Serratia Infection Epidemiology Among Very Preterm Infants in the Neonatal Intensive Care Unit

Abstract

Background:

Serratia spp. are opportunistic, multi-drug resistant, gram-negative pathogens, previously described among preterm infants in case reports or outbreaks of infection. We describe Serratia late-onset infection (LOI) in very preterm infants in a large, contemporary, nationally-representative cohort.

Methods:

In this secondary analysis of prospectively-collected data of preterm infants born 401–1500 grams and/or 22–29 weeks’ GA from 2018–2020 at 774 Vermont Oxford Network members, LOI was defined as culture-confirmed blood and/or cerebrospinal fluid infection >3 days after birth. The primary outcome was incidence of Serratia LOI. Secondary outcomes compared rates of survival and discharge morbidities between infants with Serratia and non-Serratia LOI.

Results:

Among 119,565 infants, LOI occurred in 10,687 (8.9%). Serratia was isolated in 279 cases (2.6% of all LOI; 2.3 Serratia infections per 1000 infants). Of 774 hospitals, 161 (21%) reported at least one Serratia LOI; 170/271 (63%) of cases occurred at hospitals reporting 1 or 2 Serratia infections and 53/271 (20%) occurred at hospitals reporting ≥5 Serratia infections. Serratia LOI was associated with a lower rate of survival to discharge compared to those with non-Serratia LOI (adjusted relative risk 0.88, 95% CI 0.82, 0.95). Among survivors, infants with Serratia LOI had higher rates of tracheostomy, gastrostomy, and home oxygen use compared to those with non-Serratia LOI.

Conclusions:

The incidence of Serratia LOI was 2.3 infections per 1000 very preterm infants in this cohort. Lower survival and significant morbidity among Serratia LOI survivors highlight the need for recognition and targeted prevention strategies for this opportunistic nosocomial infection.

INTRODUCTION

Serratia spp. are opportunistic gram-negative bacteria and an important cause of nosocomial infection in neonatal intensive care units (NICUs). Serratia infection has been estimated to account for 1–2% of bloodstream infections (BSIs) among NICU patients in the United States (US) and Germany(1–3), and up to 15% of BSIs among a group of NICUs in Europe(4). BSIs are the most common manifestation of neonatal Serratia disease, but meningitis, urinary tract infections (UTIs), respiratory tract infections, and conjunctivitis due to Serratia also occur(5). Although most reports of neonatal Serratia infection arise from single centers and describe infectious clusters or outbreaks in small cohorts(6,7,16,17,8–15), endemic patterns of sporadic infections are also described(18). Environmental reservoirs are heavily implicated in Serratia transmission, with outbreaks in NICUs traced to medical equipment, water sources, as well as colonized patients and hospital personnel(19).

Risk factors for neonatal Serratia infection are common to most forms of hospital-acquired infection among preterm infants, including low birth weight, mechanical ventilation, and presence of central catheters(2,6,12,18,19). Although Serratia infections can be mild in some hosts, neonatal Serratia BSIs have been associated with increased duration of hospitalization, higher healthcare costs, and higher all-cause and Serratia-attributable mortality, with an 18% case-fatality rate in one cohort of very-low birth weight infants(2,3). Ventriculitis and multiple abscess formation complicating Serratia meningitis can both result in devastating neurologic damage(20). There are limited available surveillance data to inform current patient- and center-level burdens of Serratia infections and related outcomes among very preterm infants in the United States. Therefore, the objectives of this study were to describe the epidemiology of late-onset Serratia infections among very preterm infants, and to evaluate the relationship between Serratia infections, survival, and discharge morbidities, within a large, contemporary, nationally-representative cohort.

