Transfusion Practice and Blood Stream Infections in Critically Ill Patients

clinical investigations in critical care Transfusion Practice and Blood Stream Infections in Critically Ill Patients* Andrew F. Shorr, MD, MPH; William L. Jackson, MD; Kathleen M. Kelly, MD; Min Fu, MS, MBA; and Marin H. Kollef, MD

Study objective: To examine the relationship between packed RBC (pRBC) transfusion and the development of ICU-acquired bloodstream infections (BSIs) Design: Secondary analysis of a large, prospective, observational study of transfusion practice in critically ill patients. Setting: A total of 284 adult ICUs in the United States. Patients: Critically ill adults who lacked BSIs both at ICU admission and 48 h after ICU admission. Interventions: None. Measurements and results: BSIs were prospectively tracked in this study, and diagnosis of a new BSI represented the primary end point. Transfusions administered in the ICU prior to development of a BSI were also prospectively recorded. Of 4,892 patients enrolled in this investigation, 3,502 patients lacked BSIs both at ICU admission and 48 h later. Among these individuals, 117 patients (3.3%) had an ICU-acquired BSI. In multivariate analysis adjusting for severity of illness, primary diagnosis, use of mechanical ventilation, placement of central venous catheters, and ICU length of stay, three variables were independently associated with diagnosis of a new BSI: baseline treatment with cephalosporins (odds ratio [OR], 1.84; 95% confidence interval [CI], 1.26 to 2.68), higher sequential organ failure assessment score measured on ICU days 3 to 4 (OR, 1.11; 95% CI, 1.06 to 1.16), and pRBC transfusion (OR, 2.23; 95% CI, 1.43 to 3.52). The relationship between pRBC transfusion and BSI was evident with both small transfusion volumes (OR with 1to 2-U pRBC transfusion, 1.89; 95% CI, 1.10 to 3.23) and larger transfusion volumes (OR with > 4-U pRBC transfusion, 2.63; 95% CI, 1.52 to 4.53). Conclusions: pRBC transfusion is associated with subsequent ICU-acquired BSI. Avoiding unnecessary transfusions may decrease the incidence of BSIs. (CHEST 2005; 127:1722–1728) Key words: bacteremia; blood; blood stream infection; immunosuppression; intensive care; transfusion Abbreviations: APACHE ⫽ acute physiology and chronic health evaluation; BSI ⫽ bloodstream infection; CI ⫽ confidence interval; CVC ⫽ central venous catheter; MV ⫽ mechanical ventilation; OR ⫽ odds ratio; pRBC ⫽ packed RBC; SOFA ⫽ sequential organ failure assessment

infections disproportionately H ospital-acquired affect critically ill patients, with nearly one in five nosocomial infections acquired in the ICU.1 Nosocomial bloodstream infections (BSIs) in particFrom the Pulmonary, Critical Care, and Sleep Medicine Service (Dr. Shorr), Department of Medicine, and the Critical Care Medicine Service (Dr. Jackson), Department of Surgery, Walter Reed Army Medical Center, Washington, DC; Ortho-Biotech Clinical Affairs, LLC (Dr. Kelly and Mr. Fu), Bridgewater, NJ; and the Medical Intensive Care Unit (Dr. Kollef), Barnes Jewish Hospital, Washington University School of Medicine, St. Louis, MO. Ortho Biotech Clinical Affairs, LLC (Bridgewater, NJ) sponsored the CRIT Trial. 1722

ular present a significant challenge in the care of critically ill patients. Although controversy exists regarding the attributable mortality of these infecThe opinions expressed herein are not to be construed as official or as reflecting the policy of either the Departments of the Army or the Departments of Defense. Manuscript received November 15, 2004; accepted November 16, 2004. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (www.chestjournal. org/misc/reprints.shtml). Correspondence to: Andrew Shorr, MD, MPH, Pulmonary, Critical Care, and Sleep Medicine, Walter Reed Army Medical Center, 6900 Georgia Ave NW, Washington, DC 20307; e-mail: [email protected] Clinical Investigations in Critical Care

