CLINICAL MICROBIOLOGY AND INFECTIOUS DISEASE

Transcription

1 CLINICAL MICROBIOLOGY AND INFECTIOUS DISEASE Original Article S o n i c a t e d V a s c u l a r C a t h e t e r T i p C u l t u r e s Quantitative Association With Catheter-related Sepsis and the Non-Utility of an Adjuvant Cytocentrifuge Gram Stain MICHAEL KELLY, MD, 1 LEANNE R. WUNDERLICH WCIORKA, MT, 2 SANDRA McCONICO, 2 AND LANCE R. PETERSON, MD 1 ' 2 Microbial quantification of potentially infected intravenous catheters is used to determine whether the removed device may be the source of sepsis. In this study, the authors defined a clinically significant colony count using the sonication technique to implicate the catheter as the source of sepsis. The utility of a cytocentrifuge-prepared Gram's stain of sonication broth as a rapid test for the accurate diagnosis of catheterrelated bacteremia and fungemia was also evaluated. Initially, 230 consecutive vascular catheter cultures were performed. In 14 patients, the blood culture was found to grow the same organism as the catheter culture. The eight blood-culture positive patients with probable catheterrelated sepsis had four catheters with s, and four with more than 10 s CFU derived from the catheter sonication broth. Six bacteremias not catheter related showed three catheters with <10 3, two with , and one with >10 s CFU. One catheter-related bacteremia had a Vascular catheter-related infections remain a major source of nosocomial sepsis despite a better understanding of the pathogenesis of these infections leading to increased efforts to improve techniques for their diagnosis over the past two decades. 1 " 3 An initial semiquantitative assessment of vascular catheters was described by Maki and colleagues, 4 who, using the roll- plate method, demonstrated that the growth of > 15 CFU on an agar plate correlated with local infection at the catheter site, and that > 1,000 colonies was associated with catheter-related bacteremia. Further studies have shown the utility of semiquantitative catheter tip cultures in the diagnosis of catheter-related sepsis. 5 " 6 Although very useful, several Presented in part at the Fall Meeting of the American Society of Clinical Pathologists, October 22-28, 1994, in Washington, DC. Supported by Shandon Lipshaw, Pittsburgh, Pennsylvania; Northwestern Memorial Hospital; and Northwestern University. Manuscript received August 9, 1995; revision accepted October 13, Address reprint requests to Dr. Peterson: Microbiology Division, Pathology Department, Northwestern Memorial Hospital, Wesley Pavilion, Room 565, 250 E. Superior Street, Chicago, 1L negative culture. A subsequent validation study on 175 catheter specimens using 1,000 CFU as a cut-off between a positive and negative test fully separated eight patients with catheter-related bacteremia from six patients with infection at a distant site. Cytospin Gram stain, studied in the initial evaluation, was positive in 17 catheters with <10 3 CFU, 4 at ,4 at and 9 with >10 5 CFU in the sonication broth. No correlation was found between this test and catheter-related sepsis. From these results, the authors concluded that those catheters with counts below 10 3 CFU in the catheter sonication broth do not appear to have an association with catheter-related sepsis, and that a Cytospin Gram stain done with the sonication technique does not correlate with the presence of catheter sepsis. (Key words: Cytocentrifuge; Gram stain; Vascular catheter infection; Quantitative catheter culture; Vascular catheter sonication) Am J Clin Pathol 1996; 105: limitations of the roll-plate method have been described, including its inability to quantify more than 1,000 colonies per catheter segment 7 and the fact that only the external surface is sampled. 8 Quantitative culture techniques whereby organism is suspended into a broth, culturing both the internal and external surfaces, followed by quantitative culture have been studied as a potential improvement over the rollplate method. In addition to evaluating both external and internal catheter surfaces, this approach permits quantification of > 1,000 colony forming units per catheter tip. 7 " 10 Methods used to suspend organisms in the broth include flushing, 9 centrifuging, 10 and sonicating 7 " 8 the catheter segment. These studies have demonstrated good positive correlation between catheter-related sepsis From the the 'Department of Pathology, Northwestern University and the number of organisms isolated from the broth. Medical School. Chicago, Illinois; and 1 Microbiology Division, Department of Pathology, Northwestern Memorial Hospital, Chicago. A limitation for any semiquantitative or quantitative Illinois. culture method is the necessity of an overnight incubation, leading some investigators to attempt incorporating a Gram stain with various methods in order that more rapid information might be obtained, thereby expediting appropriate clinical treatment."" 14 Investigations have looked at the utility of "impression smears" of the external surface of the catheter in conjunction with semiquantitative culture,'' a method for Gram staining the 210

