Risk of Transmission of Vaccine-Strain Rotavirus in a Neonatal Intensive Care Unit That Routinely Vaccinates

Since the introduction of second-generation oral rotavirus vaccines in 2006, rotavirus-associated morbidity and mortality have been markedly reduced among children worldwide.^1,–3 Postmarketing evaluations identified a small risk of intussusception from rotavirus vaccines in some populations, including in US infants.⁴ Some vaccinated infants shed live attenuated vaccine-strain rotavirus in stool for weeks after receipt^5,6; thus, many neonatal intensive care units (NICUs) avoid administering rotavirus vaccine during hospitalization due to the theoretical risk of transmission of vaccine strains to unvaccinated infants, including critically ill or age-ineligible preterm infants. In the 2009 US Advisory Committee on Immunization Practices (ACIP) vaccine recommendations, ACIP advised against administering rotavirus vaccine to infants during NICU hospitalization and supported rotavirus vaccination of eligible infants at the time of discharge.⁷ This decision was made with the understanding that some infants at risk of severe complications from rotavirus disease would never receive rotavirus vaccine given the maximum age for initiating rotavirus vaccination (14 weeks, 6 days).⁸ 

The few previous evaluations of rotavirus vaccination in NICUs that included systematic stool collection were limited by infrequent administration of vaccine. A report from a US hospital identified no instances of RotaTeq (RV5; Merck & Co.) strain in stools of unvaccinated infants hospitalized in ICUs (including a NICU), although only 19 RV5 doses were administered.⁹ In a Japanese NICU, no transmission was detected after 19 infants received at least 1 RV5 dose and 10 infants received at least 1 dose of Rotarix (RV1; GlaxoSmithKline) and were subsequently placed on contact precautions.¹⁰ Retrospective studies performed in NICUs found no evidence of vaccine-strain transmission to unvaccinated infants,^10,11 although case series described nosocomial transmission of vaccine-strain rotavirus between immunocompromised infants in non-NICU settings.^12,13 

This study examined the incidence of suspected transmission of RV5 strain(s) in a large cohort of infants hospitalized in a NICU that administers more than 200 RV5 doses annually to eligible inpatients, through systematic collection and testing of stool samples.

The Newborn/Infant Intensive Care Unit at Children’s Hospital of Philadelphia (CHOP) is a 100-bed, level IV NICU with ∼1200 admissions annually. Approximately one-third of the patients are admitted for surgical diagnoses. The unit includes 38 private rooms; the remainder of the beds are distributed across 14 multibed rooms (each containing 3–8 beds).

Eligibility for and prescription of RV5 is determined by clinicians caring for the patient and guided by clinician judgment and existing formulary guidance. Institutional policies for RV5 vaccination follow ACIP and Vaccines for Children Program guidelines,¹⁴ with initiation between day of life 42 and 104 (usually starting at age 2 months or as soon as eligible thereafter), administration of doses 2 and 3 at age 4 months and 6 months, respectively, (or ≥4-week intervals for patients on delayed schedules), and completion of the series by age 8 months, 0 days. The NICU’s infection prevention policies require neither contact precautions nor cohorting after vaccination.

All patients admitted to the NICU from January 2021 through January 2022 were included in the cohort and eligible for weekly stool collection. This study was approved by CHOP’s Institutional Review Board with a waiver of consent and Health Insurance Portability and Accountability Act (HIPAA) authorization and was reviewed by the Centers for Disease Control and Prevention (CDC) Human Subjects Committee.

Collection of a stool sample was attempted for every infant each week of their NICU stay. If infants were producing scant or no stool, or discharged prior to weekly collection, they were considered ineligible for sample collection that week. Nurses collected samples, either transferring stool into a sterile container or placing a patient’s diaper in a plastic bag and sealing. Samples were aliquoted, frozen at −80 °C and shipped monthly on dry ice to CDC’s Rotavirus Surveillance and Molecular Epidemiology Laboratory (Supplemental Materials).

Rotavirus testing was performed after all samples had been received at CDC. Viral RNA was extracted from clarified 10% suspensions and tested for rotavirus RNA using the nonstructural protein 3 (NSP3) real-time reverse transcription polymerase chain reaction (qRT-PCR) assay.^15,16 All samples with rotavirus NSP3 RNA were tested with RV5 vaccine–specific RT-PCR assays (Supplemental Materials). Samples were considered RV5 positive if the primary and at least 1 secondary RV5 assays were positive. If neither secondary assay was positive, the sample was re-extracted twice and the primary RV5 assay repeated—samples positive with the primary assay in at least 2/3 total runs were considered RV5 positive; samples positive with the primary assay in only 1/3 runs were considered possibly RV5 positive.

