Management and investigation of a Serratia marcescens outbreak in a neonatal unit in Switzerland – the role of hand hygiene and whole genome sequencing

Serratia marcescens has long been recognized as an important pathogen in neonatal intensive care units (NICUs). It is the third most common pathogen identified in published NICU outbreaks [1], and it has been found to account for 15% of all culture-positive nosocomial infections in this setting [2]. The large number of outbreak reports underestimates the true occurrence of S. marcescens in neonatology units and NICUs. According to the results of the mandatory surveillance of healthcare-associated infections (HAIs) in very low birth weight infants in Germany from 2006 to 2011, at least one to two Serratia outbreaks per year are expected for Germany alone [3].

S. marcescens causes a wide range of clinical manifestations in neonates, from asymptomatic colonization to infections such as urinary tract infections, pneumonia, sepsis or meningitis [4, 5]. Risk factors for Serratia spp. acquisition by neonates are related to immaturity, prolonged hospital stay, antibiotic use, and mechanical ventilation [4, 6].

The objective of this outbreak report was to summarize the investigation and successful management of a S. marcescens outbreak in neonates and to investigate the contribution of using whole genome sequencing. This report follows the ORION (Outbreak Reports and Intervention Studies Of Nosocomial infection) statement [7].

A total of 232 neonatal screenings (117 stool samples, 115 nasal swabs) [9] were performed. In addition to the 4 index cases and the 8 cases identified in the first prevalence survey, an additional three, four and one cases were identified in the second, third and fourth prevalence surveys in April and May, respectively (Fig. 2). The proportions of positive rectal and nasal swabs were 11.1% (13/117) and 6.1% (7/115), respectively. Five infants had a positive result for both rectal and nasal swabs. All cases were preterm-, and seven were very-low-birth weight infants. No new S. marcescens cases were identified until October 2016, including two further prevalence surveys, which were performed in June and October. There was no invasive infection due to S. marcescens, neither during the outbreak nor until end of December 2013. A total of 231 environmental samplings were performed (Table 1). All tested products, materials and surfaces were negative for S. marcescens. Hand hygiene compliance improved from 51% in April 2013 to 79% in May 2013 following the promotion programme. Compliance remained above 65% in the following months (Fig. 1).

Fingerprinting of 24 isolates from 16 neonates showed identical strains (Fig. 3). Whole genome sequencing was carried out on the set of 24 S. marcescens strains (Sm01-Sm24) belonging to the outbreak. The sequencing on an Illumina HiSeq produced on average a total of 13 million 150 bp reads per sample, exhibiting very high theoretical coverage values (between 300 and 900 fold). Assembly was achieved using SPAdes 3.9.0 software, after read quality trimming and filtering was applied to the raw reads. The assembly produced very similar results for each strain from the outbreak, resulting in an average genome size of 5′080’124 bp with a standard deviation of 1097 bp. This shows extreme similarity between the strains of the outbreak. Moreover, pairwise comparisons using MUMmer software were performed between each couple of strains and showed similarity above 99.999% [10]. Figure 4 shows an unrooted tree displaying all the isolates from the outbreak (Sm01-24), and one representative strain isolated in Germany during a SM outbreak in 2016 (SMB2099) [11].

The source of S. marcescens outbreaks in NICUs is rarely identified. A review summarizing 48 NICU outbreaks reported an identified source in only 40% (19/48), the most frequent being a colonized or infected index patient (8/48), followed by equipment for patient care (6/48), an (unspecified) environment (4/48), and food (1/48) [12]. Colonized or infected patients are thought to represent the most important reservoir of S. marcescens outbreaks in NICUs [4, 13]. In the currently described situation, the investigation revealed two additional NICU cases prior to the outbreak. A mother of one of them hadan amnion infection syndrome due to S. marcescens three months before the start of the outbreak, and thus, may have been the most likely source [12]. In the literature, healthcare workers were never identified as a source, but few studies investigated this specifically [14]. Like in our study, extensive environmental screening rarely yielded positive results in the reported outbreaks, suggesting that environmental contamination may play a limited role.

The clinical implication of colonization with S. marcescens in neonates and its implication in outbreaks is not always clear. A number of studies reported that S. marcescens is rather commonly isolated in stool of preterm neonates within the first weeks of life [6, 15, 16]. However, the pathogen was not isolated consistently in preterm neonates either [17], and S. marcescens has not been isolated in healthy term babies [18]. Thus, while the finding of S. marcescens in neonatal stool can be considered a regular event in NICUs, the pathogen per se is not part of the normal early gut flora of healthy newborns in the community. Once the intestines of a neonate are colonized with S. marcescens from the (pathologic) outside, they can become the “source” of an outbreak. This makes control of Serratia outbreaks very difficult, and underlines the importance of hands in the transmission of the pathogen and the possible successful management of an outbreak.

Already in 1981, contaminated hand-washing brushes were reported to be implicated in a S. marcescens outbreak, indirectly pointing to the pivotal role of hand hygiene in the spread of the pathogen [19]. Contaminated soaps were identified in another report [20]. Interventions to control outbreaks included always both hand hygiene and environmental control [9, 21‐28]. Given that all environmental samples were negative in our outbreak investigation and work had been done in the past to dedicate equipment, drugs, ointments and other cosmetic products to individual neonates (avoiding multivials) [29], our intervention focused on hand hygiene improvement. This was all the more justified because a decrease of hand hygiene compliance had been identified as part of the investigation; hand hygiene compliance was 72% in the year before the outbreak.

Whole genome sequencing technology emerged as a powerful tool to identify clonality among outbreak isolates. The genome of a S. marcescens strain that sparked an outbreak among infants in Germany was reported very recently [11, 12]. Comparison with our index case revealed a moderate difference around 200 SNPs. Other isolates identified from our institution and not related to the outbreak were sequenced and the number of SNPs was higher than 100′000 (data not shown).

There are limitations in our outbreak investigation. First, environmental sampling was large but we did not consistently use neutralizing agents for testing the different cosmetic products, resulting in potential underreporting. However, cosmetic products were strictly confined to the neonates and no multivials were in use. A positive finding of a product with a case would not have been prove of the product being the source. Second, we did not produce formal records from the audits other than hand hygiene.