American Thoracic Society Documents

Guidelines for the Management of Adults withHospital-acquired, Ventilator-associated, andHealthcare-associated PneumoniaThis official statement of the American Thoracic Society and the Infectious Diseases Society of America was approvedby the ATS Board of Directors, December 2004 and the IDSA Guideline Committee, October 2004

CONTENTS

Executive SummaryIntroductionMethodology Used to Prepare the GuidelineEpidemiology

IncidenceEtiologyMajor Epidemiologic Points

PathogenesisMajor Points for Pathogenesis

Modifiable Risk FactorsIntubation and Mechanical VentilationAspiration, Body Position, and Enteral FeedingModulation of Colonization: Oral Antiseptics and AntibioticsStress Bleeding Prophylaxis, Transfusion, and Glucose ControlMajor Points and Recommendations for Modifiable

Risk FactorsDiagnostic Testing

Major Points and Recommendations for DiagnosisDiagnostic Strategies and Approaches

Clinical StrategyBacteriologic StrategyRecommended Diagnostic StrategyMajor Points and Recommendations for Comparing

Diagnostic StrategiesAntibiotic Treatment of Hospital-acquired Pneumonia

General ApproachInitial Empiric Antibiotic TherapyAppropriate Antibiotic Selection and Adequate DosingLocal Instillation and Aerosolized AntibioticsCombination versus MonotherapyDuration of TherapyMajor Points and Recommendations for Optimal

Antibiotic TherapySpecific Antibiotic RegimensAntibiotic Heterogeneity and Antibiotic Cycling

Response to TherapyModification of Empiric Antibiotic RegimensDefining the Normal Pattern of ResolutionReasons for Deterioration or NonresolutionEvaluation of the Nonresponding PatientMajor Points and Recommendations for Assessing

Response to TherapySuggested Performance Indicators

EXECUTIVE SUMMARY

Since the initial 1996 American Thoracic Society (ATS) guide-line on nosocomial pneumonia, a number of new developments

Am J Respir Crit Care Med Vol 171. pp 388416, 2005DOI: 10.1164/rccm.200405-644STInternet address: www.atsjournals.org

have appeared, mandating a new evidence-based guideline forhospital-acquired pneumonia (HAP), including healthcare-asso-ciated pneumonia (HCAP) and ventilator-associated pneumonia(VAP). This document, prepared by a joint committee of theATS and Infectious Diseases Society of America (IDSA), fo-cuses on the epidemiology and pathogenesis of bacterial pneu-monia in adults, and emphasizes modifiable risk factors for infec-tion. In addition, the microbiology of HAP is reviewed, with anemphasis on multidrug-resistant (MDR) bacterial pathogens,such as Pseudomonas aeruginosa, Acinetobacter species, andmethicillin-resistant Staphylococcus aureus. Controversies aboutdiagnosis are discussed, emphasizing initial examination of lowerrespiratory tract samples for bacteria, and the rationale for bothclinical and bacteriologic approaches, using either semiquanti-tative or quantitative microbiologic methods that help directselection of appropriate antibiotic therapy. We also provide rec-ommendations for additional diagnostic and therapeutic evalua-tions in patients with nonresolving pneumonia. This is an evi-dence-based document that emphasizes the issues of VAP,because there are far fewer data available about HAP in nonintu-bated patients and about HCAP. By extrapolation, patients whoare not intubated and mechanically ventilated should be man-aged like patients with VAP, using the same approach to identifyrisk factors for infection with specific pathogens.

The major goals of this evidence-based guideline for the man-agement of HAP, VAP, and HCAP emphasize early, appropriateantibiotics in adequate doses, while avoiding excessive antibiot-ics by de-escalation of initial antibiotic therapy, based on micro-biologic cultures and the clinical response of the patient, andshortening the duration of therapy to the minimum effectiveperiod. The guideline recognizes the variability of bacteriologyfrom one hospital to another and from one time period to an-other and recommends taking local microbiologic data into ac-count when adapting treatment recommendations to any specificclinical setting. The initial, empiric antibiotic therapy algorithmincludes two groups of patients: one with no need for broad-spectrum therapy, because these patients have early-onset HAP,VAP, or HCAP and no risk factors for MDR pathogens, and asecond group that requires broad-spectrum therapy, because oflate-onset pneumonia or other risk factors for infection withMDR pathogens.

Some of the key recommendations and principles in this new,evidence-based guideline are as follows:

HCAP is included in the spectrum of HAP and VAP, andpatients with HCAP need therapy for MDR pathogens.

A lower respiratory tract culture needs to be collectedfrom all patients before antibiotic therapy, but collectionof cultures should not delay the initiation of therapy incritically ill patients.

Either semiquantitative or quantitative culture datacan be used for the management of patients with HAP.

Lower respiratory tract cultures can be obtained broncho-

American Thoracic Society Documents 389

scopically or nonbronchoscopically, and can be culturedquantitatively or semiquantitatively.

Quantitative cultures increase specificity of the diagnosisof HAP without deleterious consequences, and the specificquantitative technique should be chosen on the basis oflocal expertise and experience.

Negative lower respiratory tract cultures can be used tostop antibiotic therapy in a patient who has had culturesobtained in the absence of an antibiotic change in the past72 hours.

Early, appropriate, broad-spectrum, antibiotic therapyshould be prescribed with adequate doses to optimize anti-microbial efficacy.

An empiric therapy regimen should include agents that arefrom a different antibiotic class than the patient has re-cently received.

Combination therapy for a specific pathogen should beused judiciously in the therapy of HAP, and considerationshould be given to short-duration (5 days) aminoglycosidetherapy, when used in combination with a -lactam to treatP. aeruginosa pneumonia.

Linezolid is an alternative to vancomycin, and uncon-firmed, preliminary data suggest it may have an advantagefor proven VAP due to methicillin-resistant S. aureus.

Colistin should be considered as therapy for patients withVAP due to a carbapenem-resistant Acinetobacter species.

Aerosolized antibiotics may have value as adjunctive ther-apy in patients with VAP due to some MDR pathogens.

De-escalation of antibiotics should be considered once dataare available on the results of lower respiratory tract cul-tures and the patients clinical response.

A shorter duration of antibiotic therapy (7 to 8 days) isrecommended for patients with uncomplicated HAP, VAP,or HCAP who have received initially appropriate therapyand have had a good clinical response, with no evidenceof infection with nonfermenting gram-negative bacilli.

INTRODUCTION

As with all guidelines, these new recommendations, althoughevidence graded, need validation for their impact on the outcomeof patients with HAP, VAP, and HCAP. In addition, this guide-line points out areas of incomplete knowledge, which can beused to set an agenda for future research.

Hospital-acquired pneumonia (HAP), ventilator-associatedpneumonia (VAP), and healthcare-associated pneumonia (HCAP)remain important causes of morbidity and mortality despite ad-vances in antimicrobial therapy, better supportive care modal-ities, and the use of a wide-range of preventive measures (15).HAP is defined as pneumonia that occurs 48 hours or more afteradmission, which was not incubating at the time of admission(1, 3). HAP may be managed in a hospital ward or in the intensivecare unit (ICU) when the illness is more severe. VAP refers topneumonia that arises more than 4872 hours after endotrachealintubation (2, 3). Although not included in this definition, somepatients may require intubation after developing severe HAPand should be managed similar to patients with VAP. HCAPincludes any patient who was hospitalized in an acute care hospi-tal for two or more days within 90 days of the infection; residedin a nursing home or long-term care facility; received recentintravenous antibiotic therapy, chemotherapy, or wound carewithin the past 30 days of the current infection; or attended ahospital or hemodialysis clinic (3, 4, 6). Although this documentfocuses more on HAP and VAP, most of the principles overlapwith HCAP. Because most of the current data have been col-lected from patients with VAP, and microbiologic data from

nonintubated patients may be less accurate, most of our informa-tion is derived from those with VAP, but by extrapolation canbe applied to all patients with HAP, emphasizing risk factorsfor infection with specific pathogens.

This guideline is an update of the 1996 consensus statementon HAP published by the American Thoracic Society (5). Theprinciples and recommendations are largely based on data pre-sented by committee members at a conference jointly sponsoredby the American Thoracic Society (ATS) and the InfectiousDisease Society of America (IDSA). The committee was com-posed of pulmonary, critical care, and infectious disease special-ists with clinical and research interests in HAP, VAP, and HCAP.All major aspects of the epidemiology, pathogenesis, bacteriol-ogy, diagnosis, and antimicrobial treatment were reviewed bythis group. Therapy recommendations are focused on antibioticchoice and patient stratification; adjunctive, nonantibiotic ther-apy of pneumonia is not discussed, but information on this topicis available elsewhere (7). Recommendations to reduce the riskof pneumonia are limited in this document to key, modifiablerisk factors related to the pathogenesis of pneumonia to avoidredundancy with the more comprehensive Guidelines for Pre-venting Health-careassociated Pneumonia, prepared by the Cen-ters for Disease Control and Prevention (CDC) and the HospitalInfection Control Practices Advisory Committee (HICPAC) (3).

The goal of our document is to provide a framework for theinitial evaluation and management of the immunocompetent,adult patient with bacterial causes of HAP, VAP, or HCAP,and excludes patients who are known to be immunosuppressedby human immunodeficiency virus (HIV) infection, hematologicmalignancy, chemotherapy-induced neutropenia, organ trans-plantation, and so on. At the outset, the ATS/IDSA GuidelineCommittee members recognized that currently, many patientswith HAP, VAP, or HCAP are infected with multidrug-resistant(MDR) bacterial pathogens that threaten the adequacy of initial,empiric antibiotic therapy. At the same time, the committeemembers recognized that many studies have shown that exces-sive antibiotic use is a major factor contributing to increasedfrequency of antibiotic-resistant pathogens. Four major princi-ples underlie the management of HAP, VAP, and HCAP:

Avoid untreated or inadequately treated HAP, VAP, orHCAP, because the failure to initiate prompt appropriateand adequate therapy has been a consistent factor associ-ated with increased mortality.

