Etiology, Treatments, and Outcomes of Patients With Severe Community-Acquired Pneumonia in a Large U.S. Sample

Abstract

OBJECTIVES:

Compare the clinical practice and outcomes in severe community-acquired pneumonia (sCAP) patients to those in non-sCAP patients using guideline-defined criteria for sCAP.

DESIGN:

Retrospective observational cohort study.

SETTING:

One hundred seventy-seven U.S. hospitals within the Premier Healthcare Database.

PATIENTS:

Hospitalized adult (≥ 18 yr old) patients with pneumonia.

MEASUREMENTS AND MAIN RESULTS:

Adult patients (≥ 18 yr old) with a principal diagnosis of pneumonia or a secondary diagnosis of pneumonia paired with a principal diagnosis of sepsis or respiratory failure were included. Patients with at least one guideline-defined major criterion for severe pneumonia were compared with patients with nonsevere disease. Among 154,799 patients with pneumonia, 21,805 (14.1%) met criteria for sCAP. They had higher organ failure scores (1.9 vs 0.63; p < 0.001) and inpatient mortality (22.0 vs 5.0%; p < 0.001), longer lengths of stay (8 vs 5 d; p < 0.001), and higher costs ($20,046 vs $7,543; p < 0.001) than those with nonsevere disease. Patients with sCAP had twice the rate of positive blood cultures (10.0% vs 4.5%; p < 0.001) and respiratory cultures (34.2 vs 21.1%; p < 0.001) and more often had isolates resistant to first-line community-acquired pneumonia antibiotics (10% of severe vs 3.1% of nonsevere; p < 0.001). Regardless of disease severity, Streptococcus pneumoniae was the most common pathogen recovered from blood cultures and Staphylococcus aureus and Pseudomonas species were the most common pathogens recovered from the respiratory tract. Although few patients with sCAP had cultures positive for a resistant organism, 65% received vancomycin and 42.8% received piperacillin-tazobactam.

CONCLUSIONS:

sCAP patients had worse outcomes and twice the rate of culture positivity. S. aureus and S. pneumoniae were the most common organisms in respiratory and blood specimens, respectively. Although only recommended for sCAP patients, nearly all pneumonia patients received blood cultures, a quarter of nonsevere patients received sputum cultures, and treatment with broad-spectrum agents was widespread, indicating fertile ground for antimicrobial and diagnostic stewardship programs.

Although pneumonia is the eighth leading cause of death in the United States (1), the prevalence and implications of severe pneumonia are not well described. Early studies of severe pneumonia did not use a uniform definition for this condition. Most commonly, it was described as pneumonia requiring ICU level care (2), although this is fraught with imprecision because ICU admission criteria vary widely across hospitals (3) making inter-hospital comparisons difficult to interpret. Starting with the 2007 edition (4) and continuing with the 2019 edition (5), the American Thoracic Society (ATS)/Infectious Diseases Society of America (IDSA) guidelines for the treatment of community-acquired pneumonia (CAP) define severe CAP (sCAP) as meeting either one major criterion (septic shock with need for vasopressors or respiratory failure requiring mechanical ventilation) or three minor (laboratory and physiologic markers such as uremia or hypothermia) criteria. The guideline gives separate recommendations for severe and nonsevere disease and for those with and without risk factors for infection with resistant organisms (as measured by prior respiratory infection with either methicillin-resistant Staphylococcus aureus (MRSA) or Pseudomonas species or recent hospitalization with use of parenteral antibiotics). The major distinctions between these groups are in recommendations for diagnostic testing and empiric antibiotics. Sputum Gram stain and culture, blood cultures, and testing for pneumococcal and Legionella urinary antigens are reserved for patients with sCAP or those with risk factors for resistant infection with but not for patients with routine CAP. Additionally, while recommended first-line therapy for routine CAP is beta-lactam plus macrolide or quinolone, antibiotics for sCAP include beta-lactam plus macrolide or beta-lactam plus quinolone. Broad-spectrum antibiotics such as vancomycin, linezolid, piperacillin-tazobactam, cefepime, imipenem, or meropenem are reserved for patients with risk factors for resistant organisms.

Because there have been no recent or large-scale studies describing patients with sCAP using the standardized definition, we applied the ATS/IDSA major criteria to a large cohort of patients hospitalized across the United States with pneumonia in order to describe its epidemiology, microbiology, and outcomes. In addition, we sought to outline patterns of clinical practice and management of sCAP including pathogen testing and empiric antimicrobial use, with the goal of providing baseline data for use in policy and guideline development, antimicrobial and diagnostic stewardship programs, and prediction models.

