jama | originalinvestigation … · 2 days ago · t...
TRANSCRIPT
Association Between Statewide School Closure and COVID-19 Incidenceand Mortality in the USKatherine A. Auger, MD, MSc; Samir S. Shah, MD, MSCE; Troy Richardson, PhD; David Hartley, PhD, MPH;Matthew Hall, PhD; Amanda Warniment, MD; Kristen Timmons, MS; Dianna Bosse, BA; Sarah A. Ferris, BA;Patrick W. Brady, MD, MSc; Amanda C. Schondelmeyer, MD, MSc; Joanna E. Thomson, MD, MPH
IMPORTANCE In the US, states enacted nonpharmaceutical interventions, including schoolclosure, to reduce the spread of coronavirus disease 2019 (COVID-19). All 50 states closedschools in March 2020 despite uncertainty if school closure would be effective.
OBJECTIVE To determine if school closure and its timing were associated with decreasedCOVID-19 incidence and mortality.
DESIGN, SETTING, AND PARTICIPANTS US population–based observational study conductedbetween March 9, 2020, and May 7, 2020, using interrupted time series analysesincorporating a lag period to allow for potential policy-associated changes to occur. To isolatethe association of school closure with outcomes, state-level nonpharmaceutical interventionsand attributes were included in negative binomial regression models. States were examinedin quartiles based on state-level COVID-19 cumulative incidence per 100 000 residents at thetime of school closure. Models were used to derive the estimated absolute differencesbetween schools that closed and schools that remained open as well as the number of casesand deaths if states had closed schools when the cumulative incidence of COVID-19 was in thelowest quartile compared with the highest quartile.
EXPOSURES Closure of primary and secondary schools.
MAIN OUTCOMES AND MEASURES COVID-19 daily incidence and mortality per 100 000 residents.
RESULTS COVID-19 cumulative incidence in states at the time of school closure ranged from 0to 14.75 cases per 100 000 population. School closure was associated with a significantdecline in the incidence of COVID-19 (adjusted relative change per week, −62% [95% CI,−71% to −49%]) and mortality (adjusted relative change per week, −58% [95% CI, −68% to−46%]). Both of these associations were largest in states with low cumulative incidence ofCOVID-19 at the time of school closure. For example, states with the lowest incidence ofCOVID-19 had a −72% (95% CI, −79% to −62%) relative change in incidence compared with−49% (95% CI, −62% to −33%) for those states with the highest cumulative incidence. In amodel derived from this analysis, it was estimated that closing schools when the cumulativeincidence of COVID-19 was in the lowest quartile compared with the highest quartile wasassociated with 128.7 fewer cases per 100 000 population over 26 days and with 1.5 fewerdeaths per 100 000 population over 16 days.
CONCLUSIONS AND RELEVANCE Between March 9, 2020, and May 7, 2020, school closure inthe US was temporally associated with decreased COVID-19 incidence and mortality; statesthat closed schools earlier, when cumulative incidence of COVID-19 was low, had the largestrelative reduction in incidence and mortality. However, it remains possible that some of thereduction may have been related to other concurrent nonpharmaceutical interventions.
JAMA. 2020;324(9):859-870. doi:10.1001/jama.2020.14348Published online July 29, 2020.
Viewpoint page 833 andEditorial page 845
Author Audio Interview
Video and Supplementalcontent
Author Affiliations: Authoraffiliations are listed at the end of thisarticle.
Corresponding Author: Katherine A.Auger, MD, MSc, Division of HospitalMedicine, Cincinnati Children’sHospital Medical Center, 3333 BurnetAve, ML 9016, Cincinnati, OH 45229([email protected]).
Research
JAMA | Original Investigation
(Reprinted) 859
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
T he novel severe acute respiratory syndrome coronavi-rus 2 (SARS-CoV-2) causing coronavirus disease 2019(COVID-19) was first identified in the US in January 2020,
with subsequent spread throughout the country. In the ab-sence of effective treatments, governors and state health of-ficials enacted policies aimed at reducing infections throughnonpharmaceutical interventions.1 The nonpharmaceutical in-terventions included: school closure, nonessential business clo-sure, restaurant and bar closure, and prohibition of gather-ings with more than 10 people. With limited precedent and apaucity of evidence on the effectiveness of nonpharmaceuti-cal interventions, policies varied markedly state to state inscope and timing.
Children infected with SARS-CoV-2 may be asymptom-atic or have mild symptoms indistinguishable from other com-mon upper respiratory tract infections,2,3 allowing them tospread the virus when they feel well. Children are often keytransmitters in viral epidemics like influenza4 because ofspending prolonged periods in close proximity to other chil-dren at school and during physical activities. Prior studies havedemonstrated an association between school closure and re-duced transmission of viral respiratory illnesses.5-8 Given con-cerns that children represented a significant vector for SARS-CoV-2, all states closed schools despite a lack of evidencesupporting the effectiveness of school closure in curbing thespread of this virus.
Schools promote child education, growth, development,and overall well-being.9 Knowing whether school closure is ef-fective in reducing infections is critical to reduce the nega-tive effects of continued school closure on child health if schoolclosure is ineffective. This national study assessed the asso-ciation between school closure and its timing with subse-quent COVID-19 incidence and mortality, with the hypoth-esis that any association between school closure and incidenceand mortality would be strongest in states that closed schoolsearly when the cumulative incidence of disease was low.
MethodsThe study was a population-based time series analysis of all50 US states conducted between March 9, 2020, and May 7,2020. This period allowed for at least 6 weeks of data collec-tion after school closures in each state. The institutional re-view board at Cincinnati Children’s Hospital Medical Centerdeemed this study was not subject to oversight given the useof publicly available data.
Independent VariablesAssociations of primary and secondary school closure (kin-dergarten-grade 12) and its timing with outcomes of interestwere examined. Because school closure timing varied rela-tive to disease progression in the state, we examined the cu-mulative incidence of COVID-19 (defined as total number ofcases per 100 000 residents) grouped in quartiles by the datethe school closure policy went into effect.
The analysis was performed by quartiles of cumulative in-cidence of COVID-19 instead of as a continuous variable be-
cause the relationships between baseline cumulative inci-dence and outcomes were not assumed to be linear.
Outcome MeasuresDaily COVID-19 incidence and daily mortality per 100 000 resi-dents in each state were estimated using publicly available datafrom the Johns Hopkins University School of Public Health.10
This source aggregates data from the US Centers for DiseaseControl and Prevention (CDC) as well as from state and localpublic health departments. In accordance with CDC guide-lines, confirmed COVID-19 cases include presumptive posi-tive cases and probable cases, and death totals include con-firmed and probable cases. The denominator for the outcomemeasures was the state population from the 2018 AmericanCommunity Survey.11
CovariatesState characteristics were included as covariates in the mod-els to assess the independent associations with school clo-sure. For each state, the following non–school-related non-pharmaceutical intervention covariates were considered: stay-at-home or shelter-in-place order, nonessential businessclosure, restaurant and bar closure, and prohibition of gath-erings with more than 10 people. These nonpharmaceutical in-terventions were included based on the policy effective date(eTable 1 in the Supplement) plus a lag period to allow for anypotential policy-related effects on daily COVID-19 incidenceand mortality (eFigure in the Supplement). Potential effectsassociated with subsequent lifting of nonpharmaceutical in-terventions occurred outside the study period and thus are notincluded (eMethods in the Supplement).
SARS-CoV-2 testing rates varied by state and throughout thestudy period. To account for this variation, state-level COVID-19testing12 (calculated daily as the cumulative number of tests per1000 residents) was modeled as a categorical variable. Statemeasures of urban population density (a measure of the state’spopulation density combined with the percentage of residentsliving in urban areas),11 percentage of the state’s population with
Key PointsQuestion Was statewide school closure associated withdecreased incidence and mortality for coronavirus disease 2019(COVID-19)?
Findings In this US population–based time series analysisconducted between March 9, 2020, and May 7, 2020, schoolclosure was associated with a significant decline in both incidenceof COVID-19 (adjusted relative change per week, −62%) andmortality (adjusted relative change per week, −58%). In a modelderived from this analysis, it was estimated that closing schoolswhen the cumulative incidence of COVID-19 was in the lowestquartile compared with the highest quartile was associated with128.7 fewer cases per 100 000 population over 26 days and with1.5 fewer deaths per 100 000 population over 16 days.
Meaning There was a temporal association between statewideschool closure and lower COVID-19 incidence and mortality,although some of the reductions may have been related to otherconcurrent nonpharmaceutical interventions.
