Clinical risk factors for admission with Pseudomonas and multidrug-resistant Pseudomonas community-acquired pneumonia

Microbial etiology for community-acquired pneumonia (CAP) is evolving with pathogens known for high CAP mortality e.g., Pseudomonas species. Chronic obstructive pulmonary disease (COPD) patients are at risk for hospitalization for CAP. Understanding regional patterns and risk factors for multidrug-resistant (MDR) Pseudomonas acquisition has implications for antimicrobial stewardship. To evaluate the regional epidemiology of MDR Pseudomonas CAP and its association with COPD. We queried the electronic medical records of the University of Alabama at Birmingham Healthcare System to identify patients hospitalized for CAP with Pseudomonas positive respiratory samples between 01/01/2013–12/31/2019. Log binomial regression models were used to examine associations between COPD diagnosis and risk of Pseudomonas/MDR Pseudomonas CAP. Cohort consisted of 913 culture positive CAP cases aged 59-year (IQR:48–68), 61% (560) male, 60% (547) white, 65% (580) current/past smokers, and 42% (384) COPD. Prevalence of Pseudomonas CAP in culture positive CAP was 18% (167), MDR Pseudomonas CAP in Pseudomonas CAP was 22% (36), and yearly incidence of MDR Pseudomonas CAP was stable (p = 0.169). COPD was associated with Pseudomonas CAP (RR 1.39; 95% CI 1.01, 1.91; p = 0.041) but not with MDR Pseudomonas CAP (0.71; 95% CI 0.35, 1.45; p = 0.349). Stroke (RR 2.64; 95% CI 1.51, 4.61; p = 0.0006) and use of supplemental oxygen (RR 2.31; 95% CI 1.30, 4.12; p = 0.005) were associated with MDR Pseudomonas CAP. Incidence of MDR Pseudomonas CAP was stable over time. COPD was associated with Pseudomonas CAP but not with MDR Pseudomonas CAP. Larger cohort studies are needed to confirm findings.

COPD, was the fourth leading cause of death in the US in 2019 with 156,979 deaths (age-adjusted death rate of 38.2 per 100,000 population); in Alabama, it was the third leading cause of death with 3530 deaths (age-adjusted death rate of 55.6 per 100,000 population) [4]. Traditionally, Streptococcus pneumoniae (pneumococcus) is the most common bacteria isolate in CAP; other common bacterial isolates include Hemophilus influenzae, Moraxella catarrhalis, and atypical bacteria (i.e., Mycoplasma pneumoniae, Chlamydia pneumoniae) [5][6][7]. But, over the years, there has been a decline in the prevalence of pneumococcal pneumonia, especially in the US. The decline in pneumococcal pneumonia has been majorly linked to an increase in pneumococcal vaccination [8][9][10][11]. Also in the last decade, there were some reports of the emergence of CAP caused by bacteria that are conventionally not implicated in CAP including Pseudomonas and methicillin resistant Staphylococcus aureus (MRSA) [12,13]. The proportion of CAP due to these bacteria could differ by region and time such that there is a need for local and temporal risk assessment.
Lack of identifying CAP due to Pseudomonas and resistant Pseudomonas may lead to inappropriate antimicrobial treatment, which can worsen CAP morbidity and mortality, increase the risk for antimicrobial resistance, and increase healthcare utilization and cost [14]. Furthermore, considering that pneumonia caused by Pseudomonas is associated with increased mortality [15,16], understanding the regional epidemiology and antibiotic resistance profile of these bacteria is important in CAP management, especially in vulnerable population which includes those with COPD comorbidity. This study aims to assess the local epidemiology of Pseudomonas and multidrug-resistant Pseudomonas, and the association of COPD comorbidity with these bacteria in patients hospitalized with community-acquired bacterial pneumonia who had positive Pseudomonas isolates from respiratory tract samples.

