AMERICAN THORACIC SOCIETY
Guidelines for the Management of Adults with Community-acquired Pneumonia
Diagnosis, Assessment of Severity, Antimicrobial Therapy, and Prevention

	EXECUTIVE SUMMARY

TOP EXECUTIVE SUMMARY INTRODUCTION ETIOLOGY OF COMMUNITY-ACQUIRED... PATIENT STRATIFICATION DIAGNOSTIC STUDIES OF PATIENTS... THE ROLE OF CLINICAL... THE DECISION TO HOSPITALIZE... DEFINITION OF SEVERE COMMUNITY-... TREATMENT GUIDELINES FOR... DURATION OF THERAPY, RESPONSE... MANAGEMENT OF PATIENTS WHO... VACCINATION RECOMMENDATIONS FOR... SUMMARY AND RECOMMENDATIONS REFERENCES

This document is an update of the original 1993 statement on community-acquired pneumonia, incorporating new information about bacteriology, patient stratification, diagnostic evaluation, antibiotic therapy, and prevention. The statement includes a summary of the available literature, as well as evidence-based recommendations for patient management, developed by a multidisciplinary group composed of pulmonary, critical care, general internal medicine, and infectious disease specialists.

The sections of this document are as follows: an overview of the purpose of our efforts and the methodology used to collect and grade the available data; a review of the likely etiologic pathogens causing community-acquired pneumonia (CAP), including a discussion of drug-resistant Streptococcus pneumoniae (DRSP); a proposed approach to patient stratification for the purpose of predicting the likely etiologic pathogens of different patient populations with CAP; a summary of available and recommended diagnostic studies; suggestions on how to define the need for hospitalization and admission to the intensive care unit (ICU) for patients with CAP; guidelines for antibiotic therapy of CAP, including principles of therapy and specific recommendations for each patient category; an approach to the nonresponding patient, as well as a discussion of when to switch to oral therapy and when to discharge an admitted patient with CAP who is responding to initial therapy; and recommendations for the use of pneumococcal and influenza vaccines.

Likely Pathogens and Patient Stratification

All CAP patients fall into one of four groups, each with a list of likely pathogens, and suggested empiric therapy follows from this list (see Figure 1). Stratification is based on an assessment of place of therapy (outpatient, inpatient ward, or intensive care unit), the presence of cardiopulmonary disease, and the presence of "modifying factors" (see Table 1), which include risk factors for DRSP, enteric gram-negatives, and Pseudomonas aeruginosa. Not every patient should be considered as being at risk for infection with DRSP, and clinical risk factors have been defined. The role of enteric gram-negatives in CAP is controversial, but these organisms do not need to be considered unless specific risk factors are present; however, one of these risk factors includes residence in a nursing home, a population that is not excluded from this statement.

View larger version (17K): [in this window] [in a new window]   Figure 1.

For all patients with CAP, pneumococcus is the most common pathogen, and may even account for pneumonia in patients who have no pathogen identified by routine diagnostic testing. Although the incidence of DRSP is increasing, available data show that mortality in CAP is adversely affected by drug-resistant pneumococci only when minimal inhibitory concentration (MIC) values to penicillin are  4 mg/L. The impact of organisms at lower levels of resistance remains uncertain. All patients with CAP could potentially be infected with Chlamydia pneumoniae, Mycoplasma pneumoniae, and Legionella spp. (the "atypical" pathogens), either alone or as part of a mixed infection, and thus all patients should receive therapy to account for this possibility. Although the term "atypical pneumonia" is not an accurate description of the clinical features of CAP, the use of the term "atypical" was retained in this statement to refer to the specific pathogens listed above. When patients with CAP are admitted to the ICU, the organisms responsible include pneumococcus, the "atypical" pathogens (especially Legionella in some series), and enteric gram-negatives. Pseudomonas aeruginosa has been recovered from some patients with severe CAP, but this organism should be considered only when patients have well-identified risk factors present.

Diagnostic Testing

All patients with CAP should have a chest radiograph to establish the diagnosis and the presence of complications (pleural effusion, multilobar disease), although in some outpatient settings, this may be impossible. All outpatients should have a careful assessment of disease severity, but sputum culture and Gram's stain are not required. All admitted patients with CAP should have an assessment of gas exchange (oximetry or arterial blood gas), routine blood chemistry and blood counts, and a collection of two sets of blood cultures. If a drug-resistant pathogen, or an organism not covered by usual empiric therapy, is suspected, sputum culture should be obtained, and Gram's stain should be used to guide interpretation of culture results. In general, sputum Gram's stain cannot be used to focus initial empiric antibiotic therapy, but could be used to broaden initial antibiotic therapy to include organisms found on the Gram's stain that are not covered by the usual initial empiric antibiotic therapy options. Routine serologic testing is not recommended for any population with CAP. For patients with severe CAP, Legionella urinary antigen should be measured, and aggressive efforts at establishing an etiologic diagnosis should be made, including the collection of bronchoscopic samples of lower respiratory secretions in selected patients, although the benefit of such efforts has not been proven.

Admission Decision and Need for ICU Care

The admission decision remains an "art of medicine" decision, and prognostic scoring rules (the Pneumonia Patient Outcomes Research Team [PORT] and British Thoracic Society rules), are adjunctive tools to support, but not replace, this process. In general, hospitalization is needed if patients have multiple risk factors for a complicated course, and these are summarized in this document. Patients may also need to be hospitalized for a variety of nonmedical reasons, and such social factors should also be incorporated into the admission decision process.

Admission to the ICU is needed for patients with severe CAP, defined as the presence of either one of two major criteria, or the presence of two of three minor criteria. The major criteria include need for mechanical ventilation and septic shock; the minor criteria include systolic blood pressure (BP)  90 mm Hg, multilobar disease, and Pa_O2/FI_O2 ratio < 250. Patients who have two of four criteria from the British Thoracic Society rules also have more severe illness and should be considered for ICU admission. These criteria include respiratory rate  30/min, diastolic blood pressure  60 mm Hg, blood urea nitrogen (BUN) > 7.0 mM (> 19.1 mg/dl), and confusion.

Therapy Principles and Recommendations

Patients should initially be treated empirically, based on the likely pathogens for each of the four patient groups (see Tables 2345), although when culture results become available, organism-specific therapy may be possible for some patients. All populations should be treated for the possibility of atypical pathogen infection, and this should be with a macrolide (or tetracycline) alone in outpatients, or an intravenous macrolide alone in inpatients who have no risk factors for DRSP, gram-negatives, or aspiration. For outpatients or non-ICU inpatients with risk factors for these other organisms, therapy should be with either a -lactam/macrolide combination or an antipneumococcal fluoroquinolone alone. Although both regimens appear therapeutically equivalent, particularly among inpatients, in the outpatient treatment of the more complicated patient, an antipneumococcal fluoroquinolone may be more convenient than a -lactam/macrolide combination. All admitted patients should receive their first dose of antibiotic therapy within 8 h of arrival to the hospital. In the ICU-admitted patient, current data do not support the use of an antipneumococcal fluoroquinolone alone, and therapy should be with a -lactam plus either a macrolide or quinolone, using a regimen with two antipseudomonal agents in appropriate, at-risk, patients.

