Association of Rotavirus Vaccination With Inpatient and Emergency Department Visits Among Children Seeking Care for Acute Gastroenteritis, 2010-2016

IMPORTANCE RotavirusvaccineshavebeenrecommendedforuniversalUSinfantimmunizationfor more than 10 years, and understanding their effectiveness is key to the continued success of the US rotavirus vaccine immunization program. OBJECTIVE To assess the association of RotaTeq (RV5) and Rotarix (RV1) with inpatient and emergency department (ED) visits for rotavirus infection. DESIGN, SETTING, AND PARTICIPANTS This case-control vaccine effectiveness study was performed at inpatient and ED clinical settings in 7 US pediatric medical institutions from November 1, 2009, through June 30, 2016. Children younger than 5 years seeking medical care for acute gastroenteritis were enrolled. Clinical and epidemiologic data, vaccination verification, and results of stool sample tests for laboratory-confirmed rotavirus were collected. Data were analyzed from November 1, 2009, through June 30, 2016. MAIN Rotavirus vaccine effectiveness for preventing rotavirus-associated inpatient and ED visits over time for each licensed vaccine, stratified by clinical severity and age. Abstract(continued) (95% CI, 76%-84%) against moderate infections, and 91% (95% CI, 85%-95%) against severe infections. CONCLUSIONS AND RELEVANCE Evidence from this large postlicensure study of rotavirus vaccine performance in the United States from 2010 to 2016 suggests that RV5 and RV1 rotavirus vaccines continue to perform well, particularly in preventing inpatient visits and severe infections and among younger children. second (range, 82% 12 to 90% 17 ), third (range, 83% 11 to 88% 12 ), and fourth (range, 76% 12 to 79% 11 ) years of life were comparable to our findings. Our RV1 findings for these age groupings have greater study power and further clarify these previously published results. Earlier RV1 vaccine effectiveness estimates were previously derived from smaller samples, reflecting the delayed uptake of RV1 coverage across the United States.


Introduction
The New Vaccine Surveillance Network (NVSN) has continuously monitored the effects and effectiveness of licensed rotavirus vaccines since the recommendation for universal US infant rotavirus vaccination by the Advisory Committee on Immunization Practices (ACIP). 1  During the last decade, as the US rotavirus vaccination program has matured, precipitous decreases have been sustained in US hospitalizations [6][7][8] and emergency department (ED) visits 9,10 attributable to rotavirus gastroenteritis. However, this evidence of a public health effect may sometimes seem incongruous in relation to patient-level detections of rotavirus in children, particularly with the recent advent of highly sensitive multipathogen laboratory tests that potentially detect rotavirus shed from recently resolved infections and mild illnesses and even live-attenuated rotavirus vaccine virus shed naturally after immunization. Unbiased assessments of rotavirus vaccine effectiveness are important to maintain the public confidence in the US rotavirus immunization policy. Furthermore, rigorous US vaccine performance assessments inform and influence policy makers throughout the world on the initiation and continuation of their own rotavirus vaccination programs.
We conducted active rotavirus surveillance at 7 geographically diverse US pediatric medical institutions participating in the NVSN using a common protocol for participant enrollment, data collection, and laboratory confirmation from November 1, 2009, through June 30, 2016. To discern trends in rotavirus vaccine effectiveness during an extended period, we standardized and aggregated published NVSN data from 4 years (2010-2013) 11,12 with previously unpublished NVSN data from 3 years (2014)(2015)(2016). This accrual of comparable methods and case ascertainment across time allows an evaluation of rotavirus vaccine performance in the United States with large sample sizes, precision in results, and the ability to measure changes in rotavirus vaccine effectiveness over time.
(Oakland). Institutional review board approval was obtained from the CDC and each study site, and all parents or guardians of participants provided written informed consent. Analysis  The NVSN used common surveillance and laboratory protocols across all sites during the 2010 to 2016 period covered by this analysis. Children younger than 5 years were enrolled with informed consent from a parent or guardian on hospitalization or while visiting the ED with acute gastroenteritis (AGE), defined as diarrhea (Ն3 episodes within 24 hours) and/or vomiting (Ն1 episode within 24 hours). Exclusionary criteria involved indications of a noninfectious etiology, a history of immune deficiency, previous enrollment for the same AGE episode, or transfer from another hospital. Children who were originally enrolled in the ED but were subsequently hospitalized for the same illness were categorized as inpatients.

