Content uploaded by Dongying Li
Author content
All content in this area was uploaded by Dongying Li on Jun 18, 2023
Content may be subject to copyright.
CORRECTED PROOF
Review
Climatic and meteorological exposure and mental and behavioral health: A
systematic review and meta-analysis
DongyingLia, ,YueZhanga,XiaoyuLia,KaiZhangb,YiLuc,RobertD. Browna
a DepartmentofLandscapeArchitectureandUrbanPlanning,TexasA&MUniversity,CollegeStation,TX77843,USA
b DepartmentofEnvironmentalHealthSciences,SchoolofPublicHealth,UniversityatAlbany,StateUniversityofNewYork,Albany,NY12222,USA
c DepartmentofArchitectureandCivilEngineering,CityUniversityofHongKong,HongKong
A R T I C L E I N F O
Editor:SCOTTSHERIDAN
Keywords:
Climate
Mentalhealth
Schizophrenia
Mooddisorders
Neuroticdisorders
meta-analysis
A B S T R A C T
Asclimatechangeexertswideranginghealthimpacts,thereisasurgeofinterestintheassociationsbetweencli-
maticfactorsandmentalandbehavioraldisorders(MBDs).Existingquantitativesynthesesfocusmainlyonheat
andhightemperatureexposure,neglectingtheeffectsofotherclimaticfactorsandtheirsynergies.Theobjective
ofthisstudyistoconductasystematicreviewandmeta-analysisoftheevidenceofassociationsbetweenclimatic
exposureandcombinedmentalandbehavioralhealthconditionsandspecificmentaldisorders(e.g.,schizophre-
nia,dementia).
Asystematic searchwasconductedApril11
–
16,2022 usingWebofScience,Medline, ProQuest,EMBASE,
PsycINFO,CINAHL,andEnvironmentComplete.Screeningandeligibilityscreeningfollowedinclusioncriteria
basedonpopulation,exposure,comparator,andoutcomeguidelines.Riskofbiasassessmentwasperformed,a
narrativesynthesiswasfirstpresentedforallstudies,andrandom-effectmeta-analyseswereperformedwhenat
leastthreestudieswereavailableforaspecificexposure-outcomepair.Certaintyofevidencewasevaluatedfol-
lowingtheGradingofRecommendationsAssessment,DevelopmentandEvaluation(GRADE)tool.
Thesearchprocessyielded7696initialresults,fromwhichweidentified88studiestoincludeinthereview
set.Climaticfactorsreportedincludedairtemperature,solarradiation/sunshine,barometricpressure,precipita-
tion,relativehumidity,winddirection/speed,andthermalcomfortindex.OutcomesincludedMBDincidences
(e.g.,schizophrenia,mooddisorders,neuroticdisorders),mentalhealth-relatedmortality,andself-reportedpsy-
chologicalstates.Meta-analysisshowedthatheatwaves(pooledRR=1.05,95%CI=1.02
–
1.08)andextreme
hightemperatures(99thpercentile:pooledRR=1.18,95%CI=1.08
–
1.29)wereassociatedwithhigherrisk
ofMBD.Coldextremes,however,werenotassociatedwithMBDrisk.Thefindingsfurtheridentifiedanassocia-
tion between increases in a thermal index (i.e., apparent temperature) and elevated risk of MBD (pooled
RR=1.06,95%CI=1.03
–
1.12);specifically,a99thpercentilehightemperaturewasassociatedwithin-
creasedschizophreniarisk(pooledRR=1.07,95%CI=1.01
–
1.12).
Riskofbiasassessmentshowedmoststudiestohavelowormoderatelylowrisks,whileafewstudieswere
ratedprobablyhighinconfounding,selectionbias,outcomemeasurement,andreportingbias.GRADEevalua-
tionrevealedmoderatecertaintyofevidenceonthermalcomfortindexandMBD,butlowcertaintyrelatedtoair
temperatureorsunshineduration.Thesefindingscallattentiontotheheterogeneityofexposuremeasuresand
theutilityofthermalindicesthatconsiderthesynergisticeffectsofmeteorologicalfactors.Methodologicalcon-
cernssuchasthelinearityassumptionandcumulativeeffectsarediscussed.
1. Introduction
Climate extremes have adverse effects on health. With climate
change, climate extremes are expected to increase in intensity, fre-
quency,andhealthimpact(Petkovaetal.,2013;Guoetal.,2018).Such
extremeeventsarenotisolatedincidences,butarereflectiveoflarger-
scale, often persistent changes in the thermodynamic environment
acrosscitiesandcountryside(Trenberthetal.,2015).Infact,climate
changeisalreadyaffectingeveryinhabitedregionworldwide,causing
changesinecosystemservices(Shawetal.,2011)andambientfactors
towhichhumansareroutinelyexposedineverydaylife,suchastem-
perature,relativehumidity,andterrestrialradiation.Forexample,ob-
⁎
Correspondingauthor.
E-mailaddress:dli@arch.tamu.edu(D.Li).
https://doi.org/10.1016/j.scitotenv.2023.164435
Received21 January 2023;Received in revised form22 May 2023;Accepted22 May 2023
0048-9697/© 20XX
Edit Preview PDF
This is the pre-proof version of the published manuscript generated by the publisher &
recompiled by the authors. Some content may be slightly different from the final published
version.
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
servedglobalsurfacetemperatureincreasedby>1.2°Cfrom1850to
2020, with the steepest increase occurring after 1990 (Masson-
Delmotteetal.,2021).Atthemicroclimatescale,thisincreaseinsur-
facetemperatureisdirectlylinkedtorisingmeanradianttemperature
andhumanbodyheatfluxintheenvironment(GálandKántor,2020).
Strongevidencehas associated climate extremeswithcardiovascular
disease,respiratorydisease,infectiousdisease,andotherphysicaldis-
eases(Bragaetal.,2002;Medina-Ramónetal.,2006;Wuetal.,2016;
Obradovichet al.,2018a).Moreover,climatechangemay exacerbate
health disparities by affecting minority groups and socially-
disadvantagedgroupsdisproportionately(Voelkeletal.,2018;Lietal.,
2022a).Thus,acomprehensiveunderstandingoftherelationshipsbe-
tweenmeteorologicalfactorsandhealthisaresearchprioritytotheur-
gentsocietal goals of climate changeadaptationanddisparityreduc-
tion.
Intherecentdecade,epidemiologicalstudieshaverevealedcritical
links between meteorological conditions, especially temperature ex-
tremes,andmentalillness(Hayesetal.,2018;Berryetal.,2010).Re-
searchhasidentifiedimpactsofthreetypesofclimate-relatedevents
(i.e., acute, subacute, and long-lasting changes) on mental health
(PalinkasandWong,2020).Forexample,high andlowtemperatures
mightbefactorsthatcontributetomentaldisorders(Obradovichetal.,
2018b; Shiloh et al., 2005; Williams et al., 2012; Yoo et al., 2021;
Zhangetal.,2020),andheateventsthatoccurduetoincrementaltem-
peraturechangesinsummercouldposeasalientrisktohumanmental
conditions().Todate,severalsystematicreviewshaveassessedtheas-
sociationsbetweenextremeweathereventsandhumanmentalhealth.
Ratajet al.(Ratajetal.,2016)investigatedthe prevalence ofmental
disordersindevelopingcountries(exceptAfricanregions)duringex-
tremeweatherevents(i.e.,stormsandflooding)with17articlespub-
lishedby2014.Rotheretal.(Rotheretal.,2021)filledtheresearchgap
inAfricatosomeextentandexaminedexistingfindingsontheimpact
offloodingonchildandadolescentmentalhealthbasedonasystematic
reviewthat identified only two articles meeting criteria.Focusingon
Europeancountries,Cruzetal.(Cruzetal.,2020)andWeilnhammeret
al.(Weilnhammeretal.,2021)examinedthefindingsfrom21and35
articlespublishedupto2019
–
2020,respectively,andsynthesizedthe
mentalhealthimpactsarisingfromextremeweathereventsinvolving
heat/coldwaves,droughts,wildfires,andfloods.Furthermore,system-
aticreviewsfocusingonextremeheathavesummarizedtheepidemio-
logicalevidenceoftheimpactofhightemperatureandheatwaveson
mentalhealth-relatedmorbidityandmortality(Thompsonetal.,2018;
Liuetal.,2021).Thompsonand colleagues summarized thefindings
from34articlespublishedthrough2017andpresentedstronglinksbe-
tweenhightemperaturesandsuiciderisk(Thompsonetal.,2018).No-
tably,Liuetal.(Liuetal.,2021)conductedasystematicreviewof53
articlespublishedbetween1990and2020andameta-analysisof41by
poolingtheeffectsizesofhighambienttemperaturesandreporteda
0.9%increase in mentalhealth-related morbidity for every1 °C in-
creaseintemperature.
Therecent surgeofpublishedresearchin climatedeterminantsof
mentalhealthwarrantsacomprehensivequantitativesynthesis.Specif-
ically,severalmajorresearchgapsremaintobeaddressedwithasys-
tematicreview.Existingreviewsoftenfocusonhotweatherorextreme
heatconditionsasthesoleexposureofinterest.Althoughtheirfindings
providesolidgroundforarelationshipbetweenambienttemperatures
andmentalillness,wehavealimitedunderstandingoftheinfluencesof
multiple,orthecombinedeffects,ofmeteorologicalvariablesonhu-
man-environmentheatfluxandhealthoutcomes.Accordingtobiome-
teorology,humanheatandcoldregulationdependsonabalancedheat
budget(Jietal.,2022;BrownandGillespie,1986).Environmentalfac-
torsthatinfluencehumanheat/coldstressinclude:radiation,tempera-
ture,vaporpressure,anddiffusionconductancetoheatandvapor.As
such, widely accepted thermophysiological comfort and heat/cold
stressmodels consider fourmeteorologicaldeterminantsairtempera-
ture,humidity,windspeed,andsolarradiation(Jietal.,2022),along
withindividualcharacteristicssuchasBMI,physicalactivity,andcloth-
ing insulation (Zhao et al., 2021; Höppe, 1999a). Research has re-
viewedthat,controllingforairtemperature,otherclimatefactorssuch
ashumidity and radiationaffectshumanthermalsensationandcom-
fort.Therelationshipsamongtemperature,humidity,andradiationare
non-linear(Lietal.,2018)andtheirrelativeimportanceintheircontri-
butiontothermalsensationvaryacrossseasons(e.g.,airtemperature
contributesstronglyinsummerwhileradiationcontributesstronglyin
winter)(Liuetal.,2016).Assuch,acomprehensivereviewthatcriti-
callyassesses the fullrangeofmeteorologicalfactorsassociatedwith
mentalhealthiswarranted.
Itisalsoworthnotingthatpreviousempiricalarticlesandreviews
outliningtheimportanceofclimateandweatherconditionsoftendif-
ferentiatemorbidity/hospital admissions and mortality, but consider
mentaldisorderas one outcomecategory without differentiating the
specificdiagnoses/subdiagnoses(Williamsetal.,2012;Charlsonetal.,
2021).Suchanapproachmaynotcapturecomplexitiesrelatingtothe
variouscausesandsymptomologyofeachdifferenttypesofMBD(e.g.,
schizophrenia,depression,andpost-traumaticstressdisorder).Forex-
ample,inarecentreviewandmeta-analysis,Liuetal.presenteddiffer-
enteffectsizesbetweenhightemperaturesandspecificMBDs,suggest-
ingvariationsinthepresence,magnitude,andpossiblyevendirection
andmechanismoftherelationships(Liuetal., 2021). In addition to
temperature, some mental disorders (e.g., bipolar disorders, depres-
sion,andschizophrenia)maybestronglyrelatedtooneoracombina-
tionofmultipleclimaticconditionsbutnotothers,suchassunlightin-
tensity, humidity, and wind conditions (Gu et al., 2019; Kim et al.,
2021;Leeetal.,2007;Molin et al., 1996). As such,estimatedeffect
sizesmaydifferacrossdifferenttypesofmentalhealthconditions,re-
quiringamorenuancedreviewthatconsiderseachexposure-outcome
pair.Inaddition,asthecontemporarydefinitionofhealthemphasizes
notonlytheabsenceofdiseasebutalsopositivemoodstatesandhappi-
ness (World Health Organization, 2002), evidence on mental health
outcomesotherthanmorbidityandmortalityshouldbeincluded.
Furthermore,climatechangeandclimaticfactorsaffecteachregion
differently.MorerecentstudieswereconductedintheGlobalSouth,ex-
panding our understanding of the global mental health impact of
weathervariability;ithasbeensuggestedthattherelativeriskofmen-
talhealth-relatedmortalitymightvaryindifferentclimatezones,e.g.,
higherintropicalzonesthaninsubtropicalorcontinentalzones(Liuet
al.,2021).Givensuchpotentialgeographicalvariability,itiscriticalto
conductareviewthatidentifiesthegeographicalandclimatezonesthat
yetremainunderrepresented.
Toourknowledge,noreviewhasprovidedaquantitativesynthesis
regardingtherelationshipsofthefullsetofclimateandmeteorological
factorswithmentalhealthandMBD.Consideringthegrowingbodyof
evidenceonthistopic,thecurrentstudyaimstodepictacompletepic-
tureoftheserelationships,modelpooledeffectsizes,discussresearch
gapsandrisk of bias, andpropose future directions. Thespecific re-
searchquestionsexaminedinclude:
1. What are the geographical, temporal, and to pical trends in the
literatureonclimatefactorsandmentalhealth?
2. Whatarethedirectionandmagnitudeoftherelationshipbetween
each climate/meteorological factor and mental and behavioral
disorders?
3. How strong and consistent are the relationships when each
mentalandbehavioraldisorderisassessed?
4. Whataretheknowledgegapsintheexistingliterature,andwhat
arethemethodologicalconcernsrelatingtostudybias?
2
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
2. Methods
2.1. PECOframeworkandliteraturesearch
TheprotocolwasregisteredattheInternationalProspectiveRegis-
ter of Systematic Reviews database (PROSPERO, registered ID:
CRD42022321928).Wedevelopedthesystematicreviewprotocolfol-
lowingthePreferredReportingItemsforSystematicReviewsandMeta-
Analysis(PRISMA)guidelines(Moheretal.,2009),whichsetstandards
forafour-stepprocesstoconductrigoroussystematicreviews:identifi-
cation,screening,eligibility,andinclusion.
ThestudysearchwasconductedinApril11
–
16,2022,insevenelec-
tronic databases: Web of Science, Medline, ProQuest, EMBASE,
PsycINFO,CINAHL,andEnvironmentComplete.Thesedatabaseswere
selectedsoastobestcovertheliteraturefromabroadarrayoffields
suchasclimatescienceandmeteorology,environmentalscience,psy-
chologyandpsychiatry,public health, and medicine.Wedid not re-
strictthesearchperiod,sothesearchidentifiedarticlesfromtheincep-
tionofeachdatabaseuntilthesearchdate.Weadoptedthepopulation,
exposure,comparator, and outcome (PECO) framework to guidethe
searchandeligibilitycriteria.
2.1.1. Population
Weincluded studies reporting the general human population ex-
posedtovaryinglevelsofclimate/meteorologicalconditions.Animal
studieswereexcluded.Inordertoidentifyacompletesetofstudies,we
didnot restrict the geographicalareasordemographicandsocioeco-
nomiccharacteristics.
2.1.2. Exposure
Climaticandmeteorologicalindicatorsthatarerelevanttohuman
healthandwell-beingwereconsideredasexposure.Toensurecompa-
rability,studiesthatreportedobjectivelymeasuredclimate/meteorol-
ogy/weatherconditions were included,while those onlyconsidering
perceivedambientconditionswereexcluded.Toensurethereviewen-
compassedrelevantexposuretypes,weusedexistingbiometeorological
modelsofhuman-environmentenergyexchangesuchasthePETmodel
(Höppe, 1999b) and the COMFA outdoor thermal comfort model
(BrownandGillespie,1986).Theyestimatehumanheatfluxinanenvi-
ronmentbasedonambientfactorssuchas airtemperature,humidity,
windspeed,andsolarradiation.Inaddition,otherclimateconditions
suchasbarometricpressure, precipitation (e.g.,rain,snow), and fog
werealsoconsidered.Adetailedtablelistingtheexposuredomainsis
presentedinSupplementarymaterialS1.
2.1.3. Comparators
Acomparablepopulationexposedtodifferentlevelsofclimatecon-
ditionsneedstobepresented inordertoestimate therisksofmental
healthdisordersorthevaluesofreportedpsychologicalconditions.As
such,onlystudiesinvolvingacontrolpopulationorreportingvarying
exposureofthesamepopulationwereincluded.
2.1.4. Outcome
Mental health-related outcomes in this study include mortality,
morbidity,andself-reportedmentalhealthandemotionalstates.Mor-
talityandmorbidityrelatedtermswereselectedbasedontheInterna-
tionalClassificationofDiseasesTenthRevision(ICD-10)mental,behav-
ioral, and neurodevelopmental disorders (F00
–
99) classification, in-
cludingsubcategoryterms.ThesetermsencompasstheentireMBDdo-
main,includingdementiaand organic disorders(F00
–
09),psychoac-
tivesubstanceuse(F10
–
19),schizophrenia(F20
–
29),mooddisorders
(F30
–
39), neurotic disorders (F40
–
49), behavioral syndromes
(F50
–
59), personality disorder (F60
–
69), intellectual disabilities
(F70
–
79),developmentaldisorders(F80
–
89),childhoodbehavioraldis-
orders(F90
–
98),andothermentaldisorders(F99).Equivalentdefini-
tionsusingICD-9or the Diagnostic andStatisticalManualof Mental
Disorders, Fifth Edition (DSM-5) were included and cross-walked to
ICD-10classes.Inaddition,weincludedgeneraltermssuchas
“
mental
health
”
andtermsthatdescribesubjectivepsychologicalstate,suchas
“
happiness.
”
and
“
sadness
”
.A detailed table listing the outcome do-
mainsispresentedinSupplementarymaterialS1.
Thesearchsyntaxwasdeveloped based on the PECOframework.
Weusedwildcardstoaccountforvaryingformsofthekeywords.Inad-
dition,forwardandbackwardsearchesusingeligiblearticlesandprevi-
ously conducted systematic reviews were also performed. Example
searchsyntaxusedfortheWebofScienceisprovidedinSupplementary
materialS2.
2.2. Studyselection
AfterimportingtheretrievedarticlesintoEndnote20andremoving
duplicates,weperformedthescreeningandeligibilitystepsbyexamin-
ingthetitles,abstracts,andfulltextsforconcurrencewiththeeligibil-
itycriteriadefinedbasedonthePECOframework.Theselectiondeci-
sionforeachrecordwasmadeindependentlybytworesearchers,with
anydisagreementresolvedthroughdiscussionandthenbyconsultinga
thirdresearcherwhennecessary.Studieswereincludedinthereviewif
theymetthefollowingcriteria.
