RESP mobility and health

1

Colloque de l’ITMO Sante´ Publique (Aviesan)

De´terminants sociaux de la sante´ : les apports de la recherche en sante´ publique Paris, 26 octobre 2012

Neighborhood environments, mobility, and health: Towards a new generation of studies in environmental health research

Quartier, mobilite´ et sante´ : vers une nouvelle approche des recherches en sante´ environnementale B. Chaix

*

, J. Me´line

, S. Duncan

, L. Jardinier

, C. Perchoux

, J. Valle´e

, C. Merrien

,

N. Karusisi

, A. Lewin

, R. Brondeel

, Y. Kestens

aInserm,U707,faculte´ deme´decineSaint-Antoine,27,rueChaligny,75012Paris,France

bFaculte´ deme´decineSaint-Antoine,universite´ Pierre-et-Marie-Curie/Paris6,UMR-S707,27,rueChaligny,75012Paris,France

cHumanpotentialcentre,Aucklanduniversityoftechnology,17,AntaresPlace,MairangiBay,0632Auckland,NewZealand

dCentred’e´tudessurlesre´seaux,lestransports,l’urbanismeetlesconstructionspubliques,9,rueJuliette-Re´camier,69456Lyoncedex06,France

eDe´partementdeme´decinesocialeetpre´ventive,universite´ deMontre´al,1430,boulevardMont-Royal,Montre´al,H2V4P3Que´bec,Canada

fCNRS,UMRGe´ographie-cite´s,13,rueduFour,75006Paris,France Received31May2013;accepted31May2013

Abstract

Whilepublicpoliciesseektopromoteactivetransportation,thereisalackofinformationonthesocialandenvironmentalfactorsassociated withtheadoptionofactivetransportationmodes.Moreover,despitetheconsensusontheimportanceofidentifyingobesogenicenvironmental factors,mostpublishedstudiesonlytakeintoaccountresidentialneighborhoodsinthedefinitionofexposures.Thereareatleastthreemajor reasonsforincorporatingdailymobilityinpublichealthresearch:(i)toidentifyspecificpopulationgroups,includingsociallydisadvantaged populations,whoexperiencemobilityorspatialaccessibilitydeficits;(ii)tostudytheenvironmentaldeterminantsoftransportationhabitsand investigatethecomplexrelationshipsbetweentransportation(asasourceofphysicalactivity,pollutants,andaccidents)andphysicalactivityand health;and(iii)toimprovetheassessmentofspatialaccessibilitytoresourcesandexposuretoenvironmentalhazardsbyaccountingfordaily trajectoriesforabetterunderstandingoftheirhealtheffects.Thereisurgentneedtodevelopnovelmethodstobetterassessdailymobility.The RECORDStudyrelieson(i)anelectronicsurveyofregularmobilitytoassessthechronicexposuretoenvironmentalconditionsoverarelatively longperiod,and(ii)GlobalPositioningSystemtrackingtoevaluatepreciselyacuteenvironmentalexposuresoveramuchshorterperiod.The presentarticlearguesthatfutureresearchshouldcombinethesetwoapproaches.Gatheringscientificevidenceontherelationshipsbetweenthe environments,mobility/transportation,andhealthshouldallowpublichealthandurbanplanningdecisionmakerstobettertakeintoaccountthe individual and environmental barriers to the adoption of active transportation and to define innovative intervention strategies addressing obesogenicenvironmentstoreducedisparitiesinexcessweight.

#2013ElsevierMassonSAS.Allrightsreserved.

Keywords:Activetransport;Environmentalhealth;Exposureassessment

Re´sume´

Alorsquelespolitiquespubliquescherchenta` promouvoirlamobilite´ active,lesconnaissancessurlesfacteursenvironnementauxfavorisant l’adoption dutransport actifrestent insuffisantes.Deplus,malgre´ leconsensus sur l’importanced’identifierles facteursenvironnementaux obe´soge`nes,lese´tudesnetiennentsouventcomptequeduquartierdere´sidencepourde´finirlesexpositions.Pourtant,ilyaaumoinstroisraisonsde tenircomptedelamobilite´ quotidiennedansl’espace,etpasuniquementduquartierdere´sidence:afin(i)d’identifierdesgroupesdepopulation, notamment de´favorise´s, qui subissent un de´ficit de mobilite´ ou d’accessibilite´ spatiale aux ressources ; (ii) d’e´tudier les de´terminants

www.sciencedirect.com

Revued’E´ pide´miologieetdeSante´ Publique61S(2013)S139–S145

*Correspondingauthor.

