Matched Case‐Control Analysis of Polymicrobial Bloodstream Infection in a Neonatal Intensive Care Unit
Objective. To compare and contrast the epidemiology of polymicrobial and monomicrobial bloodstream infections (BSIs) in newborn intensive care unit (NICU) patients.
Design. Retrospective, matched case‐control study.
Setting. The Yale‐New Haven Hospital NICU from 1989 through 2006.
Subjects. NICU patients with BSIs.
Methods. Each neonate with polymicrobial BSI (case patient) was matched to one neonate with monomicrobial BSI (control patient), by birth date, weight, and sex; and univariate and multivariate analyses were performed.
Results. One hundred five cases of polymicrobial BSI were identified in 102 infants, representing 10% of all neonatal BSIs in our institution. Coagulase‐negative staphylococci were the most common organisms recovered from culture. Infants with polymicrobial BSI had later onset of infection than infants with monomicrobial BSI (mean day of life, 37.5 vs 24.0; 
). Polymicrobial BSI occurred more frequently among infants with a severe underlying condition than in those without such a condition (odds ratio [OR], 1.8; 95% confidence interval [CI], 1.1–3.2) and among infants requiring an indwelling central venous catheter for a prolonged duration (mean, 16.9 days, compared with 9.8 days for infants with monomicrobial BSI; 
). Multivariate analysis revealed that later onset of infection (adjusted OR [aOR], 1.02; 95% CI, 1.00–1.04) and presence of a severe underlying condition (aOR, 1.91; 95% CI, 1.12–3.38) were independent risk factors for polymicrobial BSI. No differences in outcome or mortality were observed.
Conclusions. Changes in the microbiology and epidemiology of NICU‐related polymicrobial BSI have occurred since the last North American review. In the present study, although differences were observed, most risk factors and outcomes were similar between monomicrobial BSI and polymicrobial BSI. Epidemiologic surveillance is critical to identify trends associated with neonatal polymicrobial BSI, particularly those that may impact preventative strategies, diagnostic measures, and therapeutic interventions.
Received February 21, 2008; accepted June 17, 2008; electronically published September 9, 2008.
A recent review of polymicrobial infection in adult cancer patients noted both the frequency of these infections and the paucity of data regarding their clinical, laboratory, and microbiologic manifestations.1 The scarcity of data was attributed to various factors, including lack of standardized definitions for polymicrobial infection and problems in ascertaining the pathogenic potential of various organisms in certain circumstances (eg, the role of coagulase‐negative staphylococci [CoNS] in polymicrobial bloodstream infections [BSI] in immunocompromised hosts). Polymicrobial BSIs are relatively common in hospitalized neonates, occurring in approximately 3% to 10% of cases of neonatal BSI.2‐4 Compared with neonatal monomicrobial BSI, for which clinical manifestations, risk factors, and outcomes have been well described,5‐7 informative data on polymicrobial BSI in neonates, like that in adults, are extremely limited.
In adults, polymicrobial BSI typically occurs in severely immunocompromised individuals, including those with serious underlying conditions who are hospitalized with invasive support apparatus.8‐17 Neonatal intensive care unit (NICU) patients mirror this population. The last major North American review of polymicrobial sepsis among NICU infants included 15 cases from the period 1971–1986.3 Univariate analysis revealed a significantly higher rate of mortality among infants with late‐onset (occurring at 5 days of life or later, by their definition) polymicrobial sepsis than among infants with late‐onset monomicrobial sepsis. We hypothesized that the incidence of polymicrobial BSI in our NICU was increasing, particularly among infants with significant comorbidities. In order to test this hypothesis, we performed a single‐center, matched case‐control multivariate comparison of 105 cases of neonatal polymicrobial BSI with 105 cases of monomicrobial BSI during an 18‐year period.
