Intensive Care Med (1998) 24: 1034±1039 Ó Springer-Verlag 1998

J.R. Gowardman C. Montgomery S. Thirlwell J. Shewan A. Idema P.D. Larsen J.H. Havill

Received: 26 February 1998 Accepted: 30 June 1998

)

J.R. Gowardman ( ) Intensive Care Unit, Wellington Hospital, P. O. Box 7902, Wellington South, New Zealand Tel.: + 64 (4) 385-5999 Fax: + 64 (4) 385-5962 email: [email protected] C. Montgomery × J. H. Havill Intensive Care Unit, Waikato Hospital, Private Bag, Hamilton, New Zealand S. Thirlwell × J. Shewan × A. Idema Department of Microbiology, Waikato Hospital, Private Bag, Hamilton, New Zealand P.D. Larsen Section of Anaesthesia, Wellington School of Medicine, P. O. Box 7343, Wellington South, New Zealand

O R I GI N A L

Central venous catheter-related bloodstream infections: an analysis of incidence and risk factors in a cohort of 400 patients

Abstract Objective: To determine the incidence of central catheter-related bloodstream infection (CRBSI) and to compare patient and catheter characteristics of those with and without CR-BSI from a clinically suspected subgroup. Secondly, to assess the efficacy of the acridine orange leucocyte cytospin test (AOLC) as a rapid in situ method of detecting central venous catheter (CVC) infection. Design: One-year prospective audit. Setting: Intensive care unit/high dependency unit (ICU/HDU) and general wards of a tertiary referral hospital. Patients and participants: 400 patients with non-tunnelled CVCs. Interventions: Daily surveillance, blood culture from peripheral venepuncture, blood sample from the CVC for assessment of the AOLC test and removal of suspected CVCs were carried out on patients clinically suspected of having CR-BSI. Measurements and results: CR-BSI was diagnosed using well defined criteria. Infection rate was calculated by dividing the number of definitive catheter associated infections by the total number of appropriate catheter in situ days. The AOLC test was performed on all those with suspected CR-BSI. A total of 499 CVCs in 400 patients were assessed, representing 3014 catheter in situ days. Over 80 % of patients were from our ICU/HDU, representing

404 CVCs and 1901 catheter in situ days. A total of 49/499 (9.8 %) CVCs in the same number of patients were suspected of being infected subsequently 12/499 (2.4 %) CVCs [95 % confidence interval (CI) 1.25 to 4.16] in 12 separate patients were demonstrated to be the direct cause of the patient's BSI. Rates of CR-BSI per 1000 catheter days were 3.98 (95 % CI 2.06 to 6.96) for the whole cohort and 4.20 (95 % CI 1.81 to 8.29) for the ICU/HDU subgroup. In the group suspected of having CR-BSI, CVCs were removed unnecessarily in 55 %, and no patient or catheter variables measured were predictive of the development of CR-BSI. The AOLC test was negative in all 12 catheters subsequently shown to be the definitive cause of BSI. Conclusions: We have defined the incidence of CR-BSI in a cohort of patients from a tertiary referral hospital, the rates comparing favourably with those reported for similar populations. We were unable to demonstrate significant differences in any patient or catheter variables between those with and without CRBSI. The AOLC test used alone was unhelpful as a method to diagnose in situ CVC infection in this patient population. Key words Central venous catheters × Infection × Diagnosis × Intensive care

1035

Introduction Central venous catheters (CVCs) are thought to account for about 90 % of all nosocomial bloodstream infections [1]. The literature on CVC infections is often confounded by varying terminology pertaining to definitions and lack of consistency in the expression of catheter infection rates. Recently catheter-related bloodstream infection (CR-BSI) rates varying from 2.1 per 1000 catheter days in respiratory intensive care units (ICUs) to up to 30.2 in burn ICUs have been reported [2]. Expression of CR-BSI rates in this manner has been recommended to facilitate comparison among different studies and to achieve constancy in the literature [3]. The primary aim of our study was to determine, using well-defined criteria, the baseline incidence of CRBSI in our study population and in a group of patients clinically suspected of having CR-BSI to compare patient and catheter characteristics in those with and without definite CR-BSI in an attempt to ascertain risk factors for CR-BSI in this population. In addition, we tested the efficacy in an adult population of the acridine orange leucocyte cytospin (AOLC) test, which has previously been successfully used in an infant population to diagnose catheter related sepsis [4].

