COVID-19 Vaccination and Incidence of Pediatric SARS-CoV-2 Infection and Hospitalization

Key Points Question Was implementation of the pediatric COVID-19 immunization program of California associated with reductions in the reported pediatric COVID-19 incidence and hospitalizations? Finding In this case series including 3.9 million children, pediatric vaccination was estimated to avert 146 210 cases of COVID-19 among adolescents aged 12 to 15 years during a 141-day postvaccine evaluation period and 230 134 cases among children aged 5 to 11 years during a 199-day postvaccine evaluation period. In addition, an estimated 168 hospitalizations were averted among children aged 6 to 59 months during a 225-day evaluation period. Meaning The findings of this study suggest that vaccination against SARS-CoV-2 was associated with significant reductions in COVID-19 incidence and hospitalizations among children in California.


Introduction
Vaccination is among the most important interventions to reduce the public health impact of infectious diseases. 1SARS-CoV-2 mRNA vaccines, including mRNA1273 (Moderna) and BNT162b2 (Pfizer BioNTech), were approved for adult use in December 2020. 2 On May 10, 2021, the first mRNA COVID-19 vaccine was approved for use in adolescents aged 12 to 15 years.Vaccines were subsequently approved for children aged 5 to 11 years on October 29, 2021, and for children aged 6 to 59 months on June 17, 2022 (Figure 1). 2 COVID-19 vaccines are safe for children. 3However, concerns over vaccine-related adverse events, lower vaccine effectiveness against illness in children, and perceptions of a milder disease course in children have resulted in high rates of parental vaccine hesitancy [4][5][6] and resistance to pediatric vaccine mandates. 7While California has among the highest rates of vaccination in the US, 8 pediatric vaccination coverage lags that of adults substantially, with only 8.2% of children younger than 5 years and 37.8% of children aged 5 to 11 years fully vaccinated as of May 2023. 8Severe manifestations of COVID-19 are rare among children, but can occur. 9Understanding the populationlevel impact of COVID-19 vaccinations in SARS-CoV-2 infections and hospitalizations in pediatric populations would aid in public health decision-making on pediatric vaccine and booster policy and provide pediatric-specific information on vaccine outcomes that could be applied to future SARS-CoV-2 variants.
Herein, we analyze data on 3 913 063 pediatric cases of COVID-19 and 12 740 hospitalizations from California.Using the phased introduction of the vaccine to individuals aged 12 to 15 years, 5 to 11 years, and 6 to 59 months, we estimated statewide and county-specific outcomes associated with vaccination on pediatric incidence and hospitalizations in California.

Methods Epidemiologic Data
We obtained deidentified information on all pediatric COVID-19 cases reported in California between April 1, 2020, and February 27, 2023, from the California COVID-19 Reporting System, along with the patient's county of residence, age, and hospitalization status.Each case was confirmed using a nucleic acid amplification test.Because the research constitutes a public health surveillance activity, the study did not constitute human research and does not require institutional review board review or exemption according to the Common Rule (45 CFR §46).We followed the reporting guideline for case series studies.
Daily cases were aggregated by county and age groups based on dates of vaccination eligibility: 0 to 5 months (vaccine ineligible), 6 to 59 months, 5 to 11 years, 12 to 15 years, and older than 16 years (nonpediatric).To remove variation due to differential health care seeking by day of week, we calculated 7-day moving averages of case counts per county and age group.Due to small counts for pediatric hospitalizations, we aggregated hospitalizations by week and age group within 5 Californiadesignated regions (eFigure 1 in Supplement 1).Descriptions of other covariate data are covered in the eMethods in Supplement 1).

Training and Prediction Periods
For each age group of interest (6-59 months, 5-11 years, and 12-15 years), we split data into age-eligible and age-ineligible periods.The prevaccine eligibility period encompassed data up to the date of vaccine eligibility.The evaluation period for the outcome associated with the vaccine lasted from 1 month following the date of vaccine eligibility (to allow time to complete 2 doses) until the date that the next age group became eligible or until the time of analysis (February 27, 2023) (Figure 1).Ending the evaluation period at the vaccine eligibility date of the next eligible age group permitted that age group to be selected as a control time series in our predictive models.

Candidate Model Generation and Selection
2][13] For each area-age group combination, we selected the model with the lowest out-of-sample mean square error across holdout folds.For this model, we also calculated the coefficient of determination, r 2 , a goodness-of-fit metric.The selected predictors varied by area and age group.Selected models for each area-age group combination are included in eTable 2 (for cases) and eTable 3 (for hospitalizations) in Supplement 1.