To estimate the incidence of RV5 transmission to unvaccinated infants, each stool collected from a nonvaccinated patient (“nonvaccinated stool”) was considered to contribute 1 patient-day at risk for RV5 strain transmission (denominator). For nonvaccinated patients with more than 1 RV5-positive or possibly RV5-positive sample, only the first positive sample was counted (numerator). For patients with a sample that had an invalid result (the extraction control was not valid upon repeated testing), only data from samples collected before the invalid sample were included. For the 95% CI, we assumed transmissions followed a Poisson distribution.

A presumed transmission recipient (“recipient”) was defined as a patient with an RV5-positive or possibly positive stool sample prior to receipt of any rotavirus vaccine (Supplemental Materials). For each recipient, systematic investigations were conducted to identify patients who might have been the source of the RV5 strain (“source”). To identify possible transmission from “near-neighbors,” we reviewed vaccine histories for all roommates to ascertain receipt of RV5 during the “exposure window,” defined as the period from 1 to 14 days before collection of the recipient’s first RV5-positive sample (censored at day of admission to study NICU). To identify possible transmission via shared health care worker (HCW), we enumerated all NICU patients who received RV5 during that exposure window and examined records for staffing overlap (ie, HCW documented as providing care on the same day to both source and recipient patients) (Supplemental Materials).

Using 2-dimensional blueprints of the NICU imported into digital mapping software (ArcGIS Pro 3.2.0; 2023 ESRI Inc.), we analyzed the physical and environmental characteristics of each bed space and determined the “total room” space; ratio of standard bed footprint to room area; and distance from each bed to fixed auxiliary elements in the clinical care space, including handwashing sinks (Supplemental Figure 7). We assigned all beds to 20 mutually exclusive pods; all beds within a pod were less than 10 m of direct traversable distance from the pod center point. Pod assignments were performed to maximize the number of related beds. Impacted beds and pods were defined as a bed or pod with either a presumed recipient or a presumed source patient.

We performed statistical analyses of bed and pod characteristics, comparing impacted beds and pods vs nonimpacted beds and pods. Continuous variables were examined using 2 sample t tests, and categorical variables by χ² test.

Patient characteristics were extracted from patient demographics in electronic health records (EHRs). χ² tests of independence and Fisher exact tests were performed using RStudio, version 2023.6.1.524 (Posit Software, Boston, MA) and Stata 15.1 (StataCorp, College Station, TX), to examine associations of demographic and clinical factors with receipt of RV5. Because other investigators have observed differential vaccination rates based on race and ethnicity, we examined these characteristics.¹⁷ 

Among infants who received their first RV5 dose in the NICU before the date of discharge, we ascertained the number who would have become age-ineligible to start the series (ie, age >104 days at discharge) if they had not been vaccinated in the NICU. Similarly, we ascertained if infants who received at least 1 dose of rotavirus vaccine before NICU admission and received their second or third RV5 dose in the NICU would have become age-ineligible to complete their series (age >8 months, 0 days) by the discharge date.

The study cohort included 1238 unique patients who had a median NICU length of stay (LOS) of 11.8 days (IQR: 4.0–38.9) (Table 1). The average daily census during the study period was 100 patients. Approximately half of the cohort were white (659, 53%), and 203 patients (16%) were identified as Hispanic/Latino ethnicity. Most (678, 55%) were at least 37 weeks’ gestation and had NICU stays of 8 weeks or less (999, 81%). The median NICU LOS for the 26% of patients with gastrointestinal pathology was 28.4 days (IQR: 10.1–80.4), compared with 8.3 days (IQR: 3.2–26.6) for patients without gastrointestinal pathology (Supplemental Table 1). The median NICU LOS for infants at least 37 weeks’ gestation was 7 days (IQR: 2.9–18.1); LOS increased as gestational age at birth decreased (Supplemental Table 2).

A total of 488 patients (39.4%) were in the NICU while they were eligible for any dose of RV5 (based only on age and timing since any previous dose; Table 1). During the study period, 226 RV5 doses were given to 162 patients (132 first, 67 second, 27 third doses). Among 162 patients who received at least 1 dose, 72% were born premature (<37 weeks’ gestation), and 32% had a history of gastrointestinal pathology. Prematurity and longer NICU LOS were associated with receipt of RV5 in the NICU.