Patients and methods Patients The study was carried out as a prospective audit between December 1993 and January 1995 at Waikato Hospital, Hamilton, New Zealand, a tertiary referral centre providing comprehensive specialist adult and paediatric medical and surgical services. The parent population was in the main from the general and cardiac 14bed ICU and adjacent 9-bed high dependency unit (HDU). The ICU admitted nearly 900 patients in the year of the study, of which the main admitting services were general medicine and general and cardiothoracic surgery. Data were also collected from CVCs in use on the general wards, in particular the renal, surgical and oncology units. Only patients with non-tunnelled CVCs, including pulmonary artery catheters inserted via the subclavian or jugular veins, were studied. Catheter insertion and care CVCs were supplied as a commercially available kit (Cook Critical Care, Bloomington, Ind., USA) and inserted using a Seldinger technique by either intensive care or anaesthetic registrars or consultants. Insertion was carried out under full aseptic conditions (gowns, gloves and masks). The insertion site was cleaned with either chlorhexidine 0.5 % in 70 % alcohol or 10 % povidine-iodine solution for at least 1±2 min and a sterile field ensured with large drapes. Site dressing was with a sterile gauze pad and a transparent semipermeable overdressing. Dressings were changed if the gauze became excessively damp or soiled. CVCs were not routinely changed unless they were suspected of being infected or developed a mechanical problem. When removed for suspected CR-BSI the

catheter was either not replaced (if no longer needed), inserted at a new site or guidewire exchanged (GWX). In suspected CR-BSI, appropriate investigations were carried out as delineated under Interventions below. Definitions Definite CR-BSI was diagnosed if the patient had a clinical picture compatible with bloodstream infection (temperature > 38 °C, fever, chills, hypotension and a raised or lowered peripheral white cell count) with the catheter as the only obvious source of infection. In addition, isolation of the same organism (identical species and antibiogram) from semiquantitative catheter culture and peripheral blood was required (direct evidence). Alternatively, defervescence within 48 h after removal of the catheter, in a patient with confirmed bloodstream infection previously refractory to a 72-h trial of appropriate antibiotic, was accepted as indirect evidence of CR-BSI [3]. Interventions Throughout the study period medical staff in both the ICU/HDU and general wards were asked to perform the following tests on any patient clinically suspected of having CR-BSI: culture blood from a peripheral venepuncture, routinely send concurrent blood samples from the CVC for the AOLC test, and if the catheter was being removed or GWX, send suspected tip for semiquantitative culture along with blood cultures and blood concurrently taken for the AOLC test. At the attending clinicians discretion a CVC could be left in situ after blood and AOLC cultures were taken, however, if cultures subsequently returned positive the CVC was removed and the tip sent for semiquantitative culture. Data collection When a patient was cultured for suspected CR-BSI, we recorded: age, total length of ICU/HDU or ward stay in days, diagnosis, Acute Physiology and Chronic Health Evaluation (APACHE) II score in the first 24 h, ventilated (yes, no), patient outcome, date CVC inserted and date removed, number of catheters inserted, type of catheter (lumen number), site of insertion, use (e. g. infusion, monitoring), significant catheter and blood culture results, antibiotics prescribed, reason for CVC removal, oral temperature before and after the catheter was removed (if removed for suspected CR-BSI), the presence of rigours or hypotension and documentation of the highest or lowest white cell count in the 24 h preceding blood culture of the patient. In all other patients with a CVC in situ, who were not suspected of having CR-BSI per se, the catheter was monitored closely for complications (local and systemic), otherwise data were collected only on number of catheters inserted, in situ time and diagnostic group. Microbiological methods Collection and culture of peripheral blood were performed under strict aseptic conditions: into each aerobic and anaerobic culture bottle, 10 ml of blood was inoculated and incubated for up to 6 days, before being reported as negative. Culture was done initially via the VITAL automated blood culture system (bioMerrieux Vitek, Missouri, USA). Any positive cultures during this time

1036

were subcultured manually onto blood and chocolate agar and incubated at 37 °C for 48 h.