If a -lactam/macrolide combination is used for a patient with risk factors for DRSP, only selected -lactams can be used, and these include oral therapy with cefpodoxime, amoxicillin/clavulanate, high-dose amoxicillin, or cefuroxime; or intravenous therapy with ceftriaxone, cefotaxime, ampicillin/ sulbactam, or high-dose ampicillin. Ceftriaxone can also be given intramuscularly. There are several antibiotics, such as cefepime, imipenem, meropenem, and piperacillin/tazobactam that are generally clinically active against DRSP, but since these agents are also active against P. aeruginosa, they should be reserved for patients with risk factors for this organism.

Clinical Response, Switch to Oral Therapy, and Discharge

Most patients with CAP will have an adequate clinical response within 3 d, and when the patient has met appropriate criteria, switch to oral therapy should be made. Criteria for switch include improvement in cough and dyspnea; afebrile (< 100° F) on two occasions 8 h apart; white blood cell count decreasing; and functioning gastrointestinal tract with adequate oral intake. Even if the patient is febrile, switch therapy can occur, if other clinical features are favorable. If the patient has met criteria for switch, oral therapy can be started and the patient discharged on the same day, if other medical and social factors permit.

For most patients, initial antibiotic therapy should not be changed in the first 72 h, unless there is a marked clinical deterioration. Up to 10% of all CAP patients will not respond to initial therapy, and a diagnostic evaluation is necessary to look for a drug-resistant or unusual (or unsuspected) pathogen, a nonpneumonia diagnosis (inflammatory disease or pulmonary embolus), or a pneumonia complication. This evaluation begins with a careful requestioning about epidemiologic factors that predispose to specific pathogens (see Table 6).

Prevention

Pneumonia can be prevented by the use of pneumococcal and influenza vaccines in appropriate at-risk populations. Smoking cessation should be promoted in all patients, and can also eliminate an important risk factor for CAP.

	INTRODUCTION

TOP EXECUTIVE SUMMARY INTRODUCTION ETIOLOGY OF COMMUNITY-ACQUIRED... PATIENT STRATIFICATION DIAGNOSTIC STUDIES OF PATIENTS... THE ROLE OF CLINICAL... THE DECISION TO HOSPITALIZE... DEFINITION OF SEVERE COMMUNITY-... TREATMENT GUIDELINES FOR... DURATION OF THERAPY, RESPONSE... MANAGEMENT OF PATIENTS WHO... VACCINATION RECOMMENDATIONS FOR... SUMMARY AND RECOMMENDATIONS REFERENCES

Community-acquired pneumonia (CAP) remains a common and serious illness, in spite of the availability of potent new antimicrobials and effective vaccines. In the United States, pneumonia is the sixth leading cause of death, and the number one cause of death from infectious diseases (1, 2). Because pneumonia is not a reportable illness, information about its incidence is based on crude estimates, but it appears that up to 5.6 million cases of community-acquired pneumonia occur annually, and as many as 1.1 million of these require hospitalization (1, 2). In the outpatient setting, the mortality rate of pneumonia remains low, in the range of < 1-5%, but among patients with community-acquired pneumonia who require hospitalization, the mortality rate averages 12% overall, but increases in specific populations, such as those with bacteremia, and those from nursing home settings, and approaches 40% in those who are most ill and who require admission to the intensive care unit (3).

Both the epidemiology and treatment of pneumonia have undergone changes. Pneumonia is increasingly being recognized among older patients and those with comorbidity (coexisting illness) (2, 15, 18, 19, 21). Such illnesses include chronic obstructive lung disease, diabetes mellitus, renal insufficiency, congestive heart failure, coronary artery disease, malignancy, chronic neurologic disease, and chronic liver disease (15). These individuals may become infected with a variety of newly identified, or previously unrecognized, pathogens (5, 14, 17, 27). At the same time, a number of new antimicrobial agents have become available, some with utility for community-acquired pneumonia. Paralleling the improvement in our antibiotic armamentarium has been the evolution of bacterial resistance mechanisms. In the 1990s, many of the common respiratory pathogens have become resistant, in vitro, to widely used antimicrobials. Resistance, by a variety of mechanisms, is being identified with increasing frequency among Streptococcus pneumoniae, Hemophilus influenzae, Moraxella catarrhalis, and a number of enteric gram-negative bacteria (33).

Chronic obstructive pulmonary disease (COPD) is a common illness, affecting up to 15 million persons in the United States, with more than 12 million having a component of illness characterized as chronic bronchitis, and these patients commonly develop community-acquired respiratory infections, including pneumonia (40). COPD is the fourth leading cause of death in the United States, and age-adjusted death rates in this illness have risen, whereas other common causes of death such as heart disease and cerebral vascular disease have fallen (40- 42). This population is prone to frequent acute bronchitic exacerbations of chronic bronchitis (AECB), and bacterial infection is believed to play a role in at least half of these episodes (43). Although most experts agree that antibiotics should not be used in patients with acute bronchitis in the absence of chronic lung disease, the role of antibiotic therapy in AECB is controversial, with some patients receiving such therapy and others not. At the current time, the role of antibiotic therapy in this illness is uncertain, and it remains unclear whether specific subpopulations of patients with AECB can be defined for the purpose of prescribing different therapy to different patients.

In 1993, the American Thoracic Society (ATS) published guidelines for the initial management of community-acquired pneumonia, based on available knowledge and a consensus of experts (44). Since that time, new information has become available in many areas related to this illness, including prognostic scoring to predict mortality, new knowledge of the bacteriology of this illness, and new approaches to providing care in a cost-effective and efficient manner. Since 1993, a number of new antibiotics have been approved for the therapy of CAP, in at least four different drug classes. At the same time, in vitro antibiotic resistance among the organisms causing CAP has become increasingly prevalent, and the clinical relevance of resistance is beginning to be understood.

Goals of This Document

This document is a revision of the initial CAP guidelines, intended to update and expand on the original statement, by including more recent information as well as by covering new areas such as pneumonia prevention and the importance of drug-resistant organisms. It includes not only elements from the original ATS CAP guidelines, but also takes into account the recommendations from the more recently published guidelines of the Infectious Diseases Society of America (IDSA, Alexandria, VA) and the newly published Canadian CAP document (45, 46). The discussion is limited to the apparently immunocompetent patient with community-acquired pneumonia, because this represents the population most commonly encountered. However, patients with immune suppression due to chronic corticosteroid therapy and due to nonhematologic malignancy (without neutropenia) are commonly treated by many types of physicians, and the approach to these patients is included in this document. The approach to other immunocompromised patients is different, because of the large number of potential etiologic agents for pneumonia in these individuals. Thus the discussion does not deal with the problems of pneumonia in the human immunodeficiency virus (HIV)-infected patient, or in those immunocompromised as a result of myelosuppressive chemotherapy, organ transplantation, or "traditional" immunosuppressive illnesses, such as Hodgkin's disease.