In 2 instances, first doses were given outside age guidelines (15 weeks, 0 days and 15 weeks, 1 day). In 1 case, a third dose was given outside guidelines at age 8 months, 10 days. Of the 126 infants who received their first dose at age 104 days or younger and prior to the discharge date (excluding 4 infants who received first dose on discharge date), 69 (55%) were aged older than 104 days at discharge and therefore would have been age-ineligible to start the series (30 of these infants received 1 dose in the NICU, 23 received 2 doses, and 16 received 3 doses). During the study period, 93 second or third doses were given to infants aged 8 months, 0 days or younger. Of the 3 infants who received a first rotavirus vaccine dose before their NICU admission and then received a second and third RV5 dose in the NICU, each was aged older than 8 months, 0 days at discharge (eg, age-ineligible for doses).

Sixty-three percent (774/1238) of patients provided at least 1 stool sample (median 2 samples/patient [IQR: 1–6] among those with samples), for 3448 total samples, including 2252 samples from 686 unvaccinated patients (Figure 1). Seventy-five percent (346/464) of patients who did not contribute a stool sample were admitted for less than 7 days. Each week, a mean of 32.4 (±7.4) patients were considered unlikely to produce stool and therefore ineligible for stool collection (Supplemental Methods). For all patients, the mean weekly sample collection rate was 61% (±7.5%). For patients expected to produce a sample, the mean weekly collection rate was 87% (±9.2%).

RV5 strain was detected in 212 (6%) samples and possibly detected in 6 samples (0.2%); nontypeable rotavirus was detected in 10 (0.3%) samples (Supplemental Results). No rotavirus was detected in 3165 (92%) samples (RV NSP3 negative). For 55 (1.6%) samples, a valid result was not obtained (during repeated testing, the extraction control was invalid).

RV5 was detected in 70% (59/84) of samples collected during the first week and 41% (36/87) during the second week after administration of RV5 dose 1 (Figure 2). In samples collected during the first week after doses 2 and 3, RV5 was detected in 48% (20/42) and 32% (6/19) of samples, respectively (Supplemental Figures 1–3).

Most patients (681/686, 99.3%) who contributed at least 1 nonvaccinated sample never tested positive for RV5. Five patients (<1%) had nonvaccinated stools that tested RV5 positive (n = 3) or possibly RV5 positive (n = 2) (Table 2). Four of these patients had RV5 detected in their first collected stool (1–5 days after admission). One patient had RV5 detected in their second collected stool (6 days after admission) after testing negative in a stool collected on day 2 after admission. The estimated rate of presumed RV5 strain transmission to unvaccinated infants was 5/2252 patient-days at risk (2.2 events/1000 patient-days [95% CI: 0.7–5.2]).

One patient with documentation of RV1 administration (lot number in both the primary care practice record and state immunization registry) prior to NICU admission tested RV5 positive in their first stool collected (3 days after admission). This initial stool was negative for RV1; subsequent samples were RV5 negative. By report, 5 other infants received RV1 before NICU admission; all 7 of their samples were NSP3 negative. Including data from the RV1-vaccinated infants, the estimated rate of RV5 strain transmission to unvaccinated or RV1-vaccinated infants was 6/2260 patient-days (2.7 events/1000 patient-days [95% CI: 1.0–5.8], with 686/692 (99.1%) infants never testing RV5 positive (Supplemental Materials).

Investigations of transmission events (Table 2; Supplemental Figure 9) revealed 4 of 5 unvaccinated transmission recipients shared an HCW or care team with a recently vaccinated patient. One of those recipients also had near-neighbor exposure; during their exposure window, their bedspace was immediately adjacent to that of a recently vaccinated patient. The transmission recipient who received RV1 prior to admission shared an HCW with a recent vaccinee (Supplemental Figure 8).

Based on chart review, none of the recipients had symptoms of gastroenteritis, unexplained fever, or unexplained feeding intolerance during the period from 7 days before through 7 days after collection of their first RV5-positive stool.

During the investigation of these transmission events to unvaccinated or RV1-vaccinated recipients, we identified 6 separate bed spaces that housed recipients, whereas an additional 6 bed spaces housed possible source patients. Impacted, as compared with unimpacted, bedspaces were more likely to be in open pods vs private rooms (11/12 vs 51/89, P = .02).

We assessed the relative crowdedness of impacted vs unimpacted pods; the ratio of beds to total pod area was not associated with transmission events (Supplemental Tables 3 and 4). Impacted pods tended to be larger than unimpacted pods (115.49 vs 101.50 m², P = .22). The ratio of sinks to beds was greater in impacted pods vs unimpacted (0.96 vs 0.67, P = .045). The mean distance from a bedspace to a fixed handwashing sink was not associated with transmission events.