E-mailaddress:chaix@u707.jussieu.fr(B.Chaix),julie.meline@inserm.fr(J.Me´line),scott.duncan@aut.ac.nz(S.Duncan),

laurent.jardinier@developpement-durable.gouv.fr(L.Jardinier),camille.perchoux@umontreal.ca(C.Perchoux),merrien@u707.jussieu.fr(C.Merrien), karusisi@u707.jussieu.fr(N.Karusisi),lewin@u707.jussieu.fr(A.Lewin),brondeel@u707.jussieu.fr(R.Brondeel),yan.kestens@umontreal.ca(Y.Kestens).

0398-7620/$–seefrontmatter#2013ElsevierMassonSAS.Allrightsreserved.

http://dx.doi.org/10.1016/j.respe.2013.05.017

environnementauxdutransportactifetlesrelationscomplexesentreletransport(commesourced’activite´ physique,depolluantsetd’accidents)et lasante´ ;et(iii)demieuxappre´henderl’accessibilite´ spatialeauxressourcesetlesexpositionsauxnuisancesenvironnementalesentenantcompte destrajectoiresdemobilite´ quotidienne.Ilesturgentdede´velopperdenouvellesme´thodespourmieuxtenircomptedelamobilite´ quotidienne.

L’e´tudeRECORDs’appuiesur une enqueˆtee´lectroniquedelamobilite´ re´gulie`re,afind’appre´henderl’exposition chroniqueauxconditions environnementales,ainsiquesurunsuiviparGPSafind’e´valuerpre´cise´mentlesexpositionsenvironnementalessurunepe´riodepluscourte.Le pre´sentarticlesoutientquelestravauxfutursdevraientcombinercesdeuxapproches.L’obtentiondedonne´esprobantessurlesrelationsentre environnements,mobilite´/transportetsante´ devraitpermettreauxde´cideursdesante´ publiqueetdel’ame´nagementurbaindemieuxtenircompte desbarrie`resindividuellesetenvironnementalesdansl’adoptiondemodesdetransportactif;celadevraite´galementleurpermettredede´finirdes strate´giesinnovantesvisanta` modifierlesenvironnementsobe´soge`nespourre´duirelesdisparite´sd’exce`sdepoids.

#2013ElsevierMassonSAS.Tousdroitsre´serve´s.

Motscle´s: Transportactif;Sante´ environnementale;E´ valuationdesexpositions

1. Theoreticalbackground:reviewofsocietalissuesand scientificliterature

The past 15 years have witnessed the emergence of a flourishingneighborhoodandhealthresearchfield[1].Studies havereported associations betweenawide range of environ- mentalfactorsmeasuredinlocalresidentialneighborhoodsand alargenumberofhealth-enhancingordamagingbehaviors,risk factorsofdiseases,andhealthconditions[2].However,thereis agrowingconsensusthatoneofthemoststrikinglimitationsof neighborhood and health studies is their almost systematic focusonresidentialneighborhoods[3].Thisisamajorconcern becausemostpeoplespendonlyalimitedfractionoftheirtime intheirresidentialneighborhood.

Following recent calls to incorporate daily mobility in neighborhoodandhealthresearch [4,5],thepresent articleis interestedintherelationshipsbetweensocialposition(term1), geographiclifeenvironments(term2),mobility(term3),and physical activity and metabolic health (term 4) (overall theoreticalframework reportedinFig.1).Couples orgroups ofthefourkeytermsonFig.1are successivelyconsidered.

1.1. Obesogenicenvironmentalfactorsandtheirsocial stratification

1.1.1. Environmentaldeterminantsofobesity

Recent environmental and associated behavioral changes relatedtotheagri-foodindustryandmarketing,foodprovision inrestaurants,urban sprawling,transportation habits (private car), the reduction of energy expenditure at work, and the development of sedentary leisure have created a favorable contextfortheobesityepidemic(terms2and4onFig.1).In addition to differences between countries, obesogenic envi- ronmental factors exhibit spatial variations within countries, with an unequal geographic distribution of environmental resourcesand exposures influencing obesity even on a local scale(sportfacilities,greenspaces,fastfoodrestaurants,etc.).

To propose innovative and efficient interventions, it is important to assess the extent to which such environmental factors actually influence physical activity, weight status, hypertension,lipiddisorders,andtype2diabetes.

Inasystematicliteraturereviewthatweperformed[6],only oneofthe131articlespublishedbetween1985and2009onthe relationships between geographic life environments and

metabolic risk factors was based on Frenchdata [7]. In this body of literature, the identified contextual determinants of excess weight are related both to the effects of the food environment[8,9]andtotheeffectsofthebuiltenvironmenton theadoptionofactivetransportationmodes [10].

Ninety percent of the studies examined were exclusively focusedonthe residentialenvironment andonly4%of them considered information onboth the residentialneighborhood andanothernon-residentialenvironmentsuchasthegeographic work or school environments [6].One pioneering study was abletogeocodeanumberofregulardestinationsofparticipants – where they lived, worked, shopped, sought medical care, worshipped,andspentothertime–butmainlyaccountedfor these locations to correlate neighborhood socioeconomic informationwithself-reportedhealth[11].Asanotherapproach toaccountfordailymobility,recentstudieshavestartedtorely onGPS technologiestoassesstheexposure ofindividuals to their multiple daily environments, in an attempt to explain variationsinphysicalactivitylevelsanddietarybehavior[12].