2 KELLY ET AL. 211 Vascular Cathetet Sonication Cultures catheter as a whole, 12 and Gram staining of the blood collected from total parental nutrition catheters in conjunction with quantitative blood cultures. 13 In this study, we evaluated a sonication technique for quantitative catheter tip cultures to define a clinically significant colony count that implicates the catheter as the source of sepsis for our laboratory, and we investigated the utility of performing a cytocentrifuge-prepared (Cytospin 3, Shandon Lipshaw, Pittsburgh, PA) gramstained smear on a portion of the sonication broth. The Cytospin Gram stain has been shown to be useful for various bodily fluids, providing a sensitive method for detecting bacteria, even when only a small sample is available. 15 " 17 We hypothesized that in the setting of using a sonication technique for quantitative catheter tip cultures, a Cytospin Gram stain may be a helpful tool for providing more rapid information than culture alone. MATERIALS AND METHODS We performed two sequential evaluations of the sonication technique as an assessment for confirmation of catheter-related sepsis. Our initial study material came from 230 consecutive catheter tips collected from the patient population at Northwestern Memorial Hospital and McGaw Medical Center, Chicago, Illinois, between October and December The catheters were obtained from 135 patients, primarily adult, from a variety of ward locations within the facility. The majority were from the surgical or medical intensive care units. Four patients were from the outpatient department. The catheters were not divided as to arterial or venous source, or specific catheter type. The sonication technique for quantitative culture and Cytospin Gram stain was performed on the specimens described previously. In a second (validation) evaluation, we studied 175 consecutive catheters received between July and August 1994, using only quantitative culture with a set cut-off point to validate the conclusions drawn from the initial data. Overall, a total of 405 catheters were evaluated. The sonication technique used is a modification of methods previously described. 7 " 8 Each catheter was first placed into a tube containing 4 ml of brain heart infusion (BHI, Difco Laboratories, Detroit, MI) broth. The tube was then sonicated (Branson Ultrasonic, Danberg, CT) for 1 minute, followed by vortexing for 15 seconds. This was followed by plating on chocolate and blood agar (Difco) using both 0.1 ml and ml of the sonicated broth on duplicate sets of chocolate and blood agar plates. The plates and remaining BHI broth were incubated at 35 C in 5% C0 2 and examined every 24 hours for 5 days. The plates with the 0.1 ml inoculum were read first. Positive results were interpreted using the fol- lowing equation: NX 10X4 = CFU, where N is the number of colonies on the plate, 10 is a constant for the inoculation amount of 0.1 ml, 4 is a constant to adjust for amount of media in the BHI tube, and CFU is the number of colony forming units on the catheter tip. Twenty-five colonies represent 1,000 colony forming units (CFU) in the original broth, and 250 colonies represent 10,000 CFU. If the 0.1 ml inoculated plates contained too many organisms for accurate counting (ie, organisms were not present as discrete colonies), the number of colonies from the ml inoculation were used with an appropriate modification of the equation (N X 1,000 X 4 = CFU), where 25 colonies represent 100,000 CFU from the catheter segment. A positive culture was defined as any catheter that had at least 1,000 CFU recovered from the sonication broth. A cytocentrifuge Gram stain was performed using 0.5 ml of the sonicated BHI broth in a Cytospin 3 instrument where the material was centrifuged at 2,000 rpm for 15 minutes. The grading system used in reading the slides was that described by Olson and colleagues 15 used for screening urine specimens, whereby the slides were read by scanning 12 consecutive fields under 1,000X oil immersion. The number of fields in which organisms were seen was recorded and the results interpreted as follows: No bacteria seen = no bacteria in any of the 12 fields; Few = the same morphologic type seen in fewer than six fields; Moderate = the same morphologic type seen in six fields or more; Many = more than 100 organisms of the same morphologic type seen in all oil fields. Few, moderate, or many represented a positive Gram stain result in this evaluation. Clinical correlation was done retrospectively by reviewing the records of all study patients with a positive catheter tip culture, or positive gram-stained smear, as well as those who had a positive blood culture with an accompanying catheter tip submitted for processing. The patient medical records from all positive blood and catheter cultures were evaluated looking for evidence of catheter related sepsis. The catheter was felt to be the likely source of sepsis if (1) the same organism that was isolated from the catheter was identified in one or more blood cultures taken within 48 hours before the catheter tip removal; and (2) there was no other source of infection identified. Additional evidence for catheter related infection included documentation of visible evidence of purulence at the insertion site, clinical manifestations of sepsis, resolution of the clinical infection within 48 hours following catheter removal, and a medical record diagnosis of catheter-related sepsis. 