1) Studyreportedoriginalempiricalresearchandwaspublishedina
peer-reviewedjournal
2) StudywaswritteninEnglish
3) Study reported an observational study on a human population
(seeSection2.1.1)
4) Study included objective climate measurements as the exposure
(seeSection2.1.2)
5) Study included mental health or behavior as the outcome (see
Section2.1.3)
6) Studiesinvolvedacomparisonpopulationorvaryingexposureof
thesamepopulation(seeSection2.1.4)
7) Studywasquantitativeandreportedatleastoneeffectestimateof
aclimate-MBDpair
Studieswereexcludedifthey:1)werenotpeer-reviewedjournalar-
ticlesorwerenon-empiricalsuchasarevieworcommentarypiece;2)
were not written in English; 3) reported an experimental or quasi-
experimentalstudy,4)targetednon-humanbeingsasresearchsubjects
(e.g.,animalstudies);5)comprisedacasestudy/casereportofasingle
subject;6)didnothavearelevantexposureoroutcomevariable;7)ex-
aminedclimateimpactsunderaspecificprogram,training,oroccupa-
tionalenvironment(e.g.,militarytraining);7)investigatedartificially
designed/controlledambientconditions(e.g.,hospitalsorclassrooms
withdifferent levelsoflighting);8)took theseasonorself-reportsof
perceivedclimateasexposure without objective measuresofclimate
factors;9)reportedanon-time-seriesstudywithmeteorologicalfactors
compared at a spatial scale that was too coarse (cross-continent or
cross-country);10)reportedall-causemortalityormorbidityduetocli-
mateexposurebutnotmentalhealth-relatedoutcomes;11)werequali-
tativeor did notconductanyestimateoftheexposure-outcomerela-
tionship; and 12) examined reproductive health or linked-life pairs
(e.g.,amother-childdyad).
The search strategy and selection procedure guided by PRISMA
(Moheretal.,2009)ispresentedinFig.1.
2.3. Datacollectionandnarrativesynthesis
Informationextractionwasconductedbythreeresearchersindepen-
dentlyandthencross-checked.Adescriptive informationspreadsheet
andameta-analysissheetweredevelopedinMicrosoftExceltoextract
and tabulate information from the included studies. The following
3
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Fig.1. PRIMAflowchartofthestudyselectionprocess.
study characteristics were extracted in the descriptive information
sheet:authors, citationdetails, country,population, KöppenClimate
Zoneofstudyarea,studytype,samplesize,spatialresolution,temporal
resolution,climate/meteorologicalmeasures,mentalhealthmeasures,
statisticalanalysis,effectsize,lagperiod,andmoderator/mediator.The
climatic/meteorological measures category included check boxes for
thermalcomfortindex,airtemperature,winddirection/speed,solarra-
diation,relativehumidity, barometric pressure,precipitation, among
others.Wealsotooknotesofwhetheraworkingdefinitionofaclimate/
weatherdisasterwasused,detaileddescriptionsoftheexposuremea-
sures,anddatasources.TheMBDmeasuresincludedtypeofdata(e.g.,
mortality/morbidityrecord,medicationdispensation,self/caregiverre-
port),detaileddescriptionsofoutcome measures/scales, check boxes
foreachMBDtype,timepointsofmeasurement,anddatasources.The
check boxes included categories corresponding to ICD-10 F01
–
F09,
F10
–
F19,F20
–
F29,F30
–
F39,F40
–
F48,F50
–
F59,F60
–
F69,F70
–
F79,
F80
–
F89,F90
–
F98,andF99andequivalentICD-9andDSMcategories.
WerecodedandtabulatedalldatacollectedusingMicrosoftExcel
PivotTableandRpackages.Descriptiveplotswereproducedfordata
synthesis.Amapshowinggeolocationandclimatezonewasproduced
inArcGISPro.Studiesweregroupedbasedonthespatiotemporalreso-
lutions,population characteristics, and exposure andoutcomes mea-
suresandchartssuchascircularbarchartwereproducedusingRpack-
ages.(SeeFig.2.)
4
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
2.4. Riskofbiasassessment
Astheincludedstudiesrangedintypefromecologicaltime-seriesto
individual-level cross-sectional, case-control, and cohort studies, we
developedariskofbiasassessment(RoB)rubricbyintegratingrelevant
itemsfromassessmenttoolscommonlyusedforassessingstudiesonen-
vironmentalexposureandhealthoutcomeatindividualoraggregated
populationlevels.ThesetoolsincludedtheNationalHeart,Lung,and
BloodInstitutesStudyQualityAssessmentTools(NationalHeart,Lung,
andBloodInstitute,2014a;NationalHeart,Lung,andBloodInstitute,
2014b),theJBI'sCriticalAppraisalTools(Moolaetal.,2017),andthe
Riskofabiasassessmentinstrumentforsystematicreviewsinforming
WorldHealthOrganizationglobalairqualityguidelines(WorldHealth
Organization,2020).Specifically,sixdomainswithtwelveitemswere
considered: selection bias, exposure assessment, outcome measure-
ment, confounding, missing data/attrition bias, and reporting bias.
Eachdomainincludedonetothreesub-items,whichwereratedona
four-pointscale:lowrisk,probablylowrisk,probablyhigh risk, and
highrisk. For example, for confounding, we assessed whether time-
invariantfactors(e.g.,age,sex, socioeconomic conditions) andtime-
variant factors (e.g., air pollution, noise, seasonality) potentially re-
lated to the outcomes were considered, measured using valid ap-
proaches,andstatisticallyadjustedforinthemodels.Studiesthatdid
notconsiderrelevantcontrolvariables,didnot adjust for these vari-
ablesinthe statistical modelsestimating the exposure-outcomerela-
tionshipswereratedasprobablyhigh/highinriskofbiasfortherespec-
tiverubricitemsunderconfounding.Forselectionbias,weconsidered
whetherstudy population wasclearlydefinedandrecruitment/inclu-
sioncriteriaconsistentandspecified.Individual-basedstudiesthatused
convenientsamplesoronlinepanels,andecologicalstudiesthatused
datafromasinglehospitaloragencywithoutdescribingthepopulation
servedandcoveragewereratedprobablyhigh/highinoutcomemea-
sure.Fortheoutcomemeasurement,weassessedwhetherthehealth
outcomeassessorswereblindedtotheexposurestatusofparticipants
andwhetherthemeasureswereclearlydefined,valid,reliable,andim-
plementedconsistentlyacrossstudypopulations.Studiesthatusedun-
validatedscales ofoutcomevariableswereconsideredhigh inriskof
bias.Studiesthat only reportedsignificantresults and omittedother
findingsfromplannedanalysiswereevaluatedasprobablyhighinre-
portingbias.Thedetailsoftheassessmentinstrumentarepresentedin
SupplementarymaterialS3.Thegraphicsweregeneratedusingaweb-
basedvisualizationtoolnamedRoB2(Sterneetal.,2019).
Eacharticlewas independently assessedbytwo researchers; con-
flicts were resolved by discussion and then consulting a third re-
searcher.Theitemscoreswerethenaveragedandroundedtogenerate
domainscores.Wedidnotassignoverallriskofbiasscoresduetothe
arbitrarinessofassigningweightstodifferentdomainsofrisk.Instead,
wepresentthe by-item and-domainRoB scores, whichretaintrans-
parency and better reflect the methodological strengths and weak-
nessesofeacharticle.Accordingly,wealsodidnotexcludeanystudies
basedonRoBresultsbutinsteaddiscussedbiasesandprovidedrecom-
mendationsforfuturestudies.
2.5. Meta-analysis
Meta-analysiswasperformedtoestimatethepooledeffectsizesof
theassociationsbetweenclimateexposuresandmentalhealth.Forthe
meta-analysisdataextractionsheet,wecollectedinformationoneach
5
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
exposure-outcomepairincludingexposurevariable,outcomevariable,
statisticalmodel,typeofeffectsizestatistic,effectsizeestimate,stan-
darderror,95%confidence interval lower bound andupperbound,
anddetailedstepsofdataprocessing.Thenaturallogarithmsoftherel-
ativerisk (RR) estimatesandthecorrespondingstandarderrorswere
calculatedandusedinthemeta-analysis.
Whenmultipleestimatesonthesameexposure-outcomepairwere
reportedinonearticle,weextractedonlyoneeffectsizefollowingthe
approaches adopted in previous environmental epidemiology-related
meta-analyses(ChenandHoek,2020;Khreisetal.,2017).Studiesre-
portingoddsratio(OR)wereconvertedintorelativerisk(RR)usingEq.
1whendatawereavailable,andbyassumingequalitywhenrisksinthe
controlgroupwereextremelylow(Harreretal.,2021).Whenanarticle
reportedtwoormoreestimatesforsubgroups(e.g.,byregionorpopu-
lationgroup),wecalculatedthepooledeffectsizeusingafixed-effects
meta-analysis model. When multiple outcomes that belonged to the
sameICD-10subclassweretestedseparately,weusedEq.2tocalculate
thecombinedeffectsize(Harreretal.,2021).Whenmultipleestimates
existedduetomodelspecification,function,estimator,orlagduration,
wefollowedthedecisionruleinsequentialordertoselectthemodelre-
sults:1)selecttheauthors'favoredmodel(mainmodelormodelfollow-
ing main conceptual framework/highlighted in the Abstract, High-
lights,Results,orMainfindingssectionoftheDiscussion),2)selectthe
fullyadjustedmodelratherthanthecrudemodel,andfinally3)select
themodelwith the smallestlagvalue. Pooled effectswerenot com-
putedforself-reportedhealth or psychologicalconditionsbecause of
thesmallnumberofstudiesandthehighheterogeneityofpsychometric
scalesused.
(1)
wherecrepresentstheeventsinthecontrolgroup, isthetotal
numberof participantsinthecontrolgroup, andORistheestimated
oddsratio,RRistherelativeriskcalculated.
(2)
where and arethetwooutcomesand and arethevari-
ance, isthemergedeffectsize.
Heterogeneityofeffectestimatesacrossstudieswasassessedusing
,Cochran'sQ,and statistics. isthevarianceofthedistributionof
thetrueeffectsizeacrossstudiesandiseasytointerpretbasedonthe
originalmeasurementmetric,but andCochran'sQbothheavilyre-
flectstatisticalpowerwhile isnotassensitivetothenumberofstud-
ies;assuch,presentingbothQand helpsindepictingacompletepic-
tureofbetween-studyheterogeneity.Theruleofthumbemployedfor
interpreting wasasfollows:upto25%aslowheterogeneity,50%
for moderate heterogeneity, and 75 % for substantial heterogeneity
(HigginsandThompson,2002).Duetothesmallnumberofstudieson
eachexposure-outcomepairandhighoverallheterogeneity,wedidnot
removeoutliersinthemeta-analysisorperformsensitivityanalysesbut
flaggedontheforestplotthosestudiesreportingeitherextremelysmall
effects(whentheupperboundofthe95%CIislowerthanthelower
boundofthepooledeffect)orextremelylargeeffects(whenthelower
boundoftheCIishigherthantheupperboundofthepooledeffect).
Subgroupanalysisbasedongeographyorpopulationwasnotpossible
duetosmallnumberofexposure-outcomepairs.TheRpackagemetafor
(Viechtbauer, 2010) was used to perform the random-effects meta-
analysismodeling,identifyoutliers,andproduceforestplots.
2.6. Evaluationofcertaintyofevidence
WeutilizedtheGradingofRecommendationsAssessment,Develop-
mentandEvaluation(GRADE)frameworktoassesstheoverallquality
ofevidence (Guyattetal.,2008;Balshemet al., 2011).Abaselineof
moderatecertaintywasinitiallyapplied,andsubsequentlydowngraded
orupgraded based oneach of theGRADE domains. Factorsthat de-
creasedthecertaintyofevidenceincludedD1)Riskofbiasacrossstud-
ies,D2)Indirectness,D3)Inconsistency,D4)Imprecision,andD5)Pub-
licationbias.FactorsthatincreasedcertaintyincludedU1)Largemag-
nitudeofeffects,U2)Consistentdose-responsegradient,andU3)Con-
foundingminimizeseffect.Thedetailedcriteria fordowngradingand
upgradingareasfollows.
Downgrading.D1)The certainty of evidencewas downgraded by
onelevelifatleastonestudythathadanon-negligibleweightinthe
pooledeffectsizeestimateshowedatleastthreehighorprobablyhigh
biasratingsintheRoBevaluations.Inourcase,becausethenumberof
studies in each exposure-outcome pair was small, downgrading was
performedwhenany study exhibitedatleast three highor probably
highbias ratings. D2) The certainty of evidence was downgraded if
studiesdidnot adhere tothe population, exposure,comparator, and
outcomespecifiedfortheresearchquestion,forexamplenotmeasuring
outcomesdirectlybutusingsurrogateoutcomemeasures.Here,thisap-
pliedwhenastudydefinedMBDcasesbasednotonclinicaldiagnosis,
butonoutcomesfromapreliminaryscreeningtoolorevaluationsbyre-
searcher(s)whosecredentialswerenotreported.D3)Thecertaintyof
evidencewasdowngradedwhenverylargeheterogeneityorvariability
in results was detected. Specifically, as observational studies are
demonstratedtohavemoderatetolargeheterogeneityacrossgeogra-
phies and populations (Chen and Hoek, 2020; Schwingshackl et al.,
2021),wedowngradedbyoneortwolevelsifthepredictioninterval
wasmorethantwicetheconfidenceintervalandstudiespresentedthe
associationasbidirectional(bothpositiveandnegative).D4)Thecer-
taintyofevidencewasdowngradedifthenumberofparticipantsand
person/population-time were small (n < 500, determined based on
previousstudiesdiscussingsamplesizeinepidemiology(Vergouweet
al.,2005; Rigby and Vail,1998)).D5)Thecertaintyofevidencewas
downgradedifpublicationbiaswasidentified basedonvisualassess-
mentoffunnelplots.However,funnelplotsandEgger'stesthadlimited
valueduetothefactthatthenumberofstudieswassmallerthanthe
empiricalthresholdof10.
Upgrading. U1) The certainty of evidence was upgraded by one
level if the pooled effect size was large. Defining the threshold for
“
large
”
was challenging, especially given that RRs were mostly low
comparedtothethresholdvaluesused in other medical studies.U2)
Thecertainty of evidence was upgraded by one level if studies pre-
sentedabiologicaldose-responsegradient(e.g.,higherRRassociated
withhigherexposure). U3). The certaintyofevidence was upgraded
whenpossibleresidualconfounderswouldreducethedemonstratedef-
fect.
3. Results
3.1. Literaturesearchandselectionresults
Theinitial search yielded7696 articles, to which59 were added
fromforward/backwardsearchesandotherliterature.Afterduplicate
removal,abstract/titlescreening,andfull-textreview,thefinalreview
setcomprised88articlespublished1972
–
2022(Table1).Apost-2016
surgeinpublicationswasevident.Thevastmajorityofincludedstudies
(n=84,95.5%)examinedallagesoradults(e.g.,18+);onlythree
(3.4%)focusedonolderadults(e.g.,50+),andonlyone(1.1%)on
children.
6
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1
Studycharacteristics(N=88).
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Goldstein,
1972
(Goldstein,
1972)
Collegestudents:
Acohortof
studentsin
anintroductory
psychologycourse
atacommunity
collegeinU.S.
22
persons
11days Cohort City/town
level
Daily Air
temperature,
Humidity,
Barometric
pressure
Self-reported
affect/
mental
health
F30
–
F39 Correlation No
Persinger,
1975
(Persinger,
1975)
Collegestudents:
Acohortof
university
studentsfroma
classatLaurentian
University,
Canada
10
persons
90days(9
January
–
8
April1974)
Cohort City/town
level
Daily Air
temperature,
Windspeed,
Sunshine,
Relative
humidity,
Barometric
pressure,
Self-reported
affect/
mental
health
F30
–
F39 Correlation,
(Generalized)
linearmodel
No
Howarth&
Hoffman,
1984
(Howarth
and
Hoffman,
1984)
Collegestudents:
Acohortof
university
students
participatingin
thestudyfor
course
requirementin
Canada
24
persons
11days Cohort City/town
level
Daily Wind
direction/
speed,
Sunshine,
Humidity,
Barometric
pressure,
Precipitation
Self-reported
affect/
mental
health
F30
–
F39 Correlation,
(Generalized)
linearmodel
No
Mawsonand
Smith,1981
(Mawson
andSmith,
1981)
General
population:
Mentaldiseases-
relatedhospital
admissionswithin
theareaofthe
GreaterLondon
Councilrecorded
bytheStatistics
andResearch
Departmentofthe
DHSS
2126
records
1975 Timeseries City/town
level
Daily Barometric
pressure,
Relative
humidity
Medical
records/
clinical
diagnosis
F30
–
F39 Correlation No
Barnston,
1988
(Barnston,
1988)
Collegestudents:
Undergraduate
psychologyclasses
attheUniversity
ofIllinoisat
Urbana,U.S.
62
persons
September
16
–
October
27,1974
Cohort City/town
level
Daily Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation,
Relative
humidity,
Wind
direction/
speed
Self-reported
affect/
mental
health
General
psychological
health
Correlation No
Carneyetal.,
1988
(Carneyet
al.,1988)
General
population:
Mentaldisorders-
relatedhospital
admissions
University
Departmentof
Psychiatryatthe
RegionalHospital,
Galway.
104
records
1980
–
1984 Timeseries Neighborhood/
Hospital
catchment
level
Monthly Air
temperature,
Solar
radiation/
Sunshine
Medical
records/
clinical
diagnosis
F30
–
F39 Correlation No
Molinetal.,
1996(Molin
etal.,1996)
Acohortof
patientsdiagnosed
withdepression
126
persons
1991
–
1994 Cohort City/town
level
Weekly Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation
Symptom
worsening/
Newepisode
(Generalized)
linearmodel
No
(continuedonnextpage)
7
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Salibetal.,
1999(Salib
andSharp,
1999)
General
population:
Admissiontoa
largepsychiatric
hospital(Winwick
Hospital)inNorth
Cheshire
2070
records
1993 Timeseries Countylevel Daily Air
temperature,
Solar
radiation/
Sunshine,
Relative
humidity,
Precipitation
Medical
records/
clinical
diagnosis
F00
–
F09 Correlation No
Leeetal.,
2002(Leeet
al.,2002)
General
population:
Admissiontothe
psychiatricunitof
thetwohospitals
affiliatedwiththe
KoreaUniversity
MedicalCenter
152
persons
1996
–
1999 Cohort Neighborhood/
Hospital
catchment
level
Monthly Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation,
Relative
humidity
Medical
records/
clinical
diagnosis
F30
–
F39 Correlation No
Saliband
Sharp,2002
(Saliband
Sharp,2002)
General
population:
Admissiontoa
largepsychiatric
hospital(Winwick
Hospital)inNorth
Cheshire
1084
records
1993 Timeseries City/town
level
Daily Air
temperature,
Solar
radiation/
Sunshine,
Relative
humidity,
Precipitation
Medical
records/
clinical
diagnosis
F10
–
19,F20
–
F29,F30
–
F39,F40
–
F49
Correlation No
Cornalietal.,
2004
(Cornaliet
al.,2004)
Dementia
patients:
Patientsdiagnosed
withdementia
admittedtothe
Alzheimer
Rehabilitation
Unit,Richiedei
MedicalCenter,
Palazzolos/O
–
Brescia,Italy.