1.1.2. Socialdeterminantsofspatialaccessibilityto resources

The accumulated knowledge in the transportation field suggeststhatdisadvantagedpopulationsfacenumerousbarriers tomobility,andtherebytospatialaccesstoresources(terms1, 2,and3onFig.1).Anumberofbarrierstotransportarerelated toindividualsocioeconomiccharacteristics(e.g.difficultiesto

Fig.1. Conceptualframeworkforincorporatingdailymobilityinneighborhood andhealthresearch.

access to a driving license, shifted work schedules [13]).

Additionally,disadvantagedneighborhoodsofurbanterritories areknownfortheirinadequateprovisionofpublictransportin France.Thisisproblematicbecausetheexpertsemphasizethe importance of ensuring the access to mobility to integrate disadvantaged populations and overcome the isolation of certainneighborhoods[14].

Somestudieshaveexaminedwhetherthedistancescovered duringtripsortheoverallsurfaceoftheresultingactivityspace varyaccordingtoindividualsocialcharacteristicsorcontextual factors [15]. Preliminary findings based on the regular destinationsgeocodedforeachparticipantinthesecondwave of the RECORD Study indicate that unemployed people, participants with fixed-term and short-term contracts, and participants with financial problems had a much smaller activityspacethanmoreadvantagedones.

Ideally,studies should distinguish three setsof factors to improve our understanding of the determinants of spatial accessibilitytoresourcesindisadvantagedneighborhoods:

individual factors (related to transportation resources, mobilityneeds,space-timebudgets,incapacitiesformobility, andmobilitypreferences);

the transportation system (cost, type, maintenance of the network,andservicearrangements);

thespatialdistributionofservicesandactivitiesofinterestfor theindividual.

1.2. Socialand neighborhoodcontexts,active transportation,andhealth

1.2.1. Activetransportation,physicalactivity,and health Inacontextwherephysicalactivitylevelsremaintoolowin France,leadingexpertsrecommend,inadditiontorecreational physical activity, the promotion of active transportation (walking, cycling, use of public transport that increases walkingtimes)[16,17](terms 3and4onFig.1).

Whilecertainpublicationshavereported inconclusivefindings [18],anumberofstudiesthatreliedonindividuallongitudinal datahavedocumentedbenefitsofactivetransportationforhealth.

Forexample,aBritishstudy[19]showedthatolderadultswho wereusingpublictransportationhadalowerrisktobecomeobese overfouryears(2004–2008).Arecentquantitativereviewthat concludedthatusingpublictransportationisassociatedwithan increase in physical activity [20] is coherent with this study demonstrating benefits of the use of publictransportation for metabolic health. Conversely, in a Chinese study, the use of motorizedtransportationformorethan5yearsover1997–2006 wasrelatedtoa1.2kggreaterweightgain[21].

1.2.2. Transportationhabitsbysocialcontexts

Inadditiontotheinfluenceofmobilityonthespatialaccess to resources, a second reason for socio-epidemiologists to investigate daily mobility is that social differences in transportation habits have repercussions on the physical activity levelsof the different social groups (terms 1, and 4 onFig.1).FrenchmobilitysurveyssuchastheNationalSurvey

onTransportandTravel(ENTD2008)suggestthathighwhite collarworkers’jobstendtoconcentrateinasmallnumberof urbancentersandthathighwhitecollarworkersthereforemore frequentlyusepublictransportthanothergroupstogotowork.

Ontheopposite,blue-collarworkershavetheirworkplacesand residences more frequently located in territories with poor public transportation opportunities, leading to a particular dependenceoncarsof thissocialgroup.

2. Hypothesesinthestudyoftherelationshipsbetween theenvironments, mobility,andhealth

2.1. The needforanoverall theoreticalframework

In the theoretical framework reported on Fig. 1, daily mobilityreferstotheeverydaymobilitycircumscribedbythe daily life territory and centered daily on the residence (as opposedtolifelongresidentialmobilityandtouristmobility).

The present framework is particularly interested in the two indirectrelationshipsbetweentheenvironmentormobilityand healthrepresentedonFig.1bythetwosuccessivedottedarrows andbythetwo successivedashedarrows.

Regarding the ‘‘environment ! mobility ! health’’

relationship (two successive dotted arrows on Fig. 1), the builtenvironmentandlandusesinfluencetheadoption(ornot) of active transportation modes, which in turn influences physical activity levels, weight status, and metabolic health.

Accordingly, transportation behaviormay explain partof the effects of social conditions and related geographic life environments on physical activity levels and metabolic risk factors, interveningas amediator toinvestigate.