tions, multiple reports2,3 have documented that bacteremia arising in the ICU increases both length of stay and hospital costs. For example, one study4 documented that nosocomial bacteremia adds ⬎ 7 days to the duration of ICU hospitalization and leads to ⬎ $20,000 in excess costs. The growing prevalence of antibiotic-resistant organisms coupled with the increasing severity of illness seen in ICU subjects has further complicated the management of these BSIs. Current efforts to address the burden of nosocomial infection in the ICU have focused on prevention. Common approaches emphasize more judicious use of antibiotics, improved hand hygiene, and the cohorting of patients having infection or colonization with resistant pathogens.5,6 Specific attempts to prevent catheter-related BSIs have included improved care of central venous catheters (CVCs), use of chlorhexidine rather than povoidodine for site preparation prior to CVC insertion, and routine employment of full barrier drapes.7,8 Suggestions for avoiding secondary nosocomial bacteremia (eg, bacteremias that develop subsequent to an infection at another site) generally address the prevention of other ICU-acquired infections, such as ventilator-associated pneumonia.9 Efforts at prevention originate from an improved understanding of the factors associated with the development of nosocomial infection. Transfusion of packed RBC (pRBCs) has emerged as a potential risk factor for nosocomial infection. For example, Claridge et al10 revealed that pRBC transfusion independently increased the risk for nosocomial infection in critically ill trauma patients. Similarly, Taylor and coworkers11 documented a correlation between the administration of pRBCs and nosocomial infections. Most prior investigations into the link between transfusion and infection have been limited because they describe single-center rather than multicenter experiences or have pooled all infections into a common end point. Such limitations expose these investigations to confounding and various forms of bias. More conclusive evidence of a relationship between transfusion and nosocomial infection must arise from more heterogeneous populations of critically ill patients. Similarly, since the risk factors for various types of ICU-acquired infections differ, it is also important to explore the relationship between specific nosocomial infections and transfusion practice. Because of these issues and changing patterns of pRBC utilization in the ICU, we sought to determine the relationship, if any, between pRBC transfusion and ICU-acquired bacteremia. We conducted a secondary analysis of a large, multicenter, prospective www.chestjournal.org

observational trial to explore both transfusion practice and outcomes in the ICU.

Materials and Methods Primary Study Synopsis Between August 2000 and April 2001, 284 adult ICUs in the United States participated in the CRIT study; the details and primary results from this trial have been described previously.12 The study prospectively tracked the utilization of pRBCs in critically ill patients and included nearly 5,000 individuals. Researchers frequently evaluated enrolled patients as to their clinical status and recorded outcomes including survival, ICU length of stay, need for surgical procedures, and development of infection. Subjects were followed up until death, hospital discharge, or up to 30 days after ICU admission, which ever came first. The present project represents a secondary, subgroup analysis of the larger trial. The institutional review board of each participating hospital approved the original study. Subjects and End Points All subjects expected to remain in the ICU for at least 48 h were included in this study. We excluded from evaluation all patients with a BSI on ICU admission. Additionally, to limit the confounding from BSIs related to infections arising prior to ICU admission, we further excluded those persons who acquired a BSI within the first 48 h of ICU hospitalization. In other words, only patients lacking a BSI both at admission to the ICU and 48 h later comprised the “at-risk” population of interest. Documentation of a new BSI infection represented our primary end point. We did not distinguish between primary and secondary BSIs. For patients who were admitted to the ICU more than once, only the first ICU stay was studied. Moreover, among those with more than one BSI diagnosed during their ICU admission, the initial BSI served as the end point. Documentation of a BSI required a positive blood culture result that was believed by the investigator not to represent a contaminant. The definition of a contaminant was left to the investigator in accordance with standard infectioncontrol accounting guidelines for identifying nosocomial infections. Analysis of Potential Risk Factors We compared patients who acquired a BSI while in the ICU to those who did not. Rates of pRBC transfusion and numbers of pRBC units transfused were compared between the two groups. For individuals with BSIs, we only analyzed pRBCs administered prior to the diagnosis of the BSI. Since pRBCs transfused after the discovery of a BSI could not logically be causally linked to the BSI, we specifically controlled for this potential confounder. Other variables recorded included patient demographics, primary admitting diagnosis, and ICU type (medical, surgical, or mixed). Severity of illness was measured with both the APACHE (acute physiology and chronic health evaluation) II score and the sequential organ failure assessment (SOFA) score.13,14 Because we restricted our investigation to patients who lacked BSIs at 48 h after ICU admission, we also noted severity of illness at days 3 to 4 (SOFA). Chronic comorbid medical conditions such as cardiac disease, malignancy, and renal disorders were documented. Process-of-care variables such as need for mechanical ventilation (MV), placement of a CVC, use of parenteral nutrition, and treatment with antibiotics at admission to the ICU CHEST / 127 / 5 / MAY, 2005

