Long-term Evolution of Multiple Outbreaks of Serratia marcescens Bacteremia in a Neonatal Intensive Care Unit

Abstract

The annual incidence of Serratia marcescens bacteremia in a neonatal intensive care unit increased significantly between 2002 and 2010. Molecular epidemiology studies revealed that 8 clones were responsible for 85.2% of cases. Given that these infections are potentially preventable, even the appearance of 1 case of bacteremia should be an indicator for outbreak management.

Serratia marcescens is an important nosocomial pathogen, especially in neonatal intensive care units (NICUs) where outbreaks have been frequently reported causing serious infections, including pneumonia, meningitis and bacteremia.^1–5 Different sources have been implicated in outbreaks, including contaminated liquid nonmedicated soap, hexetidine solution, pressure transducers, air conditioner duct and contaminated milk.⁶

During 2010, we observed an increase in cases of S. marcescens bacteremia in our NICU. These cases were scattered throughout the year, and all the bacterial isolates had a typical antimicrobial susceptibility pattern. We decided to study the epidemiology and clonality of S. marcescens isolates recovered from blood samples in neonates during a 9-year period and to investigate whether there were S. marcescens clones that could chronically persist in the NICU causing infections.

MATERIALS AND METHODS

The NICU at the Hospital 12 de Octubre is a 41-bed referral unit with 19 intensive care beds in a single large space and 22 intermediate intensive care beds in 3 separate rooms. In 2007, the NICU was refurbished and the 19 intensive care beds were distributed in 2 separate rooms. For hand hygiene, the NICU staff used an alcohol-based hand sanitizer. We studied retrospectively all neonates with S. marcescens bacteremia during the period from January 2002 to December 2010. Medical records were reviewed to assess the associated risk factors.

Blood samples were inoculated and processed in an automated system (BacT/ALERT, BioMérieux, Durham, NC). Identification and susceptibility testing were performed using conventional methods and the Wider System (Francisco Soria Melguizo, Madrid, Spain). Breakpoints for susceptibility were applied according to Clinical Laboratory Standard Institute.⁷ Molecular characterization of isolates was performed by pulsed-field gel electrophoresis (PFGE) with SpeI.^8,9 PFGE patterns were analyzed with BioNumerics software (Applied Maths, Kortrijk, Belgium).

The annual incidence of S. marcescens bacteremia was calculated as the number of clinical episodes per 10,000 bed-days. The time-related changes in the incidence of bacteremia were estimated by Poisson regression.

RESULTS

During the study period, 54 episodes of S. marcescens bacteremia were identified in 50 neonates. The annual incidence estimate of S. marcescens bacteremia increased from 0.8 in 2002 to 11.4 cases per 10,000 bed-days in 2010 (P for trend <0.001) (Fig. 1A). The proportion of cases was 3.74 (95% confidence interval, 1.86–7.36) higher in the period 2006–2010 than before 2006. Of the total cases, 51 (94.4%) were considered as late-onset bacteremia (>3 days) and 3 (5.6%) as early-onset bacteremia (≤3 days). In addition, 4 neonates had recurrent bacteremia. The mean gestational age was 29.8 weeks (SD, 4.95), the mean age at onset of bacteremia was 24.5 days (SD, 19.14) and 70% of neonates had low birth weights (<1500 g). Among the factors commonly predisposing to bacteremia, we found 91.8% (45/49) neonates with central intravenous catheters, 47.9% (23/48) with prior surgery and 54.1% (26/48) with mechanical ventilation. All the isolates were susceptible to piperacillin/tazobactam, cefotaxime, cefepime, aztreonam, imipenem, meropenem, ertapenem, gentamicin, amikacin and ciprofloxacin. Forty-two neonates were evaluable for antibiotic therapy: 24 received combination therapy (21 meropenem plus an aminoglycoside) and 18 had monotherapy (11 aminoglycoside, 6 meropenem and 1 cefotaxime). The overall mortality was 21.3%, and the attributable mortality was 12.7%.

FIGURE 1: Evolution of S. marcescens bacteremia in neonates. A, Annual incidence. B, Temporal distribution of clusters. The symbol represents a case of bacteremia, and the length of the lines corresponds to the months of persistence of each clone.

Molecular typing of the 54 clinical isolates by PFGE revealed the presence of 16 patterns, with 8 patterns grouping 46 (85.2%) cases (Fig. 1B) and 8 being considered unique PFGE patterns. The temporal distribution of clusters was presented in the Figure 1B and showed that 4 clones (A, C, E and F) persisted during 10, 8, 15 and 9 months, respectively, causing 7, 6, 9 and 12 cases of bacteremia, respectively.

Four neonates had recurrent bacteremia. Three cases had 2 episodes, and the initial and recurrent isolates were indistinguishable by PFGE. One neonate had 2 episodes caused by isolates with different PFGE patterns. The 3 cases with identical PFGE patterns had isolates with identical antimicrobial susceptibility patterns, and the recurrent episode was interpreted as a result of re-exposure.

DISCUSSION

The purpose of the present molecular epidemiologic study of S. marcescens isolates from blood was to gain insight into the clonal structure of S. marcescens and to determine whether particular clones of these organisms were spreading within our NICU. Although we recognize that these isolates represent the “tip of the iceberg” of the total S. marcescens population, we believe that molecular epidemiologic studies of these pathogenic organisms can provide a clearer view of the overall clonal diversity within this species. A previous study in our NICU focusing on trends of bacteremia demonstrated a decrease in the global incidence; however, an increase in late-onset bacteremia caused by Gram-negative microorganisms was observed, with S. marcescens being the most frequent Gram-negative microorganism (9.4%) (15th Spanish Congress of Clinical Microbiology and Infectious Diseases, abstract 127). This percentage is higher than the 2.2% published in very low birth weight neonates in United States.¹⁰

Our study showed that the rate of S. marcescens bloodstream infection in the NICU increased significantly during the study period. The clinical profile of the neonates was similar to other studies: low birth weight, length of stay and a history of multiple invasive procedures.^1,2,4 After PFGE analysis of all blood isolates, we determined that of the 54 episodes of bacteremia, 46 (85.2%) were caused by isolates that had identical PFGE patterns. It is important to highlight that some clones were able to persist >1 year causing bacteremia. These results suggest that most cases of S. marcescens bacteremia might be attributed to the concatenation of several outbreaks caused by genetically unrelated clones. This result has important clinical implications as most cases could be attributed to horizontal transmission in the NICU.

We do not know the source of the strains causing theses outbreaks. Neonate colonization, environment contamination and hand transmission are all recognized contributors to bacterial endemic spread.^3,5 On the other hand, S. marcescens can cause long-standing outbreaks that can remain unnoticed.³ Our current infection control program in the NICU, besides active surveillance of multidrug-resistant microorganisms, includes active surveillance of S. marcescens colonization, since 2007.

Understanding the epidemiology and clonality of S. marcescens infections has implications for future efforts to control the emergence and spread not only of multidrug-resistant strains but also of susceptible microorganisms. Given that these infections are potentially preventable, even the appearance of one case of bacteremia could be an indicator for outbreak management.