Secular Trends in Nosocomial Bloodstream Infections in a 55-Bed Cardiothoracic Intensive Care Unit Steven M. Gordon, MD, Janet M. Serkey, RN, Thomas F. Keys, MD, Thomas Ryan, MD, Cynthia A. Fatica, RN, Steven K. Schmitt, MD, Judith A. Borsh, RN, Delos M. Cosgrove III, MD, and Jean-Pierre Yared, MD Departments of Infectious Disease, Nursing, Cardiothoracic Anesthesia, and Cardiothoracic Surgery, The Cleveland Clinic Foundation, Cleveland, Ohio
Background. Although bloodstream infections (BSIs) occur more frequently in intensive care unit patients than in ward patients, most studies of nosocomial BSIs in critically ill patients have not distinguished between intensive care unit populations beyond surgical, medical, and pediatric patients. Methods. The primary objective of this study was to characterize the secular trends in rates of nosocomial BSIs for all pathogens among patients admitted to a busy cardiothoracic intensive care unit in a single tertiary care institution between January 1986 and December 1995. Patients with nosocomial BSIs were identified through continual prospective surveillance. Results. A total of 40,207 patients were admitted to the cardiothoracic intensive care unit during the 10-year study period, and 804 episodes of nosocomial BSIs among 681 patients were identified. The mean crude BSI infection rate was 6.0 per 1,000 patient-care days and increased linearly during the study period (range, 4.4 to
8.1 per 1000 patient-care days), and approached statistical significance (p value 5 0.07). The most common organisms causing BSIs were Staphylococcus aureus (12%), coagulase-negative staphylococci (11%), Candida albicans (11%), Pseudomonas aeruginosa (10%), and Enterococci (9%). The leading sources of nosocomial BSIs were primary BSIs (33%), intravascular devices (27%), lower respiratory tract infections (17%), and surgical wound infections (12%). The etiologic fraction or the proportion of deaths in cardiothoracic intensive care unit patients with BSIs was 15-fold higher than those patients without BSIs (37% versus 2.5%, p < 0.001). Conclusions. Rates of nosocomial BSIs among patients in our cardiothoracic intensive care unit have increased linearly during the past decade and patients with nosocomial BSIs have an increased risk of in hospital mortality. (Ann Thorac Surg 1998;65:95–100) © 1998 by The Society of Thoracic Surgeons
N
groups among patients admitted to a 55-bed cardiothoracic intensive care unit (CTICU) in a tertiary health care institution.
osocomial bloodstream infections (BSIs) occur two to seven times more often in intensive care unit (ICU) patients than in ward patients with concomitant increases in attributable mortality and economic costs [1– 4]. The attributable mortality rate from nosocomial BSI was recently estimated to be 35% in a cohort of critically ill surgical patients at the University of Iowa [4]. These infections were associated with a doubling of ICU stay and an excess length of hospital stay of 24 days in survivors and significant economic burden ($40,000 per survivor). Although the rates of nosocomial BSIs vary by ICU type with the highest rates found among patients in surgical ICUs [1], most studies of nosocomial BSIs in critically ill patients have not distinguished among populations beyond surgical, medical, and pediatric ICU patients. The primary objective of this study was to characterize the secular trends in nosocomial BSI rates for pathogen
Accepted for publication July 10, 1997. Address reprint requests to Dr Gordon, The Cleveland Clinic Foundation, Mailstop S-32, 9500 Euclid Ave, Cleveland, OH 44195-5066 (e-mail:
[email protected]).
