Pediatric SARS-CoV-2 Incidence in California by Vaccination Status, December 2021 to April 2023

COVID-19, caused by SARS-CoV-2 infection, was the leading infectious and respiratory disease cause of death for children and adolescents aged 0 to 19 years in the United States between April 2020 and July 2022.¹ During the pandemic, 2 messenger RNA (mRNA) vaccines were granted emergency use authorizations (EUA) for administration in children and adolescents aged 6 months to 17 years: BNT162b2 (Pfizer/BioNTech) and mRNA-1273 (Moderna). Studies assessing their effectiveness among children and adolescents have shown 4 common trends. First, a reduction in vaccine effectiveness (VE) against documented infection over time since vaccination has been demonstrated across ages and vaccine products.^2–21 Second, VE has differed by SARS-CoV-2 variant, with a drop in VE against infection between the Delta and Omicron predominant periods.^{3–6,9,14,16,17,19,22} Third, a lower VE was observed for younger age groups who receive smaller doses compared to older age groups.^{4–9,12,17,23} Lastly, VE against severe disease outcomes after completion of a primary series exhibited less waning and less variation among age groups and SARS-CoV-2 variants.^{10–12,19,20,24,25} 

To gain insight into trends related to performance of COVID-19 vaccines in children and adolescents in California (particularly in those aged < 5 years, for whom there is less information available in the literature), we compared incidence rate ratios (IRRs) between vaccinated and unvaccinated children and adolescents in California given vaccine manufacturer and time since vaccination using immunization information systems (IIS) and confirmed case surveillance data. We analyzed the period between December 31, 2021, when the Omicron BA.1/BA.2 subvariants became heavily predominant (≥ 75% of SARS-CoV-2 isolates sequenced in California), through April 18, 2023, when the Advisory Committee on Immunization Practices recommended a simplified vaccination schedule for all ages using a bivalent COVID-19 vaccine. In doing so, we provide estimates of IRRs for children aged < 5 years who received mRNA vaccines, for whom there is less information available in the literature.^15,26,27 

California began the administration of COVID-19 vaccines on December 14, 2020,²⁸ and we obtained individual-level data on COVID-19 vaccine administration for individuals aged 1 to 17 years between December 14, 2020, and April 17, 2023, as well as case and hospitalization data from February 1, 2020, through April 17, 2023. Cases were defined as individuals with a positive SARS-CoV-2 nucleic acid amplification test regardless of symptom status. For analyses conducted using data from the Omicron BA.1/BA.2 predominant period, we adjusted for previous infection. However, increased prevalence of infections and availability of home antigen testing during the Omicron BA.1/BA.2 predominant period and the subsequent Omicron BA.4/BA.5/BQ/XBB predominant periods meant that previous infection became harder to capture, and many infections went unrecognized. Consequently, although we include person-time for individuals with previous infection, we did not adjust for it (see Supplementary Materials for an analysis excluding previous infection).

The vaccines considered in this study include BNT162b2, mRNA-1273, and their respective bivalent (original strain plus Omicron BA.4/BA.5) formulations. Detailed information on dose, date of EUA (timing of availability for different age groups), and targeted age group are described in Table 1. Although the BNT162b2 and mRNA-1273 vaccines were approved for children aged 6 months and older on June 17, 2022, the population data that underpin the unvaccinated person-time estimates are only available in 1-year increments, so our analyses are limited to individuals aged 1 to 17 years. More information on data sources is available in the Supplementary Materials.

We assessed incidence among vaccinated compared to unvaccinated individuals for all analyses, allocating person days at risk to the vaccinated and unvaccinated groups by date of vaccination and infection episode date. We tracked vaccination status as unvaccinated, partially vaccinated (1 dose of an mRNA vaccine), completed primary series (2 homologous doses of any mRNA vaccine for those aged 5+ years or those aged 1 to 4 years receiving mRNA-1273 or 3 doses for those aged 1 to 4 years receiving BNT162b2), monovalent boosted (received 1 or more additional monovalent doses of any mRNA vaccine after completion of a homologous primary series), and bivalent boosted (received a bivalent booster dose of any mRNA vaccine after completion of a homologous primary series, regardless of whether or not a monovalent booster dose was also administered). Episode date is a hierarchical variable that is defined as the earliest of date of report, date of diagnosis, date of symptom onset, date of death, and date of specimen collection.