METHODS

Data Source and Study Population:

Vermont Oxford Network (VON) is a nonprofit, voluntary worldwide community of practice dedicated to improving the quality, safety, and value of newborn care through a coordinated program of data-driven quality improvement, education, and research. VON maintains a prospective database describing NICU care of infants born <1500 grams. We utilized that database to perform a secondary cohort study of infants admitted to 774 participating VON centers from January 1, 2018, to December 31, 2020. The study population included infants born with birth weight 401–1500 grams and/or gestational age of 22–29 weeks. Infants who died or were transferred from the participating center by day 3 after birth were excluded. For study infants, data were collected from birth until hospital discharge, death, or first birthday (whichever came first); infants transferred after day 3 were tracked to determine their ultimate disposition and length of stay. The Institutional Review Board at the University of Vermont determined that use of the VON database for this analysis was not human subjects research.

Study Definitions

Late-onset infection (LOI):

LOI was defined as isolation of a prespecified bacterial or fungal pathogen from blood and/or cerebrospinal fluid (CSF) cultures obtained >3 days after birth(21). We were not able to distinguish LOI due to bacteremia versus meningitis. Recurrent and/or polymicrobial LOI episodes were included in the analysis. Coagulase-negative staphylococci (CoNS) LOI episodes were included if pathogens were isolated from blood or CSF, accompanied by at least one clinical sign of generalized infection, and ≥5 days of antibiotic therapy.

Exposures:

All eligible preterm infants were included in analysis of the primary outcome (incidence of Serratia LOI). For assessment of secondary outcomes among infants with LOI, the exposure of interest was Serratia versus non-Serratia LOI.

Outcomes:

The primary outcome was incidence of late-onset Serratia infections per 1000 eligible very preterm infants during the study period. Secondary outcomes included survival to discharge and morbidities among surviving infants with Serratia vs. non-Serratia LOI, including discharge with tracheostomy, gastrostomy (or jejunostomy), supplemental oxygen (defined as fraction of inspired oxygen >0.21), and any enteral human milk feeding. Technology dependence at discharge is an important patient-oriented outcome and is associated with increased risk of hospital readmission, healthcare resource utilization, and financial burden(22–24).

Covariates:

Covariates of interest were defined per the VON Manual of Operations(21). Infant demographic characteristics included sex, race, ethnicity, gestational age at birth, birth weight, and presence of a congenital anomaly. Small for gestational age (SGA) status was defined as birth weight <10^th percentile, per the Fenton growth chart(25). Maternal and delivery characteristics included: maternal hypertensive disorders (including pre-eclampsia), maternal diabetes of any type, chorioamnionitis, delivery mode, and maternal receipt of antenatal corticosteroids. Morbidities of prematurity included chronic lung disease (CLD, receipt of supplemental oxygen at 36 weeks’ postmenstrual age), severe intraventricular hemorrhage (IVH, grade 3–4), severe retinopathy of prematurity (ROP, stage 3–5), and necrotizing enterocolitis (NEC)(21). All morbidities occurred prior to NICU discharge or death. Length of stay was defined as the number of days elapsed between birth and hospital discharge or death. NICU level of care is defined using centers’ responses to the VON membership survey as: Type A (required to transfer infants for assisted ventilation based on infant characteristics or duration of ventilation required), Type B (provide mechanical ventilation without limitations on duration, no major surgery performed), and Type C (provide mechanical ventilation without limitations on duration, and major surgery performed, excluding cardiac surgery requiring bypass), and Type D (provide mechanical ventilation without limitations on duration and major surgery performed, including cardiac surgery requiring bypass)(26,27). Geographical regions were defined by US Census Bureau classifications.

Statistical Analysis

We determined the incidence of late-onset Serratia per 1000 eligible very preterm infants, and Serratia prevalence among all infants with LOI (stratified by gestational age and study year). Demographics, clinical characteristics, and outcomes were compared between infants with Serratia and non-Serratia LOI. To study the association of Serratia vs non-Serratia LOI with the secondary study outcomes, we used logistic regression with generalized estimating equations, adjusted for clustering of infants within hospitals and for multiple covariates (gestational age, SGA status, sex, delivery mode, inborn status, and presence of congenital anomalies). Risk ratios were estimated using the Poisson distribution with a log link function. The primary non-Serratia LOI comparison group was composed of LOI due to all non-Serratia bacteria (including CoNS) and fungi. To ensure comparison between infections of similar morbidity, in sensitivity analyses we compared infants with Serratia LOI versus those with (1) bacterial non-Serratia LOI, excluding infections with CoNS (an organism that may cause less severe infection among preterm infants(28,29)) and fungi, and with (2) non-Serratia gram-negative LOI. Finally, we described the distribution and hospital characteristics associated with Serratia infections occurring at VON centers. All analyses were performed using SAS version 9.4.

