Jeffrey Tabas, MD

Data Sources and Searches

We searched the English‐language literature via MEDLINE (1966 through September 2007), MEDLINE‐In Process, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, and ACP Journal Club. Within each of these databases we used keywords and exploded Medical Subject Headings (MeSH) to produce the following search strategy: blood culture(s) (keyword), bacteriological techniques (MeSH), blood [microbiology] (MeSH), bacteremia [microbiology or drug therapy] (MeSH), or diagnostic tests, routine (MeSH) combined with pneumonia (keyword), pneumonia (MeSH), lower respiratory tract infection(s) (keyword), or community‐acquired infections (MeSH). To maximize capture of BC or bacteremia studies with subgroups of CAP patients we added the following search strategy: explode microbiological techniques [utilization] (MeSH), explode blood specimen collection [utilization] (MeSH), or focus bacteremia [drug therapy] (MeSH). We reviewed the reference lists of all included studies as well as those of important background articles. Finally, we asked experts to evaluate the completeness of our list.

Study Selection

We included studies in which: (1) subjects were adults hospitalized with CAP; (2) BCs were obtained at or near hospital admission; and (3) the effects of BCs on management (change in antibiotic therapy or other effects such as duration of parenteral therapy, length of hospitalization, or level of care) were reported. The first 2 requirements could be satisfied by a subgroup.

From retrieved citations, relevant abstracts were reviewed, and studies with any potential to meet inclusion criteria were chosen for full‐text review. Two authors (N.A., R.S.) independently analyzed each full‐text article to determine inclusion for data analysis. A third author (J.T.) analyzed all included and narrowly excluded articles to confirm the final list of included studies. Disagreements were resolved by discussion.

Data Extraction

For the included studies, 2 authors (N.A., K.A.) independently abstracted the following data using a standardized collection instrument: study design and setting, inclusion and exclusion criteria, number of hospitalized CAP patients in whom BCs were obtained, empiric antibiotic regimens, number of true‐positive and false‐positive BCs, bacteria isolated in true‐positive BCs, BC‐directed antibiotic narrowing, BC‐directed antibiotic broadening ultimately associated with a resistant organism, and any other management effects reported. Narrowing refers to coverage of fewer organisms, while broadening refers to coverage of a larger or different spectrum of organisms.

If a study included patients not meeting our selection criteria, our analysis was limited to the subset of patients meeting criteria. We also analyzed each study to determine whether a subgroup of severely ill patients was reported separately and whether such a group benefited from BCs. The 2 authors independently repeated all data abstraction to confirm accuracy. We attempted to contact authors for clarification when needed.

Data Synthesis

Data were synthesized by compilation of characteristic summary tables. In the primary analysis, the proportion of positive BCs (both true and false) and the frequency of BC‐directed changes in antimicrobial therapy (narrowing, or broadening ultimately associated with a resistant organism) were determined and reported for each study and then described as an aggregate range. This compilation required studies to provide a particular denominatorthe number of patients in whom BCs were performed. If a study did not do so, it was described separately in the secondary analysis, where we also assessed the cost of BCs as well as the impact of BCs in critically ill patients and on outcomes other than antibiotic change. Heterogeneity of subject inclusion and exclusion criteria and empiric antibiotic use were summarized qualitatively. Two authors (N.A., R.S.) assessed each study's quality.

Search Results

Our electronic database search yielded 3236 citations. From this list and the supplementary search of references, we reviewed 607 abstracts; of these, we selected 73 articles for full‐text review, and 15 were included in the final analysis (Figure 1). One study was narrowly excluded because it largely included CAP patients that had already been admitted to the hospital and failed an empiric antibiotic trial before BCs were obtained.13

Figure 1

Study Characteristics

Fifteen studies with a total of 3898 patients evaluated BC‐directed management changes in adults admitted with CAP.11, 1427 However, 2 of these, involving only patients with bacteremic pneumococcal CAP, by design could not report the number of patients that had BCs done; thus they were not included in the primary analysis.16, 25

The 13 studies amenable to the primary analysis (Table 1) all had an observational cohort design; 6 were prospective11, 18, 20, 24, 26, 27 and 7 were retrospective.14, 15, 17, 19, 2123 Sample size varied from 52 to 760 patients. Settings included university and community hospitals in the U.S. and 4 other countries, with patient enrollment spanning the years 19882003 (publication dates 19912007).

