Predictors of hospital mortality and multidrug-resistant pathogens in hospitalized pneumonia patients residing in the community

Background and objective The 2019 ATS/ADSA guidelines for adult community-acquired pneumonia (CAP) eliminated healthcare-associated pneumonia (HCAP) and considered it to be a form of CAP. This concept, however, was based on studies with relatively small sample sizes. Methods We investigated the risk factors of 30-day mortality, and methicillin-resistant Staphylococcus aureus (MRSA) and Pseudomonas aeruginosa infections in patients with pneumonia coming from the community using the Diagnosis Procedure Combination database, a nationwide discharge database of acute care hospitals. Furthermore, we compared these factors between CAP and HCAP. Results A total of 272,337 patients aged ≥20 years with pneumonia were grouped into 145,082 CAP patients and 127,255 HCAP patients. The 30-day mortality rate (8.9 % vs.3.3 %), MRSA infection (2.4 % vs. 1.4 %), and Pseudomonas aeruginosa infection (1.6 % vs. 1.0 %) were significantly higher in HCAP than in CAP patients. Multivariable logistic regression analysis showed that 12 of 13 identified predictors of mortality (i.e., high age, male, underweight, non-ambulatory status, bedsore, dehydration, respiratory failure, consciousness disturbance, hypotension, admitted in critical care, comorbidity of heart failure, and chronic obstructive pulmonary disease) were identical in CAP and HCAP patients. Similarly, five of six distinct risk factors for MRSA infection, and three of three for Pseudomonas aeruginosa infection were identical between the patients. Conclusion The risk factors for mortality and MRSA or Pseudomonas aeruginosa infection were almost identical in patients with CAP and HCAP. The assessment of individual risk factors for mortality and MRSA or Pseudomonas aeruginosa infection in CAP and abandoning categorization as HCAP can improve and simplify empiric therapy.

Background and objective: The 2019 ATS/ADSA guidelines for adult community-acquired pneumonia (CAP) eliminated healthcare-associated pneumonia (HCAP) and considered it to be a form of CAP.This concept, however, was based on studies with relatively small sample sizes.Methods: We investigated the risk factors of 30-day mortality, and methicillin-resistant Staphylococcus aureus (MRSA) and Pseudomonas aeruginosa infections in patients with pneumonia coming from the community using the Diagnosis Procedure Combination database, a nationwide discharge database of acute care hospitals.Furthermore, we compared these factors between CAP and HCAP.Results: A total of 272,337 patients aged ≥20 years with pneumonia were grouped into 145,082 CAP patients and 127,255 HCAP patients.The 30-day mortality rate (8.9 % vs.3.3 %), MRSA infection (2.4 % vs. 1.4 %), and Pseudomonas aeruginosa infection (1.6 % vs. 1.0 %) were significantly higher in HCAP than in CAP patients.Multivariable logistic regression analysis showed that 12 of 13 identified predictors of mortality (i.e., high age, male, underweight, nonambulatory status, bedsore, dehydration, respiratory failure, consciousness disturbance, hypotension, admitted in critical care, comorbidity of heart failure, and chronic obstructive pulmonary disease) were identical in CAP and HCAP patients.Similarly, five of six distinct risk factors for MRSA infection, and three of three for Pseudomonas aeruginosa infection were identical between the patients.

Conclusion:
The risk factors for mortality and MRSA or Pseudomonas aeruginosa infection were almost identical in patients with CAP and HCAP.The assessment of individual risk factors for mortality and MRSA or Pseudomonas aeruginosa infection in CAP and abandoning categorization as HCAP can improve and simplify empiric therapy.

