within different healthcare settings.

Respiratory bacteria are the major causative microorganisms, Streptococcus pneumoniae being the most common etiological CAP agent.4 Other frequently identified pathogens are Haemophilus influenzae, Gram-negative enteric bacilli and respiratory viruses.6 However, certain clinical features of patients have been associated with the risk of infection by multidrug-resistant (MDR) pathogens.7 Some of these cases have been designated as healthcare-associated pneumonia (HCAP), which includes patients living in nursing homes, who are immunosuppressed, had recent hospitalization, undergo chronic dialysis or home-based infusion therapy. This differentiation was supported by the contrasting observed outcomes, with higher mortality rates among those with pneumonia acquired in the hospital or associated with healthcare environments.8 However, this is a controversial term, given that a few European studies have shown etiology patterns in HCAP patients similar to those found in CAP patients,9 which contrasts with data from the USA.9 Moreover, etiological agents may have a seasonal distribution.10 These variables must be taken into account when approaching the patient diagnosed with pneumonia.

The primary aim of this study was to describe the etiology of CAP in adults admitted to a northern Portugal University Hospital. Secondary aims are the evaluation of the yield of the diagnostic methods applied as well as the analysis of the impact of patients’ demographic and clinical features on CAP microbiological etiology.

This study provides a comprehensive insight into the etiological and clinical profile of CAP hospitalized patients in Portugal and describes the real-life microbiological testing in this setting. Proper diagnostic testing improves clinical outcomes directly through individualization of antibiotic management, and indirectly by delivering relevant epidemiological data that influences initial empirical therapy. Although a fundamental exercise, establishing microbiological CAP diagnosis is challenging.4 In a prospective study, when conventional methods (i.e., bacterial cultures, urinary antigen assays, serology) were combined with PCR-based methods, definitive or probable etiology was established in 63% of cases.13 In our analysis, with complete standard microbiological testing (sputum, blood culture and urine antigen test), we attained a microbial confirmation in 31.5% of cases. While it represents a lower diagnostic rate than the described in prospective studies,14–16 it is comparable to previous large real-life studies.17,18 Due to the retrospective nature of this work, harvesting of all products was not possible in some cases, and 14% of the patients in the cohort did not undergo any microbiology test. We may speculate that clinicians are less prone to undergo invasive microbiological assessment in patients with the poorest performance status. Indeed, patients without microbiological studies were significantly older.

Additionally, our work confirms the low sensitivity of blood cultures described in other reports18,19 with only 7.4% positive results. The low yield of blood cultures could be possibly explained by delay in sample collection and/or prior use of antibiotics, but these findings must also depend on bacterial infection biology, as there was a higher rate of bacteremia among pneumococcal disease. Microbial yield of sputum culture was also very modest. The poor quality of sputum accounted for several missing diagnoses (data not shown), further aggravated by ineffective cough or inappropriate sputum production, particularly at the older population fringe. Finally, once serologic assays are not routinely performed in our center, Mycoplasma and Chlamydophila pneumoniae, and supposedly several other cases of viral infections, could not be detected, which also affected importantly the overall diagnosis rate. Many studies conducted prospectively, performing serum IgM antibody measurement, combined or not with PCR detection, described atypical bacteria as major causes of CAP, with variable frequencies up to 18%, and the viruses involved in up to 30% of hospitalized CAP, only surpassed by S. pneumoniae as the most frequent agents.20 In the present work, viral etiology was probably underestimated, having been established by PCR of samples collected by pharyngeal swabs in selected patients, based on clinical judgement. Nevertheless, it is acknowledged that serology testing has little impact on the routine management of the individual patient.20,21 Bearing this in mind and considering the high healthcare costs associated, there is a general notion that serologies are more useful in epidemiological studies than in clinical practice.