Calculation of Vaccine Outcomes and Association Between Averted Cases and Vaccination Coverage
Selected models were fit to prevaccine eligibility data for their age group and geographic area and then used to predict counterfactual incidence or hospitalization in the postvaccine period or the expected case or hospitalization counts had vaccination not occurred.For inference, we computed 95% prediction intervals (PIs) around the counterfactual predictions, using a sandwich estimator to account for overdispersion when computing SEs (eMethods in Supplement 1). 14Prediction intervals, which are wider than CIs, capture the uncertainty around each future predicted value.Statewide estimates were obtained by summing predictions across geographies (eMethods in Supplement 1).
We estimated the absolute and relative differences between predicted counterfactual values and observed values for each county or region during the postvaccine evaluation period.To understand the association between vaccination coverage and averted cases, we fit regression models relating the reduction in cases within each age and county to county-level vaccination coverage within the same age group, using a fixed-effects meta-analysis with weights equal to the inverse estimated SE of the estimates per county.We used segmented regression models (eMethods in Supplement 1) to examine whether there were coverages below which reductions in cases could not be identified or above which diminishing returns on vaccination were observed. 15 examine whether postvaccine predictions from a different, but well predictive model, yielded similar estimates of vaccination outcomes, we repeated model selection using the mean absolute error instead of the mean square error in our cross-validation algorithm.We conducted jackknife analyses to examine whether postvaccine predictions from any one county were driving observed effects, dropping each county in turn from the overall pool of counties and recalculating the primary analytic end point of cases averted.
All analyses were conducted in R, version 3.6.0(R Foundation for Statistical Computing). 16

Descriptive Results
Between when model selection was done using mean absolute error as the loss function for children aged 5 to 15 years.Estimated averted cases in children aged 6 to 59 months were slightly lower using mean absolute error, but hospitalization results were consistent (eResults in Supplement 1).Estimates of cases averted (eTable 6 in Supplement 1) and hospitalizations averted (eTable 7 in Supplement 1) made using the mean absolute error as the loss function for each county or region are available, along with plots of observed and counterfactual case and hospitalization series for all geographic areas (eFigures 9-14 in Supplement 1).information criterion and bayesian information criterion values than segmented regression model fits for all age groups.Across all age groups, pediatric vaccination rates in California were generally highest among Bay Area counties (eFigure 8 in Supplement 1), which also ranked highest for averted cases due to vaccination (eResults in Supplement 1).

Discussion
We provide evidence that California's pediatric COVID-19 immunization program averted 376 085 (95% PI, 348 355-417 328) reported cases and 273 (95% PI, 77-605) hospitalizations among children aged 6 months to 15 years during the 4 to 7 months following vaccine availability.This represents a reduction of 26.3% of the number of cases that would have been seen in this population absent the vaccine.Prior work has similarly reported a high impact of widespread administration of mRNA vaccines in adult populations.In California, COVID-19 vaccines were estimated to avert more than 1.5 million cases, 72 000 hospitalizations, and 19 000 deaths statewide during the first 10 months of vaccination (through October 16, 2021). 17In the US, each 10% increase in vaccination coverage among individuals aged 18 years or older at the county level was associated with an 8% reduction in mortality and a 7% reduction in incidence. 18Similarly, a study in Israel estimated that nearly 650 000 cases of COVID-19 were averted in the first 2 months following vaccination introduction, 19 Earlier studies have estimated vaccine effectiveness in pediatric populations by comparing incidence rates among vaccinated children with those in unvaccinated children using test-negative designs, [20][21][22] or retrospective 23,24 or prospective cohort studies. 25Our counterfactual case series approach, which has been used in other studies to estimate the population-level impact of interventions with a clearly specified rollout time, 26,27 enables calculation of vaccine program impact at the population level, without information on individual vaccine status.
The cumulative effect of vaccination at the population level may be meaningful even if individual vaccine effectiveness is low.While influenza vaccine effectiveness was estimated at 29% in 2017-2018, 28 it was estimated that widespread vaccination averted more than 3.1 million cases of influenza in the US. 29Nevertheless, overall impact depends on vaccine coverage.We identified positive associations between county-level vaccination coverage and averted cases in each age group, whereby each 10 additional vaccinations per 1000 children corresponded to an average reduction of 0.9 to 3.5 cases per 1000 children.Segmented regression models associating vaccine coverage with averted cases did not identify break points, suggesting that over the range of vaccination coverages examined (0%-85%), we saw neither diminishing returns on increased coverage owing to the acquisition of sufficient population-level immunity nor a threshold below which vaccination has limited public health impact.This is consistent with the persistence of SARS-CoV-2 circulation in populations with high vaccination coverage and resulting value of direct protection.
Results for individuals aged 6 to 59 months differed from those of older age groups in that we found a significant reduction in hospitalizations, but not cases, following vaccination.One explanation for this discrepancy could be that postvaccine evaluation period for children aged 6 to 59 months did not include a surge in COVID-19 cases as it did for the other age groups (Figure 2), potentially making it difficult to detect statistically significant reductions from the counterfactual.However, vaccine effectiveness of early mRNA vaccines was lower against Omicron variants compared with Alpha and Delta variants, 30,31 and the Omicron variant dominated during the postvaccine period for children aged 6 months to 11 years (Figure 1).The detection of significant reductions in hospitalization in this age group, but not others, may be due, in part, to the fact that COVID-19 mortality disproportionately affects very young children compared with older children. 32r older age groups, we also estimated reductions in hospitalizations, although the 95% PI spans 0.
However, we note that 95% CIs are narrower than PIs and may not have encompassed the null.