25
persons
June14
–
21,
2002
Case-control Neighborhood/
Hospital
catchment
level
Daily Air
temperature
Symptom
worsening/
Newepisode
&Drug
dispensation
Dementia (Generalized)
linearmodel
No
Bulbenaetal.,
2005
(Bulbenaet
al.,2005)
General
population:
Psychiatric
emergenciesatdel
MarHospital,
Barcelona
368
records
2002 Other
ecological
City/town
level
Daily Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation,
Relative
humidity,
Wind
direction/
speed
Medical
records/
clinical
diagnosis
F40
–
F49 (Generalized)
linearmodel
No
Kelleretal.,
2005(Keller
etal.,2005)
General
population:
Acohortof
populationliving
inAnnArbor,
Michigan,U.S.
605
persons
April5
–
June
15,2001,
April16
–
July27,
2003,
January
–
December
2002
Cross-
sectional
City/town
level
Daily Air
temperature,
Barometric
pressure
Self-reported
affect/
mental
health
F30
–
F39 (Generalized)
linearmodel
No
(continuedonnextpage)
8
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Shilohetal.,
2005(Shiloh
etal.,2005)
General
populationaged
18+:
Hospital
admissionsand
psychiatric
diagnoses
recordedbythe
IsraeliNational
Psychiatric
Registry(INPR),
Departmentof
Informationand
Evaluation,
MentalHealth
Services(Ministry
ofHealth,
Jerusalem,Israel)
33,614
records
1981
–
1991 Timeseries City/town
level
Daily Thermal
comfort
index,Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Relative
humidity
Medical
records/
clinical
diagnosis
F20
–
F29 Correlation No
Hartigetal.,
2007(Hartig
etal.,2007)
General
population:
Dispensationof
selectiveserotonin
reuptake
inhibitors(SSRIs)
recordedbythe
ApoteketABin
Sweden
NR 1991
–
1998 Timeseries Nationallevel Monthly Air
temperature
Drug
dispensation
Depression Auto
regressive
integrated
moving
average
No
Leeetal.,
2007(Leeet
al.,2007)
General
population:
Hospitalization
recordinthe
TaiwanNational
HealthInsurance
ResearchDatabase
15,060
records
1999
–
2003 Timeseries Nationallevel Monthly Air
temperature,
Solar
radiation/
Sunshine,
Precipitation
Medical
records/
clinical
diagnosis
F30
–
F39,F99 Auto
regressive
integrated
moving
average
No
Christensenet
al.,2008
(Christensen
etal.,2008)
Bipolarpatients:
Bipolarpatients
admittedtothe
departmentsof
psychiatryin
threeCopenhagen
University
Hospitals,aged18
and75yearsold
56
persons
1990
–
1993 Cohort Neighborhood/
Hospital
catchment
level
Multi-
month
Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation,
Relative
humidity,
Wind
direction/
speed
Medical
records/
clinical
diagnosis
F30
–
F39 Correlation No
Denissenetal.,
2008
(Denissenet
al.,2008)
General
population:
Acohortof
populationsigning
uptoanonline
diarystudyin
Germany
1233
persons
July2005
–
February
2007
Cohort Nationallevel Daily Air
temperature,
Windspeed,
Sunlight,
Barometric
pressure,
Precipitation
Self-reported
affect/
mental
health
F30
–
F39 (Generalized)
linearmodel
No
Hansenetal.,
2008
(Hansenet
al.,2008)
General
population:
Hospitalizations
formentaland
behavioral
disordersordeath
recordsassociated
withmentaland
behavioral
disordersin
Adelaide,South
Australiacollected
bytheSouth
Australian
171,
614
records
1993
–
2006 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 (Generalized)
linearmodel
Yes
(continuedonnextpage)
9
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Bulbenaetal.,
2009
(Bulbenaet
al.,2009)
General
population:
Psychiatric
emergenciesat
HospitaldelMar
hospitaland
InstitutMunicipal
Psiquiatria
hospital,
Barcelona
872
records
2003
summer
days
Timeseries Neighborhood/
Hospital
catchment
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F10
–
F19,
F20
–
F29,
F30
–
F39,
F40
–
F49,
F60
–
F69
(Generalized)
linearmodel
Yes
Huibersetal.,
2010
(Huiberset
al.,2010)
General
population(aged
18
–
65):
Reportedpresence
ofmental
disordersin
southern
Netherlands
14,478
persons
2005
–
2007 Timeseries Nationallevel Daily Air
temperature,
Solar
radiation/
Sunshine,
Precipitation
Medical
records/
clinical
diagnosis
F30
–
F39 (Generalized)
linearmodel
Yes
Khalajetal.,
2010(Khalaj
etal.,2010)
General
population:
Hospital
admissionsinfive
regions(Sydney
East;Sydney
West;Gosford
Yong;Newcastle
andIllawarra)of
NewSouthWales
1,497,
655
records
springand
summer
days
(September
–
February)of
1998
–
2006
Timeseries Regionallevel Daily Thermal
comfort
index,Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 (Generalized)
linearmodel
Yes
Raduaetal.,
2010(Radua
etal.,2010)
General
population:
Admissionstothe
AcuteUnitatthe
Departmentof
Psychiatryin
Bellvitge
University
Hospital,
Barcelona,Spain
421
persons
1997
–
2004 Timeseries City/town
level
Daily Thermal
comfortindex
Medical
records/
clinical
diagnosis
F30
–
F39 Autoregressive
Integrated
Moving
Average
No
Klimstraetal.,
2011
(Klimstraet
al.,2011)
Adolescentsand
theirmothers:
Acohortof
adolescentsand
theirmothers
enrolledinan
ongoing
longitudinal
projectinThe
Netherlands,
entitledResearch
onAdolescent
Developmentand
Relationships
(RADAR).
823
persons
6weeks Cohort Nationallevel Daily Air
temperature,
Sunlight,
Precipitation
Self-reported
affect/
mental
health
F30
–
F39 Correlation No
Sungetal.,
2011(Sung
etal.,2011)
General
population:
Medicalrecordsof
psychiatric
hospital
admissionsin
Psychiatric
InpatientMedical
Claim(PIMC)
datasetofthe
NationalHealth
Insurance
ResearchDatabase
inTaiwan
41,023
records
1996
–
2007 Timeseries Nationallevel Daily Air
temperature
Medical
records/
clinical
diagnosis
F20
–
F29 (Generalized)
linearmodel
Yes
(continuedonnextpage)
10
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Yackersonet
al.,2011
(Yackerson
etal.,2011)
General
population:
Mentaldiseases-
relatedemergency
visitsand
hospitalizationin
theregional
MentalHealth
Center(MHC)of
Ben-Gurion
University,Israel
4325
persons
2001
–
2003 Timeseries Neighborhood/
Hospital
catchment
level
Weekly Air
temperature,
Windspeed/
direction,
Relative
humidity
Medical
records/
clinical
diagnosis
F20
–
F29 Correlation No
Gasparriniet
al.,2012
(Gasparrini
etal.,2012)
General
population:
Deathrecords
associatedwith
heatrecordedby
theOfficefor
NationalStatistics
inEnglandand
Wales
92,439
records
summers
(June
–
September)
of1993
–
2006
Timeseries Regionallevel Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F10
–
F19,
F20
–
F29
(Generalized)
linearmodel
Yes
Vidaetal.,
2012(Vida
etal.,2012)
General
population:
Emergency
departmentvisits
formental
disordersinthree
geographicareas
ofQuébec,aged
15yearsandover
347,
552
records
1995
–
2007 Other
ecological
Climatezone Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F10
–
F19,
F20
–
F29,
F30
–
F39
(Generalized)
linearmodel
Yes
Williamsetal.,
2012
(Williamset
al.,2012)
General
population:
Mortality,hospital
admissions,and
emergency
departmentvisits
collectedbythe
SouthAustralian
Departmentof
Health
NR 1993
–
2009 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 GEE Yes
Alexander,
2013
(Alexander,
2013)
General
population:
Psychiatric
disease-related
callstothepublic
emergencyservice
ofthecityof
BuenosAires,
Argentina
80,724
records
1999
–
2004 Timeseries City/town
level
Monthly Thermal
comfort
index,Air
temperature,
Precipitation,
Relative
humidity
Medical
records/
clinical
diagnosis
General
mentalhealth
Correlation No
McWillimaset
al.,2013
(McWilliams
etal.,2013)
General
population:
Admissionsto
psychiatric
hospitalsinthe
Republicof
Ireland
48,347
records
1971
–
2002 Timeseries Regionallevel Daily Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation,
Wind
direction/
speed
Medical
records/
clinical
diagnosis
F20
–
F29 Autoregressive
Integrated
Moving
Average;
(Generalized)
linearmodel
No
(continuedonnextpage)
11
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Sungetal.,
2013(Sung
etal.,2013)
General
population:
Medicalrecordsof
psychiatric
hospital
admissionsin
Psychiatric
InpatientMedical
Claim(PIMC)
datasetofthe
NationalHealth
Insurance
ResearchDatabase
inTaiwan
9071
records
1996
–
2007 Timeseries Nationallevel Daily Air
temperature,
Precipitation
Medical
records/
clinical
diagnosis
F30
–
F39 (Generalized)
linearmodel
Yes
Tsutsui,2013
(Tsutsui,
2013)
Collegestudents:
Acohortof
university
studentsrecruited
oncampus
groundsaswellas
throughawebsite
atOsaka
University,Japan
75
persons
516days(1
November
2006
–
31
March2008)
Cohort City/town
level
Daily Air
temperature,
Windspeed,
Sunshine,
Humidity,
Precipitation
Self-reported
affect/
mental
health
General
psychological
health,F30
–
F39,F40
–
F49,F50
–
F59
Correlation,
(Generalized)
linearmodel
No
Vaneckovaet
al.,2013
(Vaneckova
and
Bambrick,
2013)
General
population:
Admissionstoall
privateandpublic
hospitalslocated
intheSydney
930,
322
records
1991
–
2009 Case-
crossover
City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F20
–
F29,
F40
–
F49,
F70
–
F79
(Generalized)
linearmodel
Yes
Wilsonetal.,
2013
(Wilsonet
al.,2013)
General
population:
Medicalrecordsof
hospital
admissionsor
deathrecordsin
theNewSouth
WalesDepartment
ofHealth
NR 1997
–
2007
and1997
–
2010
Case-
crossover
City/town
level
Daily Thermal
comfort
index,Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Henriquez-
Sanchezet
al.,2014
(Henriquez-
Sanchezet
al.,2014)
CollegeStudents:
Acohortof
Spanishuniversity
graduates
participatingin
theSUNproject
fromSeguimiento
Universityof
Navarra
13,938
persons
1999
–
2009 Cohort Regionallevel Multi-
year
Air
temperature,
Solar
radiation/
Sunshine,
Precipitation
Medical
records/
clinical
diagnosis
F30
–
F39 Cox
(Generalized)
linearmodel
Yes
McWillimaset
al.,2014
(McWilliams
etal.,2014)
General
population:
Admissionsto
psychiatric
hospitalsinthe
Republicof
Ireland
34,465
records
1971
–
2002 Timeseries Regionallevel Daily Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation,
Wind
direction/
speed
Medical
records/
clinical
diagnosis
F30
–
F39 Autoregressive
Integrated
Moving
Average;
(Generalized)
linearmodel
No
Obrienetal.,
2014
(Obrienet
al.,2014)
Populationaged
15+:
Acohortof
populationfrom
theHousehold,
Incomeand
LabourDynamics
inAustralia
(HILDA)Survey
5012
persons
2007
–
2008 Cross-
sectional
(onewave
ofa
longitudinal
dataset)
Nationallevel Monthly Precipitation Self-reported
affect/
mental
health
General
psychological
health
(Generalized)
linearmodel
No
(continuedonnextpage)
12
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Simonsetal.,
2014
(Simonset
al.,2014)
General
population:
Acohortof
population
admittedto3
medicalcenters
(Radboud
University
NijmegenMedical
Centrein
Nijmegen,
University
MedicalCentrein
Utrecht,and
JeroenBosch
Hospital,'s-
Hertogenbosch)in
TheNetherlands
3198
persons
2008
–
2012 Cohort Neighborhood/
Hospital
catchment
level
Monthly Solar
radiation/
Sunshine
Medical
records/
clinical
diagnosis
F00
–
F09 (Generalized)
linearmodel
Yes
Wangetal.,
2014(Wang
etal.,2014)
General
population:
Emergencyroom
visitsformental
illnessinNational
AmbulatoryCare
ReportingSystem
inTorontothat
capturedover
97%oftheER
visitsinthe
provinceof
Ontarioandhad
goodreabstraction
accuracy
271,
746
records
2002
–
2010 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Kimetal.,
2015(Kim
etal.,2015)
General
population:
Deathrecord
providedby
StatisticsKorea
50,055
records
1992
–
2009 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99,
F10
–
F19,
F20
–
F29
(Generalized)
linearmodel
Yes
Beecheretal.,
2016
(Beecheret
al.,2016)
Mentalhealth
distresspatients:
Acohortof
university
students
participatingin
mentalhealth
treatmentat
BrighamYoung
University,U.S.
16,452
persons
2008
–
2014 Other City/town
level
Daily Sunshine Self-reported
affect/
mental
health
General
mentalhealth
(Generalized)
linearmodel
No
Dingetal.,
2016(Ding
etal.,2016)
Populationaged
45+
Acohortof
residentsaged45
andoverof
thesoutheast
Australianstateof
NewSouthWales,
Australia.