Introduction
Pneumonia is a common condition that remains one of the leading causes of death worldwide [1].To achieve better outcomes, accurate assessment and classification of patients with pneumonia are imperative to provide appropriate treatment.To achieve this goal, the American Thoracic Society/Infectious Diseases Society of America (ATS/ADSA) guidelines in 2005 and 2007 introduced healthcare-associated pneumonia (HCAP), in addition to community-acquired pneumonia (CAP), as new classifications of pneumonia that need to be treated as hospital-acquired pneumonia (HAP) [2,3].
While research from the United States has shown a higher prevalence of multidrug-resistant (MDR) pathogens in HCAP patients than in CAP patients [4,5], many studies report that HCAP patients have a lower prevalence of MDR pathogens than HAP patients, and that HCAP patients have a similar prevalence of MDR compared to CAP patients [6][7][8].This misestimation of MDR pathogens can cause the overuse of antibiotics leading to adverse outcomes [9,10].Additionally, a systematic review and meta-analysis by Chalmers et al. reported that the evidence supporting the concept of HCAP is based on low-quality research, and may not accurately identify MDR pathogens [11].
Following this debate, the dependent concept of HCAP was eliminated from the 2016 ATS/IDSA guidelines for nosocomial pneumonia and the 2019 guidelines for CAP, and it was considered a form of CAP [12,13].This suggests that exposure to healthcare is not necessarily the determining factor in deciding the treatment strategy, rather the accurate assessment of individual risks for carrying MDR pathogens and mortality is even more important.
Considering the above, we aimed to update and establish the risk factors for mortality and MDR pneumonia, dividing them into Methicillin-resistant Staphylococcus aureus (MRSA) and Pseudomonas aeruginosa infections, and to re-evaluate the classification of CAP and HCAP using the Diagnosis Procedure Combination (DPC) database, which comprise all pneumonia admissions in Japan and their detailed clinical and background information.

Data source
We retrospectively studied patients who were admitted with pneumonia between 2016 and 2017 using records from the DPC database.Detailed information on the DPC database has been described elsewhere [14,15] and is provided in the online supplement.In brief, the database provides clinical and procedural information on all 82 university hospitals and other acute care hospitals in Japan, Fig. 1.Flowchart outlining the selection and categorization of study participants.
T. Ukai et al. covering approximately eight million admissions per year.Participating hospitals have an incentive to provide accurate data under the public insurance system.Physicians are responsible for ensuring the accuracy of clinical information.
Study approval was obtained from the Institutional Review Board of the Ethical Committee for Epidemiology of Hiroshima University (E− 2104).Due to the anonymous nature of the data, the requirement for patient informed consent was waived.

Study population and definitions of pneumonia
All patients aged ≥20 years who were admitted to hospitals with an admission diagnosis (defined as the disease determined to be the basis for hospitalization and necessary treatment) of pneumonia with available information on disease severity between 2016 and 2017 were identified.The admission diagnosis could by modified later at discharge.Therefore, diseases with infiltration such as idiopathic lung disease, pulmonary embolism, lung cancer or congestive heart failure were excluded in this study.The International Statistical Classification of Diseases, 10th Revision (ICD-10) codes included J10.0 (influenza due to identified influenza virus with pneumonia), J11.0 (influenza due to unidentified influenza virus with pneumonia)J12 (viral pneumonia), J13 (pneumonia due to Streptococcus pneumonia), J14 (pneumonia due to Hemophilus influenza), J15 (bacterial pneumonia, not elsewhere classified), J16 (pneumonia due to other infectious organisms, not elsewhere classified), J17 (pneumonia in diseases classified elsewhere), and J18 (pneumonia, unspecified organism) [16].Patients with missing data on variables (n = 373) were excluded from the study (Fig. 1).HCAP was defined as pneumonia in patients with any of the following conditions: hospitalization during the preceding 90 days; residence in a nursing home; or immunosuppressed conditions including cancer, immune disorders, steroid use, immunosuppressant use, and chronic dialysis based on the ATS/IDSA guidelines and previous studies [2,3,5,17].CAP was defined as pneumonia that did not meet the HCAP criteria.