The‘‘mobility!environment!health’’relationship(two successivedashedarrowsonFig.1)pointstoasecondreasonof accountingfordailymobilityinthesestudies.Thissecondpath suggeststhatmobility,as avector ofexposure tothevarious geographicenvironmentsvisited,determinestheenvironmental resourcesandhazardstowhichpeopleareexposedintheirdaily trajectories. Accordingly, an important reason for measuring mobility is to improve the assessment of environmental exposures.

2.2. Examplesofobjectivesfor neighborhoodandhealth research incorporatingdailymobility

2.2.1. Social andenvironmentaldeterminantsofdaily mobility

BasedonFig.1,animportant aimistocharacterizedaily mobilityinitsdifferentcomponentsandinvestigatetheirsocial andenvironmentaldeterminants:theoverallspatialbehaviorof individuals(whichreferstothe‘‘where’’and‘‘when’’ofspatial mobility),theirtransportationhabitsanduseoftransportation modes (‘‘how’’), and transportation-related physical activity (healthconsequenceofmobility).

2.2.1.1. Spatialbehavior. Spatialbehavior(infactspace-time behavior)referstothepatternsofdailymovementinspaceover time. Spatial behavior has been relatively neglected in the

neighborhood and health literature, which has tended to overlookgeographiclifeenvironmentsotherthantheresiden- tialenvironment.

Intertwined components of spatial behavior include the spatial extent and shape of movement (e.g. overall pattern, internalstructure of the activity space),the time structure of spatialbehavior(e.g.frequencyofmovement),andthenature ofactivitiesinvolved.Apragmaticapproachforneighborhood andhealthresearchersistoinvestigate,inahypothesis-driven perspective,thefacetsofspatialbehaviorthatmattermostfor estimatingneighborhoodeffectsonhealthorprovideimportant backgroundinformationinthesestudies.

First, it is useful to quantify the extent to which spatial behavior is circumscribed by a given area container that includes the residence (the residential neighborhood, one’s municipalityof residence,oradefinitebuffercenteredonthe residence).Suchindicatorsofspatialbehavior(ontheextentof clustering of daily trips around the residence) are useful as modifiers inregression models to assess whether residential effectsonhealtharestrongerforparticipantswhosedailylifeis anchoredintheirresidentialarea.

Second, investigating spatial behavior is useful in sub- populationswitharestrictedmobility(sociallydisadvantaged groups,children,elderlypeople).Forexample,itisimportantto determinewhethercertainsociallydisadvantagedgroupshave theiractivityspacerestrictedbyconstraintratherthanbychoice toalocalenvironmentwithalowlevelofresourcesandahigh levelof stressors[22].

Third,adetailedinvestigationoftheregulardestinationsof peopleinagiventerritorymaybeusefultolistthemaintypesof places that matter (in addition to the residence and the workplace)andthatmayserveasspatialanchorsaroundwhich toassessthe spatial accessibility toresources (alist of such meaningful anchor points likely varies from one territory to another).

2.2.1.2. Transportationhabits. Beyondspatialbehavior,itis ofinterestforepidemiologicresearchtoassessthetransporta- tionhabitsofparticipants(choicesoftransportationmodesor chain of modes) rather than to only focus on the physical activity levels that result from transport (e.g., through accelerometry). It is useful if the aim is to intervene on physicalactivitylevelsthroughtransportationpolicies.

Ofimportance,transportationdataareanalyzablebothatthe individuallevelandatthetriplevel.Individual-levelanalyses, i.e., with individuals as statistical units, typically analyze outcomevariablesontransportationhabitsdefinedonthebasis oftheobservationperiod(e.g.,7days).Examplesofindicators includethe cumulatedlengthofthetrajectoryovertheperiod and the time devoted to transportation; and the number of kilometers covered with and the time spent in the different transportationmodes.

Conversely, in analyses conducted at the trip level, the databasecomprisesoneobservationpertrip(inthetransporta- tionfield,atripreferstothetravelbetweentwoactivityplaces, whileatripstageisthecomponentofatripbasedonaunique transportation mode). Trip-level data make it possible to

investigateindividualchoicesoftransportationmodesaccord- ingtoindividualcharacteristics,accordingtothecharacteristics ofthetrip(home–worktripornot,distancecovered,hourofthe day, week or weekend trip, etc.), and according to the environmentalcharacteristicsmeasuredattheoriginofthetrip, atthedestinationof thetrip,andalong thetripitself.

2.2.1.3. Transportation-relatedphysicalactivity. Toestablish links between active transportation and physical activity or metabolic health, an innovative strategy is to collect over a follow-upperiod exhaustiveinformationonthetransportation modesthatareusedwiththestartandendtimesofthetripsand contemporaneous information on physical activity levels.