Research Original Investigation Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US
860 JAMA September 1, 2020 Volume 324, Number 9 (Reprinted) jama.com
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
obesity,13 percentage of the state’s population aged 15 years oryounger,11 percentage of the state’s population aged 65 years orolder,11 and the number of nursing home residents per 1000people were included.14 The CDC social vulnerability index alsowas included; this index accounts for multiple factors, includ-ing socioeconomic status, household composition, disability sta-tus, race and ethnicity composition, English-language profi-ciency, housing type, and transportation access, to assess acommunity’s preparedness for a natural disaster or illnessoutbreak.15 Many of these factors have been associated withCOVID-19 disease, mortality, or both. All covariates are de-scribed in detail in the eMethods in the Supplement.
Statistical AnalysisInterrupted time series analyses were used to compare the dailychange in outcomes (daily COVID-19 incidence and mortal-ity) before and after school closure. Acknowledging that schoolclosure and other nonpharmaceutical interventions would nothave immediate effects on COVID-19 incidence and mortal-ity, estimates were used to determine when school-based ex-posure could be expected to lead to changes in COVID-19 in-cidence and associated mortality (eFigure in the Supplement).A time from exposure to symptom onset of 5 days was as-sumed per Lauer et al.16 Given the early emphasis (and somestate restrictions17-25) on limiting testing to hospitalized pa-tients, time to diagnosis was defined as time between symp-tom onset and hospitalization (7 days).26 For school closure,given the low documented prevalence of COVID-19 in chil-dren, an additional period was included for a child to infect anadult, assuming a child exposed at school could expose an adultprior to symptom onset and within 4 days. The analyses forthe mortality outcome assumed 17 days from symptom onsetto death.27 For details on lag period calculations, see theeMethods in the Supplement.
Daily COVID-19 incidence and mortality were modeledusing negative binomial regression. Interactions betweenschool closure and all included covariates were explored be-cause school closure may affect at-risk communitiesdifferently.9 Given the large number of covariates and inter-actions considered, a single parsimonious model for each out-come was created selecting covariates from the primary modelusing a stepwise regression approach, with entry and re-moval criteria specified as a P value of <.20. Because factorsassociated with COVID-19 incidence and mortality may varywith school closure, covariate selection was completed inde-pendently during and after the lag period (eMethods in theSupplement).
Results are reported as the relative change in the out-come from week to week. Adjusted changes in both dailyCOVID-19 incidence and daily mortality over time are graphi-cally displayed for all states by quartile of cumulative inci-dence at the time of school closure. To estimate absolute dif-ferences associated with school closure, the projected COVID-19incidence and mortality if schools had remained open werecompared with the modeled incidence and mortality withschool closure. Both linear and exponential assumptions wereused to project COVID-19 incidence and mortality if schoolshad remained open (eMethods in the Supplement). To esti-
mate absolute differences in outcomes based on school clo-sure timing, model parameters were used to estimate the ab-solute differences in the number of COVID-19 cases and deathsfor a state that closed schools when the cumulative incidenceof COVID-19 was in the lowest quartile compared with a statethat closed schools when the cumulative incidence of COVID-19was in the highest quartile (eMethods in the Supplement).
Analyses were performed using SAS version 9.4 (SAS In-stitute Inc) and 2-sided P values of <.05 were considered sta-tistically significant.
Sensitivity Analyses Around the Lag PeriodBecause the COVID-19 incidence and mortality lag period es-timates are based on emerging evidence, the sensitivity analy-ses examined the robustness of findings if the lag period wasshorter (10 days for incidence and 20 days for mortality) or lon-ger (21 days for incidence and 33 days for mortality). Detailedmethods and rationale for the sensitivity ranges appear in theeMethods and eFigure in the Supplement.
ResultsAll 50 states closed schools between March 13, 2020, and March23, 2020. The cumulative incidence of COVID-19 at the time ofschool closure ranged from 0 to 14.75 cases per 100 000 popu-lation. State characteristics by COVID-19 incidence quartile atthe time of school closure appear in Table 1. There was wide vari-ability in the testing rate per 1000 residents and in the numberof nursing home residents per 1000 people. There was less vari-ability in the percentage of the state populations for the num-ber of nonpharmaceutical interventions enacted; 39 states en-acted all 4 nonpharmaceutical interventions examined.
States in the highest quartile of cumulative incidence ofCOVID-19 at the time of school closure enacted multiple non-pharmaceutical interventions over a shorter period. The me-dian time from school closure to the last enacted nonpharma-ceutical intervention was 5 days (interquartile range, 2.5-8days). In comparison, states in the lowest quartile of cumula-tive incidence of COVID-19 at the time of school closure en-acted nonpharmaceutical interventions over a longer period.The median time from school closure to the last enacted non-pharmaceutical intervention was 12 days (interquartile range,8-14 days; Table 1).
COVID-19 IncidenceThe observed case rates of COVID-19 in each state (relative tothe day of school closure by cumulative incidence) and the 16-day lag period are depicted in Figure 1A. In the unadjustedanalyses during the period prior to potential effects of schoolclosure (ie, during the lag period), the overall relative changein COVID-19 incidence per week was 220% (95% CI, 205% to236%). The unadjusted relative change per week associatedwith school closure was −68% (95% CI, −70% to −66%). Theunadjusted effect size associated with school closure variedby cumulative COVID-19 incidence at the time of school clo-sure, with states in the highest quartile of cumulative COVID-19incidence having the smallest relative effect size (Table 2).
Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US Original Investigation Research
jama.com (Reprinted) JAMA September 1, 2020 Volume 324, Number 9 861
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
Tabl
e1.
Stat
eCh
arac
teris
ticsb
yCo
rona
viru
sDis
ease
2019
(CO
VID
-19)I
ncid
ence
Qua
rtile
atth
eTi
me
ofSc
hool
Clos
ure
Stat
esby
COVI
D-19
inci
denc
equ
artil
eat
the
time
ofsc
hool
clos
ure
Cum
ulat
ive
inci
denc
epe
r10
000
0re
side
nts
Date
ofsc
hool
clos
ure
inM
arch
Max
imum
test
ing
rate
per1
000
resi
dent
sa
Perc
enta
geof
stat
epo
pula
tion
No.
ofnu
rsin
gho
me
resi
dent
spe
r100
0pe
ople
Perc
enta
geof
stat
epo
pula
tion
with
obes
ityb
Urba
nde
nsity
asre
side
nts/
squa
rem
ilec
Soci
alvu
lner
abili
tyin
dexd
No.
ofno
n–sc
hool
-re
late
dN
PIs
enac
tede
Days
betw
een
scho
olcl
osur
ean
den
actm
ent
ofla
stN
PIAg
ed≤1
5y
Aged
≥65
yLo
wes
tqua
rtile
Med
ian
(IQ
R)0.
48(0
.26-
0.57
)20
.5(1
6.0-
25.0
)18
.5(1
8.2-
19.5
)15
.6(1
4.6-
16.5
)3.
8(3
.3-5
.1)
33.0
(30.
4-34
.8)
46.9
(44.
5-58
.5)
4(4
-4)
12(8
-14)
Wes
tVirg
inia
016
31.4
16.9
18.3
5.1
39.5
<50
51.5
48
Nor
thDa
kota
0.13
1654
.319
.613
.97.
435
.1<5
020
.12
14
Alas
ka0.
1416
33.0
21.2
10.5
0.8
29.5
<50
45.9
412
Ariz
ona
0.26
1616
.019
.516
.51.
629
.550
-100
72.1
415
Nor
thCa
rolin
a0.
3716
16.9
18.7
15.1
3.5
3310
0-15
058
.54
14
Kent
ucky
0.47
1617
.718
.915
.15.
136
.650
-100
55.8
410
Okl
ahom
a0.
4817
22.2
20.4
14.6
4.7
34.8
<50
62.6
28
Haw
aii
0.49
1625
.018
.217
.12.
424
.910
0-15
046
.94
9
Mic
higa
n0.
5316
24.8
18.1
15.9
3.8
3350
-100
45.6
48
Sout
hCa
rolin
a0.
5615
15.6
18.4
16.4
3.4
34.3
100-
150
61.8
422
Ohi
o0.
5817
15.1
18.5
15.6
6.3
34>1
5044
.54
6
Virg
inia
0.58
1614
.618
.414
.33.
330
.410
0-15
034
.43
14
Penn
sylv
ania
0.60
1620
.517
.216
.86
30.9
>150
41.1
416
Seco
ndlo
wes
tqua
rtile
Med
ian
(IQ
R)0.
79(0
.71-
0.87
)19
.7(1
7.6-
26.8
)19
.5(1
8.4-
19.7
)15
.2(1
4.3-
16.6
)4.