Table 1 Characteristics of patients with suspected CR-BSI. Values are mean (SD) or number (%) Characteristica

Culture of the catheter tip Catheters were removed according to a set protocol under aseptic conditions. The distal 5 cm was cut off into a sterile container; in addition, a 5-cm proximal segment was also removed, providing an intracutaneous segment. The external surface of each segment was cultured semiquantitatively using the method described by Maki et al. [5]. All colony types that appeared were identified and enumerated. The plates were incubated for up to 72 h before being discarded as no growth. Semiquantitative catheter tip culture results were defined thus: catheter infection: > 15 colonies (positive semiquantitative culture) and catheter colonisation: < 15 colonies (negative semiquantitative culture). AOLC test

Mean age (years) Sex Male Female

CR-BSI (n = 12)

Non-CR-BSI (n = 37)

50 (17.3)

48.2 (21.3)

6 (50) 6 (50)

26 (70) 11 (30)

Mean APACHE II score Mean duration of admission (days)

17.7 (7.4) 23.6 (14.6)

17.8 (7.5) 21.4 (16.2)

Mean duration of catheterisation (days)

11.4 (6.1)

11.3 (6.9)

Mortality

3 (25)

6 (16)

Ventilated Location Ward ICU/HDU

7 (58)

25 (67)

4 (33) 8 (67)

13 (35) 24 (65)

The AOLC test was performed as previously described [4, 6]. A positive result was indicated by the presence of one or more viable or non-viable organisms. If the test was questionable or positive, a Gram stain was performed on a duplicate cytospin smear.

a

Statistical methods and data analysis

Table 2 Catheter characteristics of patients with suspected CRBSI. Values are number (%)

From among those suspected of having CR-BSI we compared those with definitive CR-BSI with those found not to have CRBSI. Continuous variables were analysed with the unpaired t-test, while discrete variables were analysed with the chi-square test. Analysis was performed using Statview 4.5 (Abacus Concepts, USA). CR-BSI rates were calculated by dividing the number of definite catheter associated infections by the total number of appropriate catheter in situ days, and multiplying the result to produce a figure per 1000 catheter days [2].

Results Over the period of the study, 499 CVCs were assessed in a total of 400 patients, representing 3014 catheter in situ days; 74/400 (18.5 %) patients were from the general wards and 326/400 (81.5 %) from the ICU/HDU group. In the ward group, 95 CVCs were assessed, representing 1113 catheter in situ days. In the combined ICU/HDU group, a total of 404 CVCs were assessed, representing 1901 catheter in situ days. Over the period of the study, a total of 49/499 (9.8 %) CVCs in 49 patients were strongly suspected of being infected. A total of 12 (2.4 %) [95 % confidence interval (CI) 1.25 to 4.16] of the total catheters assessed (24 % of those clinically suspected) in 12 separate patients were subsequently shown to be the direct cause of the patient's infection. Three patients met the criteria indirectly and 9 fulfilled the criteria directly; 8 of the 12 patients with CR-BSI were from the ICU/HDU group and 4 were from the ward group. Among the 49 patients suspected of having CR-BSI, the lines were removed unnecessarily in 27

Age, APACHE II score, duration of admission and catheterisation were compared using the unpaired t -test. Sex, outcome, ventilation and location were compared using the chi-square test. There were no statistically significant differences between the two groups

Catheter site Internal jugular Subclavian Catheter use Infusion and monitoring TPN only Haemodialysis Catheter lumen Single Double Triple

CR-BSI (n = 12)

Non-CR-BSI (n = 37)

6 (50) 6 (50)

21 (57) 16 (43)

8 (67) 3 (25) 1 (8)

30 (81) 4 (11) 3 (8)

2 (17) 2 (17) 8 (67)