1723

represented additional variables of interest. Baseline treatment with antibiotics was subcategorized as to antibiotic class (eg, quinolones, macrolides) To control for confounding due to differences in ICU length of stay and therefore the time observed for the development of a BSI, we measured duration of ICU stay. We measured ICU length of stay prior to the diagnosis of a BSI in the subjects who acquired a BSI. This was done for two reasons: (1) to allow us to correct for this potential confounder, and (2) to asses if any differences we observed simply reflected differences in the time the patient was considered “at risk” for a BSI. To be conservative, for those without a BSI diagnosis, we counted the entire ICU length of stay as the time at risk. Statistical Analysis For univariate analysis, continuous variables are reported as mean ⫾ SD and were assessed using univariate logistic regression. Categorical variables are reported as frequency distributions and were analyzed with the ␹2 test unless the sample size was small and then we employed the Fisher exact test. Multivariate logistic regression was used to identify independent variables associated with the diagnosis of a new BSI. Factors significant at the 0.2 level were entered into the model in a stepwise elimination process. Adjustments were made for colinearity. We report adjusted odds ratios (ORs) and their corresponding 95% confidence intervals (CIs). All tests were two tailed, and a p value ⬍ 0.05 was predetermined to represent statistical significance. Analyses were completed using statistical software (SAS 8.2; SAS Institute; Cary, NC).

Results The original CRIT trial enrolled 4,892 subjects, and 1,390 patients were excluded from the present analysis.12 Of the 1,390 patients, 1,153 were discharged early from ICU, 94 died early in the study, and 143 were either admitted with a BSI or had one diagnosed within 48 h after ICU admission. Of the remaining 3,502 individuals, 117 patients (3.3%) acquired a new BSI. The median length of stay in the ICU prior to the onset of a BSI was 11 days (range, 5 to 29 days) As shown in Table 1, gender did not correlate with BSI. However, patients with BSIs were younger than those without this complication (56.7 ⫾ 16.4 years vs 60.3 ⫾ 18.1 years, p ⫽ 0.037) [mean ⫾ SD]. There was a trend toward two admitting diagnoses being more frequent in those with BSIs: postoperative admission and trauma. Although postoperative patients accounted for only one in five persons in the data set, more than one in four BSIs occurred in this population (p ⫽ 0.052). Similarly, trauma patients were 1.53 times (95% CI, 0.93 to 2.44; p ⫽ 0.096) more likely to have an ICU-acquired, nosocomial BSI than were nontrauma patients. The most frequent comorbid illnesses in the population were hypertension (44.7%) and cardiac disease (40.1%). Approximately one in six subjects had a malignancy. The distribution of chronic medical conditions, however, did not correlate with BSIs (Table 1). 1724