The goal of this statement is to provide a framework for the evaluation and therapy of the patient with community-acquired pneumonia. The most common pathogens have been defined from published studies, and the determination of which diagnostic tests should be obtained routinely has been made on the basis of published data. While organism-directed antimicrobial therapy would be ideal because of reduced costs, reduction in adverse drug reactions, and antibiotic selection pressure, the limitations of our current diagnostic methods force us to rely on empiric antibiotic therapy in most patients with CAP. The approach to such therapy must be based on an assessment of the likelihood that a given pathogen is causing disease in a given patient, a determination guided by information from the literature. The major variables that influence the spectrum of etiologic agents and the initial approach to therapy are the severity of illness at initial presentation, the presence of coexisting illness, and the presence of identified clinical risk factors for drug-resistant and unusual pathogens (Table 1). Patients with severe community-acquired pneumonia have a distinct epidemiology and a somewhat different distribution of etiologic pathogens than do patients with other forms of pneumonia (8, 9, 16, 19, 20, 23, 24, 47, 48). Once empiric therapy has been initiated, other decisions, such as the duration of therapy and the change from parenteral to oral therapy, become relevant. Finally, it is inevitable that empiric therapy will not be successful for all patients, and thus an approach is provided for use if the patient is not responding to the selected regimen.

Methodology Used to Prepare This Document

The development of these guidelines was by a committee composed of pulmonary, critical care, infectious disease, and general internal medicine specialists, in an effort to incorporate a variety of perspectives and to create a statement that was acceptable to a wide range of physicians. The committee originally met as a group, with each individual being assigned a topic for review and presentation to the entire group, during a two-day meeting. Each topic in the guideline was reviewed by more than one committee member, and following presentation of information, the committee discussed the data and formulated its recommendations. Each section of the statement was then prepared by committee members, and a draft document incorporating all sections was written. This document was circulated to the committee for review and modification and then the committee met again in May of 2000 to deliberate on suggested changes. The manuscript was then revised and circulated to the committee for final comment. This final statement represents the results of this process and the opinions of the majority of the committee. For any topic in which there was disagreement, the majority position was adopted.

We used an evidence-based approach for making final recommendations, after review of all available and relevant peer-reviewed studies (collected by literature search and selected by the experts reviewing each topic), published until December 2000. Much of the literature on the etiology, epidemiology, and diagnostic approach to respiratory infections is observational, and only a few therapy trials have been conducted in a prospective randomized fashion. Therefore, in grading the evidence supporting our recommendations, we used the following scale, similar to the approach used in the recently updated Canadian CAP statement (46): Level I evidence comes from well-conducted randomized controlled trials; Level II evidence comes from well-designed, controlled trials without randomization (including cohort, patient series, and case control studies); Level III evidence comes from case studies and expert opinion. Level II studies included any large case series in which systematic analysis of disease patterns and/or microbial etiology was conducted, as well as reports of new therapies that were not collected in a randomized fashion. In some instances therapy recommendations come from antibiotic susceptibility data, without clinical observations, and these constitute Level III recommendations.

While numerous studies detailing the incidence and etiology of pneumonia have been published, all have limitations. The approach used in this statement is based on an evaluation of studies that were long enough to avoid seasonal bias and recent enough to include newly recognized pathogens. Therefore, we reviewed the available literature, emphasizing data from prospective studies of one or more years' duration, reported in the past 15 years, involving adults in North America and elsewhere (3, 37, 38, 47). We focused on studies that included an extensive diagnostic approach to define the etiologic pathogen, and which did not rely on sputum Gram's stain and culture alone for this determination. Most involved hospitalized patients, but a wide spectrum of patients was included, ranging from outpatients to those admitted to an intensive care unit. In some of the studies, patients were receiving antimicrobials at the time of initial diagnostic evaluation, and the committee considered information from such studies of uncertain reliability.

	ETIOLOGY OF COMMUNITY-ACQUIRED PNEUMONIA

TOP EXECUTIVE SUMMARY INTRODUCTION ETIOLOGY OF COMMUNITY-ACQUIRED... PATIENT STRATIFICATION DIAGNOSTIC STUDIES OF PATIENTS... THE ROLE OF CLINICAL... THE DECISION TO HOSPITALIZE... DEFINITION OF SEVERE COMMUNITY-... TREATMENT GUIDELINES FOR... DURATION OF THERAPY, RESPONSE... MANAGEMENT OF PATIENTS WHO... VACCINATION RECOMMENDATIONS FOR... SUMMARY AND RECOMMENDATIONS REFERENCES

Types of Data Reviewed

While a rapid diagnosis is optimal in the management of community-acquired pneumonia, the responsible pathogen is not defined in as many as 50% of patients, even when extensive diagnostic testing is performed (3, 13). No single test is presently available that can identify all potential pathogens, and each diagnostic test has limitations. For example, sputum Gram's stain and culture may be discordant for the presence of Streptococcus pneumoniae, and these tests are also not able to detect frequently encountered pathogens such as Mycoplasma pneumoniae, Chlamydia pneumoniae, Legionella spp., and respiratory viruses (49, 50). In addition, several studies have reported that some patients with CAP can have mixed infection involving both bacterial and "atypical" pathogens. This type of mixed infection may require therapy of all the identified pathogens, but cannot be diagnosed initially with readily available clinical specimens (13, 15, 17). In addition, mixed infection can involve more than one bacterial species, or can involve both a bacterial pathogen and a viral organism (13, 17, 28).

The role of "atypical" pathogens is controversial because the frequency of these organisms is largely dependent on the diagnostic tests and criteria used, and it is uncertain whether these organisms infect along with a bacterial pathogen, or if they cause an initial infection that then predisposes to secondary bacterial infection (13, 17, 28). The very term "atypical" pathogen is potentially misleading since the clinical syndrome caused by these organisms is not distinctive (see below), but in this statement the term "atypical" is used to refer to a group of organisms (Mycoplasma pneumoniae, Chlamydia pneumoniae, Legionella spp.), rather than to a clinical picture. The data supporting the presence of atypical pathogen coinfection (which has varied in frequency from as low as 3% to as high as 40%) have generally been derived by serologic testing, documenting fourfold rises in titers to M. pneumoniae, C. pneumoniae, or Legionella spp., and some of these diagnoses have even been made with single high acute titers (14, 17). Since many of these diagnoses have not involved testing for the surface antigens of these pathogens, or cultures of respiratory secretions, the clinical significance of the serologic data remains uncertain.

In defining the bacteriology of CAP, we examined the likely etiologic pathogens for each patient category, adding new information about the emerging resistance of common CAP pathogens such as pneumococcus, H. influenzae, and M. catarrhalis (33). In addition, there is now an increased awareness of the importance of newly recognized pathogens (such as hantavirus) and of "atypical" pathogens. In most studies, a large number of patients have no defined etiology. This is likely a reflection of a number of factors, including prior treatment with antibiotics, the presence of unusual pathogens that go unrecognized (fungi, Coxiella burnetii), the presence of viral infection, the presence of a noninfectious mimic of CAP, and the presence of pathogens that are currently not identified or recognized.

Organisms Causing CAP in Outpatients

Relatively few studies have been conducted in ambulatory patients with CAP and in this group an unknown diagnosis is present in 40-50% of all patients (11, 12, 30, 31, 51). When a pathogen has been identified, the nature of the organisms has reflected the population studied and the types of diagnostic tests performed. With use of sputum culture, pneumococcus is the most commonly identified pathogen (9-20% of all episodes), while M. pneumoniae is the most common organism (accounting for 13-37% of all episodes) identified when serologic testing is performed (11, 12, 51). Chlamydia pneumoniae has been reported in up to 17% of outpatients with CAP (51). In the outpatient setting, Legionella spp. have also been seen, with rates varying from 0.7 to 13% of all patients (30). The incidence of viral infection is variable, but in one series was identified in 36% of patients (30). The incidence of gram-negative infection in ambulatory patients is difficult to define from currently available studies, but the complexity of the population that is currently treated out of the hospital is increasing, and many of these patients have well-identified risk factors for colonization of the respiratory tract by gram-negative bacilli, a common predisposing factor to pneumonia with these pathogens (52).