5(3
.7-5
.7)
33.8
(30.
7-35
.1)
46.0
(34.
1-67
.2)
4(4
-4)
10(8
.5-1
4)
Kans
as0.
6217
15.4
20.5
14.5
534
.4<5
038
.64
11
Mon
tana
0.67
1619
.418
.217
.24
26.9
<50
26.2
410
Mar
ylan
d0.
6816
24.8
18.6
14.2
4.1
30.9
>150
39.1
414
Iow
a0.
7316
21.2
19.4
15.6
7.5
35.3
<50
29.1
310
Arka
nsas
0.74
1720
.119
.615
.75.
837
.1<5
065
.52
9
Loui
sian
a0.
7713
43.0
19.8
13.9
5.6
36.8
50-1
0072
410
New
Mex
ico
0.81
1640
.919
.616
2.7
32.3
<50
77.2
47
Nev
ada
0.82
1517
.619
.214
.81.
829
.5<5
068
.94
16
Dela
war
e0.
8416
28.8
17.8
17.2
4.4
33.5
>150
44.7
48
Mai
ne0.
9015
17.6
15.6
194.
530
.4<5
029
.54
17
Indi
ana
0.90
1918
.819
.614
.45.
834
.110
0-15
047
.34
5
Texa
s0.
9319
16.3
21.8
11.6
3.3
34.8
50-1
0065
.24
14
Seco
ndhi
ghes
tqua
rtile
Med
ian
(IQ
R)1.
28(1
.16-
1.60
)23
.3(1
9.8-
27.0
)18
.8(1
7.6-
20.1
)15
.1(1
4.4-
15.8
)4.
6(3
.5-5
.3)
30.7
(29.
9-32
.5)
42.6
(30.
5-57
.3)
4(3
-4)
11(7
-15)
Neb
rask
a0.
9416
19.8
20.8
14.4
634
.1<5
033
.72
18
Ore
gon
0.96
1617
.317
.616
.51.
829
.9<5
052
.53
7
(con
tinue
d)
Research Original Investigation Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US
862 JAMA September 1, 2020 Volume 324, Number 9 (Reprinted) jama.com
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
Tabl
e1.
Stat
eCh
arac
teris
ticsb
yCo
rona
viru
sDis
ease
2019
(CO
VID
-19)I
ncid
ence
Qua
rtile
atth
eTi
me
ofSc
hool
Clos
ure
(con
tinue
d)
Stat
esby
COVI
D-19
inci
denc
equ
artil
eat
the
time
ofsc
hool
clos
ure
Cum
ulat
ive
inci
denc
epe
r10
000
0re
side
nts
Date
ofsc
hool
clos
ure
inM
arch
Max
imum
test
ing
rate
per1
000
resi
dent
sa
Perc
enta
geof
stat
epo
pula
tion
No.
ofnu
rsin
gho
me
resi
dent
spe
r100
0pe
ople
Perc
enta
geof
stat
epo
pula
tion
with
obes
ityb
Urba
nde
nsity
asre
side
nts/
squa
rem
ilec
Soci
alvu
lner
abili
tyin
dexd
No.
ofno
n–sc
hool
-re
late
dN
PIs
enac
tede
Days
betw
een
scho
olcl
osur
ean
den
actm
ent
ofla
stN
PIAg
ed≤1
5y
Aged
≥65
yFl
orid
a1.
0517
23.9
16.6
19.4
3.5
30.7
>150
60.7
417
Sout
hDa
kota
1.16
1623
.320
.815
.16.
930
.1<5
032
.61
7
Illin
ois
1.26
1729
.618
.814
.35.
231
.8>1
5048
.64
4
New
Ham
pshi
re1.
2716
21.4
15.8
16.5
4.8
29.6
50-1
0013
.84
11
Utah
1.28
1644
.225
.310
.31.
727
.8<5
030
.23
11
Min
neso
ta1.
3918
17.6
19.5
14.6
4.5
30.1
<50
27.8
49
Wis
cons
in1.
5918
17.7
18.4
15.6
4.2
3250
-100
30.5
47
Alab
ama
1.60
1923
.718
.615
.74.
636
.250
-100
61.8
415
Mis
siss
ippi
1.67
1927
.020
14.5
5.3
39.5
<50
74.1
415
Conn
ectic
ut1.
9017
32.4
1715
.86.
327
.4>1
5042
.64
9
Geor
gia
1.93
1821
.120
.112
.83.
232
.510
0-15
057
.34
16
Hig
hest
quar
tile
Med
ian
(IQ
R)3.
30(2
.95-
11.3
7)30
.2(1
9.5-
43.5
)18
.7(1
7-19
.2)
15.1
(14.
6-15
.4)
4.1
(2.8
-5.5
)27
.7(2
5.8-
28.9
)41
.5(3
6.0-
51.9
)4
(4-4
)5
(2.5
-8)
Calif
orni
a2.
4319
21.5
19.3
13.3
2.6
25.8
>150
65.4
40
Verm
ont
2.88
1829
.515
.417
.83.
927
.5<5
020
.14
7
Mas
sach
uset
ts2.
8816
51.5
16.5
15.2
5.7
25.7
>150
40.1
48
New
Jers
ey3.
0118
33.0
18.3
155
25.7
>150
44.4
43
Mis
sour
i3.
0723
17.0
18.8
15.5
6.2
3550
-100
41.1
214
Wyo
min
g3.
0919
20.7
19.9
14.7
4.2
29<5
027
.82
1
Tenn
esse
e3.
5020
35.5
18.7
15.3
434
.410
0-15
052
.84
11
Idah
o4.
0323
18.3
21.6
14.8
228
.4<5
038
.94
2
Rhod
eIs
land
10.0
323
77.9
16.2
15.7
7.4
27.7
>150
514
5
New
York
12.7
218
55.6
17.5
15.1
5.2
27.6
>150
54.5
45
Colo
rado
12.7
323
16.5
1913
.12.
923
<50
334
3
Was
hing
ton
14.7
517
30.8
18.7
14.5
2.2
28.7
50-1
0041
.94
8
Abbr
evia
tions
:IQ
R,in
terq
uart
ilera
nge;
NPI
s,no
npha
rmac
eutic
alin
terv
entio
ns.
aCa
lcul
ated
daily
bydi
vidi
ngth
ecu
mul
ativ
enu
mbe
roft
ests
perf
orm
edin
ast
ate
byth
est
ate
popu
latio
n;m
odel
edas
aca
tego
rical
varia
ble.
The
min
imum
test
ing
rate
was
less
than
0.1t
ests
per1
00
0re
siden
tsan
doc
curr
edon
the
first
day
ofte
stin
gfo
reve
ryst
ate.
The
max
imum
test
ing
rate
was
the
test
ing
rate
onth
ela
stda
yof
the
stud
ype
riod.
bD
efin
edas
abo
dym
assi
ndex
(cal
cula
ted
asw
eigh
tin
kilo
gram
sdiv
ided
byhe
ight
inm
eter
ssqu
ared
)of3
0or
grea
ter.
cA
mea
sure
ofth
est
ate’
spop
ulat
ion
dens
ityco
mbi
ned
with
the
perc
enta
geof
resid
ents
livin
gin
urba
nar
eas;
exam
ined
asa
cate
goric
alva
riabl
e.
dU
SCe
nter
sfor
Dise
ase
Cont
rola
ndPr
even
tion
mea
sure
ofa
com
mun
ity’s
prep
ared
ness
fora
natu
rald
isast
eror
illne
ssou
tbre
akby
acco
untin
gfo
rsoc
ioec
onom
icst
atus
,hou
seho
ldco
mpo
sitio
n,di
sabi
lity
stat
us,r
ace
and
ethn
icity
com
posit
ion,
Engl
ish-la
ngua
gepr
ofic
ienc
y,ho
usin
gty
pe,a
ndtr
ansp
orta
tion
acce
ss.S
cale
is0
to10
0w
ithlo
wer
num
bers
indi
catin
gbe
tter
prep
ared
ness
fora
natu
rald
isast
eror
illne
ssou
tbre
ak;t
hena
tiona
lmed
ian
is50
.e
Rang
edbe
twee
n1a
nd4.
The
NPI
sexa
min
edin
clud
edst
ay-a
t-ho
me
orsh
elte
r-in-
plac
eor
der,
none
ssen
tial
busin
essc
losu
re,r
esta
uran
tand
barc
losu
re,a
ndpr
ohib
ition
ofga
ther
ings
with
mor
eth
an10
peop
le.A
llsta
tes
enac
ted
atle
ast1
non–
scho
ol-r
elat
edN
PI.
Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US Original Investigation Research
jama.com (Reprinted) JAMA September 1, 2020 Volume 324, Number 9 863
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
Figure 1. Observed Daily COVID-19 Incidence and Mortality by Cumulative Incidence Quartile at the Time of School Closure
Daily incidence by quartile of cumulative incidence at the time of school closureA
40
30
20
10
Daily
CO
VID-
19 in
cide
nce
per 1
00 0
00
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Lowest quartile
40
30
20
10
Daily
CO
VID-
19 in
cide
nce
per 1
00 0
00
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Second lowest quartile
40
30
20
10
Daily
CO
VID-
19 in
cide
nce
per 1
00 0
00
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Second highest quartile
40
30
20
10
0 0
0 0
Daily
CO
VID-
19 in
cide
nce
per 1
00 0
00
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Highest quartile
Daily mortality rate by quartile of cumulative incidence at the time of school closureB
5
4
2
1Daily
CO
VID-
19 d
eath
spe
r 100
000
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Lowest quartile
5
4
2
1Daily
CO
VID-
19 d
eath
spe
r 100
000
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Second lowest quartile
5
4
3
2
Daily
CO
VID-
19 d
eath
spe
r 100
000
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Second highest quartile
5
4
3
2
3 3
1 1
0 0
0 0
Daily
CO
VID-
19 d
eath
spe
r 100
000
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Highest quartile
COVID-19 indicates coronavirus disease 2019.
Research Original Investigation Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US
864 JAMA September 1, 2020 Volume 324, Number 9 (Reprinted) jama.com
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
Tabl
e2.
COVI
D-19
Inci
denc
ean
dM
orta
lity
and
Effe
ctSi
zeAs
soci
ated
With
Scho
olCl
osur
e
Mod
elAl
lSta
tes
COVI
D-19
cum
ulat
ive
inci
denc
equ
artil
eat
the
time
ofsc
hool
clos
ure
Low
est
Seco
ndlo
wes
tSe
cond
high
est
Hig
hest
COVI
D-19
inci
denc
e
Perio
dbe
fore
scho
olcl
osur
e(d
urin
gla
gpe
riod:
days
1-16
)
Ove
rall
new
case
sper
100
000
resi
dent
s,N
o.71
2829
3915
1
New
case
sper
100
000
resi
dent
sper
stat
e,m
edia
n(I
QR)
19(1
7to
52)
16(1
5to
19)
21(1
8to
34)
30(1
7to
45)
73(5
1to
107)
Unad
just
edre
lativ
ech
ange
perw
eek,
%(9
5%CI
)22
0(2
05to
236)
324
(284
to36
9)25
9(2
25to
296)
195
(170
to22
3)13
5(1
14to
157)
Com
posi
tea
adju
sted
bre
lativ
ech
ange
perw
eek,
%(9
5%CI
)26
5(2
31to
303)
387
(327
to45
6)33
9(2
85to
402)
240
(200
to28
6)14
3(1
17to
173)
Perio
daf
ters
choo
lclo
sure
(aft
erla
gpe
riod:
days
17-4
2)
Ove
rall
new
case
sper
100
000
resi
dent
s,N
o.23
815
716
018
338
7
New
case
sper
100
000
resi
dent
sper
stat
e,m
edia
n(I
QR)
116
(79
to21
7)79
(69
to12
2)12
4(8
4to
257)
126
(110
to19
8)12
1(9
0to
764)
Com
posi
tec
adju
sted
bre
lativ
ech
ange
perw
eek,
%(9
5%CI
)10
(1to
18)
7(–
2to
18)
16(5
to28
)20
(9to
32)
–4(–
12to
5)
Rela
tive
chan
gepe
rwee
kas
soci
ated
with
scho
olcl
osur
e,%
(95%
CI)
Unad
just
ed–6
8(–
70to
–66)
–77
(–80
to–7
3)–6
9(–
74to
–65)
–62
(–67
to–5
6)–6
1(–
66to
–55)
Adju
sted
b–6
2(–
71to
–49)
–72
(–79
to–6
2)–6
6(–
75to
–54)
–55
(–66
to–3
9)–4
9(–
62to
–33)
COVI
D-19
mor
talit
y
Perio
dbe
fore
scho
olcl
osur
e(d
urin
gla
gpe
riod:
days
1-26
)
Ove
rall
new
deat
hspe
r100
000
resi
dent
s,N
o.7
33
314
New
deat
hspe
r100
000
resi
dent
sper
stat
e,m
edia
n(I
QR)
2(1
to4)
1(1
to2)
2(1
to3)
2(1
to3)
6(2
to10
)
Unad
just
edre
lativ
ech
ange
perw
eek,
%(9
5%CI
)17
1(1
60to
184)
212
(184
to24
3)18
2(1
57to
210)
169
(148
to19
2)12
9(1
13to
146)
Com
posi
tea
adju
sted
dre
lativ
ech
ange
perw
eek,
%(9
5%CI
)18
6(1
75to
197)
243
(212
to27
7)22
4(1
95to
258)
172
(152
to19
5)12
0(1
05to
136)
Perio
daf
ters
choo
lclo
sure
time
(aft
erla
gpe
riod:
days
27-4
2)
Ove
rall
new
deat
hspe
r100
000
resi
dent
s,N
o.9
76
715
New
deat
hspe
r100
000
resi
dent
sper
stat
e,m
edia
n(I
QR)
3(2
to9)
3(2
to4)
3(2
to11
)4
(2to
5)4
(2to
27)
Com
posi
tec
adju
sted
dre
lativ
ech
ange
perw
eek,
%(9
5%CI
)2
(–8
to14
)6
(–7
to20
)10
(–4
to25
)7
(–6
to22
)–1
2(–
21to
–1)
Rela
tive
chan
gepe
rwee
kas
soci
ated
with
scho
olcl
osur
e,%
(95%
CI)
Unad
just
ed–6
4(–
67to
–61)
–69
(–73
to–6
4)–6
3(–
68to
–57)
–63
(–68
to–5
7)–6
1(–
66to
–56)
Adju
sted
d–5
8(–
67to
–46)
–64
(–73
to–5
2)–6
1(–
71to
–47)
–54
(–65
to–3
9)–5
3(–
63to
–40)
Abbr
evia
tions
:CO
VID
-19,c
oron
aviru
sdise
ase
2019
;IQ
R,in
terq
uart
ilera
nge.
aIn
dica
test
heov
eral
lincr
ease
.The
com
pone
ntso
feac
hco
varia
teat
trib
utin
gto
the
over
allc
hang
ebe
fore
scho
olcl
osur
ear
epr
esen
ted
ineT
able
2(in
cide
nce)
and
eTab
le5
(mor
talit
y)in
the
Supp
lem
ent.
bAd
just
edfo
rallm
odel
com
pone
ntsr
etai
ned
inin
cide
nce
mod
el(in
terc
ept:
perc
enta
geof
stat
e’sp
opul
atio
nag
ed�
15ye
ars,
perc
enta
geof
stat
e’sp
opul
atio
nag
ed�
65ye
ars,
and
US
Cent
ersf
orD
iseas
eCo
ntro
land
Prev
entio
n(C
DC)
soci
alvu
lner
abili
tyin
dex;
befo
resc
hool
clos
ure:
stay
-at-
hom
eor
shel
ter-i
n-pl
ace
orde
r,re
stau
rant
and
barc
losu
re,t
estin
gra
tepe
r10
00
resid
ents
,and
urba
nde
nsity
;aft
ersc
hool
clos
ure:
test
ing
rate
per1
00
0re
siden
ts,s
tay-
at-h
ome
orsh
elte
r-in-
plac
eor
der,
perc
enta
geof
stat
e’sp
opul
atio
nag
ed�
65ye
ars,
num
bero
fnur
sing
hom
ere
siden
tspe
r10
00
peop
le,a
ndur
ban
dens
ity).
cBa
sed
onth
elin
earc
ombi
natio
nof
the
follo
win
gm
odel
para
met
eres
timat
es:r
elat
ive
chan
gebe
fore
scho
olcl
osur
e,th
eef
fect
asso
ciat
edw
ithsc
hool
clos
ure,
and
allo
ther
non–
scho
olcl
osur
eef
fect
sove
rtim
e.d
Adju
sted
fora
llmod
elco
mpo
nent
sret
aine
din
mor
talit
ym
odel
(inte
rcep
t:pe
rcen
tage
ofst
ate’
spop
ulat
ion
aged
�15
year
s,pe
rcen
tage
ofst
ate’
spop
ulat
ion
aged
�65
year
s,an
dCD
Cso
cial
vuln
erab
ility
inde
x;be
fore
scho
olcl
osur
e:st
ay-a
t-ho
me
orsh
elte
r-in-
plac
eor
der,
proh
ibiti
onof
gath
erin
gsw
ith>1
0pe
ople
,res
taur
anta
ndba
rcl
osur
e,pe
rcen
tage
ofst
ate’
spop
ulat
ion
aged
�15
year
s,pe
rcen
tage
ofst
ate’
spop
ulat
ion
aged
�65
year
s,nu
mbe
rofn
ursin
gho
me
resid
ents
per1
00
0pe
ople
,and
urba
nde
nsity
;aft
ersc
hool
clos
ure:
rest
aura
ntan
dba
rcl
osur
e,nu
mbe
rofn
ursin
gho
me
resid
ents
per1
00
0pe
ople
,and
urba
nde
nsity
).
Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US Original Investigation Research
jama.com (Reprinted) JAMA September 1, 2020 Volume 324, Number 9 865
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
In the adjusted analyses during the period prior to poten-tial effects of school closure (ie, during the lag period), the rela-tive change in COVID-19 incidence per week was 265% (95%CI, 231% to 303%; Table 2). The overall combined compositerelative weekly change in COVID-19 incidence after school clo-sure was 10% (95% CI, 1% to 18%). This composite change af-ter school closure is a combination of the changes associatedwith school closure and other non–school-related changes dur-ing the period after school closure and is visually depicted asthe change after school closure in Figure 2. When examiningonly school closure, it was associated with a relative changein COVID-19 incidence per week of −62% (95% CI, −71% to−49%; Table 2).
The states that closed early, when the cumulative inci-dence of COVID-19 was lowest, had the greatest relative changeper week associated with school closure (−72% [95% CI, −79%to −62%]). States that were slowest to close schools and hadthe highest cumulative incidence of COVID-19 had a relative
change per week associated with school closure of −49% (95%CI, −62% to −33%; Table 2 and Figure 2). The full model withall covariate estimates appears in eTable 2 in the Supple-ment. The relative change associated with school closure forCOVID-19 incidence varied significantly by the testing rate per1000 residents, by the percentage of the state’s population aged65 years or older, by the number of nursing home residents per1000 people, and by urban density. Information on interpret-ing relative weekly changes appears in the eMethods in theSupplement.
The absolute effects associated with school closure dur-ing the 26-day period after school closure (days 17-42), whichwere calculated using model estimates with the assumptionof linear growth, yielded 638.7 cases per 100 000 that wouldhave occurred if schools had remained open (Table 3). Com-pared with the 214.8 cases per 100 000 estimated from theschool closure model, the absolute difference associated withschool closure was 423.9 (95% CI, 375.0 to 463.7) cases per
Figure 2. Modeled Association of School Closure With Coronavirus Disease 2019 (COVID-19) Incidence and Mortality
Daily incidence for all statesA
30
20
10
Daily
CO
VID-
19 in
cide
nce
per 1
00 0
00
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
30
20
10
Daily
CO
VID-
19 in
cide
nce
per 1
00 0
00
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
2.0
1.5
1.0
0.5
Daily
CO
VID-
19 d
eath
spe
r 100
000
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
2.0
1.5
1.0
0.5
0 0
0 0
Daily
CO
VID-
19 d
eath
spe
r 100
000
Time from school closure, d
End of lag period
–10 0 10 20 30 40 50
Mortality for all statesC
Daily incidence by cumulative incidence quartile at the timeof school closure
B
Mortality by cumulative incidence quartile at the time of school closureD
Highest quartileSecond highest quartileSecond lowest quartileLowest quartile
Highest quartileSecond highest quartileSecond lowest quartileLowest quartile
The data markers indicate the national unadjusted daily rates. The lines depictaggregated national daily rates adjusted for each state’s unique set of testingand demographic characteristics on each day of the study period with 95% CIsdepicted by gray lines. Six weeks after school closure, states in the lowestquartile had fewer new cases and fewer deaths compared with the states in thehighest quartile. Panels A and B were adjusted for all model componentsretained in the incidence model (intercept: percentage of state’s populationaged �15 years, percentage of state’s population aged �65 years, and USCenters for Disease Control and Prevention [CDC] social vulnerability index;before school closure: stay-at-home or shelter-in-place order, restaurant andbar closure, testing rate per 1000 residents, and urban density; after schoolclosure: testing rate per 1000 residents, stay-at-home or shelter-in-place order,
percentage of state’s population aged �65 years, number of nursing homeresidents per 1000 people, and urban density). Panels C and D were adjustedfor all model components retained in the mortality model (intercept:percentage of state’s population aged �15 years, percentage of state’spopulation aged �65 years, and CDC social vulnerability index; before schoolclosure: stay-at-home or shelter-in-place order, prohibition of gatherings with>10 people, restaurant and bar closure, percentage of state’s population aged�15 years, percentage of state’s population aged �65 years, number of nursinghome residents per 1000 people, and urban density; after school closure:restaurant and bar closure, number of nursing home residents per 1000 people,and urban density).
Research Original Investigation Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US
866 JAMA September 1, 2020 Volume 324, Number 9 (Reprinted) jama.com
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
100 000. States that closed schools late (in the highest quar-tile of cumulative incidence of COVID-19) had the largest ab-solute reduction in cases (621.7 [95% CI, 535.4 to 742.6] per100 000). However, states that closed schools earlier (in thelowest quartile) had fewer total cases (−128.7 [95% CI, −168.7to −74.2] per 100 000) during the period after school closure(Table 3). The absolute difference in COVID-19 incidence as-suming continued exponential growth appears in eTable 3 inthe Supplement.
Sensitivity Analyses Around the Lag PeriodThe point estimates ranged from −61% to −63% for the rela-tive change per week associated with school closure as the lagperiod varied. The point estimates for the relative change foreach quartile varied slightly across the lag period (eTable 4 inthe Supplement).
COVID-19 MortalityThe observed death rates in each state by quartile of cumula-tive incidence of COVID-19 at the time of school closure ap-pears in Figure 1B. In the unadjusted analyses during the pe-riod prior to potential effects of school closure (ie, during thelag period), the overall relative change in mortality per weekwas 171% (95% CI, 160% to 184%). In the unadjusted analy-ses, the relative mortality change per week associated withschool closure was −64% (95% CI, −67% to −61%). The unad-justed effect size associated with school closure varied byCOVID-19 cumulative incidence at the time of school closure,with states in the lowest quartile having the largest associ-ated effect size (Table 2).
In the adjusted analyses during the period prior to poten-tial effects of school closure (ie, during the lag period),COVID-19 mortality increased by 186% (95% CI, 175% to197%) per week (Table 2). The overall combined compositerelative weekly change in mortality after school closure was2% (95% CI, −8% to 14%; Table 2). This composite change inmortality after school closure is visually depicted in Figure 2.When examining only school closure, it was associated with arelative change per week in COVID-19 mortality of −58% (95%CI, −67% to −46%). This association was greatest in stateswith the lowest cumulative COVID-19 incidence at the time ofschool closure (relative change per week of −64% [95% CI,−73% to −52%]). In comparison, states that closed schoolslater when cumulative COVID-19 incidence was in the highestquartile had the smallest associated relative decline in mor-tality (−53% [95% CI, −63% to −40%]; Table 2 and Figure 2).The full model with all covariates appears in eTable 5 in theSupplement. The relative change in mortality associated withschool closure varied significantly by restaurant and bar clo-sure and urban density.
The absolute effects associated with school closure dur-ing the 16-day period after school closure (days 27-42), whichwere calculated using model estimates with the assumptionof linear growth, yielded 19.4 deaths per 100 000 that wouldhave occurred if schools had remained open (Table 4). Com-pared with the 6.8 deaths per 100 000 estimated from theschool closure model, the absolute difference associated withschool closure was 12.6 (95% CI, 11.8 to 13.6) deaths per 100 000Ta
ble
3.Es
timat
edAb
solu
teD
iffer
ence
sin
COVI
D-19
Case
saBe
twee
nPe
riod
ofSc
hool
Clos
ure
and
Scho
olsR
emai
ning
Ope
nU
sing
Line
arPr
ojec
tionb
COVI
D-19
case
sper
100
000
inav
erag
est
atec,
d
Estim
ated
abso
lute
diff
eren
ce(9
5%CI
)in
case
sper
100
000
byqu
artil
eof
COVI
D-19
cum
ulat
ive
inci
denc
eat
scho
olcl
osur
ed
Low
est
Seco
ndlo
wes
tSe
cond
high
est
Hig
hest
Scho
olsr
emai
nop
en(p
roje
cted
)
Befo
resc
hool
clos
ure
(dur
ing
lag
perio
d:da
ys1-
16)
37.0
(30.