© 1998 by The Society of Thoracic Surgeons Published by Elsevier Science Inc
Material and Methods Hospital Setting The Cleveland Clinic Foundation is a 1,100-bed tertiary referral center with an active cardiothoracic surgical service. The CTICU is a 55-bed unit that accommodated between 2,980 and 3,700 open heart surgical patients and an additional 300 to 1,000 noncardiac thoracic cases annually. Pediatric open heart patients were also cared for in the CTICU for a brief period of time before moving to a dedicated pediatric unit. Twelve cardiothoracic staff surgeons performed approximately 42,000 cardiothoracic procedures between 1986 and 1995; approximately 49% were isolated coronary artery bypass grafting procedures, 20% involved valve repair or replacement, 20% were noncardiac thoracic procedures, 4% were congenital cases, and 2% were transplant procedures (heart and lung). Patients undergoing open heart operations rou0003-4975/98/$19.00 PII S0003-4975(97)01039-4
96
GORDON ET AL SECULAR TRENDS IN NOSOCOMIAL BLOODSTREAM INFECTIONS
Ann Thorac Surg 1998;65:95–100
Sparks, MD) system was used between 1979 and 1983 when it was replaced by the BACTEC 660 system, each involving the inoculation of 3 to 5 mL of blood into aerobic and anaerobic bottles. In 1986 this system was replaced by the 10-mL Isolator system (Wampole Laboratories, Cranbury, NJ) and a bottle containing tryptic soy broth (Difco Laboratories, Detroit, MI). Beginning in 1992 the tryptic soy broth bottle was replaced successively by the ESP Aerobic 80A bottle (Difco Laboratories) and the BACTEC Plus Aerobic/F bottle (Becton Dickinson).
Statistical Methods Fig 1. Secular trends in nosocomial bloodstream infection rates. Rates of nosocomial bloodstream infections in the cardiothoracic intensive care unit at the Cleveland Clinic between 1986 and 1995 demonstrate a linear increase (r 5 0.35, p 5 0.07).
tinely received perioperative prophylactic antibiotics with either cefamandole (from January 1986 to May 1991), cefuroxime (after May 1991), or vancomycin (for penicillin-allergic patients). Prophylactic antibiotics were routinely given in the operating room by the anesthesiologist immediately before the induction of general anesthesia and were continued for 48 hours postoperatively. Body substance isolation became the standard of practice in the hospital in October 1988. Selective digestive decontamination was instituted in a study protocol for prevention of aspiration pneumonia in the CTICU during a 9-month period in 1990. Since 1993, all procedures involving vascular catheter insertions and changes have been routinely performed with sterile drapes, gowns, gloves, and masks.
Surveillance Methods Continual prospective surveillance for all nosocomial BSIs has been ongoing since 1983 using standard methods. Surveillance was performed by the same infection control practitioner during the entire study period. All patients admitted to the CTICU between January 1, 1986, and December 31, 1995, were included in the study. An episode of nosocomial BSI was defined as the occurrence of a positive blood culture(s) from any patient admitted to the CTICU for at least 48 hours with clinical signs of sepsis according to standard definitions by the Centers for Disease Control and Prevention. If blood cultures were repeatedly positive for the same organism, the occurrence was considered to be one episode of BSI. If another organism was subsequently isolated from blood cultures in the same patient, it was designated as a separate episode. The BSI was considered to be polymicrobial if more than one organism was isolated from the same blood culture. A primary BSI was defined as a bloodstream infection for which a source was not identified.
Microbiologic Techniques The blood culture methods during this 10-year study period varied. The BACTEC 460 (Becton-Dickinson,
The sum of all nosocomial BSIs (episodes) documented each year was considered in statistical analyses. Rate per 1,000 patient-care days were calculated. Linear trends were analyzed using the simple linear regression analysis (least-squares method), with year considered as an independent variable and numbers of BSI episodes as dependent variables. Odds ratios and 95% confidence intervals for death were assessed with x2 tests. Adjusted odds ratios were assessed with use of the Mantel-Haenszel procedure. Statistical analysis was performed with EpiInfo version 6.02 (CDC, Atlanta, 1994). All tests were two-tailed and a p value of less than 0.05 was considered significant.