Included patients were usually required to have clinical features of pneumonia and a confirmatory chest x‐ray. Treating physicians were required to obtain BCs (either by study or hospital protocol) in only 3 studies14, 22, 24 and in a subgroup of another study;11 otherwise the performance of BCs was left to physician discretion.

Nine studies excluded patients who were immunocompromised,11, 15, 17, 18, 2224, 26, 27 a label that was often incompletely defined. Otherwise, exclusion criteria were variable. Notably, only 3 studies excluded patients admitted to the intensive care unit (ICU),11, 18, 24 while 6 excluded patients with cancer15, 17, 2224, 27 and 6 excluded either nursing home residents15, 17, 22, 26 or the elderly (de facto exclusion of most nursing home residents).24, 27

Empiric antibiotic regimens, where reported, were predominantly cephalosporin plus macrolide combinations in 4 studies,17, 2224 fluoroquinolones in 3 studies,11, 14, 26 and penicillin or 1 of its derivatives in 1 study.27

Concerning the 2 studies not included in the primary analysis, the one by Waterer et al.25 was a retrospective review of all cases of pneumococcal bacteremia (n = 74) associated with an admission diagnosis of CAP (N = 1805) in a US urban hospital over a 3‐year period. The one by Chang et al.16 was a retrospective case‐control study of 288 randomly‐selected, immunocompetent Medicare patients with bacteremic pneumococcal CAP who survived to discharge. They were matched 1:1 with blood and sputum culture‐negative controls to study the rate of fluoroquinolone use at discharge in the 2 groups.

Study Findings

Summary of Findings

Our systematic review of the literature finds that BCs rarely alter empiric antibiotic therapy in adults hospitalized with community‐acquired pneumonia. Even when there is a change in treatment it usually is not of the type most likely to impact patient outcome, which is antibiotic broadening ultimately associated with a resistant organism. In the 13 studies that could quantify this effect, it occurred in only 0% to 1% of cases in which BCs were obtained. Antibiotic narrowing occurred in 0% to 3% of cases, with physicians often choosing not to narrow antibiotics when BC results suggested that they could do so.

Limits on BC Utility

Polymicrobial Infection

If positive BCs in CAP mostly reveal antibiotic‐sensitive pathogens, one may infer that at least they lead to narrowing of therapy. However, the studies reviewed here reveal that this usually does not happen.

One explanation for this reluctance to narrow antibiotics is that CAP is often a polymicrobial disease. When rigorous serologic testing is done, multiple pathogens are found in up to 40% of cases.39 The occult copathogen is frequently an intracellular one and thus cannot be detected by BCs. Though the evidence for empirically treating these atypical organisms is mixed,40, 41 expert guidelines recommend doing so,12 and guideline‐concordant antibiotic therapy in CAP is associated with lower mortality.42 Even in bacteremic pneumococcal CAP, monotherapy is associated with higher mortality.4346 Thus, stopping antibiotic coverage of atypical pathogens in response to BCs alone might not always be appropriate.

Limitations of the Review

Our search strategy was designed to be sensitive and included backup methods such as searching article references and querying experts. Nevertheless, we may have missed studies, especially if there were small eligible subgroups or if determining management effects was not a primary purpose. We chose not to measure instances of antibiotic broadening that were not associated with a resistant organism, though in unusual cases (eg, Pseudomonas bacteremia) this effect of BCs may be useful.

The methodologies of the included studies were adequate to measure the key outcomes with reasonable validity. Biases were evident, though they occurred both for and against BC utility.