Accelerometry is a valid and reliable method for deriving high-resolution estimates of physical activity intensity, frequency,andduration,inadditiontothetimespentsedentary, the number of steps taken, and (indirectly) the energy expenditurerelatedtophysical activity[23].

Basedonthesedata,anapproachtoinvestigateassociations betweentransportationmodesandphysicalactivitymaybeto aggregatetheseaccelerometryoutputs:

atthelevelofeachtransportation‘‘stage’’(correspondingto the useofauniquetransportationmode);

at the level of each trip (possibly comprising different transportation modes);

atthelevelofeachtransportationmodeovertheobservation period;

attheleveloftheentiretransportactivityovertheweekand;

over thewholeweek.

Such aggregated accelerometry data would permit analyses bothattheindividuallevelandatthetriplevel.Individual-level analyses could focus onthe physicalactivity related toeach transportation mode, on the overall transportation physical activity,ontheoverallphysicalactivity,andontheproportion oftheoverallphysicalactivitythatisrelatedtotransportation.

Analyses at the trip level would allow oneto determine the extent to which trips that include a travel stage by public transportation imply a larger number of steps walked and energyexpenditurethantripsthatincludeatravelstagebycar (e.g.,for each10min oftravel).

2.2.2. Incorporatingdailymobilityinanalysesofthe relationshipsbetweenenvironmentalfactorsandmetabolic condition

Incorporating daily mobility in the determination of environmental exposures shouldallow researchers tocapture inamorerealisticwaytheeffectsofenvironmentalexposures onhealth(e.g.,obesogenicenvironmentaleffects).Thereareat leasttwostrategiestoincorporatedailymobilityinthestudyof the effects of geographic environments on health. A first approach is to rely on tracking technologies to assess daily mobility,oftenonlyoverashortperiod(e.g.,7days).Another approachistorelyonasurveyofregularmobilitytogatherdata ondestinationsthataremeaningfuloveralongerperiod.Inthe second wave of the RECORD Study [4], participants are

surveyed with the VERITAS electronic survey application, which integrates interactive electronic maps to facilitate the geocodingofregularactivityplaces.InRECORD,inorderto comparethesealternativeorcomplementarystrategies,werely both on the VERITAS survey of regular mobility to assess

‘‘chronic’’environmentalexposuresandonGPStrackingover 7daystocapture‘‘acute’’environmentalexposures.

Previousanalysesof GPSdataconductedatthe individual level generally examined the relationship between environ- mentalexposuresassessedaroundGPSlocationsandphysical activityfromaccelerometry ordietarybehavior assessedover the same period (e.g., 7days). Despite the objectivity and precisionofGPStrackingdata,adebatableissueiswhetheritis theoretically relevant to correlate environmental exposures relatedtomobilityoversuch ashortperiodwithametabolic outcome such as obesity that is determined by chronic processes. A drawback of relying on GPS data to assess environmentaldeterminantsofchronichealthoutcomesisthat they may not be representative of mobility patterns over a longerperiod.Evenif7-daymobilitydataaremoreinformative thanmobilitydataoveroneday,theyareunabletocapturethe seasonalityofmobilityhabits;moreover,ifoneofthedaysof thefollow-upisnot representativeofregular mobilityhabits, there is a relatively high probability that some of the other 6days are non-representative as well (e.g., vacation week, episodeofsickness).

Ideally,astrategymaybetoassessdailymobilitythrough7- dayGPStrackingasecondtime,e.g.after6months.Suchan experimentwouldallowonetodeterminetheextenttowhicha second setof GPS data provides additional information and permit to construct more reliable environmental exposure variables.

3. Methodologies forincorporatingmobilityin environmentalheathresearch

As noted above, our RECORD project relies both on an electronic survey of regular mobility (so far 73,200 regular activityplaceshavebeengeocodedfor5080participants)and onGPS tracking.The present sectionfocusesonmethodolo- gicalissuesrelatedtothehardwareandsoftwareinfrastructure todeployforthecontinuousmonitoringofmobilitywithGPS.

3.1. Differentstrategiestocollectdataontripsand mobility

3.1.1. Limitationsofthe classicalmethodologiesofsurvey oftransportation

Retrospectivesurveysofthetripsofthepreviousday(s)area classical approach to investigate the transportation habits of populations.However,itiswellknownthatthismethodology leadstoanunderreportingofshorttrips.Moreover,cardrivers tendto underestimate their traveltime while usersof public transport tend to overestimate it [24]. To address these concerns, travel diaries (filled by the participants during the follow-up period) were increasingly used at the end of the 1990s [25]. However, due to the significant burden for the

participants,thisapproachoftenleadstoahighnon-response rate[26].

Another limitation of declarative data is that they do not provideinformationontheexactitinerariesfollowedbypeople.