1 Identical inclusion criteria were used throughout the study. Identification of microbial strains was done on those isolates growing on the agar plates using conventional Vol. I' No. 2

3 212 CLINICAL MICROBIOLOGY AND INFECTIOUS DISEASE TABLE 1. LISTING OF 137 ORGANISMS ISOLATED FROM SONICATED CULTURES OF CATHETER TIPS RECEIVED FROM 71 PATIENTS IN THE INITIAL STUDY Organism CNSS NHSS Staphylococcus aureus Enterobacteraceae Candida sp. Enterococcus sp. Pseudomonas aeruginosa a-hemolytic streptococci Corynebacterium sp. No. of Organisms (n = 137) Original Article % CNSS = coagulase negative Staphylococcus species; NHSS = nonhemolytic Streptococcus species. methods. 18 Statistical analysis was also performed using standard methods. 19 Initial Study RESULTS Of 230 catheters from 116 patients cultured in the first study, 106 catheters (46%) gave positive results; either positive culture alone (72 catheters), positive culture and Gram's stain (30 catheters), or Gram stain alone (4 catheters). From 13 catheters, more than 1 organism was identified. A total of 137 organisms were isolated from catheters removed from 71 patients, and are summarized in Table 1. The majority of these were coagulasenegative Staphylococcus species (CNSS). Forty-seven were identified to species level, and most were S epidermidis, making up 29 of the 47 isolates (62%). These isolates were equally distributed among slime (glycocalyx) producers (15) and nonproducers (14). Of the Enterobacteriaceae species, 10 were identified to the species level, including Enterobacter aerogenes (3), Enterobacter cloacae (2), Proteus mirabilis (2), Klebsiella pneumonia (1), Serratia marcescens (1), and Escherichia coli(\). Of the Candida species identified, there were 3 Candida albicans. In the 106 positive catheters, a total of 34 Gram stains were positive (32%) when at least "Few" (same organism type in fewer than 6 of 12 fields) organisms was considered a positive test. The distribution of these catheters, in regard to quantitative colony count was: 17 catheters with 0-1,000 CFU (including 4 with no growth, and 4 only with growth in the BHI broth); 4 catheters with 1,000-10,000 CFU; 4 with 10, ,000; and 9 catheters with greater than 100,000 CFU. These data are summarized in Table 2 with regard to the number of organisms identified on the slide. Of the eight patients of probable catheter-related sepsis we initially identified, three showed a positive gram stain, as did one patient of sepsis from another site that did not involve the catheter (Table 3). A computer search of microbiology laboratory records was performed for the inclusive time period looking for positive blood cultures from the patients corresponding to the culture specimens. In 23 patients, there were at least 2 positive blood cultures within 48 hours before catheter removal, with 14 of these patients demonstrating the same organism that was isolated from the catheter tip. The chart review performed identified eight patients that were likely catheter related septicemia. There were four patients due to ; two patients with S epidermidis; one patient with S haemolyticus; and one patient with bacteremia due to Enterobacter aerogenes. These patients, along with the corresponding colony counts and Gram stain results, are summarized in Table 3. In the remaining 9 of the total 23 bacteremic patients, both catheter culture and Gram stain were negative. A chart review was performed on these patients looking for possible false-negative catheter culture results (patients felt clinically to be catheter related but failed to give any positive results on culture or Gram stain). One additional patient was found in this group of nine that did appear to be catheter related, but failed to give positive culture results on the plated media. This patient had two positive blood cultures for S haemolyticus within 24 hours before catheter removal. Of the 93 blood culture negative patients, 58 had < 10 3 CFU in the catheter sonication broth, 20 had CFU in the CSB, 8 had , and 7 had >10 5 CFU in the catheter sonication broth. Evaluating the data for predicting the source of bacteremia or fungemia in septic patients by culture of soni- TABLE 2. NUMBER OF SPECIMENS FOUND WHEN COMPARING THE COLONY COUNT ON SONICATED CATHETER SEGMENT CULTURE TO THE QUANTITY OF BACTERIA SEEN ON 34 POSITIVE CYTOSPIN GRAM STAIN SLIDES Colony Forming Units/mL of Sonicated Broth Quantitation of Bacteria per Slide 0-1, ,000 > Few Moderate Many Few = same morphologic type seen in fewer than sixfields;moderate = same morphologic type seen in six or morefields; Many = more than 100 forms of the same organism seen in all 12 fields. A.J.C.P.* February 1996