Considering the low microbial detection yield of various works, some authors have claimed that current empirical antimicrobial recommendations are based on weak evidence.18 Knowing the importance of the early implementation of appropriate therapy, we derived etiological predictors from CAP patients’ features, to better suit therapy to the most probable causative agent. Our data suggest that, in older people, S. pneumoniae, Influenza and L. pneumophila are less common, and a higher incidence of CAP is observed caused by GNB, S. aureus and M. catarrhalis, pathogens that more frequently present drug resistances. This is important, since CAP is a growing problem among the elderly.1,22

Globally, S. pneumoniae remains the most common detected cause of CAP4 and our results confirm these data, with 45.7% of the identified cases. However, there was a large contribution of the urine antigen test for pneumococcal diagnoses, when a similar method is not routinely available to other microbial agents. In fact, if we consider etiological diagnosis based on respiratory samples alone, we find H. influenza as the most commonly isolated organism. It is known that H. influenzae is a common CAP pathogen in older patients and those with respiratory comorbidities, such as COPD, as is the case in our sample population. These patients usually present chronic bronchitis and are frequent sputum producers, which facilitates the noninvasive collection of appropriate sputum samples improving the diagnostic yield these cultures.

Both S. pneumoniae and H. influenzae may exist as commensal organisms of the upper respiratory tract, so quantitative multiplex nucleic acid amplification test (NAAT) would detect and differentiate the etiologic agent of CAP better. A study from the United Kingdom identified an etiologic agent by quantitative PCR in 87% of CAP patients, including S. pneumoniae in 36% and H. influenzae in 40%.23

Future studies are needed to unravel how demographic changes in European countries will impact current microbial epidemiology. Furthermore, it is expected that the incidence of CAP by S. pneumoniae will decrease due to the introduction of pneumococcal vaccines.24 Thus, upcoming empirical antimicrobial treatment recommendations may have to take into special account the rise of beta-lactamase-producing agents in this setting, including H. influenza, M. catarrhalis and GNB. We report a significant proportion of CAP cases due to GNB (∼10%), with Klebsiella pneumoniae and P. aeruginosa comprising nearly half of these cases, which is only paralleled in a few other reports.25,26 The highest incidence occurred in patients suffering from dementia and were associated with increased mortality of up to 30%. When controlling for age, dementia and other disabling comorbidities, P. aeruginosa, but not GNB as a whole, was an independent risk factor for mortality. These findings are identical to Spanish data,25 where the authors reported 11% of GNB infections, associated with 32% mortality, much higher than observed with the non-GNB group (9%). In that publication, however, P. aeruginosa failed to be an independent predictor of death.25 Similar effect on mortality related to GNB was found in Asian countries27 and South America.28 Unfortunately, there are no specific measures to prevent pneumonia caused by GNB in the community setting. Despite this, since dementia was an independent risk for GNB infections in our study, measures promoting good oral hygiene and minimizing the use of proton-pump inhibitors can help to reduce bacterial microaspiration among these patients. Additionally, hand feeding should be tried before considering tube feeding and a semi-recumbent position with the head of the bed at a 30–45° angle should be adopted in bedridden patients. Another well-recognized risk factor for GNB infection or other drug-resistant bacteria is previous antibiotic treatment. In line with this, we believe that our results may contribute to antibiotic stewardship and decrease the prevalence of multidrug-resistant organisms.

Moreover, one of the highlights of the paper was the seasonal variability of microbiological agents. Seasonality was particularly important for L. pneumophila infections that occurred almost exclusively during summer and Fall. Given the potential severity of the disease, we propose that clinicians should have higher suspicion index and undergo more intensive diagnostic workup to exclude Legionellosis during that period. A similar approach has been already taken during Influenza season with active virus search during winter, with impact on treatment choices.

Some limitations of the study must be acknowledged. This was a retrospective observational study that could not evaluate all the clinical parameters that may affect etiology. Moreover, a complete microbiological evaluation was not performed in all patients and failure to perform serology cause potential bias in agent identification. Finally, this work was conducted in a particular setting, only including hospitalized patients in a tertiary northern Portuguese hospital. Therefore, these data may not reflect cases of less severity, diagnosed at small community hospitals or in other geographical regions.

In conclusion, besides promoting smoking cessation and encouraging flu and anti-pneumococcal vaccination, investing in the etiological diagnosis, taking into account the most probable agents in a given clinical context can lead to better antibiotic selection thus improving outcomes. Our data suggest that initial empirical coverage of GNB, including P. aeruginosa, should be considered in the elderly with dementia, since these patients usually have a worse prognosis and fail to provide appropriate sputum samples for culture.

Further studies on the epidemiology and prognostic risk factors for CAP caused by GNB and other resistance-prone agents are warranted to assess if early recognition of microbial etiology can modify the outcomes.

The authors declare no conflict of interests.