Geomaticalgorithmsthatautomaticallydeterminetheitinerary betweentheoriginandthedestinationofthetriparesometimes usedtoaddressthisconcern.However,thisapproachdoesnot allow one to retrieve the actual itineraries followed by pedestrians andcyclists,overestimatingthe distancescovered onheavytrafficroads.Overall,declarativedataareparticularly inaccuratetoinvestigateactivetransportation.

3.1.2. Strengths andlimitationsofstudiesbasedonGPS tracking

GPS technologiesmay,tosomeextent, replace traditional methodologies of survey of mobility. First, the use of GPS technologiesallowsonetoimprovetheprecisionoftripdata:

the departureandarrivaltimesatthe differentplaces andthe timedevotedtotripsareobjectivelyrecorded.Second,studies have shown thatGPS tracking permits tocapture short trips (includingwalkingtrips)underreportedwithtraditionalsurvey methods.Third,GPStechnologiesmakeitpossibletocollect detailedinformationonitineraries.Fourth,datacollection by GPSdecreasestheburdenofworkfortheparticipants.Finally, thedatacollectedinthiswayareimmediatelyaccessibleina digital format.

However,therearealsodifficultiesassociatedwiththeuseof GPS[27].Afirstissueisrelatedtotheintrusivenessofthisdata collection approach andtothe resulting effect onthe rateof participationinstudiesandonstudyrepresentativeness.Apart from refusaltoparticipateinthe study, alack ofcompliance withtheprotocol(devicebatterynotreloadedeveryday,etc.) leads to missing data. Moreover, temporary losses of signal result inmissing portions of the trajectory [28]. Finally and mostimportantly,acriticalissueisthatGPSdatadonotdirectly indicatetheactivitiespracticedatthedifferent placesandthe transportation modes that were used, requiring methods of imputationand/or surveyof participants.

3.2. Selectedmethodological issuesrelatedtothe continuousmonitoringofmobility

3.2.1. Integratingthe PublicHealthapproachand TransportationapproachtoGPS studies

GPStrackingisusedindifferentfieldsofresearch,including publichealthandnutritionontheonehandandtransportation sciencesontheotherhand.Inpublichealthandnutrition,studies combine GPS tracking with accelerometry. However, these studiesoftenlackinformationontheactivitiespracticedatthe differentplacesandonthetransportationmodesthatwereused betweeneachtwoplaces.AfewstudiesfromthePublichealth/

Nutrition fields have collected information on transportation modes,butcollectedthisinformationonlyforasubsetofthetrips (e.g.,onlyforhome–schoolorhome–workjourneys)[29,30]or couldnotestablishanexact timecorrespondencebetweenthe trips and the accelerometry data [29]. Alternatively, in transportation sciences, researchers usually combine GPS

trackingwithaprecisemobilitysurvey,thoughoftenonlyover oneday.However,transportationresearchersoftendonotrelyon accelerometers,andthereforelackreliableinformationonthe physical activity performed during transportation, as well as complementaryinformationfortheidentificationoftransporta- tionmodes.Therefore,akeyobjectiveforfutureresearchisto integratethe Public health/Nutrition approachand the Trans- portationapproachtoGPSstudies[31].

3.2.2. Thehardwareinfrastructuretosurveymobility StudiesfollowingparticipantswithGPSandaccelerometers were constrained to rely on two separate devices. Even if participantsdonothavetoreloadthebatteryofaccelerometers, twodevices are less convenienttohandlethan auniqueone.

Anotherimprovementofsuchdeviceswouldbetoincorporatea GPRS module that permits real-time transmission of data throughtheGSMnetwork,allowingtodetectproblemsofdata collectionornonwearofthedevice,toimmediatelyprocessthe dataandpresenttheGPStracksinanapplication,andtoimprove geolocationbyGSMtriangulationintheabsenceofGPSsignal.

Apart from smartphones,the few devices available on the marketthatcombine aGPS andaSIMcardor aGPS andan accelerometerlackabatterythatpermitstrackingofactivityover thewholedayfromwakeuptobedtime.Adevicecombininga GPS,anaccelerometer,aSIMcard,andamorepowerfulbattery isunderdevelopmentattheUniversityofMontreal.

3.2.3. Automaticinterpretationoftripdata

Afterdatacollection,anautomaticinterpretationofthedata is generally conducted [32]. This data interpretation phase relieson algorithms thatattempt to cleanthe GPS data and segmentthetrajectoryintopieces(throughtheidentificationof activityplacesandtripsbetweentheseplaces)[33].Imputation algorithmssubsequentlyattempttogenerateinformationonthe activities practiced at the different places and on the transportation modes. For example, algorithms of detection ofactivitiescanrely ongeocodedinformationonthe regular activityplacesofeachparticipantandongeomaticdataonland uses,services andfacilities,andother pointsofinterest. The aimofsuch algorithms,attheendof thecollectionperiod or during the follow-up in case of real-time transmission, is to imputeas much information as possible so as to reduce the burdenofthe surveyfor theparticipants.