Specimen Collection and Case Determination
Whole stool specimens were obtained within 10 days after AGE symptom onset, with more than 95% of specimens obtained within 7 days after onset. Surveillance sites first tested specimens using an enzyme immunoassay (EIA) (Premier Rotaclone; Meridian Bioscience, Inc). Specimens were shipped to the CDC, where rotavirus strains were genotyped using reverse transcription-polymerase chain reaction (RT-PCR) assays and nucleotide sequencing. 15 Specimens that could not be genotyped were retested by EIA and real-time RT-PCR assay at the CDC. Any confirmed positive result led to a testpositive case designation. Specimens failing all CDC confirmatory analyses (<1%) were deemed to have indeterminate test results and were removed from the analytical data set.
Cases were defined as children with AGE symptoms who were hospitalized or treated in the ED and had a confirmed positive test result for rotavirus in the stool specimen during that AGE episode.
Data from cases were compared with those from children with AGE whose specimens had negative test results for rotavirus by EIA (controls).

Vaccine Effectiveness Analyses
Rotavirus immunization status was verified, as in previous publications, 11-13 by obtaining vaccine records from primary care physicians or health care professionals and/or regional immunization information systems. Vaccine doses were defined as valid if given at least 14 days before onset of symptoms for the cases and controls. Children included in this study were required to be born on or after April 1, 2006, for RV5 analyses and on or after August 1, 2008, for RV1 analyses to ensure vaccine age eligibility following published ACIP recommendations. 1 We restricted analyses to cases and controls who had reached the maximum ACIP-recommended age for completion of the vaccine series within the recommended age window (maximum age for the last dose, 8 months 0 days) to control for residual confounding by age at the time of last dose for both vaccine types.
We limited this analysis to children younger than 5 years across the 7-year study period. A testnegative case-control design was used to estimate the vaccine effectiveness, with cases defined as children age eligible to receive rotavirus vaccine who had a stool specimen with confirmed results positive for rotavirus and controls defined as children whose stool specimens had confirmed results negative for rotavirus. Vaccine effectiveness was calculated using the formula ( considered by the ACIP to constitute a complete course. 1 We presented an aggregate vaccine effectiveness for any dose of either rotavirus vaccine type.

Statistical Analysis
Data were analyzed from November 1, 2009, through June 30, 2016. We compared demographic and socioeconomic data for cases and controls using Wilcoxon rank sum tests for continuous variables and χ 2 tests for categorical variables. Tests of statistical significance were 2-sided, and P < .05 was considered statistically significant. We assessed the clinical severity of AGE illnesses by calculating a modified Vesikari Severity Score (VSS) 12 with severity classifications of mild (score Յ10), moderate (score [11][12][13][14][15], and severe (score Ն16) by clinical setting.

Characteristics of Cases and Controls
Our vaccine effectiveness analysis included a total of 1193 laboratory-confirmed cases and 9620 controls, but did not differ by sex (Table). Vaccination status did not significantly differ by health insurance status, a potential proxy of health care-seeking behavior.

Distribution of Rotavirus Cases by Time, Clinical Setting, and Age
The introduction of rotavirus vaccines altered the historically annual peaks in US rotavirus incidence in favor of a new, biennial epidemiologic pattern of tempered incidence, 16 a phenomenon also observed in a multicenter network of pediatric hospitals 17 (Figure 1). Accordingly, the distribution of rotavirus cases in our vaccine effectiveness analysis also varied from year to year, with a higher incidence of cases occurring during odd-numbered years. During these odd-numbered years,