Eligibility varied across studies, and most investigations excluded immunocompromised or other high‐risk patient groups, which could have biased results against BC utility. However, results of these studies were consistent with those that included all patients with CAP, suggesting the degree of bias was probably small. Still, given this concern, it would be prudent not to generalize the findings of this review to immunocompromised patients. Moreover, although the critically ill and those who today would be classified as having healthcare‐associated pneumonia (HCAP)nursing home residents, the recently hospitalized, and hemodialysis patientswere included in most studies, their numbers were small, and these groups were not analyzed separately; thus, the results might not be generalizable to these populations either. Finally, the reported studies, which enrolled patients through 2003, do not reflect more recent increases in the prevalence of resistant pathogens, such as MRSA, in the community.

BCs as a Quality Measure

The adoption of BCs as a quality measure was largely predicated on the widely‐cited study by Meehan et al.,4 which showed an association between BC obtainment and reduced mortality. This study, which associated processes of care with hard outcomes such as mortality, was limited by uncontrolled confounders, including variation in hospital quality.53 A more recent study of pneumonia processes of care found no association between BC collection and mortality.54 Another study often cited to support BC use, by Arbo and Snydman,55 showed that positive BCs were associated with changes in antibiotic therapy, but it included very few pneumonia patients and did not describe results for them separately.

The inclusion of BC acquisition in 2 quality measures in the NHQM guidelines for pneumonia impacts the clinical practice of hospitals and physicians, which may be rated and reimbursed differentially based on their compliance with such measures. One of the quality measures requires BCs in patients admitted to the ICU. The other requires that ED BCs for pneumonia, if obtained, be drawn before antibiotics are given.6

The studies we reviewed are not specific to these quality measures, but are relevant to them. With regard to the first measure, all but 3 studies included patients admitted to the ICU and found BCs to be of minimal benefit overall. Our subgroup analysis of severely ill patients was unrevealing. The ICU measure is tentative in its validity, but it is not unreasonable given that these patients have a life‐threatening infection and may be at risk for bacteremia with resistant pathogens.12

The second measure, though perhaps simply seeking to maximize the potential for BCs to turn positive, depends for its validity on BCs being useful in a large proportion of patients with CAP. Though we cannot exclude the possibility that BCs benefit certain subsets of patients, such as those who are immunocompromised or have HCAP, our findings do not support obtaining BCs in all or even most adults hospitalized with CAP. This conclusion is reflected in the 2007 Infectious Diseases Society of America/American Thoracic Society management guidelines for CAP, which state than BCs are optional except for patients with severe pneumonia, some immunocompromised states, and particular radiographic abnormalities.12

With such data and guidelines in mind, a physician seeking to minimize treatment delays in a patient with pneumonia may give antibiotics early in the ED course (the basis of another quality measure) without obtaining BCs. If she later determines that the patient is particularly high‐risk for bacteremia or a resistant pathogen, should she be discouraged from ordering BCs? Experts specifically state that BCs, even after antibiotics, are warranted for such a patient.12

With the scope of medical practice captured in quality measures being so narrow, having 2 measures based on a test with such limited benefit is itself questionable.

Data Sources and Searches

We searched the English‐language literature via MEDLINE (1966 through September 2007), MEDLINE‐In Process, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, and ACP Journal Club. Within each of these databases we used keywords and exploded Medical Subject Headings (MeSH) to produce the following search strategy: blood culture(s) (keyword), bacteriological techniques (MeSH), blood [microbiology] (MeSH), bacteremia [microbiology or drug therapy] (MeSH), or diagnostic tests, routine (MeSH) combined with pneumonia (keyword), pneumonia (MeSH), lower respiratory tract infection(s) (keyword), or community‐acquired infections (MeSH). To maximize capture of BC or bacteremia studies with subgroups of CAP patients we added the following search strategy: explode microbiological techniques [utilization] (MeSH), explode blood specimen collection [utilization] (MeSH), or focus bacteremia [drug therapy] (MeSH). We reviewed the reference lists of all included studies as well as those of important background articles. Finally, we asked experts to evaluate the completeness of our list.

Study Selection

We included studies in which: (1) subjects were adults hospitalized with CAP; (2) BCs were obtained at or near hospital admission; and (3) the effects of BCs on management (change in antibiotic therapy or other effects such as duration of parenteral therapy, length of hospitalization, or level of care) were reported. The first 2 requirements could be satisfied by a subgroup.