3.2.4. Validation,correction,and complementationoftrip datathroughanelectronicmobility survey

The automatic interpretation of GPS data only provides partial information on the transportation modes and on the activitiespracticedatthedifferentplaces.Threeoptionsexistin thetransportationliterature inrelationtothisconcern:

it is possible to rely only on the imperfect information imputed;

itispossibletosurveyonlyafractionoftheparticipantsand to use the information collected to feed imputation algorithmsapplied totheother participants(as intheOhio DepartmentofTransportationstudy inCincinnati);

it ispossible tocollectsurveyinformationonthe tripsand activities for thewholesample (as wedo inourRECORD GPS Study).

Regardingsurveymethodstocollectthesedata,anapproach isto relyon electronicsurvey applicationsthatstimulatethe recallbypresentingtotheparticipantstheirownGPStrackson amap [34]. Theseapplicationsare based ononline mapping services and allow the users to validate or correct the informationimputedbythe algorithms.

Advanced functionalities of interest in these applications includetoolstoeditandcomplementtheitineraries.Inorderto reconstructtheentiretrajectoryoverthefollow-upperiod,such edit tools allow survey technicians to add undetected stops along the path; to decompose trips into trip stages by geolocatingmissingpointsof changeof transportationmode;

and to report entirely missing portions of trajectories by geolocatingasuccessionofactivityplacesandexactitineraries betweenthem.

4. Conclusion

We believe that the incorporation of daily mobility in neighborhood and health studies and environmental health research will be one of the most striking evolutions of the fieldinthecomingyears.Thereisaclearneedtoexperiment variousstrategiestoassessdailymobilityandcomparetheir merits and weaknesses for each particular research design (e.g., electronic survey of regular mobility and short-term GPS tracking).

Foranefficientintegrationofdailymobilityinpublichealth research, health researchers will need to establish strong partnershipswithtransportation researchers(urbanplanning).

Forexample,relyingonaprecisemobilitysurveyasperformed inthetransportation fieldincombination ofadatacollection withGPS andaccelerometers isalmost anecessary develop- menttoovercomethelimitationsofcurrentGPSstudiesinthe fieldofpublic healthandnutrition.

Asconcludingconsiderations,incorporatingdailymobility instudies of environmental effectson health isa majorstep forward towards more realistic models of the interactions between people and their environments. It has important implicationsfortheidentificationofvulnerablepopulations,for investigatingthehealthbenefitsandhazardsoftransportation, andforabettercharacterizationoftheenvironmentstowhich peopleareexposed.

More broadly, the advent of a new generation of studies combining multiple sensors of geographic location, environ- mentalexposures,activityandbehavior,andhealthcondition will lead to a novel research perspective, with analyses performedataninfra-individuallevelthatpermittoinvestigate thedaily circumstancesthatinfluencebehavior.

Disclosureof interest

The authorsdeclare thattheyhave noconflicts ofinterest concerningthisarticle.

References

[1] ChaixB. Geographiclifeenvironmentsand coronaryheartdisease:a literaturereview,theoreticalcontributions,methodologicalupdates,anda researchagenda.AnnuRevPublicHealth2009;30:81–105.

[2] RivaM,GauvinL,ApparicioP,etal.Disentanglingtherelativeinfluence ofbuiltandsocioeconomicenvironmentsonwalking:thecontributionof areas homogenous along exposures of interest. Soc Sci Med 2009;

69:1296–305.

[3] ChaixB,MerloJ,EvansD,etal.Neighbourhoodsineco-epidemiologic research:delimitingpersonalexposureareas.AresponsetoRiva,Gauvin, ApparicioandBrodeur.SocSciMed2009;69:1306–10.

[4] ChaixB,KestensY,PerchouxC,etal.Aninteractivemappingtoolto assessindividualmobilitypatternsinneighborhoodstudies.AmJPrev Med2012;43:440–50.

[5] Rainham D, McDowell I, Krewski D, et al. Conceptualizing the healthscape: contributions of time geography, location technologies andspatialecologytoplaceandhealthresearch.SocSciMed2010;

70:668–76.

[6] Leal C, Chaix B. Theinfluence of geographic life environmentson cardiometabolic risk factors: a systematic review, a methodological assessmentandaresearchagenda.ObesRev2011;12:217–30.

[7] ChaixB,ChauvinP.Tobaccoandalcoholconsumption,sedentarylifestyle andoverweightnessinFrance:amultilevelanalysisofindividualandarea- leveldeterminants.EurJEpidemiol2003;18:531–8.

[8] GlanzK,SallisJF,SaelensBE,etal.Healthynutritionenvironments:

conceptsmeasures.AmJHealthPromot2005;19:330–3[ii].

[9] McKinnonRA,ReedyJ,MorrissetteMA,etal.Measuresofthefood environment:acompilationoftheliterature,1990–2007.AmJPrevMed 2009;36:S124–33.