Methods
The Yale‐New Haven Hospital NICU is a 46‐bed, level IIID tertiary care referral center for infants with complex medical and surgical conditions. There are approximately 4,500 to 5,000 live births per year at Yale‐New Haven Hospital, with 850 to 900 neonatal admissions per year; the admitted neonates are either born there or are transported to the NICU from local hospitals.
Case Selection and Data Collection
The NICU at Yale‐New Haven Hospital has maintained a record of all bacterial and fungal‐related BSIs since 1928.18 These data exclude blood cultures which have yielded organisms not known to cause infection in neonates (ie, “contaminants” such as Corynebacterium). For the purpose of this study, all positive cultures of neonatal blood samples from the period January 1, 1989, to December 31, 2006, that yielded more than 1 causative organism and that were obtained at more than 72 hours of life (ie, representing late‐onset BSI, according to our definition), were identified. If any blood culture yielded CoNS, a neonatal commensal bacterial species responsible for a high rate of false‐positive culture results, the associated episode was reviewed and included only if modified criteria from the Centers for Disease Control and Prevention for CoNS BSI were met.18,19 The medical records of all infants with polymicrobial BSI were then reviewed, and demographic, hospital course, and outcome data were collected.
Control Population
Monomicrobial blood cultures from the Yale‐New Haven Hospital NICU from the same time period were identified, and the affected infants were designated as control patients. One control patient was selected for every case patient with polymicrobial BSI. If an infant had more than 1 episode of polymicrobial BSI, a separate control patient was chosen for each episode. Control patients were matched to case patients according to birth weight (±300 g), sex, and date of birth (±1 year). The individual performing the matching was masked regarding the specific organism responsible for BSI and regarding any underlying medical condition(s) and/or outcome data from each control neonate.
Definitions
Polymicrobial BSI was considered present if there was a record of a blood culture that yielded more than 1 causative organism. Monomicrobial BSI was considered present if there was a record of a blood culture that yielded only 1 causative organism not deemed to be a contaminant.
Potential Risk Factors for BSI
The presence of a central venous catheter, including umbilical venous and arterial catheters, percutaneous central venous catheters, Arrow catheters, and Broviac Silastic catheters, was included as a risk factor only if the catheter was placed before the onset of infection and was in place at the time the blood sample for culture was drawn. Surgery was included only if the procedure occurred less than 7 days prior to the onset of bacteremia. Prolonged intubation was defined as the use of mechanical ventilation for more than 14 days prior to infection. Other potential risk factors for BSI (ie, use of total parenteral nutrition, intralipids, H2‐blockers) had to be present at the time the culture‐positive blood sample was obtained to qualify as a risk factor. Severe underlying conditions, included as risk factors only if diagnosis preceded the onset of BSI, consisted of any major congenital anomaly (eg, congenital heart disease, gastroschisis, jejunal atresia, etc.) and/or acquired life‐threatening conditions (eg, necrotizing enterocolitis, major end organ failure, intracranial hemorrhage, and/or respiratory failure of any cause) and their associated interventions.
Comorbidities
Bronchopulmonary dysplasia was considered present if there was a need for supplemental oxygen at 36 weeks postmenstrual age in association with characteristic radiographic changes.20 Bronchopulmonary dysplasia was included in the analysis as a potential comorbidity related to BSI only if diagnosed more than 7 days after the onset and treatment of the BSI.
Necrotizing enterocolitis was defined according to the modified Bell’s staging and included only those cases at stage IIA or higher.21 Necrotizing enterocolitis was included in the analysis as a potential comorbidity related to BSI only if diagnosed less than 7 days prior to the onset of the BSI.
Outcome
Death was considered related to BSI if it occurred within 7 days after the culture‐positive blood sample was obtained or if clinical signs and symptoms of infection were documented in the medical record as the direct cause of death. The BSI‐related death rate was calculated with the numerator as the number of BSIs with related death and the denominator as the total number of BSIs, times 100.
Laboratory Testing
Blood cultures were assessed using a fluorescence‐detection system for the presence of CO2 production (Bactec II or 9240; Becton Dickinson), and species were identified using standard procedures.
