Pediatric Infection-Induced SARS-CoV-2 Seroprevalence Increases and Seroprevalence by Type of Clinical Care—September 2021-February 2022

Abstract Background and Objectives Trends in estimates of US pediatric SARS-CoV-2 infection-induced seroprevalence from commercial laboratory specimens may overrepresent children with frequent healthcare needs. We examined seroprevalence trends and compared seroprevalence estimates by testing type and diagnostic coding. Methods Cross-sectional convenience samples of residual sera between September 2021 and February 2022 from 52 U.S. jurisdictions were assayed for infection-induced SARS-CoV-2 antibodies; monthly seroprevalence estimates were calculated by age group. Multivariate logistic analyses compared seroprevalence estimates for specimens associated with ICD-10 codes and laboratory orders indicating well-child care with estimates for other pediatric specimens. Results Infection-induced SARS-CoV-2 seroprevalence increased in each age group; from 30% to 68% (1-4 years), 38% to 77% (5-11 years), and 40% to 74% (12-17 years). On multivariate analysis, patients with well-child ICD-10 codes were seropositive more often than other patients aged 1-17 years (adjusted prevalence ratio [aPR] 1.04; 95% CI 1.02-1.07); children aged 9-11 years receiving standard lipid screening were seropositive more often than those receiving other laboratory tests (1.05; 1.02-1.08). Conclusions Infection-induced seroprevalence more than doubled among children under 12 between September 2021 and February 2022, and increased 85% in adolescents. Differences in seroprevalence by care type did not substantially impact US pediatric seroprevalence estimates.


A C C E P T E D M A N U S C R I P T
Infection-induced SARS-CoV-2 seroprevalence increased in each age group; from 30% to 68%

18
(1-4 years), 38% to 77% (5-11 years), and 40% to 74% (12-17 years). On multivariate analysis, 19 patients with well-child ICD-10 codes were seropositive more often than other patients aged 1-20 17 years (adjusted prevalence ratio [aPR] 1.04; 95% CI 1.02-1.07); children aged 9-11 years 21 receiving standard lipid screening were seropositive more often than those receiving other 22 laboratory tests (1.05; 1.02-1.08). The COVID-19 pandemic, caused by SARS-CoV-2, has taken a tremendous toll in the United 10 States; among children, there have been over 15 million reported cases and 1,794 deaths as of 11 October 10, 2022 [1]. The burden has been quantified in a number of ways, mostly focusing on 12 case rates, hospitalizations, and reported deaths. However, these methods do not capture 13 infections that were asymptomatic, not diagnosed, or not reported. The resulting under-capture of 14 infections is particularly pronounced among the pediatric population, who are more likely to 15 have asymptomatic or mild disease compared with older age groups [2]. 16 17 The national commercial laboratory seroprevalence study is a repeated, cross-sectional national where seroprevalence rates for the children aged 0-17 years are reported for the U.S. and for each 3 jurisdiction with sufficient sample size. High seroprevalence among children as compared to 4 other age groups had occurred prior to the study period for the present analysis; national 5 seroprevalence estimates for children aged 0-17 years were significantly higher than all adult age 6 groups between February 1 and July 11, 2021[3]. Because the national commercial laboratory seroprevalence study tests residual serum specimens 9 that had been sent for clinical laboratory testing, the sample is more likely to include children 10 with more frequent medical visits and increased healthcare access, which could potentially bias 11 national seroprevalence estimates through over-representation of children with chronic or 12 immunocompromising conditions. Since general exams and preventive or health maintenance 13 services [hereafter 'well-child care'] are recommended for all children, seroprevalence among 14 the pediatric population presenting for well-child care would likely be more representative of 15 general pediatric seroprevalence than that among those accessing healthcare for other reasons. 16 17 This study uses the ICD-10 codes and laboratory tests associated with pediatric specimens to 18 explore potential biases of using commercial lab specimens to study seroprevalence of SARS-

19
CoV-2 anti-N antibodies in this age group. ICD-10-CM codes corresponding with well-child 20 care and lipid screenings among the recommended[4] age group (9-11 years) were used as 21 proxies to identify well-child visits.
pediatric age groups corresponding to current pediatric vaccination recommendations and 2 potential future vaccination considerations, and then compare the seroprevalence among 3 specimens most likely to represent well-child care to that of other specimens. Study design 7 The national commercial laboratory seroprevalence study included a convenience sample of 8 blood specimens submitted to commercial laboratories for routine, clinical screening, or 9 diagnostic testing between September 2021 and February 2022. Specimens for each state were 10 selected by the designated commercial laboratory network, which used an internal algorithm to 11 select specimens to optimize geographic distribution across the jurisdiction and a roughly 12 equivalent sample size across five age groups (≤11, 12-17, 18-49, 50-64, and 65 years).