From retrieved citations, relevant abstracts were reviewed, and studies with any potential to meet inclusion criteria were chosen for full‐text review. Two authors (N.A., R.S.) independently analyzed each full‐text article to determine inclusion for data analysis. A third author (J.T.) analyzed all included and narrowly excluded articles to confirm the final list of included studies. Disagreements were resolved by discussion.

Data Extraction

For the included studies, 2 authors (N.A., K.A.) independently abstracted the following data using a standardized collection instrument: study design and setting, inclusion and exclusion criteria, number of hospitalized CAP patients in whom BCs were obtained, empiric antibiotic regimens, number of true‐positive and false‐positive BCs, bacteria isolated in true‐positive BCs, BC‐directed antibiotic narrowing, BC‐directed antibiotic broadening ultimately associated with a resistant organism, and any other management effects reported. Narrowing refers to coverage of fewer organisms, while broadening refers to coverage of a larger or different spectrum of organisms.

If a study included patients not meeting our selection criteria, our analysis was limited to the subset of patients meeting criteria. We also analyzed each study to determine whether a subgroup of severely ill patients was reported separately and whether such a group benefited from BCs. The 2 authors independently repeated all data abstraction to confirm accuracy. We attempted to contact authors for clarification when needed.

Data Synthesis

Data were synthesized by compilation of characteristic summary tables. In the primary analysis, the proportion of positive BCs (both true and false) and the frequency of BC‐directed changes in antimicrobial therapy (narrowing, or broadening ultimately associated with a resistant organism) were determined and reported for each study and then described as an aggregate range. This compilation required studies to provide a particular denominatorthe number of patients in whom BCs were performed. If a study did not do so, it was described separately in the secondary analysis, where we also assessed the cost of BCs as well as the impact of BCs in critically ill patients and on outcomes other than antibiotic change. Heterogeneity of subject inclusion and exclusion criteria and empiric antibiotic use were summarized qualitatively. Two authors (N.A., R.S.) assessed each study's quality.

Search Results

Our electronic database search yielded 3236 citations. From this list and the supplementary search of references, we reviewed 607 abstracts; of these, we selected 73 articles for full‐text review, and 15 were included in the final analysis (Figure 1). One study was narrowly excluded because it largely included CAP patients that had already been admitted to the hospital and failed an empiric antibiotic trial before BCs were obtained.13

Figure 1

Study Characteristics

Fifteen studies with a total of 3898 patients evaluated BC‐directed management changes in adults admitted with CAP.11, 1427 However, 2 of these, involving only patients with bacteremic pneumococcal CAP, by design could not report the number of patients that had BCs done; thus they were not included in the primary analysis.16, 25

The 13 studies amenable to the primary analysis (Table 1) all had an observational cohort design; 6 were prospective11, 18, 20, 24, 26, 27 and 7 were retrospective.14, 15, 17, 19, 2123 Sample size varied from 52 to 760 patients. Settings included university and community hospitals in the U.S. and 4 other countries, with patient enrollment spanning the years 19882003 (publication dates 19912007).

Included patients were usually required to have clinical features of pneumonia and a confirmatory chest x‐ray. Treating physicians were required to obtain BCs (either by study or hospital protocol) in only 3 studies14, 22, 24 and in a subgroup of another study;11 otherwise the performance of BCs was left to physician discretion.

Nine studies excluded patients who were immunocompromised,11, 15, 17, 18, 2224, 26, 27 a label that was often incompletely defined. Otherwise, exclusion criteria were variable. Notably, only 3 studies excluded patients admitted to the intensive care unit (ICU),11, 18, 24 while 6 excluded patients with cancer15, 17, 2224, 27 and 6 excluded either nursing home residents15, 17, 22, 26 or the elderly (de facto exclusion of most nursing home residents).24, 27

Empiric antibiotic regimens, where reported, were predominantly cephalosporin plus macrolide combinations in 4 studies,17, 2224 fluoroquinolones in 3 studies,11, 14, 26 and penicillin or 1 of its derivatives in 1 study.27

Concerning the 2 studies not included in the primary analysis, the one by Waterer et al.25 was a retrospective review of all cases of pneumococcal bacteremia (n = 74) associated with an admission diagnosis of CAP (N = 1805) in a US urban hospital over a 3‐year period. The one by Chang et al.16 was a retrospective case‐control study of 288 randomly‐selected, immunocompetent Medicare patients with bacteremic pneumococcal CAP who survived to discharge. They were matched 1:1 with blood and sputum culture‐negative controls to study the rate of fluoroquinolone use at discharge in the 2 groups.