53,144
persons
2006
–
2008 Cohort State/Province
level
Daily Air
temperature,
Relative
humidity
Self-reported
affect/
mental
health
General
psychological
health
(Generalized)
linearmodel
No
Noelkeetal.,
2016
(Noelkeet
al.,2016)
General
populationaged
18+:participants
reportingpresence
ofmental
disordersinthe
GallupG1K
dataset
1,854,
746
persons
2008
–
2013 Other Nationallevel Daily Air
temperature
Self-reported
affect/
mental
health
General
psychological
health
(Generalized)
linearmodel
No
(continuedonnextpage)
13
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
O'Hareetal.,
2016
(O'Hareet
al.,2016)
Populationaged
50+:
Acohortof
populationfrom
thefirstwaveof
TheIrish
Longitudinal
StudyonAgeing
(TILDA)
8027
persons
late2009to
early2011
Cross-
sectional
(onewave
ofa
longitudinal
dataset)
Nationallevel Monthly Air
temperature,
Solar
radiation/
Sunshine,
Precipitation
Medical
records/
clinical
diagnosis
F30
–
F39 (Generalized)
linearmodel
No
Trang,
Rocklöv,
Giang,
Kullgren,et
al.,2016
(Trangetal.,
2016a)
General
population:
Mentaldisorders-
relatedhospital
admissionsina
databasefrom
HanoiMental
Hospital(oneof
mentalhospitals
inHanoiCity)in
northernVietnam
21,443
records
2008
–
2012 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F70
–
F79
(Generalized)
linearmodel
Yes
Shiue,Perkins,
&Bearman,
2016(Shiue
etal.,2016)
General
population:
Mentalbehavioral
disorders-related
hospital
admissionsin
Germanhospitals
NR 2009
–
2011 Timeseries Regionallevel Daily Thermal
comfortindex
Medical
records/
clinical
diagnosis
F01
–
F09,
D10
–
F19,
F20
–
F29,
F30
–
F39,
F40
–
F49,
F50
–
F51,F98
Polynomial
regression
model
No
Trang,
Rocklöv,
Giang,&
Nilsson,
2016(Trang
etal.,
2016b)
Populationwith
medicalrecordsof
mentaldisorders-
relatedhospital
admissionsina
databasefrom
HanoiMental
Hospital(oneof
themental
hospitalsinHanoi
City)innorthern
Vietnam
23,525
records
2008
–
2012 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F10
–
19,F20
–
F29,F30
–
F39,F40
–
F49,F50
–
F59,F70
–
F79,F99
(Generalized)
linearmodel
Yes
Yi-Fanetal.,
2016(Yi-
Fanetal.,
2016)
General
population:
Participantsofthe
International
SocialSurvey
Program(ISSP)
reporting
psychological
conditionsfrom
29counties
5420
records
NR Cross-
sectional
Nationallevel Daily Air
temperature,
Relative
humidity,
Wind
direction/
speed
Self-reported
affect/
mental
health
General
psychological
health
(Generalized)
linearmodel
No
Bullocketal.,
2017
(Bullocket
al.,2017)
Bipolarpatients:
Patientsdiagnosed
withbipolar
disorderfroma
publichealth
serviceinregional
Victoria,Australia
11
persons
Anaverage
of130
consecutive
days(range:
14
–
231days)
Case-
crossover
Regionallevel Daily Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Relative
humidity
Symptom
worsening/
Newepisode
Mood (Generalized)
linearmodel
No
Linaresetal.,
2017
(Linareset
al.,2017)
General
population:
Mentaldisorders-
relatedemergency
admissionsto
municipal
hospitalsin
Madrid
1175
records
2001
–
2009 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09 (Generalized)
linearmodel
Yes
(continuedonnextpage)
14
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Pengetal.,
2017(Peng
etal.,2017)
General
population:
Participatedin
HealthInsurance
Systemof
Shanghaiwith
medicalrecordsof
hospital
admissionsfor
mentaldisorders
inShanghai,
China
93,971
records
2008
–
2015 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Sarranetal.,
2017
(Sarranet
al.,2017)
Patientswith
seasonalaffective
disorder
Acohortof
patientsdiagnosed
withseasonal
affectivedisorder
attheUniversity
Centerof
Psychiatryatthe
University
MedicalCenter
Groningen,
Netherlands
291
persons
2003
–
2009 Cohort Neighborhood/
Hospital
catchment
level
Weekly Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Relative
humidity
Symptom
worsening/
Newepisode
Depression (Generalized)
linearmodel
No
Basuetal.,
2018(Basu
etal.,2018)
General
population:
Medicalrecordsof
mentaldisorders-
relatedemergency
roomvisitsby
CaliforniaOffice
ofStatewide
HealthPlanning
andDevelopment
219,
942
records
2005
–
2013 Timeseries State/Province
level
Daily Thermal
comfortindex
Medical
records/
clinical
diagnosis
F20
–
F29,
F40
–
F49
(Generalized)
linearmodel
Yes
Chanetal.,
2018(Chan
etal.,2018)
General
population:
Mentaldisorders-
relatedhospital
admissions
collectedbythe
HospitalAuthority
ofHongKong
contained>99%
ofcasesduesto
mentaldisorders
inHongKong
44,600
records
2002
–
2011 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F10
–
F19,
F20
–
F29,
F30
–
F39,
F40
–
F49,F99
DLNM Yes
M. Leeetal.,
2018(Leeet
al.,2018a)
General
population:
Acorhortof
population
participatingthe
JapaneseHealth
DiaryStudyin
Japan
4548
persons
Octoberof
2013
Cohort Nationallevel Daily Air
temperature,
Relative
humidity
Medical
records/
clinical
diagnosis
F40
–
F49 (Generalized)
linearmodel
No
S. Leeetal.,
2018(Leeet
al.,2018b)
General
population:
Emergency
admissions
collectedbythe
KoreanNational
HealthInsurance
Corporation
containing
medical
informationfor
almost100%of
theKorean
population
166,
578
records
2003
–
2013 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F20
–
F29,
F30
–
F39,
F40
–
F49
DLNM Yes
(continuedonnextpage)
15
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Obradovichet
al.,2018
(Obradovich
etal.,
2018b)
General
populationaged
18+:
Participants
completingthe
BehavioralRisk
Factor
Surveillance
System(BRFSS)
undertheCenters
forDisease
Controland
Prevention(CDC)
1,961,
743
persons
2002
–
2012 Other Nationallevel Monthly Air
temperature,
Precipitation
Self-reported
affect/
mental
health
General
mentalhealth
(Generalized)
linearmodel
No
Sherbakovet
al.,2018
(Sherbakov
etal.,2018)
General
population:
Hospitalizationsin
Californiainthe
Officeof
StatewideHealth
Planningand
Development
(OSHPD)Patient
DischargeData
(PDD)
130,
065
records
May
–
Octoberof
1999
–
2009
Timeseries Climatezone Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Tapaketal.,
2018(Tapak
etal.,2018)
General
population:
hospitalizationin
Farshchian
hospital(theonly
psychiatric
hospitalacross
Hamadan
Province)in
westernIran
20,406
persons
2005
–
2017 Timeseries Neighborhood/
Hospital
catchment
level
Daily Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation
Medical
records/
clinical
diagnosis
F20
–
F29,
F30
–
F39
(Generalized)
linearmodel
Yes
Wangetal.,
2018(Wang
etal.,2018)
General
population:
Admissionstothe
AnhuiMental
HealthCenterin
China
17,744
records
warmseason
(May
–
October)in
2005
–
2014
Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F20
–
F29 DLNM Yes
Xuetal.,2019
(Xuetal.,
2019)
General
population:
Emergency
departmentvisits
recordedbythe
Queensland
Health
NR 2013
–
2015 Timeseries Neighborhood/
Hospital
catchment
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Almendraet
al.,2019
(Almendra
etal.,2019)
General
population:
Mentaldisorders-
relatedhospital
admissionsinthe
DiagnosisRelated
Groupsgeneral
databaseprovided
bythePortuguese
HealthSystem
Central
Administration
30,139
records
2008
–
2014 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F10
–
F19,
F20
–
F29,
F30
–
F39,
F40
–
F49,
F50
–
F59,
F60
–
F69
DLNM Yes
Guetal.,2019
(Guetal.,
2019)
General
population:
Mentaldisorders-
relatedhospital
admissionatthe
biggestpsychiatric
hospitalin
Ningbo,China
10,132
records
2012
–
2016 Timeseries City/town
level
Daily Solar
radiation/
Sunshine
Medical
records/
clinical
diagnosis
F20
–
F29 DLNM Yes
(continuedonnextpage)
16
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Hoetal.,2019
(Hoand
Wong,2019)
General
population:
Deathrecordsina
mortalitydataset
providing
informationofall
decedentsinHong
Kong
133,
359
records
2007
–
2014 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 (Generalized)
linearmodel
Yes
Liuetal.,
2019(Liuet
al.,2019)
General
population:
Mentaldiseases-
relatedhospital
admissionsatthe
MentalHealth
Centerof
Shandong
provinceinChina
19,569
persons
14days(14
June14
–
17,
June28
–
30,
July4
–
7,
andJuly29
–
31) in2010
Case-
crossover
City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 (Generalized)
linearmodel
No
Minetal.,
2019(Min
etal.,2019)
General
population:
Mentaldisorders-
relatedhospital
admissionsinthe
hospitalmedical
recordsystemsof
Yanchengcity,
China
8438
records
2014
–
2017 Timeseries City/town
level
Daily Thermal
comfortindex
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Mullins&
White,2019
(Mullinsand
White,
2019)
General
population:
Mentaldisorders-
relatedemergency
departmentvisits
and
hospitalizations
collectedbythe
California'sOffice
ofStatewide
HealthPlanning
andDevelopment
&participants
reportingmental
healthby
interviewsfrom
theBehavioral
RiskFactor
Surveillance
System(BRFSS)
undertheCenters
forDisease
Controland
Prevention(CDC),
aged18yearsand
older
5,996,
037
records
2005
–
2016 Timeseries Countylevel Daily Air
temperature
Medical
records/
clinical
diagnosis&
Self-reported
affect/
mental
health
F00
–
F99 (Generalized)
linearmodel
No
Panetal.,
2019(Panet
al.,2019)
General
population:
Mentaldiseases-
relatedhospital
admissions
collectedbythe
AnhuiMental
HealthCenterin
Hefei,China
30,022
records
2005
–
2014 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F20
–
F29 DLNM No
(continuedonnextpage)
17
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Xuetal.,2019
(Xuetal.,
2018)
Childrenaged6
–
11:
Acohortof
childrenaged6to
11yearsold,
participatingin
theLongitudinal
Studyof
Australian
Children
6875
records
2008
–
2014 Cohort Nationallevel Yearly Air
temperature,
Precipitation
Self-reported
affect/
mental
health
F00
–
F99 (Generalized)
linearmodel
No
Yietal.,2019
(Yietal.,
2019)
General
population:
Emergency
admissionsinthe
AnhuiMental
HealthCenterin
China
36,607
records
2005
–
2014 Timeseries City/town
level
Daily Thermal
comfortindex
Medical
records/
clinical
diagnosis
F20
–
F29 DLNM Yes
daSilvaetal.,
2020(da
Silvaetal.,
2020)
General
population:
Hospitalizations
formentaland
behavioral
disordersinthe
publicSingle
SystemofHealth
(SUS)inCuritiba,
Brazil
5397
records
2010
–
2016 Timeseries City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Lietal.,2020
(Lietal.,
2020)
General
populationaged
18+:
Participantsofthe
BehavioralRisk
Factor
Surveillance
System(BRFSS)
undertheCenters
forDisease
Controland
Prevention(CDC)
3,060,
158
persons
1993
–
2010 Other
ecological
Countylevel Daily Air
temperature
Self-reported
affect/
mental
health
General
mentalhealth
(Generalized)
linearmodel
No
Liuetal.,
2020(Liuet
al.,2020)
General
population:
Deathrecords
collectedbythe
HongKongCensus
andStatistics
Department
19,534
records
2006
–
2016 Timeseries Nationallevel Daily Air
temperature,
Relative
humidity
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Niuetal.,
2020(Niuet
al.,2020)
General
population:
Mentaldisorders-
relatedemergency
admissionsin30
hospitalsin
Beijingrecorded
byBeijing
MunicipalHealth
Commission
Information
Centerthat
coveredall
admissions
16,606
records
2016
–
2018 Timeseries City/town
level
Daily Thermal
comfortindex
Medical
records/
clinical
diagnosis
F10
–
F19,
F20
–
F29,
F30
–
F39,
F40
–
F49
DLNM Yes
(continuedonnextpage)
18
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Ohetal.,2020
(Ohetal.,
2020)
General
population:
Emergency
departmentvisits
forpanicattacks
intheNational
Emergency
Department
Information
System(NEDIS)in
Seoul,South
Korea
1926
persons
2008
–
2014 case-
crossover
City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F40
–
F49 (Generalized)
linearmodel
Yes
Zhangetal.,
2020(Zhang
etal.,2020)
General
population:
Admissionstothe
publiclyfunded
andauthoritative
psychiatric
specialisthospitals
formental
disordersinthree
Chinesecities
(Shenzhen,
Zhaoqing,and
Huizhou)
1,133,
220
records
2013
–
2018 Case-
crossover
City/town
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F20
–
F29,
F30
–
F39,
F40
–
F49
DLNM Yes
Bundoetal.,
2021(Bundo
etal.,2021)
General
population:
Mentaldisorders-
relatedhospital
admissionsinthe
University
Psychiatric
HospitalinBern,
Switzerland
88,996
records
1973
–
2017 Case-
crossover
Neighborhood/
Hospital
catchment
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09,
F10
–
F19,
F20
–
F29,
F30
–
F39,
F40
–
F49,
F60
–
F69,
F70
–
F79,
F80
–
F89
Distributedlag
linearmodel
Yes
Burdett,
Davillas,&
Etheridge,
2021
(Burdettet
al.,2021)
General
population:
Acohortof
populationfrom
theUKHousehold
Longitudinal
Study(UKHLS)on
mentalhealth
NR April
–
Julyin
2020
Cohort Nationallevel Daily Air
temperature,
Sunshine,
Precipitation
Self-reported
affect/
mental
health
General
mentalhealth
(Generalized)
linearmodel
No
Eun-hyeetal.,
2021(Eun-
hyeetal.,
2021)
General
population:
Mentaldisorders-
relatedemergency
roomvisitsinthe
Statewide
Planningand
Research
Cooperative
Systemoperated
bytheNewYork
StateDepartment
ofHealth.
92,627
records
2009
–
2015 Timeseries State/Province
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Jahan&
Wraith,
2021(Jahan
andWraith,
2021)
General
population:
Mentaldisorders-
relatedhospital
admissionsinthe
Queensland
HospitalAdmitted
PatientData
Collectionthat
collects
informationfrom
allpublicand
privatehospitals
inQueensland,
Australia
132,
088
records
1996
–
2015 Timeseries Regionallevel Daily Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation,
Relative
humidity
Medical
records/
clinical
diagnosis
F20
–
F29 DLNM No
(continuedonnextpage)
19
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Kimetal.,
2021(Kim
etal.,2021)
General
population:
Acohortthat
underwenthealth
examinations
fromtheKorean
NationalHealth
InsuranceService
25,589
persons
2002
–
2013 Case-control Nationallevel Multi-
day
Air
temperature,
Solar
radiation/
Sunshine,
Barometric
pressure,
Precipitation,
Relative
humidity
Medical
records/
clinical
diagnosis
F30
–
F39 (Generalized)
linearmodel
Yes
Middletonet
al.,2021
(Middleton
etal.,2021)
General
population:
Mentalhealth-
relatedclinic
visitsatfive
communityclinics
inNunatsiavut
228,
104
records
2012
–
2018 Timeseries Neighborhood/
Hospital
catchment
level
Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 (Generalized)
linearmodel
No
Sonetal.,
2021(Son
andShin,
2021)
General
population:
Mentaldisorders
fromHealth
InsuranceReview
andAssessment
Service-National
PatientSample
providedbythe
NationalHealth
Insuranceof
Korea
531,
342
records
2016 Timeseries Nationallevel Daily Solar
radiation/
Sunshine
Medical
records/
clinical
diagnosis
F30
–
F39 (Generalized)
linearmodel
Yes
Tangetal.,
2021(Tang
etal.,2021)
General
population:
Medicalrecordsof
hospitalizationin
thelargest
psychiatric
hospital(Anhui
MentalHealth
Center)inAnhui
Province,China
53,288
records
2005
–
2019 Timeseries Neighborhood/
Hospital
catchment
level
Daily Thermal
comfortindex
Medical
records/
clinical
diagnosis
F20
–
F29 (Generalized)
linearmodel
No
Yooetal.,
2021(Yooet
al.,2021)
General
population:
International
SocialSurvey
Progental
disorders-related
emergencyroom
visitsinthe
Statewide
Planningand
Research
Cooperative
Systemoperated
byNewYorkState
Departmentof
Health
2,893,
794
records
2009
–
2016 Timeseries Regionallevel Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F99 DLNM Yes
Zapataetal.,
2021
(Zapata,
2021)
General
population:
Participantsofthe
employment
surveyofEcuador
conductedbythe
NationalInstitute
ofStatisticsand
Census(INEC)
54,541
persons
NR Cross-
sectional
Nationallevel Monthly Thermal
comfort
index,Air
temperature,
Precipitation,
Relative
humidity
Self-reported
affect/
mental
health
General
psychological
health
(Generalized)
linearmodel
No
(continuedonnextpage)
20
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table1(continued)
Citation Population Sample
size
Study
periods
Typeof
study
Spatialunitof
analysis
Temporal
unitof
analysis
Microclimate
exposure
Typeof
mental
health
outcome
Mentalhealth
outcome
Statistical
analysis
Included
inmeta-
analysis
Gongetal.,
2022(Gong
etal.,2022)
General
population:
Mentaldisorders-
relatedemergency
roomvisitsinthe
NationalHealth
Service(NHS)
DigitalinEngland
NR 1998
–
2009 Timeseries Regionallevel Daily Air
temperature
Medical
records/
clinical
diagnosis
F00
–
F09 Distributedlag
linearmodel
Yes
Note.NR:notreported;record:thenumberofmedicalrecords(e.g.,mentaldisorder-relatedadmissions,hospitalization,oremergencyvisit);person:thenumberof
individuals.GEE:generalizedestimatingequation;DLNM:distributedlagnon-linearmodel.
3.2. Narrativesynthesisofstudycharacteristics
3.2.1. Geographicalandclimatezonedistribution
The included studies represented 23 countries/regions, predomi-
nantly in Asia (n = 29, 33.0 %) and Europe (n = 27, 30.7 %), followed
by North America (n = 17, 19.3 %). Outside of a few Asian countries
and Australia, the rest of the Global South (e.g., Latin America, Africa)
was underrepresented (n = 3, 3.4 %). In terms of climate zone, most
studies focused on temperate and dry climates (n = 74, 84.1 %); only a
few from Ecuador (Zapata, 2021), Brazil (da Silva et al., 2020), Taiwan
(Sung et al., 2011; Sung et al., 2013), northern Australia (Xu et al.,
2019; Jahan and Wraith, 2021) and Vietnam (Trang et al., 2016a) fea-
tured tropical climates, and none polar climates (Fig. 2).
3.2.2. Typesofclimaticandmeteorologicalexposuresreported
Frequently-investigated climatic and meteorological factors in-
cluded air/ambient temperature, relative humidity, precipitation, sun-
light/solar radiation, and barometric pressure (Fig. 3, left). At least two
factors were evaluated in 74 studies (84.1 %), and just one in the re-
mainder (15.9 %). Most studies (n = 74, 84.1 %) included air tempera-
ture and temperature-based metrics (e.g., heatwave occurrence),
chiefly mean daily temperature, mean of daily maximum or minimum
temperature, and daily temperature range. Heat and cold were also
commonly considered, typically defined with either a) an absolute
threshold based on local climate norms (Hansen et al., 2008; Bulbena et
al.,2009;Trangetal.,2016a;Liuetal.,2019)orb)aparticularrelative
percentile(e.g.,95%,99%)ofthetemperaturedistribution(Liuetal.,
2020;Sherbakovetal.,2018;Khalajetal.,2010).Likewise,heatwaves
orextremeheateventsweredefinedasmaximumtemperaturesexceed-
inganabsoluteorrelativethresholdforconsecutivedays.
Inadditiontotemperature,30studies(34.1%)examinedtheeffect
ofsolarradiationintermsofsunlightradiation(MJ/m2),sunshinedu-
ration (h), day length (h), cloud cover (Okta) (National Weather
Service,n.d.),andglobalradiation(MJ/m2);25studies(28.4%)evalu-
atedprecipitation-relatedfactors,includingrainfall,snowfall,andthe
numberofrainydays;23studies(26.1%)examinedhumidity,mea-
suredasrelativehumidityordewpointtemperature;19(21.6%)in-
vestigatedbarometricpressure;andnine(10.2%)studiedtheimpactof
windspeed/velocityanddirection.Several studiesalsoaccountedfor
horizontalvisibility(Yi-Fanetal.,2016;Sarranetal.,2017;Tapaket
al.,2018),mist(Sarranetal.,2017;Tapaketal.,2018),andthenum-
berofdustyandfoggydays(Carneyetal.,1988;Noelkeetal.,2016).
Inmoststudies,differentclimaticfactorswereconsideredassepa-
rateexplanatoryvariables;however, some applied biometeorological
modelstosynthesizehumanthermalcomfortorstressscores(n=13,
14.8%).Ninearticles(10.2%)usedapparenttemperatureastheheat
indicator(Khalajetal.,2010;Wilsonetal.,2013;Basuetal.,2018;Min
et al., 2019; Yi et al., 2019; Niu et al., 2020; Zapata, 2021; Oudin
Åström et al., 2015); this thermal comfort index integrates ambient
temperature,relativehumidity, and wind velocity (Steadman,1984).
21
Fig.2. Geographicandclimatezonedistributionofstudiesincludedinthereview(n=88).
Fig.3. Typesofmeteorologicalexposuresandmentalhealthoutcomes(n=88).
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Othersusedthebodydiscomfortindex(Shilohetal.,2005)orphysio-
logicallyequivalenttemperature(PET)(Shiueetal.,2016).Finally,one
studydevelopedaspecificheatindexbygeneratingacumulativetem-
peraturemeasurereflectiveofhourlyextreme-heatexposurewithina
day(Tangetal.,2021
3.2.3. Typesofmentalhealthoutcomesexamined
Themajorityofstudiesconducted before 1990 usedself-reported
psychologicalconditions,whilemostrecentstudiesutilizedhospitalad-
mission/utilizationrecordsorclinicaldiagnosisasoutcomevariables
[e.g.17,23,65,76](n=65,73.9%).Inaddition,afewstudiesused
drugdispensation (Cornali et al.,2004; Hartig etal., 2007) (n= 2,
2.3%),symptomworseningordevelopmentofanewepisode(Molinet
al.,1996;Cornalietal.,2004;Bullocketal.,2017;Sarranetal.,2017)
(n=4,4.5%).Ofthosedrawinguponhospitalrecords(Fig.3,right),
83(94.3%)examinedmentalandbehavioralmorbidity(Obradovichet
al.,2018b;Shilohetal.,2005;Carneyetal.,1988;Obrienetal.,2014)
andsix(6.8%)usedmentalhealthrelatedcause-specificmortality(Liu
etal., 2020; Gasparrini etal.,2012;Kimetal.,2015;HoandWong,
2019;OudinÅströmetal.,2015;Rocklövetal.,2014).Mostextracted
MBD status from hospital records based on the ICD-9 or ICD-10
[e.g.,23,38](n=55,62.5%),althoughsomeusedclinicaldiagnosis
byatrainedpsychiatristbasedontheDiagnosticandStatisticalManual
ofMentalDisorders(DSM-IV)(Leeetal.,2002;Huibersetal.,2010).