Results Incidence Rates of Nosocomial Bloodstream Infections Between 1986 and December 1995, 40,207 patients were admitted to the CTICU, a mean of 4,020 annually (range, 3,343 to 4,709), representing a total of 131,245 patient-care days. A total of 804 episodes of nosocomial BSIs were identified in 681 patients, which included both primary and secondary BSIs. The overall BSI rate was 6.0 per 1,000 patient-care days and increased during the study period (Fig 1; Table 1). Although the BSI rates were higher in the last 3 years of the study period the linear trend was not statistically significant (estimated slope was 0.2224 with Table 1. Nosocomial Bloodstream Infection Rate in the Cardiothoracic Intensive Care Unit, The Cleveland Clinic Foundation, 1986 –1995
Year
Episodes of Bloodstream Infections
Patient-Care Days
Bloodstream Infection Ratea
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
71 56 70 78 71 64 75 109 89 121
11,387 12,611 12,913 12,962 13,287 12,999 14,773 14,796 13,978 14,956
6.2 4.4 5.4 6 5.3 4.9 5.1 7.3 6.4 8.1
804
134,671
Total a
Per 1,000 patient-care days.
6
Ann Thorac Surg 1998;65:95–100
GORDON ET AL SECULAR TRENDS IN NOSOCOMIAL BLOODSTREAM INFECTIONS
97
Table 2. Episodes of Nosocomial Bloodstream Infection in the Cardiothoracic Intensive Care Unit, Cleveland Clinic Foundation, 1986 –1995 Episodesa Microorganism Gram-positive cocci Coagulase-negative staphylococci Staphylococcus aureus Enterococci Others Gram-negative rods Pseudomonas aeruginosa Klebsiella sp Enterobacter sp Serratia marcescens Others Fungi Candida albicans Other than Candida albicans Polymicrobial Total a
1986 –1987
1988 –1989
1990 –1991
1992–1993
1994 –1995
Total
47 (37) 11 (8.7) 11 (8.7) 20 (16) 2 (1.6) 53 (42) 12 (9.4) 2 (1.6) 13 (10.2) 12 (9.4) 12 (9.4) 25 (19.7) 20 (15.7) 5 (3.9) 2 (1.6)
55 (37) 13 (8.9) 20 (13.5) 15 (10.1) 4 (2.7) 58 (39.2) 14 (9.4) 5 (3.4) 10 (6.8) 5 (3.4) 19 (12.8) 26 (17.6) 18 (12.1) 8 (5.4) 9 (6.1)
47 (35) 12 (8.9) 16 (12) 15 (11.1) 1 (0.7) 57 (42.2) 22 (16.3) 6 (4.4) 11 (8.1) 1 (0.7) 15 (11.1) 24 (17.8) 12 (8.9) 12 (8.9) 7 (5.2)
75 (40.7) 12 (6.5) 38 (20.6) 16 (8.7) 2 (1.1) 64 (34.8) 18 (9.8) 4 (2.2) 13 (7.1) 8 (4.3) 17 (9.2) 31 (16.8) 17 (9.2) 14 (7.6) 14 (7.6)
71 (33.4) 46 (21.9) 15 (7.1) 6 (2.8) 3 (1.4) 96 (45.7) 17 (8.1) 17 (8.1) 17 (8.1) 16 (7.6) 26 (12.3) 30 (14.7) 18 (8.5) 12 (5.7) 13 (6.2)
295 (37) 94 (11.7) 100 (12.4) 72 (8.9) 12 (1.5) 328 (41) 83 (10.3) 34 (4.2) 64 (7.9) 42 (5.2) 89 (11) 136 (16.9) 85 (10.6) 51 (6.3) 45 (5.6)
127 (100)
148 (100)
135 (100)
184 (100)
210 (100)
804 (100)
Presented as number (%).