Study design and population
This was a retrospective clinical cohort study of patients that were admitted to the University of Alabama at Birmingham (UAB) Healthcare System between 01/01/2013-12/31/2019 with a bacterial pneumonia diagnosis. We used bacterial pneumonia diagnosis from International Classification of Diseases (ICD) codes which include ICD 9 (481, 482, 483) and ICD 10 (J13, J14, J15, and J16) to identify patients. Only ICD 9 and 10 codes designated as 'final' and/or 'confirmed' in the electronic medical records (EMRs) were considered for disease diagnosis. Also, we used a base cohort of hospital inpatients aged 18 years or older admitted from a physician's office or a non-healthcare facility and who had bacterial pneumonia diagnosis. A patient in the base cohort must have bacterial pneumonia diagnosis recorded in the EMRs to be present on admission. In cases where there was bacterial pneumonia diagnosis but no information about it being present on admission, the patient must have a microbiology culture sample collected within 48 h of admission. From the base cohort, we excluded those with cystic fibrosis, bronchiectasis, no respiratory samples (sputum, bronchoalveolar lavage, bronchial wash, or tracheal aspirate), no culture isolates, and those with isolates from samples collected after 48 h from admission. For patients with multiple episodes of hospitalization, only the first episode was included. The University of Alabama at Birmingham Institutional Review Board approved this study.

Data and data source
Data was obtained from EMRs through the UAB Informatics for Integrating Biology and the Bedside (i2b2) program. The i2b2 program is an NIH-funded National Center for Biomedical Computing based at the Partners HealthCare System. We obtained data on patients' socio-demographic characteristics; microbial culture and susceptibility; hospitalization, comorbidities, and other clinical records. We used patients' comorbidities and validated weights to calculate Charlson comorbidity index (CCI) [17,18].

Risk factors
The primary risk factor was a COPD diagnosis before or during hospitalization with bacterial pneumonia diagnosis. Comorbidity with COPD was identified with ICD 9 codes (490, 491, 492, 495, 496, 506, 506.4) and ICD 10 codes (J40, J41, J42, J43, J44) designated as 'final' and/or 'confirmed' . We classified COPD based on time of diagnosis. Patients who had their first diagnosis of COPD during the current admission were classified as non-preexisting COPD; those diagnosed with COPD before the current admission were classified as pre-existing COPD.

Isolates
A total of 1986 patients were admitted from community settings (home or physician office) with bacterial pneumonia diagnosis present on admission, or microbiology culture sample collected within 48 h of admission in cases where there was bacterial pneumonia diagnosis but no information about it being present on admission ( Fig. 1