Organisms Causing CAP in Non-ICU-Hospitalized Patients

On the basis of a review of 15 published studies from North America, over 3 decades in primarily hospitalized patients, Bartlett and Mundy (22) concluded that S. pneumoniae was the most commonly identified pathogen (20-60% of all episodes), followed by H. influenzae (3-10% of all episodes), and then by Staphylococcus aureus, enteric gram-negatives, Legionella, M. pneumoniae, C. pneumoniae, and viruses (up to 10% of episodes for each of these latter agents). In addition, some patients (3-6%) have pneumonia due to aspiration. In all studies, an etiologic agent was not found in 20-70% of patients (4, 5, 8, 13, 14, 18, 21). For many years, patients with an unknown diagnosis were assumed to have the same distribution of pathogens as those with an established diagnosis, since the outcomes in both groups were similar, but one study of hospitalized patients suggested that many patients without a known diagnosis actually have pneumococcal infection (53).

In several studies of hospitalized patients with CAP, there has been a high incidence of atypical pathogen infection, primarily M. pneumoniae and C. pneumoniae among those outside the ICU, while the incidence of Legionella infection has been low in patients who are not admitted to the ICU (8, 13, 14, 15, 16, 17). The incidence of infection with these "atypical" organisms has been as high as 40-60% of all admitted patients, often as part of a mixed infection, but the findings have not been corroborated by all investigators (13, 17, 54). This high incidence was identified primarily by serologic testing that included single high acute titers as well as a 4-fold rise between acute and convalescent titers, but the serologic criteria and diagnostic tests used for these organisms are not standardized, and include the use of IgG and IgM titers. When atypical pathogens have been identified, they have not been confined to the population of young and healthy patients, but rather have been found in patients of all age groups (14). Even in series that do not identify a high incidence of atypical pathogens, they are often part of a mixed infection when they are identified (13, 54). The importance of mixed infection is also uncertain, with some investigators reporting that coinfection with bacterial and atypical pathogens leads to a more complicated course than monomicrobial infection, while others report no impact on the clinical course (28, 54). On the basis of these data, it is difficult to define the importance of these organisms and the need for specific therapy. However, several outcome studies show that both inpatients and outpatients have a less complicated clinical course if a macrolide is used as part of the therapy regimen, or if a quinolone is used alone (55).

Enteric gram-negative bacteria are not common in CAP, but may be present in up to 10% of non-ICU-hospitalized patients. They have been found most commonly in those who have underlying comorbid illness (particularly COPD) on previous oral antibiotic therapy, in those coming from nursing homes, and those with hematologic malignancy or immune suppression (the latter not being covered in this statement) (8, 15, 16, 18, 19). In one study, enteric gram-negatives were identified in 9% of patients, and in 11% of all pathogens, and the presence of any of the following comorbidities was associated with an increased risk of infection (odds ratio 4.4) with these organisms: cardiac illness, chronic lung disease, renal insufficiency, toxic liver disease, chronic neurologic illness, diabetes, and malignancy active within the last year (15). Although the incidence of P. aeruginosa infection is not high in most patients with CAP, this organism was found in 4% of all CAP patients with an established etiologic diagnosis (14, 15). There is still controversy about the true incidence of gram-negative infection in patients with CAP, since diagnostic testing that involves sputum culture cannot always distinguish between colonization by these organisms and true infection. The incidence of gram-negative infection is not as high in all admitted patients with CAP, but rises among those admitted to the ICU, as discussed below (8, 16, 20, 47).

Organisms Causing CAP in Hospitalized Patients Requiring ICU Admission

While gram-negative aerobic organisms have been identified with an increased frequency in patients with CAP requiring intensive care, the most common organisms in patients falling into this category are pneumococcus, Legionella, and H. influenzae, with some series reporting S. aureus as a common pathogen (8, 9, 16, 23). In addition, atypical pathogens such as C. pneumoniae and M. pneumoniae can lead to severe illness, and in at least one study, these organisms were more common than Legionella in causing severe CAP (16). Overall, up to 10% of admitted patients with CAP are brought to the ICU, and pneumococcus is present in up to one-third of all patients (8, 9, 16, 20). Among patients admitted to the ICU, organisms such as P. aeruginosa have been identified, particularly in individuals with underlying bronchiectasis (8, 16, 20, 47). In this population, the Enterobacteriaceae have been found in up to 22% of patients, and up to an additional 10-15% of ICU patients in some series have infection with P. aeruginosa (20, 23, 47). In all of these series, 50-60% of patients with severe CAP have an unknown etiology, and the failure to define a pathogen in these patients has not been associated with a different outcome than if a pathogen is identified (20, 24).

Drug-resistant Pneumococcus in CAP

The emergence of DRSP is an increasingly common problem in the United States and elsewhere, with more than 40% of all pneumococci falling into this category by current in vitro definitions of resistance (35, 59-60a). Controversy continues, however, about the clinical relevance of in vitro resistance in the absence of meningitis, and whether the problem, as currently defined, requires new therapeutic approaches or whether the presence of resistance influences the outcome of CAP (33, 34, 37, 38, 59, 60). The current definitions of resistance include "intermediate-level" resistance with penicillin MIC values of 0.12-1.0 µg/ml, while "high-level" resistance is defined as MIC values of  2.0 µg/ml (60). When resistance to penicillin is present, there is often in vitro resistance to other agents, including cephalosporins, macrolides, doxycycline, and trimethoprim/sulfamethoxizole (35, 60a). In one survey, with isolates collected as recently as 1998, when high-level penicillin resistance was present, in vitro resistance to cefotaxime was 42%, to meropenem 52%, to erythromycin 61%, and to trimethoprim/sulfamethoxizole 92% (60a). The newer antipneumococcal fluoroquinolones (gatifloxacin, gemifloxacin, levofloxacin, moxifloxacin, trovafloxacin, and sparfloxacin), the ketolides, and vancomycin are active agents for DRSP, and linezolid, a newly available oxazolidinone, is also active against DRSP. Although quinolone resistant pneumococci have been uncommon, one recent report found that 2.9% of pneumococcal isolates from adults were ciprofloxacin resistant and that 4.1% of isolates with high level penicillin resistance were also quinolone (ciprofloxacin) resistant (61). When ciprofloxacin resistance was present, in vitro resistance to the newer quinolones was also present.