MATERIALS AND METHODS

In this retrospective observational study, we used the Premier Healthcare Database (Premier, Charlotte, NC) to identify patients. This database is derived from an inpatient data service developed for measuring quality and healthcare utilization and is frequently used for healthcare research (6–9). Hospitals that participate in Premier are located in geographically diverse regions of the United States, vary widely in size, include both urban and rural as well as community and academic medical centers and represent approximately 25% of annual inpatient admissions in the United States (7). Data elements collected include sociodemographic information, International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes, hospital and physician information, treatments received, source of admission, and discharge status. Our dataset was drawn from the 177 Premier hospitals that use Safety Surveillor, a Premier software product that manages microbiology data, including cultures and antibiotic sensitivity testing. We excluded hospitals that did not provide microbiology data. Because the dataset excluded patient-identifiers, the Cleveland Clinic Institutional Review Board (IRB) determined that this study did not constitute human subjects research (IRB Number 15–1254).

Patients included in this study were greater than or equal to 18 years old and admitted to any of the 177 hospitals that submitted microbiology data to Premier between July 2010 and June 2015. For patients with multiple pneumonia hospitalizations during the study period, we randomly selected one admission. We included patients who were assigned a principal diagnosis of pneumonia or a secondary diagnosis of pneumonia paired with a principal diagnosis of respiratory failure, acute respiratory distress syndrome, respiratory arrest, or sepsis. Patients with CAP and those who met the previous definition for healthcare-associated pneumonia (HCAP) were included, while patients with hospital-acquired pneumonia or ventilator-associated pneumonia (10) were excluded. To improve specificity of the diagnosis codes, we required patients to have chest imaging and antibiotic administration within the first 2 days in hospital. Patients were also excluded if there was evidence of ventilator dependence prior to admission, the patient had cystic fibrosis, or if they were transferred to or from another acute care hospital since the treatments and outcomes at the other hospitals could not be verified. We also excluded patients with evidence of an alternate site of infection (e.g., central line-associated bloodstream infection, endocarditis, intra-abdominal infection, abscess, or cellulitis); for complete exclusions, see Figure 1.

Figure 1.

Patient inclusion/exclusion flowchart. Flow chart of patient inclusions and exclusions used to derive the final n of 21,805 patients with severe pneumonia. IMV = invasive mechanical ventilation.

For each hospitalization, we extracted demographic information, principal and secondary diagnoses, comorbidities, treatments, and microbiologic data. Comorbidities were identified from ICD-9-CM secondary diagnosis codes and Diagnosis-Related Groups using Healthcare Cost and Utilization Project Comorbidity Software, Version 3.1 (https://www.healthypeople.gov/2020/data-source/healthcare-cost-and-utilization-project-state-inpatient-databases), based on the work of Elixhauser (11). Combined comorbidity scores were also calculated using the method described by Gagne et al (12). Organ failure was derived from ICD-9-CM codes for these diagnoses, for example, 518.8 for acute respiratory failure. We categorized patients as having sCAP if they met at least one of the IDSA guideline’s two major criteria, that is, those patients who received invasive mechanical ventilation (IMV) or vasopressors on day 0, 1, or 2 of hospitalization. Because the dataset does not contain vital signs or results of hematologic or chemical blood test results, we were unable to determine patients who may have had sCAP defined by minor criteria. We used guideline-defined categories (5) for risk of infection with resistant organisms, including either hospitalization accompanied by treatment with parenteral antibiotics or a positive respiratory tract culture with MRSA or Pseudomonas within the prior 90 days.

Data are summarized by means and sds of relatively symmetrically distributed quantitative variables, medians, and quartiles of the asymmetrically distributed variables (cost and length of hospital stay), and category counts and proportions of totals for categorical variables. Differences between baseline characteristics, utilization and results of testing, organisms detected, treatments, and outcomes of patients with severe and nonsevere pneumonia were tested for statistical significance by t test for relatively symmetrically distributed quantitative variables, Wilcoxon rank-sum test for asymmetrically distributed quantitative variables, and Pearson chi-square test (without continuity correction) for categorical variables. Since almost all inferential comparisons between patients with severe and nonsevere pneumonia were statistically significant (p < 0.001), for simplicity and readability, only exceptions to this are noted in the tables.

