Available online at www.sciencedirect.

com

Journal of Sport and Health Science 9 (2020) 302
312

www.jshs.org.cn

Original article
Projecting the impact of the coronavirus disease-2019 pandemic on
childhood obesity in the United States: A microsimulation model
Ruopeng An
Brown School, Washington University in St. Louis, St. Louis, MO 63130, USA
Received 9 May 2020; revised 11 May 2020; accepted 14 May 2020
Available online 23 May 2020

2095-2546/Ó 2020 Published by Elsevier B.V. on behalf of Shanghai University of Sport. This is an open access article under the CC BY-NC-ND license.
(http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract
Purpose: The coronavirus disease-2019 (COVID-19) pandemic in the United States led to nationwide stay-at-home orders and school closures.
Declines in energy expenditure resulting from canceled physical education classes and reduced physical activity may elevate childhood obesity
risk. This study estimated the impact of COVID-19 on childhood obesity.
Methods: A microsimulation model simulated the trajectory of a nationally representative kindergarten cohort’s body mass index z-scores and
childhood obesity prevalence from April 2020 to March 2021 under the control scenario without COVID-19 and under the 4 alternative scenarios
with COVID-19—Scenario 1: 2-month nationwide school closure in April and May 2020; Scenario 2: Scenario 1 followed by a 10% reduction in
daily physical activity in the summer from June to August; Scenario 3: Scenario 2 followed by 2-month school closure in September and Octo-
ber; and Scenario 4: Scenario 3 followed by an additional 2-month school closure in November and December.
Results: Relative to the control scenario without COVID-19, Scenarios 1, 2, 3, and 4 were associated with an increase in the mean body mass
index z-scores by 0.056 (95% confidence interval (95%CI): 0.055
0.056), 0.084 (95%CI: 0.084
0.085), 0.141 (95%CI: 0.140
0.142), and
0.198 (95%CI: 0.197
0.199), respectively, and an increase in childhood obesity prevalence by 0.640 (95%CI: 0.515
0.765), 0.972 (95%CI:
0.819
1.126), 1.676 (95%CI: 1.475
1.877), and 2.373 (95%CI: 2.135
2.612) percentage points, respectively. Compared to girls and non-His-
panic whites and Asians, the impact of COVID-19 on childhood obesity was modestly larger among boys and non-Hispanic blacks and His-
panics, respectively.
Conclusion: Public health interventions are urgently called to promote an active lifestyle and engagement in physical activity among children to
mitigate the adverse impact of COVID-19 on unhealthy weight gains and childhood obesity.
Keywords: Childhood obesity; Coronavirus; COVID-19; Microsimulation; Physical activity

1. Introduction Control and Prevention recommended school closures in

response to community transmission of COVID-19.9 All 50
The coronavirus disease-2019 (COVID-19) pandemic has
states and Washington, DC, have issued executive orders on
been reported in more than 200 countries, claiming massive ill-
school closures, and most states issued these orders during
nesses, hospitalizations, and death tolls.1,2 Following the dec-
March 13
23.7,8 To date, no state has lifted school closure
laration of a national emergency in the United States in
orders, and it is likely for the school closures in most places to
response to COVID-19,3 42 states and Washington, DC, issued
continue for the rest of the 2019
2020 academic year.10,11 It
stay-at-home orders from March 19 to April 7.4 These orders
remains unclear whether and on what scale schools will reopen
covered approximately 95% of the U.S. population.4 Closely
after the summer break, as such decisions by states are primar-
related to the stay-at-home orders are various non-pharmaceu-
ily contingent upon how the COVID-19 pandemic evolves in
tical interventions that aim to promote social distancing.5,6
the coming months.12,13
One of the most widely adopted policy interventions across
The COVID-19 pandemic may expose children to an
the states is school closures.7,8 The U.S. Centers for Disease
increased risk of childhood obesity.14,15 Childhood obesity is
associated with various immediate and long-term illnesses
Peer review under responsibility of Shanghai University of Sport. such as sleep apnea, hypertension, type 2 diabetes, heart dis-
E-mail address: [email protected] ease, stroke, osteoarthritis, and certain types of cancer.16,17 In
https://doi.org/10.1016/j.jshs.2020.05.006
Cite this article: An R. Projecting the impact of the coronavirus disease-2019 pandemic on childhood obesity in the United States: a microsimulation model.
J Sport Health Sci 2020;9:302
12.
Impact of COVID-19 on childhood obesity 303