Statistical Analysis
The SPSS statistical software, version 15.0 (SPSS) and GraphPad Prism, version 3.0 (GraphPad Software) were utilized for data analyses. Trends in the incidence of polymicrobial BSI were analyzed using linear regression. In the univariate analyses, continuous data were analyzed using the Student t test for comparison of mean values, and dichotomous variables were analyzed using the Fisher exact test for cells with expected counts of 5 or fewer, and the χ2 test for expected counts greater than 5. Variables with a P value of less than .10 on univariate analysis were included in a multivariate, forward conditional logistic regression analysis. If an infant had more than 1 episode of polymicrobial BSI, only the first episode was included in the multivariate analysis. Polymicrobial BSI was analyzed as both the outcome of interest (when assessing the effects of demographic and hospital course data) and as a covariate (when assessing its effect on specific neonatal outcomes, eg, bronchopulmonary dysplasia, necrotizing enterocolitis, death, BSI‐related death). Linear regression analysis was utilized when hospital length of stay was analyzed as the outcome of interest. A P value of less than .05 was considered statistically significant.
This study was approved by the human investigation committee of the Yale University School of Medicine.
Results
Polymicrobial BSI
There were a total of 1,105 BSIs in the Yale‐New Haven Hospital NICU during the study period, of which 108 (10%) were polymicrobial BSI (in 105 neonates). One hundred five of these BSIs (in 102 neonates) occurred after 72 hours of life and therefore met our inclusion criteria. The number of cases of polymicrobial BSI per 1,000 NICU admissions increased significantly during the study period (
; 
) (Figure). Each case was differentiated by classification of the responsible organisms (Table 1). Gram‐positive cocci were recovered from culture in 81 (77%) of the 105 cases of polymicrobial BSI. CoNS were the most common organisms recovered overall and were present in 47 cases (45%). The proportion of cases of polymicrobial BSI in which CoNS were recovered increased over the study period (
; 
), although this increase was not statistically significant. Species of Enterococcus were present in 46 cases (44%); gram‐negative rods were present in 57 cases (54%); Escherichia coli was the most common gram‐negative rod species recovered from culture and was present in 20 cases (19%). The number of cases of gram‐negative rod–related polymicrobial BSI did not change over the study period (
; 
). Fungi were present in 19 cases of polymicrobial BSI (18%); Candida albicans was the causative agent in 13 cases (12%).
Figure. Annual incidence rate of polymicrobial bloodstream infection in a neonatal intensive care unit 1988–2006. The increase was statistically significant (
; 
).
Monomicrobial BSI
Of the 105 cases of monomicrobial BSI among the control patients, 62 (59%) were caused by gram‐positive cocci, 38 (36%) were caused by gram‐negative rods, and 5 (5%) were caused by fungi. CoNS were responsible for 35 cases (33%), E. coli were responsible for 18 (17%), and Staphylococcus aureus were responsible for 13 (12%).
Univariate Analyses of Case and Control Patients
Neonates with polymicrobial BSI and those with monomicrobial BSI were well matched with respect to gestational age, birth weight, and sex (Table 2). In comparison with monomicrobial BSI, polymicrobial BSI had a later onset and occurred in infants who were more likely to have a severe underlying condition (Table 2). Polymicrobial BSI was more common among patients who required an indwelling central venous catheter for a more prolonged duration and was associated with a longer duration of postnatal antibiotic use prior to the onset of infection (Table 2).
No statistically significant differences were observed between case patients (with polymicrobial BSI) and control patients (with monomicrobial BSI) in the proportion subsequently diagnosed with bronchopulmonary dysplasia or necrotizing enterocolitis. Death occurred more frequently and the duration of hospital stay was longer in patients with polymicrobial BSI, compared with the monomicrobial BSI group, although these differences were not statistically significant (Table 2).