Limitations
This study has limitations.Case data represented individuals who sought testing, which may be differential across unvaccinated and vaccinated groups, geographies, and time.Access to at-home testing likely resulted in further case underascertainment.If individuals were, on average, less likely to seek care for mild illness following vaccination, our analysis could have overestimated the absolute effect of the vaccine on cases averted.Overestimation of the relative effect of the vaccine may have resulted if vaccine recipients were disproportionately represented in the surveillance record both before and after vaccine eligibility compared with never-vaccinated individuals being more connected to care.Data on hospitalizations are less likely to be subjected to biases from differential case ascertainment.We estimated significant reductions in hospitalizations following vaccine introduction compared with counterfactual predictions.
Several considerations could lead to underestimates of the association between vaccination and child long-term health.4][35] Second, we were unable to estimate indirect outcomes associated with the vaccine in other age groups or control for social contacts.If children increased social contacts following receipt of the vaccine, as has been shown elsewhere, 36 they may be challenged more frequently with SARS-CoV-2.Third, we assessed the outcomes of the vaccine over a short postvaccination period, limiting our ability to examine vaccine responses under waning immunity.
Two important limitations relate to model functional form.First, attributing differences between the observed and the predicted counterfactual cases to the vaccine assumes that the associations between incidence in the age group being modeled and incidence in the age groups selected as model predictors would, absent the vaccine, remain constant over the pre-to-post vaccine periods.This would not occur if one age group developed increased immunity or if different variants had differential age-disease associations.This is especially salient for the 5-to 11-year age group, as the models were primarily trained on data from the period when the Delta variant predominated, yet the Omicron variant, which is less reliant on angiotensin-converting enzyme 2 binding for entry 37 and disproportionately influenced children younger than 5 years, prevailed in the evaluation period.Accordingly, the effect of vaccination may have been overestimated for this age group in counties where the incidence in children younger than 5 years was selected as a predictor (eTable 2 in the Supplement).
Second, there is potential for unstable predictions in the evaluation period if the predictive model was faced with values of selected predictors that fell outside the range of data used to fit the model.Our time series with a gap cross-validation approach guards against both of these limitations by prioritizing selection of generalized linear models that do well predicting values in periods that

Figure 1 .
Figure 1.Timeline of Major Circulating Variants and Vaccine Eligibility by Age Group

Figure 2 .
Figure 2. Comparison of Observed and Counterfactual Model Predictions for COVID-19 Cases by Age Statewide

JAMA Network Open | Pediatrics COVID
-19 Vaccination and Incidence of Pediatric SARS-CoV-2 Infection and Hospitalization

Attributable Averted Cases and Hospitalizations by Pediatric Age Group As
April 1, 2020, and February 27, 2023, a total of 3 913 063 COVID-19 cases were reported in shown in eFigure 3 in Supplement 1, r 2 values for models fit to daily case data were 0.92 (IQR, 0.79-0.96)for children aged 6 to 59 months, 0.89 (IQR, 0.78-0.95)for children aged 5 to 11 years, and 0.79 (IQR, 0.62-0.90)for adolescents aged 12 to 15 years.eFigure 4 in Supplement 1 shows the model fit for hospitalizations.More details on model fit are included in the eResults in Supplement 1.