the United States, childhood obesity prevalence has increased nationally representative sample drawn from both public and
by 3-fold over the past few decades.18,19 In 2017
2018, private schools, with participants attending both full-day and
approximately 18.5% of children aged 2
19 years (13.7 mil- part-day kindergarten in 2010
2011.31 A total of 15,631 chil-
lion) had obesity.20 Non-Hispanic blacks and Hispanics have dren with diverse socioeconomic and racial/ethnic back-
higher childhood obesity prevalence than non-Hispanic whites grounds participated in the ECLS-K:2011 and were followed
and Asians.20 Physical inactivity among children is one of the up in 9 waves from kindergarten through 5th grade.31
major behavioral risk factors for childhood obesity.21 The
Physical Activity Guidelines for Americans (2nd edition) rec-
2.2. Overview of microsimulation model
ommends 60 min or more of moderate-to-vigorous physical
activity (MVPA) daily among children.22 However, more than A microsimulation model was built to project the impact of
3-quarters (76%) of children in the United States do not meet COVID-19 on body mass index z-scores (BMIz) and childhood
the daily PA level recommended in the guidelines,23 and obesity in the United States. In particular, the model used data
nearly one-half (47%) of U.S. children exceed 2 h per day in from the ECLS-K:2011 kindergarten cohort (the data were first
sedentary behavior.24 Physical education (PE) is a mandated collected in 2010
2011) to simulate changes in BMIz and
part of the school curricula in most states.25 PE contributes to obesity prevalence from April 2020 to March 2021. The study
children’s daily accumulation of PA and is of particular impor- sample’s trajectory of BMIz and childhood obesity prevalence
tance for those who have obesity or lack access to PA opportu- was compared between the control scenario without the
nities in their home environment.26 Because schools have been COVID-19 pandemic and 4 alternative scenarios with
closed nationwide, children in the United States have missed COVID-19. Scenario 1 assumed a 2-month nationwide school
their opportunity to participate in PE classes and other school- closure due to COVID-19 in April and May 2020. Scenario 2
based PA (e.g., recess, after-school sports programs).27 They assumed that Scenario 1 would be followed by a 10% reduc-
are also less likely to make up for it under the stay-at-home tion in daily PA in the summer from June to August. Scenario
orders, as parks and outdoor recreation areas across the states 3 assumed that Scenario 2 would be followed by a 2-month
are closed, thus limiting PA opportunities outside.27,28 In sum, school closure in September and October. Scenario 4 assumed
the COVID-19 pandemic has reinforced the obesogenic envi- that Scenario 3 would be followed by an additional 2-month
ronment conducive to physical inactivity and sedentary behav- school closure in November and December 2020. For each of
ior.29 Children living under such conditions are likely to be the 4 alternative scenarios, a full resumption of routine life and
exposed to an elevated risk for energy imbalance and social operations immediately following the event, and until
unhealthy weight gain.29,30 the end of the study period, i.e., March 2021, was assumed.
Given the looming influence of COVID-19 on PA and the Under the 4 alternative scenarios, the COVID-19 pandemic is
deep uncertainties about how the pandemic may evolve in the expected to positively impact kindergarteners’ body weight
future,29,30 it is critical to assess the impact of COVID-19 on (and consequently BMIz and childhood obesity prevalence)
childhood obesity and project its trajectory in response to the through reduced energy expenditure (EE) resulting from can-
pandemic. Such information will be valuable to policymakers celed PE classes during the academic months and decreased
at federal, state, and local levels, as well as to various stake- PA during the summer break.
holders (e.g., schools, communities, and families), in designing Five technical details regarding model specifications are
and implementing countermeasures to minimize the detrimen- noted. First, the model runs by an increment of a month, but
tal impact of COVID-19 on PA and prevent children from obe- a 2-month time interval during the spring and fall academic
sity onset. This study aimed to build a microsimulation model terms was adopted to simplify the reporting and illustration
to project the impact of COVID-19 on childhood obesity in the of modeling results. Second, states and school districts follow
United States. different academic calendars, so that the exact start and end
dates for academic terms and the summer break differ. How-
2. Methods ever, such geographical and school information was unavail-
able in the ECLS-K:2011; thus, we were unable to model
2.1. Study sample
child-specific school schedules. Nevertheless, it should have
The study sample of the microsimulation model was little influence on the model estimates because the compari-
retrieved from the Early Childhood Longitudinal Study, Kin- sons in BMIz and childhood obesity prevalence between the
dergarten Class of 2010
2011 (ECLS-K:2011). The ECLS- control and 4 alternative scenarios were made in March 2021,
K:2011 is a longitudinal study sponsored by the U.S. Depart- 3 months after the end of the assumed impact of COVID-19.
ment of Education.31 The same children (n = 15,631) were fol- Therefore, as long as the overall impact duration of COVID-19
lowed from kindergarten through 5th grade.31 Most children remains the same for all children, the variations in academic
were 5
6 years old when they first entered the study. Informa- calendars across schools should not impact the estimated differ-
tion was collected in the fall and spring of kindergarten ences in the accumulated impact of COVID-19 on BMIz and
(2010
2011), fall and spring of 1st grade (2011
2012), fall childhood obesity prevalence. Third, a 10% reduction in PA
and spring of 2nd grade (2012
2013), spring of 3rd grade during the summer break was assumed based on a study of
(2014), spring of 4th grade (2015), and spring of 5th grade adult Fitbit users in the United States (i.e., a 12% reduction
(2016).31 The children in the ECLS-K:2011 comprise a in PA due to COVID-19).32 In the sensitivity analyses, an 8%
304 R. An