Discussion
In a clinical cohort of patients hospitalized with bacterial CAP between 2013 and 2019 at a tertiary hospital in the southeastern USA, we examined the epidemiology of CAP with Pseudomonas and MDR Pseudomonas isolates, and the association of this CAP with COPD comorbidity. In patients with culture-positive respiratory samples, the estimated prevalence of CAP with Pseudomonas isolates was 18%; among CAP patients with Pseudomonas isolates, the estimated prevalence of CAP with MDR Pseudomonas was 22%. There was no significant trend in the yearly incidence of CAP with MDR Pseudomonas isolates over the years. Lastly, though COPD was associated with the risk of isolating Pseudomonas isolates, it was not associated with the risk of isolating MDR Pseudomonas isolates.
There are variations in the estimated prevalence of CAP due to Pseudomonas and MDR Pseudomonas in different studies. This difference is possibly due to different populations [13,[20][21][22][23][24]. In an observational study by Cilloniz et al. in an European population, the estimated 15-year prevalence of CAP due to Pseudomonas isolates was 4% among patients with culture positive CAP [16]. This was lower than the 7-year estimated prevalence of 18% in the current study. However, the estimated prevalence of MDR Pseudomonas aeruginosa among  Pseudomonas aeruginosa isolates reported by Cilloniz et al. was higher than the current study (32% vs 22%) [23]. In a multinational study by Restrepo et al., the estimated global prevalence of CAP due to Pseudomonas aeruginosa was 11.3% in isolate-positive CAP cases. Again, this was lower than the estimate (18.3%) for the current study which drew its population from the southeastern US. The estimated global prevalence for MDR Pseudomonas aeruginosa, irrespective of positive culture isolates, reported in the multinational study was 2.8%. This, again, was lower than the estimated prevalence reported in the current study (3.9%) [24]. Though the current study focused on any Pseudomonas isolates while the multinational study focused on Pseudomonas aeruginosa isolates, 98% of Pseudomonas isolates in the current study were Pseudomonas aeruginosa. There may be more cases of Pseudomonas CAP in southeastern United States when compared to the global average. Understanding validated regional differences in Pseudomonas CAP could help guide the clinical treatment of CAP at the regional level. It will be informative to explore region-specific characteristics that drive regional differences in the prevalence of Pseudomonas and MDR Pseudomonas CAP in future research; specifically, how regional differences in   vaccination (e.g., pneumococcal vaccines), antimicrobial prescribing practices, and other factors affect regional differences in MDR Pseudomonas CAP epidemiology. According to the Centers for Disease Control and Prevention 2020 data on outpatient prescription of fluoroquinolones dispensed in US pharmacies, Southern US (e.g., Alabama, Mississippi) accounts for the highest prescription rates [25]. Alabama had the second highest rate of outpatient prescriptions of fluoroquinolones with 77 prescriptions per 1000 population, after Mississippi (1 st ) with 82 prescriptions per 1000 population. A lot of effort is geared toward improving rational use of antibiotics through antimicrobial stewardship [26]. Cilloniz et al., like the current study, found that COPD was associated with the risk of CAP due to Pseudomonas, and it was not associated with the risk of MDR Pseudomonas [16]. In a meta-analysis that examined risk factors for MDR Pseudomonas aeruginosa, previous antibiotics use and hospital admission, including intensive-care unit, were the risk factors identified [27]. It is important to note that most of the studies included in the meta-analysis had base cohorts that were not CAP. In patients with intensive care unitacquired pneumonia due to Pseudomonas aeruginosa, Barat et al. found no association between COPD and risk of MDR Pseudomonas in adjusted models [28]. Restrepo et al., also found that chronic lung disease   was associated with Pseudomonas CAP and MDR Pseudomonas CAP in their multi-nation study [24]. Different results observed by various studies with the association between COPD and MDR Pseudomonas suggest that a regional approach to the assessment of MDR Pseudomonas may be the effective means to manage this public health burden. Diagnosis for dependence on supplemental oxygen during the current or a previous hospital visit was one of the risk factors for MDR Pseudomonas identified in the present study in an adjusted model. Supplemental oxygen is a medical device that provides oxygen supply to patients with low oxygen levels. Regular use of respiratory devices like supplemental oxygen could promote the formation of bacterial biofilm which could predispose patients to infections when there is poor hygiene in the handling of the device [29,30]. This may be a potential explanation for the association between dependence on supplemental oxygen and the risk of MDR Pseudomonas. Also, a patient who depends on supplemental oxygen could have had a previous hospitalization which required mechanical ventilation. This is also a potential source of acquisition of MDR Pseudomonas [28,[31][32][33]. There are different indications for the use of supplemental oxygen, and COPD is one of them, especially severe one. Though we did not find any association between COPD and risk of MDR Pseudomonas, future research could examine the interaction between COPD and the use of supplemental oxygen and the risk of MDR Pseudomonas.
We found that among CAP patients with Pseudomonas isolates, those with current or previous stroke diagnosis were more likely to have MDR Pseudomonas isolates in their respiratory samples when compared to those with no stroke diagnosis. Stroke clinical management may involve frequent hospital admissions and prolong length of hospital stay [34][35][36]. Such frequent and prolonged patient interactions with healthcare facilities have been shown to be risk factors for acquisition of drug-resistant bacteria like Pseudomonas [27]. This may be the mechanism which makes stroke comorbidity to be associated with the risk of MDR Pseudomonas CAP. We also found that increasing age was associated with lower risk of MDR Pseudomonas CAP. It is possible that in this study setting known risk factors for MDR bacteria, like exposure to broad spectrum antibiotics, are lower in older patients.
While the current study had a substantial population to estimate the prevalence, incidence, and risk of Pseudomonas CAP, we had little sample size (161) for the analysis of the risk of MDR Pseudomonas among CAP patients with Pseudomonas isolates. We also relied on EMR for patient information, so the data we used for analysis was limited to what was obtainable in the EMR. Though EMR data might have some weaknesses, it is inexpensive and provides real-world information.

Conclusions
In summary, we found that the incidence of MDR Pseudomonas CAP was stable over time, and prevalences of Pseudomonas and MDR Pseudomonas communityacquired pneumonia were different in this study population when compared to other regions, highlighting the importance of leveraging local epidemiology and validated risk factors for antimicrobial stewardship guidance. Chronic obstructively pulmonary disease was associated with Pseudomonas CAP but not with MDR Pseudomonas CAP. Larger cohort studies are needed to confirm these findings.