5to
45.0
)23
.3(1
7.2
to31
.7)
31.8
(23.
1to
43.7
)32
.6(2
4.1
to44
.1)
81.2
(59.
5to
110.
7)
Afte
rlag
perio
d(d
ays1
7-42
)63
8.7
(604
.5to
686.
9)54
6.2
(509
.0to
598.
1)67
0.5
(620
.6to
741.
2)51
8.0
(475
.1to
578.
0)83
6.5
(750
.2to
957.
4)
Scho
olsc
lose
Befo
resc
hool
clos
ure
(dur
ing
lag
perio
d:da
ys1-
16)
37.0
(30.
5to
45.0
)23
.3(1
6.1
to30
.6)
31.8
(23.
1to
43.7
)32
.6(2
4.1
to44
.1)
81.2
(55.
5to
106.
8)
Afte
rlag
perio
d(d
ays1
7-42
)21
4.8
(175
.0to
263.
7)15
1.2
(103
.8to
198.
5)23
0.8
(165
.9to
321.
2)22
2.0
(163
.0to
302.
4)27
9.9
(189
.4to
370.
2)
Diff
eren
cein
inci
denc
edu
ring
26d
afte
rlag
perio
d
Betw
een
stat
esaf
ters
choo
lclo
sure
and
hypo
thet
ical
stat
esw
here
scho
olsr
emai
ned
open
423.
9(3
75.0
to46
3.7)
331.
4(2
94.2
to38
3.3)
455.
6(4
05.7
to52
6.4)
303.
2(2
60.3
to36
3.2)
621.
7(5
35.4
to74
2.6)
Betw
een
stat
esth
atcl
osed
scho
olsa
fter
the
lag
perio
dan
dst
ates
inth
ehi
ghes
tqua
rtile
NA
–128
.7(–
168.
7to
–74.
2)–4
9.1
(–11
4.0
to41
.3)
–57.
8(–
116.
8to
22.6
)0
[Ref
eren
ce]
Abbr
evia
tions
:CO
VID
-19,c
oron
aviru
sdise
ase
2019
;NA,
nota
pplic
able
.a
Adju
sted
fora
llmod
elco
mpo
nent
sret
aine
din
incid
ence
mod
el(in
terc
ept:
perc
enta
geof
stat
e’spo
pula
tion
aged
�15
year
s,pe
rcen
tage
ofst
ate’s
popu
latio
nag
ed�
65ye
ars,
and
socia
lvul
nera
bilit
yin
dex;
befo
resc
hool
closu
re:
stay
-at-h
ome
orsh
elte
r-in-
plac
eor
der,
rest
aura
ntan
dba
rclo
sure
,tes
ting
rate
per1
000
resid
ents
,and
urba
nde
nsity
;afte
rsch
oolc
losu
re:t
estin
gra
tepe
r100
0re
siden
ts,s
tay-
at-h
ome
orsh
elte
r-in-
plac
eor
der,
perc
enta
geof
stat
e’spo
pula
tion
aged
�65
year
s,nu
mbe
rofn
ursin
gho
me
resid
ents
per1
000
peop
le,a
ndur
ban
dens
ity).
bU
sed
slope
from
line
tang
entia
lto
the
curv
eat
the
end
ofth
ela
gpe
riod.
cCu
mul
ativ
ein
cide
nce
atth
etim
eof
scho
olcl
osur
efo
rsta
teis
sett
oth
ead
just
edm
ean.
dSt
ate
with
allc
hara
cter
istic
sset
toth
em
ean
valu
efo
rallc
ovar
iate
sret
aine
din
the
mod
el.
Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US Original Investigation Research
jama.com (Reprinted) JAMA September 1, 2020 Volume 324, Number 9 867
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
(Table 4). States that closed schools late (in the highest quar-tile of COVID-19 cumulative incidence) had the largest abso-lute reduction in deaths (15.8 [95% CI, 13.9 to 18.1] per 100 000).However, states that closed schools earlier (in the lowest quar-tile) had fewer estimated total deaths (−1.5 [95% CI, −2.7 to −0.1]per 100 000) during the period after school closure (Table 4).The absolute difference in deaths assuming continued expo-nential growth appears in eTable 3 in the Supplement.
Sensitivity Analyses Around the Lag PeriodThe point estimates ranged from −55% to −61% for the rela-tive change associated with school closure as the lag period var-ied. The point estimates for the relative change for each quar-tile varied slightly across the lag period (eTable 4 in theSupplement).
DiscussionBetween March 9, 2020, and May 7, 2020, school closure inthe US was temporally associated with decreased COVID-19 in-cidence and mortality. States that closed schools earlier (whenthe state’s cumulative incidence was lower) had the largest rela-tive reduction in overall incidence and mortality.
In March 2020, states enacted multiple nonpharmaceu-tical interventions, including closing schools, nonessentialbusinesses, and restaurants and bars, and prohibiting largegatherings, to curb SARS-CoV-2 spread and prevent death.Completely isolating the effects of any single nonpharmaceu-tical intervention is impossible because recommendations forincreased handwashing, cleaning, and wearing of masksevolved simultaneously. Measured COVID-19 incidence alsowas affected by testing availability, which was limited early inthe pandemic and varied nationally.
In this study, changes in COVID-19 incidence and mortal-ity associated with school closure were isolated to the extentpossible by adjusting for other state-enacted policies and test-ing rates. In adjusted models, school closure was associatedwith decreased COVID-19 incidence and deaths. These analy-ses do not incorporate the risks of school closure on child edu-cation and development or from a societal perspective. How-ever, the analyses suggest that school closure may be effectivein curbing SARS-CoV-2 spread and preventing deaths duringfuture outbreaks.
These findings complement evolving evidence on the roleof children in the transmission of SARS-CoV-2. Studies havedocumented lower attack rates for children,28 and childrencomprise a small proportion of documented infections.29 Chil-dren may be less susceptible to SARS-CoV-2 infection30; how-ever, studies have documented viral shedding in asymptom-atic children.31 Recent studies suggest school closure may haveonly modest effects on COVID-19 deaths.32-35 School closurein this study was associated with a −62% relative change inCOVID-19 incidence per week. A decline of 62% was equiva-lent to 39% of the projected value with schools open. So, perweek, the incidence was estimated to have been 39% of whatit would have been had schools remained open. Extrapolat-ing the absolute differences of 423.9 cases and 12.6 deaths perTa
ble
4.Es
timat
edAb
solu
teD
iffer
ence
sin
COVI
D-19
Dea
thsa
Betw
een
Scho
olCl
osur
ean
dSc
hool
sRem
aini
ngO
pen
Usi
ngLi
near
Proj
ectio
nb
COVI
D-19
deat
hspe
r100
000
inav
erag
est
atec,
d
Estim
ated
abso
lute
diff
eren
ce(9
5%CI
)in
deat
hspe
r100
000
byqu
artil
eof
COVI
D-19
cum
ulat
ive
inci
denc
eat
scho
olcl
osur
ed
Low
est
Seco
ndlo
wes
tSe
cond
high
est
Hig
hest
Scho
olsr
emai
nop
en(p
roje
cted
)
Befo
resc
hool
clos
ure
(dur
ing
lag
perio
d:da
ys1-
26)
3.2
(2.9
to3.
5)2.
5(2
.2to
3.0)
3.1
(2.6
to3.
7)2.
6(2
.2to
3.0)
5.5
(4.7
to6.
5)
Afte
rlag
perio
d(d
ays2
7-42
)19
.4(1
8.6
to20
.4)
19.2
(17.
9to
20.8
)22
.1(2
0.5
to24
.1)
14.6
(13.
5to
15.9
)22
.6(2
0.6
to24
.9)
Scho
olsc
lose
Befo
resc
hool
clos
ure
(dur
ing
lag
perio
d:da
ys1-
26)
3.2
(2.9
to3.
5)2.
6(2
.1to
3.0)
3.1
(2.6
to3.
7)2.
6(2
.2to
3.0)
5.5
(4.7
to6.
5)
Afte
rlag
perio
d(d
ays2
7-42
)6.
8(5
.8to
7.9)
6.3
(5.0
to7.
7)7.
6(6
.1to
9.4)
5.6
(4.7
to6.
8)7.
9(6
.3to
9.5)
Diff
eren
cein
inci
denc
edu
ring
16d
afte
rlag
perio
d
Betw
een
stat
esaf
ters
choo
lclo
sure
and
hypo
thet
ical
stat
esw
here
scho
olsr
emai
ned
open
12.6
(11.