4 KELLY ET AL. 213 Vascular Catheter Sonication Cultures TABLE 3. POSITIVE BLOOD CULTURES WHERE THE SAME ORGANISM WAS ISOLATED FROM THE CATHETER TIP CULTURE Organism Likely catheter related 5 aureus S epidermidis S haemotyticus S epidermidis Eaerogenes S haemotyticus Other source identified, or indeterminate 5 aureus S epidermidis K pneumoniae Eaerogenes CFU 28, , ,000 24,000 >400,000 48,000 12, , ,000 2, GBO 8,000 GBO Gram Stain G+ cocci, many G+ cocci, moderate G+ cocci, few G+ cocci, few CFU «colony forming unitsrecovered from the catheter: G+ = gram positive: GBO Analysis = of Combined Data growth in brain heart infusion broth only. cated fluid using 1,000 CFU as a cut-off for positive found the sensitivity, specificity, positive predictive value, and negative predictive value to be 89%, 50%, 73%, and 75%, respectively. Using 10,000 CFU as the cut-off point, the same parameters were 89%, 83%, 89%, and 75%, respectively. The ability of the cytocentrifuge Gram stain in predicting a positive quantitative culture (ie, a culture growing greater than 1,000 CFU), is reflected in the following parameters: sensitivity 37%; specificity 72%; positive predictive value 50%; and a negative predictive value of 60%. If the Gram stain was read more conservatively, with only patients showing a "Moderate" number of organisms or more read as positive, the same parameters would be 20%, 88%, 50%, and 58%, respectively. In assessing our positive culture results, we found a reliable cut-off for colony counts using the sonication technique is CFU, with those counts greater than 10 4 CFU indicating a significant correlation with catheter-related sepsis. In this first portion of our study, there were a total of 24 catheters with more than 10 4 CFU grown from them. One-third of these catheters with counts >10 4 CFU in the catheter sonication broth were associated with the development of bacteremia. Validation Study After analyzing the initial results, we selected a colony count cut-off that should maximize our sensitivity. Therefore, the validation study was done with a cut-off for a positive quantitative test result set at 1,000 CFU. For the 175 catheters from July and August 1994 used for validation of our initial evaluation, we found 14 patients in which there was a positive blood culture and catheter culture for the same organism. By chart review, six of these patients' infections appeared catheter related, and all grew > 1,000 CFU (range l,400->40,000). The remaining eight patients were determined not to be catheter related and all were negative, growing < 1,000 CFU. Therefore, in the validation study, the sensitivity, specificity, positive predictive value, and negative predictive value were all 100% when 1,000 CFU was used as the cut-off to separate likely catheter-related sepsis from infection due to another source. In this portion of the study, there were 28 catheters that grew > 1,000 CFU, and 6 of these (21 %) were associated with catheter-related sepsis. Overall, the utility of sonicated cultures for demonstrating the catheter as the source of infection in septic patients, using 1,000 CFU as a cut-off for a positive association, found the sensitivity, specificity, positive predictive value, and negative predictive value to be 94%, 75%, 84% and 90%, respectively. Using 10,000 as the cut-off point, the same parameters were 88%, 92%, 94%, and 85%, respectively. In the overall study, choosing the lower cut-off enhances sensitivity, whereas the higher cut-off enhances specificity. DISCUSSION Accurate diagnosis of sepsis being related to a vascular catheter is important for at least two reasons. Because this is considered a nosocomial infection, the tracking of catheter-related bloodstream infections is crucial in monitoring how well appropriate catheter management minimizes their rate of occurrence. More importantly, the demonstration that a bloodstream infection is or is not catheter related has a direct impact on the length of antimicrobial agent therapy that must be given to adequately treat the septic episode. Infection due to a removable focus (catheter) can usually be treated in fewer than half the number of days than bloodstream infections related to endocarditis or other infections within the vascular space that cannot be removed. 20 It has been demonstrated in previous studies that using quantitative catheter tip culture techniques to extend the upper limit of quantification over the level obtainable by the roll-plate technique provides a positive correlation between the number of organisms isolated and catheterrelated sepsis. Cleri and colleagues 9 using a technique of Vol. 105 No. 2

