In this convenience sample of residual serum collected from Colorado children 1–17 years of age during spring/summer 2021, seroprevalence of SARS-CoV-2 IgG was 36.7%. Our results show higher seroprevalence than that found from rounds 20 and 21 of the nationwide antibody seroprevalence survey (https://covid.cdc.gov/covid-data-tracker/#national-lab), the only contemporaneous rounds with an estimate for seroprevalence among children, which estimated seroprevalence of 13.4% (95% CI 6.2%–23.8%) for round 20 and 17.5% (95% CI 12.4%–22.8%) for round 21 (20). A few factors may explain those differences: the nationwide surveys were conducted just before and during the first few weeks of our study, the nationwide survey used a different nucleocapsid assay and did not test for spike antibodies, and the relatively small pediatric populations in these convenience samples may demographically differ from those in our sample. In addition, although adjusted rates, such as those calculated in the nationwide survey, are comparable to other similarly adjusted rates across different populations, they may not be comparable to unadjusted rates such as those in our study. However, results for round 25, conducted a few months after the conclusion of our sample selection, was much more consistent with our results: seroprevalence of 40.1% (95% CI 33.7%–46.6%).

Among our study cohort, SARS-CoV-2 prevalence determined by positive molecular amplification testing was only 6.5% and seropositive case ascertainment was 16%. This finding indicates a potential underestimation of cases among this pediatric population by 84% when relying solely on confirmed case counts, either because of underascertainment or underreporting. If the unadjusted study results were generalized to the larger Colorado pediatric population, confirmed case counts statewide could require a multiplication factor of 6 to more accurately reflect the prevalence of SARS-CoV-2 infection among children. This result is consistent with findings from other states that showed multiplication factors from 4.7 to 8.9 during a similar time frame (21). Our results may represent opportunities to improve individual-level awareness of SARS-CoV-2 infection and increase mitigation measures, including isolation and quarantine. On a population level, our results can be used to corroborate modeled SARS-CoV-2 prevalence estimates based on hospitalization rates.

In addition, we found statistically significant differences in seroprevalence and seropositive case ascertainment across racial/ethnic groups. Compared with non-Hispanic White children, seroprevalence was significantly higher among Hispanic and non-Hispanic Black children, indicating higher rates of previous SARS-CoV-2 infection, consistent with results of several published studies on molecular test positivity and seroprevalence studies in adults (11,22–27). Our study also found that seropositive case ascertainment was lower among Hispanic and non-Hispanic Black children than among non-Hispanic White children; cases were detected in Hispanic children half as often as in non-Hispanic White children, and cases in non-Hispanic Black children were detected less than one third as often as in non-Hispanic White children. This disparity is supported by studies showing reduced rates of testing in many racial/ethnic minority groups (22,27). Our results show that reliance on reported case counts would lead to the erroneous assumption that rates of SARS-CoV-2 infection in the study period were highest among non-Hispanic White children. Those results further demonstrate the value of complementary surveillance mechanisms such as serologic and wastewater testing that can avoid testing and reporting biases.

Seroprevalence measured in our study reflects infections among children occurring during the 2020–21 school year and early summer before the peaks of the Delta and Omicron variants in Colorado. During that period, several school-based SARS-CoV-2 mitigation measures were mandatory in Colorado (e.g., masking and classroom cohorting), and in-person learning was interspersed with remote learning for intermittent school closures or hybrid learning models. Despite those measures and questions of whether children were at lower risk for infection with and transmission of SARS-CoV-2, compared with reported viral test results, our study shows that SARS-CoV-2 prevalence among children was higher. Those findings may have broad implications for estimating the severity of SARS-CoV-2 among children, levels of potential immunity, and the effectiveness of various mitigation measures in the pediatric population, and they may help guide future public health recommendations.

Figure 2

Figure 2. Boxplots showing number of previous tests, by racial/ethnic and age groups, among 829 children included in seroprevalence and case ascertainment analyses, Colorado, USA, May 12–July 13, 2021. NH, non-Hispanic.

Our study did not evaluate causal mechanisms for the observed racial/ethnic disparities in previous SARS-CoV-2 infection or seropositive case ascertainment. However, those findings could be explained by the existing literature, which suggests differences in risk for exposure to SARS-CoV-2 because of occupation or living setting, test-seeking behavior, underlying health and healthcare disparities, or differential barriers to viral testing (22,28). Despite extensive efforts at the federal, state, and local levels to improve access to free SARS-CoV-2 testing throughout Colorado during the study period, some barriers to access (e.g., time and transportation) may persist in some communities. In addition, differences in individual factors (e.g., perceived vulnerability and variability in access to childcare) could guide testing-seeking behaviors. The discrepancy between the number of seropositive children with any prior test (162, 53.3%; Appendix Table 3) and the number with previous positive test results (50, 16.4%) suggests that timing of viral testing or changes in access over time may also complicate case identification efforts. That 53% of seropositive persons had been previously tested but only 16% ever received a positive test result suggests that access to testing is necessary but not sufficient for identifying cases. The overall number of previous tests was skewed toward zero across all racial/ethnic and age groups; however, each group had outliers with high numbers of prior testing (Figure 2). The only group with a median number of tests >0 was non-Hispanic White children. To improve case finding within this population, greater knowledge of drivers for and access to viral testing is needed. Among children, for whom cases are often asymptomatic or mildly symptomatic, pursuing regular screening and postexposure testing may be needed to improve case ascertainment or better incorporate pediatric specimens in serosurveys to provide more information about disease prevalence. These considerations may also apply to other populations with largely asymptomatic illness, possible future SARS-CoV-2 variants that may cause fewer symptoms, or emergence of novel respiratory viruses with similar symptoms in children.