Regression Analysis
A conditional logistic regression analysis was performed utilizing polymicrobial BSI as the binary dependent variable and selected covariates identified from the univariate analysis (ie, those with 
) as the independent variables. The presence of a severe underlying condition and later onset of infection were determined to be independent risk factors for polymicrobial BSI (Table 3).
Cases of polymicrobial BSI and monomicrobial BSI were also assessed via regression analyses as independent variables for outcome data and comorbidities, including bronchopulmonary dysplasia, necrotizing enterocolitis, length of stay, death, and BSI‐related death. Polymicrobial BSI, as compared with monomicrobial BSI, did not appear to have a significant effect on any of the selected outcomes of interest.
Discussion
Data on polymicrobial BSI in neonates are extremely limited. Faix et al.3 identified 15 cases of polymicrobial BSI (3.9% of patients with BSI) in Ann Arbor, Michigan, from 1971 through 1986. A retrospective, univariate, unmatched case‐control analysis from India identified 40 cases of neonatal polymicrobial BSI during 1 year.4 Our study is, to our knowledge, the first matched case‐control, multivariate comparison of polymicrobial and monomicrobial BSI, it has the largest neonatal cohort and longest data collection period of any comparison of polymicrobial and monomicrobial BSI, and it is the first North American analysis of polymicrobial BSI performed in a NICU population in the last 20 years.3
Our hypothesis that the incidence of polymicrobial BSI among infants hospitalized in the NICU at our hospital had increased was confirmed by the data, as was the association of polymicrobial BSI in infants with the presence of severe underlying conditions. Polymicrobial BSI frequently occurs in adult and pediatric patients with serious underlying conditions.12‐16,22,23 In adults, these conditions include cancer,12 hospitalization in the intensive care unit,13 chronic hemodialysis,14 burns,15 and acquired immune deficiency syndrome.16 Reports of BSI in pediatric patients are less common, but associated underlying conditions are similar to those described in adults.22,23 Hospitalized neonates represent an intriguing subset of the pediatric population. Although cancer, burns, and renal disease requiring hemodialysis are extremely rare diagnoses in the NICU, all newborn infants have developmentally immature immune systems and increased susceptibility to invasive infections.24 Infants in the NICU are particularly at high risk of infection because of the prevalence of severe underlying conditions that often necessitate invasive intervention, the prolonged need for indwelling support apparatus, and, in extremely premature infants, severe immunological immaturity. In our cohort of infants with polymicrobial BSI, the mean gestational age was 30.7 weeks; the mean birth weight was 1,660 g; 84 (80%) had a central venous catheter in place prior to infection; and when prematurity was included in the definition, 97 (95%) of 102 had a severe underlying condition.
Although the incidence of polymicrobial BSI has increased at our NICU over the past 18 years, there has not been a concurrent increase in NICU BSI‐related mortality. Furthermore, no statistically significant difference in the rate of BSI‐related death was determined to exist between the polymicrobial BSI group and the monomicrobial BSI group. Studies of the mortality associated with polymicrobial BSI in all patient populations have been conflicting; some have found increased mortality, compared with monomicrobial BSI, some decreased mortality, and others no difference.3,13,25‐27 These different findings may be attributed to various factors, such as study design, advances in medical care, differences in the patient populations being studied, infection with less virulent organisms, and other, unidentified factors.
The study design of Faix et al.3 utilized univariate analyses to demonstrate that increased mortality is associated with late‐onset polymicrobial sepsis, compared with late‐onset monomicrobial sepsis. Our analyses utilized a forward conditional logistic regression model to analyze the same association, but we found no such association. Disparities in mortality can also be attributed to significant advances in antenatal and neonatal care. Faix et al.3 reported infection‐related death among 70% of infants with late‐onset polymicrobial sepsis and 17% of infants with monomicrobial sepsis. In contrast, BSI‐related death occurred in 16% of cases of polymicrobial BSI and 8% of cases of monomicrobial BSI in our cohort. The use of antenatal corticosteroid therapy, exogenous surfactant treatment, and novel methods of mechanical ventilation have been widely adopted since Faix et al.3 published their data. These developments have resulted in an overall improvement in the survival of NICU infants, including those with BSI.