5 (14) 5 (14) 27 (72)

Comparison were made using the chi-square test. There were no statistically significant differences between the two groups

(55 %) on the suspicion of CR-BSI. The rate of CRBSI per 1000 catheter days was 3.98 (95 % CI 2.06 to 6.96) for the whole cohort and 4.20 (95 % CI 1.81 to 8.29) for the ICU/HDU subgroup. The AOLC test was negative in all cases and therefore of no value in identifying the 12 cases of CR-BSI. Analysis of patients and catheter characteristics in those with and without CR-BSI among the 49 suspected cases can be seen in Tables 1 and 2. No statistically significant differences were noted between the two groups (t-test, chi-square test). The primary diagnostic groupings of patients suspected of and with CR-BSI are given in Table 3 and the microbial isolates from blood and catheters in those with CR-BSI in Table 4.

1037

Table 3 Primary diagnosis of patients with suspected and definitive CR-BSI

Trauma Sepsis/organ system failure Ruptured aortic aneurysms Gastrointestinal emergency Post-cardiac surgery Pancreatitis Burns Postoperative Other

CR-BSI (n = 12)

Non CR-BSI (n = 37)

1 (8) 2 (16) 2 (16) 2 (16) ± ± 1 (8) 4 (33) ±

11 (30) 8 (22) 4 (11) 2 (5) 1 (3) 3 (8) 0 3 (8) 5 (14)

a

Patients classified by primary diagnosis. Total number in each class is given followed by percentage in parentheses. Frequency of CR-BSI by diagnostic group was analysed using the chi-square test. No diagnostic group was represented in the CR-BSI group statistically more than in the suspected group

Discussion In this study we have demonstrated that, in a cohort of 400 patients from a tertiary referral hospital, the incidence of CR-BSI was 3.98 per 1000 catheter in situ days, and in our ICU/HDU population, which comprised approximately 80 % of the total patient population, the incidence of CR-BSI was 4.20 per 1000 catheter in situ days. We identified a subgroup of 49 patients clinically suspected of having CR-BSI. Subsequently, 12 catheters, 2.4 % of the total number studied in 12 separate patients, were identified as the definitive cause of the patient's infection. We were unable to distinguish these 12 from the 37 others suspected of CR-BSI on the basis of any data collected. We also demonstrated that in the 12 patients with CR-BSI the AOLC test did not prove useful as an aid to the in situ diagnosis of catheter infection. Based on North American data compiled from the national nosocomial infection surveillance system, CR-BSI rates ranging from 2.1 per 1000 catheter days for respira-

Table 4 Microbial isolates from catheter and blood in patients with CR-BSI (n = 12) and the primary diagnosis

tory ICUs, through 5.1 and 5.8 for combination medical/ surgical and trauma ICUs, respectively, to 30.2 for burn units have been previously reported [2]. Collignon [7] has also recently estimated the incidence of central catheter infections by pooling data from 15 Australian tertiary referral hospitals, finding a catheter sepsis rate of 2.3 %, while Gosbell et al. [8] reported 6.7 % in a large Australian general hospital population. The rates of CR-BSI in our ICU/HDU subgroup would therefore compare favourably with other units of similar case mix. The majority (65 %) of patients suspected of having CR-BSI were from our ICU/HDU group. The clinical judgement of a CVC infection in this complex group of patients continues to remain difficult, with a low predictive value [9]. Based on clinical judgement alone, over half of all CVC removals are unnecessary [6, 10, 11], which is supported by our data. This high rate of unnecessary CVC removal is in part due to the absence of a reliable test that detects an infected catheter in situ, such a test would indeed be a very useful clinical tool. Quantitative cultures drawn simultaneously through the catheter and a peripheral vein have been advocated by several authors [12±14] as fulfilling this role. Bozzeti et al. [9] have also proposed a model based on clinical assessment and catheter hub cultures and suggested that, if performed accurately, predictive value could be as high as 89 %. The AOLC test, which detects bacteria from a small quantity (50 ml) of catheter blood, can be performed in less than 1 h, unlike quantitative cultures, thus making it a potentially viable method of rapidly detecting an infected catheter in situ. Rushforth et al. [4], in an infant population, found the test 87 % sensitive and 94 % specific in the diagnosis of central catheter-related sepsis, as defined by quantitative blood culture. More recently, Tighe et al. [6] have reported using the test alone, and combined with an endoluminal brush technique in an adult population. Alone the test was positive in only 11 % of patients with infected catheter tips, quantified