5 214 CLINICAL MICROBIOLOGY AND INFECTIOUS DISEASE Article mation to possibly guide clinical treatment. Cooper and colleagues found that a direct staining of the catheter tip was 100% sensitive and 96.9% specific for picking up colonization of the tip. 12 Their study used the roll-plate method for culturing the catheter tip before performing the Gram stain. Therefore, this method would not be applicable for the sonication technique that requires catheter immersion into broth. Collington and colleagues," using an impression Gram stain, were able to pick up 3 of 3 patients with catheter-related bacteremia (including one patient that was negative by semiquantitative culture), with negative Gram stains in all five patients where another source was identified. However, because of the low prevalence of catheter-related bacteremia in the study group, there was a very low positive predictive value reported of only 44%. A Cytospin centrifuge Gram stain has been shown to be very useful in concentrating small quantities ( mL) of body fluids, revealing enhanced organism detection and better leukocyte morphology than unconcentrated or conventionally concentrated specimens. This method would seem to lend itself well to concentrating sonicated broth. The advantage would be that both the culture and Gram stain could be done on a portion of the same broth, possibly providing more rapid information with a high degree of correlation. However, the results of our study do not support the usefulness of the Cytospin Gram stain for early detection of catheterrelated sepsis. Of the nine patients with catheter-related sepsis we identified, only three had a positive Gram stain. In one of six patients where the sepsis was thought to be non-catheter related, it was also positive. The ability of the Gram stain to predict a positive result on culture (ie, > 1,000 CFU) was also limited, with a positive predictive value of only 53%. The results were not improved significantly if the Gram stain was interpreted more conservatively (ie, only patients where a "Moderate" number of organisms were seen read as positive. The specificity went from 73% to 90%, but the sensitivity decreased concomitantly from 40% to 22%. Also, there were a number of false-positive Gram stains where the culture was negative (4 patients), or there was growth in the BHI broth only (4 patients). This brings up the question could the Gram stain be useful in picking up organisms that would not grow in culture because of treatment or fastidiousness of the organism, but still be involved in catheterrelated sepsis? A chart review of any possibly false-negative catheter tip cultures (from the positive blood culture, negative catheter tip cultures) found one patient of septicemia with S haemolyticus that was thought to be catheter related, but was negative for culture. However, the cytocentrifuge Gram stain was also negative in this patient (Table 3, patient 9). flushing the catheter with broth followed by serial dilutions, found no patients with associated bacteremia when counts were < 1,000 CFU. Bjornson and coworkers 10 using a centrifugation technique to remove organisms followed by serial dilutions, found 5 of 5 cases of proven catheter-related bacteremia grew > 1,000 CFU. Sheretz and associates 7 using the sonication technique, found that among 207 catheters that grew > 100 CFU and had an associated positive blood culture within 48 hours for the same organism, there was a positive linear correlation of 0.93 between frequency of concordant positive blood cultures and the number of organisms recovered after sonication. Also, they found catheter tips that grew <100 CFU were strongly associated with proven infection at a site distant from the catheter site (P =.001). Because of their method of enumeration, it appears that the cut-off they proposed of 100 CFU is identical to the one we chose at 1,000 CFU for a total organism count from the sonicated catheter. Our current studies support these previous studies. Of the initial nine patients with catheter-related sepsis in this study, four grew > 10,000 CFU, four grew > 100,000, and one was negative for growth. The negative growth catheter is consistent with the fact that an occasional catheter-related sepsis may be due to infection of the catheter hub, and these would not be detected by cultures of the catheter itself. 