The clinical relevance of DRSP has been debated, but in the absence of meningitis, clinical failure with high-dose -lactam therapy is currently unlikely (60). Using currently defined levels of resistance, most investigators have found no difference in mortality for patients infected with resistant or sensitive organisms, after controlling for comorbid illness, although patients with resistant organisms may have a more prolonged hospital stay (33, 34). One study has shown that suppurative complications (such as empyema) are more common in patients with penicillin-nonsusceptible organisms than in patients with susceptible organisms, even though the majority of patients received apparently adequate therapy (62). In Spain, where the incidence of high-level DRSP is higher than in the United States, the presence of resistance has been reported to cause a rise in mortality, which was not statistically significant (38). In another study of a population with a high incidence of HIV infection, the presence of high-level penicillin resistance (as defined above) was associated with increased mortality, in spite of most patients receiving therapy that appeared to be appropriate (63). A Centers for Disease Control and Prevention (CDC, Atlanta, GA) study has shown that the breakpoint for clinically relevant resistance to penicillin is an MIC value of  4.0 µg/ml (37). At these levels, resistance was associated with increased mortality in patients with invasive disease (primarily bacteremia), provided that patients dying in the first 4 d of therapy were excluded from the analysis (37). When these levels of resistance are suspected (or documented), alternative agents to penicillin should be used, and these are discussed below, although routine therapy with vancomycin is rarely needed (60).

Not all patients in areas with high geographic rates of DRSP are likely to be infected with these organisms, and even in areas with high rates of resistance, organisms isolated from sputum and blood cultures are less commonly resistant than organisms isolated from the upper respiratory tract (64). Identified risk factors for DRSP include age > 65 years (odds ratio [OR], 3.8), alcoholism (OR, 5.2), noninvasive disease (suggesting possibly reduced virulence of resistant organisms) (OR, 4.5), -lactam therapy within 3 mo (OR, 2.8), multiple medical comorbidities, exposure to children in a day care center, and immunosuppressive illness (38, 59, 65, 66). In one study, the effect of age was less clear, with individuals  age 65 yr having an odds ratio of 1.2 (95% confidence interval [CI], 1.0-1.5) corresponding to an incidence of DRSP of 24%, compared with an incidence of 19% in those aged 18-64 yr (60a).

	PATIENT STRATIFICATION

TOP EXECUTIVE SUMMARY INTRODUCTION ETIOLOGY OF COMMUNITY-ACQUIRED... PATIENT STRATIFICATION DIAGNOSTIC STUDIES OF PATIENTS... THE ROLE OF CLINICAL... THE DECISION TO HOSPITALIZE... DEFINITION OF SEVERE COMMUNITY-... TREATMENT GUIDELINES FOR... DURATION OF THERAPY, RESPONSE... MANAGEMENT OF PATIENTS WHO... VACCINATION RECOMMENDATIONS FOR... SUMMARY AND RECOMMENDATIONS REFERENCES

Features Used to Define Patient Subsets

We divided patients into four groups on the basis of place of therapy (outpatient, hospital ward, or intensive care unit); the presence of coexisting cardiopulmonary disease (chronic obstructive pulmonary disease, congestive heart failure); and the presence of "modifying factors," which included the presence of risk factors for drug-resistant pneumococcus, the presence of risk factors for gram-negative infection (including nursing home residence), and the presence of risk factors for Pseudomonas aeruginosa (specifically in patients requiring ICU admission) (Table 1) (15, 19, 27). The history of cigarette smoking was not used to classify patients, since all of the recommended therapy regimens account for H. influenzae, the organism that is more likely to occur in smokers than in nonsmokers. In this approach to stratification, the place of therapy is a reflection of severity of illness, with the need for hospitalization and the need for ICU admission being defined by the criteria described in subsequent sections of this statement.

In the previous version of the ATS CAP guidelines, age was used as a major discriminating factor among patients to define bacterial etiology. This concept has not been corroborated by studies that have shown that age alone, in the absence of comorbid illness, has little impact on the bacterial etiology of CAP (18, 19, 67). As discussed above, the elderly patient can have infection by "atypical" pathogens, and enteric gram-negatives are common primarily in those with comorbid illness (particularly underlying COPD), recent antibiotic therapy, and in patients residing in nursing homes (8, 15). One pathogen whose presence may be impacted by age alone is drug-resistant Streptococcus pneumoniae (DRSP), with several studies showing that age > 65 is, by itself, a specific epidemiologic risk for CAP due to this organism, but is not an independent risk factor for other organisms (38, 59).

Risk factors for penicillin and drug-resistant pneumococcus were defined from the literature, and are summarized in Table 1. Risk factors for enteric gram-negatives include residence in a nursing home, underlying cardiopulmonary disease, multiple medical comorbidities, and recent antibiotic therapy (15, 18, 19, 52). The risk factors for P. aeruginosa include the presence of any of the following: structural lung disease such as bronchiectasis, corticosteroid therapy (> 10 mg of prednisone per day), broad-spectrum antibiotic therapy for > 7 d in the past month, malnutrition, and leukopenic immune suppression (the latter is not included in this statement) (8, 15, 16, 68).

Patient Subsets

Using these factors, the four patient groups were defined (Tables 2345 and Figure 1) as the following:

I. Outpatients with no history of cardiopulmonary disease, and no modifying factors (Table 2)

II. Outpatients with cardiopulmonary disease (congestive heart failure or COPD) and/or other modifying factors (risk factors for DRSP or gram-negative bacteria) (Table 3)

III. Inpatients, not admitted to the ICU, who have the following (Table 4):

    a. Cardiopulmonary disease, and/or other modifying factors (including being from a nursing home)

    b. No cardiopulmonary disease, and no other modifying factors

IV.  ICU-admitted patients who have the following (Table 5):

    a. No risks for Pseudomonas aeruginosa

    b. Risks for Pseudomonas aeruginosa

For each group, results from available studies were combined to identify the most common pathogens associated with pneumonia and an attempt was made to rank the incidence of pathogens broadly, but a precise numeric incidence or percentage was not included. Patients with nursing home- acquired pneumonias were included, with the realization that this population is unique, and that a knowledge of local (institution-specific) epidemics and antibiotic susceptibility patterns is necessary to choose optimal empiric therapy. However, the following pathogens are recognized more frequently in nursing home patients than in patients with the same coexisting illnesses who are residing in the community: methicillin-resistant S. aureus (MRSA), enteric gram-negative bacteria, Mycobacterium tuberculosis, and certain viral agents (i.e., adenovirus, respiratory syncytial virus [RSV], and influenza) (18, 69). A miscellaneous group is included in each of Tables 2345 and represents organisms that were present in about 1% of patients in these studies, or pathogens that have been otherwise reported to occur in this setting.

Specific Pathogens for Each Patient Subset

The most common pathogens in Group I, that is, outpatients with no cardiopulmonary disease and no risks for DRSP or gram-negatives, include S. pneumoniae, M. pneumoniae, C. pneumoniae, and respiratory viruses (Level II evidence). Miscellaneous pathogens include Legionella sp. (usually a more severe illness), M. tuberculosis, and endemic fungi. Although H. influenzae can be seen in this group of patients, it is a particular concern if the patient has a history of cigarette smoking. This population has a mortality rate of < 1-5% (10,72).