RESULTS

Among 154,799 patients with pneumonia included in the cohort (Fig. 1), 21,805 (14.1%) met at least one major criterion for sCAP. Among those, 71.6% received IMV, 60.3% received vasopressors, and 31.8% received both. Compared with patients with nonsevere pneumonia, those who had sCAP were younger and more likely to have a principal diagnosis of sepsis or respiratory failure, to have risk factors associated with the concept of HCAP, to receive a diagnosis of organ failure of the respiratory, cardiovascular, renal, hepatic, or central nervous systems, and to have higher combined organ failure scores (Supplemental Table 1, http://links.lww.com/CCM/G1000). Patients with sCAP more often had immunosuppression, diabetes, fluid, and electrolyte disorders and anemia than those with nonsevere pneumonia. The proportion of sCAP patients admitted to an ICU (89.5%) was far higher than in the nonsevere group (17.3%). Unadjusted outcomes between the two groups also differed substantially with worse outcomes among those with sCAP compared with nonsevere. Patients with sCAP incurred higher mortality (22% vs 5%; p < 0.001), longer lengths of stay (8 vs 5 d; p < 0.001), and higher median cost ($20,046 vs $7,543; p < 0.001).

Table 1 shows the microbiologic and molecular tests performed. While over 90% of patients in both groups had blood cultures obtained, the rate of positivity in the severe group was double that in the nonsevere (10.0% vs 4.5%; p < 0.001). Respiratory culture sampling was twice as likely to occur among patients with severe disease than in those with nonsevere disease, and a pathogen was identified more often among the former than among the latter (34.2% vs 21.1%; p < 0.001). Testing for influenza occurred in approximately one of five patients in each group, and although uncommonly positive, the test was twice as likely to be positive among patients with severe disease. Compared with patients with nonsevere pneumonia, patients with sCAP were slightly more likely to have urinary antigen tests for pneumococcus or Legionella performed, and these tests were more often positive for pneumococcus but not Legionella. Nasal swabs for S. aureus were collected more commonly on patients with sCAP than those with nonsevere pneumonia, likely a reflection of policies regarding collection of these specimens for ICU patients, although there was no difference in the rate of positivity between the two groups.

TABLE 1.

Cultures and Tests Performed on Patients With Severe Pneumonia and Nonsevere Pneumonia (n = 154,799)

Type of Test	Nonsevere		Severe
	Test (%)	Positive Test (Row %)	Test (%)	Positive Test (Row %)
Blood culture	122,402 (92.0)	5,460 (4.5)	20,641 (94.7)	2,055 (10.0)
Respiratory culture	34,108 (25.6)	7,212 (21.1)	11,770 (54.0)	4,029 (34.2)
Influenza test	28,532 (21.5)	471 (1.7)	4,539 (20.8)	153 (3.4)
UAT for Streptococcus	20,264 (15.2)	1,365 (6.7)	3,621 (16.6)	340 (9.4)
UAT for Legionella	27,429 (20.6)	403 (1.5)	5,069 (23.2)	77 (1.5)
Nasal swab for Staphylococcus	7,586 (5.7)	1,391 (18.3)	3,674 (16.8)	725 (19.7)
PCR for Legionella	95 (0.07)	1 (1.1)	40 (0.18)	1 (2.5)
PCR for Chlamydia	2,023 (1.5)	2 (0.10)	505 (2.3)	0 (0.0)
PCR for Mycoplasma	2,183 (1.6)	15 (0.69)	559 (2.6)	0 (0.0)

PCR = polymerase chain reaction, UAT = urine antigen test.

All comparisons of test utilization between severe and nonsevere pneumonia except for influenza test (p = 0.033), and all comparisons of test prevalence except influenza test (also p = 0.033), UAT for Legionella (p = 0.79), nasal swab for Staphylococcus (p = 0.075), PCR for Legionella (p = 0.51), PCR for Chlamydia (p = 0.99), and PCR for Mycoplasma (p = 0.049) are statistically significant (p < 0.001) by Pearson χ² test.