and 12% reduction in PA during the summer break were used 2.4. EE
to assess their respective influence on model estimates.
In comparison to the control scenario without the COVID-19
Fourth, the model assumed that all children would take PE
pandemic, the reduction in EE in the 4 alternative scenarios
classes if offered. However, in reality, some children may be
with COVID-19 is assumed to result from 2 independent sour-
absent from PE due to health or other reasons. In such a case,
ces—canceled PE classes during the academic months and
the forgone PE classes would serve as a proxy for other
decreased PA during the summer break (June
August). More
missed PA opportunities due to school closures (e.g., recess,
specifically, the model assumes that no PE classes would be
before-, and after-school sports programs). Finally, the
offered if schools were closed during a regular academic term
ECLS-K:2011 followed study participants for 6 years from
(e.g., April and May in the spring of 2020, and September
De-
their kindergarten entrance in 2010
2011 to 5th grade in
cember in the fall). Two recent review studies of national data
2016. It would not be appropriate to extrapolate their BMIz
provided estimates on the weekly instruction time of PE classes
and childhood obesity to 2020
2021. Instead, the microsimu-
among U.S. kindergarteners.37,38 A weekly PE instruction time
lation model assumed that the ECLS-K:2011 study sample all
of 131.81 min was used in the microsimulation model, which
started school in September 2019 (i.e., kindergarteners in
averaged the 2 estimates of 110.10 min by Kahan and McKen-
2019 would share the same height and weight distribution as
zie37 and 153.52 min by Bednar and Rouse.38 Following Kahan
those in 2011), interpolated their height and weight using
and McKenzie,37 Eq. 1 was used to calculate EE (kcal/month)
restricted cubic spline regressions until March 2020, and then
resulting from PE class participation:
simulated changes in BMIz and childhood obesity prevalence
from April 2020 to March 2021 under the control and 4 alter- EEðkcal=monthÞ ¼
131:81 min=week
native scenarios. Based on the U.S. Centers for Disease Con-  4 weeks=month
60 min=h
trol and Prevention reports, the childhood obesity rate among  ð0:503  4:5 MET þ 0:497  1:8 MET Þ
ðEq:1Þ
children aged 6
11 years increased only modestly from  0:453592 kg=lb  weight in lb
17.7% in 2011
2012 to 18.4% in 2017
2018.20,33 Since the in which; 1 MET ¼ 1 kcal=ðkg  hÞ
microsimulation model aimed to project the change in BMIz
and childhood obesity rate rather than their baseline level, the In Eq. 1, MVPA and light PA are assumed to occupy 50.3%
discrepancy resulting from a modest difference in childhood and 49.7%, respectively, of the weekly PE instruction time
obesity rate between 2010
2011 (i.e., the baseline of the and have a metabolic equivalent (MET) of task of 4.5 and 1.8
ECLS-K:2011) and 2019
2020 (i.e., the baseline of the kcal/kg/h, respectively.37
microsimulation model) is unlikely to cause significant esti- Besides PE classes, kindergarteners typically engage in
mation bias. additional PA during recess and before- or after-school pro-
grams. Therefore, counting reduced EE exclusively from can-
celed PE classes during school closure is likely to provide a
2.3. Anthropometric measurement conservative estimate regarding the impact of COVID-19 on
EE among kindergarteners.
In the ECLS-K:2011, children’s height and weight were
Due to nationwide stay-at-home restrictions and closures of
measured during each wave of data collection.34 A Shorr board
exercise facilities (e.g., YMCA, gyms, and recreation cen-
(i.e., a tall wooden stand with a ruled edge used for measuring
ters),29,30 kindergarteners’ PA is likely to be adversely affected
height) and a digital scale were used to obtain the measure-
during the 2020 summer break. On average, elementary school
ments.34 Trained assessors recorded children’s height (in
children in the United States participate in 88 min of PA per
inches to the nearest 1-quarter inch) and weight (in pounds to
day.23 Although no published work to date has estimated the
the nearest one-half pound) on a standardized recording form
influence of COVID-19 on PA among U.S. children, data col-
and then entered the measurements into a laptop computer.34
lected from U.S. adult Fitbit users revealed a reduction in daily
Children were asked to remove their shoes before measure-
PA by about 12%.32 The microsimulation model thus assumes
ment.34 Each measurement was taken and recorded twice to
a 10%, or 8.8-min, reduction in daily PA among U.S. kinder-
ensure reliability.34
garteners, equally divided between MVPA (4.5 METs) and
During the 6-year study period of the ECLS-K:2011, child-
light PA (1.8 MET) time. Eq. 2 was used to calculate EE
ren’s height and weight were measured in 9 waves—once per
(kcal/month) resulting from reduced PA during the summer
academic semester from kindergarten to 2nd grade and once
break:37
per academic year from 3rd to 5th grade.31,34 Restricted cubic
spline regressions were performed to interpolate each child’s EEðkcal=monthÞ ¼
height and weight by month. Using the interpolated height and 8:8 min=day
 30 days=month
60 min=h
weight together with children’s recorded sex and age (in ðEq:2Þ
 ð0:5  4:5 MET þ 0:5  1:8 MET Þ
months), BMIz were calculated using the 2000 U.S. Centers
for Disease Control and Prevention age- and sex-specific  0:453592 kg=lb  weight in lb
in which; 1 MET ¼ 1 kcal=ðkg  hÞ
growth chart.35 Childhood obesity is defined as BMIz in the
95th percentile or higher in the growth chart.36
Impact of COVID-19 on childhood obesity 305

