Pathogens Isolated in Respiratory Tract Samples of Recently Hospitalized Patients: Implications for Primary Care Physicians

Little is known about the etiology of community-acquired lower respiratory tract infections (CA-LRTI), and treatment is largely empirical. We aimed to provide clinicians with microbiologic data of the more severe cases, i.e. those that result in hospital admission. A retrospective observational cohort study was conducted at a tertiary care hospital in Haifa, Israel. Consecutive respiratory tract samples obtained from admitted patients older than 12 years between 2014 and 2020 were included. Pathogen distribution and drug susceptibility were described, and factors associated with 14 day mortality rates were analyzed using a multivariable logistic regression with a stepwise model reduction. A total of 1,395; 2,212; and 2,760 samples were included in the community-acquired, healthcare-associated and hospital-acquired LRTI groups, respectively. Gram negative bacteria were the most common bacteria isolated. Streptococcus pneumoniae was over represented in patients admitted to the intensive care unit (ICU) with CA-LRTI and healthcare-associated LRTI (accounting for 20% and 10.1% of patients admitted to the ICU versus 12.6% and 6.4% of patients in the entire cohort, p < 0.001). Streptococcus pneumoniae was susceptible to penicillin in around 50% of cases; to erythromycin in 65% of cases; and to uoroquinolones and 3rd generation cephalosporins in more than 95% of cases. Twenty-ve percent of other typical community-acquired pathogens (Hemophillus inuenza, Moraxella spp., and Bordatella spp.) were resistant to ampicillin. Susceptibility of Gram negative bacteria to penicillins with and without a β-lactamase inhibitor and to second generation cephalosporins was lower than 50%. In the multivariable analysis, factors that were associated with increased mortality in the entire cohort were: age, admission to an ICU or surgical department, healthcare-associated or hospital-acquired infections, and infections with gram negative bacteria, Staphylococcus aureus, and Stenotrophomonas or Acinetobacter baummannii. multiple comorbidities. Streptococcus pneumoniae and other typical community acquired pathogens are proportionally more common in younger patients, although most absolute cases occur in elderly patients. The susceptibility pattern of these community pathogens suggests that for high-risk patients the choice of effective oral antibiotic agents in the community is limited.

Our data shows that gram negative bacteria are common causative agents among elderly patients with multiple comorbidities. Streptococcus pneumoniae and other typical community acquired pathogens are proportionally more common in younger patients, although most absolute cases occur in elderly patients. The susceptibility pattern of these community pathogens suggests that for high-risk patients the choice of effective oral antibiotic agents in the community is limited. Background A community-acquired lower respiratory tract infection (CA-LRTI) is one of the most common infections attended by primary care physicians and results in considerable use of antibiotics [1,2]. As microbiological analysis of sputum cultures is rarely performed in community settings, little is known about the etiology of CA-LRTI, and bacterial susceptibilities of causative pathogens are unknown in most areas. Treatment of LRTI by primary care physicians is thus largely empirical with penicillins, 2nd and 3rd generation cephalosporins, uoroquinolones, doxycycline and macrolides often used. On the one hand, such empirical treatment may be super uous resulting in more antibiotic resistance in the community [3]. On the other hand, for the few patients whose treatment fails and results in eventual hospital admission, antibiotic treatment may be inadequate.
In the few studies that have examined the microbiology of community acquired pneumonia among relatively young patients, the only culturable pathogens that were found in a signi cant proportion of cases (i.e. > 5%) are Streptococcus pneumoniae and Hemophillus in uenza [4]. Other pathogens, such as Moraxella spp., Bordatella spp. and Staphylococcus aureus are uncommon [5]. Whereas, common pathogens such as Legionella pneuumophila, Mycoplasma pneumoniae or Chlamydia pneumoniae are only diagnosed through molecular methods or serological tests [6]. Gram negative bacteria are detected in 2.8%-14% of community acquired pneumonia cases, particularly in the elderly and patients who are admitted from a long-term care facility (LTCF) [7,8].
By analyzing the results of respiratory tract cultures of recently admitted patients, we aimed to provide physicians with the relevant microbiologic data of the more severe CA-LRTI cases, i.e. those that resulted in hospital admission. Such microbiological data would provide primary care and emergency room physicians with additional insights as to the causative agents of severe CA-LRTI, thereby improving the process of selecting empirical antibiotic therapy for patients with high-risk LRTI in the community.