Vaccine Effectiveness for Those Receiving at Least 1 Dose of Rotavirus Vaccine
The 2010 to 2016 aggregate effectiveness of at least 1 dose of rotavirus vaccine to prevent a rotavirus-associated hospitalization or ED visit for AGE among US children younger than 5 years was 79% (95% CI, 76%-82%) (Figure 2). Stratified by clinical setting, any dose of rotavirus vaccine was 82% (95% CI, 77%-86%) protective against inpatient visits during these 7 years (Figure 1). Inpatient rotavirus vaccine effectiveness remained consistently high, ranging from 78% (in 2015) to 92% (in 2016), with no statistically significant downward trend detected over time for either rotavirus vaccine. Oscillations in inpatient vaccine effectiveness estimates were observed to occur biennially but were not statistically different, with 84% vaccine effectiveness in even-numbered years (95% CI, 76%-89%) compared with 81% vaccine effectiveness in odd-numbered years (95% CI, 73%-86%) (Figure 1).
Because the VSS classification incorporates a severity value for hospitalization, we calculated vaccine effectiveness estimates separately for each clinical setting, as well as combined inpatient and ED estimates (Figure 3). A consistent gradation in vaccine effectiveness by severity existed, with vaccine protection greatest for more severe infections. In the combined clinical setting, any dose of rotavirus vaccine was 65% (95% CI, 56%-73%) effective against mild infections, 81% (95% CI, 76%-84%) against moderate infections, and 91% (95% CI, 85%-95%) against severe infections.

Vaccine Effectiveness for a Mixed Course of RV5 and RV1
We analyzed vaccine effectiveness for a mixed series of RV5 and RV1 vaccines pursuant to the ACIP recommendation that a child's rotavirus vaccine series be completed with the same product whenever possible, 1 but allowing interchanging vaccine types if the product used for a previous dose(s) is not available or is unknown. Among children younger than 5 years, the vaccine effectiveness of these mixed courses was 86% (95% CI, 74%-93%) against inpatient and ED visits ( Figure 4). Aggregate vaccine performance for RV5 and RV1 was statistically indistinguishable, and mixed courses of both vaccines appeared to be highly effective. Previously published RV5 (3-dose) vaccine effectiveness results from the NVSN for the first (range, 85% 11 to 91% 12 ), second (range, 82% 12 to 90% 17 ), third (range, 83% 11 to 88% 12 ), and fourth (range, 76% 12 to 79% 11 ) years of life were comparable to our findings. Our RV1 findings for these age groupings have greater study power and further clarify these previously published results. Earlier RV1 vaccine effectiveness estimates were previously derived from smaller samples, reflecting the delayed uptake of RV1 coverage across the United States.

Discussion
We found that vaccine effectiveness for mixed-dose rotavirus vaccination course was 86% and was therefore potentially more protective, but this finding did not reach statistical significance.
However, when added to previously published estimates (vaccine effectiveness, 80%), 21 this further demonstrates that the estimated 100 000 US children who annually receive a mixed course of rotavirus vaccines are adequately protected. Although our data do not explain why a mixed course of rotavirus vaccines might be associated with higher overall protection than either vaccine alone, one hypothesis warranting further evaluation is whether receiving both vaccine types accrues specific immunologic benefits from each. Given the nonrandomized nature of our study, differences in estimated vaccine effectiveness might also be due to unmeasured, and thus unadjusted, confounding.
In longitudinal cohort studies conducted in Mexico City and Vellore, India, 22  Our study detected oscillations in inpatient and ED vaccine effectiveness estimates corresponding inversely with the biennial epidemiologic pattern of rotavirus incidence. In vaccine effectiveness analyses using pooled inpatient and ED data, these oscillations by clinical setting may be masked. Even-numbered years had the highest inpatient vaccine effectiveness (84% vs 81%), and odd-numbered years had the highest ED vaccine effectiveness (77% vs 71%) ( Figure 1). We hypothesize that children who receive a vaccination and acquire a rotavirus infection of diminished severity or at an older age may be less likely to be hospitalized and may instead be rehydrated in the

Limitations
This study has limitations. Some test-negative case-control studies are affected by case misclassification and by differential health care-seeking behaviors. We have attempted to minimize the misclassification of cases with rotavirus through extensive additional laboratory confirmations.
Our methods determined true rotavirus case status of children through EIA analyses, followed by confirmatory RT-PCR assays and nucleotide sequencing. With these laboratory confirmations, we estimate our laboratory results to have 100% sensitivity in detecting true rotavirus cases and 98.4% specificity, within the Rotaclone specificity estimate of 92% to 100%. 26 Previous analyses on a subset of these data 11 have indicated that the test-negative controls with AGE whom we used are the optimal control group for our design, despite the possibility of some residual differences in health care-seeking behavior.