Pathogen-Specific Clustering of Nosocomial Blood Stream Infections in Very Preterm Infants
BACKGROUND AND OBJECTIVES: Nosocomial infections in NICUs tend to cluster, sometimes as devastating outbreaks, but pathogen-specific transmission probabilities are unknown. We aimed to quantify the pathogen-specific risk of a blood stream infection (BSI) in preterm infants after an index case with that pathogen in the same department.
METHODS: Data of 44 818 infants below 1500 g birth weight of the German NICU surveillance system (2000–2011) were used to calculate the probability of a BSI in the presence or absence of another infant in the same unit with a same-pathogen BSI.
RESULTS: The relative risk was similar for the more common pathogens, Enterococcus spp (4.3; 95% confidence interval: 2.7–6.9; n = 243), Enterobacter spp (7.9, 5.4–11.4; n = 246), Escherichia coli (7.9; 5.1–12.1; n = 210), Candida albicans (8.7; 5.0–15.4; n = 138), Staphylococcus aureus (9.5; 7.6–12.1; n = 407) and Klebsiella spp (13.1; 9.0–19.1; n = 190) but markedly elevated for Serratia spp (77.5; 41.1–146.1; n = 58) and Pseudomonas aeruginosa (64.5; 25.7–162.1; n = 38). Rates of BSI per 100 exposed infants ranged between 2.21 (Enterococcus) and 8.15 (Serratia). The same pattern emerged after adjustments were made for patients’ characteristics or when the analysis was restricted to positive blood cultures during the preceding 30 days.
CONCLUSIONS: Although BSIs with P aeruginosa or Serratia spp in preterm infants are rare, they are associated with a markedly elevated risk of secondary same-pathogen BSI and should prompt intensified active surveillance and infection control measures.
- BSI —
- blood stream infection
- CI —
- confidence interval
- CoNS —
- coagulase-negative staphylococci
- GEE —
- generalized estimating equation
- OR —
- odds ratio
- RR —
- relative risk
What’s Known on This Subject:
Nosocomial infections in neonatal intensive care tend to run in clusters. Calculating pathogen-specific transmission probabilities requires large databases.
What This Study Adds:
The relative risk for a same-pathogen blood stream infection after an index case in the same unit is markedly elevated for Pseudomonas aeruginosa and Serratia spp, as compared with other pathogens.
In very low birth weight preterm infants, nosocomial blood stream infections (BSIs) are associated with substantial morbidity and mortality. BSI with the most commonly isolated organisms in BSI, coagulase-negative staphylococci (CoNS) display a rather benign clinical course.1,2 CoNS are part of the normal skin flora, and rates of BSI with CoNS have been found to decline after implementation of improved handling of central venous lines.2,3 In contrast, BSIs with Staphylococcus aureus, gram-negative rods, or fungi have a high mortality risk,4 with Pseudomonas aeruginosa and Serratia marcescens figuring prominently.2,3,5,6 Horizontal transmissions of these pathogens in NICUs may cause difficult-to-control outbreaks.
BSIs and outbreaks with non-CoNS pathogens, while potentially devastating for the affected infants or units, respectively,7–11 still occur mostly in a sporadic fashion. Thus, there is a need for a sufficiently large database to calculate and compare the BSI clustering probability for specific pathogens. We used the data from the German national neonatal infections surveillance system1,10,12,13 to estimate the probability of a hospitalized very low birth weight (<1500 g) infant to develop a BSI with a particular pathogen when another infant previously diagnosed with a BSI of the same pathogen was being cared for in the same unit. We also calculated the probability of contracting a BSI with a particular pathogen after an index case during the preceding 30 days in the same department.
Methods and definitions employed by the German national neonatal infection surveillance system have been described previously.1,10,12,13 It has become the preferred surveillance system for neonatal nosocomial infections after participation in an infection surveillance system was made mandatory for all NICUs in Germany caring for infants below 1500 g birth weight. All analyses are done using anonymized data within the legal framework of quality improvement, obviating the need for ethical approval and informed consent.
The surveillance records pneumonia, necrotizing enterocolitis, and BSI beyond 72 hours of age in all preterm infants with a birth weight <1500 g until discharge, until death, or until they reach 1800 g. Trained hospital nurses and physicians collect demographic and obstetric data of all infants surveyed, and an array of clinical data of infected neonates.12 Non-CoNS BSIs are being defined as any case with a positive blood and/or cerebrospinal fluid culture with a recognized pathogen, unrelated to an infection at another site, in the presence of at least 2 clinical signs of systemic infection (increased [>38°C] or decreased [<36.5°C] body temperature or temperature instability, tachycardia, increased rates of apnea/bradycardia episodes, prolonged capillary refill, metabolic acidosis, hyperglycemia, apathy, or seizures). Pathogens isolated from cerebrospinal fluid are included in the BSI definition as most cases of meningitis in preterm infants result from hematogenous dissemination, and microbiological findings in cerebrospinal fluid may identify pathogens missed when culturing inadequate amounts of blood.