2.5. Modeling COVID-19 impact on BMIz and childhood 2.9. Simulation setting
obesity
The microsimulation model was built in R (Version 3.6.3; R
Following Hennessy et al.,39 the microsimulation model Development Core Team, Vienna, Austria). Each model was
assumes that an energy surplus or deficit translates into simulated 1000 times using bootstrapped samples of the same
weight change at a rate of 3492.66 kcal per pound. All size as the original ECLS-K:2011 (n = 15,631), based on which
agents (i.e., 15,631 kindergarteners in the ECLS-K:2011) the mean, standard error, and 95% confidence interval (CI) of
start their academic year in September 2019, with their the impacts of COVID-19 under different scenarios were esti-
height and weight updated by the end of each month. mated. The ECLS-K:2011 sampling weights were incorpo-
Monthly BMIz and childhood obesity status are also calcu- rated in the effect estimation to facilitate modeling result
lated accordingly. From April 2020 to March 2021, the generalization to the national kindergarten population.34
mean BMIz scores and childhood obesity prevalence among
agents are compared between the control scenario without 3. Results
the COVID-19 pandemic and the 4 alternative scenarios
Fig. 1 shows the simulated kernel density distributions of
with COVID-19.
BMIz in March 2021 under the control and 4 alternative sce-
narios. In comparison to the BMIz distribution under the con-
2.6. Modeling COVID-19 impact by population subgroup trol scenario without COVID-19, model estimates under the 4
alternative scenarios gradually shifted to the right with higher
The microsimulation model estimated the impact of peaks, reflecting a monotonic growth in the mean BMIz when
COVID-19 on BMIz and childhood obesity prevalence among the duration of COVID-19 increased (i.e., from April
May to
all agents and by sex (i.e., boys and girls) and race/ethnicity November
December 2020).
(i.e., non-Hispanic whites, non-Hispanic blacks, non-Hispanic Fig. 2 shows the simulated change in BMIz from April 2020
Asians, non-Hispanic other race, and Hispanics). to March 2021 under the control and 4 alternative scenarios. In
the absence of the COVID-19 pandemic, the mean BMIz
2.7. Model assumptions among U.S. kindergarteners gradually increased from 0.497
(95%CI: 0.480
0.513) in April 2020 to 0.530 (95%CI:
The microsimulation model is based on 2 primary assump- 0.514
0.546) in September 2020 and subsequently declined
tions. First, children’s reduction in PA during the academic to 0.487 (95%CI: 0.470
0.504) in March 2021. In contrast,
months is assumed to be entirely attributable to the cancelation under Scenarios 1, 2, 3, and 4 with COVID-19, the mean
of PE classes. In reality, schools provide many other PA BMIz started at 0.531 (95%CI: 0.515
0.547) in April 2020,
opportunities besides PE classes, such as recess, walk or bike further increased to a value of 0.591 (95%CI: 0.576
0.607) in
to school programs, and after-school sports programs. In such September, 0.623 (95%CI: 0.607
0.638) in August, 0.679
cases, the microsimulation model would provide a conserva- (95%CI: 0.664
0.695) in October, and 0.718 (95%CI:
tive estimate for the impact of COVID-19 on PA reduction 0.703
0.733) in December, and subsequently declined to
and childhood obesity. Admittedly, children may make up for 0.543 (95%CI: 0.526
0.560), 0.572 (95%CI: 0.555
0.588),
their lost PA opportunities at school by engaging in additional 0.629 (95%CI: 0.612
0.645), and 0.685 (95%CI:
PA at home. However, given the widespread community trans- 0.669
0.701) in March 2021, respectively.
mission of COVID-19 and nationwide stay-at-home orders, it Fig. 3 shows the simulated change in childhood obesity
is unlikely for the majority of children to adequately compen- prevalence from April 2020 to March 2021 under the control
sate for their reduced PA due to school closures.29,30 Second, and 4 alternative scenarios. In the absence of COVID-19, obe-
children’s energy imbalance (i.e., surplus) is influenced by sity prevalence among U.S. kindergarteners gradually
reduced PA only. In reality, many other factors related to the increased from 13.52% (95%CI: 12.99%
14.06%) in April
COVID-19 pandemic may impact children’s EE, such as a 2020 to 14.77% (95%CI: 14.22%
15.33%) in March 2021. In
change in diet (e.g., school meals vs. family meals), sleeping, contrast, under Scenario 1, 2, 3, and 4 with COVID-19, despite
and stress or anxiety level.40,41 some modest periodical fluctuations, childhood obesity preva-
lence in large observed a steady increase from 13.86%
(95%CI: 13.32%
14.41%) in April 2020 to 15.41% (95%CI:
2.8. Sensitivity analyses
14.85%
15.98%), 15.74% (95%CI: 15.17%
16.32%),
Two sets of sensitivity analyses were conducted to assess 16.45% (95%CI: 15.87%
17.03%), and 17.15% (95%CI:
changes in the estimated impact of COVID-19 on BMIz and 16.55%
17.74%) in March 2021, respectively.
childhood obesity in response to changes in model assump- Fig. 4A and Fig. 4B report the simulated increase in BMIz
tions. First, the reduction in PA resulting from PE class cancel- between the control scenario and the 4 alternative scenarios in
lations was allowed to vary from 90% to 110% of its mean. March 2021 by sex and race/ethnicity. Compared to the control
Second, the reduction in PA during the summer break was scenario without COVID-19, Scenarios 1, 2, 3, and 4 were asso-
allowed to vary from 8% to 12% of the total daily PA duration ciated with an increase in the mean BMIz among all kindergar-
(i.e., 88 min). The microsimulation model was re-estimated teners by 0.056 (95%CI: 0.055
0.056), 0.084 (95%CI:
for each specific case accordingly. 0.084
0.085), 0.141 (95%CI: 0.140
0.142), and 0.198 (95%CI:
306 R. An

Fig. 1. Simulated kernel density distributions of body mass index z-scores (BMIz) in March 2021 under the control and 4 alternative scenarios. A microsimulation
model was built to simulate the trajectory of a U.S. nationally representative kindergarten cohort’s BMIz from April 2020 to March 2021 under the control scenario
without coronavirus disease-2019 (COVID-19) and under the 4 alternative scenarios with COVID-19—Scenario 1: a 2-month nationwide school closure in April
and May 2020; Scenario 2: Scenario 1 followed by a 10% reduction in daily physical activity (PA) in the summer from June to August; Scenario 3: Scenario 2 fol-
lowed by a 2-month school closure in September and October; and Scenario 4: Scenario 3 followed by an additional 2-month school closure in November and
December.

Fig. 2. Simulated trend in body mass index z-scores (BMIz) from April 2020 to March 2021 under the control and 4 alternative scenarios. A microsimulation
model was built to simulate the trajectory of a U.S. nationally representative kindergarten cohort’s BMIz from April 2020 to March 2021 under the control scenario
without coronavirus disease-2019 (COVID-19) and under the 4 alternative scenarios with COVID-19—Scenario 1: a 2-month nationwide school closure in April
and May 2020; Scenario 2: Scenario 1 followed by a 10% reduction in daily physical activity (PA) in the summer from June to August; Scenario 3: Scenario 2 fol-
lowed by a 2-month school closure in September and October; and Scenario 4: Scenario 3 followed by an additional 2-month school closure in November and
December. The error bars denote the estimated standard errors.
Impact of COVID-19 on childhood obesity 307

Fig. 3. Simulated trend in childhood obesity prevalence from April 2020 to March 2021 under the control and 4 alternative scenarios. A microsimulation model
was built to simulate the trajectory of a U.S. nationally representative kindergarten cohort’s childhood obesity prevalence from April 2020 to March 2021 under
the control scenario without coronavirus disease-2019 (COVID-19) and under the 4 alternative scenarios with COVID-19—Scenario 1: a 2-month nationwide
school closure in April and May 2020; Scenario 2: Scenario 1 followed by a 10% reduction in daily physical activity (PA) in the summer from June to August; Sce-
nario 3: Scenario 2 followed by a 2-month school closure in September and October; and Scenario 4: Scenario 3 followed by an additional 2-month school closure
in November and December. Childhood obesity is defined as body mass index (BMI) z-scores at the 95th percentile or higher in the 2000 U.S. Centers for Disease
Control and Prevention age- and sex-specific growth chart. The error bars denote the estimated standard errors.