Recognize the variability of bacteriology from one hospitalto another, specific sites within the hospital, and from onetime period to another, and use this information to alter theselection of an appropriate antibiotic treatment regimen forany specific clinical setting.

Avoid the overuse of antibiotics by focusing on accuratediagnosis, tailoring therapy to the results of lower respira-tory tract cultures, and shortening duration of therapy tothe minimal effective period.

Apply prevention strategies aimed at modifiable risk fac-tors.

The ATS/IDSA guideline was established for use in the initialmanagement of patients in whom HAP, VAP, or HCAP is sus-pected. Therapeutic algorithms are presented that are based onthe expected antimicrobial susceptibility of the common bacte-rial pathogens, and with therapeutic regimens that can commonlylead to initial adequate antibiotic management.

This guideline is not meant to replace clinical judgment, butrather to give an organizational framework to patient manage-ment. Individual clinical situations can be highly complex andthe judgment of a knowledgeable physician with all availableinformation about a specific patient is essential for optimal clini-

390 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 171 2005

TABLE 1. EVIDENCE-BASED GRADING SYSTEM USED TO RANK RECOMMENDATIONS

Evidence Level Definition

Level I (high) Evidence comes from well conducted, randomized controlled trialsLevel II (moderate) Evidence comes from well designed, controlled trials without randomization (including cohort,

patient series, and case-control studies). Level II studies also include any large case seriesin 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

Level III (low) Evidence comes from case studies and expert opinion. In some instances therapy recommendationscome from antibiotic susceptibility data without clinical observations

Adapted from American Thoracic Society guidelines for the management of adults with community-acquired pneumonia (8).

cal management. As more laboratory and clinical data becomeavailable, therapy often needs to be streamlined or altered. Fi-nally, our committee realizes that these guidelines will changeover time, and that our current recommendations will need tobe updated as new information becomes available.

METHODOLOGY USED TO PREPARE THE GUIDELINE

The ATS/IDSA Guideline Committee originally met as a group,with each individual being assigned a topic for review and pre-sentation to the entire group. Each topic in the guideline wasreviewed by more than one committee member, and after pre-sentation of information, the committee discussed the data andformulated recommendations. Two committee members pre-pared each section of the document, and a draft document incor-porating all sections was written and distributed to the committeefor review and suggestions. The guideline was then revised andcirculated to the committee for final comment. This final state-ment represents the results of this process and the opinions ofthe majority of committee members.

The grading system for our evidence-based recommendationswas previously used for the updated ATS Community-acquiredPneumonia (CAP) statement, and the definitions of high-level(Level I), moderate-level (Level II), and low-level (Level III)evidence are summarized in Table 1 (8). All available and rele-vant, peer-reviewed studies published until July 2004 were con-sidered. Much of the literature is observational, and only a fewtherapy trials have been conducted in a prospective, randomizedfashion.

Nearly all of the evidence-based data on risk factors for bacte-rial HAP have been collected from observational studies, whichcannot distinguish causation from noncausal association. Mostof the studies have focused on patients with VAP, but the com-mittee extrapolated the relationship between risk factors andbacteriology to all patients with HAP, including those withHCAP. Ultimate proof of causality, and ideally the best strate-gies for prevention of HAP, VAP, and HCAP, should be basedon prospective, randomized trials. However, recommendationsare further compromised when such trials provide conflictingresults, often as a result of differences in definitions, study design,and the specific population studied. In addition, evidence-basedrecommendations are dynamic and may change as new therapiesbecome available and as new interventions alter the naturalhistory of the disease.

EPIDEMIOLOGY

Incidence

HAP is usually caused by bacteria, is currently the second mostcommon nosocomial infection in the United States, and is associ-ated with high mortality and morbidity (3). The presence ofHAP increases hospital stay by an average of 7 to 9 days perpatient and has been reported to produce an excess cost of

more than $40,000 per patient (911). Although HAP is not areportable illness, available data suggest that it occurs at a rateof between 5 and 10 cases per 1,000 hospital admissions, with theincidence increasing by as much as 6- to 20-fold in mechanicallyventilated patients (9, 12, 13). It is often difficult to define theexact incidence of VAP, because there may be an overlap withother lower respiratory tract infections, such as infectious tra-cheobronchitis in mechanically ventilated patients. The exactincidence varies widely depending on the case definition of pneu-monia and the population being evaluated (14). For example,the incidence of VAP may be up to two times higher in patientsdiagnosed by qualitative or semiquantitative sputum culturescompared with quantitative cultures of lower respiratory tractsecretions (9, 15).

HAP accounts for up to 25% of all ICU infections and formore than 50% of the antibiotics prescribed (16). VAP occursin 927% of all intubated patients (9, 11). In ICU patients, nearly90% of episodes of HAP occur during mechanical ventilation.In mechanically ventilated patients, the incidence increases withduration of ventilation. The risk of VAP is highest early in thecourse of hospital stay, and is estimated to be 3%/day duringthe first 5 days of ventilation, 2%/day during Days 5 to 10 ofventilation, and 1%/day after this (17). Because most mechanicalventilation is short term, approximately half of all episodes ofVAP occur within the first 4 days of mechanical ventilation. Theintubation process itself contributes to the risk of infection, andwhen patients with acute respiratory failure are managed withnoninvasive ventilation, nosocomial pneumonia is less common(1820).

Time of onset of pneumonia is an important epidemiologicvariable and risk factor for specific pathogens and outcomes inpatients with HAP and VAP . Early-onset HAP and VAP, definedas occurring within the first 4 days of hospitalization, usually carrya better prognosis, and are more likely to be caused by antibiotic-sensitive bacteria. Late-onset HAP and VAP (5 days or more) aremore likely to be caused by multidrug-resistant (MDR) pathogens,and are associated with increased patient mortality and morbidity.However, patients with early-onset HAP who have received priorantibiotics or who have had prior hospitalization within the past90 days are at greater risk for colonization and infection withMDR pathogens and should be treated similar to patients withlate-onset HAP or VAP (Table 2) (21).

The crude mortality rate for HAP may be as high as 30 to70%, but many of these critically ill patients with HAP die oftheir underlying disease rather than pneumonia. The mortalityrelated to the HAP or attributable mortality has been estimatedto be between 33 and 50% in several case-matching studies ofVAP. Increased mortality rates were associated with bacteremia,especially with Pseudomonas aeruginosa or Acinetobacter species,medical rather than surgical illness, and treatment with ineffectiveantibiotic therapy (22, 23). Other studies using similar methodol-ogy failed to identify any attributable mortality due to VAP,

American Thoracic Society Documents 391

TABLE 2. RISK FACTORS FOR MULTIDRUG-RESISTANTPATHOGENS CAUSING HOSPITAL-ACQUIRED PNEUMONIA,HEALTHCARE-ASSOCIATED PNEUMONIA, ANDVENTILATOR-ASSOCIATED PNEUMONIA

Antimicrobial therapy in preceding 90 d Current hospitalization of 5 d or more High frequency of antibiotic resistance in the community or

in the specific hospital unit Presence of risk factors for HCAP:

Hospitalization for 2 d or more in the preceding 90 dResidence in a nursing home or extended care facilityHome infusion therapy (including antibiotics)Chronic dialysis within 30 dHome wound careFamily member with multidrug-resistant pathogen

Immunosuppressive disease and/or therapy

suggesting a variable outcome impact, according to the severityof underlying medical conditions (2426).

Etiology

HAP, VAP, and HCAP may be caused by a wide spectrum ofbacterial pathogens, may be polymicrobial, and are rarely dueto viral or fungal pathogens in immunocompetent hosts (9, 12,2732). Common pathogens include aerobic gram-negative ba-cilli, such as P. aeruginosa, Escherichia coli, Klebsiella pneumo-niae, and Acinetobacter species. Infections due to gram-positivecocci, such as Staphylococcus aureus, particularly methicillin-resistant S. aureus (MRSA), have been rapidly emerging in theUnited States (16, 33). Pneumonia due to S. aureus is morecommon in patients with diabetes mellitus, head trauma, andthose hospitalized in ICUs (34).

Significant growth of oropharyngeal commensals (viridansgroup streptococci, coagulase-negative staphylococci, Neisseriaspecies, and Corynebacterium species) from distal bronchialspecimens is difficult to interpret, but these organisms can pro-duce infection in immunocompromised hosts and some immuno-competent patients (35). Rates of polymicrobial infection varywidely, but appear to be increasing, and are especially high inpatients with adult respiratory distress syndrome (ARDS) (9, 12,3638).

The frequency of specific MDR pathogens causing HAP mayvary by hospital, patient population, exposure to antibiotics, typeof ICU patient, and changes over time, emphasizing the needfor timely, local surveillance data (3, 8, 10, 21, 3941). HAPinvolving anaerobic organisms may follow aspiration in nonintu-bated patients, but is rare in patients with VAP (28, 42).

Elderly patients represent a diverse population of patientswith pneumonia, particularly HCAP. Elderly residents of long-term care facilities have been found to have a spectrum of patho-gens that more closely resemble late-onset HAP and VAP (30, 31).In a study of 104 patients age 75 years and older with severepneumonia, El-Solh found S. aureus (29%), enteric gram-nega-tive rods (15%), Streptococcus pneumoniae (9%), and Pseudo-monas species (4%) as the most frequent causes of nursing home-acquired pneumonia (30). In another study of 52 long-term careresidents aged 70 years and above who failed to respond to72 hours of antibiotics, MRSA (33%), gram-negative enterics(24%), and Pseudomonas species (14%) were the most frequentpathogens isolated by invasive diagnostics (bronchoscopy) (31).In the latter study, 72% had at least two comorbidities whereas23% had three or more.