21 In the six patients where another source was identified or the source was indeterminate, three had < 1,000, two had < 10,000, and one had > 100,000 CFU. These studies suggest that a clinically useful cut-off point for this technique is either 1,000 and 10,000 CFU. When validating the results by evaluating more patients, we found 1,000 to be the most reliable cut-off point. All of the six patients of sepsis that were catheter related grew > 1,000 CFU, whereas all eight of the patients where another source was identified grew < 1,000 CFU. Combining the results of these two studies gives a sensitivity of 94% and a specificity of 75% when using a cut-off of > 1,000 CFU versus 88% and 92% at > 10,000 as an indication that a catheter site is the likely source of bacteremia or fungemia in patients with positive blood cultures. This indicates that either cut-off could be used, with one sacrificing some sensitivity and the other some specificity. Our experience would suggest that the results of our second "validation" study, done after the laboratory technologists were more experienced with the procedure, indicates the cut-off of 1,000 CFU is very useful in a clinical laboratory setting. Previous studies that have looked at incorporating a Gram stain into the procedure for detecting catheter-related sepsis have suggested a stained smear may be used in conjunction with culture to provide more rapid infor- A.J.C.P- 1996

6 From this study, we concur with previous studies that the sonication technique of quantitative catheter tip culture is a useful technique, with a positive correlation between the number of organisms cultured and the presence of catheter-related sepsis. We have found a practical and reliable cut-off to be 1,000 CFU, with cultures below this level reported as negative, with an accompanying comment that the catheter is an unlikely source for bacteremia. We also conclude that a Cytospin Gram stain done in conjunction with the sonication technique is not a sensitive or specific method, with little correlation between a positive culture and the presence of catheter-related sepsis. REFERENCES 1. Raad II, Bodey GP. Infectious complications of indwelling catheters. Clin Infect Dis 1992; 15: Salzman MB, Rubin LG. Intravenous catheter-related infections. Adv Pediat Infect Dis 1995; 10: Goldman DA, PierGB. Pathogenesis of infections related to intravascular catheterization. Clin Microbiol Rev 1993;6: Maki DG, Weise CE, Sarafin HW. A semiquantitative culture for identifying intravenous catheter-related infections. N Engl J Med 1977; 296: Moyer MA, Edwards LD, Farley L. Comparative culture methods on 101 intravenous catheters. Arch Intern Med 1983; 143: Collignon PJ, Soni N, Pearson IY, Woods WP, Munro R, Sorrell TC. Is semiquantitative culture of central vein catheter tips useful in the diagnosis of catheter-associated bacteremia? J Clin Microbiol 1986;24: Sherertz RJ, Raad II, Belani A, et al. Three-year experience with sonicated vascular catheter cultures in a clinical microbiology laboratory. J Clin Microbiol 1990; 28: Raad II, Sabbagh MF, Rand K.H, Sherertz RJ. Quantitative tip culture methods and the diagnosis of central venous catheterrelated infections. Diagn Microbiol Infect Dis 1992; 15: KELLY ET AL. 215 Vascular Cathetei Sonication Cultures 9. Cleri DJ, Corrado ML, Seligman SJ. Quantitative culture of intravenous catheters and other intravascular inserts. J Infect Dis 1980;141: Bjornson HS, Colley R, Bower RH, Duty VP, Schwartz-Fulton JT, Fischer JE. Association between microorganism growth at the catheter insertion site and colonization of the catheter in patients receiving total parental nutrition. Surgerv 1982; 92: Collignon MB, Chan R, Munro R. Rapid diagnosis of intravascular catheter-related sepsis. Arch Intern Med 1987; 147: Cooper GL, Hopkins CC. Rapid diagnosis of intravascular catheter-associated infection gram staining of the catheter segment. N Engl J Med 1985;312: Moonens F, Alami SE, Gossum AV, Struelens MJ, Serruys E. Usefulness of gram staining of blood collected from total parenteral nutrition catheter for rapid diagnosis of catheter-related sepsis. J Clin Microbiol 1994; 32: Collignon PJ, Munro R. Laboratory diagnosis of intravascular catheter associated sepsis. Eur J Clin Microbiol Infect Dis 1989;8: Shanholtzer CJ, Schaper PJ, Peterson LR. Concentrated Gram stain smears prepared with a cytospin centrifuge. J Clin Microbiol 1982;16: Olson ML, Shanholtzer CJ, Willard K.E, Peterson LR. The slide centrifuge Gram stain as a urine screening method. Am J Clin Pathol 1991;96: Goswitz JI, Willard K.E, Eastep SJ, et al. Utility of slide centrifuge Gram's stain versus quantitative culture for diagnosis of urinary tract infection. Am J Clin Pathol 1993:99: Baron EJ, Peterson LR, Finegold SM. Bailey and Scott's Diagnostic Microbiology, ed 9. St. Louis, MO: Mosby-Year Book Company, Ilstrup DM. Statistical methods in microbiology. Clin Microbiol Rev 1990;3: Iannini PB, Crossley, K. Therapy of Staphylococcus aureus bacteremia associated with a removable focus of infection. Ann Intern Med 1976; Salzman MB, Isenberg HD, Shapiro JF. Lipsitz PJ, Rubin LG. A prospective study of the catheter hub as the portal of entry for microorganisms causing catheter-related sepsis in neonates. J Infect Dis 1993; 167: