Overview
A state-wide longitudinal cohort at the biweekly level during the period from 9/5/2021-6/26/2022 was created by combining: testing data and data about school populations collected by the Massachusetts Department of Elementary and Secondary Education (DESE), vaccination and town-level community COVID-19 case rates from the Massachusetts Department of Public Health12, and data from the National Center for Education Statistics (NCES)13 about school type and demographics. Consent rates for the testing programs were obtained from CIC Health6, the private public health company that partnered with the state to facilitate in-school testing during the 2021–22 academic year.
Key policy milestones that occurred throughout the academic year included: testing program availability at the start of the academic school year, vaccine availability to 5-11-year-olds on 11/2/2021, switch to decreased state support for TTS (with a pivot to distribution of weekly at-home antigen tests) on 1/31/2022 and lifting of the state-wide in-school mask mandate on 2/28/2022 (Fig. 1).
Testing programs & consent process
During the 2021-2022 academic year, three different optional testing programs were available to students and staff in all Massachusetts schools (Table 1)5,14. First, a weekly pooled testing (surveillance) program8 in which samples from 2–10 individuals were pooled and tested using PCR to identify positive cases. If a positive pool was identified, additional testing was undertaken to identify the positive case among the pool; thus, individuals included in the pool were tested twice, often resulting in multiple positive tests for each positive individual. The pooled testing program was designed for longitudinal repeated testing. Most commonly, individuals participating in this program underwent pre-planned weekly testing. Because testing was pre-planned, repeated, and independent of exposures or other policies, pooled testing positivity results provide information about case rates over time that is not directly impacted by other policies or testing based on contact tracing. Vaccinated students were not excluded from participation.
Table 1 Details of In-school testing programs
The second testing program was a TTS modified quarantine program for asymptomatic students meeting criteria for in-school COVID-19 exposure9. Rather than automatically being excluded from school, as they might otherwise have been given national guidelines concerning exposures at that time, these students took daily antigen tests on site at the beginning of the school day, and, if negative, continued to participate in in-person learning. At the start of the program, per contemporaneous CDC guidance, in-school exposure was defined as: within 3 feet of distance for >15 min if both masked, or 3–6 feet of distance if 1 or both unmasked. Also per contemporaneous CDC guidance, vaccinated students were excluded from the close contact definition. If students or staff declined to participate in the TTS program, or if TTS was not available in their district, then a 10-day at home quarantine period was recommended. Individual districts and schools made determinations about referrals to and participation in testing programs based on local contact tracing efforts and determination of in-school exposures and thus decisions about testing vaccinated contacts were made at the school level. During the active phase of the program, only in-school exposures were eligible for participation in TTS; exposures that occurred outside of classroom settings were recommended to undergo a 10-day quarantine period. On-site antigen symptomatic testing was a third type of testing program, available for individuals who became sick during the school day.
Effective 1/31/2022, due to ongoing frequent exposures in and out of the classroom, straining contact tracing resources, there was a state-wide shift away from the TTS program to an at-home weekly self-testing program with distribution of antigen tests at school5. The pooled testing program was available for the duration of the school year.
School districts and individual schools selected which testing program or combinations of three programs were locally offered and available; within a school, anyone interested in participating in the program had the option to opt-in. Individual participants were only offered the option of enrolling in all the testing programs that were offered by the district; there was not the option to selectively enroll in testing programs (e.g., one could not consent to test-to-stay but not surveillance testing). Parental or legal guardian consent for participation was required for minor participants and was obtained via both paper and electronically. Participants were able to opt-in and opt-out of the programs throughout the school year.
Cohort creation
A longitudinal dataset at the bi-weekly level for schools nested within towns was created (Supplementary Table 1, Data Sources). School-level COVID-19 testing and consent data were obtained from CIC Health6. Key variables included: proportion of students in the school consented to participate, total number of tests performed, type of test performed, total number of tests positive, total number of surveillance pools performed, total number of pools positive, and information about reflex testing results. For antigen testing, the age of the individual tested was available. Reflex test results, which represent repeated testing, were excluded to avoid double-counting of cases. For pooled testing results, because the pools contained multiple individuals, age was not calculable. Percentage of consented students was calculated by dividing the number of individuals participating in the program by the total school population, which, along with information about school type and demographics, was obtained from DESE and NCES15. Limited individual demographic data were available, so individuals ≤19 years of age were classified as students and individuals >19 years of age were classified as staff members. Schools were excluded if they were missing consent data, typically due to paper consenting processes or if they could not be matched to a CIC Health-test results file, which occurred in approximately 2% of all schools state-wide, primarily due to utilization of a separate tracking system. Data linked to the district only and not a specific school were also excluded.
Cumulative town-level SARS-CoV-2 case rates per 100,000 residents and vaccination rates stratified by age group on a biweekly basis were collected from the Massachusetts Department of Public Health16,17. Town-level cumulative prior infection and vaccination rates were both included as measures of immunity.
Outcome
The school-level testing program initiation date was defined as the date of the first test result documented in the system.
Impact of pooled testing program
To assess the impact of pooled testing, the primary outcome was school-level biweekly test-to-stay positivity rate, defined as the number of positive test-to-stay tests divided by the consented population. The secondary outcome of the first analysis was school-level total test positivity rate, defined as the total number of positive tests (pooled tests, test-to-stay tests, symptomatic tests) divided by the consented school population.
Impacts of vaccination and immunity
To assess the impact of vaccination and immunity, the primary outcome was school-level biweekly surveillance testing positivity rate, defined as the number of positive surveillance pools divided by the consented population. All test positivity rate was evaluated secondarily.