Significant changes have also occurred in the prevalence in the NICU of certain pathogens over the past 20 years, with less‐virulent species of bacteria (ie, CoNS) having recently replaced traditional neonatal pathogens (ie, group B Streptococcus and E. coli) as the most predominant type of organisms recovered from culture.18 Faix et al.3 reported no cases of polymicrobial sepsis involving CoNS, but it was the predominant organism isolated from our cohort. These differences may also contribute to the observed discrepancies in mortality in reports of polymicrobial BSI. They also exemplify the need for ongoing surveillance in the NICU to identify changing patterns of polymicrobial BSI, particularly those that may impact diagnostic measures, therapeutic interventions, and development of strategies for prevention.
Despite its design, this study has its limitations, including the pitfalls inherent in retrospective, single‐center studies. A limitation of polymicrobial BSI studies is the potential for inclusion of false‐positive blood cultures, particularly those containing commensal species. In our study, stringent clinical and laboratory criteria were applied to each episode of BSI involving CoNS, the most common commensal species recovered from culture.18 Despite these measures, it is possible that some of our cases of polymicrobial BSI were monomicrobial, with CoNS present as a contaminant. This could falsely inflate our incidence of polymicrobial BSI and interfere with data interpretation. Alternatively, exclusion of polymicrobial cultures that yielded commensal species and were deemed to be false‐positive findings, without proper identification of the organism and antibiotic susceptibility testing, is equally problematic and may have been a factor in previous reports of polymicrobial infection in immunocompromised cancer patients.1 Efforts to establish a consensus definition for polymicrobial BSI are needed to eliminate these discrepancies in data reporting and interpretation, particularly in the neonatal population, to whom standard definitions developed for adult patients are often difficult to apply.
Polymicrobial BSI poses a significant threat to the already vulnerable hospitalized neonate. As advances in medical therapies continue to improve survival in this patient population, polymicrobial BSI is likely to become a more prevalent problem. Efforts aimed at understanding and updating the epidemiology of polymicrobial BSI in all at‐risk populations, including the hospitalized infant, should continue in order to facilitate diagnosis, treatment, and prevention.
Acknowledgments
Potential conflicts of interest. All authors report no conflicts of interests relevant to this study.
References
- 1. Rolston K VI, Bodey GP, Safdar A. Polymicrobial infection in patients with cancer: an underappreciated and underreported entity. Clin Infect Dis 2007; 45:228‐233.
- 2. Jarvis WR, Highsmith AK, Allen JR, Haley RW. Polymicrobial bacteremia associated with lipid emulsion in a neonatal intensive care unit. Pediatr Infect Dis 1983; 2:203‐208.
- 3. Faix RG, Kovarik SM. Polymicrobial sepsis among intensive care nursery infants. J Perinatol 1989; 9:131‐136.
- 4. Gupta P, Kumhar GD, Kaur G, Ramachandran VG. Clinical significance of polymicrobial bacteremia in newborns. J Paediatr Child Health 2005; 41:365‐368.
- 5. Stoll BJ, Hansen N, Fanaroff AA, et al. Late‐onset sepsis in very low birth weight neonates: the experience of the NICHD Neonatal Research Network. Pediatrics 2002; 110:285‐291.
- 6. Christie C, Hammond J, Reising S, Evans‐Patterson J. Clinical and molecular epidemiology of enterococcal bacteremia in a pediatric teaching hospital. J Pediatr 1994; 125:392‐399.
- 7. Fairchild KD, Tomkoria S, Sharp EC, Mena FV. Neonatal Candida glabrata sepsis: clinical and laboratory features compared with other Candida species. Pediatr Infect Dis J 2002; 21:39‐43.