Table 1—Study Cohort Characteristics*

Variables

Patients Patients With Without BSI BSI p (n ⫽ 117) (n ⫽ 3,385) Value

Demographics Male gender 57.3 54.8 0.854 Age, yr 56.7 ⫾ 16.4 60.3 ⫾ 18.1 0.037 Major admitting diagnosis Postoperative 26.5 19.3 0.052 Respiratory failure 32.5 34.4 0.664 Primary hematologic disease 1.7 0.8 0.253 Trauma 18.0 12.7 0.096 Pneumonia 17.1 16.6 0.888 Hemodynamic instability 12.8 9.5 0.237 Primary neurologic diagnosis 8.6 8.2 0.879 Acute respiratory distress 4.3 3.4 0.622 syndrome Cardiovascular 12.0 12.5 0.857 GI bleeding 5.1 6.7 0.494 Sepsis 13.7 10.3 0.242 Comorbid illnesses Cardiac disease 35.0 40.3 0.312 Hypertension 45.3 44.7 0.692 Peripheral vascular disease 7.7 10.5 0.625 Diabetes mellitus 28.2 23.3 0.273 Pulmonary disease 35.0 34.9 0.670 Thromboembolic disease 4.3 4.4 0.522 Malignancy 18.0 17.6 0.624 Severity of illness APACHE II score 22.3 ⫾ 8.1 20.5 ⫾ 8.0 0.013 SOFA score on admission 8.0 ⫾ 4.1 6.5 ⫾ 3.6 ⬍ 0.001 SOFA score day 3–4 7.7 ⫾ 4.3 5.4 ⫾ 3.8 ⬍ 0.001 MV on admission 58.1 50.8 0.120 ICU length of stay 11.5 ⫾ 5.7 8.9 ⫾ 7.2 ⬍ 0.001 ICU type 0.996 Surgical ICU 23.9 23.8 Medical ICU 35.0 35.4 Combined SICU and MICU 41.0 40.8 Process of care Central line 30.8 22.3 0.030 Parenteral nutrition during ICU 12.0 7.4 0.067 stay Any baseline antibiotic use at 82.1 74.8 0.074 admission Particular antibiotics administered Aminoglycosides 5.1 8.1 0.245 ␤-Lactams 21.4 20.7 0.851 Cephalosporins 52.1 35.5 ⬍ 0.001 Macrolides 6.8 9.9 0.274 Quinolones 21.4 23.1 0.667 Sulfonamides 2.6 2.7 0.920 Tetracyclines 0 0.5 0.429 Vancomycin 14.5 13.6 0.771 pRBC transfusion Transfused 76.1 48.7 ⬍ 0.001 Units of pRBCs transfused 4.0 ⫾ 4.6 2.3 ⫾ 4.3 ⬍ 0.001 *Data are presented as % or mean ⫾ SD.

With respect to severity of illness, the baseline APACHE II score was higher among persons with an ICU-acquired BSI (22.3 ⫾ 8.1 vs 20.5 ⫾ 8.0, Clinical Investigations in Critical Care

p ⫽ 0.013). The day 3– 4 SOFA score was also significantly higher in patients who acquired a BSI subsequently (7.7 ⫾ 4.3 vs 5.4 ⫾ 3.8, p ⬍ 0.001). On average (Table 1), subjects with a BSI were hospitalized in the ICU longer before the diagnosis of the BSI than those without infection (11.5 ⫾ 5.7 days vs 8.9 ⫾ 7.2 days, p ⬍ 0.001). Several process-of-care variables correlated in univariate analysis with the development of a nosocomial BSI. Having a CVC placed, for instance, heightened the risk for bacteremia nearly 50% (OR, 1.56; 95% CI, 1.04 to 2.32; p ⫽ 0.030). Need for MV on admission, however, was not associated with BSI. The relationship between BSI and either use of parenteral nutrition or treatment with antibiotics at admission to the ICU only approached statistical significance (p ⫽ 0.067 and 0.074, respectively). When analyzing classes of antibiotics, though, there was a positive relationship between utilization of cephalosporins and later bacteremia. Half of the study population received pRBC transfusions at some point during their ICU stay. More than three fourths of patients with an ICU-related BSI received pRBCs, compared to 48.7% of patients without this complication (p ⬍ 0.001). In addition to being more likely to be administered pRBCs, patients with later BSIs received on average more units of pRBCs (4.0 ⫾ 4.6 U vs 2.3 ⫾ 4.3 U, p ⬍ 0.001). Figure 1 reveals the unadjusted relationship between transfusion amount and BSI and demonstrates that with increasing rates of transfusion there was a related rise in the incidence of BSIs. In multivariate analysis, three variables remained independently associated with ICU-acquired BSIs (Table 2). Subjects treated with cephalosporins faced an increased risk for subsequent ICU BSIs (OR, 1.84; 95% CI, 1.26 to 2.68; p ⫽ 0.002). Severity of illness on days 3 to 4 as measured by SOFA score was also positively related to the diagnosis of later BSI. For each 1-point increase in SOFA score, the probability for BSI increased by ⬎ 10% (OR, 1.11; 95% CI, 1.06 to 1.16; p ⬍ 0.001). Transfusion remained

Figure 1. Transfusion volume and ICU-acquired bloodstream infections. www.chestjournal.org

Table 2—Independent Predictors of ICU-Acquired Bloodstream Infection Variables

OR

95% CI

p Value

Baseline cephalosporin use Day 3–4 SOFA score Any pRBC transfusion Amount transfused, U* 1–2 3–4 ⬎4

1.84 1.11 2.23

1.26–2.68 1.06–1.16 1.43–3.52

0.002 ⬍ 0.001 ⬍ 0.001

1.89 2.41 2.63

1.10–3.23 1.33–4.35 1.52–4.53

0.021 0.004 ⬍ 0.001

*Estimates from a separate model in which transfusion categories replace the yes/no categorical transfusion variables.