RESULTS

Of 119,565 eligible very preterm infants, 10,687 (8.9%) had at least one episode of LOI, and of those, 279 (2.6%) were infected with Serratia. The overall incidence of Serratia LOI was 2.3 infections per 1000 very preterm infants (95% confidence interval [CI] 2.1, 2.6). Cohort demographic and clinical characteristics are presented in Table 1. Median gestational age was similar among infants with Serratia and non-Serratia LOI (25 weeks, interquartile range [IQR] 24, 27), and lower compared to infants with no LOI (30 weeks, IQR 29, 32). Median birth weight was also similar among infants with Serratia and non-Serratia LOI, and was lower compared to infants with no LOI. Infants with Serratia LOI had a higher prevalence of CLD (73% vs 62%) and severe ROP (72% vs 62%) compared to infants with non-Serratia LOI, with similar rates of NEC and severe IVH between the two groups.

Table 1:

Maternal and infant demographics and clinical characteristics

	Serratia LOI			Non-Serratia LOI*			No LOI
	N	n	%	N	n	%	N	n	%
Maternal Characteristics
Maternal Race/Ethnicity
Black non-Hispanic, %	278	107	38.5	10,281	3,552	34.5	107,685	33,532	31.1
Hispanic, %	278	53	19.1	10,281	2,121	20.6	107,685	20,885	19.4
White non-Hispanic, %	278	108	38.8	10,281	3,834	37.3	107,685	44,453	41.3
Other non-Hispanic**, %	278	10	3.6	10,281	774	7.5	107,685	8,815	8.2
Antenatal steroids, %	278	228	82.0	10,346	9,052	87.5	108,459	95,690	88.2
Chorioamnionitis, %	272	47	17.3	10,274	1,823	17.7	107,973	13,148	12.2
Hypertension, %	274	86	31.4	10,317	3,118	30.2	108,322	43,555	40.2
Diabetes, %	268	27	10.1	10,280	939	9.1	108,120	12,625	11.7
Multiple gestation, %	279	75	26.9	10,408	2,186	21.0	108,874	26,562	24.4
Vaginal delivery, %	279	67	24.0	10,406	3,459	33.2	108,861	26,722	24.5
Infant Characteristics
Gestational age
≤23 weeks, %	279	54	19.4	10,408	1,833	17.6	108,874	3,946	3.6
24–25 weeks, %	279	115	41.2	10,408	3,628	34.9	108,874	13,312	12.2
26–27 weeks, %	279	69	24.7	10,408	2,554	24.5	108,874	20,968	19.3
28–29 weeks, %	279	23	8.2	10,408	1,568	15.1	108,874	31,148	28.6
>29 weeks, %	279	18	6.5	10,408	825	7.9	108,874	39,500	36.3
Gestational age (median, Q1, Q3)			25 (24, 27)			25 (24, 27)			30 (29, 32)
Birth weight, grams (median [Q1, Q3])	279		705 (590, 900)	10,406		760 (610, 985)	108,873		1,140 (870, 1,350)
Small for gestational age, %	278	53	19.1	10,311	1,622	15.7	108,644	21,183	19.5
Male, %	279	145	52.0	10,405	5,719	55.0	108,857	54,065	49.7
Inborn, %	279	227	81.4	10,408	8,344	80.2	108,878	94,698	87.0
Congenital anomaly, %	279	30	10.8	10,404	846	8.1	108,866	5,613	5.2
Morbidities of Prematurity
Necrotizing enterocolitis, %	279	46	16.5	10,400	1,683	16.2	108,862	4,161	3.8
Chronic lung disease, %	211	153	72.5	8,041	4,982	62.0	93,338	25,212	27.0
Retinopathy of prematurity, %	217	156	71.9	8,228	5,075	61.7	87,365	25,444	29.1
Intraventricular hemorrhage, %	276	123	44.6	10,136	4,353	42.9	100,148	23,466	23.4
Morbidities at Discharge
Tracheostomy, %	279	17	6.1	10,396	351	3.4	108,851	683	0.6
Gastrostomy or jejunostomy, %	279	48	17.2	10,396	1,254	12.1	108,851	4,031	3.7
Oxygen at discharge, %	166	75	45.2	6,730	2,529	37.6	94,482	12,487	13.2
Human milk at discharge, %	174	46	26.4	6,792	2,538	37.4	94,580	49,079	51.9
Survival, %	274	206	75.2	10,258	8,030	78.3	108,323	102,789	94.9
Length of Stay, days [median, (Q1, Q3)]
Overall	274		117 (72, 178)	10,235		102 (62, 140)	108,135		62 (41, 89)
Among survivors	206		136 (102, 187)	7,999		113 (86, 149)	102,596		64 (44, 91)
Among non-survivors	68		29 (13, 59)	2,226		21 (11, 46)	5,521		13 (6, 31)