Study Setting and Cohort Description
A retrospective observational cohort study was performed at the Rambam Health Care Campus (RHCC), a 970-bed primary and tertiary care hospital in Haifa,

Israel.
The cohort included all consecutive respiratory tract samples obtained from patients older than 12 years who were admitted to the RHCC between January 2014 and January 2020. The Institutional Review Board of the RHCC approved the study for waiver of informed consent.

Data Collection
For each patient's admission throughout the study period, all unique bacterial species were included (i.e. the rst bacteria of each species for each admission). Data were retrieved with the use of the institution's electronic medical records. Data pertaining to demographics, source of admission (home, LTCF), hospitalization at the RHCC within the past 90 days, Charlson Comorbidity Index (CCI), and admission unit were collected for each admission. Microbiology data included dates of sample collection, dates of results, name of bacteria species, and antimicrobial susceptibility.
We de ned a group of pathogens as 'typical community pathogens' based on previous large studies that showed Streptococcus pneumoniae and Hemophillus in uenza to be by far the most common bacterial pathogens to cause CA-LRTI [4]. Moraxella and Bordatella spp., although uncommon, were also included in this group as they rarely cause hospital-acquired pulmonary infections [9,10].
Drug resistance was de ned according to the Clinical and Laboratory Standards Institute (CLSI) breakpoints for susceptibility. Bacteria with intermediate drug susceptibility were de ned as resistant. Finally, fourteen-day mortality was assessed in relation to the date the index culture was obtained.

Variable De nitions
The place of acquisition of each culture was classi ed with the use of three factors: 1) the time elapsed from date of admission to the date the sample was collected; 2) whether the admission was from a LTCF; and 3) if the patient had been hospitalized within the past 3 months. Accordingly, the study consisted of three cohorts as follows: i) Community-acquired cultures which were de ned as cultures obtained within less than 48 hours of admission of patients who were not residing in an LTCF and were not hospitalized in the RHCC within the past 3 months; ii) Healthcare-associated (but community-acquired) cultures which were de ned as cultures collected within 48 hours of admission of patients who were either admitted from an LTCF or hospitalized in the RHCC within the past 3 months; and iii) Hospital-acquired cultures which were de ned as cultures collected from patients who were hospitalized for no less than 3 days but for no more than 7 days.
The patients' ages at the time of culture collection were strati ed into 3 groups as follows: 12-18 years; >18 years to 40 years; and older than 40 years. The CCI was strati ed into 3 groups: 1) a CCI of 0; 2) a CCI between 1 and 4; and 3) a CCI of > 4.

Statistical Analysis
Continuous variables were described with the use of median and 25-75% interquartile ranges (IQR). We analyzed factors associated with a 14-day mortality in the entire cohort, and in the cohort of patients with typical community pathogens (Streptococcus pneumoniae, Hemophillus in uenza, Bordatella spp., Moraxella spp.) using a multivariate logistic regression with a stepwise model reduction, and accounting for age, gender, referring ward, and place of acquisition. A p-value of 0.05 was considered statistically signi cant.