Study Findings

Summary of Findings

Our systematic review of the literature finds that BCs rarely alter empiric antibiotic therapy in adults hospitalized with community‐acquired pneumonia. Even when there is a change in treatment it usually is not of the type most likely to impact patient outcome, which is antibiotic broadening ultimately associated with a resistant organism. In the 13 studies that could quantify this effect, it occurred in only 0% to 1% of cases in which BCs were obtained. Antibiotic narrowing occurred in 0% to 3% of cases, with physicians often choosing not to narrow antibiotics when BC results suggested that they could do so.

Limits on BC Utility

Polymicrobial Infection

If positive BCs in CAP mostly reveal antibiotic‐sensitive pathogens, one may infer that at least they lead to narrowing of therapy. However, the studies reviewed here reveal that this usually does not happen.

One explanation for this reluctance to narrow antibiotics is that CAP is often a polymicrobial disease. When rigorous serologic testing is done, multiple pathogens are found in up to 40% of cases.39 The occult copathogen is frequently an intracellular one and thus cannot be detected by BCs. Though the evidence for empirically treating these atypical organisms is mixed,40, 41 expert guidelines recommend doing so,12 and guideline‐concordant antibiotic therapy in CAP is associated with lower mortality.42 Even in bacteremic pneumococcal CAP, monotherapy is associated with higher mortality.4346 Thus, stopping antibiotic coverage of atypical pathogens in response to BCs alone might not always be appropriate.

Limitations of the Review

Our search strategy was designed to be sensitive and included backup methods such as searching article references and querying experts. Nevertheless, we may have missed studies, especially if there were small eligible subgroups or if determining management effects was not a primary purpose. We chose not to measure instances of antibiotic broadening that were not associated with a resistant organism, though in unusual cases (eg, Pseudomonas bacteremia) this effect of BCs may be useful.

The methodologies of the included studies were adequate to measure the key outcomes with reasonable validity. Biases were evident, though they occurred both for and against BC utility.

Eligibility varied across studies, and most investigations excluded immunocompromised or other high‐risk patient groups, which could have biased results against BC utility. However, results of these studies were consistent with those that included all patients with CAP, suggesting the degree of bias was probably small. Still, given this concern, it would be prudent not to generalize the findings of this review to immunocompromised patients. Moreover, although the critically ill and those who today would be classified as having healthcare‐associated pneumonia (HCAP)nursing home residents, the recently hospitalized, and hemodialysis patientswere included in most studies, their numbers were small, and these groups were not analyzed separately; thus, the results might not be generalizable to these populations either. Finally, the reported studies, which enrolled patients through 2003, do not reflect more recent increases in the prevalence of resistant pathogens, such as MRSA, in the community.

BCs as a Quality Measure

The adoption of BCs as a quality measure was largely predicated on the widely‐cited study by Meehan et al.,4 which showed an association between BC obtainment and reduced mortality. This study, which associated processes of care with hard outcomes such as mortality, was limited by uncontrolled confounders, including variation in hospital quality.53 A more recent study of pneumonia processes of care found no association between BC collection and mortality.54 Another study often cited to support BC use, by Arbo and Snydman,55 showed that positive BCs were associated with changes in antibiotic therapy, but it included very few pneumonia patients and did not describe results for them separately.