Outcome measures and clinical assessment
The outcomes of interest were in-hospital mortality within 30-days of admission and development of MDR pneumonia.MRSA (ICD code J15.2) and P. aeruginosa (ICD code J15.1) were defined as MDR pathogens based on ATS/IDSA guidelines [2].Pneumonia due to other pathogens, including Acinetobacter baumannii and extended-spectrum beta-lactamase (ESBL)-producing organisms, were not included because they were not identified from the ICD codes and their incidences were considerably smaller than those of MRSA or P. aeruginosa in the Japanese population [18][19][20][21].
The following patient information was extracted from the database: age, sex, body mass index, smoking status (Brickman index), place of residence before admission (home, nursing home, or extended care facility), non-ambulatory status (yes or no), bedsore (yes or no), dialysis (yes or no), immunosuppression defined as having cancer or immunity disorder or using steroids or immunosuppressants (yes or no), intensive care defined as admission to the intensive care unit, date of admission and discharge, date of the previous admission, diagnoses and comorbidities recorded with ICD-10 codes, and medications.
We also used dehydration (blood urea nitrogen ≥21 mg/mL), respiratory failure (pulse oximetry saturation, SpO2≤90 %), consciousness disturbance, and low blood pressure (systolic blood pressure ≤90 mmHg) included in the A-DROP scoring system [15] proposed by the Japanese Respiratory Society.Concurrently, comorbidities of heart failure, liver failure, and chronic obstructive pulmonary disease (COPD), assessed by the Hugh-Jones grade.The variables were chosen from the database based on a previous study [19,22,23]; however, some were excluded because of multicollinearity.
We compared the baseline characteristics, 30-day hospital mortality, and prevalence of MRSA or P. aeruginosa infections, predictors for 30-day hospital mortality, and risk factors for MDR (MRSA or P. aeruginosa) infections between patients with CAP and HCAP.

Statistical analysis
Demographic and clinical characteristics are described as frequencies and percentages according to pneumonia classification (all pneumonia [CAP + HCAP], CAP, or HCAP).The difference between CAP and HCAP was compared using chi-square test.Univariable logistic regression was used to test the association between demographic or clinical characteristics and all pneumonia, CAP, or HCAP.Multivariable logistic regression analyses were performed to evaluate the predictors for 30-day hospital mortality and risk factors for MDR (MRSA and P. aeruginosa) infection.For the 30-day hospital mortality, we adjusted for age, sex, body mass index, smoking status, residence in a nursing home or extended care facility (yes or no), non-ambulatory status (yes or no), bedsore (yes or no), dialysis (yes or no), immunosuppression (yes or no), intensive care (yes or no), dehydration, respiratory failure, orientation disturbance, low blood pressure (systolic blood pressure ≤90 mmHg), the Hugh-Jones grade for patients with COPD, heart failure, and liver failure.For MDR development, we adjusted for variables associated with patients' condition before admission (i.e., age, sex, body mass index, smoking status, residence in a nursing home or extended care facility, non-ambulatory status, bedsore, dialysis, immunosuppression, the Hugh-Jones grade, heart failure, and liver failure).These covariates were selected from previous studies [19,23,24], and also selected on the basis of clinical relevance.The results of univariable logistic regression analysis for the covariates are described in Table S1 and Table S2.Before conducting the analysis, we examined the correlations between the variables, removing one of two variables if the correlation was high to avoid multicollinearity.Odds ratios (ORs) and 95 % confidence intervals (CIs) were calculated.Statistical analyses were performed using STATA version 12.All tests were two-tailed, and a P value less than 0.05 was considered statistically significant.
T. Ukai et al.

Patients' characteristics
A cohort of 272,710 patients was obtained from the database.Individuals with missing data concerning specific variables (n = 373); no available data on residence status before admission [n = 1], bedsores [n = 4], and daily living activity [n = 368]) were excluded from the study.Subsequently, 272,337 patients diagnosed with pneumonia were included in the analysis.Among them, 145,082 (53.3 %) patients were categorized as having CAP, and 127,255 (46.7 %) patients were categorized as having HCAP.The total number of 30-day hospital mortality was 16,119 (5.9 %), and 8638 (3.2 %) developed MDR pneumonia, including 5141 with MRSA (1.9 %) and 3497 with P. aeruginosa (1.3 %) infections.There were significant differences between patients with CAP and HCAP for all variables extracted from the database.Patients with HCAP were more likely to be older, female, underweight, and non-ambulatory and have bedsores and comorbidities such as heart failure, liver failure, and advanced COPD than those with CAP.The condition on admission was more severe for HCAP patients than for CAP patients regarding dehydration, respiratory failure, consciousness disturbance, and hypotension.MDR pathogens were more prevalent (4.0 % vs. 2.4 %, P < 0.001) in HCAP patients than in CAP patients, including MRSA (2.4 % vs. 1.4 %, P < 0.001) and P. aeruginosa (1.6 % vs. 1.0 %, P < 0.001).The 30-day mortality was significantly higher for HCAP than that for CAP (8.9 % vs. 3.3 %, P < 0.001; Table 1).

Table 1
Baseline characteristics of all pneumonia (CAP + HCAP), CAP and HCAP patients.T. Ukai et al.