8to
13.6
)12
.4(1
1.1
to14
.0)
15.3
(13.
7to
17.3
)7.
8(6
.7to
9.1)
15.8
(13.
9to
18.1
)
Betw
een
stat
esth
atcl
osed
scho
olsa
fter
the
lag
perio
dan
dst
ates
inth
ehi
ghes
tqua
rtile
NA
–1.5
(–2.
7to
–0.1
)–0
.3(–
1.8
to1.
5)–2
.3(–
3.2
to–1
.1)
0[R
efer
ence
]
Abbr
evia
tions
:CO
VID
-19,c
oron
aviru
sdise
ase
2019
;NA,
nota
pplic
able
.a
Adju
sted
fora
llmod
elco
mpo
nent
sret
aine
din
mor
talit
ym
odel
(inte
rcep
t:pe
rcen
tage
ofst
ate’
spop
ulat
ion
aged
�15
year
s,pe
rcen
tage
ofst
ate’
spop
ulat
ion
aged
�65
year
s,an
dso
cial
vuln
erab
ility
inde
x;be
fore
scho
olcl
osur
e:st
ay-a
t-ho
me
orsh
elte
r-in-
plac
eor
der,
proh
ibiti
onof
gath
erin
gsw
ith>1
0pe
ople
,res
taur
anta
ndba
rcl
osur
e,pe
rcen
tage
ofst
ate’
spop
ulat
ion
aged
�15
year
s,pe
rcen
tage
ofst
ate’
spop
ulat
ion
aged
�65
year
s,nu
mbe
rofn
ursin
gho
me
resid
ents
per1
00
0pe
ople
,and
urba
nde
nsity
;aft
ersc
hool
clos
ure:
rest
aura
ntan
dba
rcl
osur
e,nu
mbe
rofn
ursin
gho
me
resid
ents
per1
00
0pe
ople
,and
urba
nde
nsity
).
bU
sed
slope
from
line
tang
entia
lto
the
curv
eat
the
end
ofth
ela
gpe
riod.
cCu
mul
ativ
ein
cide
nce
atth
etim
eof
scho
olcl
osur
efo
rsta
teis
sett
oth
ead
just
edm
ean.
dSt
ate
with
allc
hara
cter
istic
sset
toth
em
ean
valu
efo
rallc
ovar
iate
sret
aine
din
the
mod
el.
Research Original Investigation Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US
868 JAMA September 1, 2020 Volume 324, Number 9 (Reprinted) jama.com
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
100 000 to 322.2 million residents nationally suggests thatschool closure may have been associated with approximately1.37 million fewer cases of COVID-19 over a 26-day period and40 600 fewer deaths over a 16-day period; however, these fig-ures do not account for uncertainty in the model assump-tions and the resulting estimates.
The analyses presented here suggest that the timing ofschool closure plays a role in the magnitude of changes asso-ciated with school closure. As hypothesized, school closure instates that enacted this intervention early (when the cumula-tive incidence of COVID-19 was low) had greater associated rela-tive decreases in incidence and mortality. Although these rela-tive differences translate into smaller absolute differencesassociated with school closure, states that closed schools later(in the highest quartile of COVID-19 cumulative incidence) hadmore new cases and deaths from COVID-19 during the periodafter school closure. Thus, this study can inform future deci-sions about optimal timing for state and local officials to con-sider school closure to curb SARS-CoV-2 spread in the high like-lihood that the pandemic continues.
The mechanism by which school closure could affectCOVID-19 spread is not only through disrupting spread by oramong children. School closure affects family routines,necessitating alternative childcare and modified work sched-ules. These changes are evident by the number of telework-ers more than doubling.36,37 The disruption in everyday lifelikely influenced how people engaged in group activities,traveled, and conducted business. If the primary effect asso-ciated with school closure is related to altered adult behavior,and not children spreading the virus to adults, the primarylag period considered in these analyses should be adjusted.Eliminating the 4 days for a child to adult transmission wouldresult in a COVID-19 incidence lag period of 12 days and amortality lag period of 22 days. In sensitivity analyses, theeffect sizes associated with school closure at these shorter lagperiods were similar to the primary analysis effect sizes. Thedegree to which the associations with school closure relate todecreased spread of SARS-CoV-2 by children or a combina-tion of child and adult factors is unclear. Because school clo-sure was the first nonpharmaceutical intervention in moststates, the effects associated with school closure may belarger than if school closure had followed other nonpharma-ceutical interventions.
It is unclear how COVID-19 spread would be affected ifschools remained open while states enacted other policies torestrict movement. It is possible school-related spread may be
mitigated with infection-control interventions recom-mended by the CDC and the American Academy of Pediat-rics, including frequent handwashing, universal mask poli-cies, physical distancing measures, and increased sanitationprocedures.38,39 However, given that school closure also al-ters adult behavior, decreasing COVID-19 spread within schoolsmay be inadequate as a stand-alone intervention and may re-quire continued alteration of adult interactions.
LimitationsThis study has several limitations. First, many states enactedadditional nonpharmaceutical interventions concurrently withor shortly after school closure, making it impossible to fullyisolate potential effects of school closure. Some nonpharma-ceutical interventions, such as increased handwashing, couldnot be included due to lack of available data.
Second, analyses were conducted at the state level. Theanalyses did not account for resident travel leading to viralspread between states. Even though the study modeled state-level policies, some states had more restrictive policies lo-cally (ie, by county). Nevertheless, these analyses are usefulto understand the practical implications of state policy in con-taining spread.
Third, inadequate testing has impeded COVID-19 diagno-sis. Testing variability was accounted for with the use of state-level testing rates as a model covariate; however, testing ratesdo not fully capture a state’s testing capability, infrastruc-ture, and strictness of testing guidelines.
Fourth, the completeness and accuracy of the Johns Hop-kins University database with respect to COVID-19 incidenceand mortality has not been established. This data source ag-gregates publicly available data and accuracy may vary stateto state. As with limitations in testing, inconsistencies in re-porting are unavoidable limitations of all COVID-19 US popu-lation-based studies.
ConclusionsBetween March 9, 2020, and May 7, 2020, school closure inthe US was temporally associated with decreased COVID-19 in-cidence and mortality; states that closed schools earlier, whencumulative incidence of COVID-19 was low, had the largest rela-tive reduction in incidence and mortality. However, it re-mains possible that some of the reduction may have been re-lated to other concurrent nonpharmaceutical interventions.
ARTICLE INFORMATION
Accepted for Publication: July 17, 2020.
Published Online: July 29, 2020.doi:10.1001/jama.2020.14348
Author Affiliations: Division of Hospital Medicine,Cincinnati Children’s Hospital Medical Center,Cincinnati, Ohio (Auger, Shah, Warniment,Timmons, Bosse, Ferris, Brady, Schondelmeyer,Thomson); James M. Anderson Center for HealthSystems Excellence, Cincinnati Children’s HospitalMedical Center, Cincinnati, Ohio (Auger, Shah,Hartley, Brady, Schondelmeyer); Department of
Pediatrics, University of Cincinnati College ofMedicine, Cincinnati, Ohio (Auger, Shah, Hartley,Brady, Schondelmeyer, Thomson); PediatricResearch in Inpatient Settings Network, Cincinnati,Ohio (Auger, Shah, Richardson, Hall, Brady,Thomson); Division of Infectious Diseases,Cincinnati Children’s Hospital Medical Center,Cincinnati, Ohio (Shah); Division of EmergencyMedicine, Cincinnati Children’s Hospital MedicalCenter, Cincinnati, Ohio (Warniment).
Author Contributions: Dr Richardson had fullaccess to all of the data in the study and takes
responsibility for the integrity of the data and theaccuracy of the data analysis. Drs Brady,Schondelmeyer, and Thomson made equalsubstantial contributions to the manuscript and areco-last authors.Concept and design: Auger, Shah, Richardson,Hartley, Hall, Warniment, Bosse, Ferris, Brady,Schondelmeyer, Thomson.Acquisition, analysis, or interpretation of data: Allauthors.Drafting of the manuscript: Auger, Richardson,Timmons, Brady, Schondelmeyer.
Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US Original Investigation Research
jama.com (Reprinted) JAMA September 1, 2020 Volume 324, Number 9 869
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021
Critical revision of the manuscript for importantintellectual content: Shah, Richardson, Hartley, Hall,Warniment, Bosse, Ferris, Brady, Schondelmeyer,Thomson.Statistical analysis: Auger, Richardson, Hartley, Hall.Administrative, technical, or material support:Hartley, Warniment, Timmons, Bosse, Ferris,Thomson.Supervision: Brady, Schondelmeyer, Thomson.