N: total number of infants with data available

n: number of infants with given characteristic among total

^*Infants with other late bacterial, CoNS, or fungal infection

^**Includes Asian/Pacific Islander, Native American/Alaska Native, Other

Rates of LOI due to Serratia increased with decreasing gestational age and birth weight, but did not differ by year of birth (Table 2). Infants with Serratia were more likely to have multiple distinct episodes of LOI or episodes of polymicrobial LOI; 35% of infants with Serratia had LOI due to ≥2 pathogens, compared to 13% of infants with non-Serratia LOI.

Table 2:

Serratia incidence by study year and gestational age category among infants with any late infection

Category	All infants with LOI (n)*	Infants with Serratia LOI (n, %)	Serratia incidence rate per 1000 infants with any late infection (95% CI)
Overall	10679	279 (2.6)	26.1 (22.4, 30.4)
Year of Birth
2018	3550	95 (2.7)	26.8 (20.6, 34.7)
2019	3651	98 (2.7)	26.8 (20.8, 34.6)
2020	3478	86 (2.5)	24.7 (18.8, 32.5)
Gestational Age (weeks)
≤23	1884	54 (2.9)	28.7 (20.3, 40.3)
24–25	3739	115 (3.1)	30.8 (24.3, 38.9)
26–27	2623	69 (2.6)	26.3 (19.4, 35.6)
28–29	1591	23 (1.5)	14.5 (8.5, 24.4)
>29	842	18 (2.1)	21.4 (11.8, 38.4)
Birth Weight (grams)
≤500	875	25 (2.8)	27.8 (16.8, 45.7)
501–750	4228	139 (3.2)	31.8 (25.7, 39.4)
751–1000	2844	70 (2.4)	24.0 (17.7, 32.5)
1001–1250	1465	34 (2.3)	22.7 (14.7, 34.9)
1251–1500	854	11 (1.3)	12.7 (6.0, 26.9)
≥1500	132	0 (0.0)	0.0

^*Infants with Serratia, other late bacterial infections (including coagulase-negative Staphylococcus), or fungal infections.

Of 774 participating hospitals, 161 (21%) reported at least one late-onset Serratia infection during the three-year study period (see Table, Supplemental Digital Content 1). Of 279 Serratia LOI episodes, 271 occurred at VON centers and were included in the center-level analysis. The majority of Serratia LOI episodes (170/271, 63%) occurred at 137 hospitals reporting only 1 or 2 Serratia infections during the study period; 54/271 (20%) Serratia LOI episodes occurred at 9 hospitals reporting ≥5 Serratia infections. Hospitals reporting Serratia infections were more frequently located in the southern United States, were teaching institutions offering higher levels of neonatal and surgical care, and had higher annual admissions (Table 3).