The inclusion of BC acquisition in 2 quality measures in the NHQM guidelines for pneumonia impacts the clinical practice of hospitals and physicians, which may be rated and reimbursed differentially based on their compliance with such measures. One of the quality measures requires BCs in patients admitted to the ICU. The other requires that ED BCs for pneumonia, if obtained, be drawn before antibiotics are given.6

The studies we reviewed are not specific to these quality measures, but are relevant to them. With regard to the first measure, all but 3 studies included patients admitted to the ICU and found BCs to be of minimal benefit overall. Our subgroup analysis of severely ill patients was unrevealing. The ICU measure is tentative in its validity, but it is not unreasonable given that these patients have a life‐threatening infection and may be at risk for bacteremia with resistant pathogens.12

The second measure, though perhaps simply seeking to maximize the potential for BCs to turn positive, depends for its validity on BCs being useful in a large proportion of patients with CAP. Though we cannot exclude the possibility that BCs benefit certain subsets of patients, such as those who are immunocompromised or have HCAP, our findings do not support obtaining BCs in all or even most adults hospitalized with CAP. This conclusion is reflected in the 2007 Infectious Diseases Society of America/American Thoracic Society management guidelines for CAP, which state than BCs are optional except for patients with severe pneumonia, some immunocompromised states, and particular radiographic abnormalities.12

With such data and guidelines in mind, a physician seeking to minimize treatment delays in a patient with pneumonia may give antibiotics early in the ED course (the basis of another quality measure) without obtaining BCs. If she later determines that the patient is particularly high‐risk for bacteremia or a resistant pathogen, should she be discouraged from ordering BCs? Experts specifically state that BCs, even after antibiotics, are warranted for such a patient.12

With the scope of medical practice captured in quality measures being so narrow, having 2 measures based on a test with such limited benefit is itself questionable.

Data Sources and Searches

We searched the English‐language literature via MEDLINE (1966 through September 2007), MEDLINE‐In Process, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, and ACP Journal Club. Within each of these databases we used keywords and exploded Medical Subject Headings (MeSH) to produce the following search strategy: blood culture(s) (keyword), bacteriological techniques (MeSH), blood [microbiology] (MeSH), bacteremia [microbiology or drug therapy] (MeSH), or diagnostic tests, routine (MeSH) combined with pneumonia (keyword), pneumonia (MeSH), lower respiratory tract infection(s) (keyword), or community‐acquired infections (MeSH). To maximize capture of BC or bacteremia studies with subgroups of CAP patients we added the following search strategy: explode microbiological techniques [utilization] (MeSH), explode blood specimen collection [utilization] (MeSH), or focus bacteremia [drug therapy] (MeSH). We reviewed the reference lists of all included studies as well as those of important background articles. Finally, we asked experts to evaluate the completeness of our list.

Study Selection

We included studies in which: (1) subjects were adults hospitalized with CAP; (2) BCs were obtained at or near hospital admission; and (3) the effects of BCs on management (change in antibiotic therapy or other effects such as duration of parenteral therapy, length of hospitalization, or level of care) were reported. The first 2 requirements could be satisfied by a subgroup.

From retrieved citations, relevant abstracts were reviewed, and studies with any potential to meet inclusion criteria were chosen for full‐text review. Two authors (N.A., R.S.) independently analyzed each full‐text article to determine inclusion for data analysis. A third author (J.T.) analyzed all included and narrowly excluded articles to confirm the final list of included studies. Disagreements were resolved by discussion.

Data Extraction

For the included studies, 2 authors (N.A., K.A.) independently abstracted the following data using a standardized collection instrument: study design and setting, inclusion and exclusion criteria, number of hospitalized CAP patients in whom BCs were obtained, empiric antibiotic regimens, number of true‐positive and false‐positive BCs, bacteria isolated in true‐positive BCs, BC‐directed antibiotic narrowing, BC‐directed antibiotic broadening ultimately associated with a resistant organism, and any other management effects reported. Narrowing refers to coverage of fewer organisms, while broadening refers to coverage of a larger or different spectrum of organisms.

If a study included patients not meeting our selection criteria, our analysis was limited to the subset of patients meeting criteria. We also analyzed each study to determine whether a subgroup of severely ill patients was reported separately and whether such a group benefited from BCs. The 2 authors independently repeated all data abstraction to confirm accuracy. We attempted to contact authors for clarification when needed.

