Impact of Previous Antimicrobial Therapy on the Etiology and Outcome of Ventilator-associated Pneumonia

Impact of Previous Antimicrobial Therapy on the Etiology and Outcome of Ventilatorassociated Pneumonia* jordi Rello, M .D. , Ph .D. ; Vicenr Ausina, .\J .D., Ph.D.; Maite Ricart, R.N.; joan Castella , M.D., Ph.D. ; and Guillem Prats, M.D., Ph .D.

Objective: To define the influence of prior antibiotic use on

the etiology and mortality of ventilator-associated pneumonia (VAP). Setting: A university hospital medical-surgical ICU. Design: Prospective clinical study. Methods: Over a 35-month period, we prospectively studied 129 consecutive episodes of VAP. Etiologic diagnosis was established using a protected specimen brush and quantitative culture techniques. We examined prognostic factors by univariate and multivariate analyses using a statistical software package (SPSS). Results: The rate of VAP caused by Gram-positive cocci or Haemophilus in.fluenzae was statistically lower (p<0.05) in the patients who had received antibiotics previously, while the rate of VAP caused by PseutWfflQfUUI aeruginosa was statistically higher (p<0.01). Patients died of causes directly related to the infection in 18 (14.0 percent) episodes, P aeruginosa being isolated in 9 of these fatal cases. Indeed, we found that 27.7 percent (15/54) of patients who had received prior antimicrobial therapy before the onset of pneumonia died, compared with only 4.0 percent (3n5) of

Nosocomial pneumonia occurs in 0.5 to 2 percent of patients hospitalized 1.z and is the leading cause of death from nosocomial infection. a.v Moreover, intubated patients receiving mechanical ventilation (MV) have an important increase in risk for pneumonia,5 with the incidence of ventilator-associated pneumonia (VAP) ranging between 9 and 68 percent, depending on the populat.i!>ll studied . 1-" -Y The impact on mortality is the most important aspect of any clinical problem. Identification of factors

For editorial comment see page 993 influencing mortality offers the possibility of improving prognosis through the identification of patients with "high risk of death" and the modification of some of these factors by medical intervention. The use of multiple regression analysis helps to identify independent prognostic factors for a disease as complex and multifactorial as nosocomial pneumonia.

those who did not. In the univariate analysis, the variables significantly associated with attributable mortality were age older than 45 years, use of corticosteroids, presence of shock, hospital day of VAP over 9, antecedent COPD, and a prior antibiotic use. A step-forward logistic regression analysis defined only prior antibiotic use (p<0.0001, OR= 9.2) as significantly influencing the risk of death from VAP. The same result was obtained when severity was included in the model. However, prior antibiotic use entirely dropped out as a significant risk factor when the etiologic agent was included in the regression equation. Conclusions: Distribution of infecting microorganisms responsible for VAP differs in patients who ~eceived prior antimicrobial therapy, and this factor determines a higher mortality rate. We suggest a restrictive antibiotic policy in mechanically ventilated patients with the purpose of reducing the risk of death from VAP. (Chest 1993; 104:1230-35) MV = mechanical ventilation; VAP =ventilator-associated pneumonia

A high percentage of intubated patients have concomitant infections, which often leads to broad-spectrum antimicrobial therapy. Moreover, some authors11~12 have recommended the use of prophylactic antibiotics in intubated patients, because of the high morbidity and mortality associated with VAP. In contrast, Fagon et aJH have recently reported that previous antimicrobial therapy prior to the onset of VAP may be associated with higher mortality. However, to our knowledge, the influence of this factor in the prognosis of VAP has not been studied previously using a multivariate analysis. Theref(>re, we performed a prospective study of 129 consecutive episodes of VAP to determine the influence of a variety of epidemiologic and clinical variables on the outcome, by means of univariate and multivariate analyses. Our specific goal was to provide current information on the influence of antimicrobial therapy before the onset of pneumonia on the frequency of microbial etiologies and subsequent mortality. METHODS

*From the Intensive Care . Microhiolo~·. and Respiratory Depart· ments. Hospital de Ia S. Creu i S. Pau. Universitat Autonoma de Barcelona, Barcelona, Spain. Manuscript received Oetoher 13, 1992; revision accepted Fehruary 23, 1993

1230

Pormlation Study Our intensi,-e care unit (ICL1) is a 16-bed medical-sur~ieal unit, with an avera~e of600 admissions per year, in a 1,000-bed university hospital that serves as hoth a referral center and a first-line hospital.