Table 3:

Characteristics of Hospitals Reporting Neonatal Serratia LOI

	Hospitals with ≥1 Serratia Case		Hospitals with No Serratia Cases
	N	%	N	%
NICU Type
A	161	0.6	613	13.5
B	161	19.9	613	44.4
C	161	42.2	613	34.3
D	161	37.3	613	7.8
NICU beds - med (Q1, Q3)	161	40 (24, 60)	610	20 (12, 30)
Total NICU admissions - med (Q1, Q3)	159	691 (445, 964)	592	327 (218, 537)
Teaching hospital	161	70.2	601	44.3
Single family rooms
≤ 10%	159	45.9	609	55.2
11–50%	159	8.2	609	4.4
51–90%	159	12.6	609	5.9
≥ 91%	159	33.3	609	34.5
Region
Northeast	161	13.7	613	15.2
Midwest	161	18.6	613	21.4
South	161	53.4	613	33.6
West	161	14.3	613	29.9

N: total number of infants with data available.

The overall proportion of infants surviving to discharge was similar between infants with Serratia and non-Serratia LOI (75% vs 78%, respectively). However, survival was significantly lower among infants with Serratia after adjusting for potential confounders, including gestational age (adjusted risk ratio [aRR] 0.88, 95% CI 0.82, 0.95) (Table 4).

Table 4:

Outcomes of Serratia vs non-Serratia late-onset infection

	Serratia LOI			Non-Serratia LOI1			Adjusted risk ratio (95% CI)2
	N	n	%	N	n	%
Survival to discharge	274	206	75.2	10,258	8,030	78.3	0.88 (0.82, 0.95)
Among survivors discharged home:
Tracheostomy placement	279	17	6.1	10,396	351	3.4	2.78 (1.68, 4.59)
Gastrostomy or jejunostomy placement	279	48	17.2	10,396	1,254	12.1	1.83 (1.41, 2.38)
Oxygen at discharge	166	75	45.2	6,730	2,529	37.6	1.29 (1.08, 1.54)
Human milk at discharge	174	46	26.4	6,792	2,538	37.4	0.64 (0.50, 0.83)

N: total number of infants with data available

n: number of infants with given characteristic among total

^1:Includes all non-Serratia late bacterial LOI (including coagulase-negative Staphylococcus) and fungal LOI

^2:Adjusted for clustering of infants within hospitals, and for gestational age in weeks, small for gestational age status, sex, mode of delivery, inborn status, and presence of a congenital anomaly

In adjusted analyses describing discharge morbidities among LOI survivors, infants with Serratia LOI were more likely to be discharged with tracheostomy (aRR 2.8, 95% CI 1.7, 4.6) and gastrostomy (aRR 1.8, 95% CI 1.4, 2.4), compared to infants with non-Serratia LOI due to any bacteria or fungi. At discharge, Serratia LOI was associated with increased risk for supplemental oxygen use (aRR 1.3, 95% CI 1.1, 1.5) and decreased likelihood of any enteral human milk feeding (aRR 0.64, 95% CI 0.5, 0.8) among survivors, compared to surviving infants with non-Serratia LOI (Table 4).

In a sensitivity analysis, adjusted survival among infants with Serratia LOI was not different compared to infants with non-CoNS bacterial LOI, and was higher compared to infants with other gram-negative LOI (see Tables, Supplemental Digital Content 2 and 3). Compared to infants with non-CoNS bacterial LOI, infants with Serratia had significantly higher associated risk of discharge with gastrostomy. There were no statistical differences in the risk of discharge morbidities when comparing infants with Serratia LOI versus other gram-negative LOI.

DISCUSSION

In this large, contemporary, nationally-representative cohort, Serratia accounted for 2.6% of LOI episodes among very preterm infants, similar to rates reported in earlier US-based cohorts(1,2). We further demonstrate that invasive Serratia LOI is not rare, and that infection with this opportunistic pathogen is associated with lower survival and increased morbidity at discharge, compared to infants infected with non-Serratia pathogens.