To understand the potential variance among hospitals in the utilization of diagnostic testing and antibiotic prescribing, we repeated several analyses at the hospital level rather than the patient level. These analyses are important to understand variability that may be driven by local factors such as hospital culture, teaching status, or affiliation with a parent organization that dictates clinical protocols rather than relying on individual clinician judgment and can be a marker of the use of quality measures in routine practice. Supplemental Table 2 (http://links.lww.com/CCM/H2) shows hospital-level utilization of diagnostic testing, demonstrating a significantly higher rate of obtaining blood and respiratory cultures for patients with severe pneumonia than those with nonsevere, but no differences for other diagnostic tests including urinary antigen tests and polymerase chain reaction. Supplemental Table 3 (http://links.lww.com/CCM/H3) shows hospital-level antibiotic utilization for patients with severe versus nonsevere pneumonia, demonstrating significantly greater use of broad-spectrum antimicrobials for patients with severe pneumonia.

The most commonly recovered microorganisms by the site of culture (blood or respiratory) are shown in Table 2. Overall, each organism was detected infrequently, but growth of every organism was more common among patients with sCAP than among those with nonsevere pneumonia. Methicillin-sensitive S. aureus (MSSA) and MRSA were the most frequently isolated organisms among sCAP patients overall, driven largely by respiratory cultures, whereas Streptococcus pneumoniae was the most commonly isolated organism among patients with nonsevere CAP. Interestingly, when considering the source of culture only, S. pneumoniae was the most commonly isolated organism from blood cultures regardless of disease severity, and MSSA, MRSA, and Pseudomonas species were most common when considering only respiratory cultures. Overall, however, compared with those with nonsevere disease, patients with sCAP were nearly three times as likely to have an organism that was resistant to IDSA guideline-recommended empiric antibiotics for CAP (10.0% vs 3.1%; p < 0.001), predominantly among organisms recovered from respiratory cultures.

TABLE 2.

Most Common Organisms Recovered From Patients With Severe Pneumonia and Nonsevere Pneumonia by Source of Cultures (Blood Cultures vs Respiratory Cultures) (n = 154,799)

Organism	Nonsevere			Severe
	Blood Culture (n = 122,402)	Respiratory Culture (n = 34,108)	Overall (n = 125,089)	Blood Culture (n = 20,641)	Respiratory Culture (n = 11,770)	Overall (n = 21,230)
	Count (column %)
Methicillin-sensitive Staphylococcus aureus	709 (0.58)	1,681 (4.9)	2,312 (1.8)	326 (1.6)	1,068 (9.1)	1,279 (6.0)
Methicillin-resistant S. aureus	509 (0.42)	1,386 (4.1)	1,823 (1.5)	224 (1.1)	829 (7.0)	961 (4.5)
Streptococcus pneumoniae	1,801 (1.5)	1,091 (3.2)	2,810 (2.2)	489 (2.4)	520 (4.4)	916 (4.3)
Escherichia coli	1,162 (0.95)	515 (1.5)	1,672 (1.3)	426 (2.1)	353 (3.0)	739 (3.5)
Pseudomonas species	211 (0.17)	1,655 (4.9)	1,844 (1.5)	114 (0.55)	720 (6.1)	786 (3.7)
Klebsiella species	378 (0.31)	691 (2.0)	1,054 (0.84)	166 (0.80)	454 (3.9)	591 (2.8)
Other bacteria	502 (0.41)	641 (1.9)	1,134 (0.91)	217 (1.1)	379 (3.2)	577 (2.7)
Haemophilus influenzae	114 (0.09)	266 (0.78)	379 (0.30)	56 (0.27)	147 (1.2)	195 (0.92)
Proteus mirabilis	168 (0.14)	172 (0.50)	338 (0.27)	112 (0.54)	134 (1.1)	239 (1.1)
Serratia species	20 (0.02)	148 (0.43)	165 (0.13)	18 (0.09)	96 (0.82)	108 (0.51)
Any bacteria resistant to ceftriaxone + macrolide	948 (0.77)	1,948 (5.7)	2,828 (2.3)	477 (2.3)	1,188 (10.1)	1,545 (7.3)
Any bacteria resistant to ceftriaxone + quinolone	497 (0.41)	942 (2.8)	1,396 (1.1)	254 (1.2)	693 (5.9)	883 (4.2)
Any bacteria resistant to quinolone alone	909 (0.74)	1,319 (3.9)	2,173 (1.7)	479 (2.3)	932 (7.9)	1,316 (6.2)
Any bacteria resistant to any community- acquired pneumonia therapy option	1,534 (1.3)	2,461 (7.2)	3,891 (3.1)	795 (3.9)	1,524 (12.9)	2,131 (10.0)

Detection of every organism class listed was statistically significantly higher (p < 0.001) among severe than among nonsevere pneumonia patients.