Cases and vaccinated individuals were excluded from the analysis if they received a heterologous primary series (combination of BNT162b2 and mRNA-1273 doses) or received a booster dose not in accordance with the authorized vaccine schedule. Although heterologous vaccine priming (mix-and-matching primary vaccine doses between different vaccines) was internationally supported in scenarios in which homologous priming would be difficult to efficiently achieve during the study period, it was not supported by clinical guidance in the United States.²⁹ We did not exclude doses received before the EUA dates for each given vaccine and age group.

Age for vaccinated individuals was determined by age at most recent vaccine dose received for the calendar date assessed. Age for unvaccinated individuals was based on date of birth and episode date (if a case) and 2020 California Department of Finance population estimates (for non-cases). We subtracted the number of vaccinated individuals (cases and non-cases) and unvaccinated cases for each 1-year age strata from the total yearly population size to obtain an estimate of the number of non-case unvaccinated children and adolescents each calendar date. Once an individual received a vaccine at ≥ 1 year of age, they were included in the study. Individuals who received the dose being evaluated at ≥ 18 years of age were censored. We excluded individuals who (1) had reported residency outside of California, (2) received vaccines other than monovalent or bivalent BNT162b2 or mRNA-1273, or (3) had fewer than 15 days between receipt of their first and second doses.

We present our main results by Omicron subvariant period (BA.1/BA.2 and BA.4/BA.5/BQ/XBB) as well as by age group and vaccine manufacturer. When assessing BNT162b2, the age groupings are 1–4 years, 5–11 years, and 12–17 years. For any mRNA vaccine, we assess 1–4 years, 6–11 years, and 12–17 years. Children aged 5 years are excluded from analyses combining both vaccines, as they received the second smallest dose among BNT162b2 doses but the smallest dose among mRNA-1273 doses. The 2 vaccines are assessed separately, and 5 year olds are included in their respective vaccine schedule age groupings in the Supplementary Materials. We also present a supplementary analysis for the Omicron BA.1/BA.2 period comparing 11 year olds to 12 year olds (Supplementary Materials Figure S2) to determine whether patterns seen in the broader age groups were reflected for these specific ages; these adolescents, although similar in age, received different doses for BNT162b2 and mRNA-1273 vaccines (10 μg vs 30 μg for BNT162b2 and 25 μg vs 50 μg for mRNA-1273; see Table 1).

We used a quasi-Poisson generalized additive model with an offset equal to the natural logarithm of person-time to obtain the adjusted IRR for each age group using the R package mgcv. The number of adjustment variables depended on a sufficient case or hospitalization and person-time data for each stratum. Smoothing with a cubic regression spline basis was applied to age and calendar month (months since December 12, 2021) with a basis dimension of 3. If there were only 2 possible values for month, the month variable was fitted without smoothing. IRR estimates were considered too imprecise to report if there were no cases in the analysis or if the 95% CI was wider than 10. CIs were constructed using the delta method, and analyses were performed using R version 4.2.3.

This study received a non-research determination from the California Committee for the Protection of Human Subjects.

During winter 2021 to 2022, the Omicron SARS-CoV-2 variant (BA.1/BA.2 subvariants) became predominant in California (Figure 1A), and its emergence was associated with large surges in reported cases. The subsequent introduction of BA.4/BA.5 subvariants in the summer of 2022 led to persistently high case numbers (Figure 1C and 1D), even as increased availability of home antigen testing led to a decline in the proportion of infections reported to local public health authorities.³⁰ 

Omicron BA.1 and BA.2 subvariants were predominant (≥ 75% of all sequences) in California between December 31, 2021, and June 9, 2022 (Figure 1A). During this period, case rates among children and adolescents were the highest of the pandemic in California, reaching a peak of 423 cases per 100 000 population per day among those aged 12 to 17 years (Figure 1C, limited to cases included in the analysis dataset; see Methods). Case rates were lowest for 1 to 4 year olds until June 2022, coinciding with the end of widespread school-based testing for 5 to 17 year olds (Figure 1B).³¹ In contrast, during the BA.1/BA.2 predominant period, hospitalization rates were as high for 1 to 4 year olds as for 12 to 17 year olds and lowest for 5 to 11 year olds (Figure 1D), and unvaccinated case rates among all age groups were high (Figure 1E).