This study provides new insights into the association of Serratia LOI with morbidity and mortality risks in very preterm infants. Technology dependence at discharge is an important patient-oriented outcome with a multifactorial risk profile. LOI is associated with (and may modify) risk of severe BPD, via inflammatory-mediated pathways (30–32), and has been independently associated with tracheostomy placement in preterm infants(33). BPD, growth failure, and neurodevelopmental impairment are all identified risk factors for gastrostomy placement in preterm infants; all of these conditions have are further associated with antecedent LOI (22,34–36). Although Serratia accounts for a low proportion of neonatal infections and has variable pathogenicity in infants (ranging from asymptomatic colonization to invasive disease)(5,19,37), we identified lower survival and higher risks of discharge morbidities (tracheostomy, gastrostomy, and supplemental oxygen) associated with Serratia LOI, compared with non-Serratia LOI. These findings were attenuated when restricting the comparison group to non-CoNS bacterial LOI; although mortality was equivalent, excess morbidity risk associated with Serratia LOI remained. A final comparison was restricted to infants with Serratia LOI versus other gram-negative LOI, given higher mortality associated with gram-negative LOI (compared to gram-positive or CoNS LOI)(3,38–40). Infants with Serratia LOI had slightly higher survival and similar risks of discharge morbidities compared to infants with other gram-negative LOI, underscoring the virulence of this opportunistic pathogen among preterm infants.

While Serratia infections account for a small percentage of all invasive neonatal LOI, our center-level analysis demonstrates that this infection burden is distributed across 21% of all hospitals reporting data to VON. Our study reinforces prior work identifying Serratia infections as more frequently occurring in large, academic NICUs offering complex medical and surgical services(2), likely reflecting a more chronically ill patient population with longer lengths of stay and higher technical care utilization(2,6,15,18,41). Although most published neonatal Serratia literature describes infections occurring in outbreak patterns, center-level data from our study suggest that Serratia LOI largely occurs sporadically. Most Serratia LOI (170/271, 63%) occurred in hospitals that reported ≤2 Serratia infections over the three-year study period; however, the nine highest-burden Serratia hospitals (≥5 episodes of Serratia LOI, 1% of all 774 hospitals) accounted for 20% (54/271) of infections. Serratia cases appeared to occur more commonly in NICUs offering higher levels of service. Among 108 type D NICUs, 48 (44%) did not have a Serratia LOI during the study period – though given the sporadic infection pattern, it is difficult to identify type D units with no Serratia LOI due to chance, versus those without Serratia LOI due to particularly effective infection prevention procedures. Hospitals reporting Serratia or non-Serratia LOI were proportionally more commonly located in the southern US; this geographic distribution has been similarly reported at the patient level(2), though it should be noted that neither study reports geographic distribution of patients and hospitals without LOI. Further elucidation of Serratia epidemiology by US geographic region may require national surveillance, including consideration of hospital-wide Serratia burden in addition to NICUs. We acknowledge that this study may underestimate outbreak patterns, as we exclusively studied Serratia bacteremia and meningitis and thus did not capture other infections (e.g., conjunctivitis, UTIs) that are also reported in Serratia outbreaks.

Serratia species are intrinsically resistant to beta-lactam antibiotics and carry inducible beta-lactamase resistance via chromosomal ampC genes. Depending on the degree of ampC expression, beta-lactamases variably reduce the bactericidal capacity of third-generation cephalosporins and piperacillin-tazobactam (commonly-used empiric antibiotics in NICUs) against Serratia(42). In a large cohort study of neonatal sepsis in low- and middle-income countries, Serratia was the second-most common gram-negative pathogen and accounted for 6% of all neonatal sepsis cases. Whole-genome sequencing of bacterial isolates in that study demonstrated concerning rates of antimicrobial resistance genes, including extended-spectrum beta-lactamases (ESBLs) and carbapenemases(43). In the face of rising Serratia incidence among pediatric(2) and adult(44) populations, as well as increasing antimicrobial resistance rates due to ESBL-producing Enterobacterales, renewed attention to Serratia infection prevention is needed to mitigate infectious burdens and use of reserve antibiotics for resistant isolates.