Children Aged 6 to 59 Months Children
The postvaccine evaluation period for this age group did not include a surge in COVID-19 cases as it did for the other age groups (Figure2A).However, we estimated that 168 (95% PI, 42-324) hospitalizations were averted following vaccination, or a reduction of 24.4% (95% PI, 6.1%-47.1%)fromcounterfactualexpectations(TableandFigure3A).-19Vaccination and Incidence of Pediatric SARS-CoV-2 Infection and HospitalizationTable. Statewide Estimates of Averted COVID-19 Cases and Hospitalizations in California Due to Vaccination Among Children Aged 6 Months to 15 Years a 8ged 6 to 59 months were eligible for vaccination on June 17, 2022.By February 27, 2023, 177 087 (7.9%) individuals had received both doses of the primary series, with a range of 0.7% to 38.5% across counties.8Inthe225daysfollowingJuly17,2022,we did not detect any significant changes in cases from counterfactual expectations in the postvaccine period (estimated averted cases: −259; 95% PI, −1938 -1019) (Table).Sensitivity AnalysesAs indicated in the eResults and eFigures 5 and 6 in Supplement 1, results for individuals aged 5 to 15 years were not sensitive to the inclusion of any single county, although results for children aged 6 to 59 months were sensitive to the inclusion of Los Angeles (eFigure 7 in Supplement 1).Estimated cases (eTable 4 in Supplement 1) and hospitalizations (eTable 5 in Supplement 1) were consistentJAMA Network Open | PediatricsCOVIDaThe expected number of cases and hospitalizations for each age group is calculated according to the counterfactual predictions.
Predictors Eligible for Selection in Candidate Models for Each Age Group and Outcome Combination eTable 2. Best Predictive Models Selected for Each County-Age Group Combination Using Incidence as Outcome eTable 3. Best Predictive Models Selected for Each Region-Age Group Combination Using Hospitalization as Outcome eTable 4. Statewide Estimates of Averted COVID-19 Cases Due to Vaccination Among Children Aged 6 Months to 15 Years of Age Using Mean Absolute Error as the Loss Function JAMA Network Open | Pediatrics COVID-19 Vaccination and Incidence of Pediatric SARS-CoV-2 Infection and Hospitalization JAMA Network Open.2024;7(4):e247822.doi:10.1001/jamanetworkopen.2024.7822(Reprinted) April 23, 2024 13/14 Downloaded from jamanetwork.comby guest on 04/25/2024 eTable 5. Statewide Estimates of Averted COVID-19 Hospitalizations Due to Vaccination Among Children Aged 6 Months to 15 Years of Age Using Mean Absolute Error as the Loss Function eTable 6.Estimated Number of COVID-19 Cases Averted by County and Age Group During the Post-Vaccination Period eTable 7.Estimated Number of COVID-19 Hospitalizations Averted by Region and Age Group During the Post-Vaccination Period eFigure 1. Map of the Five Regions Used for Estimating the Effect of Pediatric SARS-CoV-2 Vaccines on Childhood Hospitalizations for COVID-19 eFigure 2. Schematic of the Time Series With Gap Cross-Validation Algorithm Used to Select the Best Predictive Algorithm During the Pre-Period eFigure 3. Model Performance of Selected Generalized Linear Model Predicting Cases During the Pre-Vaccine Period for Each County-Age Group Combination eFigure 4. Model Performance of Selected Generalized Linear Model Predicting Hospitalizations During the Pre-Vaccine Period for Each Region-Age Group Combination eFigure 5. Jackknife Analysis for Children 12-15 Years eFigure 6. Jackknife Analysis for Children 5-11 Years eFigure 7. Jackknife Analysis for Children 6-59 Months eFigure 8. Maps of Pediatric COVID-19 Cases Averted and County-Level Pediatric Vaccination Coverage in California.eFigure 9. County-Specific Model Predictions for COVID-19 Cases Among Children Aged 12-15 Years eFigure 10.County-Specific Model Predictions for COVID-19 Cases Among Children Aged 5-11 Years eFigure 11.County-Specific Model Predictions for COVID-19 Cases Among Children Aged 6-59 Months eFigure 12. Region-Specific Model Predictions for COVID-19 Hospitalizations Among Children Aged 12-15 Years eFigure 13.Region-Specific Model Predictions for COVID-19 Hospitalizations Among Children Aged 5-11 Years eFigure 14.Region-Specific Model Predictions for COVID-19 Hospitalizations Among Children Aged 6-59 Months