- 8. Kiani D, Quinn EL, Burch KH, Madhavan T, Saravolatz LD, Neblett TR. The increasing importance of polymicrobial sepsis. JAMA 1979; 242:1044‐1047.
- 9. Weinstein MP, Reller LB, Murphy JR. Clinical importance of polymicrobial bacteremia. Diagn Microbiol Infect Dis 1986; 5:185‐196.
- 10. Reuben AG, Musher DM, Hamill RJ, Broucke I. Polymicrobial bacteremia: clinical and microbiologic patterns. Rev Infect Dis 1989; 11:161‐183.
- 11. Pittet D, Li N, Wenzel RP. Association of secondary and polymicrobial nosocomial bloodstream infections with higher mortality. Eur J Clin Microbiol Infect Dis 1993; 12:813‐819.
- 12. Huang WT, Chen TY, Su WC, Yen CJ, Tsao CJ. Implantable venous port‐related infections in cancer patients. Support Care Cancer 2004; 12:197‐201.
- 13. Rello J, Quintana E, Mirelis B, Gurgui M, Net A, Prats G. Polymicrobial bacteremia in critically ill patients. Intensive Care Med 1993; 19:22‐25.
- 14. Shmuely H, Pitlik S, Yahav J, Samra Z, Leibovici L. Seven‐year study of bacteremia in hospitalized patients on chronic hemodialysis in a single tertiary hospital. Ren Fail 2003; 25:579‐588.
- 15. Bang RL, Sharma PN, Sanyal SC, Bang S, Ebrahim MK. Burn septicaemia in Kuwait: associated demographic and clinical factors. Med Princ Pract 2004; 13:136‐141.
- 16. Lortholary O, Jehl F, Petitjean O, Cohen P, Tarral E, Guillevin L. Polymicrobial pyomyositis and bacteremia in a patient with AIDS. Clin Infect Dis 1994; 19:552‐553.
- 17. Price CS, Hacek D, Noskin GA, Peterson LR. An outbreak of bloodstream infections in an outpatient hemodialysis center. Infect Control Hosp Epidemiol 2002; 23:713‐715.
- 18. Bizzarro MJ, Raskind C, Baltimore RS, et al. Seventy‐five years of neonatal sepsis at Yale: 1928–2003. Pediatrics 2005; 116:595‐602.
- 19. Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. CDC definitions for nosocomial infections, 1988. Am J Infect Control 1988; 16:128‐140.
- 20. Shennan AT, Dunn MS, Ohlsson A, Lennox K, Hoskins EM. Abnormal pulmonary outcomes in premature infants: prediction from oxygen requirement in the neonatal period. Pediatrics 1988; 82:527‐532.
- 21. Kliegman RM, Walsh MC. Neonatal necrotizing enterocolitis: pathogenesis, classification, and spectrum of illness. Curr Probl Pediatr 1987; 17:213‐288.
- 22. Bonadio WA. Polymicrobial bacteremia in children: an 11‐year experience. Am J Dis Child 1988; 142:1158‐1160.
- 23. Trad O, Jumaa PA, Afify Z. The changing pattern of bloodstream infection in pediatric oncology patients in the United Arab Emirates. Pediatr Hematol Oncol 2003; 20:281‐289.
- 24. Clapp DW. Developmental regulation of the immune system. Semin Perinatol 2006; 30:69‐72.
- 25. McKenzie FE. Case mortality in polymicrobial bloodstream infections. J Clin Epidemiol 2006; 59:760‐761.
- 26. Vales EC, Abraira V, Sanchez JC, et al. A predictive model for mortality of bloodstream infections: bedside analysis with the Weibull function. J Clin Epidemiol 2002; 55:563‐572.
- 27. Wisplinghoff H, Bischoff T, Tallent SM, Seifert H, Wenzel RP, Edmond MB. Nosocomial bloodstream infections in U.S. hospitals: analysis of 24,179 cases from a prospective nationwide surveillance study. Clin Infect Dis 2004; 39:309‐317.