For patients in Group II, the presence of cardiopulmonary disease (congestive heart failure or COPD), or the presence of risk factors for DRSP (including age > 65 yr) or gram-negatives (including being from a nursing home), changes the likely pathogens. Although pneumococcus remains the most likely pathogen, resistance to penicillin and other agents (macrolides, trimethoprim/sulfamethoxizole) is more likely, and this should be considered in antibiotic selection (below). In addition, if the patient is from a nursing home, then aerobic gram-negative infection is possible and can include the Enterobacteriaceae such as Escherichia coli, or Klebsiella spp., and even P. aeruginosa (if bronchiectasis is present) (Level II evidence). Also, in this population, aspiration with anaerobes should be considered in the presence of poor dentition and if the patient has a history of neurologic illness, impaired consciousness, or a swallowing disorder. Less common pathogens include Moraxella catarrhalis, Legionella sp., Mycobacterium sp., and endemic fungi. Mortality in this setting is also < 5%, but as many as 20% of patients initially treated as outpatients may require hospitalization (72).

When the patient is hospitalized, there are usually risks for DRSP and enteric gram-negatives, or underlying cardiopulmonary disease, and these factors influence the likely pathogens (Group IIIa). These patients are at risk for infection with pneumococcus, H. influenzae, atypical pathogens (alone or as a mixed infection), as well as enteric gram-negatives such as the Enterobacteriaceae, and also a polymicrobial bacterial flora including anaerobes associated with aspiration (if risk factors are present). All admitted patients are also at risk for M. tuberculosis and endemic fungi, but these are less commonly identified than the other organisms listed above. Tuberculosis is a particular concern in patients who have been born in foreign countries with high rates of endemic illness, in the alcoholic, and in the elderly who reside in nursing homes. Mortality rates reported for these patients ranged from 5 to 25%, and most of the deaths occurred within the first 7 d (3, 10). If, however, the admitted patient has no cardiopulmonary disease, and no risks for DRSP or gram-negatives (Group IIIb), then the most likely pathogens are S. pneumoniae, H. influenzae, M. pneumoniae, C. pneumoniae, viruses, and possibly Legionella sp. (Level II evidence).

In some studies of admitted patients with CAP, the etiology may be polymicrobial. The incidence of "mixed" infection, usually a bacterial pathogen and an "atypical" pathogen, varies from < 10% to up to 40% (10, 12, 13, 17, 28). Atypical pathogens have been frequently identified in admitted patients of all age groups. The difference in relative incidence of "mixed" infections may relate to how aggressively investigators collected both acute and convalescent titers, and to the type of criteria used to define the presence of these pathogens (single-titer versus 4-fold rise in titers) (13, 17).

Severe community-acquired pneumonia (defined below) has been separated from cases of less severe pneumonia requiring hospitalization, because of the high mortality rate (up to 50%) and the need for immediate recognition of patients with this severity of illness (8, 16, 10, 24, 47, 48). Although severe pneumonia was defined differently by the various investigators, a practical approach defines all patients admitted to the ICU because of respiratory infection as having severe illness (see below) (48). The pathogens most frequently identified among patients with severe pneumonia (Group IVa) include S. pneumoniae, Legionella sp., H. influenzae, enteric gram-negative bacilli, S. aureus, M. pneumoniae, respiratory tract viruses, and a group of miscellaneous pathogens (C. pneumoniae, M. tuberculosis, and endemic fungi) (8, 9, 16, 19, 20, 23, 47). In one study, the incidence of Legionella sp. in patients with severe CAP decreased over time, in one hospital, but was replaced by other atypical pathogens, such as C. pneumoniae and M. pneumoniae (16). There has been some debate about whether P. aeruginosa can lead to severe CAP, and although this organism has been reported in some studies (in 1.5-5% of such patients), the committee felt that this pathogen should be considered only when specific risk factors are present (Group IVb) (8, 16, 20, 47) (Level III evidence). These risks include chronic or prolonged (> 7 d within the past month) broad-spectrum antibiotic therapy, or the presence of bronchiectasis, malnutrition, or diseases and therapies associated with neutrophil dysfunction (such as > 10 mg of prednisone per day). Undiagnosed HIV infection has been identified as a risk factor for CAP due to P. aeruginosa (73). The frequency of S. aureus as a severe CAP pathogen is also variable, being present in anywhere from 1 to 22% of all patients. Risks for infection with this organism include recent influenza infection, diabetes, and renal failure (74).

	DIAGNOSTIC STUDIES OF PATIENTS WITH COMMUNITY-ACQUIRED PNEUMONIA

TOP EXECUTIVE SUMMARY INTRODUCTION ETIOLOGY OF COMMUNITY-ACQUIRED... PATIENT STRATIFICATION DIAGNOSTIC STUDIES OF PATIENTS... THE ROLE OF CLINICAL... THE DECISION TO HOSPITALIZE... DEFINITION OF SEVERE COMMUNITY-... TREATMENT GUIDELINES FOR... DURATION OF THERAPY, RESPONSE... MANAGEMENT OF PATIENTS WHO... VACCINATION RECOMMENDATIONS FOR... SUMMARY AND RECOMMENDATIONS REFERENCES

The diagnosis of pneumonia should be considered in any patient who has newly acquired respiratory symptoms (cough, sputum production, and/or dyspnea), especially if accompanied by fever and auscultatory findings of abnormal breath sounds and crackles. In a patient with advanced age or an inadequate immune response, pneumonia may present with nonrespiratory symptoms such as confusion, failure to thrive, worsening of an underlying chronic illness, or falling down (21, 71). In these patients, fever may be absent, but tachypnea is usually present, along with an abnormal physical examination of the chest (75). In the initial evaluation of the patient with CAP, the history may on occasion help to identify patients at risk for infection with specific organisms, as outlined in Table 6.

Standard posteroanterior (PA) and lateral chest radiographs are valuable in patients whose symptoms and physical examination suggest the possibility of pneumonia, and every effort should be made to obtain this information. The radiograph can be useful in differentiating pneumonia from other conditions that may mimic it. In addition, the radiographic findings may suggest specific etiologies or conditions which as lung abscess, or tuberculosis. The radiograph can also identify coexisting conditions such as bronchial obstruction or pleural effusion. In some patients, the history and physical examination suggest the presence of pneumonia, but the radiograph is negative. One study has shown that some of these radiographically negative patients do have lung infiltrates if a high-resolution computed tomography (CT) scan of the chest is done (76). However, the clinical relevance of these findings is uncertain, since most studies of CAP have required the presence of a lung infiltrate on a routine chest radiograph to define the presence of pneumonia. Radiography is also useful for evaluating severity of illness by identifying multilobar involvement (below). However, in certain outpatient settings, depending on the time of day and the availability of a radiology facility, it may be difficult to obtain a chest radiograph.

Once the diagnosis of CAP is established, an effort should be made to identify a specific etiologic diagnosis in a timely manner, with focused and appropriate diagnostic testing. However, even with extensive diagnostic testing, most investigators cannot identify a specific etiology for community-acquired pneumonia in up to half, or more, of all patients. If an exact etiology is identified, then therapy can be focused and cost- effective, but this goal needs to be tempered by two findings. First, if diagnostic testing leads to delays in the initiation of appropriate therapy, it may have an adverse outcome. One large Medicare study showed that 30-d CAP mortality was increased when administration of the first dose of antibiotic therapy was delayed more than 8 h from the time of arrival to the hospital (77). Second, since the possibility of coinfection is a consideration, with a bacteria and an atypical pathogen (which may take days or weeks to identify), the value of focused therapy, directed at a rapidly identified bacterial etiology, is uncertain. In fact, in large population studies, treatment that accounted for atypical pathogen coinfection led to a better outcome than treatment that did not account for this possibility (55) (Level II evidence).