For the purpose of this analysis, exposure to a particular pathogen was defined in 2 ways:
An infant was considered exposed to a pathogen if a BSI with that particular species had caused a BSI in another infant (index case) cared for simultaneously in the same department (date of admission of the exposed infant preceding the date of discharge of the index case, date of BSI of the index case preceding the date of discharge of the exposed infant, and date of BSI of the index case preceding the date of BSI of the exposed infant).
An infant was considered exposed to a pathogen if a BSI with that particular species had been diagnosed in an index case cared for in the same department during the length of stay of the exposed infant or up to 30 days before admission of the exposed infant, with the date of BSI of the index case preceding the date of BSI of the exposed infant.
The analysis was restricted to pathogens implicated in >30 BSIs during the study period. As CoNS are part of the normal skin flora of virtually every infant and often considered contaminations when grown from blood cultures, no calculations were carried out for CoNS.
For the various pathogens, adjusted odds ratios (ORs) were calculated by using multivariable logistic regression models. The patient-based parameters birth weight (<500/500–749/750–999/1000–1249/1250–1499 g), gestational age (≤26/27–28/29–30/≥31 weeks), gender (boy/girl), delivery mode (planned cesarean, vaginal delivery, emergency cesarean), multiple birth (no/yes), and length of stay were considered in all models as confounders. Missing values were considered as a separate category. Furthermore, generalized estimating equations (GEEs) were calculated to consider clustering effects within a department by exchangeable correlation structure. In the GEE models, only birth weight and gender were considered in all models because the limited number of events restricted adjusting for more confounders.
P values < .05 were considered significant. Calculations were carried out using SPSS 22.0 (IBM SPSS Statistics, IBM Corporation) and SAS 9.3 (SAS Institute, Inc, Cary, NC).
Records of 44 818 infants with a birth weight <1500 g born between January 1, 2000, and December 31, 2011, and being cared for in 229 hospitals across Germany revealed a total of 2004 non-CoNS culture-positive BSIs. Pathogens identified (in descending order) were methicillin-sensitive S aureus (n = 407), Enterobacter spp (n = 246), vancomycin-sensitive Enterococcus spp (n = 243), cefotaxime-sensitive Escherichia coli (n = 210), cefotaxime-sensitive Klebsiella spp (n = 190), Candida albicans (n = 138), Serratia spp (n = 58), and P aeruginosa (n = 38). Pathogens with 30 or less BSIs not analyzed further included Citrobacter spp (n = 17), Acinetobacter baumannii (n = 28), and Group B Streptococcus (n = 30).
Numbers of BSI acquired while another infant with a same-pathogen BSI was being cared for in the unit are shown in Table 1. Rates of BSI per 100 exposed infants varied between 2.2 (Enterococcus spp) and 8.2 (Serratia spp). The relative risk (RR) for acquiring a BSI in the presence or absence of an infant with a preceding same-pathogen BSI varied between 4.3 (Enterococcus) and 77.5 (Serratia). In logistic regression analysis, adjusted ORs varied between 2.4 (Enterococcus) and 63.3 (Serratia). Adjusted OR did not differ substantially with the results calculated by using GEE models (Table 1).
We also looked at numbers of BSI when a same-pathogen BSI was observed in the same unit during the preceding 30 days (Table 2). By this definition, rates of BSI per 100 exposed infants varied between 1.4 (C albicans) and 6.5 (Serratia spp). The RR for acquiring a BSI while a same pathogen-BSI had been diagnosed in the preceding 30 days in the same department varied between 2.3 (Enterococcus) and 59.5 (Serratia). Adjusted ORs using logistic regression analysis or GEE are shown in Table 2.
This analysis of 2004 BSIs in 44 818 very low birth weight infants demonstrates that the probability of an infant developing a BSI with a particular pathogen increases while another infant cared for in the same department has a BSI with the same pathogen. Rates of pathogen-specific BSI in exposed infants ranged from 0.022 for Enterococcus spp to 0.082 for Serratia spp. The RR of contracting a BSI in exposed versus nonexposed infants reveals even more variation. Judged by overlapping 95% confidence intervals (CIs), a pattern with 2 groups emerges. The first 1 contains C albicans, Enterococcus, S aureus, Enterobacter, E coli, and Klebsiella, whereas the second group comprises Serratia spp and P aeruginosa. The same pattern emerged when calculating the probabilities of contracting a BSI with a particular pathogen in a neonatal department within 30 days after that pathogen had been isolated from a blood culture in another infant cared for in that department.