Results
A total of 1,395; 2,212; and 2,760 samples were included in the community-acquired, healthcare-associated and hospital-acquired LRTI groups, respectively. Of the 2,212 healthcare-associated samples, 295 were obtained from patients coming from LTCFs, while 1917 were obtained from patients with previous hospital admissions. Samples taken from female patients accounted for 1884 / 6367 (29.6%) of the entire cohort. The mean age was 60.6 ± 19.9 years (median 65, IQR = 50-75), and the mean CCI was 5.7 ± 4.1 (median 5, IQR = 2-8.5). Samples were obtained in the pediatric ward (267, 4.2%); ICU (1823, 28.6%); emergency room and internal medicine wards (3190, 50.1%); oncology and hemato-oncology wards (190, 2.5%); and surgery and orthopedics wards (897, 14.1%). Sputum samples, tracheal aspirations or broncho-alveolar lavage were used in 3912 (61.4%), 1879 (29.5%), and 576 (9%) of all specimens, respectively. These demographic data are presented in Table 1. Table 1 Characteristics of patients with LRTI Broncho-alveolar lavage (N (%)) 576 (9) LRTI, lower respiratory tract infection Pathogen distribution among the three study cohorts is shown in Table 2. The probability of isolating bacteria which we initially expected to be "typical" community-acquired pathogens, namely Streptococcus pneumoniae, Hemophillus in uenza, Bordatella spp., Moraxella spp., was higher among patients with CA-LRTI (p < 0.0001 for all comparisons). Enterobacteriacea, non-fermenting gram negative bacteria, and other gram negative bacteria, were more frequently isolated from patients with healthcare-associated and hospital-acquired LRTI, but were by far the most common pathogens isolated in all cohorts (p < 0.0001 for all comparisons), while Staphylococcus aureus was relatively equally distributed among the three study cohorts. These observations were also true when patients were divided into age groups (12-  Overall, the number of respiratory cultures was higher among patients older than 40 years compared to other age groups. The relative proportion of samples with a growth of Streptococcus pneumoniae was signi cantly higher (p < 0.001) among patients aged 18 to 40 years (9.8% of all samples) when compared to younger patients (7.1%), or to adults aged more than 40 years (5.9%), although the absolute number of samples positive for this bacterium was much higher among older patients. Similarly, Hemophillus in uenza, Moraxella spp., and Bordatella spp., were isolated in 15.1%, 18.1% and 12.6% of samples from patients aged 12 to 18 years, > 18 to 40, and > 40 years, respectively (p < 0.001 for all comparisons), but the vast majority of samples positive for these bacteria were taken from patients older than 40 years.
Pathogen distribution in samples taken from patients in the ICU is shown in Table 3. Among patients with CA-LRTI and healthcare-associated LRTI, Streptococcus pneumoniae was over-represented among patients with CA-LRTI and healthcare-associated LRTI admitted to the ICU (accounting for 20% and 10.1% among patients admitted to the ICU versus 12.6% and 6.4% among patients in these cohort, p < 0.001 for both comparisons). Other typical community pathogens (Hemophillus in uenza, Moraxella spp. and Bordatella spp.), gram negative bacteria and Staphylococcus aureus were not more prevalent among patients in the ICU when compared to their relative frequency in the entire cohort. The table shows the frequency of pathogens taken from patients hospitalized in all departments and in the intensive care unit (ICU) at the time of sample acquisition.
Antibiotic susceptibilities of various bacteria are shown in Table 4. Streptococcus pneumoniae was susceptible to penicillin in only around 50% of cases, and to erythromycin in 65% of cases. Streptococcus pneumoniae bacteria were susceptible in more than 95% of cases to uoroquinolones (both cipro oxacin and levo oxacin) and to 3rd generation cephalosporins (ceftriaxone). Other typical community-acquired pathogens (Hemophillus in uenza, Moraxella spp., and Bordatella spp.) were also nearly universally susceptible to uoroquinolones and 3rd generation cephalosporins, but about 25% were resistant to ampicillin. Gram negative bacteria were resistant to penicillins, penicillin and β-lactamase combinations, and 2nd generation cephalosporins in more than 50% of cases. Susceptibility rates to uoroquinolones and 3rd generation cephalosporins were around 80% and 75% respectively (Table 4).   Factors associated with a 14 day mortality were analyzed in the entire cohort, including gram negative bacteria and typical community pathogens (Streptococcus pneumoniae, Hemophillus in uenza, Moraxella spp., and Bordatella spp.). In the multivariable analysis, factors that were associated with increased mortality in the entire cohort were: age, admission to an ICU, admission to a surgical department, healthcare-associated or hospital-acquired infections (when compared to community acquired infections), and infections with Staphylococcus aureus, Enterobacteriaceae, or Stenotrophomonas and Acinetobacter baummannii (Tables 6a and 6b).  *Increase for each year ¶ An interaction was found between the Charlson Comorbidity Index and the place of acquisition. Subgroup analysis for healthcare-associated infection is described in the text above.
In the multivariable analysis that included only typical community-acquired pathogens (Streptococcus pneumoniae, Hemophillus in uenza, Moraxella spp., and Bordatella spp.), factors that were associated with an increased risk of mortality included ICU and surgical ward admission. Because of a strong interaction between the place of acquisition and the CCI, the cohort was strati ed into subgroups according to the place of acquisition and the multivariable analysis for a 14-day mortality was repeated. In the subgroup of healthcare-associated LRTI, a CCI between 1 and 4, as well as a CCI greater than 4 were independent predictors for a 14-day mortality (OR 12.94, 95% CI 1.65, 101.15, p = 0.01; OR 14.84, 95% CI 1.70, 129.15, p = 0.01, respectively). The CCI was not associated with a 14-day mortality in the other subgroups.