Impact of Previous Antimicrobial Therapy in Ventilator-associated Pneumonia (Rallo at a/)

Table 1- Summary of Indications for Antibiotic Therapy Indication

Cases, No.

Previous nos
19 10 8 5 4

3 2 3

*Prolonged for more than 48 h. tNFI =nosocomial febrile illness. :f;Tetanus, peritonitis, and fever of unknown origin. It is located in an urban area with a population of over 3 million. During a 35-month period, all patients who developed pneumonia in our ICU were considered eligible and were prospectively evaluated by one of the investigators (J.R.). We did not use any selective decontamination regimen. A diagnosis of pneumonia was ~'msidered when new and persistent pulmonary infiltrates not otherwise explained appeared on chest radiographs. Moreover, at least two of the following criteria were also required: (1) fever 2:38°C; (2) leukocytosis 2:10,000/mm' ; and (3) purulent respiratory secretions. Pneumonia was mnsidered ventilator associated when its onset occurred after 48 h of MV and was thus judged not to have been incubated before starting MV. To analyze factors that might influence outmme, the following variables were remrded: age, sex , severity, prior trauma or surgery, underlying disease, the presence of COPD, diabetes, renal failure requiring dialysis, cardiopathy, persistence of coma during MV, the duration of MV prior to the development ofVAP, corticosteroid use, and presence of shock (either prior as subsequent to the diagnosis of VAP). All antimicrobials received for more than 48 h during the 10 days preceding the episode of VAP were recorded . Antibiotics administered for less than 48 h , mainly for perioperative prophylaxis, were not considered and the corresponding episodes of VAP were classified as "free of exposition." Antecedent COPD was diagnosed using the standard criteria recommended by the American Thoracic Society." Coma was diagnosed when a Glasgow Coma Smre lower than 9 was obtained" for more than 24 h. Criteria for cortimsteroid use have been defined previously." Surgery or trauma was considered to he present if remrded within 2 weeks prior to the episode. Presence of cardiopathy was mnsidered if the patient had undergone surgery or if an episode of angina or cardiac failure was previously diagnosed. Diabetes was diagnosed if the patient received insulin previously to hospital admission. Shock was defined as a systolic blood pressure below 90 mm Hg accompanied by clinical signs of hypoperfusion, such as oliguria and peripheral vasoconstriction. Severity of underlying medical mnditions was stratified in three categories according to the criteria proposed by McCabe and Jackson•• and in five categories acmrding to NNIS ICU severity-of-illness clinical classification codes." Crude mortality included all deaths that occurred in the ICU among patients with VAP. Mortality was considered as attributable to the pulmonary infection if death ensued before any objective response to antimicrobial therapy or if the pulmonary infection was considered a mntributing factor to death in patients with additional conditions.•

Microbiology Fiberoptic bronchost'Opic examination using a protected specimen brush"·'" was perfi1rmed on each of these patients within the first 12 h after the development of a new pulmonary infiltrate, except on weekends. A telescoping canula brush with a distal polyethylene glycol occlusion (model BWF/IOn0/90; Meditech Inc,