There are several limitations to this study. First, the reporting system made no distinction between various strains of Enterococcus, Enterobacter, or Serratia, and there was no genotyping of the pathogens involved. Thus, 2 temporally related BSIs in the same department might well be a coincidence. Even in outbreaks, genotyping has revealed unexpected polyclonality of the pathogens isolated.14–19 The likelihood of this type of error increases alongside the incidence density, overestimating the transmission probabilities of more common pathogens such as S aureus, and rendering this kind of analysis virtually pointless for CoNS, as virtually all preterm infants become colonized with some CoNS during the first week of life.20 Retrieval of CoNS from the blood culture of a sick infant indicates that the colonization of the infant’s skin with CoNS has turned into an invasive disease, usually triggered by indwelling catheters, but it does not suggest spread of CoNS from 1 infant to another infant.20 Rates of BSI by CoNS but not by gram-negative pathogens have been found to decrease following implementation of improved handling of indwelling central lines.2,3 Second, while looking for in-hospital transmissions, we had no data about the actual physical proximity of infants contracting a BSI with the same pathogen. Infants being cared for in the same department may or may not have shared the same room, the same nurse, or the same ward. Third, the analysis was restricted to very low birth weight infants, but NICUs are being populated also by more mature infants who are not subjects to this type of infection surveillance. Fourth, we only looked at culture-positive BSI, whereas pathogen transmissions may involve chains of colonized or infected infants without clinical symptoms of systemic infection.21 Focusing on BSI does not take into account merely colonized infants as a source of exposure. However, when we looked at BSI in the presence or absence of a BSI with the same pathogen in the same department during the preceding 30 days, a highly similar pattern emerged. Fifth, while the national neonatal infection surveillance system covers the vast majority of neonatal intensive units in Germany, the data are all from the same country. Therefore, the analysis presented here calls for similar studies in elsewhere. Although our results are thought to be applicable to hospitals with comparable resources and hygiene measures, pathogen-specific transmission rates might be influenced by different hygiene standards, use of antibiotics, and local prevalence of pathogens.22
The differences between the pathogens analyzed for involvement in temporally and spatially linked BSI may reflect differences of virulence, the ease of spreading in the NICU environment, or both. Reported ratios of sterile site infections to colonizations are 1:6 for S marcescens, as opposed to 1:27 for Klebsiella pneumonia.23 Calculated attack rates, as defined as the ratio of infants with invasive infections and all infants positive for a certain pathogen, have been reported to be over 20% for P aeruginosa24 and may exceed 50% for S marcescens during outbreaks.7 S marcescens and P aeruginosa have also been found to figure prominently in NICU-acquired bacterial conjunctivitis,25 emphasizing their potential for patient-to-patient spreading in this environment.
In addition to transmission via the hands of staff,24,26–29 pathogens may also spread from inanimate point sources. An increased role for environmental sources has indeed been documented for S marcescens and P aeruginosa, as compared with other pathogens. Numerous reports have identified soaps and shampoos as reservoirs for S marcescens30–33 and P aeruginosa.34,35 Water systems can act as a source of S marcescens36 and P aeruginosa.19,35,37,38
Although we found P aeruginosa and Serratia spp to have the highest rate of BSI in exposed infants of all pathogens analyzed, they were also those with the lowest incidence density. The rate of BSI in exposed infants, defined by a BSI in another infant, does not reflect the general contagiousness of a pathogen but the risk for a cluster of BSI after an index case with the specific pathogen. In contrast to other potential pathogens, P aeruginosa and Serratia spp are usually not present in hospitalized preterm infants, and the advent of a colonized infant in the NICU is pivotal as a source for further infections. NICUs should be alerted in any case of a BSI with P aeruginosa or Serratia, prompting active surveillance cultures to detect noninfected carriers and place them separate from other infants.39
P aeruginosa and Serratia spp are exceptional in their potential to spread in the NICU and attack very low birth weight infants. Because they are also those pathogens with the highest reported BSI-related mortality rates,4–6 vigorous attempts should be made to intensify infection control measures whenever P aeruginosa or Serratia spp have been isolated from a patient in the NICU.
The authors thank the health care and infection control staff for their continued support of the neonatal nosocomial infection surveillance project.
- Accepted January 5, 2016.
- Address correspondence to Christoph Bührer, MD, Klinik für Neonatologie, Charité Universitätsmedizin Berlin, D-13344 Berlin, Germany. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: The German National Neonatal Infection Surveillance System is being supported by the Federal Department of Health and intramural funds from the Charité University Medical Center.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Bizzarro MJ,
- Sabo B,
- Noonan M,
- Bonfiglio MP,
- Northrup V,
- Diefenbach K; Central Venous Catheter Initiative Committee
- Stoll BJ,
- Hansen NI,
- Higgins RD, et al; National Institute of Child Health and Human Development
- Bizzarro MJ,
- Dembry LM,
- Baltimore RS,
- Gallagher PG
- Quick J,
- Cumley N,
- Wearn CM, et al
- Almuneef MA,
- Baltimore RS,
- Farrel PA,
- Reagan-Cirincione P,
- Dembry LM
- Dias C,
- Gonçalves M,
- João A
- Copyright © 2016 by the American Academy of Pediatrics