The most commonly received antimicrobial agents among patients with severe and nonsevere pneumonia are shown in Table 3. Patients with severe pneumonia were twice as likely to receive vancomycin or piperacillin-tazobactam compared with patients with nonsevere pneumonia, and less commonly received ceftriaxone or azithromycin, although there was little difference in the rate of use of levofloxacin between the two groups. Supplemental Table 4 (http://links.lww.com/CCM/H4) demonstrates the percentage of sCAP and nonsevere pneumonia patients who had an infection with MRSA or Pseudomonas and also met the CAP guideline’s suggested criteria for risk of infection with a resistant organism. Only 7.1% of these patients had either a hospitalization with parenteral antibiotics in the past 3 months or a positive respiratory culture for MRSA or Pseudomonas species, with a higher percentage in the sCAP group having one of these risk factors than patients with nonsevere pneumonia.

TABLE 3.

Most Common Antibiotics Received on Day 0, 1, or 2 of Hospitalization Among All Patients With Severe and Nonsevere Pneumonia

Factor	Total (n = 154,799)	Nonsevere (n = 132,994)	Severe (n = 21,805)
	Count (column %)
Vancomycin	59,152 (38.2)	44,668 (33.6)	14,484 (66.4)
Piperacillin/tazobactam	40,903 (26.4)	31,413 (23.6)	9,490 (43.5)
Levofloxacin	57,619 (37.2)	48,896 (36.8)	8,723 (40.0)
Ceftriaxone	75,190 (48.6)	67,372 (50.7)	7,818 (35.9)
Azithromycin	69,073 (44.6)	62,126 (46.7)	6,947 (31.9)
Cefepime	15,720 (10.2)	12,740 (9.6)	2,980 (13.7)
Any antipseudomonal excluding quinolones	63,751 (41.2)	49,579 (37.3)	14,172 (65.0)
Any antipseudomonal including quinolones	97,565 (63.0)	80,160 (60.3)	17,405 (79.8)
Any antiviral	3,112 (2.0)	2,335 (1.8)	777 (3.6)

Comparisons between severe and not severe pneumonia are statistically significant (p < 0.001) by Pearson χ² test for each antibiotic class.

DISCUSSION

In this large national sample from 177 U.S. hospitals of 154,799 patients with pneumonia, 14.1% met at least one of the major criteria for sCAP outlined in the CAP guideline, and their hospital outcomes were commensurately worse than among patients with nonsevere disease. Despite nearly ubiquitous blood culture sampling, the yield of these was only 10% in sCAP and half of that in nonsevere. Among those patients who had positive cultures, patients with sCAP had higher rates of positivity overall but the organisms differed by source of the culture, with S. aureus accounting for the greatest percentage of positive respiratory cultures and S. pneumoniae accounting for a higher percentage among positive blood cultures. The prevalence of resistance to first-line CAP regimens was 13.1%, largely as a result of resistance among respiratory samples, which often cannot be used to distinguish between infection and colonization. In spite of low rates of resistance, patients with sCAP had double the rate of treatment with broad-spectrum antibiotics including piperacillin-tazobactam and cefepime, and nearly two-thirds of sCAP patients received vancomycin.

Our findings build on those of prior studies on the microbiology and outcomes of patients with sCAP. In one of the largest studies of critically ill CAP patients, Woodhead et al (13) described 17,869 patients in Great Britain in 2006 and found inhospital mortality of nearly 50%. Rello et al (14) assessed the microbiology and outcomes of patients admitted to two ICUs in Spain and found that microbiologic investigation resulted in changes in management in nearly half of the patients. Like many studies of sCAP, these studies were hampered by the lack of a standardized definition of “severe” and relied upon ICU admission as the sole marker. Liapikou et al (15) used the ATS/IDSA guideline’s major and minor criteria to predict ICU admission and mortality and found that these criteria had high sensitivity and specificity for predicting these outcomes. Our study adds to the literature by providing the largest analysis to date of the demographics, outcomes, tests, microbiologic results, and treatments of patients with sCAP using the CAP guideline’s major criteria. Our data demonstrate widespread use of blood and sputum cultures and broad-spectrum antibiotics regardless of severity, and informs our understanding of the major causative pathogens among both severe and nonsevere cases. These data provide robust evidence to assist in the development of models, prediction tools and guidance for the management of patients with sCAP.