Rates of documented infection were lower for adolescents aged 12 to 17 years who completed a BNT162b2 primary series compared to unvaccinated adolescents aged 12 to 17 years when adjusted by age, sex, testing volume, previous infection, and calendar month. Case IRRs by time since vaccination ranged from 0.09 (95% CI, 0.08–0.10) at 14–89 days after completion of primary series to 0.16 (95% CI, 0.16–0.17) at 180–269 days after completion of primary series for 12 to 17 year olds (Figure 2). Overall case IRRs were 2.5 times higher for children aged 5 to 11 years when compared to 12 to 17 year olds ≥ 14 days after completing a BNT162b2 primary series, and the case IRRs for 5 to 11 year olds ranged from 0.31 (95% CI, 0.30–0.33) at 14–89 days after vaccination to 0.76 (95% CI, 0.69–0.84) at 90–179 days after vaccination. Hospitalization IRRs for 5 to 11 year olds were more similar to those of 12 to 17 year olds than case IRRs, at 0.24 (95% CI, 0.17–0.32) compared to 0.12 (95% CI, 0.10–0.14) for ≥ 14 days since completion of primary series. Vaccines were not available to children under the age of 5 years during the BA.1/BA.2 predominant period.

We additionally compared individuals aged 11 and 12 years as they received different doses (10 μg vs 30 μg; see Table 1) for BNT162b2.²³ Overall case IRRs for vaccinated compared to unvaccinated 11 year olds were 2.4 times higher compared with 12 year olds ≥ 14 days after completing a BNT162b2 primary series when adjusting for sex, testing volume, previous infection, and calendar month (Supplementary Materials Figure S2). Trends in IRRs for these ages reflected the results for ages 5 to 11 years and 12 to 17 years (Figure 2). IRRs of cases for 11 year olds increased with time since completion of primary series, from 0.29 (95% CI, 0.26–0.32) at 14–89 days after completion of primary series to 0.61 (95% CI, 0.48–0.79) at 90–179 days following completion of their primary series, whereas IRRs for 12 year olds remained relatively stable between 0.10 (95% CI, 0.09–0.11) at 14–89 days after completion of primary series and 0.18 (95% CI, 0.15–0.21) at 270–359 days following completion of their primary series (Supplementary Materials Figure S2). Likewise, hospitalization IRRs for 11 year olds were similar to those of 12 year olds, at 0.18 (95% CI, 0.08–0.37) compared to 0.14 (95% CI, 0.09–0.22) for ≥ 14 days since completion of primary series.

BA.4/BA.5 became predominant (≥ 75% of all sequences) in California July 5, 2022, but by December 15, 2022, predominance had shifted to BQ/XBB subvariants. During the period of BA.4/BA.5/BQ/XBB predominance, case rates drastically decreased for all age groups compared to the BA.1/BA.2 predominant period (Figure 1A and 1C), with 7-day average mean case rates per 100 000 population of 9.5 (range: 1.5–32.7) among children aged 1–4 years, 7.4 (range: 0.9–26.0) among children aged 5–11 years, and 7.6 (range: 1.1–24.2) among children aged 12–17 years.

Between June 17, 2022, and April 17, 2023, when children under 5 years of age were eligible for vaccination, case IRR was 1.10 (95% CI, 0.98–1.24) among vaccinated compared to unvaccinated 1 to 4 year olds who received any mRNA vaccine at ≥ 14 days since completion of primary series. In contrast, estimated vaccinated compared to unvaccinated case IRR was 0.58 (95% CI, 0.54–0.62) for 6 to 11 year olds and 0.14 (95% CI, 0.13–0.15) for 12 to 17 year olds at ≥ 14 days since completion of primary series during the same time period (Figure 3). Overall hospitalization IRR at ≥ 14 days since completion of primary series for vaccinated compared to unvaccinated children and adolescents remained lower than case IRR for each age group, at 0.32 (95% CI, 0.17–0.61) among 1 to 4 year olds, 0.31 (95% CI, 0.24–0.41) for 6 to 11 year olds, and 0.11 (95% CI, 0.08–0.13) for 12 to 17 year olds at ≥ 14 days since completion of primary series. There were fewer hospitalizations during this period compared to the BA.1/BA.2 period. Results for bivalent booster case and hospitalization IRR are available in the Supplementary Materials.

Children and adolescents (< 18 years of age) are at risk of severe SARS-CoV-2 infection and its complications, including multisystem inflammatory syndrome in children, long COVID, and death.^{21,24,32–34} Evidence indicates that it is safer to obtain immunity against SARS-CoV-2 through vaccination compared to infection,³⁵ and our study of COVID-19 case and hospitalization rates among children and adolescents by vaccination status in California supports the use of vaccines to prevent severe COVID-19 disease in this population, although vaccination was less protective against infection for children aged ≤ 11 years compared to adolescents aged 12 years and older.