Organism

¦

Primary diagnosis

S. epidermidis

3

S. warneri S. haemolyticus

1 1

Ruptured aortic aneurysm Sepsis Postoperative Postoperative Gastrointestinal emergency

Coagulase positive Staphylococcus

S. aureus

1

Postoperative

Enterobacteriaceae

Serratia marcescens Enterobacter aerogenes Enterobacter cloacae

1 1 1

Trauma Sepsis Gastrointestinal emergency

Coagulase negative Staphylococcus

Corynebacterium

JK

1

Burns

Enterococci Yeast

Enterococcus faecalis Candida parapsilosis

1 1

Ruptured aortic aneurysm Postoperative

1038

by the Maki roll method. When used in conjunction with an endoluminal brush technique, the sensitivity of the test was greatly enhanced towards the values reported by Rushforth et al. We performed the AOLC test on all those suspected of having CR-BSI; however, we were particularly interested in the performance of the test in those patients with definite CR-BSI as defined by our criteria. In this context, the test proved insensitive as an in situ method of detecting an infected catheter and therefore CR-BSI, consistent with the findings of Tighe et al., but in sharp contrast to those reported by Rushforth et al. Although we did not use the test prospectively to ªscreenº for catheter infection, only receiving the results ªafter the factº, we envisaged that if the test had proved sensitive and specific it might have provided a substantive basis for CVC removal in patients with suspected CR-BSI. Failure of the AOLC test to identify adult subjects with CR-BSI may relate to bacterial density, which is significantly higher in infants [15], or to the physical properties of the catheter. With a smaller lumen catheter it is possible that the ªdragº effect of drawing a blood sample enhances the removal of bacteria from the catheter wall, making the test sensitivity inversely related to catheter bore. Several variables have been quoted as contributing to the infectious complication rate of CVCs. These include catheter selection (composition, lumen number and whether antibiotic- or antiseptic-bonded), insertion site (location, care and technique of insertion), skill of the individual placing the catheter, emergent versus elective placement, patient characteristics including severity of underlying illness and the duration of catheterisation and, finally, type of infusate and apparatus used. We examined some of these variables in our suspect population to see which if any were predictive of the development of CR-BSI. Tables 1 and 2 provide details of the patients and catheters in those with and without CR-BSI from the initial 49 patients clinically suspected. Multilumen lines have in a number of studies [16±18] been associated with a higher incidence of CR-BSI. Additionally, several studies [19±21] have shown a significantly higher colonisation and infection rate with catheters inserted into the internal jugular vein as compared with those placed in the subclavian vein. In this study, 35 of 49 (71 %) patients with suspected CR-BSI had a triple lumen CVC; the majority of these CVs were in patients from our ICU/HDU. The principal sites of catheter insertion were either the internal jugular or subclavian veins, with the internal jugular site slightly preferred. We found no statistically significant difference between the patients with CR-BSI and those with suspected but not proved infection, with respect to these two variables. Although we did not specifically examine femoral catheters, it has been shown that the infectious complication rate of well maintained femoral catheters is no