Conflict of Interest Disclosures: None reported.
Funding/Support: Funding for this work wasprovided by Agency for Healthcare Research andQuality awards K08HS024735 (Dr Auger),K08HS023827 (Dr Brady), K08HS026763 (DrSchondelmeyer), and K08HS025138 (Dr Thomson)and award 5UL1TR001425-04 from the NationalCenter for Advancing Translational Sciences,National Institutes of Health.
Role of the Funder/Sponsor: The Agency forHealthcare Research and Quality and the NationalInstitutes of Health had no role in the design andconduct of the study; collection, management,analysis, and interpretation of the data;preparation, review, or approval of the manuscript;and decision to submit the manuscript forpublication.
Disclaimer: The content is solely the responsibilityof the authors and does not represent the officialviews of the Agency for Healthcare Research andQuality, the National Institutes of Health, CincinnatiChildren’s Hospital Medical Center, the Children’sHospital Association, or the Pediatric Research inInpatient Settings Network.
Additional Information: Drs Richardson and Hallare affiliated with the Children’s HospitalAssociation, Lenexa, Kansas.
REFERENCES
1. Pan A, Liu L, Wang C, et al. Association of publichealth interventions with the epidemiology of theCOVID-19 outbreak in Wuhan, China. JAMA. 2020;323(19):1915-1923.
2. Dong Y, Mo X, Hu Y, et al. Epidemiology ofCOVID-19 among children in China. Pediatrics.2020;145(6):e20200702.
3. Lu X, Zhang L, Du H, et al. SARS-CoV-2 infectionin children. N Engl J Med. 2020;382(17):1663-1665.
4. Worby CJ, Chaves SS, Wallinga J, et al. On therelative role of different age groups in influenzaepidemics. Epidemics. 2015;13:10-16.
5. Cauchemez S, Valleron AJ, Boëlle PY, et al.Estimating the impact of school closure oninfluenza transmission from sentinel data. Nature.2008;452(7188):750-754.
6. Schanzer D, Vachon J, Pelletier L. Age-specificdifferences in influenza A epidemic curves. Am JEpidemiol. 2011;174(1):109-117.
7. Luca G, Kerckhove KV, Coletti P, et al. The impactof regular school closure on seasonal influenzaepidemics. BMC Infect Dis. 2018;18(1):29.
8. Chu Y, Wu Z, Ji J, et al. Effects of school breakson influenza-like illness incidence in a temperateChinese region. BMJ Open. 2017;7(3):e013159.
9. United Nations. Policy brief: the impact ofCOVID-19 on children. Accessed May 11, 2020.https://www.un.org/sites/un2.un.org/files/policy_brief_on_covid_impact_on_children_16_april_2020.pdf
10. Dong E, Du H, Gardner L. An interactiveweb-based dashboard to track COVID-19 in realtime. Lancet Infect Dis. 2020;20(5):533-534.
11. US Census Bureau. American CommunitySurvey. Accessed April 2, 2020. https://www.census.gov/programs-surveys/acs
12. COVID Tracking Project. Data API. AccessedMay 8, 2020. https://covidtracking.com/api
13. US Centers for Disease Control and Prevention.Behavioral Risk Factor Surveillance System: 2018data. Accessed June 15, 2020. https://www.cdc.gov/brfss/annual_data/annual_2018.html
14. Kaiser Family Foundation. Total number ofresidents in certified nursing facilities. AccessedJune 15, 2020. https://www.kff.org/f23fa35/
15. US Centers for Disease Control and Prevention.Social vulnerability index. Accessed June 15, 2020.https://svi.cdc.gov
16. Lauer SA, Grantz KH, Bi Q, et al. The incubationperiod of coronavirus disease 2019 (COVID-19)from publicly reported confirmed cases. Ann InternMed. 2020;172(9):577-582.
17. Colorado Department of Public Health andEnvironment. Get tested. Accessed March 23,2020. https://covid19.colorado.gov
18. State of Connecticut. Connecticut COVID-19response. Accessed March 30, 2020. https://portal.ct.gov/Coronavirus
19. Georgia Department of Public Health.COVID-19. Accessed March 23, 2020. https://dph.georgia.gov
20. Indiana State Department of Health. 2019novel coronavirus (COVID-19). Accessed March 31,2020. https://coronavirus.in.gov
21. Iowa Department of Public Health. Novelcoronavirus acute respiratory disease: COVID-19.Accessed March 26, 2020. https://idph.iowa.gov
22. Minnesota Department of Health. Coronavirusdisease 2019 (COVID-19). Accessed March 24,2020. https://www.health.state.mn.us/diseases/coronavirus/index.html
23. Montana Department of Public Health andHuman Services. 2019 novel coronavirus(COVID-19). Accessed March 27, 2020. https://dphhs.mt.gov
24. New Hampshire Department of Health andHuman Services. Novel coronavirus 2019(COVID-19). Accessed March 25, 2020. https://www.nh.gov/covid19
25. South Carolina Department of Health andEnvironmental Control. Coronavirus disease 2019(COVID-19). Accessed April 3, 2020. https://www.scdhec.gov/infectious-diseases/viruses/coronavirus-disease-2019-covid-19
26. Garg S, Kim L, Whitaker M, et al. Hospitalizationrates and characteristics of patients hospitalizedwith laboratory-confirmed coronavirus disease
2019. MMWR Morb Mortal Wkly Rep. 2020;69(15):458-464.
27. Beigel JH, Tomashek KM, Dodd LE, et al.Remdesivir for the treatment of Covid-19. N Engl JMed. Published online May 22, 2020. doi:10.1056/NEJMoa2007764
28. Li W, Zhang B, Lu J, et al. The characteristics ofhousehold transmission of COVID-19. Clin Infect Dis.2020;ciaa450.
29. Gudbjartsson DF, Helgason A, Jonsson H, et al.Spread of SARS-CoV-2 in the Icelandic population.N Engl J Med. 2020;382(24):2302-2315.
30. Bunyavanich S, Do A, Vicencio A. Nasal geneexpression of angiotensin-converting enzyme 2 inchildren and adults. JAMA. 2020;323(23):2427-2429.
31. Zhang J, Litvinova M, Liang Y, et al. Changes incontact patterns shape the dynamics of theCOVID-19 outbreak in China. Science. 2020;368(6498):1481-1486.
32. Ferguson NM, Laydon D, Nedjati-Gilana G, et al.Report 9: impact of non-pharmaceuticalinterventions (NPIs) to reduce COVID-19 mortalityand healthcare demand. Accessed July 20, 2020.https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf
33. Viner RM, Russell SJ, Croker H, et al. Schoolclosure and management practices duringcoronavirus outbreaks including COVID-19. LancetChild Adolesc Health. 2020;4(5):397-404.
34. Esposito S, Principi N. School closure duringthe coronavirus disease 2019 (COVID-19)pandemic. JAMA Pediatr. Published online May 13,2020. doi:10.1001/jamapediatrics.2020.1892
35. Hsiang S, Allen D, Annan-Phan S, et al. Theeffect of large-scale anti-contagion policies on theCOVID-19 pandemic. Nature. Published online June8, 2020. doi:10.1038/s41586-020-2404-8
36. Koetsier J. 58% of American knowledgeworkers are now working remotely. Accessed June9, 2020. https://www.forbes.com/sites/johnkoetsier/2020/03/20/58-of-american-knowledge-workers-are-now-working-remotely/#517af4b13303
37. Rosalsky G. COVID-19 forces more people towork from home: how’s it going? Accessed June 9,2020. https://www.npr.org/2020/05/08/852527736/covid-19-forces-more-people-to-work-from-home-hows-it-going
38. US Centers for Disease Control and Prevention.Schools during the COVID-19 pandemic. AccessedMay 15, 2020. https://www.cdc.gov/coronavirus/2019-ncov/community/schools-childcare/Schools-Decision-Tree.pdf
39. American Academy of Pediatrics. COVID-19planning considerations: guidance for schoolre-entry. Accessed July 20, 2020. https://services.aap.org/en/pages/2019-novel-coronavirus-covid-19-infections/clinical-guidance/covid-19-planning-considerations-return-to-in-person-education-in-schools/
Research Original Investigation Association Between US Statewide School Closure and COVID-19 Incidence and Mortality in the US
870 JAMA September 1, 2020 Volume 324, Number 9 (Reprinted) jama.com
© 2020 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ by a Non-Human Traffic (NHT) User on 08/26/2021