Our results are consistent with studies that found differences in outcomes between adolescents who received the same antigen dose as adults and children who received a lower dose and experienced greater increases in hazard of infection over time since vaccination.^7,23 We also provide estimates of IRRs for children aged < 5 years who received mRNA vaccines, for whom there is less information available in the literature. Despite higher case IRRs for young children compared to adolescents, the point estimates for hospitalization IRRs remained substantially < 1 across all ages and variant periods. However, in the Omicron BA.4/BA.5/BQ/XBB period, the 95% CIs for young children sometimes exceeded 1 due to the low analysis power from fewer total hospitalizations reported. These findings are consistent with previous studies that demonstrated greater vaccine effectiveness against severe disease than documented infection (Supplementary Materials Figure S1). While we noted increasing case IRR by time since completion of primary series for 5 to 11 year olds during the BA.1/BA.2 predominant period (Figure 2), this pattern was not as clearly replicated during the BA.4/BA.5/BQ/XBB predominant period (Figure 3). Given the higher case IRR and increases in case IRR over time since vaccination among younger age groups, additional studies may help inform future vaccine recommendations.

This study relied on public health surveillance data; complementary clinical data for vaccine recipients and unvaccinated individuals were not available, and thus, there was a limited ability to control for potential sources of confounding. For example, eligibility for vaccination was determined by age and did not include considerations such as medical contraindications or comorbidities, both of which may influence likelihood of vaccination and disease risk. (See Supplementary Materials for an additional analysis quantifying the potential effect of unmeasured confounding using E-values). Individuals who received a third dose as a booster or as part of their primary series due to being immunocompromised were not identifiable as such in the vaccine registry, so all vaccine recipients were assumed to have received doses in accordance with the general vaccine schedule. Additionally, we did not examine the interval between administration of the first and second primary series doses, so the effects of dose timing were not evaluated and remain an area for further study.^36–38 Other potential confounders, such as unobservable characteristics underlying the decision by parents and guardians to vaccinate their children, differences in home testing practices, adoption of nonpharmaceutical interventions by children and adolescents and their households, and access to health care, may vary by age and may have contributed to differences in estimated IRRs between age groups.

Other limitations include the potential for misclassification of the exposures, outcomes, and covariate status. Laboratory-confirmed polymerase chain reaction–positive tests represent a subset of all SARS-CoV-2 infections, and many infections are not detected or go unreported to public health. Case ascertainment for those aged 0 to 17 years was estimated to be 12.7% (95% CI, 10.9%-14.3%) in California in May 2021³⁹ and is likely to have further decreased during the study period, especially as at-home antigen tests became more readily available.³⁷ Although it is difficult to disentangle changes in case ascertainment over time, one source of confounding may be testing volume, as fewer tests lead to fewer opportunities to identify infections. Statewide testing volume was used as an adjustment in our analyses to address this issue. In California, testing volume markedly decreased in June 2022, coinciding with the end of school-based testing for 5 to 17 year olds. Differences in testing policies and practices between unvaccinated and vaccinated individuals may have additionally resulted in differential ascertainment of cases between these 2 groups.^40–43 If misclassification of infection status is nondifferential and independent of the vaccination status, then the resulting estimates would be biased toward the null, but if propensity for testing is associated with vaccination status, then the impact on our estimates is uncertain. Such misclassification could be a contributor to the age-related case IRR differences observed in this study. In comparison, hospitalization IRRs are likely to be more reliable than case IRRs, since severe outcomes are more likely to be reported than mild infections and are therefore less susceptible to misclassification bias.

We estimated low hospitalization IRRs regardless of vaccine dose received, supporting the use and recommendation of COVID-19 vaccines to prevent severe COVID-19 among children and adolescents in California. Low COVID-19 vaccine coverage rates among all pediatric age groups resulted in missed opportunities for protection against both infection and severe disease. Continued emphasis on the use of layered protections against SARS-CoV-2 infection, including vaccination and nonpharmaceutical interventions, is important for managing SARS-CoV-2 transmission and reducing infection risk, with the latter providing the additional benefit of protection against other pathogens.^44,45 Although further optimization of vaccine schedules and dosing for children may provide additional protection against infection, current age-based dosing confers sustained protection against hospitalization.

The authors thank the California local health jurisdictions, laboratories, and hospitals as well as the California Department of Public Health Coronavirus Science Branch and Immunization Branch. In addition, we specifically thank former and current members of the California Department of Public Health Modeling Team, including Chris Hoover, Ryan McCorvie, Mugdha Thakur, Lauren White, Phoebe Lu, Sindhu Ravuri, and Sophie Zhu for conversations and insights that improved these analyses.

EUA

emergency use authorization

IRR

incidence rate ratio

mRNA

messenger RNA

VE

vaccine effectiveness