AmongstudiesutilizingICDstandards,thetopfivemostexamined
outcomecategorieswereschizophrenia(n=25,45.5%),mooddisor-
ders (n = 20, 36.4 %), organic mental disorders such as dementia
(n=15,27.3%),neuroticdisorderssuchasanxietyanddepression
(n=13,23.6%),andpsychoactivesubstanceuse(n=10,18.2%).A
fewstudiesinvestigatedintellectualdisabilities(n=5,9.1%),behav-
ioralsyndromes(n= 3,5.5%),personalitydisorder(n=2,3.6%),
developmentaldisorders(n=2,3.6%),andchildhoodbehavioraldis-
orders(n=2,3.6%).Finally,31studies(56.4%)examinedMBDout-
comesthatincludedmultiplecategories.
3.2.4. Studytype,spatiotemporalscale,andstatisticalanalysis
The reviewed studies most commonly used a time series design
(n = 51, 57.9.2 %), followed by cohort (n = 17, 19.3 %), case-
crossover/case-control(n=9,10.2%),cross-sectionaldataorasingle
wavefromlongitudinaldata(n=5,5.6%),andotherdesigns(n=6,
6.8%).Time-seriesstudiesmostlyreportedalargenumberofrecords
(median=34,000),whilecohortstudiesshowedsmallersamplesizes
(median=480).Fig.4illustratesthespatialandtemporalresolutions
ofthestudies. Concerning spatial resolution, most studieswere con-
ductedatmacro-ormeso-scaleduetotheresolutionofthedatasource;
forexample,dataonhospitalizationandemergencyroomvisitswere
oftenreportedataggregatedlevels(Bulbenaetal.,2005;Bulbenaetal.,
2009;daSilvaetal.,2020)rangingfromahospitalcatchmentorneigh-
borhood(n=13,14.8 %)toanentire nation(n=19,21.6%).Re-
gardingtemporalresolution,moststudies(n=71,80.7%)examined
daily climatic factors, with a few considering monthly patterns
(10.2%).
Whenestimatingexposure-response associations, the mostwidely
usedstatisticaltechnique,especiallyinearlierstudies,waslinearmod-
eling (including generalized linear and linear mixed modeling)
(n=49,55.7%).Poissonornegativebinomiallinkfunctionswerefre-
Fig.4. Spatiotemporalresolutionofthestudies(n=88).
22
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
quentlyemployed to account for MBDincidencecountorprevalence
data;linearmixedmodelswereappliedforthenesteddatastructureof
repeated measurements from individual participants (Bullock et al.,
2017;Sarranetal.,2017). Recentstudiesfavoredthetechniquenon-
lineardistributedlagmodeling(DLNM,n=22,25.0%),whichcansi-
multaneouslyfitnon-linearexposure-responserelationships and non-
lineardelayed effects using a bi-dimensional matrix(i.e.,cross-basis)
(GasparriniandArmstrong,2010;Gasparrini,2011).Includedstudies
oftencomparedmultiplelagvaluesoftheeffects,includingcumulative
lageffects,inthemainanalysesorsensitivityanalyses.Themaximum
lagvalueswereoftendeterminedbasedontheliterature,modelfitsta-
tistics,orRR-lagplots.Asidefromlag0,themostfrequentlyusedlag
valuewas0
–
7(n=15,68.2%),followedby0
–
3(n=11,50%),0
–
2
(n=8,36.4%),and0
–
21(n=6,27.3%).Manystudiesconsidered
sevendaysasthedelayperiodforshort-termexposuretoambientcon-
ditions(Almendraetal.,2019),and21daysformedium-longexposure
(Eun-hyeetal.,2021).However,findingsrelatedtoappropriatelagval-
uesvaried.Table2providesdetailsaboutthecurvefunctions,lagval-
ues, and single-day versus cumulative effect estimates. In addition,
othermethods included Pearsonand Spearman rankcorrelation, au-
toregressiveintegratedmovingaverage(ARIMA),Coxregression,and
generalizedestimationequation(GEE)methods.
3.3. Meta-analysisoftherelationshipbetweenclimatecharacteristicsand
mentalhealth
3.3.1. ClimaticfactorsandMBD
Ofthe76studiesinthesystematicreviewset,42wereincludedin
themeta-analysis, whichprovidedestimatedreportedcomparable ef-
fectsizesthatcouldbeconvertedtorelativerisk.Asdifferentstudies
examineddifferenttypesofMBD,andsomeonlyreportedcasesofMBD
withoutdifferentiatingsubclasses(e.g.,schizophrenia,mooddisorder),
wefirstestimatedrandom-effectmodelsbycombiningallsubclassesof
MBDandconsideringMBDriskasasingleoutcomevariable(Fig.5).
Whenmultipleoutcomesormodelswerereported,weadheredtothe
rulesdetailedinSection2.5toselectonlyoneeffectsize.Asthemajor-
ityofthereviewedstudiesreportedmorbidityandonlyafewinvesti-
gatedmortalityrisk,wereportthefindingsrelatedtomorbidityfirst,
followedbymortality.Pooledeffectswerecreatedforeachexposure,
suchasthermalindex, hot/cold air temperature,andsunshine dura-
tion.Whenlessthanthreestudieswereavailableforacertainexposure-
outcomepair,wepresentedtheeffectsizesfromtheindividualstudies
forthesakeoftransparencyandcompletenessbutdidnotincludethem
in meta-analysis models. The exposure variables omitted from the
meta-analysiswerewinddirection/speed,barometricpressure,humid-
ity,andprecipitation.
3.3.1.1. Thermal comfort index and MBD. Apparent temperature is a
summaryindexthatconsiderspositivecontributionstothehumanen-
ergy budget from air temperature, relative humidity, and radiation,
alongwiththenegativecontributionofwindspeed.Drawingonthree
studiesthatexaminedapparenttemperature,ourmeta-analysisresults
suggested a heightened risk of MBD during energy budget overload
butnot energy loss conditions. Namely, when apparent temperature
was at the 90th percentile (heat overload), MBD risk was elevated
(pooledRR= 1.08, 95%CI = 1.03,1.12) compared tomedianor
minimum risk temperatures. Heterogeneity across the studies was
small (I2 = 5.3 %). Meanwhile, apparent temperature in the 10th
percentile (heat loss) was not associated with MBD (pooled
RR = 1.02, 95 % CI = 0.99, 1.05) and heterogeneity was high
(I2= 80.4%).
3.3.1.2. AirtemperatureandMBD. Themeta-analysisresultssuggested
thatheatconditionsexceedingcertainlocalthresholds(e.g.,consecu-
tivedailytemperaturesof35°Corexceedingthe97.5thor99thlocal
Table2
Studiesusingdistributedlagmodelsandselectionoflagandcumulativeef-
fects(N=22).
Author,
year
Linearity Statistical
analysis
Immediate
or
cumulative
effect
Lagtime/
exposurepriorto
outcome
Temporal
unit
Wilsonet
al.,2013
Non-
linear
Distributed
lagnon-
linear
model
Immediate
&
cumulative
0,1,2,3,0
–
2 Daily
Wanget
al.,2014
Non-
linear
Distributed
lagnon-
linear
model
Immediate
&
cumulative
0,1,2,3,4,5,6,
7,8,9,10,11,
12,13,14,15,
16,17,18,19,
20,21,22,23,
24,25,26,27,
28,29,30,0
–
7
Daily
Penget
al.,2017
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0,0
–
1,0
–
2,0
–
3,
0
–
4,0
–
5,0
–
6,0
–
7,0
–
14,0
–
21
Daily
Chanet
al.,2018
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0
–
2,0
–
8 Daily
Leeetal.,
2018
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0
–
7 Daily
Sherbakov
etal.,
2018
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0
–
3 Daily
Wanget
al.,2018
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0,0
–
1,0
–
2,0
–
3,
0
–
4,0
–
5,0
–
6
Daily
Almendra
etal.,
2019
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0,0
–
1,0
–
2,0
–
3,
0
–
4,0
–
5,0
–
6,0
–
7
Daily
Guetal.,
2019
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0
–
7,0
–
14,0
–
21 Daily
Minetal.,
2019
Non-
linear
Distributed
lagnon-
linear
model
Immediate
&
cumulative
0,1,2,3,4,5,6,
7,8,9,10,11,
12,13,14,15,
16,17,18,19,
20,21,0
–
1,0
–
2,
0
–
3,0
–
4,0
–
5,0
–
6,0
–
7,0
–
8,0
–
9,
0
–
10,0
–
11,0
–
12,0
–
13,0
–
14,
0
–
15,0
–
16,0
–
17,0
–
18,0
–
19,
0
–
20,0
–
21
Daily
Panetal.,
2019
Non-
linear
Distributed
lagnon-
linear
model
Immediate 0,1,2,3,4,5,6,
7,8,9,10,11,
12,13,14,15,
16,17,18,19,
20,21
Daily
Xuetal.,
2019
Non-
linear
Distributed
lagnon-
linear
model
Immediate 0,1,2 Daily
Yietal.,
2019
Non-
linear
Distributed
lagnon-
linear
model
Immediate
&
cumulative
0,1,2,3,4,5,6,
7,8,9,10,11,
12,13,14,0
–
1,
0
–
2,0
–
3,0
–
4,0
–
5,0
–
6,0
–
7,0
–
8,
0
–
9,0
–
10,0
–
11,
0
–
12,0
–
1,0
–
14
Daily,0
–
14
cumulative
days
(continuedonnextpage)
23
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table2(continued)
Author,
year
Linearity Statistical
analysis
Immediate
or
cumulative
effect
Lagtime/
exposurepriorto
outcome
Temporal
unit
daSilvaet
al.,2020
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0,0
–
6,0
–
7 Daily
Liuetal.,
2020
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0,1
–
5,6
–
21,0
–
21
Daily
Niuetal.,
2020
Non-
linear
Distributed
lagnon-
linear
model
Immediate
&
cumulative
0,1,2,3,4,5,6,
7,0
–
1,0
–
2,0
–
3,
0
–
4,0
–
5,0
–
6,0
–
7
Daily,0
–
7
cumulative
days
Zhanget
al.,2020
Non-
linear
Distributed
lagnon-
linear
model
Immediate
&
cumulative
0,1,2,3,4,5,6,
7,8,9,10,0
–
1,
0
–
3,0
–
5,0
–
7,0
–
9
Daily
Bundoet
al.,2021
Linear Distributed
laglinear
model
Cumulative 0
–
3,0
–
7 Daily
Eun-hyeet
al.,2021
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0,0
–
3,0
–
7,0
–
14,0
–
21
Daily
Jahan&
Wraith,
2021
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0
–
3,0
–
7,0
–
10,
0
–
15,0
–
21,
−
0-
28,0
–
30
Daily
Yooetal.,
2021
Non-
linear
Distributed
lagnon-
linear
model
Cumulative 0
–
7 Daily
percentile)were consistently associated with increased risk of MBD:
heatwave,pooledRR=1.05,95%CI=1.02,1.08;97.5thpercentile
airtemperature,pooledRR=1.18,95%CI=1.07,1.30;99th per-
centileairtemperature,pooledRR=1.18,95%CI=1.08,1.29.In
contrast,therelationshipofcoldweatherandMBDwaslessconsistent
acrossmeasures;temperaturesinthe1stpercentile(pooledRR=0.97,
95%CI=0.86,1.09)and2.5thpercentile(pooledRR=1.14,95%
CI=0.94,1.35)were not associated with MBD.Poolingoffindings
fromstudiesthatassumedalog-linearmonotonalrelationshipbetween
1°CincreaseintemperatureandMBDriskalsoyieldedinsignificantre-
sults(pooledRR=1.01,95%CI=0.99,1.03),althoughresultswith
a linearity assumption needed to be interpreted with caution (see
Section 4.3). Furthermore, Four studies among them considered the
year-roundtemperaturerange(Kimetal.,2021;Huibersetal.,2010;
Trangetal.,2016b;Oh et al., 2020), (e.g., May to September in the
Northern Hemisphere or October to March in the Southern Hemi-
sphere)(Williamsetal.,2012;Vidaetal.,2012;Bundoet al.,2021),
whiletheotherfiveuseddatayear-round.Finally,whileheterogeneity
waslowforheatwave
–
MBD(I2=17.2%),itwashigh(I2>80%)
forallothermodels,likelyduetothevastdifferencesinpopulations,
climateconditions,andhightemperatureexposurevariables.
Studiesoncause-specificmortalityrelated toMBDmostlyfocused
onairtemperaturemetrics.Poolingtheeffectsofthethreestudiesthat
assumedlog-linearity,wefounda1°Cincreaseintemperaturetobeas-
sociated with 3 % increased risk of MBD-related mortality (pooled
RR = 1.03, 95 % CI = 1.02, 1.04), and low heterogeneity
(I2=0.0%).
3.3.1.3 Solar radiation/sunshine, barometric pressure, precipita-
tion,andhumidity,andMBD.
Amongthevariousexposuremeasuresrelatedtosunshineandsolar
radiation,onlysunshinedurationbyhourwasusedbythreestudiesand
therefore amenable to risk estimation using a random effects meta-
analysismodel.This estimation didnot show anassociation (pooled
RR= 0.98, 95% CI = 0.93, 1.03). Regardingbarometric pressure,
onlytwostudieswereavailable.Likewise,forprecipitationandrelative
humidity,avarietyofvariableswereexamined,eachinasinglestudy;
theseincludedrainvolume(mmorpercentilethreshold),rain/snow/
foggyday,andrelativehumidity(%).Thus,meta-analysescouldnotbe
performedfortheseexposures.
3.3.2. Outcome-specificanalysis
Tounderstandthedirectionandmagnitudeoftheassociationsbe-
tweenclimatefactorsandspecificsubcategoriesofMBD,wetalliedef-
fect sizes according to the specific MBD subclass reported and per-
formedameta-analysisoneachsubclass.Aftersortingtheclimatefac-
tor-mentaldisorderpairs, only threeoutcomecategories were repre-
sented in at least three studies with the same exposure-outcome
pairs
—
schizophrenia (F20
–
F29, n = 4), mood disorders (F30
–
F39,
n=7),andneuroticdisorders(F40
–
F49,n=6)
—
andthusabletobe
pooledusingrandom-effectmodels.Otheroutcomessuchasdementia
andbehavioralsyndromeswerenotincludedinthemeta-analysisdue
tofeaturinginlessthanthreestudies.Wepresentforestplotsshowing
thepooledeffects of the threeanalyzableexposure-outcome pairs in
Fig.6,whilethecompletesetofindividualeffectsizesarepresentedin
SupplementarymaterialFigs.S4-1toS4-6.
3.3.2.1. Schizophrenia. Four studies fitted models to compare 99th
percentile and median/minimum air temperatures for relation to
schizophrenia.Basedonthesedata,wefound a 99th percentile high
temperature to be associated with higher risk of schizophrenia
(pooled RR = 1.07, 95 % CI = 1.01, 1.12), with low to moderate
heterogeneity (I2 = 40.4 %).
3.3.2.2. Mood disorders. Pooled results from four studies estimating
theeffect of a 1 °C increase in temperature yielded insignificant re-
sults(pooled RR = 1.00, 95 % CI = 0.97, 1.03), while the pooled
associationof 99thpercentiletemperatureandelevatedrisk ofmood
disorderapproachedstatisticalsignificance(pooledRR=1.23,95%
CI = 1.00, 1.51). These models featured high heterogeneity scores
(I2 > 80 %) due to geographical and population differences, along
with variation in outcome classification: three studies defined their
outcomevariableusingICD-9or10,wherethefourthusedtheDSM-
IV.Moreover, Kim et al. (2015) only considered ICD codes F31
–
33,
whilethe other twoconsidered F31
–
40.
3.3.2.3. Neuroticdisorders. Pooledresultsrelatedtoneuroticdisorders
(e.g., depression) were generally not significant. The association of
1°Cincreaseintemperaturewithneuroticdisordersdidapproachsig-
nificance (pooled RR = 1.02, 95 % CI = 1.00, 1.04), but that of
99th percentile temperature did not (pooled RR = 1.11, 95 %
CI = 0.92, 1.33). Heterogeneity scores were high for these models
(I2> 75 %) for reasons similar to those described above.
3.4. Riskofbiasandcertaintyofevidence
Fig.7showstheresultsoftheriskofbiasassessmentbystudyand
assessmentcategory.Theindividualscoresforeachitemevaluatedare
givenin Supplementary material S5Fig.S5
–
1.Across all studies,the
majorbiasconcernssamplingselection, outcome measurements,and
confounding factors (Supplementary material S5 Fig. S5
–
2). Studies
thatutilizedrecordsfromafewhospitalswithoutexplainingthesam-
plingmethod or theirrepresentativenessofthepopulationwerecon-
sideredtohaveelevatedbias.Afewearlyindividual-levelstudiesuti-
lized convenient samples from students, hospitals, or online panels
(Barnston,1988),whichmightbesubjecttovolunteerbiasandhence
wereconsideredprobablyhighbias.Regardingoutcomemeasurement,
studiesthatextractedoutcomesfromhospitalrecordsusingstandard
classificationsystems,suchasICD-9or10orDSM-IV,wereconsidered
24
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Fig.5. Forestplotofclimate/meteorologicalfactorsandriskofmentalandbehavioraldisorders(*denotesoutliers;REModel).
tohavelowriskofbias,aswerethoseinvolvingclinicaldiagnosisbya
trained psychiatrist or self-report based on validated diagnostic/
screeninginstruments.Studiesthatusedmeasuresnotvalidatedforthe
specificconstructtobeexaminedweregradedashavingprobablyhigh
bias.Inaddition,iftheoutcomewasclinicallydiagnosed,weconsid-
eredwhether the assessors wereblindedtoexposureconditions;out-
comesmeasuredby self-reported scale wereuniformlyconsidered to
haveelevated bias. Forconfounding factors, themajority of popula-
tion-andindividual-levelstudiesincludedtime-invariantfactorssuch
asage,sex,socioeconomicconditions,andpre-existing health condi-
tions,time-variantfactors such asseason, day ofweek, time of day,
andother exposurefactorssuchasair pollutionandnoiselevel.Sev-
25
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Fig.5.(continued)
eral studies that referenced morbidity records used auxiliary place-
baseddatasetsthatprovidedpopulation-leveldemographicconditions.
Studiesthatonlyadjustedfor temporalfactors(season,day of week,
diurnal)but didnotincludestudypopulation/individual- orenviron-
ment-relatedconfoundingfactorsweregradedashavinghigherriskof
bias,while thosethatdidnotconsideror adjustforanyconfounding
factorsintheirexposure-response estimates were consideredtohave
probablyhigh risk of bias.Finally, selective reportingof results was
observedinsomestudieswhenmultipleexposure,outcome,orstatisti-
cal models were mentioned in the planned analyses, but only those
withsignificantresultswerepresentedanddiscussed.