Few data are available about the bacteriology and risk factorsfor specific pathogens in patients with HAP and HCAP, andwho are not mechanically ventilated. Data from comprehensive

hospital-wide surveillance of nosocomial infections at the Uni-versity of North Carolina have described the pathogens causingboth VAP and nosocomial pneumonia in nonintubated patientsduring the years 20002003 (D. Weber and W. Rutala, unpub-lished data). Pathogens were isolated from 92% of mechanicallyventilated patients with infection, and from 77% of nonventi-lated patients with infection. In general, the bacteriology ofnonventilated patients was similar to that of ventilated patients,including infection with MDR pathogens such as methicillin-resistant S. aureus (MRSA), P. aeruginosa, Acinetobacter spe-cies, and K. pneumoniae. In fact, some organisms (MRSA andK. pneumoniae) were more common in nonventilated than venti-lated patients, whereas certain resistant gram-negative bacilliwere more common in patients with VAP (P. aeruginosa, Steno-trophomonas maltophilia, and Acinetobacter species). However,the latter group of more resistant gram-negative bacilli occurredwith sufficient frequency in nonventilated patients that theyshould be considered when designing an empiric therapy regi-men. Studies in nonventilated patients have not determinedwhether this population has risk factors for MDR pathogensthat differ from the risk factors present in ventilated patients.

Emergence of selected multidrug-resistant bacteria. Rates ofHAP due to MDR pathogens have increased dramatically inhospitalized patients, especially in intensive care and transplantpatients (16). Risk factors for colonization and infection withMDR pathogens are summarized in Table 2 (21, 43). Data onmechanisms of antibiotic resistance for specific bacterial patho-gens have provided new insight into the adaptability of thesepathogens.

Pseudomonas aeruginosa. P. aeruginosa, the most commonMDR gram-negative bacterial pathogen causing HAP/VAP, hasintrinsic resistance to many antimicrobial agents (4446). Thisresistance is mediated by multiple efflux pumps, which maybe expressed all the time or may be upregulated by mutation(47). Resistance to piperacillin, ceftazidime, cefepime, other oxy-imino--lactams, imipenem and meropenem, aminoglycosides,or fluoroquinolones is increasing in the United States (16). De-creased expression of an outer membrane porin channel (OprD)can cause resistance to both imipenem and meropenem or, de-pending on the alteration in OprD, specific resistance to imi-penem, but not other -lactams (48). At present, some MDRisolates of P. aeruginosa are susceptible only to polymyxin B.

Although currently uncommon in the United States, thereis concern about the acquisition of plasmid-mediated metallo--lactamases active against carbapenems and antipseudomonalpenicillins and cephalosporins (49). The first such enzyme, IMP-1,appeared in Japan in 1991 and spread among P. aeruginosa andSerratia marcescens, and then to other gram-negative pathogens.Resistant strains of P. aeruginosa with IMP-type enzymes andother carbapenemases have been reported from additional coun-tries in the Far East, Europe, Canada, Brazil, and recently inthe United States (50).

Klebsiella, Enterobacter, and Serratia species. Klebsiellaspecies are intrinsically resistant to ampicillin and other amino-penicillins and can acquire resistance to cephalosporins and az-treonam by the production of extended-spectrum -lactamases(ESBLs) (51). Plasmids encoding ESBLs often carry resistanceto aminoglycosides and other drugs, but ESBL-producing strainsremain susceptible to carbapenems. Five to 10% of oxyimino--lactam-resistant K. pneumoniae do not produce an ESBL, butrather a plasmid-mediated AmpC-type enzyme (52). Such strainsusually are carbapenem susceptible, but may become resistantby loss of an outer membrane porin (53). Enterobacter specieshave a chromosomal AmpC -lactamase that is inducible andalso easily expressed at a high level by mutation with consequentresistance to oxyimino--lactams and -methoxy--lactams,

392 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 171 2005

such as cefoxitin and cefotetan, but continued susceptibility tocarbapenems. Citrobacter and Serratia species have the sameinducible AmpC -lactamase and the same potential for resis-tance development. Although the AmpC enzyme of E. coli isnot inducible, it can occasionally be hyperexpressed. Plasmid-mediated resistance, such as ESBL production, is a more commonmechanism for -lactam resistance in nosocomial isolates, and isincreasingly recognized not only in isolates of K. pneumoniae andE. coli, but also Enterobacter species (54).

Acinetobacter species, Stenotrophomonas maltophilia,and Burkholderia cepacia. Although generally less virulentthan P. aeruginosa, Acinetobacter species have nonetheless be-come problem pathogens because of increasing resistance tocommonly used antimicrobial agents (55). More than 85% ofisolates are susceptible to carbapenems, but resistance is increas-ing due either to IMP-type metalloenzymes or carbapenemasesof the OXA type (49). An alternative for therapy is sulbactam,usually employed as an enzyme inhibitor, but with direct antibac-terial activity against Acinetobacter species (56). S. maltophilia,which shares with B. cepacia a tendency to colonize the respira-tory tract rather than cause invasive disease, is uniformly resistantto carbapenems, because of a ubiquitous metallo--lactamase.S. maltophilia and B. cepacia are most likely to be susceptibleto trimethoprimsulfamethoxazole, ticarcillinclavulanate, or afluoroquinolone (55). B. cepacia is also usually susceptible toceftazidime and carbapenems.

Methicillin-resistant Staphylococcus aureus. In theUnited States, more than 50% of the ICU infections caused byS. aureus are with methicillin-resistant organisms (16, 33). MRSAproduces a penicillin-binding protein with reduced affinity for-lactam antibiotics that is encoded by the mecA gene, whichis carried by one of a family of four mobile genetic elements(57, 58). Strains with mecA are resistant to all commerciallyavailable -lactams and many other antistaphylococcal drugs,with considerable country-to-country variability (59, 60). Al-though vancomycin-intermediate S. aureus, with a minimal inhib-itory concentration (MIC) of 816 g/ml, and high-level vanco-mycin-resistant S. aureus, with an MIC of 321,024 g/ml ormore, have been isolated from clinical specimens, none to datehave caused respiratory tract infection and all have been sensi-tive to linezolid (61, 62). Unfortunately, linezolid resistance hasemerged in S. aureus, but is currently rare (63).

Streptococcus pneumoniae and Haemophilus influenzae.S. pneumoniae and H. influenzae cause early-onset HAP in pa-tients without other risk factors, are uncommon in late-onsetinfection, and frequently are community acquired. At present,many strains of S. pneumoniae are penicillin resistant due toaltered penicillin-binding proteins. Some such strains are resis-tant as well to cephalosporins, macrolides, tetracyclines, andclindamycin (64). Despite low and moderate levels of resistanceto penicillins and cephalosporins in vitro, clinical outcomes inpatients with pneumococcal pneumonia and bacteremia treatedwith these agents have been satisfactory (65). All of the multi-drug-resistant strains in the United States are currently sensitiveto vancomycin or linezolid, and most remain sensitive to broad-spectrum quinolones. Resistance of H. influenzae to antibioticsother than penicillin and ampicillin is sufficiently rare so as notto present a problem in therapy.

Legionella pneumophila. The evidence for Legionella pneu-mophila as a cause of HAP is variable, but is increased in immu-nocompromised patients, such as organ transplant recipients orpatients with HIV disease, as well as those with diabetes mellitus,underlying lung disease, or end-stage renal disease (29, 6669).HAP due to Legionella species is more common in hospitalswhere the organism is present in the hospital water supply orwhere there is ongoing construction (3, 29, 6669). Because de-

tection is based on the widespread use of Legionella urinaryantigen, rather than culture for Legionella, disease due to sero-groups other than serogroup 1 may be underdiagnosed. Detailedstrategies for prevention of Legionella infections and eradicationprocedures for Legionella species in cooling towers and the hos-pital water supply are outlined in the CDC/HICPAC Guidelinesfor Preventing Health-careassociated Pneumonia (3).

Fungal pathogens. Nosocomial pneumonia due to fungi, suchas Candida species and Aspergillus fumigatus, may occur in organtransplant or immunocompromised, neutropenic patients, but isuncommon in immunocompetent patients (7075). NosocomialAspergillus species infections suggest possible airborne transmis-sion by spores, and may be associated with an environmentalsource such as contaminated air ducts or hospital construction.By comparison, isolation of Candida albicans and other Candidaspecies from endotracheal aspirates is common, but usually rep-resents colonization of the airways, rather than pneumonia inimmunocompetent patients, and rarely requires treatment withantifungal therapy (70).

Viral pathogens. The incidence of HAP and VAP due toviruses is also low in immunocompetent hosts. Outbreaks ofHAP, VAP, and HCAP due to viruses, such as influenza, parainflu-enza, adenovirus, measles, and respiratory syncytial virus havebeen reported and are usually seasonal. Influenza, pararinflu-enza, adenovirus, and respiratory syncytial virus account for 70%of the nosocomial viral cases of HAP, VAP, and HCAP (3, 7678).Respiratory syncytial virus outbreaks of bronchiolitis and pneu-monia are more common in childrens wards and rare in immuno-competent adults (76). Diagnosis of these viral infections is oftenmade by rapid antigen testing and viral culture or serologic assays.