an R2 of 0.35 and correlation coefficient of 0.59; p 5 0.07). The 804 episodes of nosocomial BSIs in the CTICU accounted for 23.5% of the episodes of nosocomial BSIs among all patients hospitalized at the Cleveland Clinic during the study period (804 of 3,421 episodes). The hospital-wide rate of nosocomial BSIs remained stable year to year during the study period, with a mean rate of 1.5 episodes of BSIs per 1,000 patient-care days (3,421 episodes in 2,241,948 patient-care days). The most common pathogen groups causing nosocomial BSIs in the CTICU were aerobic gram-negative rods 41% (328 episodes), gram-positive cocci 37% (295 episodes), fungi 17% (136 episodes), and polymicrobial 5% (45 episodes) (Table 1). The most common organisms causing nosocomial BSIs were Staphylococcus aureus (12%), coagulase-negative staphylococci (CNS) (11%), Candida albicans (11%), Pseudomonas aeruginosa (10%), and Enterococci (9%) (Table 1). During the study period the only notable change in the proportion of BSIs between pathogen groups occurred in the 1992 to 1993 time period, when the incidence rate of BSIs caused by gram-positive cocci was higher than aerobic gram-negative rods (75 episodes per 29,569 patient-care days versus 64 episodes per 29,569 patient-care days). The increase in nosocomial BSIs caused by grampositive cocci was attributed to an outbreak of S aureus infections in 1993 in our CTICU. Staphylococcus aureus was the most common bloodstream pathogen during 1992 to 1993 (in part related to the outbreak), accounting for 21% of the total episodes and 51% of the gram-positive coccal episodes of BSIs during this period. Several changes within pathogen-specific groups were observed during the study period (Table 2). Coagulasenegative staphylococci emerged as the most common pathogen causing BSIs during 1994 to 1995, accounting
for 22% of the total episodes and 62% of the grampositive episodes during this period. The prevalence of enterococcal BSIs decreased during the study period, accounting for 8.9% of all episodes, but only 2.8% in 1994 to 1995. Pseudomonas aeruginosa was the most prevalent pathogen among aerobic gram-negative rods, accounting for 10.3% (83 episodes) of the total BSIs, and 25% of the BSIs caused by gram-negative rods during the study period. Other prevalent pathogens within this group included Enterobacter sp. (64 episodes, 7.9% of total BSIs), Serratia marcescens (42 episodes, 5.2%), and Klebsiella sp. (34 episodes, 4.2%) (Table 2). There was no significant increase in secular trends in rates of nosocomial BSIs caused by P aeruginosa, Enterobacter sp, or Serratia marcescens. There was a sharp increase in the isolates of Klebsiella pneumoniae during 1994 to 1995 (17 episodes or 8.1% of the total BSIs) compared with the prior 8 years (17 episodes, 2.9% of total BSIs). There was no change in the proportion of nosocomial BSIs attributed to fungi during the study period (136 episodes or 10.6% of the total BSIs). Notably, the proportion of nosocomial fungemias attributed to Candida sp. other than C albicans did increase during the study period, accounting for 37% of the total nosocomial fungemias (Table 2).
Primary and Secondary Bloodstream Infections Sources were documented for 677 episodes of nosocomial BSIs (those occurring between 1988 and 1995, Table 3). The most common sources for nosocomial BSIs were primary BSIs, 33% (223 episodes); catheter-associated BSIs, 27.1% (184 episodes); lower respiratory tract infections, 17.1% (116 episodes); and wound infections, 11.7% (79 episodes). Genitourinary tract, gastrointestinal tract,
98
GORDON ET AL SECULAR TRENDS IN NOSOCOMIAL BLOODSTREAM INFECTIONS
Ann Thorac Surg 1998;65:95–100
Table 3. Leading Pathogens and Sources of Infection Responsible for 677 Episodes of Nosocomial Bloodstream Infections in the Cardiothoracic Intensive Care Unit Patients, The Cleveland Clinic Foundation, 1988 –1995 Sources of Infection, No. Microorganism Gram-positive cocci Staphylococcal aureus Coagulase-negative staphylococci Enterococci Gram-negative rods Pseudomonas aeruginosa Enterobacter sp Klebsiella sp Serratia Acinetobacter Fungi Candida albicans Other than C. albicans Polymicrobial Total a
Episodesa
Primary
Pneumonia
Surgical Wounds
IV Catheter
UTI
Other
246 (36) 89 (13) 83 (12)
53 14 14
12 6 0
41 32 6
118 32 58
4 0 0
18 5 5
51 (8) 277 (41) 68 (10) 55 (8) 34 (5) 31 (5) 76 (11) 118 (17) 70 (10) 48 (7) 36 (5)
18 81 22 18 11 8 20 77 45 30 14
0 93 25 12 9 17 19 4 1 3 5
8 20 6 6 2 0 8 6 4 2 7
16 39 7 12 6 5 7 25 14 10 7
4 16 5 4 1 0 7 1 1 1 1
5 28 3 3 5 1 0 5 5 2 2
677 (100)
225
114
74
189
22
53
Presented as number (%).