UsingtheGRADEcriteria,weassessedthecertaintyofevidencefor
eachcategorywherepooledeffectswereproduced.Weassignedabase-
lineofmoderatecertaintybasedonthestudydesign,followedbydeci-
sionstodowngradeorupgradeforeachoftheGRADEdomains.Effect
estimatesforwhichmorethanhalfofthecontributingstudiesexhibited
26
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Fig.6. Forestplotofclimate/meteorologicalfactorsandschizophrenia,mooddisorders,andneuroticdisorders.(*denotesoutliers;REModel).
greaterthanlowriskofbiasinatleastonedomainweredowngraded
byonelevel,suchastheassociationsbetweenMBDandheatwaveand
sunshineduration.Allstudiesmetthespecifieddesiredpopulation,ex-
posure,comparator,and outcome criteriaand were notdowngraded
basedonthosecategories.Forexample,nostudyusedasoutcomeasur-
rogateofaclinicalendpoint.Whenweidentifiedinconsistencies,such
asapredictioninterval more than twicetheconfidenceinterval and
studiesdisagreeinginthedirectionoftheassociation(i.e.,positiveand
negativeassociationforthesame exposure-outcome pair), wedown-
gradedtheevidence;acaseinpointistheassociationofhighairtem-
peratureandMBD.Forthenumberofparticipantsandspecificallyper-
son-time,allbutonestudyusedatime-seriesdesignorfeaturedalarge
cohort.Theonlystudythatdidnotreportthenumberofparticipants
usedhospitaladmissionsrecordsduringa16-yearspanandwas thus
consideredtohaveadequatesamplesize.Therelativerisksyieldedby
therandomeffectsmeta-analysismodelsweresmall,andthereforeno
evidence was upgraded for large magnitude of effect. As meta-
regressionwasnotpossibleduetothesmallnumberofstudies, there
was insufficient evidence for a dose-response gradient. Confounders
mayeitherhavepositiveornegativeimpactsontheassociations,and
thereforeno upgradingwasperformedforthe criteriaregardingcon-
founding.
Ultimately,basedontheoutcomesofthestructuredGRADEprocess,
thereisamoderatelevelofevidencesupportingthatMBDiselevated
whenthe thermal index(apparent temperature) increases.Similarly,
theassociationbetweenriskofschizophreniaandhighairtemperature
hasmoderatecertainty.Otheridentifiedassociationssuchasbetween
heatwaveor high temperature and MBDstill require more attention
duetohavingonlylowcertaintyofevidence.Table3presentsthede-
tailsoftheevaluationoutcome.
4. Discussion
4.1. Mainfindings
Inthisstudy,weperformedasystematicreviewandmeta-analysis
toexaminetheassociationsbetweenclimaticandmeteorologicalfac-
torsandmentalhealth.Theresultsrevealedunevengeographicaland
climatezonedistributions and uncoveredunderrepresentedclimates.
Temporally, earlier studies focused on psychological states using
smallersamples,recentstudiesemphasizedpsychiatricmorbiditywith
longitudinal/time-seriessurveillancedata.Extremeheatbasedonheat
thresholdsandhighthermalindexintegratingtemperature,humidity,
andwindvelocity were identifiedas risk factorsforMBD, while ex-
treme cold was not. Evidence was limited to generate medium-high
confidencepooledeffectssizesforhumidity,wind,andsolarradiation.
Regarding different subclasses of MBD, schizophrenia risk increased
whentemperatureroseabovethe99thpercentile,andmoodandneu-
27
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Fig.7. Riskofbiasassessmentresults.
roticdisorderriskswereapproachingsignificancewithtemperaturein-
crease.
Ourmeta-analysisrevealedthat,ingeneral,heatextremesandtem-
peraturesexceedinglocalthresholdswererelatedtomentalandbehav-
ioraldisorders;specifically,theseassociationswereidentifiedamong
studiescomparingheatwavewithnon-heatwavedaysandalsoamong
studiesestimatingMBDriskin99thor97.5thpercentiletemperatures
relativetomedianorminimumtemperatures.Theseresultsareconsis-
tent with findings from prior systematic reviews and meta-analyses,
whichindicatedhightemperaturesandheatwavestoberiskfactorsfor
mentaldisorders(Thompsonetal., 2018;Liuetal.,2021).Neverthe-
less,thecertaintyofevidenceisconsideredloworverylowformostex-
posure-outcomepairsduetoriskofbiasissuesandhighheterogeneity.
Thompsonetal.(Thompsonetal.,2018)reportedasimilarfindingbe-
tweenheatandmentalhealthoutcomes,althoughameta-analysiswas
notconducted due toheterogeneity. Liu et al. (Liuet al., 2021)ob-
tainedpooledeffectsizes(RR)forvariousdefinitionsofheatwave,and
reported pooled RRs between 1.048 and 1.753. Our effect estimate
28
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Table3
Evaluationofcertaintyofevidence.
Ratingfactors D1 D2 D3 D4 D5 U1 U2 U3 Finalrating
Apparenttemperature(heat)
–
MBD
0 0 0 0 0 0 0 0 Moderate
(+++)
Apparenttemperature(cold)
–
MBD
0 0
−
1 0 0 0 0 0 Low(++)
Heatwave
–
MBD
−
1 0 0 0 0 0 0 0 Low(++)
Airtemperature(high
temperature)
–
MBD
0 0
−
1 0 0 0 0 0 Low(++)
Airtemperature(low
temperature)
–
MBD
0 0
−
1 0 0 0 0 0 Low(++)
Airtemperature(linear
increase)
–
MBD
−
1 0
−
1 0 0 0 0 0 Verylow
(+)
Sunshineduration
–
MBD
−
1 0
−
1 0 0 0 0 0 Verylow
(+)
Airtemperature(high
temperature)
–
Schizophrenia
0 0 0 0 0 0 0 0 Moderate
(+++)
Airtemperature(high
temperature)
–
mood
disorders
0 0
−
1 0 0 0 0 0 Low(++)
Airtemperature(linear
increase)
–
mooddisorders
−
1 0
−
1 0 0 0 0 0 Verylow
(+)
Airtemperature(high
temperature)
–
neurotic
disorders
0 0
−
1 0 0 0 0 0 Low(++)
Airtemperature(linear
increase)
–
neurotic
disorders
−
1 0
−
1 0 0 0 0 0 Verylow
(+)
overlapwith the lower endof the RRrange, for estimatesincluding
heatwave(RR=1.05,95%CI=1.02
–
1.08),97.5thpercentiletem-
perature(RR=1.18,95%CI=1.07
–
1.30),and99thpercentiletem-
perature(RR=1.18,95%CI=1.08
–
1.29).Thedifferenceontheup-
perendoftheRRswaslargelyattributabletothefactthatthepresent
studydid notconsideroutcomessuchas suicideideationorattempts
whileLiuetal.did.Coldextremes,ontheotherhand,werenotsignifi-
cantlyassociatedwithMBDrisk.Nopreviousstudiesareavailableon
thisrelationshiptocomparetheeffectestimate.
Beyondtemperaturemetricsalone,wealsoidentifiedathermalin-
dex,which combines temperature,humidity, wind speed, andradia-
tion,tobefactorforelevatedrisksofMBD.Temperatureisthetopcli-
maticfactorcontributingtothermalstress,butnottheonlyone;humid-
ityandradiation variations canhave as largeeffectsas temperature
change.Whilepopularmetricsusing99thor97.5thpercentiletempera-
turesweresignificantbutlow-certaintycorrelatesofMBD,thermalin-
dexatthe90thpercentileyieldedresultswithmoderatecertainty.This
indicatestheimportanceofconsideringmultiplesourcesofheatstrain
inthe course of human-environment heatexchange.Studiesonsport
andoccupationalmedicineusingestablishedthermalmodels such as
physiological equivalent temperature (PET), the Comfort Formula
(COMFA),andwet-bulbglobetemperature(WBGT)havedemonstrated
thesemodelstohavestrongexplanatorypowerassummariesofther-
malstress(Budd,2008).Meanwhile,wefoundnosignificantrelation-
shipswhenconsideringsunshinedurationalone,likelyduetothesmall
numberofstudiesthatusedthismetric.Otherclimateexposures,such
ashumidity,windspeed,barometricpressure,andsolarradiation,were
measuredheterogeneously and couldnot be pooled,thus the results
wereinconclusive.
Whenitcomestospecificmentaldisorders,ingeneral,studiesusing
homogeneousclimate/meteorological metrics were lacking. Previous
systematic reviews that provided narrative summaries of patterns
mostlynotedthathightemperaturesmightbeatriggerofschizophre-
nia,bipolardisorder,anddementia(Thompsonetal.,2018;Monteset
al., 2021; Bongioanni et al., 2021). To our knowledge, our study is
amongthefirstto conduct a meta-analysisonvariousmental health
outcomes,especiallyforeachtypeofMBDbasedonICDclassifications
and other self-reported psychological states. The only significant
pooledeffectidentifiedwasanassociationof99thpercentiletempera-
turewithschizophrenia,whichalsoshowedmoderatecertaintyofevi-
dence.Therelationshipsoftemperaturewithmoodandneuroticdisor-
dersapproached but did not achievestatistical significance, and the
confidenceinthebodyofevidencewasassessedaslowatbest.
Theheterogeneityof effect estimates was high on most exposure-
outcomepairs,likelyduetodifferencesinclimateconditions,popula-
tioncharacteristics,andheatadaptationandmitigationcapacities.It
haspreviouslybeennoted that general populationstudiesmay yield
heterogeneityvaluesintherangeof75%andabove,higherthanthose
reportedinrandomizedcontrolledtrialsorothersmallerstudies(Chen
andHoek,2020).
4.2. Mechanismsoftheobservedassociationsbetweenclimateconditions
andmentalhealth
Themechanismsunderlyingtheobservedassociationsbetweencli-
mate/meteorological factors and mental health may be multifaceted
andcomplex.First,the associations between heatandmentalhealth
may be explained by dysregulation of the serotonin system (5-
hydroxytryptamine,5-HT),whichis involved in moodandcognition
modulation.Studieshaveshownthatheatandhumiditymaycausewa-
terdeprivationanddehydration,whichleadtoadecreaseinserotonin;
suchdeficiencyisrelatedtoincreasedmooddisorders,depression,anx-
ietydisorders,schizophrenia,andotherMBDs(Linetal.,2014).Ithas
alsobeensuggestedthat,asserotoninplaysanessentialroleinthermal
regulation,acuteambienttemperaturechangesmaycausesystemdys-
regulation,therebyaggravatingmentalhealthsymptoms(Craneetal.,
2015).Ontheotherhand,manyantipsychoticandpsychotropicmed-
icationshavethermoregulatorysideeffects,andintakeofsuchmedica-
tionsmay berelatedtocompromisedthermoregulationandperspira-
tion(Zammitetal.,2021).Second,recentresearchhasindicatedthat
heatstresscaninduceneurodegeneration;inparticular,heatstrokecan
cause excitotoxicity, necrosis, and apoptotic death of neuronal cells
(Gongetal.,2022;Zammitetal.,2021;Kourtisetal.,2012).Clinical
studiessuggestthatneuronalcellandneuralnetworkalterationscon-
tributeto the pathologyofmentaldisorders(Quachetal., 2016). Fi-
nally,inadditiontobiologicalmechanisms,theremaybeasocioeco-
logicalexplanationofthelink:extendedperiodsofheatmayinfluence
individualand family daily routines andsocial networks,presenting
challengesfor commute and childcare routines, and alsoincur emo-
tionalstrainrelatedtofamilymemberswithpre-existinghealthcondi-
tions;alloftheseareexternalstressorstomentalhealth.Adversemicro-
climateconditionsalsolimitopportunitiestoengageinoutdooractivi-
tiesthatofferrestorativeeffects(e.g.,visitingaparkorabeach)(Hartig
etal.,2007;Lietal.,2019).Finally,disastersituationssuchasheatand
coldstormsoftenalsocomewithdisruptionoftransportationanden-
ergy/water infrastructures (Li et al., 2022b), exacerbating the psy-
chosocialstressexperiencedbyindividuals.
Onepotential pathway that was notsubstantiated in the current
studyishumansunlightexposureandthecircadianclock.Specifically,
researchon seasonalaffectivedisorderanddepressionhas outlineda
circadiantimingmechanismviathehypothalamicsuprachiasmaticnu-
cleusthatisdependentonambientconditions,withthelight-darkcycle
beingconsidered themostrelevantsynchronizerforthebody's circa-
diantiming(e.g.,sleep-wakecycle,locomotoractivity,hormones);dis-
turbanceofthattimingisrelatedtodepressionanddistress(Mendoza,
2019).
4.3. Futuredirectionsandlimitations
Whiletherehasbeenarecentsurgeofresearchinterest,giventhe
broadarrayofclimaticfactors,thevarioustypesofmentalhealthcon-
ditions,andtheinsufficientevidenceofassociationdeterminedinthis
studyformostexposure-responsepairs,moreresearchiswarrantedre-
29
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
gardingeachexposure-response pair and alsothejoint effects of cli-
maticexposures.Here,wediscusspotentialfuturedirectionsforfilling
critical knowledge gaps and addressing the risk of bias and other
methodologicalconcerns.
4.3.1. Standardizingmeteorologicalmetricsandthresholds
Thewidevarietyofmeteorologicalmetricsusedinincludedstudies
reducestheabilitytoperformmeta-analysisandpromoteshighhetero-
geneity.Forexample,measuresusedtorepresentsunshineandsolarra-
diationincludeddailysunshineavailability,duration,solarradiation,
daylength,cloudcover,anddirectandglobalradiation.Likewise,for
airtemperature,studiesestimatedtheeffectsof1°Cincreaseindaily
mean or maximum temperature, 1 °C increase in daily temperature
range,variouspercentilethresholds(e.g.,99,97.5,95,90,75),andvar-
iousdefinitionsofaheatwave.Werecommendthatfuturestudiesse-
lectmetricsthatprovideadequategranularityandreflectbiometeoro-
logicalmechanisms. Regarding temperature, for instance,dailymean
ormaximum temperature mayfail to capturethe critical impactsof
nighttimetemperatureonmental health, especially duringmulti-day
heatevents.Theliteraturealsoindicatesthathighnighttimetempera-
ture,especiallyifpersistent,isrelatedtosleepingdisturbances(Haskell
etal.,1981;LibertandBach,2005),whichcouldtriggerepisodesof
mentaldisorders(LiewandAung,2021).Italsoremainsnecessaryto
considerthejointeffectofmultiplemeteorologicalconditions.Asther-
malstressisrelatedtomost recognizedbiologicalneurodegeneration
pathwaysinvolvedinmentalhealth,futurestudiesmayconsiderusing
establishedthermalindicesthatconsidermetabolic,conductive,con-
vective,andradiativeheatflux,eitherasthemainexposurevariableor
inthesensitivityanalysis.
4.3.2. Addressingscalemismatchesandattentiontomicroclimateexposure
Although the majority of studies included in this review utilized
ground-measuredmeteorologicalconditions,spatialmismatchesoften
existbetweenthemeasurementstakenatweatherstationsandthemi-
cro-environmentsexperiencedbyindividualparticipants.Studiesthat
considerclimateratherthanmicroclimatemaybesubjecttotheuncer-
taingeographiccontextproblem(UGCoP):namelythatarealmeasures
donotmatchtheenvironmentalexposurethatexertsimpactsonhealth
andbehavior,aproblemthathasbeendiscussedingeographyandspa-
tialepidemiology(Kwan,2012;Jiaetal.,2020).Nevertheless,studies
onhuman exposure to heat, wind,andsolarradiationreviewedhere
stillmostly usedareal-basedclimatemeasuresderivedfrom sparsely-
locatedweatherstationswhichmaybetoocoarsetoestablishexposure-
responserelationships.Infact,microclimateconditionscanbemodi-
fiedbyurbanheatislandeffects,site-levelstructuralelementsandveg-
etation,waterbodies,andthealbedoofmaterials.Allofthesefactors
directlyimpacthumanphysiology,thermalsensation,andbehavioral
adaptationatagivenlocation.Futurestudiesarewarrantedthatpre-
ciselymeasurethemicroclimatestowhichindividualsareexposedin
residentialneighborhoodsandthirdplaces,suchasthroughtechnolo-
gieslikedownscalingand/ormicroclimateloggernetworks(Georgeet
al.,2015).
4.3.3. Studyingdisease-specificexposuresandpathways
Morestudiesareneededonspecificmentaldisorders(e.g.,schizo-
phrenia,depression,anxiety,dementia).Inthesetofstudiesreviewed
here,manyutilizedcombineddiseaseprevalencefrommultipleMBDs
thatmayreflectdifferentriskpathways,leadingtoresultsthatarelev-
eledoutandchallengingtointerpret.Forstudiesthatconsideredspe-
cificmentaloutcomes,ICDandDSMcodesweretypicallyused;how-
ever,ICDclassificationsdependongoodclinicaljudgmentandcanbe
lessaccurate,whiletheDSMislesscommonlyusedoutsidetheU.S.In
addition,admission/dischargerecordswereoftenaggregated without
distinctionofclinicalstages.Astheimpactsofriskfactorsandinterven-
tionsmaydifferoverthediseasecourse,itiscriticaltoconsiderdisease
trajectoryand climateat differentstages suchas in prodromal, first
episode,persistent,andremittedcases.Additionally,morestudiesthat
addressneurotransmittersandothermechanismswouldhelpadvance
knowledgeandpolicy.
4.3.4. Expandingoutcomesfrommorbiditytowell-being
Publichealthpoliciesacrosstheworld,suchastheHealthyPeople
2030frameworkoftheU.S.,havebeenadvocatingfornotonlydisease
prevention/treatmentbut also forpromoting people to achievetheir
fullpotential.Althoughthisreviewsetouttoframementalhealthina
broaddefinitionthatencompasseshappinessandwell-being,thenum-
berofstudies employing such constructswas limited. Therefore,the
quantitativesynthesismostlyaddressedMBD;obtainingpooledeffects
foraffectandotherpositiveemotionswasnotpossibleduetothesmall
numberofstudiesandhighheterogeneity.Futurestudiesarewarranted
thatusevalidinstrumentstoexaminewhetherclimate-relatedrisksim-
pactsubjective well-being. In addition toidentifying modifiablerisk
factors,thereisvaluein adoptingasalutogenicview thatdetermines
theambientconditionsthatprovideforaqualityexperienceandsup-
portsubjectivewell-being.