Influenza A is probably the most common viral cause ofHAP and HCAP in adult patients. Pneumonia in patients withinfluenza A or B may be due to the virus, to secondary bacterialinfection, or both. Influenza is transmitted directly from personto person when infected persons sneeze, cough, or talk or indi-rectly by personfomiteperson transmission (3, 7981). The useof influenza vaccine along with prophylaxis and early antiviraltherapy among at-risk healthcare workers and high-risk patientswith amantadine, rimantadine, or one of the neuraminidase in-hibitors (oseltamivir and zanamivir) dramatically reduces thespread of influenza within hospital and healthcare facilities (3,8190). Amantadine and rimantadine are effective only for treat-ment and prophylaxis against influenza A strains, whereas neura-minidase inhibitors are effective against both influenza A and B.

Major Epidemiologic Points

1. Many patients with HAP, VAP, and HCAP are at in-creased risk for colonization and infection with MDRpathogens (Level II) (24, 6, 9, 1113, 21, 22).

2. It is often difficult to define the exact incidence of HAPand VAP, because there may be an overlap with otherlower respiratory tract infections, such as tracheobronchi-tis, especially in mechanically ventilated patients (LevelIII) (9, 1214).

3. The exact incidence of HAP is usually between 5 and 15cases per 1,000 hospital admissions depending on the casedefinition and study population; the exact incidence ofVAP is 6- to 20-fold greater than in nonventilated patients(Level II) (9, 1214).

4. HAP and VAP are a frequent cause of nosocomial infec-tion that is associated with a higher crude mortality thanother hospital-acquired infections (Level II) (3, 9, 16).

5. Patients with late-onset HAP and VAP are more likelyto be infected with MDR pathogens and have higher

American Thoracic Society Documents 393

crude mortality than patients with early-onset disease;patients with early-onset HAP who have recently re-ceived antibiotics or had an admission to a healthcarefacility are at risk for colonization and infection withMDR pathogens (Level II) (3, 9, 21, 22).

6. An increase in crude and attributable mortality for HAPand VAP is associated with the presence of MDR patho-gens (Level II) (3, 5, 913, 2123).

7. Bacteria cause most cases of HAP, VAP, and HCAP andmany infections are polymicrobial; rates are especiallyhigh in patients with ARDS (Level I) (2, 4, 6, 9, 12,3638).

8. HAP, VAP, and HCAP are commonly caused by aerobicgram-negative bacilli, such as P. aeruginosa, K. pneumo-niae, and Acinetobacter species, or by gram-positive cocci,such as S. aureus, much of which is MRSA; anaerobesare an uncommon cause of VAP (Level II) (9, 12, 28,3640, 42, 91).

9. Rates of L. pneumophila vary considerably between hos-pitals and disease occurs more commonly with serogroup1 when the water supply is colonized or there is ongoingconstruction (Level II) (29, 6669).

10. Nosocomial virus and fungal infections are uncommoncauses of HAP and VAP in immunocompetent patients.Outbreaks of influenza have occurred sporadically andrisk of infection can be substantially reduced with wide-spread effective infection control, vaccination, and useof antiinfluenza agents (Level I) (3, 7075, 7990).

11. The prevalence of MDR pathogens varies by patient pop-ulation, hospital, and type of ICU, which underscores theneed for local surveillance data (Level II) (3, 9, 41).

12. MDR pathogens are more commonly isolated from pa-tients with severe, chronic underlying disease, those withrisk factors for HCAP, and patients with late-onset HAPor VAP (Level II) (9, 21, 22, 30, 31, 39, 40, 91).

PATHOGENESIS

For HAP to occur, the delicate balance between host defensesand microbial propensity for colonization and invasion must shiftin favor of the ability of the pathogens to persist and invade thelower respiratory tract. Sources of infection for HAP includehealthcare devices or the environment (air, water, equipment,and fomites) and can occur with transfer of microorganismsbetween staff and patients (3, 9, 12, 13, 27, 66, 92, 93). A numberof host- and treatment-related colonization factors, such as theseverity of the patients underlying disease, prior surgery, expo-sure to antibiotics, other medications, and exposure to invasiverespiratory devices and equipment, are important in the patho-genesis of HAP and VAP (40, 93, 94)

HAP requires the entry of microbial pathogens into the lowerrespiratory tract, followed by colonization, which can then over-whelm the hosts mechanical (ciliated epithelium and mucus),humoral (antibody and complement), and cellular (polymorpho-nuclear leukocytes, macrophages, and lymphocytes and theirrespective cytokines) defenses to establish infection (9, 94).

Aspiration of oropharyngeal pathogens or leakage of bacteriaaround the endotracheal tube cuff is the primary route of bacte-rial entry into the trachea (9598). The stomach and sinuseshave been suggested as potential reservoirs for certain bacteriacolonizing the oropharynx and trachea, but their importanceremains controversial (99104). Some investigators postulatethat colonization of the endotracheal tube with bacteria encasedin biofilm may result in embolization into the alveoli duringsuctioning or bronchoscopy (105, 106). Inhalation of pathogensfrom contaminated aerosols, and direct inoculation, are less com-

mon (107, 108). Hematogenous spread from infected intravascu-lar catheters or bacterial translocation from the gastrointestinaltract lumen are quite rare.

Major Points for Pathogenesis

1. Sources of pathogens for HAP include healthcare devices,the environment (air, water, equipment, and fomites), andcommonly the transfer of microorganisms between thepatient and staff or other patients (Level II) (3, 9, 12, 13,27, 66, 92, 93).

2. A number of host- and treatment-related colonization fac-tors, such as the severity of the patients underlying disease,prior surgery, exposure to antibiotics, other medications,and exposure to invasive respiratory devices and equip-ment, are important in the pathogenesis of HAP and VAP(Level II) (40, 93, 94).

3. Aspiration of oropharyngeal pathogens, or leakage of se-cretions containing bacteria around the endotracheal tubecuff, are the primary routes of bacterial entry into thelower respiratory tract (Level II) (9598).

4. Inhalation or direct inoculation of pathogens into the lowerairway, hematogenous spread from infected intravenouscatheters, and bacterial translocation from the gastrointes-tinal tract lumen are uncommon pathogenic mechanisms(Level II) (107, 108).

5. Infected biofilm in the endotracheal tube, with subsequentembolization to distal airways, may be important in thepathogenesis of VAP (Level III) (105, 106).

6. The stomach and sinuses may be potential reservoirs ofnosocomial pathogens that contribute to bacterial coloni-zation of the oropharynx, but their contribution is contro-versial, may vary by the population at risk, and may bedecreasing with the changing natural history and manage-ment of HAP (Level II) (94, 99104, 109).

MODIFIABLE RISK FACTORS

Risk factors for the development of HAP can be differentiatedinto modifiable and nonmodifiable conditions. Risk factors mayalso be patient related (male sex, preexisting pulmonary disease,or multiple organ system failure) or treatment related (intuba-tion or enteral feeding). Modifiable risk factors for HAP areobvious targets for improved management and prophylaxis inseveral studies and in the comprehensive Guidelines for Pre-venting Health-careassociated Pneumonia, published by theCenters for Disease Control (3, 93, 110). Effective strategiesinclude strict infection control, alcohol-based hand disinfection,use of microbiologic surveillance with timely availability of dataon local MDR pathogens, monitoring and early removal of inva-sive devices, and programs to reduce or alter antibiotic-prescrib-ing practices (3, 92, 93, 100, 110113).

Intubation and Mechanical Ventilation

Intubation and mechanical ventilation increase the risk of HAP6- to 21-fold and therefore should be avoided whenever possible(3, 94, 110, 114). Noninvasive positive-pressure ventilation, usinga face mask, is an attractive alternative for patients with acuteexacerbations of chronic obstructive pulmonary disease or acutehypoxemic respiratory failure, and for some immunosuppressedpatients with pulmonary infiltrates and respiratory failure (18, 20,115119). Data suggest that use of noninvasive ventilation toavoid reintubation after initial extubation may not be a goodstrategy (115).

Specific strategies have been recommended to reduce theduration of mechanical ventilation, such as improved methods

394 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 171 2005

of sedation and the use of protocols to facilitate and accelerateweaning (120124). These interventions are dependent on ade-quate ICU staffing. Reintubation should be avoided, if possible,as it increases the risk of VAP (114).

Attention to the specific type of endotracheal tube, its mainte-nance, and the site of insertion may also be valuable. The useof oral endotracheal and orogastric tubes, rather than nasotra-cheal and nasogastric tubes, can reduce the frequency of nosoco-mial sinusitis and possibly HAP, although causality betweensinusitis and HAP has not been firmly established (109, 125).Efforts to reduce the likelihood of aspiration of oropharyngealbacteria around the endotracheal tube cuff and into the lowerrespiratory tract include limiting the use of sedative and paralyticagents that depress cough and other host-protective mechanisms,and maintaining endotracheal cuff pressure at greater than 20 cmH2O (98, 126). Continuous aspiration of subglottic secretions,through the use of a specially designed endotracheal tube, hassignificantly reduced the incidence of early-onset VAP in severalstudies (97, 127130).

VAP may also be related to colonization of the ventilatorcircuit (131). A large number of prospective, randomized trialshave shown that the frequency of ventilator circuit change doesnot affect the incidence of HAP, but condensate collecting inthe ventilator circuit can become contaminated from patientsecretions (98, 132135). Therefore, vigilance is needed to pre-vent inadvertently flushing the condensate into the lower airwayor to in-line medication nebulizers when the patient turns orthe bedrail is raised (98, 131134, 136). Passive humidifiers orheatmoisture exchangers decrease ventilator circuit coloniza-tion but have not significantly reduced the incidence of VAP(128, 135139).

Aspiration, Body Position, and Enteral Feeding

Supine patient positioning may also facilitate aspiration, whichmay be decreased by a semirecumbent positioning (140142).Using radioactive labeled enteral feeding, cumulative numbersof endotracheal counts were higher when patients were placedin the completely supine position (0 ) as compared with a semire-cumbent position (45 ) (140, 141). One randomized trial demon-strated a threefold reduction in the incidence of ICU-acquiredHAP in patients treated in the semirecumbent position com-pared with patients treated completely supine (143). Infectionin patients in the supine position was strongly associated withthe simultaneous administration of enteral nutrition. Thus, intu-bated patients should be managed in a semirecumbent position,particularly during feeding.