Results from this study are also consistent with literature suggesting that the SARS-CoV-2 spike IgG response may be more durable than the nucleocapsid IgG response because children who were spike IgG positive but nucleocapsid IgG negative were further from their positive test result at the time of serum sampling (29–34). As with any laboratory test, knowledge of the test characteristics, including seroreversion over time in the case of serology tests, is relevant for study design and interpreting results. The differential kinetics of IgG response after SARS-CoV-2 infection require further study, should inform serosurveillance efforts, and may complicate the ability to estimate SARS-CoV-2 prevalence among vaccinated populations.

Among the limitations of this study, the use of reported COVID-19 PCR and other molecular viral testing may miss some positive results from viral tests, such as antigen tests and testing performed out of state. However, during the study period, antigen tests were not widely available in Colorado, and molecular tests were the recommended test for schools. Free testing services provided by the state throughout Colorado used PCR testing, as did other large testing providers serving schools during the study period. Similarly, use of the Colorado Immunization Information System to determine vaccination status might lead to misclassification of some children as unvaccinated, although the time between specimen collection and data analysis should greatly limit the potential for missing vaccine data.

Second, unadjusted results from the study convenience sample may not be generalizable to other pediatric populations. Exclusion of previously vaccinated children could introduce some bias because only certain age groups were eligible for vaccination and vaccinated children may differ in relevant ways from their unvaccinated peers. Specifically, half of excluded vaccinees were non-Hispanic White children, and all were 12–17 years of age. However, we believe that exclusion of this group (Appendix Table 1) is likely to bias the sample toward the null and may lead to underestimation of the disparities noted. Although residence was not included in the study, this convenience sample may overrepresent children from the Denver-metro area, who account for a larger percentage of the patient population in this healthcare system, reducing generalizability. Because specimens were selected from children within a large pediatric healthcare system, this study cohort may represent children with greater access to care and families who are more likely to seek care, which could be associated with fewer barriers to testing. This population may also have a greater number of underlying health conditions, although our study sought to limit some of this bias by excluding children from subspecialty care clinics, which could lead to higher case ascertainment in this than in other populations. Similarly, SARS-CoV-2 positivity may be higher in the study cohort because symptomatic children seeking care and for whom providers ordered serum-based laboratory tests may be overrepresented, which could lead to higher seroprevalence.

Third, because of low numbers of members of certain racial/ethnic groups, including children in the non-Hispanic Asian, American Indian or Alaskan Native, Native Hawaiian or other Pacific Islander, >1 race, or other race categories, we were unable to analyze completely disaggregated racial/ethnic groups. This limitation is notable because there are known health disparities in some of the aggregated racial/ethnic groups, and a sample powered for analysis of these groups may identify similar disparities in SARS-CoV-2 serologic results and case ascertainment.

Last, seroconversion is an imperfect proxy for measuring disease prevalence because not all persons seroconvert after SARS-CoV-2 infection, and some literature suggests that rates of seroconversion may be lower or that antibody waning after infection may be more rapid for children or those with asymptomatic or mild illness (34–38). However, because some data on this issue are conflicting, the potential effect on our results is unclear (39). Similarly, the exact test characteristics over time for the assays used are not well known.

Strengths of the study include our ability to individually match serologic results to associated vaccination and reported test results, which enabled us to make direct comparisons and avoids ecological fallacy, which can arise when inferring information about individuals based on results from population-based data. Also, oversampling of children 1–4 years of age and Black children improved the representativeness of our sample across subgroups, which enabled a more nuanced analysis of racial/ethnic differences in serologic results and case ascertainment.

In conclusion, before the Delta variant peak, we found evidence of previous SARS-CoV-2 infection in more than one third of Colorado children 1–17 years of age based on seroprevalence, despite far fewer confirmed cases in this study population. Repeating this analysis after the Delta and Omicron variant peaks would be prudent. Racial/ethnic disparities identified in this pediatric population are consistent with trends identified among adults; higher seroprevalence, lower rates of reported viral testing, and lower seropositive case ascertainment were found among Hispanic and Non-Hispanic Black children than among non-Hispanic White children. Continued efforts to address racial/ethnic differences in illness burden and potential barriers to viral testing in pediatric populations are warranted.