IV 5 intravenous;
UTI 5 urinary tract infection.
infective endocarditis, and other miscellaneous sources accounted for the remaining 11% (75 episodes). The proportion of primary BSIs increased progressively from 24% in 1988 (1.8 per 1,000 patient-care days) to 39% in 1995 (2.6 per 1,000 patient-care days). The proportion of infections caused by intravascular catheters remained stable during the study period with a mean rate of 1.7 per 1,000 patient-care days. The proportion and rate of secondary BSIs caused by lower respiratory tract infections increased during the study period from 11% (0.9 per 1,000 patient-care days) in 1988 to 24% (1.9 per 1,000 patientcare days) in 1995. Gram-positive cocci were responsible for the largest proportion of secondary BSIs caused by intravascular catheters (118 of 187 episodes), wound infections (41 of 71 episodes), and infective endocarditis (10 of 10 episodes). Staphylococcus aureus was the most common cause of seondary BSI episodes from surgical wound infections (32 episodes) and CNS was the most common pathogen associated with intravascular catheters (58 episodes). Aerobic gram-negative bacilli were responsible for the largest proportion of primary BSIs (79 of 218 episodes), as well as secondary BSIs caused by lower respiratory tract infections (90 of 111 episodes), genitourinary tract infections (16 of 22 episodes), and gastrointestinal infections (17 of 25 episodes). Pseudomonas aeruginosa was the most common pathogen associated with secondary BSIs from pneumonias (22 episodes) and urinary tract infections (5 episodes). Fungi were most commonly identified as pathogens in primary BSIs (77 of 118 episodes) and secondary BSIs caused by intravascular catheters (25 episodes).
Crude and Population Attributable Mortality Crude mortality and population-attributable mortality were calculated for all patients with and without nosocomial BSIs between January 1993 and September 1995. Among 246 patients with BSIs (291 episodes), 91 patients died, giving a crude mortality rate of 37%. In contrast, the in-hospital mortality among the 9,313 patients who did not acquire BSIs was 2.5%. Therefore, patients who had nosocomial BSIs in the CTICU showed a risk ratio for death 15-fold higher than those patients in the CTICU who did not have nosocomial BSIs (p , 0.001; risk ratio 5 14.7; 95% confidence interval, 11 to 18).
Comment Although previous studies have reported on the incidence of nosocomial BSI in patients after open heart operations [6 –9] and nosocomial BSIs in patients with implanted prosthetic valves [10 –16], none have studied only patients in a cardiovascular ICU. The primary objective of our study was to characterize pathogen-specific rates of nosocomial BSIs among a large cohort of patients in an ICU who have undergone major cardiothoracic operations at our institution. The mean incidence of nosocomial BSIs reported in our study (6.0 per 1,000 patient-care days) corresponds to the median nosocomial BSI rate for surgical intensive care patients in the Center for Disease Control and Prevention’s National Nosocomial Infections Surveillance hospitals from 1986 to 1990 (6.3 per 1,000 patientcare days) [1], and the mean rate reported for 4,002 patients admitted to a surgical ICU at another tertiary care hospital between 1988 and 1990 (8.6 episodes per
Ann Thorac Surg 1998;65:95–100
GORDON ET AL SECULAR TRENDS IN NOSOCOMIAL BLOODSTREAM INFECTIONS
1,000 patient-care days) [4]. The nosocomial BSI rate for patients in our CTICU was higher than those reported in hospital-wide studies for primary and secondary nosocomial BSI (approximately 1 per 1,000 patient-care days) [17, 18]. The secular increase in nosocomial BSIs we observed among patients in the CTICU, although not statistically significant, occurred during a time when neither our definitions of infection nor our surveillance activities changed. The use of patient-care days in the denominator (versus patients) helps to control for variation in the average length of patient ICU stay (the average CTICU stay for patients undergoing open heart operations at the Cleveland Clinic is less than 72 hours). Some trends in pathogen-specific rates of nosocomial BSIs at our institution parallel other recent studies of hospital-wide primary and secondary nosocomial BSIs. We observed an increase in the importance of CNS as a cause of BSIs, as observed in the Center for Disease Control and Prevention’s National Nosocomial Infections Surveillance hospitals from 1980 to 1989 [18] and at the University of Iowa between 1980 and 1992 [17]. Coagulase-negative staphylococci currently account for onefourth of nosocomial BSIs and several reasons may account for the increase. The first is a consequence of surveillance with greater recognition and reporting of CNS as a pathogen instead of as a contaminant, as well as a consequence of the increased frequency of blood cultures. Although we had no significant changes in our surveillance technique, increased recognition of the clinical importance of CNS may have contributed to the increase in rates. Coagulase-negative staphylococci have emerged as important surgical site pathogens among patients undergoing open heart operations at the Cleveland Clinic, accounting for approximately 23% of surgical wound infections between 1988 and 1994 [19], and 57% of early onset prosthetic valve endocarditis since 1992 (unpublished data). The increasing role of CNS in nosocomial BSIs may partially be explained by the increased use of more invasive intravascular devices, including multilumen intravenous catheters, intraaortic balloon pumps, and left and right heart ventricular assist devices [20]. In our study, virtually all patients in the CTICU would have at least one intravascular device, so patient-care days is equivalent to device-days. Other factors that may contribute to the increased rate of CNS include widespread use of broad-spectrum antibiotics that increase the proportion of CNS resistant to methicillin. Among 100 isolates of CNS causing surgical wound infections in patients undergoing open heart operations at the Cleveland Clinic, 92% were methicillin-resistant [19]. As a group, gram-positive cocci accounted for 37% of the total episodes of BSIs during the study period. They were the leading pathogens causing secondary BSIs related to surgical wound infections (55%) and vascular catheter-associated infections (62%) (Table 3). Notably, S aureus continues to be an important nosocomial bloodstream pathogen for our CTICU patients and was the most prevalent pathogen during the study period, accounting for 12% of the total nosocomial BSIs. As previ-
99
ously noted, the marked increase in episodes and rates of S aureus BSIs in 1993 was related, in part, to an outbreak in the CTICU. Aerobic gram-negative bacilli were the most prevalent pathogen group causing nosocomial BSIs in CTICU patients during the study period. They were the leading pathogens causing secondary BSIs from lower respiratory tract infections (81%), genitourinary tract infections (73%), and gastrointestinal and intraabdominal infections (69%). Pseudomonas aeruginosa, Enterobacter sp., and Serratia sp. were the most common gram-negative pathogens causing BSIs in the CTICU. All of these organisms can be induced to produce b-lactamases, making them resistant to many broad-spectrum penicillin and cephalosporin antibiotics and underscoring the problems of antibiotic resistance in our CTICU. Fungemias accounted for almost 10% of nosocomial BSIs in our patients and for the majority of these episodes no source was identified. Although most episodes were caused by C albicans, the proportion of BSIs caused by species other than C albicans increased during the study period. The adverse effects of nosocomial infections have been measured by either increased mortality rates or excess health care costs. Although the association between nosocomial infection and mortality is well documented, the precise relationship in terms of cause and effect has yet to be defined [22, 23]. The 37% crude mortality rate and the 15-fold higher risk ratio for death in patients with BSIs observed among our CTICU patients with nosocomial BSIs correspond to those noted for patients at the University of Iowa [4, 17]. It should be emphasized that we did not determine whether the nosocomial BSIs contributed to or caused death, and thus attributable mortality rates for nosocomial BSIs were not reported. In summary, the prevention of infection in patients hospitalized in ICUs requires a knowledge of the rates and sources of nosocomial infections, the pathogens causing infections, and their antimicrobial resistance patterns. Hospital-wide ICU BSI rates are frequently reported, but these rates may not reflect differences in factors that may influence the risk of nosocomial BSIs among ICU patients. These may include underlying diseases, severity of illness, duration of ICU stay, and number, type, and duration of invasive devices and procedures [1]. Our data should allow us to determine the potential impact of efforts aimed at prevention of nosocomial BSIs in CTICU patients in the future.