Watertown, Mass) was inserted through the inner suction channel of the bronchoscope (model BF 10; Olympus Corp of America, New Hyde Park , NY). This protected brush was advanced and aseptically cut into a sterile tuhe t'mtaining 1 ml of sterile saline solution. Specimens were taken to the microhiology lahoratory immediately after collection. The vial was then vortexed vigorously for at least 60 s to suspend all material from the brush. Two serial 100-fold dilutions were carried out , and 0 . 1-ml aliquots of the original suspension and each dilution were inoculated on agar plates for aerobic and anaerohic culture; one agar plate with selective huffered charmal-yeast extract medium for isolation of Legionella species was included . ~· Bacterial t'Ount 2:1000 cfu/ml was the cutoff point to diagnose pulmonary infections, ac~"rding to the standards adopted in previous studies!' Bacterial identification and susceptibility testing were performed by standard methods."' Statistical Analysis Contingency tables were analyzed using the two-tailed x' test. Multivariate analysis was perfi1rmed using the stepwise logistic regression model of a software package (SPSS), using 0.05 as a limit for entering new terms. The cutoff point for those variables with more than two categories was chosen depending on the univariate analysis results. Odds ratios (OR) were calculated using standard methods.,, RESULTS

During the period of the study, 150 episodes of nosocomial pneumonia were identified in our ICU. One hundred twenty-nine of them met criteria ofVAP and were prospectively followed . The mean± SD age of these patients was 49.5 ± 20.3 years; 91 were men and 38 were women. Patients had received antimicrobial therapy for more than 48 h, in the 10 days preceding the onset of VAP, in 54 of these episodes. Indications for these antimicrobials are listed in Table 1. Table 2 shows which type of antibiotics were used prior to development of VAP, separated by etiologies, for episodes with etiologic diagnosis. Crude mortality was 34.1 percent. Patients died of causes directly related to the pulmonary infection in 18 (14.0 percent) episodes. All patients with VAP caused by Gram-positive cocci or Haemophilus irifluenzae recovered. Pseudomonas aeruginosa was the miTable 2-Class of Antibiotics Used Prior to Development of Pneumonia in 89 Episodes With Established Etiowgic Diagnosis* Microorganism

c

Staphylococcus aureus CNS Haemophilus injluen::.ae Pseudomonas aeruginosa Alcaligenes faecalis Serratia marcescens Proteus mirabilis Acinetobacter calcoaceticus Anaerobic flora Fungi

3 1 2 6 1 2 2 2 2 2

A

v

Cl

2

9

Other 2

5

6

3

2

1 7 2

2 2

*CNS =coagulase-negative staphylococci; C = cephalosporins; l=imipenem; A=aminoglycosides ; V=vancomyein; Cl =cloxacillin. CHEST I 104 I 4 I OCTOBER, 1993

1231

Table 3-Characteristics of Patients Who Died From VAP* Case/ Age . yr/Sex

Diagnosis

pATB

Microorganisms Aspergillus species Candida species P.w-udonumc~' ae,.uginosa P ae,.uginosa Se,.mtia IIUJ,.Cl'SCens P aen•ginosa Aspergillus species P ae,.uginosa

1/43/M

Heart transplant

Yes

2159/M 3/33/M

COPD Heart transplant

Yes Yes

4n6/M 5175/M 6/62/M

CET Cardiogenic slux:k CAP

'tes 'tes Yes

7fi0/M

COPD

'tes

COPD COPD Asthma Cardiac surgery Pancreatitis Septic shock Multiple trauma Neurosurgery Cardiac surgery Multiple trauma Thoracic surgery

't<·s Yes 'tes Yes 'tes Yes Yes Yes No No No

S

&74/M 9nl/F 10/46/M 11/65/M 12172/F 13/54/M 14121/M 15/48/F 16/68/M 17151/M 18f71/M

IIUJ ,.Cl'SCl'llS

Acinetobacta calcoaceticus P lll'I"Ugirwsa A cctlcoaceticus P aerogirwsa P aen•ginosa P aeroginosa Proteus mimbilis Uncertain Uncertain Uncertain 5 marcescens P aeroginosa

*YAP= ventilator-associated pneumonia: CET = cranioencephalic trauma; CAP= eommunity-acctuired pneumonia; COPD =chronic obstructive pulmonary disease; pATB =previous antibiotic therapy.