0.197
0.199), respectively. The estimated impact of COVID-19 0.183 (95%CI: 0.181
0.186) under Scenario 4, respectively
on BMIz was modestly larger among boys than among girls. (p < 0.05).
Compared to the control scenario without COVID-19, Scenarios Fig. 5A and Fig. 5B report the simulated difference in child-
1, 2, 3, and 4 were associated with an increase in the mean BMIz hood obesity prevalence between the control scenario and the 4
by 0.059 (95%CI: 0.058
0.059), 0.089 (95%CI: 0.088
0.090), alternative scenarios in March 2021 by sex and race/ethnicity.
0.149 (95%CI: 0.148
0.150), and 0.209 (95%CI: 0.207
0.211) Compared to the control scenario without COVID-19, Scenarios
among boys, and 0.052 (95%CI: 0.052
0.053), 0.079 (95%CI: 1, 2, 3, and 4 were associated with an increase in childhood obe-
0.078
0.080), 0.133 (95%CI: 0.132
0.134), and 0.187 (95%CI: sity prevalence among all kindergarteners by 0.640 (95%CI:
0.185
0.188) among girls, respectively (p < 0.05). The esti- 0.515
0.765), 0.972 (95%CI: 0.819
1.126), 1.676 (95%CI:
mated impact of COVID-19 on BMIz was largest among non- 1.475
1.877), and 2.373 (95%CI: 2.135
2.612) percentage
Hispanic Asians, followed by non-Hispanic whites, non-Hispanic points, respectively. The estimated impact of COVID-19 on
other race, non-Hispanic blacks, and Hispanics/Latinos. The childhood obesity prevalence was modestly larger among
increase in mean BMIz among non-Hispanic Asians, non-His- boys than among girls. Compared to the control without the
panic whites, non-Hispanic other race, non-Hispanic blacks, and COVID-19 pandemic, Scenarios 1, 2, 3, and 4 were associated
Hispanics/Latinos was 0.062 (95%CI: 0.061
0.064), 0.058 with an increase in childhood obesity prevalence of 0.664
(95%CI: 0.057
0.058), 0.055 (95%CI: 0.053
0.056), 0.052 (95%CI: 0.486
0.842), 1.090 (95%CI: 0.862
1.318), 1.892
(95%CI: 0.051
0.052), and 0.052 (95%CI: 0.051
0.052) under (95%CI: 1.593
2.191), and 2.719 (95%CI: 2.362
3.075) per-
Scenario 1; 0.094 (95%CI: 0.092
0.096), 0.087 (95%CI: centage points among boys, and 0.614 (95%CI: 0.439
0.790),
0.086
0.088), 0.083 (95%CI: 0.081
0.085), 0.079 (95%CI: 0.850 (95%CI: 0.644
1.056), 1.451 (95%CI: 1.183
1.719),
0.078
0.081), and 0.078 (95%CI: 0.077
0.079) under Scenario and 2.013 (95%CI: 1.699
2.328) percentage points among
2; 0.157 (95%CI: 0.154
0.161), 0.146 (95%CI: 0.145
0.147), girls, respectively (p > 0.05 for Scenarios 1 and 2; and p < 0.05
0.139 (95%CI: 0.136
0.143), 0.133 (95%CI: 0.130
0.136), for Scenarios 3 and 4). In general, the estimated impact of
and 0.131 (95%CI: 0.129
0.133) under Scenario 3; and 0.220 COVID-19 on childhood obesity prevalence tended to be larger
(95%CI: 0.215
0.225), 0.205 (95%CI: 0.203
0.207), among non-Hispanic blacks and Hispanics, whereas it was
0.196 (95%CI: 0.190
0.201), 0.187 (95%CI: 0.183
0.190), and smaller among non-Hispanic whites and Asians. Compared to
308 R. An

Fig. 4. Simulated increase in body mass index z-scores (BMIz) between the control and 4 alternative scenarios in March 2021 by (A) sex and (B) race/ethnicity. A
microsimulation model was built to simulate the trajectory of a U.S. nationally representative kindergarten cohort’s BMIz from April 2020 to March 2021 under
the control scenario without coronavirus disease-2019 (COVID-19) and under the four alternative scenarios with COVID-19—Scenario 1: a 2-month nationwide
school closure in April and May 2020; Scenario 2: Scenario 1 followed by a 10% reduction in daily physical activity (PA) in the summer from June to August; Sce-
nario 3: Scenario 2 followed by a 2-month school closure in September and October; and Scenario 4: Scenario 3 followed by an additional 2-month school closure
in November and December. The error bars denote the estimated standard errors.

the control scenario without COVID-19, Scenarios 1, 2, 3, and 4 In the sensitivity analyses, when the reduction in PA result-
were associated with an increase in childhood obesity preva- ing from PE class cancellations varied from 90% to 110% of
lence of 0.903 (95%CI: 0.499
1.307), 1.331 (95%CI: its mean, Scenarios 1, 2, 3, and 4 were associated with an
0.841
1.821), 2.234 (95%CI: 1.602
2.865), and 3.137 increase in childhood obesity prevalence among all kindergarten-
(95%CI: 2.392
3.882) percentage points among non-Hispanic ers from 0.557 (95%CI: 0.440
0.673) to 0.697 (95%CI:
blacks; 0.657 (95%CI: 0.400
0.914), 1.131 (95%CI: 0.567
0.828), 0.928 (95%CI: 0.777
1.078) to 1.024 (95%CI:
0.795
1.467), 2.025 (95%CI: 1.577
2.472), and 2.892 0.866
1.181), 1.541 (95%CI: 1.349
1.735) to 1.798 (95%CI:
(95%CI: 2.360
3.425) percentage points among Hispanics/ 1.589
2.006), and 2.226 (95%CI: 1.995
2.458) to 2.617
Latinos; 0.566 (95%CI: 0.397
0.735), 0.803 (95%CI: (95%CI: 2.366
2.867) percentage points in comparison to the
0.602
1.004), 1.343 (95%CI: 1.084
1.602), and 1.988 control scenario without COVID-19, respectively. When the
(95%CI: 1.674
2.302) percentage points among non-Hispanic reduction in PA during the summer break varied from 8% to
whites; and 0.254 (95%CI:
0.033 to 0.541), 0.677 (95%CI: 12% of the total daily PA duration, Scenarios 2, 3, and 4 were
0.209
1.145), 1.609 (95%CI: 0.891
2.327), and 1.948 associated with an increase in childhood obesity prevalence
(95%CI: 1.159
2.736) percentage points among non-Hispanic among all kindergarteners from 0.928 (95%CI: 0.777
1.078) to
Asians, respectively. 1.024 (95%CI: 0.866
1.181), 1.606 (95%CI: 1.409
1.803) to
Impact of COVID-19 on childhood obesity 309