References 1. Jarvis WR, Edwards JR, Culver DH, et al. Nosocomial infection rates in adult and pediatric intensive care units in the United States. Am J Med 1991;91 (Suppl 3B):S185–91. 2. Donowitz LG, Wenzel RP, Hoyt JW. High risk of hospitalacquired infection in the ICU patient. Crit Care Med 1982;10: 355–7. 3. Dashner FP, Frey P, Wolff G, Suter P. Nosocomial infections in intensive care wards: a multicenter prospective study. Intensive Care Med 1982;8:5–9.
100
GORDON ET AL SECULAR TRENDS IN NOSOCOMIAL BLOODSTREAM INFECTIONS
4. Pittet D, Tarara D, Wenzel RP. Nosocomial bloodstream infection in critically ill patients. JAMA 1994;271:1598 – 601. 5. Garner JS, Jarvis WR, Emori TG. CDC definition for nosocomial infections. Am J Infect Control 1988;16:128– 40. 6. Kohman LJ, Coleman MJ, Parker FB. Bacteremia and sternal infection after coronary artery bypass grafting. Ann Thorac Surg 1990;49:454–7. 7. Ferrazzi P, Allen R, Crupi G, Reyes I, Parenzan L, Maisonnet M. Reduction of infection after cardiac surgery: a clinical trial. Ann Thorac Surg 1986;42:321–5. 8. Lockey E, Gonzalez-Lavin L, Ray I, Chen R. Bacteremia after open-heart surgery. Thorax 1973;28:183–7. 9. Dandalides PC, Rutala WA, Sarubbi FA. Postoperative infections following cardiac surgery: association with an environmental reservoir in a cardiothoracic intensive care unit. Infect Control 1984;5:378– 84. 10. Gordon SM, Keys TF. Bloodstream infections in patients with implanted prosthetic cardiac valves. Semin Thorac Cardiovasc Surg 1995;1:2– 6. 11. Sande MA, Johnson WD, Hook EW, Kaye D. Sustained bacteremia in patients with prosthetic cardiac valves. N Engl J Med 1972;286:1067–70. 12. Parker FP, Greiner-Hayes C, Tomar RH, Markowitz AH, Bove EL, Marvasti MA. Bacteremia following prosthetic valve replacement. Ann Surg 1983;197:147–51. 13. Goldmann DA, Hopkins CC, Karchmer AW, et al. Cephalothin prophylaxis in cardiac valve surgery: a prospective,
14. 15. 16. 17. 18. 19.
20. 21. 22.
Ann Thorac Surg 1998;65:95–100
double-blind comparison of two-day and six-day regimens. J Thorac Cardiovasc Surg 1977;73:470–9. Magilligan DJ, Quinn EL, Davila JC. Bacteremia, endocarditis, and the Hancock valve. Ann Thorac Surg 1977;24:508–17. Fang G, Keys TF, Gentry LO, et al. Prosthetic valve endocarditis resulting from nosocomial bacteremia: a prospective multicenter study. Ann Intern Med 1993;119:560–7. Keys TF. Early-onset prosthetic valve endocarditis. Cleveland Clin J Med 1993;60:455–9. Pittet D, Wenzel RP. Nosocomial bloodstream infections: secular trends in rates, mortality, and contribution to total hospital deaths. Arch Intern Med 1995;155:1177– 84. Banerjee SN, Emori TG, Culver DH, et al. Secular trends in nosocomial primary bloodstream infections in the United States, 1980 –1989. Am J Med 1991;91(Suppl 3B):86S–9. Mossad SB, Serkey JM, Longworth DL, Cosgrove DM, Gordon SM. Coagulase-negative staphylococcal sternal wound infections after open heart operations. Ann Thorac Surg 1997;63:395– 401. McCarthy PM, Schmitt SK, Vargo RL, et al. Implantable LVAD infections: implications for permanent use of the device. Ann Thorac Surg 1996;61:359– 65. Salemi C, Morgan J, Padilla S, Morrissey R. Association between severity of illness and mortality from nosocomial infection. Am J Infect Control 1995;23:188–93. Vincent JL, Bihari DL, Suter PM, et al. The prevalence of nosocomial infection in intensive care units in Europe. JAMA 1995;274:639– 44.