croorganism most frequently isolated among fatal episodes, being present in nine of them. Details of these 18 episodes are shown in Table 3. A total of 113 microorganisms were cultured from

protected specimen brushing in significant (~ 1000 cfu!ml) concentration. Twenty-three episodes were polymicrobial infections. Table 4 catalogues all episodes by whether or not receiving antibiotics, the number of positive cultures in each group, and the outcomes. Gram-positive cocci represented 49.1 percent of isolates in the group free of antimicrobial exposure, compared with 13.4 percent in the group with previous antibiotic therapy (p<0.05). Staphylococcus au reus accounted for 75.6 percent of all Grampositive cocci isolated, and the most frequent microorganism in patients without prior antibiotic use. All S aureus strains were penicillinase producing but sensitive to all other antibiotics tested. Haenwphilus influen::.ae was the second most frequent microorganism (27.8 percent of isolates) among patients without prior antimicrobial therapy but was an infrequent finding (5.7 percent of isolates) in episodes with previous antimicrobial exposure (p<0.05). In contrast, Gram-negative bacilli-excluding H injluen;:,ae-represented 69.2 percent of isolates among episodes with antibiotic exposure prior to the onset of VAP but only represented 19.6 percent of isolates in the opposite group. Indeed, P aeruginosa, the Gram-negative bacilli most often isolated, represented only 4.9 percent of isolates in patients without previous antibiotic exposure but represented 40.3 percent of isolates in patients subjected to prior antimicrobial treatment (p<0.01), being present in up to 50 percent of VAPs with etiologic diagnosis developed in this group of patients.

Table 4-Microorganisms Isolated in 129 Episocks ofVAP* Isolates Microorganism Gram-positive t"cci:j: StaJihylococcu., tiUil'Us Stn•ptococcus Jmeunwniae f:ntemcoccus faecalis C.>agulase-negative staphylo<.,lcci Gram-negative bacilli Hcwnwphilus injluen~e:j: Pseudcmwna., aen•girwsa§ Alcaligerws faecal is Sc·,.mtia nul,.c<•scens Protc'us mimbilis Acinetobcu:tc'l" calcoan•ticu.v f:sclwrichia coli Citmbacte,. fn•tmdii

pATB

No pATB

li>tal

6

22 5 2

2H 5

()

0

I 3 21 3

17 3 4 ()

4

()

3 0

2

2

5

()

Mortality. t No.

3 I

20 24 7 5 4 4 3

9

3 I

2

I

.t

Ana~rohic Hora

3

Fungi Candida speeies Aspergillus species Uncertain

I 2

0

13

27

2 2 40

I 2 3

*pATB =previous antihiotie use. tDirectly related to the ventilator-associated pneumonia. :j:p<0.05. §pmpare isolates in the group with pATB in respect to the group with no ATB.)

1232

Impact o1 Previous Antimicrobial Therapy in Ventilator-associated Pneumonia (Rella at el)

Table 5-Prognostic Factors for VAP: Vnimriate Analysis*

\'ariahle Age, yr ,;; 45 >45 Gender Male Female Undt•rlying eondition NF L'F/ HF Se,·erityll Codes A. B. C Codes D. E Traunw Yes No Surgery Yes No Prior antibiotic therapy )t-s No Corticosteroid use )t-s No Hospital day of \i\ I' ,;;9d >9d Dialll'tes )t-s No Renal failun· \t•s No Siwek )t-s 1'\o Coma )t-s l\;o Cardiopathy Yes No CO I'D anteeedt·nt \t•s 1\o

Episodt>s of\i\P. 1'/o. (%)+

Attrihntahlt• 1'\o. (%):(:

Odds Hatio§

.til (37.2)

Rl (62.7)

.1 (6 .2) IS (IR.S)

3.6

91 (70.5\ 311 (29.4)

IS (16 ..5\ .1 (7.9)

22 (17.0} 107 (R2.\J\

I (.t.S) 17 (1.5 .9)

:1.9

64 (49.6) 6.'5 (S0.:3)

7 (10.9) II (16 .9)

1.4

44 (34.1) R.'5 (6.'5.11\

3 (6 .R) 1.'5 (17.6)

2.9

61 (47 .2) 6R (.'52 .7)

7 (11..5) II (16 .2)

1.4

.'54 (4l.R) 7.5 (SR.!)