the only other significant correlate of ICU-acquired bacteremia. Transfusion more than doubled the likelihood of BSI (OR, 2.23; 95% CI, 1.43 to 3.52; p ⬍ 0.001). This effect was independent of the time hospitalized in the ICU. Additionally, as show in Table 2, the probability of a BSI increased as the number of units of pRBCs administered escalated. Discussion This analysis of a large, multicenter, prospective, observational registry of critically ill patients reveals that pRBC transfusion is independently associated with ICU-acquired BSI. This relationship is independent of a number of potential confounders, including severity of illness both at admission and later during the ICU stay, and is also independent of ICU length of stay. Additionally, we detected a correlation between pRBC transfusion and BSI with both small and large transfusion volumes. This suggests that there was no clear “threshold effect” at which transfusion amount increased the probability of ICU-acquired bacteremia. Although prior investigations10,11 have documented a link between transfusion and nosocomial infection, few reports have specifically focused on BSIs. Taylor and colleagues,11 for example, observed an increased risk for nosocomial infection in a mixed cohort of critically ill patients. Their retrospective study of ⬎ 1,700 patients suggested that pRBC transfusion heightened the risk for nosocomial infection; however, they pooled several types of infections into a combined end point and did not break out specific rates for BSIs. Among trauma patients, multiple authors have implicated transfusion as a potential contributor to the relatively high rates of nosocomial infection seen in this population. Agarwal et al15 retrospectively analyzed 5,366 consecutive trauma admissions and explored predictors of infection. In their population, BSI was the third-mostprevalent infection. After controlling for multiple CHEST / 127 / 5 / MAY, 2005

1725

confounders to include type of trauma, presence or absence of shock, and Glasgow coma score, only severity of illness (as measured by the injury severity score) and transfusion remained significant predictors of infection. The finding that both severity of illness and transfusion relate to subsequent BSI is consistent with our results. Similarly, Claridge and coworkers10 evaluated the effects of pRBC transfusion on infection in trauma patients and documented that transfusion not only increases the likelihood of later infection but does so in a dose-dependent fashion. Limited information exists regarding transfusion and ICU-acquired BSIs. Most descriptions of predictors of nosocomial bacteremia have not examined a potential role for transfusion. Laupland and coworkers4 documented an ICU-acquired BSI rate of 4.4%, similar to our rate of approximately 3%, in a large data set from multiple hospitals in Canada. They also illustrated the impact of BSIs as it relates to increasing both morbidity and mortality.4 Although they stated that a lower hematocrit was an independent predictor of BSI, they lacked information on transfusion. However, some researchers have observed a nexus between transfusion and BSI but have not specifically focused on the ICU. In a case-control study of all hospital-acquired bacteremias by Duggan et al,16 the adjusted OR for BSI for transfusion measured 2.74 (95% CI, 1.28 to 5.88). This magnitude of effect parallels what we have noted in a more selected population. In the outpatient setting, transfusion may also alter the risk for bacteremia. Among cancer patients with long-term venous catheters, Hanna and Raad17 reported that use of blood products through the catheter was the only independent predictor of catheter-related BSIs. In short, several analyses indicated a strong association between transfusion and both non–ICUrelated bacteremias and nosocomial infections in general. Our observations bolster the hypothesis that transfusion practice is associated with rates of nosocomial infection. Unlike many earlier reports, we relied on prospective data from a large cohort of critically ill subjects. Additionally, we explored a specific type of infection. In contrast to other infections that rely on clinical criteria for diagnosis, and thus may be prone to overdiagnosis (eg, as ventilatorassociated pneumonia), bacteremia has a precise definition. Other unique aspects of our effort include the relatively large sample size and the fact that we had information regarding severity of illness both at admission and later in the ICU course. Beyond transfusion, use of cephalosporins and severity of illness were linked with BSI. As noted above, severity of illness has previously been shown to correlate with a number of nosocomial infections. 1726