Data Synthesis

Data were synthesized by compilation of characteristic summary tables. In the primary analysis, the proportion of positive BCs (both true and false) and the frequency of BC‐directed changes in antimicrobial therapy (narrowing, or broadening ultimately associated with a resistant organism) were determined and reported for each study and then described as an aggregate range. This compilation required studies to provide a particular denominatorthe number of patients in whom BCs were performed. If a study did not do so, it was described separately in the secondary analysis, where we also assessed the cost of BCs as well as the impact of BCs in critically ill patients and on outcomes other than antibiotic change. Heterogeneity of subject inclusion and exclusion criteria and empiric antibiotic use were summarized qualitatively. Two authors (N.A., R.S.) assessed each study's quality.

Search Results

Our electronic database search yielded 3236 citations. From this list and the supplementary search of references, we reviewed 607 abstracts; of these, we selected 73 articles for full‐text review, and 15 were included in the final analysis (Figure 1). One study was narrowly excluded because it largely included CAP patients that had already been admitted to the hospital and failed an empiric antibiotic trial before BCs were obtained.13

Figure 1

Study Characteristics

Fifteen studies with a total of 3898 patients evaluated BC‐directed management changes in adults admitted with CAP.11, 1427 However, 2 of these, involving only patients with bacteremic pneumococcal CAP, by design could not report the number of patients that had BCs done; thus they were not included in the primary analysis.16, 25

The 13 studies amenable to the primary analysis (Table 1) all had an observational cohort design; 6 were prospective11, 18, 20, 24, 26, 27 and 7 were retrospective.14, 15, 17, 19, 2123 Sample size varied from 52 to 760 patients. Settings included university and community hospitals in the U.S. and 4 other countries, with patient enrollment spanning the years 19882003 (publication dates 19912007).

Included patients were usually required to have clinical features of pneumonia and a confirmatory chest x‐ray. Treating physicians were required to obtain BCs (either by study or hospital protocol) in only 3 studies14, 22, 24 and in a subgroup of another study;11 otherwise the performance of BCs was left to physician discretion.

Nine studies excluded patients who were immunocompromised,11, 15, 17, 18, 2224, 26, 27 a label that was often incompletely defined. Otherwise, exclusion criteria were variable. Notably, only 3 studies excluded patients admitted to the intensive care unit (ICU),11, 18, 24 while 6 excluded patients with cancer15, 17, 2224, 27 and 6 excluded either nursing home residents15, 17, 22, 26 or the elderly (de facto exclusion of most nursing home residents).24, 27

Empiric antibiotic regimens, where reported, were predominantly cephalosporin plus macrolide combinations in 4 studies,17, 2224 fluoroquinolones in 3 studies,11, 14, 26 and penicillin or 1 of its derivatives in 1 study.27

Concerning the 2 studies not included in the primary analysis, the one by Waterer et al.25 was a retrospective review of all cases of pneumococcal bacteremia (n = 74) associated with an admission diagnosis of CAP (N = 1805) in a US urban hospital over a 3‐year period. The one by Chang et al.16 was a retrospective case‐control study of 288 randomly‐selected, immunocompetent Medicare patients with bacteremic pneumococcal CAP who survived to discharge. They were matched 1:1 with blood and sputum culture‐negative controls to study the rate of fluoroquinolone use at discharge in the 2 groups.

Study Findings

Summary of Findings

Our systematic review of the literature finds that BCs rarely alter empiric antibiotic therapy in adults hospitalized with community‐acquired pneumonia. Even when there is a change in treatment it usually is not of the type most likely to impact patient outcome, which is antibiotic broadening ultimately associated with a resistant organism. In the 13 studies that could quantify this effect, it occurred in only 0% to 1% of cases in which BCs were obtained. Antibiotic narrowing occurred in 0% to 3% of cases, with physicians often choosing not to narrow antibiotics when BC results suggested that they could do so.

Limits on BC Utility

Polymicrobial Infection

If positive BCs in CAP mostly reveal antibiotic‐sensitive pathogens, one may infer that at least they lead to narrowing of therapy. However, the studies reviewed here reveal that this usually does not happen.