1.5 (27.8) 3 (4.())

9.2

0.0001

26 (20.1) 103 (79.11)

7 (26.9) II (10.7)

3.0

0.03

60 (.'51.1) (4R.R)

4 (6.1) 14 (22.2)

5 n.R) 12-t (\!0 1\

I (20.0) 17 (1:3.7)

1..'5

1'\S

14 ( IO.R) 115 (119 I)

.t (211.6) 1-t (12.2)

2.R I

NS

15 ( ll.fi) 114 (SIU)

5 (:3:3.:3) 13 (12.R\

3.11

0.02

.'59 (45 .7\ 70 (54 .2)

.'5 (R.S) 1:3 (IR.h\

2.-t

w (14.7) 110 (R.5 .2)

:3 (1.'5 R) 1.5 (1:1(;)

l.l I

1\S

24 (IR.6) 105 (11 1.3)

11 (;1:3.:3) (H.5\

3.0

0.1Kl2

!()

f~3

1\lortalit\~

p \'a low 0.05

2.3

1'\S

NS

!liS

1'\S

NS

IUKIH 4.4

NS

*COPD = ehronit· ohstrudin• pulmmutn· diseas• VAP = n·ntilatorasS
Univariate analysis of tlw 15 clinical and epidemiologic variables included in this study is shown in Table 5. The category with the lowest mortality rate was arbitrarily assigned to have a n•lative risk of death of

I. Age older than 45 years, corticosteroid use, presence of shock, hospital day of VAP over 9, antecedent COPD , and prior antibiotic use were asso<:iated with a significantly greater risk of death in the univariate analysis. In contrast, the remaining nine variables studied were not significant . The data were further analyzed using a stepwise logistic regression model. The dependent variable was attributable mortality; the independent variables were the dichotomous variables found to he significant hy univariate analysis. After adjustment for confounding factors, only prior antibiotic use (p
In this study, we analyze the relation of a variety of clinical-epidemiologic factors to outcome of VAP. Although mortality in patients Y.ith nosocomial pneumonia has been reported very high, there is evidence to suggest that the crude mortality may he more a function of the severity of the underlying disease than lung infection per se. 24 Several severity of illness indices are available for critically ill patients and we stratify our patients according to the criteria used by McCabe and Jackson 16 and to the codes defined by Emori et al. 17 The effect of alternative indices based on organ failure, such as the APACHE score, were not evaluated. Since crude mortality appears to he an inappropriate criterion to evaluate outcome for VAP, we used only the attributable mortality for statistical analysis. According to the univariate analysis, age older than 45 years (p = 0. 05), use of corticosteroids (p = 0.03), presence of shock (p = 0.02), hospital day of VAP over 9 (p = 0.008), antecedent COPD (p = 0.002), and prior exposure to antibiotics (p = 0 .0001) were significantly associated with fatality by VAP. None of the other variables examined exerted a statistically significant effect on mortality from VAP. However, most of these variables are interrelated and, therefore, a multivariate statistical study should be performed to identify those factors that may independently influence the mortality. In both multivariate analyses of the 129 episodes studied, the presence of a prior CHEST I 104 I 4 I OCTOBER, 1993