The relationship between cephalosporins and later BSI may reflect issues with selection pressure or the fact that certain cephalosporins such as ceftazadime are associated with the emergence of resistant pathogens and superinfection.18 Lack of precise information regarding which specific agents were employed, however, limits our ability to speculate. Cephalosporins as a class were the most commonly administered antimicrobials, so that treatment with cephalosporins may be a marker for some other factor we were not able to expressly control for in our modeling. Biologically, pRBC transfusion might alter the risk for BSI by perturbing the host’s cytokine profile. In vitro models have revealed that transfusion leads to activation of various parts of the cytokine cascade.19 Conversely, in vivo, in patients undergoing cardiopulmonary bypass, transfusion results in a significant rise in proinflammatory cytokines such as interleukin-6.20 Direct measurement of cytokines in stored pRBCs reveals high levels of interleukin-6 that persist throughout the storage of the product.21 Furthermore WBCs from the donor that are not removed from the pRBCs could promote T-cell activation.22 In turn, this could result in subtle changes in the host’s immune status, which predispose him/her to infection. The recent Canadian experience with universal leukoreduction, however, suggests that more than the donor’s WBCs are involved in this process.23 In a retrospective analysis of ⬎ 14,000 patients before and after institution of a universal leukoreduction program, Hebert et al23 failed to note a change in the rate of serious nosocomial infections. Our study has several important limitations. First, although the data for the project were prospectively collected, this is a retrospective assessment with all the attendant limitations of this approach. Specifically, our methods only allow us to demonstrate association and not causation. As such, our findings should be viewed as hypothesis generating. However, unlike most retrospective reports, including those from other large administrative data sets, our primary end point was prospectively defined. Second, our results may be confounded by some variable(s) we could not take into account. For example, we lacked information on some process-of-care issues such as types and locations of CVCs placed. Additionally, ICU nurse-to-patient ratios have been shown to predict nosocomial infection and we could not adjust for this.24 Severity of illness represents another potential confounder. We attempted to ensure that our findings simply do not reflect the fact that more severely ill patients remain in the ICU longer and are exposed to more transfusions. Unfortunately, we did not have daily severity-of-illness scores, and hence cannot entirely exclude that our results may have been affected by this possibility. Clinical Investigations in Critical Care

Daily data on severity of illness would have allowed us to more precisely control for changes in severity of illness over time. Third, events outside the ICU might affect the rate of BSIs. We only possessed data regarding the patient’s ICU course. To attempt to address this, we precisely defined the at-risk cohort as only patients who had been in the ICU for at least 48 h and who lacked a BSI up until that point. Although this approach should help control for non-ICU factors, it does not guarantee that such issues did not bias our observations. Relatedly, we did not have access to information regarding transfusion prior to ICU admission. We also did not have microbiology results from cultures. Finally, use of other blood products was not recorded. Although not thought to affect the immune system to the extent that pRBC transfusion can, both platelet transfusion and administration of fresh frozen plasma may influence the immune system.25 Of note, one author concluded that there “are no studies that substantiate the concept that platelets inherently facilitate microbial pathogenesis . . . or inhibit the immune response.”26 In summary, we noted a positive relationship between pRBC transfusion and ICU-acquired BSI. Patients receiving pRBCs were more than twice as likely as those who did not receive pRBCs to acquire a BSI. This association was independent of multiple confounders, including severity of illness, underlying diagnosis, use of antimicrobials, and ICU length of stay. As evidence mounts that pRBC transfusions are associated with higher rates of subsequent nosocomial infection, physicians need to more carefully consider their decisions about when and if their patients should receive transfusions. Future prospective trials of transfusion practice need to capture nosocomial infection as a specific end point.

6

7

8

9

10

11

12

13 14 15 16 17

18

References 1 Vincent JL, Bihari DJ, Suter PM, et al. The prevalence of nosocomial infection in intensive care units in Europe: results of the European Prevalence of Infection in Intensive Care (EPIC) Study; EPIC International Advisory Committee. JAMA 1995; 274:639 – 644 2 Renaud B, Brun-Buisson C, ICU-Bacteremia Study Group. Outcomes of primary and catheter-related bacteremia: a cohort and case-control study in critically ill patients. Am J Respir Crit Care Med 2001; 163:1584 –1590 3 Warren DK, Zack JE, Elward AM, et al. Nosocomial primary bloodstream infections in intensive care unit patients in a nonteaching community medical center: a 21-month prospective study. Clin Infect Dis 2001; 33:1329 –1335 4 Laupland KB, Zygun DA, Davies HD, et al. Population-based assessment of intensive care unit-acquired bloodstream infections in adults: incidence, risk factors, and associated mortality rate. Crit Care Med 2002; 30:2462–2467 5 Jarvis WR. Benchmarking for prevention: the Centers for Disease Control and Prevention’s National Nosocomial Inwww.chestjournal.org