This prospective surveillance study suggests that transmission of RV5-strain rotavirus is infrequent and without detectable clinical consequence when administered to eligible NICU patients. In the absence of postvaccination contact precautions, we estimated the rate of transmission of RV5 strain was 2.2 events/1000 patient-days-at-risk. In this population of critically ill and premature infants, in-NICU administration of RV5 ensured vaccination of many infants at high risk of severe rotavirus disease.

After administration of the first RV5 dose, viral genome was detected in 70% of stools and persisted for more than 7 days in approximately one-third of vaccinated infants. These rates are similar to those detected by RT-PCR after dose 1 administration to healthy infants.^18,19 These data suggest viral replication occurred and that at any given time there may have been multiple infants who were shedding live RV5 strains.

Five of 6 possible transmission events were detected in the first collected stool, all of which were obtained within a few days of admission (including 3 patients who had been home prior). It is possible that some recipients acquired RV5 strain before admission to our NICU, thus inflating estimates of the in-NICU transmission rate. Alternatively, processes of care associated with admission might have increased the transmission risk (eg, breaks in hand hygiene while stabilizing acutely ill neonates). Although not possible to confirm sources, we identified possible transmission sources for 4 of 5 nonvaccinated infants; 4 shared an HCW with a source patient and 1 also had a recently vaccinated near-neighbor. Additionally, the RV1-vaccinated infant with a stool positive for RV5 shared an HCW with an infant who was recently vaccinated with RV5. We hypothesize lapses in hand hygiene after handling the vaccine dosing tube or inadequate cleaning and disinfection of shared equipment might have contributed to these transmissions. Alternatively, RV5 strain detection in a sample from a non–RV5-vaccinated infant could reflect contamination of that stool sample (eg, from HCW hand or environmental surface).

In geospatial analyses, all presumed transmission recipients and probable source patients were in multibed rooms. Crowdedness and distance to handwashing facilities were not associated with transmission, although our analysis was limited by the small number of transmission events. The clustering of events in 6 geographically proximate pods may suggest additional colinear patient-level factors that increased the likelihood of transmission, including geographic clustering of patients with similar comorbidities.

From our chart reviews, transmission events were infrequent and without clinical consequences, confirming findings of prior inpatient studies performed at other institutions.^9,11,20,21 We found that more than half of infants who initiated their RV5 series would have become age-ineligible by their discharge date if in-NICU vaccination had not occurred. Together, these findings suggest that the benefits of in-NICU administration of RV5 vaccine to stable, age-eligible infants, including premature infants, outweigh the risks. These findings provide additional data to support the American Academy of Pediatrics’ (AAP) Red Book stating that individual institutions may consider administering rotavirus vaccine at the recommended chronologic ages, including in NICUs.²² 

This study has limitations. Stools were collected once weekly; thus, infants with transient RV5 strain shedding may have been missed. Systematic documentation of eligibility for RV5 was not available, so we were unable to determine the proportion of clinically eligible patients who received RV5. The lack of samples from infants not producing stool during a week may have contributed to under-ascertainment of transmission events; the large number of samples tested and vaccine doses administered provides estimates more robust than those derived from smaller studies. Conversely, we did not collect stools at admission, which might have led to overestimation of the rates of in-NICU transmission. Lastly, this study was conducted while COVID-19 precautions were in place. Although precautions such as universal masking are unlikely to have affected RV5 strain transmission, other unmeasured COVID-19–related changes, such as heightened emphasis on hand hygiene, could have reduced transmission as compared with prepandemic rates.

Transmission of RV5 strain occurred infrequently and was without clinical consequence when RV5 was administered to stable, age-eligible NICU patients. Furthermore, in-NICU RV5 administration allowed vaccine-induced protection against rotavirus disease of a substantial proportion of long-stay NICU patients who would have become age-ineligible to ever start the series if rotavirus vaccination had been delayed until discharge. These findings add to the existing safety data and suggest that the known benefits of NICU administration of RV5 outweigh the low risks of vaccine-strain transmission.

We thank Amy Hopkins (CDC) for her assistance with specimens in the laboratory, Matt Devine (CHOP) for assistance with compiling from the EHR, and the CHOP infection prevention and control team and NICU nursing staff for their support in ensuring safe collection of specimens.

ACIP

Advisory Committee on Immunization Practices

CDC

Centers for Disease Control and Prevention

CHOP

Children’s Hospital of Philadelphia

EHR

Electronic health record

HCW

Health care worker

NICU

Neonatal intensive care unit

NSP3

Nonstructural protein 3

RT-PCR

Real-time quantitative reverse-transcription polymerase chain reaction

RV1

Monovalent rotavirus vaccine

RV5

Pentavalent rotavirus vaccine