Fig. 5. Simulated increase in childhood obesity prevalence between the control and 4 alternative scenarios in March 2021 by (A) sex and (B) race/ethnicity.
A microsimulation model was built to simulate the trajectory of a U.S. nationally representative kindergarten cohort’s childhood obesity prevalence from April
2020 to March 2021 under the control scenario without coronavirus disease-2019 (COVID-19) and under the 4 alternative scenarios with COVID-19—Scenario 1:
a 2-month nationwide school closure in April and May 2020; Scenario 2: Scenario 1 followed by a 10% reduction in daily physical activity (PA) in the summer
from June to August; Scenario 3: Scenario 2 followed by a 2-month school closure in September and October; and Scenario 4: Scenario 3 followed by an additional
2-month school closure in November and December. Childhood obesity is defined as body mass index (BMI) z-scores at the 95th percentile or higher in the 2000
U.S. Centers for Disease Control and Prevention age- and sex-specific growth chart. The error bars denote the estimated standard errors.

1.740 (95%CI: 1.535
1.945), and 2.316 (95%CI: 2.080
2.552) impact of COVID-19 on childhood obesity was modestly
to 2.431 (95%CI: 2.190
2.673) percentage points in comparison larger among boys and non-Hispanic blacks and Hispanics,
to the control scenario without COVID-19, respectively. Scenario respectively.
1 incurred no change because no reduction in PA was assumed To date, the vast majority of research on COVID-19 has
during the summer break. focused on disease pathology and clinical or pharmaceutical
interventions.1,2,42 In contrast, the social impact of the
pandemic and resulting mitigation policies are much less
4. Discussion
studied.43 The social effects of COVID-19, if unaddressed by
Relative to the control scenario without COVID-19, Scenar- timely, effective, targeted countermeasures, may lead to pro-
ios 1, 2, 3, and 4 were associated with an increase in the mean found, long-term health and economic consequences on a
BMIz by 0.056, 0.084, 0.141, and 0.198 units, respectively, scale much larger than the disease infection.43 Childhood obe-
and an increase in childhood obesity prevalence by 0.640, sity has been a leading public health concern in the United
0.972, 1.676, and 2.373 percentage points, respectively. States, and its prevalence is at a record high.16,17 The model
Compared to girls and non-Hispanic whites and Asians, the predicts that even a 2-month school closure alone could result
310 R. An

in an increase in the childhood obesity rate by 0.640 percent- objectively measured. Third, it provides the impact estimates
age points among U.S. kindergarteners. If school closures specific to child population subgroups by sex and race/ethnic-
continue to the end of 2020 due to unsubdued community ity, which may inform targeted interventions.
transmission of COVID-19, the childhood obesity rate in the Despite these contributions, several limitations of this study
United States might further increase by 2.373%. If such an should be noted. First, it is a simulation study based on obser-
impact is universal among all U.S. children aged 5
17 years, vational data. Due to the lack of sample randomization, the
by March 2021 approximately 1.27 million new childhood modeling results do not infer causality. Second, the scope of
obesity cases will develop under the COVID-19 pandemic this study is limited to the kindergarten years of a nationally
than otherwise. representative study cohort so that the findings may not be
The estimated sex difference in the impact of COVID-19 fully generalizable to all children and adolescents in the United
on childhood obesity is mainly attributable to the differen- States. Third, the microsimulation model assumes that child-
tial EE reduction between boys and girls. EE is a positive ren’s energy surplus under COVID-19 entirely results from PE
function of body weight.36 Because boys, in general, are class cancellations due to school closures during the academic
heavier than girls,44 the same decrease in PA duration will months. In reality, a multitude of factors may impact child-
result in a more substantial energy surplus among boys than ren’s EE and weight gain under the pandemic, including, but
among girls. not limited to, elevated stress and change in diet, daily routine,
The primary contributing factor to the estimated racial/eth- and sleeping patterns.40,41 Nevertheless, given the variety of
nic differences in the impact of COVID-19 on childhood obe- PA opportunities typically offered at school and given the
sity pertains to their differential baseline BMIz distribution. nationwide stay-at-home orders,27,28 the model assumption on
Compared to non-Hispanic whites (0.404) and Asians (0.218), forgone PE classes could only produce a conservative esti-
the mean BMIz in April 2020 is significantly higher among mate. The sensitivity analyses indicated that within a reason-
non-Hispanic blacks (0.569) and Hispanics (0.726). Therefore, able range, changes in the assumptions on PA reductions
the same reduction in PA duration across races/ethnicities is during the academic months and the summer break did not
more likely to induce childhood obesity among non-Hispanic cause substantial changes in modeling results. Fourth, the
blacks and Hispanics. Customized policy interventions are model assumes that the impact of COVID-19 on PA is uni-
warranted to prevent further deterioration of weight-related formly distributed among children, which does not capture
health outcomes among these 2 socioeconomically disadvan- the population heterogeneity in response to the pandemic. In
taged racial/ethnic minority groups. reality, some children may cope with the pandemic better or
For the coming months or longer, the majority of children worse than average and engage in more or less PA to mitigate
in the United States may not fully resume their daily PA rou- their energy surplus due to school and facility closures. In
tines due to the nationwide closure of schools, gyms, recrea- this case, the model estimates the average response and out-
tion centers, and parks and cancellations of sports come at the population level but not at the level of individual
activities.12
15 Policy countermeasures that aim to mitigate differences.
the adverse impact of COVID-19 need to consider innovative,
robust, and highly adaptable strategies to promote PA at home 5. Conclusion
or in residential settings that minimize social gathering. Some
researchers have made a call to action for PA engagement as a Using a microsimulation modeling approach, this study
valuable tool for controlling COVID-19 infections and main- projected the change in U.S. kindergarteners’ BMIz and
taining quality of life.45 childhood obesity under the COVID-19-induced uncertainties.
Answers to the following research questions may better Our simulation results indicate that compared to the control
inform relevant policy countermeasures. What PAs can be per- scenario without COVID-19, both BMIz and childhood obe-
formed individually or jointly by family members that have sity prevalence under COVID-19 are expected to rise, and the
few or no restrictions on location, facility, and equipment, and magnitude of the increase is proportional to the length and
how can we promote them? How can we integrate PA into severity of the pandemic. Public health interventions are
everyday stay-at-home routines to meet the daily PA level rec- urgently called for if we wish to promote an active lifestyle
ommended in the guidelines? 22 How can we best utilize avail- and engagement in PA among children and want to mitigate
able resources at home and in close surroundings to optimize the adverse impact of COVID-19 on unhealthy weight gains
PA? How can we educate children and other family members and childhood obesity.
about the importance of daily PA, ensure safety, and prevent
PA-related injuries? Competing interests
The contributions of this study are 3-fold. First, it serves as The author declares that he has no competing interests.
the first attempt to project the impact of COVID-19 on PA and
childhood obesity. Second, it constructs a microsimulation
Supplementary materials
model based on data from a nationally representative longitu-
dinal survey of a kindergarten cohort first interviewed in Supplementary material associated with this article can be
2010
2011, whose height and weight was repeatedly and found, in the online version, at doi:10.1016/j.jshs.2020.05.006.
Impact of COVID-19 on childhood obesity 311