Reduction of Serratia infection burden in NICUs requires both patient- and unit-level strategies for surveillance and decontamination. The prevalence of invasive Serratia disease in NICUs across the US likely reflects even higher underlying colonization rates that are potentially amenable to infection control measures. Epidemiologic surveillance within NICUs experiencing Serratia outbreaks has identified neonatal Serratia colonization, particularly in the gastrointestinal tract, as the most important infectious reservoir. Single-unit colonization rates amidst Serratia outbreaks are reported to range widely (28–69% of screened infants)(10,12,45,46), with isolates identified in feces and in swabs of eye, throat, umbilicus, and rectum(10,45–47). Conversion rates from Serratia colonization to clinical infection, which ranged from conjunctivitis to bacteremia/meningitis, varied from 11–28%(10,12,47,48). Active surveillance for Serratia may enable rapid initiation of enhanced hand hygiene and environmental cleaning procedures in the setting of colonization or clinical outbreaks. Reichert et al.(49) calculated pathogen-specific risks of additional BSIs among preterm infants in the same NICU once an index case was isolated; though its incidence density was low, Serratia had the highest associated risk of producing additional incident BSIs among all organisms analyzed (relative risk 77.5, 95% CI 41, 146), suggesting the potential value of initiating enhanced containment measures if index cases are detected. Approaches to reducing Serratia transmission within NICUs have largely focused on improving hand hygiene; efforts to cohort colonized/infected infants, institute contact precautions, and reduce nursery overcrowding are also described(10,12,47,50). Enhanced environmental cleaning is generally indicated; no consistent environmental reservoir is reported, but Serratia outbreaks in NICUs have been linked to sinks and water sources(51,52), incubator doors(51,53), laryngoscope blades(8), air conditioning ducts(17), and contaminated breast pumps(9), human milk(54), soap(7,55), and parenteral nutrition(56).

Strengths of this study include analysis of a large, nationally representative, contemporary cohort of very preterm infants admitted to both academic and community NICUs, supporting generalizability of our findings. However, we acknowledge study limitations. Infants with Serratia LOI were more likely to have LOI due to multiple organisms; however, the data reporting structure does not differentiate whether this was due to multiple distinct episodes of LOI growing different organisms, or a single polymicrobial LOI episode. It is therefore possible that multiple infectious episodes could have contributed to excess morbidity identified among patients with Serratia LOI. Our definition of LOI included bacteremia and meningitis, though we could not distinguish between these two types of infections. This definition did not include infections of the respiratory tract, urinary tract, and conjunctivae (all reported Serratia infection sites in preterm neonates(57)); therefore, the Serratia infection burden in this cohort could be underestimated. Because the VON dataset does not record the timing of positive blood cultures relative to infant birth, associations of LOI with some morbidities should be interpreted cautiously. For example, we cannot comment on potential causal associations of Serratia infection relative to onset of any morbidities of prematurity (e.g., CLD, ROP). Serratia LOI was associated with increased risks of discharge with tracheostomy and gastrostomy; it is possible that these procedures could have occurred prior to LOI onset, though the majority (90%) of LOI in very preterm infants is reported to occur within the first two postnatal months(58,59), while tracheostomy and gastrostomy placements usually occur late in the NICU course among infants corrected beyond term postmenstrual age(60,61). As this study focuses on Serratia LOI among very preterm infants, we are unable to comment on Serratia epidemiology among infants born at higher gestations and birth weights. No antimicrobial susceptibility data were available, so we are unable to describe Serratia antibiotic resistance patterns in this cohort.

CONCLUSIONS

Serratia is a persistent cause of opportunistic invasive late-onset infections in NICUs, affecting 2.3 infants per 1000 very preterm births and occurring in 21% of neonatal units within this large cohort from the United States. Increased risks of death and morbidity among very preterm infants with Serratia LOI reinforce the need for recognition and targeted prevention strategies for this nosocomial infection.

DATA AVAILABILITY

The dataset analyzed during the current study derives from the Vermont Oxford Network database and is not publicly available.