higher than that of CVCs located elsewhere [10, 22] and we would commend this site as being a valuable and safe alternative to the great veins of the upper extremities. From our study we were unable to make specific conclusions regarding the skill of the operator in the development of CR-BSI; however, all personnel involved in the placement of CVCs in this study were well trained in percutaneous techniques. Likewise, we were unable to demonstrate an effect relating to CVC use or infusate on the development of CR-BSI. Duration of catheterisation has also been suggested as an important risk factor in the development of CRBSI [3, 10, 22], studies in general suggesting a relationship between prolonged catheterisation and an increased incidence of infection. We were able to demonstrate that in a group of patients clinically suspected of having CR-BSI, prolonged duration of catheterisation (11.4 vs 11.3 days for those with and without CR-BSI, respectively), in a setting of reasonable catheter insertion and care, was not a factor in the observed incidence of CR-BSI. Eyer et al. [23] in a cohort of 112 surgical ICU patients likewise found no relationship between catheter in situ time and CR-BSI rate, suggesting instead that severity of illness was more important. Prolonged catheterisation may be a cofactor with severity of illness in the determination of overall CR-BSI rates [24]. From among the 49 patients with clinically suspected CR-BSI, severity of underlying illness as measured by APACHE II scores was not statistically different between patients with and without CR-BSI [17.7 (range 6±27) and 17.8 (range 7±31), respectively], and therefore in this population it appeared unhelpful as a predictor of CR-BSI. Hence, in this subgroup of nearly 50 patients clinically suspected of having CR-BSI, on direct comparison between those who did and did not have definite CR-BSI as judged by our criteria, we were unable to conclude that neither severity of illness, duration of catheterisation, use of multilumen lines, nor CVC position represented risk factors for CR-BSI. The organisms isolated from blood and catheters in patients with CR-BSI were consistent with results in other published studies [3, 11, 16]. Speciation of organisms is given in Table 4. The three patients with enteric gram negative organisms were all from the ICU group; the organisms isolated may be typical of those acquired from the hospital environment, or may be of endogenous origin. Infection caused by gram-negative organisms such as those isolated has also been associated with contaminated pressure monitoring equipment and intravenous solutions [25, 26]; however, these were not implicated as sources of infection in any of our cases. Although it has been stated that many patients with a CVC in situ die, the role of catheter infection in these deaths is often unclear [8]. Three, or 25 %, of patients in our series with CR-BSI died, two in the ICU group

1039

and one from the ward group. However, it could not be said with confidence that in any of these deaths the catheter infection contributed directly, two patients having complex multiple organ failure secondary to their underlying disease, and one dying a cerebral death secondary to a severe head injury. Notwithstanding the increased morbidity and mortality directly associated with CR-BSI our views are in keeping with those of those of Tacconelli et al. [10] that deaths in this group may represent the coexistence of severe underlying diseases in patients with CVCs, and may not be due to the CVC per se. In conclusion, we have defined the incidence of CRBSI in a cohort of patients from a tertiary referral hospi-

tal, our infection rates comparing favourably with those previously reported. From a subgroup of 49 patients with suspected CR-BSI, no patient or catheter variables were predictive of the development of CR-BSI. Over 50 % of CVCs removed on suspicion of CR-BSI were removed inappropriately, underscoring the poor predictive value of clinical judgement alone in this complex patient group. We found the AOLC test to be labour intensive and demanding of skilled personnel for accurate performance and interpretation. Based on our findings, we cannot recommend that this test be used alone as an aid to diagnose in situ CR-BSI. A fast and reliable method of diagnosing in situ CVC infection in the setting of critical illness is still not available.

References 1. Pearson ML, Abrutyne E (1997) Reducing the risk for catheter related infections: a new strategy. Ann Intern Med 127: 304±306 2. Jarvis WR, Edwards JR, Culver DH et al (1991) Nosocomial infection rates in adult and pediatric intensive care units in the United States. Am J Med 91: 185S±191S 3. Pearson ML, Hierholzer WJ, Garner JS et al (1996) Guidelines for prevention of intravascular device related infections. Am J Infect Control 24: 262±293 4. Rushforth JA, Hoy CM, Kite P, Puntis JWL (1993) Rapid diagnosis of central venous catheter sepsis. Lancet 342: 402±403 5. Maki DG, Weise CE, Sarafin MS (1977) A semiquantitative culture method for identifying intravenous catheter-related infection. N Engl J Med 296: 1305±1309 6. Tighe MJ, Kite P, Thomas D, Fawley WN, McMahon MJ (1996) Rapid diagnosis of catheter-related sepsis using the acridine orange leukocyte cytospin test and an endoluminal brush. JPEN J Parenter Enteral Nutr 20: 215±218 7. Collignon PJ (1994) Intravascular catheter associated sepsis: a common problem. Med J Aust 161: 374±378 8. Gosbell IB, Duggan D, Breust M et al (1995) Infection associated with central venous catheters: a prospective survey. Med J Aust 162: 210±213 9. Bozzeti F, Bonfanti G, Regalia E et al (1991) A new approach to the diagnosis of central venous catheter sepsis. JPEN J Parenter Enteral Nutr 15: 412±416 10. Tacconelli E, Tumbarello M, Pittiruti M, Leone F, Lucia MB, Cauda R, Ortona L (1997) Central venous catheterrelated sepsis in a cohort of 366 hospitalised patients. Eur J Clin Microbiol Infect Dis 16: 203±209

