To identify factors associated with the decision to provide in-person, hybrid, and remote learning in kindergarten through 12th grade school districts during the 2020–2021 school year.
We performed a retrospective study evaluating school district mode of learning and community coronavirus 2019 (COVID-19) incidence and percentage positivity rates at 3 time points during the pandemic: (1) September 15, 2020 (the beginning of the school year, before Centers for Disease Control and Prevention guidance); (2) November 15, 2020 (midsemester after the release of Centers for Disease Control and Prevention guidance and an increase of COVID-19 cases); and (3) January 15, 2021 (start of the second semester and peak COVID-19 rates). Five states were included in the analysis: Michigan, Missouri, North Carolina, Ohio, and Wisconsin. The primary outcome was mode of learning in elementary, middle, and high schools during 3 time points. The measures included community COVID-19 incidence and percentage positivity rates, school and student demographics, and county size classification of school location.
No relationship between mode of learning and community COVID-19 rates was observed. County urban classification of school location was associated with mode of learning with school districts in nonmetropolitan and small metropolitan counties more likely to be in-person.
Community COVID-19 rates did not appear to influence the decision of when to provide in-person learning. Further understanding of factors driving the decisions to bring children back into the classroom are needed. Standardizing policies on how schools apply national guidance to local decision-making may decrease disparities in emergent crises.
Factors guiding decisions related to mode of learning are not well understood; it is unknown whether COVID-19 community rates, as incorporated into Centers for Disease Control and Prevention school operational guidance, or other factors were associated with mode of school learning during the 2020–2021 school year.
Community coronavirus 2019 rates did not appear to influence the decision of when to provide in-person learning. Standardizing policies on how schools apply national guidance to local decision-making may decrease disparities in emergent crises.
In March 2020, US schools closed to in-person education when stay-at-home orders were issued because of the coronavirus 2019 (COVID-19) pandemic. Although many cities lifted these orders in May and June 2020, most schools remained closed for the summer.1 As schools determined reopening plans during summer 2020, little guidance was available on when and in what learning mode (ie, remote, hybrid, or in-person) to reopen. On September 15, 2020, the Centers for Disease Control and Prevention (CDC) provided indicators and thresholds for the risk of introduction and transmission of COVID-19 in US kindergarten through 12th grade (K–12) schools.2 These indicators were primarily based on the level of community COVID-19 infections; specifically, the number of new COVID-19 cases per 100 000 persons and the percentage of positive severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) real-time polymerase chain reaction tests during the previous 14 days in the community determined in-school transmission risk. Nevertheless, this guidance was issued after the start of the school year, and the majority of schools had already determined their initial reopening plans.
Despite CDC indicators for the risk of COVID-19 transmission in K–12 school, consensus guidance recommending a specific mode of learning based on community rates was not initially available. Some state and local jurisdictions developed conflicting guidance about the rates in which in-person learning could occur, and different groups (eg, school officials versus local government) had decision-making power in different jurisdictions, resulting in variable modes of learning for US K–12 schools during the 2020–2021 school year. As of December 7, 2020, only 62% of US K–12 school districts were operating with full or hybrid in-person learning.3 Compared with in-person learning, remote learning is associated with worse academic outcomes and increased mental health problems. Parents report higher risk of negative parental and child physical and emotional well-being when children are receiving remote or virtual learning, compared with full in-person instruction.4 Even relatively short periods of remote learning result in learning loss, which is more pronounced among students from disadvantaged homes.5
To date, the factors guiding decisions related to mode of learning are not well understood. Whether COVID-19 community rates, as incorporated into CDC school operational guidance, or other factors are associated with mode of school learning is unknown. Therefore, we performed a multistate review of K–12 school mode of learning at 3 different time points to identify factors associated with in-person, hybrid, and remote learning.
Methods
We performed a retrospective study evaluating school district mode of learning and community COVID-19 incidence and percentage positivity rates at 3 time points during the pandemic: September 15, 2020, November 15, 2020, and January 15, 2021. These time points were selected to capture (1) the beginning of the school year (before CDC guidance); (2) midsemester after the release of CDC guidance and an increase of COVID-19 cases; and (3) the start of the second semester and peak COVID-19 rates. Five states were included in the analysis: Michigan, Missouri, North Carolina, Ohio, and Wisconsin. States were selected on the basis of the state of residence of the participating investigators. All investigators are participants in the ABC Science Collaborative, which is a program that pairs scientists and physicians with school and community leaders to help understand the most current and relevant information about COVID-19.6 Twenty percent of each participating state’s public and charter school districts were randomly selected for this study.