One of the most controversial recommendations in the 1993 ATS guidelines for CAP was that a sputum Gram's stain and culture not be performed routinely in all admitted patients (44). Although its value is debated, some experts, including the IDSA consensus group, believe that a properly collected and examined Gram's stain of expectorated sputum is helpful for focusing initial empiric therapy in CAP (45). A lower respiratory tract sample that is not heavily contaminated by oral secretions will typically have fewer than 10 squamous epithelial cells, and > 25 neutrophils per low-power field (45). Studies of the sputum Gram's stain have shown limitations, which include the following: not all patients can provide an adequate sample (either because of an inability to produce a sample, or because the sample is of poor quality), interpretation is observer dependent, atypical pathogens (which are common either singly or as coinfecting agents, as discussed above) cannot be seen, the definition of "positive" varies from study to study, and a positive result for pneumococcus is poorly predictive of the ability to recover that organism from a sputum or blood culture (49, 50). In addition, there are no studies correlating data from Gram's stain of expectorated sputum with cultures of alveolar material in large numbers of patients with community-acquired pneumonia. However, direct staining of sputum (non-Gram stain methods) may be diagnostic for some pulmonary infections including those due to Mycobacterium sp., endemic fungi, Legionella sp. (direct fluorescent antibody staining is required), and Pneumocytis carinii.

If a sputum Gram's stain is used to determine initial therapy, the clinician must decide whether liberal criteria are being used to increase the sensitivity of the test (with a corresponding drop in specificity), or whether more stringent criteria are being used to increase the specificity of the test (with a corresponding drop in sensitivity) (49). The debate about the value of Gram's stain may be less relevant with the advent of the Clinical Laboratory Improvement Act, which has limited the use of this test to laboratory personnel who often interpret the results without knowledge of the clinical scenario and suspected pathogens.

Routine bacterial cultures of sputum often demonstrate pathogenic organisms, but sensitivity and specificity are poor, and findings should be correlated with the predominant organism identified on Gram's stain (50). However, recovery from cultures of organisms that are not usually part of the normal respiratory flora may be meaningful. Specialized cultures for Mycobacterium sp., Legionella sp., and endemic fungi may be valuable in the appropriate clinical circumstance. When drug-resistant pneumococci, other resistant pathogens, or organisms not covered by the usual empiric therapy options (such as S. aureus) are anticipated (particularly if the patient has risk factors, or is receiving antibiotics at the time of admission) sputum culture and sensitivity results may be useful. Viral cultures are not useful in the initial evaluation of patients with community-acquired pneumonia and should not be routinely performed (3). However, in the appropriate season, testing of respiratory secretions for influenza antigens, using rapid detection methods, may be helpful in guiding decisions about the use of new antiviral agents.

Recommended Testing

For the patient with CAP initially managed out of the hospital, diagnostic testing should include a chest radiograph and may include a sputum Gram's stain and culture, if drug-resistant bacteria, or an organism not covered by the usual empiric therapy options, are suspected (Level II evidence). In addition, it is necessary to assess severity of illness, relying on radiographic findings (multilobar pneumonia, pleural effusion), and physical findings (respiratory rate, systolic and diastolic blood pressure, signs of dehydration, and mental status) (78). If the patient has underlying chronic heart or lung disease, then assessment of oxygenation by pulse oximetry may help define the need for hospitalization and supplemental oxygen. Routine laboratory tests (complete blood counts, serum electrolytes, hepatic enzymes, and tests of renal function) are of little value in determining the etiology of pneumonia, but may have prognostic significance and influence the decision to hospitalize. They should be considered in patients who may need hospitalization, and in patients > age 65 yr or with coexisting illness (Level II evidence).

For the patient who is admitted, diagnostic testing should be performed rapidly, avoiding delays in the administration of initial empiric therapy, for the reasons stated above. In addition to a chest radiograph, an admitted patient should have a complete blood count and differential, and routine blood chemistry testing (including glucose, serum sodium, liver and renal function tests, and electrolytes) (Level III evidence). All admitted patients should have oxygen saturation assessed by oximetry. Arterial blood gas should be obtained in any patient with severe illness, or in any patient with chronic lung disease, to assess both the level oxygenation and the degree of carbon dioxide retention. Sputum cultures are recommended if a drug-resistant pathogen, or an organism not covered by usual empiric therapy, is suspected. If a sputum culture is obtained, efforts should be made to collect it prior to antibiotic administration. Any culture result should be correlated with the predominant organism identified on Gram's stain of an appropriate specimen, which should be performed in conjuction with a sputum culture (Level III evidence). Otherwise, in the absence of a culture, a sputum Gram's stain is optional. However, it is the consensus of the majority of the CAP statement committee that if a sputum Gram's stain is used to guide initial therapy, it should be with highly sensitive criteria (any gram-positive diplococci, rather than a predominance of such organisms), with the primary purpose being to visualize a bacterial morphology of an organism that was not anticipated, so that appropriate drugs can be added to the initial antibiotic regimen (e.g., S. aureus, or enteric gram-negatives) (Level III evidence). This conclusion differs from that of the IDSA consensus group, which recommended using Gram's stain to narrow initial empiric therapy in patients with certain organism-specific findings (45). Two sets of blood cultures should be drawn before initiation of antibiotic therapy, and may help to identify the presence of bacteremia and of a resistant pathogen, with the overall yield being approximately 11%, and with S. pneumoniae being the most common pathogen identified by this method (46). Any significant pleural effusion (> 10-mm thickness on lateral decubitus film) or any loculated pleural effusion should be sampled, preferably prior to the initiation of antibiotic therapy, to rule out the possibility of empyema or complicated parapneumonic effusion; however, there are no data showing an outcomes benefit to delaying antibiotic therapy for the purpose of performing a thoracentesis. Pleural fluid examination should include white blood cell count and differential; measurement of protein, glucose, lactate dehydrogenase (LDH) and pH; Gram's stain and acid-fast stain; as well as culture for bacteria, fungi, and mycobacteria.

Serologic testing and cold agglutinin measurements are not useful in the initial evaluation of patients with community-acquired pneumonia and should not be routinely performed (Level II evidence). However, acute and convalescent serologic testing may occasionally be useful for a retrospective confirmation of a suspected diagnosis and may be useful in epidemiologic studies. When Legionella is suspected (patients with severe CAP), measurement of urinary antigen is valuable, being positive in the majority of patients with acute Legionella pneumophila serogroup 1 infection, but the test can remain positive for many months after the acute infection (81, 82). Serial complement-fixing antibody titers may be useful in monitoring patients with extensive coccidioidomycosis, and sputum cultures for endemic fungi should be collected in at-risk patients with the proper epidemiologic history. Although patients known to be immune suppressed are excluded from this statement, HIV testing should be done (after informed consent) in any CAP patient with risk factors and should be considered in any patient aged 15-54 yr who is admitted for CAP (83). Specialized tests that measure microbial antigens by monoclonal antibodies, DNA probes, and polymerase chain reaction amplification are being developed, but have not been shown to be valuable for routine use in patients with CAP.