One explanation for this reluctance to narrow antibiotics is that CAP is often a polymicrobial disease. When rigorous serologic testing is done, multiple pathogens are found in up to 40% of cases.39 The occult copathogen is frequently an intracellular one and thus cannot be detected by BCs. Though the evidence for empirically treating these atypical organisms is mixed,40, 41 expert guidelines recommend doing so,12 and guideline‐concordant antibiotic therapy in CAP is associated with lower mortality.42 Even in bacteremic pneumococcal CAP, monotherapy is associated with higher mortality.4346 Thus, stopping antibiotic coverage of atypical pathogens in response to BCs alone might not always be appropriate.

Limitations of the Review

Our search strategy was designed to be sensitive and included backup methods such as searching article references and querying experts. Nevertheless, we may have missed studies, especially if there were small eligible subgroups or if determining management effects was not a primary purpose. We chose not to measure instances of antibiotic broadening that were not associated with a resistant organism, though in unusual cases (eg, Pseudomonas bacteremia) this effect of BCs may be useful.

The methodologies of the included studies were adequate to measure the key outcomes with reasonable validity. Biases were evident, though they occurred both for and against BC utility.

Eligibility varied across studies, and most investigations excluded immunocompromised or other high‐risk patient groups, which could have biased results against BC utility. However, results of these studies were consistent with those that included all patients with CAP, suggesting the degree of bias was probably small. Still, given this concern, it would be prudent not to generalize the findings of this review to immunocompromised patients. Moreover, although the critically ill and those who today would be classified as having healthcare‐associated pneumonia (HCAP)nursing home residents, the recently hospitalized, and hemodialysis patientswere included in most studies, their numbers were small, and these groups were not analyzed separately; thus, the results might not be generalizable to these populations either. Finally, the reported studies, which enrolled patients through 2003, do not reflect more recent increases in the prevalence of resistant pathogens, such as MRSA, in the community.

BCs as a Quality Measure

The adoption of BCs as a quality measure was largely predicated on the widely‐cited study by Meehan et al.,4 which showed an association between BC obtainment and reduced mortality. This study, which associated processes of care with hard outcomes such as mortality, was limited by uncontrolled confounders, including variation in hospital quality.53 A more recent study of pneumonia processes of care found no association between BC collection and mortality.54 Another study often cited to support BC use, by Arbo and Snydman,55 showed that positive BCs were associated with changes in antibiotic therapy, but it included very few pneumonia patients and did not describe results for them separately.

The inclusion of BC acquisition in 2 quality measures in the NHQM guidelines for pneumonia impacts the clinical practice of hospitals and physicians, which may be rated and reimbursed differentially based on their compliance with such measures. One of the quality measures requires BCs in patients admitted to the ICU. The other requires that ED BCs for pneumonia, if obtained, be drawn before antibiotics are given.6

The studies we reviewed are not specific to these quality measures, but are relevant to them. With regard to the first measure, all but 3 studies included patients admitted to the ICU and found BCs to be of minimal benefit overall. Our subgroup analysis of severely ill patients was unrevealing. The ICU measure is tentative in its validity, but it is not unreasonable given that these patients have a life‐threatening infection and may be at risk for bacteremia with resistant pathogens.12

The second measure, though perhaps simply seeking to maximize the potential for BCs to turn positive, depends for its validity on BCs being useful in a large proportion of patients with CAP. Though we cannot exclude the possibility that BCs benefit certain subsets of patients, such as those who are immunocompromised or have HCAP, our findings do not support obtaining BCs in all or even most adults hospitalized with CAP. This conclusion is reflected in the 2007 Infectious Diseases Society of America/American Thoracic Society management guidelines for CAP, which state than BCs are optional except for patients with severe pneumonia, some immunocompromised states, and particular radiographic abnormalities.12

With such data and guidelines in mind, a physician seeking to minimize treatment delays in a patient with pneumonia may give antibiotics early in the ED course (the basis of another quality measure) without obtaining BCs. If she later determines that the patient is particularly high‐risk for bacteremia or a resistant pathogen, should she be discouraged from ordering BCs? Experts specifically state that BCs, even after antibiotics, are warranted for such a patient.12

With the scope of medical practice captured in quality measures being so narrow, having 2 measures based on a test with such limited benefit is itself questionable.