<%BANNER%>

UFIR



Annual economic impacts of seasonal influenza on US counties: Spatial heterogeneity and patterns
http://www.biomedcentral.com/ ( Publisher's URL )
CITATION PDF VIEWER
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/IR00001230/00001
 Material Information
Title: Annual economic impacts of seasonal influenza on US counties: Spatial heterogeneity and patterns
Series Title: Mao et al.: Annual economic impacts of seasonal influenza on US counties: Spatial heterogeneity and patterns. International Journal of Health Geographics 201211:16.
Physical Description: Journal Article
Creator: Mao, Liang
Yang, Yang
Qiu, Youliang
Yang, Yan
Publisher: BioMed Central
Publication Date: May 17 2012
 Notes
Abstract: Economic impacts of seasonal influenza vary across US counties, but little estimation has been conducted at the county level. This research computed annual economic costs of seasonal influenza for 3143 US counties based on Census 2010, identified inherent spatial patterns, and investigated cost-benefits of vaccination strategies. The computing model modified existing methods for national level estimation, and further emphasized spatial variations between counties, in terms of population size, age structure, influenza activity, and income level. Upon such a model, four vaccination strategies that prioritize different types of counties were simulated and their net returns were examined. The results indicate that the annual economic costs of influenza varied from $13.9 thousand to $957.5 million across US counties, with a median of $2.47 million. Prioritizing vaccines to counties with high influenza attack rates produces the lowest influenza cases and highest net returns. This research fills the current knowledge gap by downscaling the estimation to a county level, and adds spatial variability into studies of influenza economics and interventions. Compared to the national estimates, the presented statistics and maps will offer detailed guidance for local health agencies to fight against influenza.
Acquisition: Collected for University of Florida's Institutional Repository by the UFIR Self-Submittal tool. Submitted by Liang Mao.
Publication Status: Published
 Record Information
Source Institution: University of Florida Institutional Repository
Holding Location: University of Florida
Rights Management: All rights reserved by the submitter.
System ID: IR00001230:00001

Downloads

This item is only available as the following downloads:

Annual influenza costs by county ( PDF )


Full Text

PAGE 1

RESEARCHOpenAccessAnnualeconomicimpactsofseasonalinfluenza onUScounties:Spatialheterogeneityand patternsLiangMao1*,YangYang1,YouliangQiu1andYanYang2AbstractEconomicimpactsofseasonalinfluenzavaryacrossUScounties,butlittleestimationhasbeenconductedatthe countylevel.Thisresearchcomputedannualeconomiccostsofseasonalinfluenzafor3143UScountiesbasedon Census2010,identifiedinherentspatialpatterns,andinvestigatedcost-benefitsofvaccinationstrategies.The computingmodelmodifiedexistingmethodsfornationallevelestimation,andfurtheremphasizedspatial variationsbetweencounties,intermsofpopulationsize,agestructure,influenzaactivity,andincomelevel.Upon suchamodel,fourvaccinationstrategiesthatprioritizedifferenttypesofcountiesweresimulatedandtheirnet returnswereexamined.Theresultsindicatethattheannualeconomiccostsofinfluenzavariedfrom$13.9 thousandto$957.5millionacrossUScounties,withamedianof$2.47million.Prioritizingvaccinestocountieswith highinfluenzaattackratesproducesthelowestinfluenzacasesandhighestnetreturns.Thisresearchfillsthe currentknowledgegapbydownscalingtheestimationtoacountylevel,andaddsspatialvariabilityintostudiesof influenzaeconomicsandinterventions.Comparedtothenationalestimates,thepresentedstatisticsandmapswill offerdetailedguidanceforlocalhealthagenciestofightagainstinfluenza. Keywords: Influenza,Economiccosts,USCounties,Vaccination,SpatialheterogeneityIntroductionEveryyearintheUS,influenzavirusesposeremarkable impactsonsocio-economy,suchascostsofmedicalcare, lossofproductivity,anddeaths[1].Sinceeconomicconsiderationsareessentialforinfluenzacontrol,decision makersoftenneedtoexaminefollowingquestionsfor healthinterventions.Howmuchwillaninfluenzaseason costtheUS?Whichstatesorcountiesbearhighcosts? Wheretodistributevaccinestoachievethemaximum returns?Todate,onlyasmallnumberofstudieshaveestimatedtheeconomicimpactsofinfluenzaintheUS.The OfficeofTechnologyAssessmentreportedthattheinfluenzaaccountsfor$1~3billionperyearinmedicalcosts [2].Meltzer,etal.arguedthattheannualeconomicburdenofpandemicinfluenzacouldrangefrom$71.3~166.5 billion[3].ThelatestestimationbyMolinarietal.indicatedthattheshort-termcostsandlong-termburdenof seasonalinfluenzacanbeamountedto$26.8~$87.1 billionayear[4].Thesestudieshaveestablishedsystematic methodstoanalyzeinfluenzaeconomicsandofferedvaluableguidanceforinterventions. Previousstudies,however,havefocusedonthenationallevelestimation,whilefewhavedrilleddowntoacounty levelandtakenintoaccountspatialheterogeneitybetween counties.Manyfactorswereassumedtobehomogenous acrosscountiesbutinfactvaryremarkably,suchastheinfluenzaactivity,populationsize,agestructure,income level,andsoon.CurrentestimatesfortheentireUSfailto differentiateinfluenzaimpactsbetweencounties,andthus offerlittleinformationforstate/county-levelhealthplanning.Forinstance,thenationalestimatescannotinform thedesignofcounty-basedvaccinationstrategies,i.e., where(orwhichcounties)shouldreceivevaccinesfirstfor abestcost-effectiveness.Inaddition,thelackofcountylevelknowledgemaycloudtheidentificationofcontributingfactorstoinfluenzacosts,duetothemodifiableareal unitproblem(MAUP).Thatis,differentlevelsofaggregation,suchasthecounty-,state-,andnationallevels,may producevariationinstatisticalassociations[5].Although *Correspondence: liangmao@ufl.edu1DepartmentofGeography,UniversityofFlorida,Gainesville,FL32611,USA Fulllistofauthorinformationisavailableattheendofthearticle INTERNATIONAL JOURNAL OF HEALTH GEOGRAPHICS 2012Maoetal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginalworkisproperlycited.Mao etal.InternationalJournalofHealthGeographics 2012, 11 :16 http://www.ij-healthgeographics.com/content/11/1/16