Enteral nutrition has been considered a risk factor for thedevelopment of HAP, mainly because of an increased risk ofaspiration of gastric contents (3, 144). However, its alternative,parenteral nutrition, is associated with higher risks for intravas-cular device-associated infections, complications of line inser-tions, higher costs, and loss of intestinal villous architecture,which may facilitate enteral microbial translocation. Althoughsome have advised feeding critically ill patients enterally as earlyas possible, a strategy of early (i.e., Day 1 of intubation andventilation) enteral feeding was, when compared with late ad-ministration (i.e., Day 5 of intubation), associated with a higherrisk for ICU-acquired VAP (145, 146). Seven studies have evalu-ated the risks for ICU-acquired HAP in patients randomized toeither gastric or postpyloric feeding (147). Although significantdifferences were not demonstrated in any individual study, post-pyloric feeding was associated with a significant reduction inICU-acquired HAP in metaanalysis (relative risk, 0.76; 95%confidence interval, 0.59 to 0.99) (147).

Modulation of Colonization: Oral Antiseptics and Antibiotics

The progression from colonization to tracheobronchitis to pneu-monia is a dynamic equilibrium and the possibility to discernthe different entities depends on the specificity of diagnostictools. Oropharyngeal colonization, either present on admissionor acquired during ICU stay, has been identified as an indepen-dent risk factor for the development of ICU-acquired HAPcaused by enteric gram-negative bacteria and P. aeruginosa(101). In a randomized trial, DeRiso and coworkers demon-strated that the use of the oral antiseptic chlorhexidine signifi-cantly reduced rates of nosocomial infection in patients undergo-ing coronary artery bypass surgery (148).

Modulation of oropharyngeal colonization, by combinationsof oral antibiotics, with or without systemic therapy, or by selec-tive decontamination of the digestive tract (SDD), is also effec-tive in significantly reducing the frequency of HAP, althoughmethodologic study quality appeared to be inversely related tothe magnitude of the preventive effects (93, 149155).

In two prospective randomized trials SDD was associatedwith higher ICU survival among patients receiving SDD (156,157). In the first study patients with a midrange APACHE IIscore on admission had a lower ICU mortality, although ICUmortality rates of all patients included did not differ significantly(156). In the largest study performed so far, SDD administeredto 466 patients in one unit was associated with a relative riskfor ICU mortality of 0.65 and with a relative risk of hospitalmortality of 0.78, when compared with 472 patients admitted ina control ward (157). In addition, infections due to antibiotic-resistant microorganisms occurred more frequently in the controlward. Importantly, levels of antibiotic-resistant pathogens werelow in both wards, with complete absence of MRSA. Moreover,a small preexisting difference in outcome between two wardsand the absence of a cross-over design warrant confirmation ofthese beneficial effects of SDD.

The preventive effects of selective decontamination of thedigestive tract for HAP have also been considerably lower inICUs with high endemic levels of antibiotic resistance. In sucha setting, selective decontamination of the digestive tract mayincrease the selective pressure for antibiotic-resistant micro-organisms (158164). Although selective decontamination of thedigestive tract reduces HAP, routine prophylactic use of antibiot-ics should be discouraged, especially in hospital settings wherethere are high levels of antibiotic resistance.

The role of systemic antibiotics in the development of HAPis less clear. In one study, prior administration of antibiotics hadan adjusted odds ratio of 3.1 (95% confidence interval, 1.46.9)for development of late-onset ICU-acquired HAP (165). More-over, antibiotics clearly predispose patients to subsequent coloni-zation and infection with antibiotic-resistant pathogens (21). Incontrast, prior antibiotic exposure conferred protection (riskratio, 0.37; 95% confidence interval, 0.270.51) for ICU-acquiredHAP in another study (17). In addition, antibiotic use at thetime of emergent intubation may prevent pneumonia within thefirst 48 hours of intubation (166). Preventive effects of intrave-nous antibiotics were evaluated in only one randomized trial:administration of cefuroxime for 24 hours, at the time of intuba-tion; and it reduced the incidence of early-onset, ICU-acquiredHAP in patients with closed head injury (167). However, circum-stantial evidence of the efficacy of systemic antibiotics also followsfrom the results of metaanalyses of selective decontaminationof the digestive tract, which have suggested that the intravenouscomponent of the regimens was largely responsible for improvedsurvival (149). In summary, prior administration of antibiotics forshort duration may be beneficial in some patient groups, but when

American Thoracic Society Documents 395

given for prolonged periods may well place others at risk forsubsequent infection with antibiotic-resistant microorganisms.

Stress Bleeding Prophylaxis, Transfusion, and Glucose Control

Both histamine Type 2 (H2) antagonists and antacids have beenidentified as independent risk factors for ICU-acquired HAP.Sucralfate has been used for stress bleeding prophylaxis, as itdoes not decrease intragastric acidity or significantly increasegastric volume. Numerous randomized trials, using different dosesand various study populations, have provided controversial re-sults on the benefits of specific stress bleeding prophylaxis agentsin relation to the increased risk of VAP (38, 99, 103, 104, 155,168). One large randomized trial comparing antacids, H2 block-ers, and sucralfate reported no differences in rates of early-onsetVAP, but rates of late-onset VAP were lower among patientstreated with sucralfate (103). In one multicenter study of VAPin patients with ARDS, sucralfate and duration of exposure tosucralfate were associated with an increased risk of VAP (38).A large, double-blind, randomized trial comparing ranitidinewith sucralfate demonstrated a trend to toward lower rates ofVAP with sucralfate, but clinically significant gastrointestinalbleeding was 4% higher in the sucralfate group (104). Thus, ifstress ulcer prophylaxis is indicated, the risks and benefits ofeach regimen should be weighed before prescribing either H2blockers or sucralfate.

A landmark prospective randomized trial comparing liberaland conservative triggers to transfusion in ICU patients notexhibiting active bleeding and without underlying cardiac diseasedemonstrated that awaiting a hemoglobin level of 7.0 g/dl asopposed to a level of 9.0 g/dl before initiating transfusion resultedin less transfusion and no adverse effects on outcome (169). Infact, in those patients less severely ill, as judged by low APACHEII scores, mortality was improved in the restricted transfusiongroup, a result thought to result from immunosuppressive effectsof nonleukocyte-depleted red blood cell units with consequentincreased risk for infection. Multiple studies have identified ex-posure to allogeneic blood products as a risk factor for postoper-ative infection and postoperative pneumonia, and the length oftime of blood storage as another factor modulating risk (170174). In one prospective randomized control trial the use ofleukocyte-depleted red blood cell transfusions resulted in a re-duced incidence of postoperative infections, and specifically areduced incidence of pneumonia in patients undergoing colo-rectal surgery (172). Routine red blood cell transfusion should beconducted with a restricted transfusion trigger policy. Whetherleukocyte-depleted red blood cell transfusions will further re-duce the incidence of pneumonia in broad populations of pa-tients at risk remains to be determined.

Hyperglycemia, relative insulin deficiency, or both may di-rectly or indirectly increase the risk of complications and pooroutcomes in critically ill patients. van den Berghe and coworkersrandomized surgical intensive care unit patients to receive eitherintensive insulin therapy to maintain blood glucose levels be-tween 80 and 110 mg/dl or to receive conventional treatment(175). The group receiving intensive insulin therapy had reducedmortality (4.6 versus 8%, p 0.04) and the difference was greaterin patients who remained in the intensive care unit more than5 days (10.6 versus 20.2%, p 0.005). When compared with thecontrol group, those treated with intensive insulin therapy hada 46% reduction of bloodstream infections, decreased frequencyof acute renal failure requiring dialysis by 41%, fewer antibiotictreatment days, and significantly shorter length of mechanicalventilation and ICU stay. Although the same degree of benefitmay not be seen among patients with VAP as in other popula-tions, aggressive treatment of hyperglycemia has both theoreticaland clinical support.

Major Points and Recommendations for ModifiableRisk Factors

General prophylaxis.

1. Effective infection control measures: staff education, com-pliance with alcohol-based hand disinfection, and isolationto reduce cross-infection with MDR pathogens should beused routinely (Level I) (3, 93, 100, 110, 111).

2. Surveillance of ICU infections, to identify and quantifyendemic and new MDR pathogens, and preparation oftimely data for infection control and to guide appropriate,antimicrobial therapy in patients with suspected HAP orother nosocomial infections, are recommended (Level II)(3, 92, 93, 100, 110113).

Intubation and mechanical ventilation.

1. Intubation and reintubation should be avoided, if possible,as it increases the risk of VAP (Level I) (3, 12, 93, 94,114).

2. Noninvasive ventilation should be used whenever possiblein selected patients with respiratory failure (Level I) (18,20, 115119).

3. Orotracheal intubation and orogastric tubes are preferredover nasotracheal intubation and nasogastric tubes to pre-vent nosocomial sinusitis and to reduce the risk of VAP,although direct causality has not been proved (Level II)(3, 93, 94, 109, 125).

4. Continuous aspiration of subglottic secretions can reducethe risk of early-onset VAP, and should be used, if avail-able (Level I) (97, 128, 130).

5. The endotracheal tube cuff pressure should be maintainedat greater than 20 cm H2O to prevent leakage of bacterialpathogens around the cuff into the lower respiratory tract(Level II) (98, 126).

6. Contaminated condensate should be carefully emptiedfrom ventilator circuits and condensate should be pre-vented from entering either the endotracheal tube or in-line medication nebulizers (Level II) (98, 131, 132).

7. Passive humidifiers or heatmoisture exchangers decreaseventilator circuit colonization, but have not consistentlyreduced the incidence of VAP, and thus they cannot beregarded as a pneumonia prevention tool (Level I) (135139).

8. Reduced duration of intubation and mechanical ventila-tion may prevent VAP and can be achieved by protocolsto improve the use of sedation and to accelerate weaning(Level II) (93, 120122, 124).

9. Maintaining adequate staffing levels in the ICU can reducelength of stay, improve infection control practices, andreduce duration of mechanical ventilation (Level II) (121124).

Aspiration, body position, and enteral feeding.

1. Patients should be kept in the semirecumbent position(3045 ) rather than supine to prevent aspiration, espe-cially when receiving enteral feeding (Level I) (140144).

2. Enteral nutrition is preferred over parenteral nutrition toreduce the risk of complications related to central intrave-nous catheters and to prevent reflux villous atrophy of theintestinal mucosa that may increase the risk of bacterialtranslocation (Level I) (3, 93, 145, 146).

Modulation of colonization: oral antiseptics and antibiotics.

1. Routine prophylaxis of HAP with oral antibiotics (selec-tive decontamination of the digestive tract or SDD), withor without systemic antibiotics, reduces the incidence ofICU-acquired VAP, has helped contain outbreaks of

396 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 171 2005

MDR bacteria (Level I), but is not recommended for rou-tine use, especially in patients who may be colonized withMDR pathogens (Level II) (149154, 156159, 161164,176).

2. Prior administration of systemic antibiotics has reducedthe risk of nosocomial pneumonia in some patient groups,but if a history of prior administration is present at the timeof onset of infection, there should be increased suspicionof infection with MDR pathogens (Level II) (157159,161164).

3. Prophylactic administration of systemic antibiotics for 24hours at the time of emergent intubation has been demon-strated to prevent ICU-acquired HAP in patients withclosed head injury in one study, but its routine use is notrecommended until more data become available (Level I)(167).

4. Modulation of oropharyngeal colonization by the use oforal chlorhexidine has prevented ICU-acquired HAP inselected patient populations such as those undergoing cor-onary bypass grafting, but its routine use is not recom-mended until more data become available (Level I) (148).

5. Use daily interruption or lightening of sedation to avoidconstant heavy sedation and try to avoid paralytic agents,both of which can depress cough and thereby increase therisk of HAP (Level II) (120).

Stress bleeding prophylaxis, transfusion, and hyperglycemia.

1. Comparative data from randomized trials suggest a trendtoward reduced VAP with sucralfate, but there is a slightlyhigher rate of clinically significant gastric bleeding, com-pared with H2 antagonists. If needed, stress bleeding pro-phylaxis with either H2 antagonists or sucralfate is accept-able (Level I) (99104, 155, 177179).

2. Transfusion of red blood cell and other allogeneic bloodproducts should follow a restricted transfusion trigger pol-icy; leukocyte-depleted red blood cell transfusions can helpto reduce HAP in selected patient populations (Level I)(169174).

3. Intensive insulin therapy is recommended to maintain se-rum glucose levels between 80 and 110 mg/dl in ICU pa-tients to reduce nosocomial blood stream infections, dura-tion of mechanical ventilation, ICU stay, morbidity, andmortality (Level I) (175).

DIAGNOSTIC TESTING

Diagnostic testing is ordered for two purposes: to define whethera patient has pneumonia as the explanation for a constellation ofnew signs and symptoms and to determine the etiologic pathogenwhen pneumonia is present. Unfortunately, currently availabletools cannot always reliably provide this information.

The diagnosis of HAP is suspected if the patient has a radio-graphic infiltrate that is new or progressive, along with clinicalfindings suggesting infection, which include the new onset offever, purulent sputum, leukocytosis, and decline in oxygenation.When fever, leukocytosis, purulent sputum, and a positive cul-ture of a sputum or tracheal aspirate are present without a newlung infiltrate, the diagnosis of nosocomial tracheobronchitisshould be considered (180). When this definition has been ap-plied to mechanically ventilated patients, nosocomial tracheo-bronchitis has been associated with a longer length of ICU stayand mechanical ventilation, without increased mortality (180).Antibiotic therapy may be beneficial in this group of patients(180, 181). In one prospective randomized trial of intubatedpatients with community-acquired bronchial infection, the use

of antibiotic therapy led to a reduced incidence of subsequentpneumonia and mortality (181).

The diagnosis of HAP is difficult, and most studies of nonintu-bated patients have involved clinical diagnosis, with sputum cul-ture, but bronchoscopy has been used less often, making the reliabil-ity of the bacteriologic information uncertain and the specificityof the diagnosis undefined (182). The accuracy of the clinicaldiagnosis of VAP has been investigated on the basis of autopsyfindings or quantitative cultures of either protected specimenbrush (PSB) or bronchoalveolar lavage (BAL) samples as thestandard for comparison (183186). Some studies have investi-gated the accuracy of a single clinical finding, whereas othersincluded multiple criteria in their definition of pneumonia. Thesestudies indicate that the diagnostic criteria of a radiographicinfiltrate and at least one clinical feature (fever, leukocytosis,or purulent tracheal secretions) have high sensitivity but lowspecificity (especially for VAP). Combinations of signs and symp-toms may increase the specificity. A study in which the diagnosticstandard was histology plus positive microbiologic cultures ofimmediate postmortem lung samples, the presence of chest infil-trates, plus two of three clinical criteria resulted in 69% sensitiv-ity and 75% specificity (187). When the three clinical variableswere used the sensitivity declined, whereas the use of only onevariable led to a decline in specificity.

For patients diagnosed with ARDS, suspicion of pneumoniashould be high and the presence of only one of the three clinicalcriteria described should lead to more diagnostic testing (188).A high index of suspicion should also be present in patients whohave unexplained hemodynamic instability or deterioration ofblood gases during mechanical ventilation. In the absence of anyof these findings, no further investigations are required. Theincidence of colonization in hospitalized patients in general andeven more in patients requiring endotracheal intubation is high(107). Antibiotic treatment of simple colonization is stronglydiscouraged. Routine monitoring of tracheal aspirate cultures toanticipate the etiology of a subsequent pneumonia has also beenfound to be misleading in a significant percentage of cases (189).

Although these criteria should raise suspicion of HAP, con-firmation of the presence of pneumonia is much more difficult,and clinical parameters cannot be used to define the microbio-logic etiology of pneumonia. The etiologic diagnosis generallyrequires a lower respiratory tract culture, but rarely may bemade from blood or pleural fluid cultures. Respiratory tractcultures can include endotracheal aspirates, BAL or PSB speci-mens. Overall, the sensitivity of blood cultures is less than 25%,and when positive, the organisms may originate from an extra-pulmonary source in a large percentage, even if VAP is alsopresent (190). Although an etiologic diagnosis is made from arespiratory tract culture, colonization of the trachea precedesdevelopment of pneumonia in almost all cases of VAP, and thusa positive culture cannot always distinguish a pathogen from acolonizing organism. However, a sterile culture from the lowerrespiratory tract of an intubated patient, in the absence of arecent change in antibiotic therapy, is strong evidence that pneu-monia is not present, and an extrapulmonary site of infectionshould be considered (191, 192). In addition, the absence ofMDR microorganisms from any lower respiratory specimen inintubated patients, in the absence of a change in antibioticswithin the last 72 hours, is strong evidence that they are notthe causative pathogen. The time course of clearance of thesedifficult-to-treat microorganisms is usually slow, so even in theface of a recent change in antibiotic therapy sterile cultures mayindicate that these organisms are not present (193). For thesereasons, a lower respiratory tract sample for culture should becollected from all intubated patients when the diagnosis of pneu-monia is being considered. The diagnostic yield and negative

American Thoracic Society Documents 397

predictive value of expectorated sputum in nonintubated pa-tients have not been determined.

Major Points and Recommendations for Diagnosis

1. All patients should have a comprehensive medical historyobtained and undergo physical examination to define theseverity of HAP, to exclude other potential sources ofinfection, and to reveal the presence of specific conditionsthat can influence the likely etiologic pathogens (Level II)(9, 16, 194).

2. All patients should have a chest radiograph, preferablyposteroanterior and lateral if not intubated, as portablechest radiographs have limited accuracy. The radiographcan help to define the severity of pneumonia (multilobaror not) and the presence of complications, such as effu-sions or cavitation (Level II) (5, 195).

3. Purulent tracheobronchitis may mimic many of the clini-cal signs of HAP and VAP, and may require antibiotictherapy, but prospective, randomized trials are needed(Level III) (180). Tracheal colonization is common inintubated patients, but in the absence of clinical findingsis not a sign of infection, and does not require therapyor diagnostic evaluation (Level II) (40, 107).

4. Arterial oxygenation saturation should be measured in allpatients to determine the need for supplemental oxygen.Arterial blood gas should be determined if concern existsregarding either metabolic or respiratory acidosis, andthis test generally is needed to manage patients who re-quire mechanical ventilation. These results, along with otherlaboratory studies (complete blood count, serum electro-lytes, renal and liver function), can point to the presenceof multiple organ dysfunction and thus help define theseverity of illness (Level II) (38, 188).

5. All patients with suspected VAP should have blood cul-tures collected, recognizing that a positive result can indi-cate the presence of either pneumonia or extrapulmonaryinfection (Level II) (190).

6. A diagnostic thoracentesis to rule out a complicating em-pyema or parapneumonic effusion should be performedif the patient has a large pleural effusion or if the patientwith a pleural effusion appears toxic (Level III) (5).

7. Samples of lower respiratory tract secretions should beobtained from all patients with suspected HAP, andshould be collected before antibiotic changes. Samplescan include an endotracheal aspirate, bronchoalveolarlavage sample, or protected specimen brush sample(Level II) (183, 184, 192, 196, 197).

8. In the absence of any clinical suspicion of HAP or nosoco-mial tracheobronchitis, no respiratory tract cultures shouldbe obtained (Level III).

9. A sterile culture of respiratory secretions in the absenceof a new antibiotic in the past 72 hours virtually rulesout the presence of bacterial pneumonia, but viral orLegionella infection is still possible (Level II) (192). Ifthese patients have clinical signs of infection, an extrapul-monary site of infection should be investigated (Level II)(190, 198).

10. For patients with ARDS, for whom it is difficult to demon-strate deterioration of radiographic images, at least one ofthe three clinical criteria or other signs of pneumonia, suchas hemodynamic instability or deterioration of blood gases,should lead to more diagnostic testing (Level II) (38).

DIAGNOSTIC STRATEGIES AND APPROACHES

Because clinical suspicion of HAP/VAP is overly sensitive, fur-ther diagnostic strategies are required for optimal management.

The goals of diagnostic approaches in patients with suspectedHAP are to identify which patients have pulmonary infection;to ensure collection of appropriate cultures; to promote the useof early, effective antibiotic therapy, while allowing for stream-lining or de-escalation when possible; and to identify patientswho have extrapulmonary infection (Figure 1). The committeeconsidered two different approaches to management, a clinicalstrategy and a bacteriologic strategy, and have incorporated fea-tures from both in the final recommendations.

Clinical Strategy

When the clinical approach is used, the presence of pneumoniais defined by new lung infiltrate plus clinical evidence that theinfiltrate is of an infectious origin. The presence of a new orprogressive radiographic infiltrate plus at least two of three clini-cal features (fever greater than 38 C, leukocytosis or leukopenia,and purulent secretions) represents the most accurate combina-tion of criteria for starting empiric antibiotic therapy (187). Al-though sensitivity for the presence of pneumonia is increased ifonly one criterion is used, this occurs at the expense of specificity,leading to significantly more antibiotic treatment. Requiring allthree clinical criteria is too insensitive and will result in manypatients with true pneumonia not receiving therapy.

The etiologic cause of pneumonia is defined by semiquantita-tive cultures of endotracheal aspirates or sputum with initialmicroscopic examination. Tracheal aspirate cultures consistentlygrow more microorganisms than do invasive quantitative cul-tures, and most microbiology laboratories report the results ina semiquantitative fashion, describing growth as light, moderate,or heavy. In general, it is rare that a tracheal aspirate culturedoes not contain the pathogen(s) found in invasive quantitativecultures (191, 199, 200). Gram staining of polymorphonuclearleukocytes and macrophages and careful examination of themorphology of any bacteria found to be present, may improvediagnostic accuracy when correlated with culture results (201,202). Conversely, a negative tracheal aspirate (absence of bacte-ria or inflammatory cells) in a patient without a recent (within72 hours) change in antibiotics has a strong negative predictivevalue (94%) for VAP (203). A reliably performed Gram stainof tracheal aspirates has been demonstrated to result in a lowincidence of inappropriate therapy when used to guide initialempiric antibiotic therapy (9, 198).

The clinical strategy emphasizes prompt empiric therapy forall patients suspected of having HAP. The driving force behindthis strategy is the consistent finding that delay in the initiationof appropriate antibiotic therapy for patients with HAP is associ-ated with increased mortality (37, 112, 204). The selection ofinitial antibiotic therapy is based on risk factors for specificpathogens, modified by knowledge of local patterns of antibioticresistance and organism prevalence. Therapy is modified on thebasis of the clinical response on Days 2 and 3, and the findingsof semiquantitative cultures of lower respiratory tract secretions.This approach requires no specialized microbiologic methods,and all patients suspected of having pneumonia are treated. Thisavoids the problem of not treating some infected individuals.Use of an ICU-specific, broad-spectrum empiric therapy regi-ment can reduce the incidence of inappropriate initial therapyto less than 10% (198, 205).

The major limitation to the clinical approach is that it consis-tently leads to more antibiotic therapy than when therapy deci-sions are based on the findings (microscopy and quantitativecultures) of invasive (bronchoscopic) lower respiratory tract sam-ples (198). The clinical approach is overly sensitive, and patientscan be treated for pneumonia when another noninfectious pro-cess is responsible for the clinical findings. These processes mayinclude congestive heart failure, atelectasis, pulmonary thrombo-

398 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 171 2005

Figure 1. Summary of the managementstrategies for a patient with suspectedhospital-acquired pneumonia (HAP), ven-tilator-associated pneumonia (VAP), orhealthcare-associated pneumonia (HCAP).The decision about antibiotic discontinu-ation may differ depending on the typeof sample collected (PSB, BAL, or endotra-cheal aspirate), and whether the resultsare reported in quantitative or semiquan-titative terms (see text for details).

embolism, pulmonary drug reactions, pulmonary hemorrhage,or ARDS. Even if the patient has pneumonia, reliance on semi-quantitative cultures, which may not reliably separate true patho-gens from colonizers, can lead to either more or broader spec-trum antibiotic therapy than with a quantitative approach (198).These cultures have their greatest value if they are negative andthe patient has not received new antibiotics within the past 72hours. One other concern is that reliance on nonquantitative cul-tures could lead to a failure to recognize extrapulmonary infectionat an early time point.

In an effort to improve the specificity of clinical diagnosis,Pugin and coworkers developed the clinical pulmonary infectionscore (CPIS), which combines clinical, radiographic, physiologi-cal (PaO2/FiO2), and microbiologic data into a single numericalresult (206). When the CPIS exceeded 6, good correlation withthe presence of pneumonia, as defined by quantitative culturesof bronchoscopic and nonbronchoscopic BAL specimens, wasfound. However, in a subsequent study that used histology plusimmediate postmortem quantitative lung cultures as the refer-ence standard, the CPIS had a sensitivity of 77% and a specificityof 42% (187). One prospective study evaluated 79 episodes ofsuspected VAP, using the CPIS, and compared the findings withdiagnoses established by BAL culture. Overall, the sensitivityand specificity of the score were low, although it improved if aGram stain of a deep respiratory tract culture was added to theevaluation (201).

The original description of the CPIS required microbiologicdata, and thus could not be used to screen for HAP. Singh andcolleagues used a modified CPIS that did not rely on culturedata to guide clinical management (207). Another approach wasto calculate the score by using the results of a Gram stain of aBAL specimen or blind protected telescoping catheter sample,and score the findings as either positive or negative. Using thisapproach, the CPIS for patients with confirmed VAP was signifi-cantly higher than the value for nonconfirmed VAP (201).

If a clinical strategy is used, reevaluation of the decision touse antibiotics based on serial clinical evaluations, by Day 3 or

sooner, is necessary, because patients who are improving willhave signs of a good clinical response by this time point (193,208). Singh and coworkers have shown that some patients witha low clinical suspicion of VAP (CPIS of 6 or less) can haveantibiotics safely discontinued after 3 days, if the subsequentcourse suggests that the probability of pneumonia is still low(207). The modified CPIS used by Singh and coworkers appearsto be an objective measure to define patients who can receivea short duration of therapy.

Major points and recommendations for the clinical strategy.

1. A reliable tracheal aspirate Gram stain can be used to directinitial empiric antimicrobial therapy and may increase thediagnostic value of the CPIS (Level II) (191, 199, 201, 209).

2. A negative tracheal aspirate (absence of bacteria or in-flammatory cells) in a patient without a recent (within 72hours) change in antibiotics has a strong negative pre-dictive value (94%) for VAP and should lead to a searchfor alternative sources of fever (Level II) (203).

3. The presence of a new or progressive radiographic infil-trate plus at least two of three clinical features (fevergreater than 38 C, leukocytosis or leukopenia, and puru-lent secretions) represent the most accurate clinical criteriafor starting empiric antibiotic therapy (Level II) (187).

4. If a clinical strategy is used, reevaluation of the decisionto use antibiotics based on the results of semiquantitativelower respiratory tract cultures and serial clinical evalua-tions, by Day 3 or sooner, is necessary (Level II) (193,205, 207, 208).

5. A modified CPIS of 6 or less for 3 days, proposed by Singhand coworkers, is an objective criterion to select patientsat low risk for early discontinuation of empiric treatmentof HAP, but still requires validation in patients with moresevere forms of VAP (Level I) (201, 207).

Bacteriologic Strategy

The bacteriologic strategy uses quantitative cultures of lowerrespiratory secretions (endotracheal aspirates, BAL or PSB speci-

American Thoracic Society Documents 399

mens collected with or without a bronchoscope) to define boththe presence of pneumonia and the etiologic pathogen. Growthabove a threshold concentration is required to diagnose VAP/HAP and to determine the causative microorganism(s). Growthbelow the threshold is assumed to be due to colonization orcontamination. The bacteriologic strategy has been used to guidedecisions about whether to start antibiotic therapy, which patho-gens are responsible for infection, which antimicrobial agents touse, and whether to continue therapy.

Because the bacteriologic approach emphasizes avoidance ofthe problem of overtreatment with antibiotics by trying to sepa-rate colonizing from infecting pathogens, use of this method hasconsistently led to finding fewer microorganisms growing abovethe diagnostic threshold than are present in nonqualitative cul-tures of tracheal aspirates. When therapy decisions have beenbased on these data, fewer patients have been treated with antibi-otics, and a potentially narrower spectrum of therapy was used,compared with the clinical approach (198, 210). Quantitativecultures have been demonstrated to have good diagnostic utilityfor the presence of pneumonia, especially in patients with a lowor equivocal clinical suspicion of infection (211, 212).

The major concern with the bacteriologic approach is that afalse negative culture can lead to a failure to treat either a specificpatient or a specific pathogen, and that the results are not alwaysconsistent and reproducible (213215). A major factor causingfalse negative quantitative cultures is a recent starting of orchange in antibiotic therapy, especially in the preceding 24 hours,but up to 72 hours (192, 212). Therefore, ideally all quantitativecultures should be obtained before any antibiotic manipulation.This may not be possible in all situations, and in this setting achange in the diagnostic threshold may be helpful (212). ForBAL, use of a threshold 10-fold lower than usual may avoidsome false negative results in patients given antibiotics beforetesting. However, some patients with pneumonia will have cul-ture growth below threshold, even without recent antibioticchanges, especially in early forms of infection (215217).

Methodologic issues involved in the inconsistent results ofpublished studies have been summarized in a meta-analysis(184). These include the evaluation of patients who did notmeet recognized clinical criteria for the presence of pneumonia;prolonged time between the performance of a diagnostic testand the collection of confirmatory histopathologic information;inclusion of patients who had received antibiotic therapy beforediagnostic testing, often without correcting for the duration ofantibiotic therapy; and inclusion of patients studied by BALperformed with insufficient lavage volume (less than 140 ml). Amajor problem with all studies of HAP diagnosis is the absence ofa gold standard with which diagnostic results can be compared.Even the best criteria for the presence of pneumonia, immediatepostmortem histologic evaluation with microbiologic confirma-tion of infection, can be inaccurate. In addition, only a subgroupof patients with severe VAP is included in these types of studies.

In a prospective study of 148 patients receiving mechanicalventilation and in whom infectious pneumonia was suspected,Gibot and coworkers used a rapid immunoblot technique onBAL fluid, and found that levels of soluble triggering receptorexpressed on myeloid cells (sTREM-1) were the strongest inde-pendent predictor of pneumonia (odds ratio, 41.5) (218). Whencommercially available, this marker, coupled with the classicclinical criteria and results of microbiologic cultures, may be avaluable tool with which to increase the specificity and maintainthe sensitivity of HAP diagnosis (197).

Histologic data have demonstrated several characteristics ofVAP pertinent to diagnostic testing, such as the finding thatthe process is often multifocal, frequently involving both lungs,generally in the posterior and lower segments (191, 215, 216).

Postmortem studies have also demonstrated that VAP is oftenin multiple different phases of evolution at different sites atthe same time (216). Prior antibiotic therapy can influence thenumber of bacteria found in lung tissue, and patients who havedied in spite of prolonged therapy are likely to have organismsresistant to the agents used, whereas patients started on therapywithin 24 (and up to 72) hours may have negative cultures,especially if the therapy is adequate (192). The multifocal natureof VAP suggests that BAL and endotracheal aspirates can pro-vide more representative samples than the protected specimenbrush (PSB), which samples only a single bronchial segment.Because of the diffuse bilateral nature of VAP and predomi-nance in dependent lung segments, blind BAL and PSB maybe as accurate as bronchoscopic sampling in some patients (219).

Another issue with the bacteriologic strategy is that cultureresults are not available immediately. Ancillary tests such asGiemsa stain for intracellular microorganisms, Gram stain, ordifferential cell counts can be used to increase the likelihood ofa subsequent positive culture and can be used to guide the needfor antibiotic therapy before culture results. In some studies,this approach has led to less use of antibiotics with no adverseoutcomes, and a tendency to improved mortality (198, 201). Notall investigators agree about the safety of withholding therapyuntil quantitative results are available, and are positive, or towithdrawing therapy if cultures are negative, after empiricallystarting antimicrobials for suspected infection (198, 220222).Clinically, these decisions have been guided by the degree ofcertainty of the diagnosis of pneumonia at the time of testing(pretest probability), and on the severity of illness of the patient(198). Thus, most investigators agree that patients with signs ofinfection, who are clinically unstable, should receive therapy,regardless of the initial bronchoscopic findings (198, 212).

The diagnostic threshold to discriminate infection from colo-nization varies with the technique used, and possibly by theclinical probability of infection (212). The threshold may belowered if the patient has recently had a change in antibiotictherapy or if the probability of infection is high. Endotrachealaspirates can be cultured quantitatively, and with a threshold of106 cfu/ml or more the sensitivity of this method for the presenceof pneumonia has varied from 38 to 82%, with a mean of 76 9%, and with a specificity ranging from 72 to 85%, with a meanof 75 28% (209).

Bronchoscopic BAL studies have typically used a diagnosticthreshold of 104 or 105 cfu/ml. Samples contaminated by upperairway secretions, as reflected by a high percentage of squamousepithelial cells, should be used with caution. A few studies haveshown the technique to be reproducible, but not all bacteria arerecovered above the diagnostic threshold when the procedurehas been repeated in the same patient at the same site (223).An evidence-based review of 23 prospective studies of BAL insuspected VAP showed a sensitivity of 4293%, with a mean of73 18% (186), and a specificity of 45100%, with a mean of82 19%. In 12 studies, the detection of intracellular organismsin 25% of recovered cells was used to diagnose pneumonia,with a mean sensitivity of 69 20% and a specificity of 75 28% (186). The advantage of looking for intracellular organismsis the ability to obtain information of high predictive value in arapid time frame, without waiting for the results of cultures todefine the presence of pneumonia, although not the specificidentity of the etiologic pathogen.

Quantitative cultures of PSB samples have used a diagnosticthreshold of 103 cfu/ml or more. The quality of the PSB sampleis difficult to measure, and the reproducibility is not exact, withas many as 25% of results on different sides of the diagnosticthreshold, when comparing two samples collected from the samesite in the same patient (183). The sensitivity and specificity range

400 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 171 2005

from 33 to 100% (mean, 66 19%) and from 50 to 100% (mean,90 15%). PSB appears to be more specific than sensitive forthe presence of pneumonia, and a positive result greatly increasesthe likelihood of pneumonia being present (186).

The bacteriologic strategy does require specialized laboratoryand clinical skills. In many clinical settings, bronchoscopy isnot immediately available, especially in the evenings, and thecollection of blind, nonbronchoscopic samples is an appealingalternative. Blind sampling can be done by BAL or PSB, or ablind bronchial suction sample can be taken. When BAL samplesare obtained nonbronchoscopically, the threshold varies by tech-nique and may be different from that of bronchoscopic BAL.The sensitivities of blind bronchial suction, blind mini-BAL, andblind PSB are 7497, 63100, and 5886%, respectively (224).The specificity of these methods has varied from 74 to 100%for blind bronchial suction, from 66 to 96% for mini-BAL, andfrom 71 to 100% for blind PSB. In general, these techniquesprovide data similar to those of samples collected bronchoscopi-cally, although with a trend toward more cultures above thediagnostic threshold. Side effects should be no greater and possi-bly less than with bronchoscopically collected samples.

Recommendation for the bacteriologic strategy. Quantitativecultures can be performed on endotracheal aspirates or samplescollected either bronchoscopically or nonbronchoscopically, andeach technique has its own diagnostic threshold and methodo-logic limitations. The choice of method depends on local expertise,experience, availability, and cost (Level II) (197, 198, 214, 224).

Recommended Diagnostic Strategy

To date, several decision analyses, one retrospective study, andfour prospective studies have evaluated the impact of diagnosticstrategies on the use of antibiotics and the outcomes of patientswith suspected VAP (198, 211, 212, 220222, 225). In three ran-domized single-center studies, no differences in mortality werefound when invasive techniques (PSB and/or BAL) were com-pared with either quantitative or semiquantitative endotrachealaspirate culture techniques (220222). However, these studiesincluded few patients (51, 76, and 88, respectively) and antibioticswere continued in all patients, even those with negative cultures,thereby negating one of the potential advantages of the bacterio-logic strategy. In fact, several prospective studies have concludedthat antibiotics can be safely stopped in patients with negativequantitative cultures, with no adverse impact on mortality (15,198, 226).

One large, prospective randomized trial did show an advan-tage to the quantitative bronchoscopic approach, when com-pared with a clinical approach in a multicenter study of 413patients suspected of having HAP (198). Compared with patientsmanaged clinically, those receiving invasive management had alower mortality rate on Day 14 (16 and 25%; p 0.02), butnot on Day 28, and lower mean sepsis-related organ failureassessment scores on Days 3 and 7 (p 0.04). At 28 days, thequantitative culture group had significantly more antibiotic-freedays (11 9 versus 7 7 days; p 0.001), but only a multivariateanalysis showed a significant difference in mortality (hazard ra-tio, 1.54; 95% confidence interval, 1.10 to 2.16; p 0.01). Onestrength of the study was that a high percentage of patients inboth arms received adequate initial antibiotics, although morepatients in the invasive group received adequate therapy thanin the clinical group, and the impact of this difference on theobserved mortality differences was uncertain. Another impor-tant consequence of quantitative culture results was that thepresence of clinical signs of infection in patients with negativecultures was often an indication that an extrapulmonary siteof infection was present. This study clearly showed that thequantitative approach could be applied safely, leading to less

antibiotic use, and potentially reducing mortality. In the trial,about 10% of the patients managed with a quantitative strategyreceived antibiotic therapy regardless of bronchoscopic findingsbecause of the presence of clinical instability and signs of sepsis.

Considering the available methods for diagnostic testing andthe goals of using appropriate therapy in a timely manner, with-out overusing antibiotics, the committee has combined featuresof the clinical and bacteriologic approach into an algorithmshown in Figure 1. The decision to discontinue antibiotics, usingthis algorithm, may differ depending on the type of respiratorytract sample that is collected and whether the culture results arereported in quantitative or semiquantitative terms. Advocatesof the bacteriologic approach support the discontinuation ofantibiotics in clinically stable patients whose quantitative cultureresults of deep lung samples (BAL or PSB) f