[10] LimtanakoolN,DijstM,SchwanenT.Theinfluenceofsocioeconomic characteristics,landuseandtraveltimeconsiderationsonmodechoice for medium- and longer-distancetrips. J Transport Geogr 2006;14:

327–41.

[11] InagamiS,CohenDA,FinchBK.Non-residentialneighborhoodexpo- suressuppressneighborhoodeffectsonself-ratedhealth.SocSciMed 2007;65:1779–91.

[12] ZenkSN,SchulzAJ,MatthewsSA,etal.Activityspaceenvironmentand dietary and physical activity behaviors: a pilot study. Health Place 2011;17:1150–61.

[13] CoutardO,DupuyG,FolS.Lapauvrete´ pe´riurbaine:de´pendancelocale oude´pendanceautomobile?Espacesetsocie´te´s2002;108:155–76.

[14] PrestonJ, Raje F.Accessibility, mobilityand transport-related social exclusion.JTransportGeogr2007;15:151–60.

[15] HaldenD,JonesP,WixeyS.Measuringaccessibilityasexperiencedby differentsociallydisadvantagedgroups–Accessibilityanalysisliterature review.Workingpaper3–EPSRCFITProgramme.2005.

[16] ShephardRJ.Isactivecommutingtheanswertopopulationhealth?Sports Med2008;38:751–8.

[17] Butler GP, Orpana HM, Wiens AJ. By your own two feet: factors associated with active transportation in Canada. CanJ PublicHealth 2007;98:259–64.

[18] WannerM,GotschiT,Martin-DienerE,etal.Activetransport,physical activity,andbodyweightinadults:asystematicreview.AmJPrevMed 2012;42:493–502.

[19] WebbE,NetuveliG,MillettC.Freebuspasses,useofpublictransportand obesityamongolderpeopleinEngland.JEpidemiolCommunityHealth 2012;66:176–80.

[20] RisselC,CuracN,GreenawayM,etal.Physicalactivityassociatedwith publictransportuse–areviewandmodellingofpotentialbenefits.IntJ EnvironResPublicHealth2012;9:2454–78.

[21] QinL,StolkRP,CorpeleijnE.Motorizedtransportation,socialstatus,and adiposity:theChinaHealthandNutritionSurvey.AmJPrevMed2012;

43:1–10.

[22] Scho¨nfelderS,AxhausenKW.Activityspaces:measuresofsocialexclu- sion?TransportPolicy2003;10:273–86.

[23] WardDS,EvensonKR,VaughnA,etal.Accelerometeruseinphysical activity:bestpracticesandresearchrecommendations.MedSciSports Exerc2005;37:S582.

[24] StopherPR.Useofanactivity-baseddiarytocollecthouseholdtraveldata.

Transportation1992;19:159–76.

[25] SchlichR,AxhausenKW.Habitualtravelbehaviour:evidencefromasix- weektraveldiary.Transportation2003;30:13–36.

[26] ArentzeTA,DijstM,DugundijE,etal.Newactivitydiaryformat:design andlimitedempiricalevidence.TranspResRec2001;1768:79–88.

[27] KerrJ,DuncanS,Schipperijn J.Usingglobalpositioningsystems in healthresearch:apracticalapproachtodatacollectionandprocessing.Am JPrevMed2011;41:532–40.

[28] DuncanS,StewartTI,OliverM,etal.Portableglobalpositioningsystem receivers:staticvalidityandenvironmentalconditions.AmJPrevMed 2013;44:e19–29.

[29] SouthwardEF,PageAS,WheelerBW,etal.Contributionoftheschool journeytodailyphysicalactivityinchildrenaged11-12years.AmJPrev Med2012;43:201–4.

[30] Oliver M, Badland H, Mavoa S, et al. Combining GPS. GIS, and accelerometry:methodologicalissuesintheassessmentoflocationand intensityoftravelbehaviors.JPhysActHealth2010;7:102–8.

[31] ChaixB,MelineJ,DuncanS,etal.GPStrackinginneighborhoodand healthstudies:astepforwardforenvironmentalexposureassessment,a stepbackwardforcausalinference?HealthPlace2013;21C:46–51.

[32] BohteW,MattK.Derivingandvalidatingtrippurposesandtravelmodes formulti-dayGPS-basedtravelsurveys:alarge-scaleapplicationinthe Netherlands.TranspResPartC2009;17:285–97.

[33] ThierryB,ChaixB,KestensY.DetectingactivitylocationsfromrawGPS data:anovelkernel-basedalgorithm.IntJHealthGeogr2013;12:14.

[34] StopherPR,CollinsA.ConductingaGPSpromptedrecallsurveyoverthe internet.Paperpresentedatthe84thAnnualMeetingoftheTransportation ResearchBoard.Washington,DC,2005.