11. Raad II, Bodey GP (1992) Infectious complications of indwelling vascular catheters. Clin Infect Dis 15: 197±210 12. Wing EJ, Norden CW, Shadduck RK, Winkelstein A (1979) Use of quantitative bacteriological techniques to diagnose catheter-related sepsis. Arch Intern Med 139: 482±483 13. Wormser GP, Onoroto IM, Preminger TJ, Culver D, Marlone WJ (1990) Sensitivity and specificity of blood cultures obtained through intravascular catheters. Crit Care Med 18: 152±156 14. Fan ST, Teoh-Chan CH, Lau KF (1989) Evaluation of central venous catheter sepsis by differential quantitative blood culture. Eur J Clin Microbiol Infect Dis 8: 142±144 15. Arpi M, Bentzon MW, Jensen J, Frederiksen W (1989) Importance of blood volume cultured in the detection of bacteraemia. Eur J Clin Microbiol Infect Dis 8: 838±842 16. Pemberton LB, Lyman B, Lander V, Covinsky J (1989) Sepsis from triple vs single lumen catheters during total parenteral nutrition in surgical or critically ill patients. Arch Surg 121: 591±594 17. Hilton E, Haslet T, Borenstein MT, Isenberg JD et al (1988) Central catheter infections: single versus triple lumen catheters; influence of guide wires on infection rates when used for replacement of catheters. Am J Med 84: 667±672 18. McCarthy MC, Shives JK, Robison RJ, Broadie RA (1987) Prospective evaluation of single and triple lumen catheters in total parenteral nutrition. JPEN J Parenter Enteral Nutr 11: 259±262

19. Richet H, Hubert B, Nitemberg G et al (1990) Prospective multicentre study of vascular catheter-related complications and risk factors for positive centralcatheter cultures in intensive care unit patients. J Clin Microbiol 28: 2520±2525 20. Gil RT, Kruse JA, Thill-Baharozian MC, Carson RW (1989) Triple vs single lumen central venous catheters. A prospective study in a critically ill population. Arch Intern Med 149: 1139±1143 21. Collignon PJ, Soni N, Pearson IY, Woods WP, Munro R, Sorrell TC (1990) Is semiquantitative culture of central vein catheter tips useful in the diagnosis of catheter-associated bacteraemia? J Clin Microbiol 24: 532±535 22. Norwood S, Ruby A, Civetta J, Cortes V (1991) Catheter-related infections and associated septicemia. Chest 99: 968±975 23. Eyer S, Brummitt C, Crossley K, Siegel R, Cerra F (1990) Catheter-related sepsis: prospective, randomized study of three methods of long-term catheter maintenance. Crit Care Med 18: 1073±1079 24. Norwood SH, Jenkins G (1990) An evaluation of triple-lumen catheter infections using a guidewire exchange technique. J Trauma 30: 706 25. Phillips I, Eykyn S, Curtis MA, Snell JJ (1971) Pseudomonas cepacia (multivorans) septicemia in an intensive care unit. Lancet I:375±377 26. Hekker TA, van Overhagen W, Schneider AJ (1990) Pressure transducers: an overlooked source of sepsis in the intensive care unit. Intensive Care Med 16: 511±512