1233

antimicrobial treatment (p<0.0001) was the only factor that independently influenced mortality. Previously, Fagon et ai,H in a prospective study of 52 episodes of VAP, reported that the overall mortality in a group of 31 patients who had received antimicrobial therapy before the onset of pneumonia was 83 percent, compared with a mortality of 48 percent in the group of 21 patients who had not received prior antibiotics (p<0.01). Our data agree with their suggestion that antimicrobial therapy before onset of VAP may be associated with a negative impact on outcome . Furthermore, a strong relationship between the use of antibiotics prior to the onset ofVAP and thP etiologic agent is clearly observed in our study. Our finding that prior antibiotic use entirely dropped out as a significant risk factor when the etiologic agent was included in the logistic regression analysis demonstrated that the increased mortality associated with prior antibiotic use is due to the selection of more lethal organisms. Indeed, Gram-positive cocci and H injluen;:,ae represented more than three quarters of isolates among episodes free of antimicrobial exposure. However, these agents were an infrequent etiology among patients subjected to prior antimicrobial therapy, and a high proportion of episodes in this group were due to P aeruginosa or multiresistant Gram-negative bacilli . Suppression ofthe normal bacterial flora by antibiotics, with the subsequent emergence of a resistant Gramnegative flora, may explain this observation. Fagon et a)H also reported that pneumonia with P aeruginosa or Acinetobacter species was higher in the 31 patients who had received prior antibiotics (65 vs 19 percent, p
antibiotic use. Antibiotic therapy has a great ecologic impact. Broad-spectrum antibiotic therapy, used in most ICUs, may alter patients' microflora, leading to airway colonization and ultimately infection with nwthicillinresistant staphylococci, multiple resistant Enterobacteriaceae, pseudomonads, and yeasts. uuz Most nosocomial outbreaks caused by resistant pathogens began in the small fraction of patients hospitalized in the ICU and antibiotic pressure was usually found to he one of the most important factors predisposing these patients to infection with resistant organisms.:tul Antibiotics not only alter colonization in the oropharynx and gastrointestinal tract, hut they may also have deleterious effects on the host immune response to infection.:)!; Some clinicians have opted fi1r the liberal use of broad-spectrum antimicrobial agents in mechanically ventilated patients when they developed a febrile nosocomial episode, or even fi1r prophylaxis of nosocomial pneumonia. We agree with Johanson 11' that such practices are expensive, may cause avoidable side effects, and compromise efforts to document causativp agents precisely in future infections. Furthennore, our current results suggPst that this stratPhry may bP associated with a higher mortality, directly related to infection if such patients developed a VAP. Consequently, we suggPst the development of spPcific guidelines for antimicrobial therapy of infections in the ICU and an antibiotic control policy based on thP restriction of hroad-spPctrum agents. We support the proposition of Flaherty and Weinstein 2 that the use of broad-spectrum antibiotics in critically ill patients should he surveyed and appmved by a designated antibiotic expPrt who should be an infectious disease physician or a critical care physician with expertise in antimicrobial therapy and infection control. In our opinion, avoidance of unnecessary administration of antibiotics should form part of the recommended procedurps in the treatment of mechanically ventilated patients. ACKNOWLEDGMENT: We thank \V. C . Johanson. M.D. , M.P.H . , and G. Verger. M.D .. fi>r helpful sU).(!.(t'Stions concerning the manuscript. \Ve also thank members of the ICU for their cooperation . REFERENCES Craven DE. Kunches LM. Kilinsky V. Lichtenberg DA. Make BJ. ~k{;ahe \VR . Risk factors for pnt>umonia and fatality in patients receiving continuous mechanical ventilation. Am Rev RI' Spir Dis 1986: 1.3.1:792-96 2 Flaherty JP, Weinstein RA. Infection control and pnt•umonia prophylaxis strall•gies in the inknsin· cart• unit. Se min Respir Infl'ct 1990:5:191-203 3 P1'nnington JE. NoSspiratory infection. In: Mandell GL, Douglas RG Jr. Bennett JE. eds. Principlt's and practice of infeetious diseases. New \i>rk: John Wiley & Sons. 19H5: 16202.5 4 (;arihaldi RA , Britt ~IR. Coleman ML. Reading JC. Pace NL. Risk faetors fc>r postoperatin· pneumonia. Am J Med 19Hl: 70:677-RO

Impact of Previous Antimicrobial Therapy in Ventilator-associated Pneumonia (Rallo eta/)

5 Celis R, Tc)rres A, Gatell JM , Almela M, Rodriguez-Roisin R. Agusti-Vidal A. Noso<_,,mial pneumonia: a multivariate analysis of risk and prognosis. Chest 1988; 93:318-24 6 Seidenfdd JJ, Pohl OF, Bell RC, Harrys GO. Incidence, site and outcome of infections in patients with the adult respiratory distress syndrome. Am Rev Respir Dis 1986; 134:12-6 7 Tc,rres A, Aznar R, Gatell JM, Jimenez P. Gonzalez J, Ferrer A, et al. Incidence, risk and prognosis factors of noso<.1)mial pne umon ia in mechanically ventilated patients. Am Rev Respir Dis 1990; 142:523-28 8 Fagon JY, Chastre J, Domart Y, Trouillet JL , Pierre J, Dame CH , et al. Noso<.• nnial pneumonia in patients receiving <:ontinuous mechanical ventilation: prospective analysis of 52 episodes with use of a protected specime n brush and quantitative culture techniques. Am Rev Respir Dis 1989; l39:877-ll4 9 Meduri CU . Ve ntilator associated pneumonia in patients with respiratory failure: a diagnostic approach. Chest 1990; 97:120819 10 \Veinstein RA. Selective intestinal de<.1mtamination: an infection mntrol measure whose time has mme? Ann Intern Med 1989; 110:853-55 li Van Saene HKF. Stoutenbeek CP. Lawin P. Ledingham IMcA . Update in intensive eare and emergeney medicine: infeetion <.1mtrol by selective demntamination. 1st ed. Berlin: SpringerVerlag, 1989 12 Comite de Consensus. Seventh Conference de <.1msensus en reanimation et medicine d 'urgence: microbial selective de<.•mtamination in intensive care patients. Paris, December 13-14 , 1991 13 Ameriean Thoraeic Society. Standards for the diagnosis and care of patients with chronic obstructive pulmonary disease (COPD) and asthma. Am Rev Respir Dis 1987; 136:225-44 14 Rimel R\V, Tyson GW. The neurologic examination in patients with eentral nervous system trauma. J Neurosurg Nurs 1979; 12:148-52 15 Rello J, Quintana E. Ausina V. Puzo C, Net A. Prats G. Risk factors for StaphylococL~ls a.un?us nos
20

21

22

23 24

25

26

27

28

29 30

31 32

33

34

35 36

for the diagnosis of pulmonary infeetions. Chest 1982; 82:55662 Edelstein PH . Improved semiselective medium for isolation of Legionella pneumophila from contaminated clinical and environmental specimens. J Clin Microbiol 1981; 14:298-303 Chastre J, Viau F, Brun P, Pierre J, Dange MC, Bouchame A. Prospeetive evaluation of the protected specimen brush for the diagnosis of pulmonary infections in ventilated patients. Am Rev Respir Dis 1984; 130:924-29 Lennette EH , Bullows A, Hausler WJ Jr, Shadomy HJ. Manual of elinical microbiology. 4th ed. Washington : American Society for Mierohiology, 1985 Schlesselman JJ. Case-<.1mtrol studies: design, conduCt, analysis. New York: Oxford University Press, 1982 A'Court CH, Garrard ChS. Nos0<.1Jmial pneumonia in the intensive care unit: mechanisms and signifieance. Thorax 1992; 47:465-73 Louria DB, Kaminski T. The effects of four antimicrobial drng regimens on sputum superinfection in hospitalized patients. Am Rev Re spir Dis 1982; 85:649-85 Carratala J, Pallares R, Verdaguer R, Santin M, Mascaro J, Manresa F, et al. Risk factors for nosphilus influen;;ae among mechanically ventilated patients: incidence , outmme and risk factors. Chest 1992; 102:1562-65 Graybill JR , Marshall L\V, Charache P, Wallace CK, Melvin VB. NoS
CHEST 1104 I 4 I OCTOBER, 1993

1235