Risk factors for the mortality
In all pneumonia patients, multivariable logistic regression analysis showed that the independent predictors for 30-day mortality were as follows: older age; male sex; being underweight; non-ambulatory status; having a bed sore; residence in a nursing home or extended care facility; hospitalization within 90 days prior to admission; dialysis; immunosuppression; comorbidities such as heart failure, liver failure, and COPD; and pneumonia severity on admission, including dehydration, respiratory failure, consciousness disturbance, and hypotension.These risk factors were shared between CAP and HCAP patients, except for the comorbidity of liver failure (Table 2).

Risk factors for the MRSA and P. aeruginosa infection
Table 3 shows the risk factors for pneumonia caused by MDR.Underweight, non-ambulatory status, bedsores, and COPD were common dependent risk factors for MDR pneumonia.Table 4 and Table 5 show the risk factors for pneumonia due to MRSA and P. aeruginosa infection separately.Underweight, prior hospitalization, and COPD were associated with both MRSA and P. aeruginosa pneumonia development.Age (>80 years old), male, non-ambulatory status, bedsores, residence in a nursing home or extended care facility, dialysis, and immunosuppression were significantly associated with MRSA pneumonia development but not with Pseudomonas pneumonia.When we compared CAP and HCAP, five out of six risk factors for MRSA pneumonia (male, underweight, non-ambulatory status, bedsore, and COPD), and three out of three risk factors for P. aeruginosa infection (female, underweight, COPD) were identical between CAP and HCAP patients.

Discussion
This study identified the individual risk factors for 30-day mortality and infection with MDR pathogens, including MRSA and P. aeruginosa, in pneumonia patients who resided in the community using a national discharge database of acute hospitals in Japan.By classifying patients into CAP and HCAP groups, we found that patients with HCAP were associated with significantly higher mortality and infection with MDR pathogens than patients with CAP, but they shared almost the same risk factors for 30-day hospital mortality