Data Collection
The Department of Elementary and Secondary Education, Department of Education, Department of Public Instruction, and Center for Educational Performance and Information7–9 Web sites were accessed to obtain a complete list of public school districts for each state. Michigan, Missouri, and Wisconsin also included charter schools. Each school district location was classified by county size.10 School demographic data, including student numbers, race, ethnicity, and percentage of free and reduced lunch were obtained from the publicly available databases. Community COVID-19 incidence was measured by using the case daily average from the CDC by county and date of interest, multiplying by 14, dividing by the population for the county, and multiplying by 100 000.11 Positivity rates were determined by using the average of COVID-19 viral laboratory 14-day positive rates in the county where the district was located.12 COVID-19 incidence and percentage positivity were evaluated over a 14-day time frame to align with the CDC September 2020 thresholds for risk of the introduction and transmission of COVID-19 in US K–12 schools.2
A standardized definition of mode of learning was used. In-person learning consisted of in-person classroom instruction 5 full days per week and available for all students. Hybrid learning provided some in-person learning for the general student population, but learning was not full-time, ranging from 1 to 4 days per week, every other week in-person learning, or a partial in-person day (ie, am/pm schedule). Remote learning only provided no in-person instruction for the general student population. Mode of learning by district for the dates of interest were determined from Department of Elementary and Secondary Education/Department of Education, local news coverage, school district Web sites, and by directly contacting schools.
Statistical Analysis
Study variables, including (1) how schools reported on demographic, socioeconomic, and geographic characteristics; and (2) county-level COVID-19 incidence and testing positivity rates, were summarized and synthesized by using summary statistics. For analysis, county size was reclassified into 4 categories: large metropolitan (large central and large fringe), medium metropolitan, small metropolitan, and nonmetropolitan (micropolitan and noncore). We compared the proportion of schools falling into various categories (location, urban density), the proportion of studies reporting on key demographic variables (race, ethnicity), and the proportion of schools reporting on free and reduced lunches stratified by mode of learning and school level (elementary, middle, high). Furthermore, we compared proportion of schools and the median COVID-19 incident cases per 100 000 by urban class, school level, and time points. Multivariable logistic regression was used to estimate the association between county-level COVID cases and school characteristics including mode of learning, school demographics, and location.
Results
A total of 445 elementary, 379 middle, and 390 high schools were included in this study across 5 US states. In-person was the most common learning mode during the study period. Elementary schools were more frequently in-person (September: 63.3%, November: 58.0%, January: 57.3%) compared with middle (September: 53.7%, November: 48.5%, January: 45.3%) or high school (September: 58.4%, November: 48.7%, January: 49.0%), and access to in-person education decreased throughout the pandemic (Table 1, Supplemental Table 1). Remote learning increased over time, with the highest rates in January 2021: elementary 22.9%, middle 20.0%, high 24.6%. The mean percentage of schools fully remote during the study period was 18% for elementary, 22% for middle, and 21% for high school.
More schools in nonmetropolitan counties provided in-person learning as compared with schools in higher density areas. On September 15, 79% of elementary schools in both nonmetropolitan and small metropolitan areas provided full in-person learning compared with 50% in medium metropolitan and 40% in large metropolitan areas. Similar trends were observed in middle and high schools (Table 1, Supplemental Table 1) and across the 3 time points.
On average, schools that offered full in-person learning had fewer enrolled students. Full in-person schools also had the highest percentage of White students. On September 15, 79% of full in-person elementary students were White, 8% were Hispanic, and 7% were Black as compared with schools providing virtual mode of learning were 40% White, 15% were Hispanic, and 36% were Black. Fewer students in in-person elementary (44%) as compared with virtual (67%) were eligible for free and reduced lunch. (Table 1). Similar trends were observed across school level (ie, elementary, middle, and high) and time periods.
The mean incidence of SARS-CoV-2 cases per 100 000 people and percentage positivity of SARS-CoV-2 tests in counties were not different across modes of learning at any time point. Throughout the evaluation period, case rates were almost always at a level considered to be in the higher or highest risk of school transmission on the basis of the available CDC classification, with November and January community rates in the highest risk level of transmission, exceeding >200 cases per 100 000 persons and >10% positivity (Fig 1, Supplemental Fig 1).
In a multiple regression analysis, urban classification was the factor most strongly associated with in-person mode of learning. In September, before CDC guidance was available, elementary schools in nonmetropolitan and small metropolitan counties were 3.4 and 1.9 times more likely to be in person. Similar trends were observed in middle and high schools. COVID-19 community rates and percentage positive were not associated with mode of learning across the study period (Fig 2). Race and ethnicity, as well as free and reduced lunch, were not consistently associated with mode of learning.
Discussion
During the 2020–2021 school year, CDC guidance indicated that community COVID-19 incidence rates influenced the risk of in-school transmission; however, the results from our study suggest that schools did not uniformly incorporate these metrics into their decision-making given the lack of association between community incidence rates and mode of learning. Recent research suggests that community rates have a minimal influence on the rate of in-school transmission when proper mitigation strategies are strictly followed.13 Although we observed no relationship between mode of learning and community COVID-19 incidence rates, we did find that urban classification of school location was associated with mode of learning with less in-person learning in large metropolitan areas. This highlights the importance of providing clear, scientific-based guidance around school decision-making.