Discussion
This retrospective cohort study provides data on respiratory tract specimens of patients who were recently admitted to a hospital. Samples were divided into three groups: those taken within 48 hours of hospital admission from patients without known healthcare exposure; those taken within 48 hours of hospital admission from patients with known healthcare exposure; and samples taken 3-7 days after hospital admission. Gram negative bacteria were surprisingly the most common pathogens isolated in all of these cohorts. As expected, typical community-acquired pathogens (i.e. Streptococcus pneumoniae, Hemophillus in uenza, Bordatella spp., and Moraxella spp.) were more frequently isolated among patients with community-acquired infections.
Although among young and relatively healthy patients, a larger proportion of samples had growth of typical community pathogens (Streptococcus pneumoniae, Hemophillus in uenza, Bordatella spp., Moraxella spp.), in terms of absolute numbers the majority of these cases occurred in patients older than 40 years. Community-acquired LRTI cases caused by Streptococcus pneumoniae, but not other typical community pathogens, were more likely to be admitted to an ICU and to die.
Our results highlight the limited choice of available oral antibiotics for patients with CA-LRTIs. Primary care physicians often use penicillins with or without β lactamase inhibitors, 2nd generation cephalosporins, macrolides, doxycycline, and uoroquinolones for LRTIs. Typical community pathogens were nearly always susceptible to 3rd generation cephalosporins that are not universally available in an oral formulation, and to uoroquinolones. Susceptibility to penicillin, ampicillin, and macrolides was consistently much lower among all study groups, and lower than those described in several previous European and North American studies [4,[11][12][13]. Gram negative bacteria, isolated mostly from frail elderly patients, were resistant to 2nd generation cephalosporins, penicillins and β-lactamase inhibitors in more than 50% of cases. Although Hemophillus in uenza, Bordatella spp., Moraxella spp. were nearly universally susceptible to amoxicillin and clavulanic acid, such combinations are unlikely to provide additional bene t over penicillins for the treatment of penicillinresistant Streptococcus pneumoniae, as resistance in these bacteria is mediated by mutations in the penicillin-binding protein rather than by the production of β-lactamases [14]. Other antibiotic agents may also not be adequate for the treatment of LRTI. For instance, trimethoprim sulphamethoxazole is not part of treatment guidelines. Furthermore, 2nd generation cephalosporins in this study were active against only about 50% of gram negative bacteria, are probably less e cacious, and are mostly available in oral formulations with relatively low bioavailability [15,16].
The therapeutic implications of these data are not easy to determine, as many patients with CA-LRTI are treated empirically, and microbiologic data is either received late in the course of treatment or is unavailable altogether. However, it is likely that for the vast majority of younger patients with a low number of comorbid conditions, the likelihood of an adverse outcome from a CA-LRTI is very low, while the ecological price of antibiotic overuse is high [17]. Since CA-LRTIs are very common, resistance to administered antibiotics develops quickly [18]. Treatment with macrolides, 2nd generation cephalosporins, and penicillin and β-lactamase inhibitor combinations is likely to result in a good outcome in low-risk patients. Such low-risk patients who are infected with Streptococcus pneumoniae with intermediate resistance to penicillin can probably be adequately treated by higher doses of penicillin. Conversely, for patients who are at a higher risk of complications, i.e. older patients with multiple comorbid conditions, failure of a 1st line antibiotic treatment administered in the community setting may result in increased use of 2nd line antibiotics, and increased rates of hospital