4.3.5. Modelingtechniquethataccountsfornon-linearanddelayed
dependencies
Thetwomainfactorsthatpertainedtostatisticalmodelselectionfor
estimatingclimateandmentalhealthrelationshipswerenon-linearity
andlagged andcumulativeeffects.Recentevidencesuggeststhat the
relationshipsbetweencontinuousclimaticand meteorological factors
andmentalhealthoutcomesarenotlinearormonotonicallyincreas-
ing/decreasing(Almendraetal.,2019).Infact,humanphysiologysug-
geststhatforclimaticconditionswithinareasonablerange,theenergy
budgetismaintained,meaningthatextremevaluesatbothendsmay
havenegativeimpacts(Eun-hyeetal.,2021).Toaccountforthenonlin-
earity,studies can use heatandcoldthresholdsoronlyrecordsfrom
warmseasons.Additionally,recent studies that considerednonlinear
exposure-responserelationships haveoften fittedmodels with spline
functions(Yooetal.,2021;Leeetal.,2018b;Bundoetal.,2021).Ofall
studiesthatconsiderednon-linearexposure-responsecurvesusingdis-
tributedlagmodels,onlyonefoundalinearrelationship;theremainder
presentedcurvilinear functions for both the predictor and lags. This
raisesaquestionastothevalidityofthefindingsofstudiesthatassume
linearity/log-linearitywithout appropriate theoreticalgroundsorsta-
tistical tests. In the absence of specified geographic/seasonal condi-
tions,interpretationsofMBDriskasincreasingwithevery1°Cincrease
in air temperature can be misleading. Future meta-analysis articles
shouldalsousecautionwhenassuminglinear/log-linearormonotonic
relationships.Standardizationandmethodologicalcomparisonsacross
modelspecificationsandsingle-dayandcumulativelagvaluesmaybe
informative.Asmorestudiesaccumulatethatemployatwo-stagetime-
seriesdesign to estimate location-specific exposure-response relation-
ships,advancedmultivariatemeta-analysismethodsshouldbeconsid-
eredforfuturereviews(GasparriniandArmstrong,2013).
Limitationsofthisreviewalsoneedtobenoted.First, one major
limitationis the small numberof studies foreach exposure-outcome
pair.Oftenonlythreestudieswereavailable,renderingitimpossibleto
performsensitivityanalysis,excludeoutliersandstudiesbasedonspe-
cificconsiderations,andexaminetheimpactsofsuchexclusionsonthe
meta-analysisresults.Second,ourreviewdidnotincludeexperimental
andquasi-experimentalstudies,andhencethefindings,dependingon
theconfidenceofevidence,contributetobutcannotdirectlyestablish
causality. Third, we did not include outcomes related to suicide
ideationand attempts. Althoughpreviously suicide has been consid-
eredasymptomassociatedwithotherpsychiatricdisorders,theDSM-5
positionedsuicidalbehavioraldisorderasanindependentconditionfor
furtherstudy(DSM-IV-TR,2000).Fourth,weexcludedstudiesconsid-
eringmaternalexposureand child mental health conditions, because
30
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
theremaybeadditionalbiophysiologicalmechanismsinvolved.Never-
theless,we recognize theimportanceofstudyingreproductivehealth
andlinked lives in climate exposure. Fifth, althoughtheGRADEtool
providesa greatframework for evaluating the certainty of evidence
(Balshemetal.,2011;Schünemannetal.,2019a),ithaslimitationsin
assessing evidence from nonrandomized/observational research
(Schünemannetal.,2019b).Finally,inlightofourresearchquestions,
weconducted outcome-specific analysis, but could not perform sub-
groupanalysisbasedonage,gender,orgeographyduetothehighhet-
erogeneityandsmall number ofstudies available for eachexposure-
outcomepair.Narrativereviewevidence,suchasSharpeandDavison
(Sharpe and Davison, 2021), highlighted the links between climate-
relateddisastersandmentaldisordersspecificallyforlow-andmiddle-
incomecountries.Astheevidencebasecontinuestogrow,futurestud-
iesshould focus onhealthdisparitiesacrosspopulationgroups,espe-
ciallyvulnerablegroupssuchasolderadults,low-incomegroups,and
sociallyisolatedpopulations,whichmayofferscientificinsightsandin-
formtargetedpolicyinterventions.
5. Conclusion
Astheimpactofclimatechangeonmentalhealthgainsincreasing
recognition,acomprehensiveinvestigationoftherelationshipbetween
variousclimaticandmeteorologicalfactorsandmentalhealthcanre-
vealgapsinknowledgeandinformpolicyrelatedtopublichealthand
climateadaptation.Accordingly,utilizingthePRISMAframework,this
meta-analysisaimedtoidentifyfactorsassociatedwithincreasedriskof
mental and behavioral disorders. Our random-effects meta-analysis
modelsrevealedthathigherthermalindexvalues,heatwaves,andex-
tremetemperaturessurpassing certain thresholdswerelinked to ele-
vatedrisksofMBD.Furthermore,whenconsideringspecificsubtypesof
MBD,hightemperatureswerefoundtobeassociatedwithanincreased
riskofschizophrenia.Notably,theresultsunderscoredtheheterogene-
ityof exposuremeasuresandascarcityof evidenceregardingtheef-
fectsof climatefactorsotherthanairtemperature, suchashumidity,
wind,solarradiation,andbarometricpressure.Thefindingsalsounder-
scoredtheimportanceofinvestigatingthesynergisticeffectsofmulti-
pleclimatefactorsusingthermophysiologicalmodels,non-linearexpo-
sure-outcomerelationships,andcumulativeandlagged effects of cli-
mateexposure.
Funding
ThisworkwassupportedbytheNationalAcademiesofSciencesEn-
gineering and Medicine Gulf Research Program [grant numbers
#2000012329and2000013443]; and theNationalInstitute of Envi-
ronmentalHealthSciences[grantnumber#P42ES027704-01],andthe
HoustonMethodistResearchInstitute.
CRediTauthorshipcontributionstatement
DongyingLi1:Conceptualization,methodology,studyscreeningand
eligibility,biasassessment,synthesis&analysis,visualization,writing
–
originaldraft,writing
–
review&editing,andprojectadministration.
YueZhang:Conceptualization, methodology, studyscreening and
eligibility,biasassessment,synthesis&analysis,visualization,writing
–
originaldraft,andwriting
–
review&editing.
XiaoyuLi:Conceptualization,methodology,studyscreeningandeli-
gibility,biasassessment,synthesis&analysis,visualization,writing
–
originaldraft,andwriting
–
review&editing.
KaiZhang:Conceptualization,methodology,studyscreeningandel-
igibility,biasassessment,synthesis&analysis,visualization,writing
–
originaldraft,andwriting
–
review&editing.
1Guarantorofthereviewprotocol.
YiLu: Conceptualization, methodology, study screening andeligi-
bility, bias assessment, synthesis & analysis, visualization, writing
–
originaldraft,andwriting
–
review&editing.
Robert Brown: Conceptualization, methodology, study screening
and eligibility, bias assessment, synthesis & analysis, visualization,
writing
–
originaldraft,andwriting
–
review&editing.
Uncitedreference
Declarationofcompetinginterest
Theauthorsdeclarethefollowingfinancialinterests/personalrela-
tionshipswhichmaybeconsideredaspotentialcompetinginterests:
DongyingLireportsfinancialsupportthatwasprovidedbytheNa-
tionalAcademiesofSciencesEngineeringandMedicineGulfResearch
Program.Dongying Lireportsthatfinancialsupportwas providedby
theNational Institute ofEnvironmentalHealthSciences.DongyingLi
reportsthatfinancialsupportwasprovidedbytheHoustonMethodist
Hospital.
Dataavailability
Dataaresharedinthesupplementarydocuments
AppendixA. Supplementarydata
Supplementarydata tothisarticlecanbefoundonlineathttps://
doi.org/10.1016/j.scitotenv.2023.164435.
References
Alexander, P .,2013.Associationofmonthlyfrequenciesofdiversediseasesinthecallsto
thepublicemergencyserviceofthecityofBuenosAiresduring1999
–
2004with
meteorologicalvariablesandseasons.Int.J.Biometeorol.57(1),83
–
90.
Almendra, R .,etal.,2019.Short-termimpactsofairtemperatureonhospitalizationsfor
mentaldisordersinLisbon.Sci.TotalEnviron.647,127
–
133.
Balshem, H .,etal.,2011.GRADEguidelines:3.Ratingthequalityofevidence.J.Clin.
Epidemiol.64(4),401
–
406.
Barnston, A . G .,1988.Theeffectofweatheronmood,productivity,andfrequencyof
emotionalcrisisinatemperatecontinentalclimate.Int.J.Biometeorol.32(2),
134
–
143.
Basu, R .,etal.,2018.Examiningtheassociationbetweenapparenttemperatureand
mentalhealth-relatedemergencyroomvisitsinCalifornia.Am.J.Epidemiol.187(4),
726
–
735.
Beecher, M . E .,etal.,2016.Sunshineonmyshoulders:Weather,pollution,andemotional
distress.J.Affect.Disord.205,234
–
238.
Berry, H . L .,Bowen, K .,Kjellstrom, T .,2010.Climatechangeandmentalhealth:acausal
pathwaysframework.Int.J.PublicHealth55(2),123
–
132.
Bongioanni, P .,etal.,2021.Climatechangeandneurodegenerativediseases.Environ.
Res.201,111511.
Braga, A . L .,Zanobetti, A .,Schwartz, J .,2002.Theeffectofweatheronrespiratoryand
cardiovasculardeathsin12UScities.Environ.HealthPerspect.110(9),859
–
863.
Brown, R . D .,Gillespie, T . J .,1986.Estimatingoutdoorthermalcomfortusingacylindrical
radiationthermometerandanenergybudgetmodel.Int.J.Biometeorol.30(1),
43
–
52.
Budd, G . M .,2008.Wet-bulbglobetemperature(WBGT)
—
itshistoryanditslimitations.J.
Sci.Med.Sport11(1),20
–
32.
Bulbena, A .,etal.,2005.Panicanxiety,undertheweather?Int.J.Biometeorol.49(4),
238
–
243.
Bulbena, A .,etal.,2009.Impactofthesummer2003heatwaveontheactivityoftwo
psychiatricemergencydepartments.ActasEspanolasDePsiquiatria37(3),158
–
165.
Bullock, B .,Murray, G .,Meyer, D .,2017.Highsandlows,upsanddowns:meteorology
andmoodinbipolardisorder.PLoSOne12(3),e0173431.
Bundo, M .,etal.,2021.AmbienttemperatureandmentalhealthhospitalizationsinBern,
Switzerland:A45-yeartime-seriesstudy.PLoSOne16(10),e0258302.
Burdett, A .,Davillas, A .,Etheridge, B .,2021.Weather,mentalhealth,andmobilityduring
thefirstwaveoftheCOVID-19pandemic.HealthEcon.30(9),2296
–
2306.
Carney, P .,Fitzgerald, C .,Monaghan, C .,1988.Influenceofclimateontheprevalenceof
mania.Br.J.Psychiatry152(6),820
–
823.
Chan, E . Y .,etal.,2018.Associationbetweenambienttemperaturesandmentaldisorder
hospitalizationsinasubtropicalcity:atime-seriesstudyofHongKongspecial
administrativeregion.Int.J.Environ.Res.PublicHealth15(4),754.
Charlson, F .,etal.,2021.Climatechangeandmentalhealth:ascopingreview.Int.J.
Environ.Res.PublicHealth18(9),4486.
Chen, J .,Hoek, G .,2020.Long-termexposuretoPMandall-causeandcause-specific
mortality:asystematicreviewandmeta-analysis.Environ.Int.143,105974.
31
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Christensen, E . M .,etal.,2008.Climaticfactorsandbipolaraffectivedisorder.NordicJ.
Psychiatry62(1),55
–
58.
Cornali, C .,etal.,2004.Effectofhighclimatetemperatureonthebehavioraland
psychologicalsymptomsofdementia.J.Am.Med.Dir.Assoc.5(3),161
–
166.
Crane, J . D .,etal.,2015.Inhibitingperipheralserotoninsynthesisreducesobesityand
metabolicdysfunctionbypromotingbrownadiposetissuethermogenesis.Nat.Med.
21(2),166
–
172.
Cruz, J .,etal.,2020.Effectofextremeweathereventsonmentalhealth:anarrative
synthesisandmeta-analysisfortheUK.Int.J.Environ.Res.PublicHealth17(22),
8581.
Denissen, J . J .,etal.,2008.Theeffectsofweatherondailymood:amultilevelapproach.
Emotion8(5),662
–
667.
Ding, N .,Berry, H . L .,Bennett, C . M .,2016.Theimportanceofhumidityintherelationship
betweenheatandpopulationmentalhealth:evidencefromAustralia.PLoSOne11
(10),15.
DSM-IV-TR,2000.DiagnosticandStatisticalManualofMentalDisorders.American
PsychiatricAssociation.
Eun-hye, Y .,etal.,2021.Effectofextremetemperaturesondailyemergencyroomvisits
formentaldisorders.Environ.Sci.Pollut.Res.28(29),39243
–
39256.
Gál, C . V .,Kántor, N .,2020.Modelingmeanradianttemperatureinoutdoorspaces,A
comparativenumericalsimulationandvalidationstudy.UrbanClim.32,100571.
Gasparrini, A .,2011.Distributedlaglinearandnon-linearmodelsinR:thepackagedlnm.
J.Stat.Softw.43(8),1.
Gasparrini, A .,Armstrong, B .,2010.Timeseriesanalysisonthehealtheffectsof
temperature:advancementsandlimitations.Environ.Res.110(6),633
–
638.
Gasparrini, A .,Armstrong, B .,2013.Reducingandmeta-analysingestimatesfrom
distributedlagnon-linearmodels.BMCMed.Res.Methodol.13(1),1
–
10.
Gasparrini, A .,etal.,2012.Theeffectofhightemperaturesoncause-specificmortalityin
EnglandandWales.Occup.Environ.Med.69(1),56
–
61.
George, A . D .,Thompson,III, F . R .,Faaborg, J .,2015.UsingLiDARandremote
microclimateloggerstodownscalenear-surfaceairtemperaturesforsite-levelstudies.
RemoteSens.Lett.6(12),924
–
932.
Goldstein, K . M .,1972.Weather,mood,andinternal-externalcontrol.Percept.Mot.Skills
35(3),786.
Gong, J .,Part, C .,Hajat, S .,2022.Currentandfutureburdensofheat-relateddementia
hospitaladmissionsinEngland.Environ.Int.159,7.
Gu, S . H .,etal.,2019.Exposure-lag-responseassociationbetweensunlightand
schizophreniainNingbo,China.Environ.Pollut.247,285
–
292.
Guo, Y .,etal.,2018.Quantifyingexcessdeathsrelatedtoheatwavesunderclimatechange
scenarios:Amulticountrytimeseriesmodellingstudy.PLoSMed.15(7).
Guyatt, G . H .,etal.,2008.Whatis
“
qualityofevidence
”
andwhyisitimportantto
clinicians?Bmj336(7651),995
–
998.
Hansen, A .,etal.,2008.Theeffectofheatwavesonmentalhealthinatemperate
Australiancity.Environ.HealthPerspect.116(10),1369
–
1375.
Harrer, M .,etal.,2021.DoingMeta-analysisWithR:AHands-onGuide.CRCPress.
Hartig, T .,Catalano, R .,Ong, M .,2007.Coldsummerweather,constrainedrestoration,
andtheuseofantidepressantsinSweden.J.Environ.Psychol.27(2),107
–
116.
Haskell, E .,etal.,1981.Metabolismandthermoregulationduringstagesofsleepin
humansexposedtoheatandcold.J.Appl.Physiol.51(4),948
–
954.
Hayes, K .,etal.,2018.Climatechangeandmentalhealth:Risks,impactsandpriority
actions.Int.J.Ment.Heal.Syst.12(1),1
–
12.
Henriquez-Sanchez, P .,etal.,2014.Geographicalandclimaticfactorsanddepressionrisk
intheSUNproject.Eur.J.Pub.Health24(4),626
–
631.
Higgins, J . P .,Thompson, S . G .,2002.Quantifyingheterogeneityinameta-analysis.Stat.
Med.21(11),1539
–
1558.
Ho, H . C .,Wong, M . S .,2019.Urbanenvironmentalinfluencesonthe
temperature
–
mortalityrelationshipassociatedmentaldisordersand
cardiorespiratorydiseasesduringnormalsummerdaysinasubtropicalcity.Environ.
Sci.Pollut.Res.26(23),24272
–
24285.
Höppe, P .,1999a.Thephysiologicalequivalenttemperature
–
auniversalindexforthe
biometeorologicalassessmentofthethermalenvironment.Int.J.Biometeorol.43,
71
–
75.
Höppe, P .,1999b.Thephysiologicalequivalenttemperature
–
auniversalindexforthe
biometeorologicalassessmentofthethermalenvironment.Int.J.Biometeorol.43(2),
71
–
75.
Howarth, E .,Hoffman, M . S .,1984.Amultidimensionalapproachtotherelationship
betweenmoodandweather.Br.J.Psychol.75(FEB),15
–
23.
Huibers, M . J . H .,etal.,2010.Doestheweathermakeussad?Meteorologicaldeterminants
ofmoodanddepressioninthegeneralpopulation.PsychiatryRes.180(2
–
3),
143
–
146.
Jahan, S .,Wraith, D .,2021.Immediateanddelayedeffectsofclimaticfactorsonhospital
admissionsforschizophreniainQueenslandAustralia:atimeseriesanalysis.Environ.
Res.197,10.
Ji, Y .,Song, J .,Shen, P .,2022.Areviewofstudiesandmodellingofsolarradiationon
humanthermalcomfortinoutdoorenvironment.Build.Environ.214,108891.
Jia, P .,etal.,2020.Spatiallifecourseepidemiologyreportingstandards(ISLE-ReSt)
statement.HealthPlace61,102243.
Keller, M . C .,etal.,2005.Awarmheartandaclearhead-Thecontingenteffectsof
weatheronmoodandcognition.Psychol.Sci.16(9),724
–
731.
Khalaj, B .,etal.,2010.ThehealthimpactsofheatwavesinfiveregionsofNewSouth
Wales,Australia:acase-onlyanalysis.Int.Arch.Occup.Environ.Health83(7),
833
–
842.
Khreis, H .,etal.,2017.Exposuretotraffic-relatedairpollutionandriskofdevelopmentof
childhoodasthma:asystematicreviewandmeta-analysis.Environ.Int.100,1
–
31.
Kim, C . T .,etal.,2015.Heat-attributabledeathsbetween1992and2009inSeoul,South
Korea.PLoSOne10(2),e0118577.
Kim, S . Y .,etal.,2021.Short-andlong-termexposuretoairpollutionandlackofsunlight
areassociatedwithanincreasedriskofdepression:Anestedcase-controlstudyusing
meteorologicaldataandnationalsamplecohortdata.Sci.TotalEnviron.757,9.
Klimstra, T . A .,etal.,2011.Comerainorcomeshine:individualdifferencesinhow
weatheraffectsmood.Emotion11(6),1495
–
1499.
Kourtis, N .,Nikoletopoulou, V .,Tavernarakis, N .,2012.Smallheat-shockproteinsprotect
fromheat-stroke-associatedneurodegeneration.Nature490(7419),213
–
218.
Kwan, M .- P .,2012.Theuncertaingeographiccontextproblem.Ann.Assoc.Am.Geogr.
102(5),958
–
968.
Lee, H .- C .,Tsai, S .- Y .,Lin, H .- C .,2007.Seasonalvariationsinbipolardisorderadmissions
andtheassociationwithclimate:apopulation-basedstudy.J.Affect.Disord.97
(1
–
3),61
–
69.
Lee, H . J .,etal.,2002.Effectsofseasonandclimateonthefirstmanicepisodeofbipolar
affectivedisorderinKorea.PsychiatryRes.113(1
–
2),151
–
159.