Exposures
The exposure variables evaluated were different in-school mitigation factors. The impact of the pooled testing program was evaluated as a dichotomous variable, with the primary outcome assessed the TTS positivity rate. TTS positivity was selected as the primary outcome because, in theory, early identification of asymptomatic cases through the pooled testing program would reduce onward transmission to exposed classmates, therefore reducing the positivity rate among identified in-school exposures. Impact on all-test positivity rate, which included pooled testing results, was assessed secondarily. Due to near universal participation, the TTS program is presented descriptively.
Assessments of the impact of vaccination rates were performed in two ways. First, school-level vaccination rate was modeled as a continuous variable and assessed in multivariate models with pooled testing positivity rate as the primary outcome. Pooled testing positivity rate was selected as the primary outcome for the impact of vaccination and prior immunity analysis because these tests were collected independent of exposure, unlike TTS. Exposures at the time that determined referral to the TTS program were defined in part by vaccination status. Second, schools were classified into high (> 80%) vaccination or low (< 50%) vaccination uptake, and pooled testing positivity rate was estimated as the primary outcome. For high schools, high vaccination uptake was defined based upon receipt of at least two doses of vaccine on or before 1/1/2022. For elementary schools, where vaccines were not available until 11/2021, high vaccination uptake was defined based upon receipt of at least one dose of vaccine on or before 1/1/2022. Given vaccine availability, impact of vaccination was evaluated in elementary and high schools during the omicron period before and after lifting of the mask mandate (e.g., before and after 2/28/2022). To evaluate the impact of immunity acquired from prior infection, the association between existing cumulative town-level (e.g., community) incidence rate and in-school positivity rate was assessed and stratified by circulating variant (delta, omicron).
Covariates
Additional independent variables included in regression models were town-level vaccination rates. School-level variables included: level of school (elementary, middle, high, mixed age), school type (public, private, charter), school size, and school demographic data (student race and ethnicity, Title I status, rurality), obtained from NCES.
Statistics and reproducibility
Surveillance testing program impact
The impact of surveillance testing on TTS positivity rate was estimated using a matched retrospective observational cohort with a target trial emulation approach, a causal inference method in which observational data are used to “mimic” a randomized controlled trial. Details of the emulation approach are presented in Table 2. Massachusetts schools with a testing program implemented during the 2021-22 academic year were potentially included. Schools were excluded if they initiated surveillance testing prior to 9/19/2021.
Table 2 Matched observational retrospective cohort with a target trial emulation design
Schools that initiated a surveillance testing program on 9/19/2021 were then propensity-matched 1:1 to any school meeting eligibility criteria that did not have a surveillance testing program implemented by 9/19/2021. Propensity matching was based on average vaccination rate in the school population at the town-level, in-school consent rate, race, urban/rural status, title I status, school type (e.g., charter, public, private). The biweekly difference in TTS positivity rate comparing schools with and without pooled testing programs was then assessed using the following multivariable regression specification estimated with generalized estimating equations (GEE) with a gaussian family and an identity link function:
$${{{\rm{E}}}}\left({{{{\rm{y}}}}}_{{{{\rm{it}}}}}\right) = \,{{{{\rm{\beta }}}}}_{1}({{{\rm{Treatment}}}})_{{{{\rm{it}}}}}+{{{{\rm{\beta }}}}}_{2}\left({{{\rm{Weeks}}}}\; {{{\rm{from}}}}\; {{{\rm{program}}}}\; {{{\rm{initiation}}}}\right)_{{{{\rm{it}}}}} \\ + \,{{{{\rm{\beta }}}}}_{3}\left({{{\rm{Proportion}}}}\; {{{\rm{vaccinated}}}}\right)_{{{{\rm{it}}}}}+\,{{{{\rm{\beta }}}}}_{4}({{{{\rm{Community}}}}\; {{{\rm{cases}}}}})_{{{{\rm{it}}}}}$$
where the dependent variable, yit, is the in-school TTS positivity rate (positive tests/school-level consented population/biweekly period) in school i at time t. For the secondary analysis, the dependent variable was the all-test positivity rate in school i at time t. All test positivity rate includes results from all TTS and pooled tests performed in school i. The independent variable of interest, Treatmentit, is an indicator for the presence of a pooled testing program. We also included covariates measuring the number of weeks since 9/19/2021 (Weeks from program initiation), the school vaccination rates (Proportion vaccinated), and the SARS-CoV-2 case rate in the surrounding town (Community cases). The analysis for estimated Eq. (1) was completed using an intention-to-treat approach.
Assessments of the impact of vaccine and natural immunity
The impacts of immunity, vaccination, and testing program availability on pooled testing positivity rates at the school-level were estimated using random effects linear regression adjusted for district demographic variables and stratified by school type (elementary/middle/high):
$${y}_{i,t}=\,{\beta }_{0}+\,{\beta }_{1}{X}_{i,t}+\,{\mu }_{i}+\,{v}_{i,t}+\,{\epsilon }_{t}\,$$
(1)
Where the dependent variable, yi,t, represents the in-school pooled testing positivity rate or all test positivity rate (positive tests/school-level consented population/biweekly period) in school i at time t in as the primary and secondary outcome, respectively. All test positivity rate includes results from all TTS and pooled tests performed in school i.
Due to vaccines being available at different times for different age groups of children (5/18/2021 for ages 12–17 and 11/5/2021 for ages 5–11) and inherent differences in in-school contact patterns, the linear regression models were stratified by school level (elementary, middle, high, mixed grade levels). Additionally, given the scale-up of the program during the early weeks of the school year, we conducted an additional sensitivity analysis in which the first 6 weeks of the school year were excluded to test the robustness of study findings.
All data were analyzed using STATA v.17.
Ethics approval and consent to participate
This study was approved by the VA Boston Research and Development Committee as non-human subjects research prior to data collection and analysis.