A number of invasive diagnostic techniques to obtain lower airway specimens, uncontaminated by oropharyngeal flora, have been described (45, 84, 85). These include transtracheal aspiration, bronchoscopy with a protected brush catheter, bronchoalveolar lavage with or without balloon protection, and direct percutaneous fine needle aspiration of the lung. These procedures are not indicated in most patients with community-acquired pneumonia (Level III evidence). It may be desirable to have an early accurate diagnosis in occasional patients who are severely ill, although retrospective data have shown that with severe illness, outcome is not improved by establishing a specific etiologic diagnosis (20, 24). In such patients, bronchoscopy with a protected brush catheter or bronchoalveolar lavage has a reasonable sensitivity and specificity when performed correctly. These procedures carry less risk and are usually more acceptable to patients and physicians than transtracheal aspiration and direct needle aspiration of the lung, although some physicians have special expertise in using ultrathin needles for direct lung aspiration.

Although the committee recommended limited initial diagnostic testing, it is necessary to do more extensive testing in patients whose illness is not resolving in spite of apparently appropriate empiric therapy (see below). As discussed above, this statement is not directed at patients known to be immune suppressed. Such patients require a different and more extensive diagnostic approach that reflects the wide range of potential pathogens in this population.

	THE ROLE OF CLINICAL SIGNS AND SYMPTOMS IN PREDICTING THE MICROBIAL ETIOLOGY OF CAP

TOP EXECUTIVE SUMMARY INTRODUCTION ETIOLOGY OF COMMUNITY-ACQUIRED... PATIENT STRATIFICATION DIAGNOSTIC STUDIES OF PATIENTS... THE ROLE OF CLINICAL... THE DECISION TO HOSPITALIZE... DEFINITION OF SEVERE COMMUNITY-... TREATMENT GUIDELINES FOR... DURATION OF THERAPY, RESPONSE... MANAGEMENT OF PATIENTS WHO... VACCINATION RECOMMENDATIONS FOR... SUMMARY AND RECOMMENDATIONS REFERENCES

The clinical features of CAP (symptoms, signs and radiographic findings) cannot be reliably used to establish the etiologic diagnosis of pneumonia with adequate sensitivity and specificity (Level II evidence). Although, in some circumstances, clinicians can confidently use clinical features to establish a specific etiologic diagnosis, in the majority of cases this is not possible. This relates not only to variations in virulence factors of particular pathogens, but also to the presence of coexisting illnesses, resulting in an overlap of clinical symptoms among various etiologic pathogens.

Typical versus Atypical Pneumonia Syndromes

Originally, the classification of pneumonia into "atypical" and "typical" forms arose from the observation that the presentation and natural history of some patients with pneumonia were different compared with those of patients with pneumococcal infection (86, 87). Some pathogens, such as H. influenzae, S. aureus, and gram-negative enteric bacteria, caused clinical syndromes identical to those produced by S. pneumoniae (88). However, other pathogens caused an "atypical" pneumonia syndrome that was initially attributed to M. pneumoniae (87), but other bacterial and viral agents have been identified that can produce a subacute illness indistinguishable from that caused by M. pneumoniae (89, 90). Some of these agents, however, like Legionella species and influenza, can cause a wide spectrum of illness, ranging from a fulminant life-threatening pneumonia to a more subacute presentation (90). Thus the term "atypical" pneumonia represents a clinical syndrome that includes diverse entities, and has limited clinical value.

The attribution of specific clinical features to an etiologic agent is a common clinical practice, particularly for patients suspected of having pneumonia caused by Legionella species. However, data have cast doubt on the specificity of these observations (4, 15, 18), leading to the conclusion that the diagnosis of Legionella sp. infection could not be made on clinical grounds alone. Other comparative studies involving both pediatric and adult populations, have concluded that an etiologic diagnosis could not be established by clinical criteria alone (91). In addition, no roentgenographic pattern is sufficiently distinctive to allow classification of individual cases (94, 95).

Advanced age and coexisting illness are important factors that affect the clinical presentation of pneumonia. Individuals over the age of 65 yr are particularly at risk for mortality from bacteremic pneumococcal disease, and among the elderly, the expression of common clinical features of pneumonia is often atypical, obscured, or even absent (71, 96, 97). Thus, because host factors are often just as important as the identity of the etiologic pathogen in defining the presenting signs and symptoms of pneumonia, the committee felt that it is not possible to reliably use clinical features, including history, physical examination, and routine laboratory and roentgenographic evaluation, to make a specific etiologic diagnosis of community- acquired pneumonia (Level II evidence).

	THE DECISION TO HOSPITALIZE PATIENTS WITH COMMUNITY-ACQUIRED PNEUMONIA

TOP EXECUTIVE SUMMARY INTRODUCTION ETIOLOGY OF COMMUNITY-ACQUIRED... PATIENT STRATIFICATION DIAGNOSTIC STUDIES OF PATIENTS... THE ROLE OF CLINICAL... THE DECISION TO HOSPITALIZE... DEFINITION OF SEVERE COMMUNITY-... TREATMENT GUIDELINES FOR... DURATION OF THERAPY, RESPONSE... MANAGEMENT OF PATIENTS WHO... VACCINATION RECOMMENDATIONS FOR... SUMMARY AND RECOMMENDATIONS REFERENCES

The initial site-of-care decision is perhaps the single most important clinical decision made by physicians during the entire course of illness for patients with community-acquired pneumonia (CAP). It has a direct bearing on the intensity of laboratory testing, microbiologic evaluation, antibiotic therapy, and costs of treating this illness. The estimated average cost of inpatient care for CAP is U.S. $7,500, compared with U.S. $150-$350 for outpatient care (2, 98).

Defining Risk Factors for a Complicated Course of CAP to Determine the Need for Admission

Studies have identified a series of risk factors that increase either the likelihood of death or the risk of a complicated course for community-acquired pneumonia (10, 72). When multiple risk factors coexist, the committee believed that hospitalization should be strongly considered (Level II evidence). The decision to hospitalize is not necessarily a commitment to long-term inpatient care but, rather, a decision that certain patients should be observed closely until it is clear that therapy can be safely continued out of the hospital. The admission decision may also be influenced by the availability of outpatient support services (home nursing, home intravenous therapy), and alternative sites for care (subacute care facilities). Since criteria for admission have not been uniformly applied by clinicians, studies have reported wide geographic variation in hospital admission rates for CAP (99). In addition, in some studies physicians have overestimated the risk of death for patients with CAP, leading to unnecessary admissions, while in other studies they have failed to recognize patients as being severely ill at the time of initial evaluation (72, 80).

Risk factors associated with an increased risk of death or a complicated course have been identified, and in one study, admitted patients had a mean of five risk factors present (57). All of these studies were observational, and the risk factors for an adverse outcome were defined only in hospitalized patients who had received antibiotics and supportive care for their illness, and the risk factors have not been studied in large numbers of outpatients. These factors include a number of features listed below, and those with an asterisk (*) are factors that have been identified to predict mortality in the PORT prediction rule model (72):

1. Age over 65 yr

2. Presence of coexisting illnesses such as chronic obstructive lung disease, bronchiectasis, malignancy (*), diabetes mellitus, chronic renal failure (*), congestive heart failure (*), chronic liver disease (*), chronic alcohol abuse, malnutrition, cerebrovascular disease (*), and postsplenectomy. A history of hospitalization within the past year is also a risk factor