13
Seroprevalence estimates for each monthly data collection period were adjusted for year of age, 14 sex, and metropolitan status of county of residence using raking[5] to create weights based on 15 data from the 2018 American Community Survey 5-year population totals [6] and bootstrap 16 resampling [7] to create confidence intervals. Differences in seroprevalence were considered to 17 be statistically significant where confidence intervals did not overlap. Additional details of this 18 methodology are published elsewhere [8.9]. The data associated with each specimen result 19 included patient age, sex, jurisdiction and county of residence, specimen collection date, 20 laboratory tests ordered, and associated ICD-10-CM codes from the clinical encounter. Data on 21 race, ethnicity, and vaccination status were not available. To decrease bias, specimens for which 22 SARS-CoV-2 antibody testing had been ordered by the clinician were excluded from the study  The distributions of the most common ICD codes associated with pediatric specimens were 9 examined with descriptive methods. Two analyses were conducted to assess whether pediatric 10 seroprevalence estimates produced by this study are similar between children seeking well-child 11 care and those seeking other types of care. First, we defined children being tested for well-child 12 care visits as those with one or more of the following ICD-10-CM codes: Z00 (General 13 examination without complaint), Z71 (Encounter for health service, not elsewhere classified), 14 and Z23 (Encounter for Immunization). We compared the infection-induced seroprevalence of 15 specimens associated with well-child care visits to that of other pediatric specimens. Specimens 16 that were associated with codes for other health concerns or conditions in addition to general 17 health codes were included in the comparison group, since the laboratories drawn could have 18 been for a reason besides well-child care, such as monitoring of a chronic condition. These 19 groups were compared in an unweighted chi-square analysis, followed by a multivariate Poisson 20 regression with robust variance adjusted for jurisdiction, sex, and metropolitan status. The same 21 multivariate model was then applied to each pediatric age group. These analyses excluded 22 specimens from children residing in one jurisdiction who did not submit pediatric specimens and those in which more than 50% of pediatric specimens were submitted with no associated ICD-1 10-CM codes (n=8). Specimens with no associated ICD-10-CM codes in the remaining 2 jurisdictions (n=5,338, representing 7% of pediatric specimens) were excluded from this 3 analysis.

5
The second analysis compared seroprevalence among children aged 9-11 who were getting a 6 lipid screening test to those whose specimens were associated with other types of testing. A lipid 7 screening for all children aged 9-11 years is strongly recommended by the National Heart, Lung, 8 and Blood Institute[4] and the American Academy of Pediatrics. Similar to the previous 9 analysis, an unweighted analysis was followed by a multivariate Poisson regression with robust 10 variance adjusted for jurisdiction, sex, and rural/urban status. Children with an ICD-10-CM code 11 for hyperlipidemia (n=603) were excluded from both groups since lipid panels in this group may 12 have been conducted due to monitoring of a chronic condition rather than as a well-child   The largest increase in national seroprevalence for each age group was seen during a period 3 corresponding with the surge in cases due to the Omicron variant. Between December 2021 and 4 February 2022, the seroprevalence among children aged 1-4 years increased from 33% (95% CI: 5 30-37%) to 68% (95% CI: 63-72%), seroprevalence among children aged 5-11 years increased 6 from 47% (95% CI: 45-49%) to 77% (95% CI: 75-79%), and seroprevalence among adolescents 7 aged 12-17 years increased from 46% (95% CI: 44-47%) to 74% (95% CI: 73-75%) ( Figure 1). 8 Examining relative increases between December 2021 and February 2022, seroprevalence 9 among children aged 1-4 years increased by 106%, while seroprevalence in children aged 5-11 10 years and adolescents aged 12-17 years increased by 64% and 63%, respectively.