References 20. Centers for Disease Control and Prevention. Childhood obesity facts.
Prevalence of childhood obesity in the United States. Available at: https://
1. Kakodkar P, Kaka N, Baig MN. A comprehensive literature review on the www.cdc.gov/obesity/data/childhood.html. [accessed 12.05.2020].
clinical presentation, and management of the pandemic coronavirus dis- 21. Chi DL, Luu M, Chu F. A scoping review of epidemiologic risk factors for
ease 2019 (COVID-19). Cureus 2020;12:e7560. doi:10.7759/cureus.7560. pediatric obesity: implications for future childhood obesity and dental car-
2. Helmy YA, Fawzy M, Elaswad A, Sobieh A, Kenney SP, Shehata AA. ies prevention research. J Public Health Dent 2017;77(Suppl. 1):S8–31.
The COVID-19 pandemic: a comprehensive review of taxonomy, genet- 22. U.S. Department of Health and Human Services. Physical activity guide-
ics, epidemiology, diagnosis, treatment, and control. J Clin Med 2020;9: lines for Americans, 2nd edition. Available at: https://health.gov/sites/
E1225. doi:10.3390/jcm9041225. default/files/2019-09/Physical_Activity_Guidelines_2nd_edition.pdf.
3. White House. Proclamation on declaring a national emergency concern- [accessed 12.05.2020].
ing the novel coronavirus disease (COVID-19) outbreak. Available at: 23. National Physical Activity Plan Alliance. The 2018 United States report
https://www.whitehouse.gov/presidential-actions/proclamation-declaring- card on physical activity for children and youth. Available at: http://
national-emergency-concerning-novel-coronavirus-disease-covid-19-out www.physicalactivityplan.org/projects/reportcard.html. [accessed
break. [accessed 12.05.2020]. 12.05.2020].
4. New York Times. See which states and cities have told residents to stay at 24. Sisson SB, Church TS, Martin CK, Tudor-Locke C, Smith SR, Bouchard
home. Available at: https://www.nytimes.com/interactive/2020/us/corona C, et al. Profiles of sedentary behavior in children and adolescents: the US
virus-stay-at-home-order.html. [accessed 12.05.2020]. National Health and Nutrition Examination Survey, 2001
2006. Int J
5. Hartley DM, Perencevich EN. Public health interventions for COVID-19: Pediatr Obes 2009;4:353–9.
emerging evidence and implications for an evolving public health crisis. 25. An R, Ji M, Clarke C, Guan C. Impact of state laws governing physical
JAMA 2020. doi:10.1001/jama.2020.5910. education on attendance among US high school students, 2003 to 2017.
6. National Conference of State Legislatures. State action on coronavirus Am J Health Promot 2019;33:1144–51.
(COVID-19). COVID-19 related legislation. Available at: https://www. 26. National Association for Sport and Physical Education. Guidance docu-
ncsl.org/research/health/state-action-on-coronavirus-covid-19.aspx. ment: instructional framework for fitness education in physical education.
[accessed 12.05.2020]. Available at: https://www.shapeamerica.org/upload/Instructional-Frame
7. Institute for Public Policy and Social Research, Michigan State Univer- work-for-Fitness-Education-in-Physical-Education.pdf. [accessed
sity. State policies to address COVID-19 school closure. Available at: 12.05.2020].
http://ippsr.msu.edu/state-policies-address-covid-19-school-closure. 27. Guan H, Okely AD, Aguilar-Farias N, del Pozo Cruz B, Draper CE, Ham-
[accessed 12.05.2020]. douchi AE, et al. Promoting healthy movement behaviors among children
8. National Conference of State Legislatures. Public education’s response to during the COVID-19 pandemic. Lancet Child Adolesc Health
the coronavirus (COVID-19) pandemic. Available at: https://www.ncsl. 2020;4:416–8.
org/research/education/public-education-response-to-coronavirus-covid- 28. Fawkner S, Niven A, Hanson S, Williamson C, Hanson CL. Physical
19.aspx. [accessed 12.05.2020]. activity for children and young people aged 5
18 years during COVID-
9. Centers for Disease Control and Prevention. Considerations for school 19. Stay safe; be active. Available at: https://blogs.bmj.com/bjsm/2020/
closure. Available at: https://www.cdc.gov/coronavirus/2019-ncov/down 04/13/physical-activity-for-children-and-young-people-aged-5-18-years-
loads/considerations-for-school-closure.pdf. [accessed 12.05.2020]. during-covid-19-stay-safe-be-active. [accessed 12.05.2020].
10. National Broadcasting Company. When will school open? Here’s a state- 29. Pinto A, Dunstan DW, Owen N, Bonfa E, Gualano B. Combating physical
by-state list. Available at: https://www.today.com/parents/when-will- inactivity during the COVID-19 pandemic. Available at: https://www.
school-open-here-s-state-state-list-t179718. [accessed 12.05.2020]. nature.com/articles/s41584-020-0427-z. [accessed 12.05.2020].