19

20 21 22

23

fections Surveillance (NNIS) system experience. Infection 2003; 31(Suppl 2):44 – 48 Healthcare Infection Control Practices Advisory Committee and Hand-Hygiene Task Force; Society for Healthcare Epidemiology of America; Association for Professionals in Infection Control and Epidemiology; Infection Diseases Society of America. Guideline for hand hygiene in healthcare settings. J Am Coll Surg 2004; 198:121–127 Eggimann P, Pittet D. Overview of catheter-related infections with special emphasis on prevention based on educational programs. Clin Microbiol Infect 2002; 8:295–309 O’Grady NP, Alexander M, Dellinger EP, et al. Guidelines for the prevention of intravascular catheter-related infections. Infect Control Hosp Epidemiol 2002; 23:759 –769 Kollef MH. Prevention of hospital-associated pneumonia and ventilator-associated pneumonia. Crit Care Med 2004; 32: 1396 – 405 Claridge JA, Sawyer RG, Schulman AM, et al. Blood transfusions correlate with infections in trauma patients in a dose-dependent manner. Am Surg 2002; 68:566 –572 Taylor RW, Manganaro L, O’Brien J, et al. Impact of allogenic packed red blood cell transfusion on nosocomial infection rates in the critically ill patient. Crit Care Med 2002; 30:2249 –2254 Corwin HL, Gettinger A, Pearl RG, et al. The CRIT Study: Anemia and blood transfusion in the critically ill current; clinical practice in the United States. Crit Care Med 2004; 32:39 –52 Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of disease classification system. Crit Care Med 1985; 13:818 – 829 Ferreira FL, Bota DP, Bross A, et al. Serial evaluation of the SOFA score to predict outcome in critically ill patients. JAMA 2001; 286:1754 –1758 Agarwal N, Murphy JG, Cayten CG, et al. Blood transfusion increases the risk of infection after trauma. Arch Surg 1993; 128:171–176 Duggan J, O’Connell D, Heller R, et al. Causes of hospitalacquired septicaemia: a case control study. Q J Med 1993; 86:479 – 483 Hanna HA, Raad I. Blood products: a significant risk factor for long-term catheter-related bloodstream infections in cancer patients. Infect Control Hosp Epidemiol 2001; 22:165– 166 Paterson DL, Ko WC, Von Gottberg A, et al. International prospective study of Klebsiella pneumoniae bacteremia: implications of extended-spectrum ␤-lactamase production in nosocomial infections. Ann Intern Med 2004; 140:26 –32 Biedler AE, Schneider SO, Seyfert U, et al. Impact of alloantigens and storage-associated factors on stimulated cytokine response in an in vitro model of blood transfusion. Anesthesiology 2002; 97:1102–1109 Fransen E, Maessen J, Dentener M, et al. Impact of blood transfusions on inflammatory mediator release in patients undergoing cardiac surgery. Chest 1999; 116:1233–1239 Vamvakas EC, Blajchman MA. Deleterious clinical effects of transfusion-associated immunomodulation: fact or fiction? Blood 2001; 97:1180 –1195 Kristiansson M, Soop M, Saraste L, et al. Cytokines in stored red blood cell concentrates: promoters of systemic inflammation and simulators of acute transfusion reactions? Acta Anaesthesiol Scand 1996; 40:496 –501 Hebert PC, Fergusson D, Blajchman MA, et al. Clinical outcomes following institution of the Canadian universal leukoreduction program for red blood cell transfusions. JAMA 2003; 289:1941–1949 CHEST / 127 / 5 / MAY, 2005

1727

24 Hugonnet S, Harbarth S, Sax H, et al. Nursing resources: a major determinant of nosocomial infection? Curr Opin Infect Dis 2004; 17:329 –333 25 Elzey BD, Tian J, Jensen RJ, et al. Platelet-mediated modulation of adaptive immunity: a communication link between

1728

innate and adaptive immune compartments. Immunity 2003; 19:9 –19 26 Yearman MR, Bayer AS. Antimicrobial host defense. In: Michelson AD, ed. Platelets. New York, NY: Academic Press, 2002; 470

Clinical Investigations in Critical Care