admission, mechanical ventilation, and death. A similar population includes older patients discharged from hospital to complete antibiotic treatment for an LRTI in the community. These patients nearly always have multiple risk factors for treatment failure and microbiologic data are usually unavailable. In such patients an initial or follow up treatment using penicillins with or without β-lactamase inhibitors, macrolides, or 2nd generation cephalosporins may be inadequate. Fluoroquinolones, and more speci cally the so called "respiratory" uoroquinolones, provide excellent coverage according to our data and those of previous studies [19][20][21]. The use of uoroquinolones, however, has several important drawbacks for patients with signi cant comorbidities. Namely, they are associated with an increased risk of Closridioides di cile associated diarrhea and probably more so than most other antibiotics [22]; they can cause QT prolongation that may lead to arrhythmias; and their use is associated with tendinopathies, peripheral neuropathy, and rupture of aortic aneurysms [22]. Treatment with uoroquinolones quickly leads to colonization with resistant pathogens in the treated patients [18,23], and increases the risk of uoroquinolones resistance in certain geographic areas even in the absence of personal antibiotic use [24].
Improving diagnostics with the use of molecular tools for identi cation of respiratory pathogens and / or their resistance genes may facilitate antibiotic choice in the near future. In addition, newer antibiotics such as lefamulin, and oral formulations of 3rd generation cephalosporins (i.e. cefditoren pivoxil or cefpodoxime) may prove useful for high-risk patients with relative contraindications to uoroquinolones.
There are several important limitations for this study. The data are retrospective, and are based on microbiological sampling results. Occasional patients with no active infections could have been included, although RHCC policy is that respiratory tract samples are taken only for patients with suspected infections.
Conversely, samples may not have been obtained for various reasons from patients with clinically signi cant LRTI. We may have missed some CA-LRTI cases caused by less common pathogens. Empirical treatment, however, should be based on the antibiotic susceptibility pro le, safety, ecological cost and economic cost of speci c agents, and targeted towards the most common community pathogens. Susceptibility tests were performed according to CLSI requirements. Therefore, there are some gaps in potentially useful data, i.e. Streptococcus pneumoniae susceptibility rates to 2nd generation cephalosporins.

Conclusions
Our data shows that gram negative bacteria are very common among patients admitted because of a LRTI, and typical community acquired pathogens (Streptococcus pneumoniae, Hemophillus in uenza, Bordatella spp., Moraxella spp.) are especially common among older patients with multiple comorbidities. The isolation of these bacteria in respiratory tract samples is associated with appreciable rates of ICU admission and 14-day mortality rates. The susceptibility pattern of these community pathogens suggests that for high-risk patients, the choice of effective oral antibiotic agents is limited. Both improvements in diagnostic procedures aimed at providing targeted therapy and newer oral antibiotic formulations are needed. Authors contributions: OH -data analysis, writing of manuscript. ARB -data analysis, review of manuscript, AN -study concepts, writing of manuscript Ethics approval and consent to participate: This study protocol was exempted for review by the Institutional ReviewBoard of the Rambam Health Care Center according to the exemptioncriteria. In this study, anonymized and deidenti ed information was used forthe analyses; thus, informed consent was not required. Figure 1 Frequency of bacteria isolated in respiratory tract cultures taken from hospitalized patients by age groups Figure 2 Frequency of bacteria isolated in respiratory tract cultures taken from patients by Charlson Comorbidity Index