11. National Public Radio. Half of U.S. public school students are home for 30. American Psychological Association. How and why to get children mov-
the school year. Available at: https://www.npr.org/sections/coronavirus- ing now. Available at: https://www.apa.org/topics/covid-19/children-exer
live-updates/2020/04/16/835941050/nearly-half-of-u-s-public-school-stu cise-strategies. [accessed 12.05.2020].
dents-are-home-for-the-school-year. [accessed 12.05.2020]. 31. National Center for Education Statistics. Kindergarten Class of 2010-11
12. USA Today. When will schools reopen? Not soon, education leaders say, (ECLS-K:2011). Available at: https://nces.ed.gov/ecls/kindergarten2011.
despite Trump’s declarations. Available at: https://www.usatoday.com/ asp. [accessed 12.05.2020].
story/news/education/2020/04/21/coronavirus-schools-reopen-closures- 32. Fitbit. The impact of coronavirus on global activity. Available at: https://
fall-2020/5167726002. [accessed 12.05.2020]. blog.fitbit.com/covid-19-global-activity. [accessed 12.05.2020].
13. Centers for Disease Control and Prevention. Guidance for cleaning and 33. U.S. Department of Health and Human Services. Health Resources and
disinfecting public spaces, workplaces, businesses, schools, and homes. Services Administration, Maternal and Child Health Bureau. Child Health
Available at: https://www.cdc.gov/coronavirus/2019-ncov/community/ USA. Available at: https://mchb.hrsa.gov/chusa14. [accessed 12.05.2020].
pdf/Reopening_America_Guidance.pdf. [accessed 12.05.2020]. 34. National Center for Education Statistics. Early childhood longitudinal
14. Rundle AG, Park Y, Herbstman JB, Kinsey EW, Wang YC. COVID-19- study, kindergarten class of 2010
11 (ECLS-K:2011). Available at:
related school closings and risk of weight gain among children. Obesity https://nces.ed.gov/pubs2017/2017285.pdf. [accessed 12.05.2020].
(Silver Spring) 2020;28:1008–9. 35. Centers for Disease Control and Prevention. Overview of the CDC growth
15. Pietrobelli A, Pecoraro L, Ferruzzi A, Heo M, Faith M, Zoller T, et al. charts. Available at: https://www.cdc.gov/nccdphp/dnpa/growthcharts/
Effects of COVID
19 lockdown on lifestyle behaviors in children with training/modules/module2/text/module2print.pdf. [accessed 12.05.2020].
obesity living in Verona, Italy: a longitudinal study. Obesity (Silver 36. Centers for Disease Control and Prevention. Defining childhood obesity.
Spring) 2020. doi:10.1002/oby.22861. Available at: https://www.cdc.gov/obesity/childhood/defining.html.
16. Kumar S, Kelly AS. Review of childhood obesity: from epidemiology, eti- [accessed 12.05.2020].
ology, and comorbidities to clinical assessment and treatment. Mayo Clin 37. Kahan D, McKenzie TL. Energy expenditure estimates during school
Proc 2017;92:251–65. physical education: potential vs. reality? Prev Med 2017;95:82–8.
17. Hruby A, Hu FB. The epidemiology of obesity: a big picture. Pharmacoe- 38. Bednar S, Rouse K. The effect of physical education on children’s body
conomics 2015;33:673–89. weight and human capital: new evidence from the ECLS-K:2011. Health
18. Ogden CL, Carroll MD, Lawman HG, Fryar CD, Kruszon-Moran D, Kit Econ 2020;29:393–405.
BK, et al. Trends in obesity prevalence among children and adolescents 39. Hennessy E, Ornstein JT, Economos CD, Herzog JB, Lynskey V, Coffield
in the United States, 1988-1994 through 2013-2014. JAMA 2016;315: E, et al. Designing an agent-based model for childhood obesity interven-
2292–9. tions: a case study of ChildObesity180. Prev Chronic Dis 2016;13:E04.
19. Lifshitz F. Obesity in children. J Clin Res Pediatr Endocrinol 2008;1:53–60. doi:10.5888/pcd13.150414.
312 R. An

40. Sasaki JE, Howe C, John D, Hickey A, Steeves J, Conger S, et al. Energy 43. Nicola M, Alsafi Z, Sohrabi C, Kerwan A, Al-Jabir A, Iosifidis C, et al.
expenditure for 70 activities in children and adolescents. J Phys Act The socioeconomic implications of the coronavirus and COVID-19 pan-
Health 2016;13(Suppl. 1):S24–8. demic: a review. Int J Surg 2020;78:185–93.
 Saunders TJ, Chaput JP. Daily energy
41. Thivel D, Aucouturier J, Doucet E, 44. Kirchengast S. Gender differences in body composition from child-
balance in children and adolescents. Does energy expenditure predict sub- hood to old age: an evolutionary point of view. J Life Sci 2010;
sequent energy intake? Appetite 2013;60:58–64. 2:1–10.
42. Sanders JM, Monogue ML, Jodlowski TZ, Cutrell JB. Pharmacologic 45. Sallis J, Pratt M. A call to action: physical activity and COVID-19. Avail-
treatments for coronavirus disease 2019 (COVID-19): a review. JAMA able at: https://www.exerciseismedicine.org/support_page.php/stories/?
2020. doi:10.1001/jama.2020.6019. b=896. [accessed 12.05.2020].