Our study has several important limitations. We were not able to capture patients who had minor criteria for severe pneumonia, however, inclusion of patients who required either mechanical ventilation or vasopressors likely captured the sickest of patients. In a meta-analysis, Marti et al (16) showed that the minor criteria have a specificity of 90.5% and an area under the curve of 0.85% to predict severe pneumonia that required ICU admission, but no study to date has compared the sensitivity and specificity of minor criteria to major criteria. Cultures in this retrospective study were obtained in real world settings during the routine care of patients, and thus, we were unable to determine reasons why some patients did or did not have cultures performed, which could have introduced selection bias into our comparisons of organisms and clinician selections of antibiotics. Some patients could have received outpatient antibiotics that could have decreased culture yields, although we have somewhat mitigated this problem by excluding patients who were transferred from other hospitals. We were unable to accurately determine which patients may have had advanced directives to limit care, and thus, some mortality may not have been preventable. We presented only the empiric antimicrobials given to patients within the first 3 days of hospitalization because this is the focus of the guidelines, however, there may have been unmeasured effects due to later de-escalation or changes in antibiotics that may have impacted outcomes. The impact of de-escalation of antibiotics on outcomes in CAP is beyond the scope of this study. Because pneumonia is ultimately a clinical diagnosis, it is possible that some of the patients could have had an alternative cause for their symptoms such as congestive heart failure, however, all patients had a diagnosis of pneumonia, and thus, the clinicians managing these patients believed they were treating pneumonia. To counter this, we only included patients if they also had chest imaging and received antibiotics, although the exact diagnostic accuracy of this approach has not been quantified. The identification of comorbidities and organ failure by ICD-9 codes may over or underestimate the actual rates of these diagnoses in comparison to chart review. Finally, the data set encompasses a time period when the concept of HCAP was still in use and reflect the associated antimicrobial prescribing patterns. Thus, these data represent a baseline and further study is needed to assess the impact of abandonment of the HCAP concept on antimicrobial prescribing practices, which was not the focus of this study.

Our findings of relatively low rates of culture positivity even among patients with severe disease support the need for new advanced diagnostic tools that could provide rapid information on the microbial etiology of pneumonia whether bacterial or viral. In addition, greater use of invasive lower respiratory tract cultures in patients with sCAP, especially those that require intubation, could increase the yield and specificity of respiratory cultures in this population. Although the guidelines recommend Legionella and pneumococcal urinary antigen testing in sCAP, we found low rates of utilization of these tests, which could reduce broad-spectrum antimicrobial utilization when positive, and should be included in care sets and protocols for sCAP patients. Although viral infection accounts for 24–40% of CAP (17), very few patients in our study underwent even the most widely available viral test (influenza). Increase utilization of viral testing in CAP could support antibiotic de-escalation. Our finding of higher rates of resistant organisms among sCAP patients, as well as more sCAP patients meeting risk factors for resistant organisms also supports the recommendation in the latest guideline to obtain blood and sputum cultures in this population. Finally, our findings support the guideline recommendation that broad-spectrum antibiotics be reserved for patients with risk factors for resistant organisms. Only 4% of sCAP patients had MRSA or Pseudomonas in cultures and only 7% of patients who grew MRSA or Pseudomonas had one of the two guideline-defined risk factors for these multidrug-resistant organisms, yet 66% of sCAP patients received vancomycin and 65 % received an antipseudomonal agent.

CONCLUSIONS

In summary, using a guideline-recommended standard definition, we found that compared with patients with nonsevere disease, those with sCAP had higher mortality, length of stay, costs, clinical deterioration, and greater receipt of empiric broad-spectrum antibiotics. Blood and sputum cultures were more often positive among patients with sCAP, with S. aureus and S. pneumoniae found most commonly in respiratory and blood specimens, respectively. Although recommended for all patients with sCAP, only 16% and 23% of severe pneumonia hospitalizations included pneumococcal and Legionella urinary antigen testing, respectively, indicating a marked under-utilization of these important diagnostic modalities. The CAP guideline outlines a relatively narrow subset of pneumonia patients who should undergo blood and sputum cultures and be treated with empiric broad-spectrum antimicrobial agents. In contrast, we found that nearly all pneumonia patients received blood cultures, a quarter of nonsevere patients received sputum cultures, and treatment with broad-spectrum agents was widespread, indicating an area that is fertile ground for antimicrobial and diagnostic stewardship programs.