Table 4
Risk factors for pneumonia due to MRSA in all pneumonia (CAP + HCAP), CAP, and HCAP patients.and the development of MRSA and P. aeruginosa infection.In addition, the risk factors for MRSA infection were different from those for P. aeruginosa infection.This study is the first to comprehensively assess the risk factors for pneumonia by studying the largest number of pneumonia patients from the community compared to previous studies.We found that patients' background characteristics such as age; male sex; being underweight; residence in a nursing home or extended care facility; being bedridden; prior admission; immunosuppression; and comorbidities of heart failure, liver failure, and COPD, as well as severity of conditions at admission, including dehydration, respiratory failure, conscious disturbance, and hypotension were significantly associated with 30-day hospital mortality.These factors have been consistently reported in older studies in the late 1990s and in more recent studies [25][26][27][28][29]. Bedsores were further identified as an independent risk factor for mortality.Predisposing factors for pressure ulcers include immobility, malnutrition, reduced skin perfusion, and sensory loss [30], and these factors overlap with the known risk factors for pneumonia mortality [31,32].After classifying pneumonia patients into CAP and HCAP groups, we found that the risk factors for 30-day mortality were similar between them, except for liver failure.
The 2005 ATS/ADSA guidelines defined the following factors to identify HCAP patients who may have been exposed to healthcare and thus, were at risk of carrying MDR pathogens: patients with hospitalization within 90 days prior to admission; nursing home residents; patients receiving home infusion therapy, hemodialysis and/or wound care; and patients with family members with MDR pathogens [3].However, later studies have reported that the weight of individual risk factors may not be equal [27] and can lead to misuse of antibiotics.Our study identified that hospitalization within 90 days prior to admission (OR 1.66, 95 % CI 1.58-1.74),residing in a nursing home (OR 1.20, 95 % CI 1.13-1.27),hemodialysis (OR 1.31, 95 % CI 1.12-1.52),and immunosuppression (OR 1.06, 95 % CI 1.01-1.12)were significantly associated with the development of MDR pneumonia.While these are independent risk factors for MDR pneumonia, the estimated ORs vary widely.Hence, treating all HCAP patients equally with broad-spectrum antimicrobials may lead to overtreatment of patients with low OR for MDR risk factors.
In addition to the MDR pneumonia risk factors included in the HCAP definition, we identified that being underweight and having a non-ambulatory status, bedsores, and COPD were also independent risk factors for the development of MDR pneumonia.Nonambulatory status and COPD have also been shown to be risk factors in previous studies [33].In our study, COPD grade showed a clear dose-response association with MDR pneumonia development; therefore, the association may be robust [34,35].Being underweight and having bedsores are indications of malnutrition in patients, making them prone to infection with MDR pathogens.These factors were similar in patients with CAP and HCAP and could be added to the individual risk factors to predict MDR pneumonia.Although the incidence of MDR pneumonia was significantly higher in HCAP patients than in CAP patients (4.0 % vs. 2.4 %), the incidence was not high compared with other types of pneumonia, such as HAP or Ventilator-Associated Pneumonia [24].If the 2005 ATS/ADSA guidelines are followed and broad-spectrum antimicrobial agents are equally used in all HCAP patients, overtreatment would be likely in the majority of patients with HCAP.Therefore, it is important to assess patients' individual risk factors for MDR judiciously to reserve the use of broad-spectrum antimicrobial agents.The ATS/IDSA, consequently removed the concept of HCAP from their new guidelines [12,13].
Previous cohort studies evaluating the predictors of MDR pneumonia have combined MRSA and P. aeruginosa [19,21,27].However, it is important to identify the predictors for MRSA and P. aeruginosa infection separately to avoid overuse of broad-spectrum antibiotics because of differences in their antibacterial spectrum.This study revealed that the risk factors for MRSA and P. aeruginosa infections were not identical.We discerned that dialysis (OR 1.88, 95%CI, 1.58-2.24)for MRSA pneumonia, and low body weight (OR 2.09, 95% CI, 1.95-2.23)for P. aeruginosa pneumonia constitute distinctive risks.Clinicians are encouraged to ascertain the pathogen of the infection as comprehensively as possible utilizing this evidence, whilst referring to locally validated risks as suggested by the 2019 guidelines for CAP [13].Furthermore, an exploration of pathogenesis concerning the rationale behind the differences between MRSA and P. aeruginosa is warranted.
This study had several limitations.First, the DPC database lacks comprehensive details in areas such as socioeconomic status, laboratory tests, radiographic images, and microbiological tests.Based on these data, it can be deduced that clinicians had comprehensibly confirmed the diagnosis; however, the specific criteria they used are not clear.This extends to how they differentiated between colonization and infection in patients with MDR infections.Second, while this study includes all patients from acute hospitals using DPC, its findings may not apply universally as all participants were Japanese.Third, patients were selected based on specific ICD-10 codes (J10 to J18).Previous studies showed that these codes reliably identify pneumonia [16], but this method may have excluded some pneumonia patients with other codes, and included some misdiagnosed patients.Lastly, the study uses data collected between 2016 and 2017, which, while not the most current, we believe is still significant.We contend that the risks for 30-day mortality and MDR infection remain fairly consistent over time.

Conclusions
This observational study from a nationwide discharge database of acute care hospitals described the clinical characteristics and frequency of MDR infections including MRSA and P. aeruginosa, and their outcomes in patients with CAP and HCAP.The risk factors for MDR infection, including MRSA and P. aeruginosa, and mortality were nearly identical in patients with CAP and HCAP.The risk factors for MRSA and P. aeruginosa infections were not identical.The assessment of individual risk factors for mortality and MDR infection, and including HCAP in CAP could contribute to the improvement and simplification of empiric therapy.This may also be useful to assess predictors for MRSA and P. aeruginosa infections separately to minimize the overuse of broad-spectrum antibiotics.

Financial contribution
This research was supported by the World Health Organization Center for Health Development (WHO Kobe Center, WKC: K18003).

Table 2
Risk factors for 30-day in hospital mortality among all pneumonia patients (CAP + HCAP), CAP and HCAP.

Table 3
Risk factors for pneumonia due to multidrug-resistant pathogens (MRSA and P. Aeruginosa) in all pneumonia (CAP + HCAP), CAP, and HCAP patients.

.27(1.11-1.46)
a Immunosuppression: patients who have cancer or immunodeficiency, or who use immunosuppressants.b Adjusted for all variables listed in this table.

Table 5
Risk factors for pneumonia due to P. Aeruginosa in all pneumonia (CAP + HCAP), CAP, and HCAP patients.