12
The precision of the estimate varied by age group. Over the study period, the confidence 13 intervals for the seroprevalence estimates for the children aged 5-11 years and adolescents aged age groups was Z001 (Encounter for pediatric health examination), which was recorded for 18% 2 of visits associated with specimens among children aged 1-4 years, 20% of visits for children 3 aged 5-11 years, and 14% of visits for adolescents aged 12-17 years. Other ICD-10-CM codes 4 varied by age group, but most corresponded to common pediatric complaints or concerns seen in 5 outpatient general practice (supplemental table 1). Laboratory tests associated with each group 6 were similar (supplemental table 2). In unweighted analysis, the infection-induced 7 seroprevalence among pediatric patients whose specimen was drawn in association with well-8 child care codes exclusively (n=6279; 53.1%) differed by only 1.2 percentage points from that of 9 pediatric patients whose specimens were associated with other codes (n=65,211; 51.9%; p= 10 0.052). Seroprevalence was between 1.0 and 2.3 percentage points higher among children 11 seeking well-child care in each pediatric age subgroup studied, though differences were not 12 significant: 1-4 years (46.2% and 43.9%, p=0.42), 5-11 years (54.9% and 53.8%, p=0.34), and 13 12-17 years (52.6% and 51.6%, p=0.25) ( Table 1). On a multivariate analysis adjusting for sex, 14 jurisdiction, and metropolitan status, pediatric patients with well-child care codes had a higher 15 prevalence of seropositivity compared with pediatric patients whose laboratory tests were 16 ordered in association with other ICD-10-CM codes (adjusted prevalence ratio [aPR] =1.04; 95% 17 CI 1.02-1.07). On analysis by pediatric age subgroup, the only age group with a significant 18 difference was among children aged 12-17 years (aPR= 1.05; 95% CI 1.01-1.08) ( Table 2). 19 While the prevalence ratio was above 1 for children aged 1-4 years and 5-11 years, the sample 20 sizes were smaller and the results were not significant.  (supplemental table 3). On analysis of pediatric patients aged 9-11 years without 12 hyperlipidemia on whom a lipid screening test had been ordered, seroprevalence (n=5224; 13 57.9%) was higher than that among the comparison group (n=6301; 52.4%; p<0.001) ( Table 3).
14 On a multivariate analysis adjusting for sex, jurisdiction, and metropolitan status, those whose 15 blood specimen was associated with a routine lipid screening had higher prevalence of 16 seropositivity for infection-induced antibodies compared with those whose blood was drawn for 17 other types of laboratory tests (aPR=1.05; 95% CI 1.02-1.08) ( Table 4). Overall, children whose laboratory tests were ordered in association with well-child care codes 12 were slightly more likely to test positive for infection-induced SARS-CoV-2 antibodies 13 compared with those whose laboratory tests were associated with other ICD-10-CM codes.
14 Adolescents (12-17 years) presenting for well-child care were also significantly more likely to 15 test seropositive; the lack of a significant difference in the two younger age groups may be 16 related to the smaller sample sizes in these groups. Children aged 9-11 years who received a lipid 17 screening test had significantly higher seroprevalence than other children in this age group. One  This study has six main limitations. First, we were restricted to an internal comparison due to 11 the lack of a nationally representative serosurvey. The population studied may still differ from 12 the general population in terms of access to medical care, frequency of care-seeking, and other 13 factors, as most pediatric well-child care visits outside the 9-11 year age group do not require lab 14 draws in the absence of clinical concerns. Second, it is difficult to understand the full clinical 15 picture of a visit based on ICD-10-CM codes alone, as codes are occasionally omitted in error. 16 This may have resulted in misclassification of some visits as well-child care as opposed to other 17 types of care. Third, there are no race or ethnicity data available for specimens, and therefore we 18 could not stratify or control for these important demographic variables, which are known to be Taken together, these findings indicate that the high seroprevalence estimates among pediatric 15 age groups in the national seroprevalence study likely approximate seroprevalence in the U.S. 16 pediatric population seeking well-child care. There are only small differences between children 17 likely to be seeking care for general exams and preventive or health maintenance services as 18 compared with those seeking services due to acute symptoms or chronic conditions, indicating    robust variance used to estimate association included jurisdiction of residence, sex, and 7 urbanicity of residence as covariates.  specimens collected by commercial laboratories from patients aged 9-11 years. 6 Multivariate Poisson regression with robust variance used to estimate association 7 included jurisdiction of residence, sex, and urbanicity of residence as covariates. 8 § Comparison between seroprevalence of specimens associated with lipid screening 9 panels and specimens associated only with other laboratory tests