Previously, researchers have found that during the COVID-19 pandemic, access to full in-person learning varied by race, ethnicity, geography, and grade level.14–16 Remote learning has been more commonly reported in children attending public school as compared with those attending private school.4 In a sample of North Carolina school districts, the probability and duration of remote-only learning was higher in school districts located in counties with a high proportion of Black residents, whereas Republican voting areas had decreased odds and shorter duration of remote-only learning.14
To date, the relationship of mode of learning with community COVID-19 incidence rates and county size has not been assessed. Our data suggest that county size was most associated with mode of learning. There is a known relationship between geographic location and population diversity. In 2018, racial and ethnic minority groups made up 22% of the rural population compared with 43% of the urban population.10 Urban school districts were less commonly in person during this study period, and the percentage of in-person schools was inversely proportional to the county population.
Metropolitan and nonmetropolitan schools may have prioritized different factors in their mode of learning decisions. Throughout the study period, the vast majority of included districts had community COVID-19 case rates that were classified as being the higher and highest risk of SARS-CoV-2 transmission in schools. Metropolitan districts may have chosen to focus more on community COVID-19 rates and may have been less able to implement mitigation strategies, specifically the ability to maintain physical distancing for larger school districts. Recent research has found that SARS-CoV-2 transmission in schools can be limited when mitigation strategies are in place, regardless of community transmission rates. With high adherence to mitigation strategies in schools, viral transmission is consistently lower than in the community.13 Physical distancing may have been more difficult to achieve in large school districts. More recent data highlight that masking can limit SARS-CoV-2 transmission in classrooms, even when physical distancing cannot be maintained.17–19 Continuous re-evaluation of ongoing science should drive changes to local and national guidance and policies to inform school decision-making during a pandemic.
In July 2021, CDC released updated guidance highlighting the importance of offering in-person learning, regardless of whether all mitigation strategies can be implemented.20 Schools are encouraged to work with local public health authorities to determine the necessary prevention strategies. Clear, scientific-based guidance as to when and how to institute these strategies will lead to less variability in uptake and, therefore, decrease the risk that schools have in-school transmission, particularly those with large unvaccinated populations (eg, elementary school).
Limitations of this study include that only 5 states were assessed, and the results may not be generalizable to the overall US population. Some districts did not have readily available public-facing data. We were unable to assess the mitigation strategies implemented by the districts. Schools may have had changes to mode of learning that were beyond their control, including excessive quarantines and local and/or state requirements. Differences in parental concern as related to return to in-person school have been associated with race and ethnicity, with racial and ethnic minority parents being more concerned about aspects of school reopening when compared with non-Hispanic White parents,21 and parental perspectives could have influenced reopening policies. The decision-making around mode of learning is complex and is influenced by state and local guidance and regulations, as well as teachers’ union, parent, and staff concerns. Furthermore, the level of decision-making is not uniform throughout the country, with reopening decisions being made at the state- or district-level in different areas. We were unable to assess all factors that may have contributed to a district’s decisions in this process.
Given that community COVID-19 rates did not appear to influence the decision of when to provide in-person learning, further understanding of factors driving the decisions to bring children back into the classroom is needed. In-person learning is the optimal mode of instruction in most circumstances, because schools provide education and critical services such as meals and therapies necessary for children to survive and thrive. Standardizing policies on how schools apply national guidance to local decision-making may decrease disparities in emergent crises.
Acknowledgment
Erin Campbell, MS, provided editorial review and submission.
Dr Brookhart and Ms Pak performed the data analyses and reviewed and revised the manuscript, Drs Maier, Hill, Butteris, DeMuri, Benjamin, and Zimmerman, Mr Anand, and Ms Omidfar designed the data collection instruments, collected the data, reviewed and revised the manuscript; Drs Goldman and Schuster designed the study, drafted the initial manuscript, conceptualized and designed the study, coordinated and supervised data collection, reviewed and revised the manuscript, and critically reviewed the manuscript for important intellectual content; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FUNDING: Funded in part by the Rapid Acceleration of Diagnostics Underserved Populations (U24 MD016258; National Institutes of Health [NIH] Agreement No.’s 1 OT2 HD107543-01, 1 OT2 HD107544-01, 1 OT2 HD107553-01, 1 OT2 HD107555-01, 1 OT2 HD107556-01, 1 OT2 HD107557-01, 1 OT2 HD107558-01, 1 OT2 HD107559-01); the Trial Innovation Network, which is an innovative collaboration addressing critical roadblocks in clinical research and accelerating the translation of novel interventions into life-saving therapies; and the National Institute of Child Health and Human Development contract (HHSN275201000003I) for the Pediatric Trials Network (principal investigator, Daniel Benjamin). The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the National Institutes of Health. Funded by the National Institutes of Health (NIH).
CONFLICTS OF INTEREST DISCLOSURES Dr Brookhart serves on scientific advisory committees for American Academy of Allergy, Asthma & Immunology, AbbVie, Amgen, Atara Biotherapeutics, Brigham and Women’s Hospital, Gilead, US Renal Data System, and Vertex; he receives consulting fees and own equity in NoviSci/Target RWE. Dr Benjamin reports consultancy for Allergan, Melinta Therapeutics, Sun Pharma Advanced Research Co. Dr Zimmerman reports funding from the National Institutes of Health and US Food and Drug Administration.