Lee, M .,etal.,2018a.Weatherandhealthsymptoms.Int.J.Environ.Res.PublicHealth
15(8),15.
Lee, S .,etal.,2018b.Mentaldisease-relatedemergencyadmissionsattributabletohot
temperatures.Sci.TotalEnviron.616,688
–
694.
Li, B .,etal.,2018.Amodifiedmethodofevaluatingtheimpactofairhumidityonhuman
acceptableairtemperaturesinhot-humidenvironments.EnergyBuild.158,393
–
405.
Li, D .,etal.,2019.Subtypesofparkuseandself-reportedpsychologicalbenefitsamong
olderadults:Amultilevellatentclassanalysisapproach.Landsc.UrbanPlan.190,
103605.
Li, D .,etal.,2022a.Modelingtherelationshipsbetweenhistoricalredlining,urbanheat,
andheat-relatedemergencydepartmentvisits:anexaminationof11Texascities.
Environ.Plann.B.Plann.Des.49(3),933
–
952.
Li, M .,Ferreira, S .,Smith, T . A .,2020.Temperatureandself-reportedmentalhealthinthe
UnitedStates.PLoSOne15(3),e0230316.
Li, X .,etal.,2022b.Amelioratingcoldstressinahotclimate:EffectofWinterStormUri
onresidentsofsubsidizedhousingneighborhoods.Build.Environ.209,108646.
Libert, J .- P .,Bach, V .,2005.Thermoregulationandsleepinthehuman.In:The
PhysiologicNatureofSleep.ImperialCollegePress,London,UK,pp.407
–
431.
Liew, S . C .,Aung, T .,2021.Sleepdeprivationanditsassociationwithdiseases-areview.
SleepMed.77,192
–
204.
Lin, S .- H .,Lee, L .- T .,Yang, Y . K .,2014.Serotoninandmentaldisorders:aconcisereview
onmolecularneuroimagingevidence.Clin.Psychopharmacol.Neurosci.12(3),196.
Linares, C .,etal.,2017.Short-termassociationbetweenenvironmentalfactorsand
hospitaladmissionsduetodementiainMadrid.Environ.Res.152,214
–
220.
Liu, J .,etal.,2020.Cause-specificmortalityattributabletocoldandhotambient
temperaturesinHongKong:atime-seriesstudy,2006
–
2016.Sustain.CitiesSoc.57,
102131.
Liu, J .,etal.,2021.Isthereanassociationbetweenhotweatherandpoormentalhealth
outcomes?Asystematicreviewandmeta-analysis.Environ.Int.153,106533.
Liu, W .,Zhang, Y .,Deng, Q .,2016.Theeffectsofurbanmicroclimateonoutdoorthermal
sensationandneutraltemperatureinhot-summerandcold-winterclimate.Energy
Build.128,190
–
197.
Liu, X .,etal.,2019.Influenceofheatwavesondailyhospitalvisitsformentalillnessin
Jinan,China-acase-crossoverstudy.Int.J.Environ.Res.PublicHealth16(1),87.
Masson-Delmotte, V .,etal.,2021.Climatechange2021:thephysicalsciencebasis.In:
ContributionofWorkingGroupItotheSixthAssessmentReportofthe
IntergovernmentalPanelonClimateChange.p.2.
Mawson, D .,Smith, A .,1981.RelativehumidityandmanicadmissionsintheLondonarea.
Br.J.Psychiatry138,134
–
138.
McWilliams, S .,Kinsella, A .,O
’
Callaghan, E .,2013.Theeffectsofdailyweathervariables
onpsychosisadmissionstopsychiatrichospitals.Int.J.Biometeorol.57(4),497
–
508.
McWilliams, S .,Kinsella, A .,O
’
Callaghan, E .,2014.Dailyweathervariablesandaffective
disorderadmissionstopsychiatrichospitals.Int.J.Biometeorol.58(10),2045
–
2057.
Medina-Ramón, M .,etal.,2006.Extremetemperaturesandmortality:assessingeffect
modificationbypersonalcharacteristicsandspecificcauseofdeathinamulti-city
case-onlyanalysis.Environ.HealthPerspect.114(9),1331
–
1336.
Mendoza, J .,2019.Circadianinsightsintothebiologyofdepression:Symptoms,
treatmentsandanimalmodels.Behav.BrainRes.376,112186.
Middleton, J .,etal.,2021.TemperatureandplaceassociationswithInuitmentalhealthin
thecontextofclimatechange.Environ.Res.198,11.
Min, M .,etal.,2019.Effectofapparenttemperatureondailyemergencyadmissionsfor
mentalandbehavioraldisordersinYancheng,China:atime-seriesstudy.Environ.
Health18(1),1
–
12.
Moher, D .,etal.,2009.Reprint
—
preferredreportingitemsforsystematicreviewsand
meta-analyses:thePRISMAstatement.Phys.Ther.89(9),873
–
880.
Molin, J .,etal.,1996.Theinfluenceofclimateondevelopmentofwinterdepression.J.
Affect.Disord.37(2
–
3),151
–
155.
Montes, J .,Serrano, C .,Pascual-Sanchez, A .,2021.Theinfluenceofweatheronthecourse
ofbipolardisorder:asystematicreview.Eur.J.Psychiatry35(4),261
–
273.
Moola, S .,etal.,2017.JoannaBriggsInstituteReviewer
’
sManual.TheJoannaBriggs
Institute,p.5.
Mullins, J . T .,White, C .,2019.Temperatureandmentalhealth:Evidencefromthe
spectrumofmentalhealthoutcomes.J.HealthEcon.68,102240.
NationalHeart,Lung,andBloodInstitute,2014a.QualityAssessmentofCase-Control
Studies.
NationalHeart,Lung,andBloodInstitute,2014b.QualityAssessmentToolfor
ObservationalCohortandCross-sectionalStudies.
NationalWeatherService.Glossary.Availablefrom:https://w1.weather.gov/glossary/
index.php?letter=o.
Niu, Y .,etal.,2020.Short-termeffectofapparenttemperatureondailyemergencyvisits
formentalandbehavioraldisordersinBeijing,China:atime-seriesstudy.Sci.Total
Environ.733,139040.
32
CORRECTED PROOF
D. Li et al. Science of the Total Environment xxx (xxxx) 164435
Noelke, C .,etal.,2016.Increasingambienttemperaturereducesemotionalwell-being.
Environ.Res.151,124
–
129.
Obradovich, N .,etal.,2018a.Empiricalevidenceofmentalhealthrisksposedbyclimate
change.Proc.Natl.Acad.Sci.115(43),10953
–
10958.
Obradovich, N .,etal.,2018b.Empiricalevidenceofmentalhealthrisksposedbyclimate
change.Proc.Natl.Acad.Sci.U.S.A.115(43),10953
–
10958.
Obrien, L . V .,etal.,2014.Droughtasamentalhealthexposure.Environ.Res.131,
181
–
187.
Oh, S .,etal.,2020.Emergencydepartmentvisitsforpanicattacksandambient
temperature:atime-stratifiedcase-crossoveranalysis.Depress.Anxiety37(11),
1099
–
1107.
O
’
Hare, C .,etal.,2016.Seasonalandmeteorologicalassociationswithdepressive
symptomsinolderadults:Ageo-epidemiologicalstudy.J.Affect.Disord.191,
172
–
179.
OudinÅström, D .,etal.,2015.Theeffectofheatwavesonmortalityinsusceptiblegroups:
acohortstudyofamediterraneanandanorthernEuropeanCity.Environ.Health14
(1),30.
Palinkas, L . A .,Wong, M .,2020.Globalclimatechangeandmentalhealth.Curr.Opin.
Psychol.32,12
–
16.
Pan, R . B .,etal.,2019.Impactsofheatandcoldonhospitalizationsforschizophreniain
Hefei,China:anassessmentofdiseaseburden.Sci.TotalEnviron.694,8.
Peng, Z . X .,etal.,2017.Effectsofambienttemperatureondailyhospitaladmissionsfor
mentaldisordersinShanghai,China:atime-seriesanalysis.Sci.TotalEnviron.590,
281
–
286.
Persinger, M . A .,1975.Lagresponsesinmoodreportstochangesintheweathermatrix.
Int.J.Biometeorol.19(2),108
–
114.
Petkova, E . P .,etal.,2013.Projectedheat-relatedmortalityintheUSurbannortheast.Int.
J.Environ.Res.PublicHealth10(12),6734
–
6747.
Quach, T . T .,etal.,2016.Neuronalnetworksinmentaldiseasesandneuropathicpain:
beyondbrainderivedneurotrophicfactorandcollapsinresponsemediatorproteins.
WorldJ.Psychiatry6(1),18.
Radua, J .,Pertusa, A .,Cardoner, N .,2010.Climaticrelationshipswithspecificclinical
subtypesofdepression.PsychiatryRes.175(3),217
–
220.
Rataj, E .,Kunzweiler, K .,Garthus-Niegel, S .,2016.Extremeweathereventsindeveloping
countriesandrelatedinjuriesandmentalhealthdisorders-asystematicreview.BMC
PublicHealth16(1),1
–
12.
Rigby, A . S .,Vail, A .,1998.Statisticalmethodsinepidemiology.II:Acommonsense
approachtosamplesizeestimation.Disabil.Rehabil.20(11),405
–
410.
Rocklöv, J .,etal.,2014.Susceptibilitytomortalityrelatedtotemperatureandheatand
coldwavedurationinthepopulationofStockholmCounty,Sweden.Glob.Health
Action7(1),22737.
Rother, H .- A .,etal.,2021.ImpactofextremeweathereventsonSub-SaharanAfrican
childandadolescentmentalhealth:theimplicationsofasystematicreviewofsparse
researchfindings.J.Clim.Chang.Health,5,100087.
Salib, E .,Sharp, N .,1999.Doestheweatherinfluencedementiaadmissions?Int.J.
Geriatr.Psychiatry14(11),925
–
935.
Salib, E .,Sharp, N .,2002.Relativehumidityandaffectivedisorders.Int.J.Psychiatry
Clin.Pract.6(3),147
–
153.
Sarran, C .,etal.,2017.Meteorologicalanalysisofsymptomdataforpeoplewithseasonal
affectivedisorder.PsychiatryRes.257,501
–
505.
Schünemann, H . J .,etal.,2019a.Completing
‘
Summaryoffindings
’
tablesandgradingthe
certaintyoftheevidence.In:CochraneHandbookforSystematicReviewsof
Interventions.pp.375
–
402.
Schünemann, H . J .,etal.,2019b.GRADEguidelines:18.HowROBINS-Iandothertoolsto
assessriskofbiasinnonrandomizedstudiesshouldbeusedtoratethecertaintyofa
bodyofevidence.J.Clin.Epidemiol.111,105
–
114.
Schwingshackl, L .,etal.,2021.Evaluatingagreementbetweenbodiesofevidencefrom
randomisedcontrolledtrialsandcohortstudiesinnutritionresearch:meta-
epidemiologicalstudy.BMJ374.
Sharpe, I .,Davison, C . M .,2021.Climatechange,climate-relateddisastersandmental
disorderinlow-andmiddle-incomecountries:ascopingreview.BMJOpen11(10),
e051908.
Shaw, M . R .,etal.,2011.TheimpactofclimatechangeonCalifornia
’
secosystemservices.
Clim.Chang.109(1),465
–
484.
Sherbakov, T .,etal.,2018.Ambienttemperatureandaddedheatwaveeffectson
hospitalizationsinCaliforniafrom1999to2009.Environ.Res.160,83
–
90.
Shiloh, R .,etal.,2005.Effectsofclimateonadmissionratesofschizophreniapatientsto
psychiatrichospitals.Eur.Psychiatry20(1),61
–
64.
Shiue, I .,Perkins, D . R .,Bearman, N .,2016.Physicallyequivalenttemperatureandmental
andbehaviouraldisordersinGermanyin2009
–
2011.J.Ment.Health25(2),
148
–
153.
daSilva, I .,etal.,2020.Riskassessmentoftemperatureandairpollutantson
hospitalizationsformentalandbehavioraldisordersinCuritiba,Brazil.Environ.
Health19(1),1
–
11.
Simons, K . S .,etal.,2014.Effectofpreadmissionsunlightexposureonintensivecareunit-
acquireddelirium:Amulticenterstudy.J.Crit.Care29(2),283
–
286.
Son, J .,Shin, J .,2021.Bimodaleffectsofsunlightonmajordepressivedisorder.Compr.
Psychiatry108,6.
Steadman, R . G .,1984.Auniversalscaleofapparenttemperature.J.Appl.Meteorol.
Climatol.23(12),1674
–
1687.
Sterne, J . A .,etal.,2019.RoB2:arevisedtoolforassessingriskofbiasinrandomised
trials.Bmj366.
Sung, T . I .,etal.,2011.Relationshipbetweenmeandailyambienttemperaturerangeand
hospitaladmissionsforschizophrenia:Resultsfromanationalcohortofpsychiatric
inpatients.Sci.TotalEnviron.410,41
–
46.
Sung, T . I .,Chen, M . J .,Su, H . J .,2013.Apositiverelationshipbetweenambient
temperatureandbipolardisorderidentifiedusinganationalcohortofpsychiatric
inpatients.Soc.PsychiatryPsychiatr.Epidemiol.48(2),295
–
302.
Tang, C .,etal.,2021.Effectsofdifferentheatexposurepatterns(accumulatedand
transient)andschizophreniahospitalizations:atime-seriesanalysisonhourly
temperaturebasis.Environ.Sci.Pollut.Res.11.
Tapak, L .,etal.,2018.Investigatingtheeffectofclimaticparametersonmentaldisorder
admissions.Int.J.Biometeorol.62(12),2109
–
2118.
Thompson, R .,etal.,2018.Associationsbetweenhighambienttemperaturesandheat
waveswithmentalhealthoutcomes:asystematicreview.PublicHealth161,
171
–
191.
Trang, P . M .,etal.,2016a.Heatwavesandhospitaladmissionsformentaldisordersin
northernVietnam.PLoSOne11(5),e0155609.
Trang, P . M .,etal.,2016b.Seasonalityofhospitaladmissionsformentaldisordersin
Hanoi,Vietnam.Glob.HealthAct.9(1),32116.
Trenberth, K . E .,Fasullo, J . T .,Shepherd, T . G .,2015.Attributionofclimateextremeevents.
Nat.Clim.Chang.5(8),725
–
730.
Tsutsui, Y .,2013.WeatherandIndividualHappiness.WeatherClim.Soc.5(1),70
–
82.
Vaneckova, P .,Bambrick, H .,2013.Cause-SpecificHospitalAdmissionsonHotDaysin
Sydney,Australia.PLoSOne8(2),e55459.
Vergouwe, Y .,etal.,2005.Substantialeffectivesamplesizeswererequiredforexternal
validationstudiesofpredictivelogisticregressionmodels.J.Clin.Epidemiol.58(5),
475
–
483.
Vida, S .,etal.,2012.Relationshipbetweenambienttemperatureandhumidityandvisits
tomentalhealthemergencydepartmentsinQuebec.Psychiatr.Serv.63(11),
1150
–
1153.
Viechtbauer, W .,2010.Conductingmeta-analysesin{R}withthe{metafor}package.J.
Stat.Softw.36(3),1
–
48.
Voelkel, J .,etal.,2018.Assessingvulnerabilitytourbanheat:astudyofdisproportionate
heatexposureandaccesstorefugebysocio-demographicstatusinPortland,Oregon.
Int.J.Environ.Res.PublicHealth15(4),640.
Wang, S .,etal.,2018.Effectofincreasingtemperatureondailyhospitaladmissionsfor
schizophreniainHefei,China:atime-seriesanalysis.PublicHealth159,70
–
77.
Wang, X .,etal.,2014.Acuteimpactsofextremetemperatureexposureonemergency
roomadmissionsrelatedtomentalandbehaviordisordersinToronto,Canada.J.
Affect.Disord.155,154
–
161.
Weilnhammer, V .,etal.,2021.Extremeweathereventsineuropeandtheirhealth
consequences
–
Asystematicreview.Int.J.Hyg.Environ.Health233,113688.
Williams, S .,etal.,2012.HeatandhealthinAdelaide,SouthAustralia:assessmentofheat
thresholdsandtemperaturerelationships.Sci.TotalEnviron.414,126
–
133.
Wilson, L . A .,etal.,2013.TheimpactofheatonmortalityandmorbidityintheGreater
MetropolitanSydneyRegion:acasecrossoveranalysis.Environ.Health12(1),1
–
14.
WorldHealthOrganization,2002.PreventionandPromotioninMentalHealth.World
HealthOrganization.
WorldHealthOrganization,2020.RiskofBiasAssessmentInstrumentforSystematic
ReviewsInformingWHOGlobalAirQualityGuidelines.WorldHealthOrganization.
RegionalOfficeforEurope.
Wu, X .,etal.,2016.Impactofclimatechangeonhumaninfectiousdiseases:Empirical
evidenceandhumanadaptation.Environ.Int.86,14
–
23.
Xu, Y .,Wheeler, S . A .,Zuo, A .,2018.Willboys
’
mentalhealthfareworseunderahotter
climateinAustralia?Popul.Environ.40(2),158
–
181.
Xu, Z .,etal.,2019.Assessingheatwaveimpactsoncause-specificemergencydepartment
visitsinurbanandruralcommunitiesofQueensland,Australia.Environ.Res.168,
414
–
419.
Yackerson, N . S .,etal.,2011.Theinfluenceofseveralchangesinatmosphericstatesover
semi-aridareasontheincidenceofmentalhealthdisorders.Int.J.Biometeorol.55
(3),403
–
410.
Yi, W .,etal.,2019.Examiningtheassociationbetweenapparenttemperatureand
admissionsforschizophreniainHefei,China,2005
–
2014:atime-seriesanalysis.Sci.
TotalEnviron.672,1
–
6.
Yi-Fan, P .,etal.,2016.Analyzingpersonalhappinessfromglobalsurveyandweather
data:ageospatialapproach.PLoSOne11(4),e0153638.
Yoo, E . H .,etal.,2021.Associationbetweenextremetemperaturesandemergencyroom
visitsrelatedtomentaldisorders:Amulti-regiontime-seriesstudyinNewYork,USA.
Sci.TotalEnviron.792,11.
Zammit, C .,etal.,2021.Neurologicaldisordersvis-à-visclimatechange.EarlyHum.Dev.
155,105217.
Zapata, O .,2021.Happinessinthetropics:climatevariablesandsubjectivewellbeing.
Environ.Dev.Econ.22.
Zhang, S . Y .,etal.,2020.Theeffectoftemperatureoncause-specificmentaldisordersin
threesubtropicalcities:acase-crossoverstudyinChina.Environ.Int.143,7.
Zhao, Q .,Lian, Z .,Lai, D .,2021.Thermalcomfortmodelsandtheirdevelopments:A
review.EnergyBuiltEnviron.2(1),21
–
33.
33