PAGE 2

theCentersforDiseaseControlandPrevention(CDC) haveofferedaFluAidsoftwaretohelpestimatelocaleconomicimpacts[6],thistoolstillemploysnationwide homogeneousparameters,andcannotcharacterizeany inter-countyvariationsexceptfordemographics. Thisarticleaimstoestimateannualeconomicimpactof seasonalinfluenzafor3143UScountiesbasedonUSCensus2010,characterizetheinter-countyvariationofsuch impact,andinvestigatecost-effectivenessofvaccination strategies.Theestimationmodifiedexistingmethodsinthe literature,andfurtheremphasizedspatialvariationsamong counties,intermsofpopulati onsize,agestructure,influenzaactivity,andincomelevel.Spatialandstatisticalanalyseswereconductedtoiden tifyspatialpatternsof influenzaimpactsacrosscounties.Futhermore,four county-basedvaccinationstr ategiesweresimulatedand theircost-effectivenesswascomparedtoidentifythe optimal.MaterialsandmethodsForaspecificcounty,theeconomicimpactsofinfluenza arethesumofmonetarycoststoeachinfluenzacasein thiscounty(Equation1).Totaleconomicimpactsi Xn jMonterycostsi ; j Pn jDirectCostsi ; j IndirectCostsi ; j Pn jMedicalCostsi ; j LossofProductivityi ; j 1 (where i representsacounty, n denotesallinfluenza casesincounty i ,and j indicatesaninfluenzacasein county i) Themonetarycostsofaninfluenzacasecanbefurther dividedintodirectandindirectcosts.Thedirectcostsresultfromexpensesofhealthcareresources,e.g., hospitalizationandantiviraltreatment,whiletheindirect costscomefromthelossofproductivityfromschool/work absenteeismanddeath[3,7].AccordingtoEquation1,followingthreesub-sectionsdescribethedatasetsandmethodstoestimatethecounty-levelinfluenzacases,directand indirectcosts,respectively.County-levelinfluenzacasesandhealthoutcomes CountypopulationbyagegroupThepopulationofacountyisabasisforcalculatingthe numberofinfluenzacases.Sinceinfluenzainfectionsvary byage,thecountypopulationwasdividedintofiveage groups:under5,5 – 17,18 – 49,50 – 64,and 65years.The countypopulationbyagegroupwasextractedfromtheUS Census2010SummaryFile1[8],themostrecentdemographicdataavailable.Forma ppingpurposes,thecounty populationwasfurthergeo-referencedtoitsadministrative boundaryfromtheUSCensusTopologicallyIntegrated GeographicEncodingandReferencingsystem(TIGER) Products[9].CountyinfluenzaattackratebyagegroupForeachagegroupinacounty,thenumberofinfluenza casesisaproductoftheage-grouppopulationandtheagespecificattackrate.Sincenodatahasbeenpublishedon theinfluenzaattackratesbyc ounty,threepiecesofinformationwereusedforestimation:thenationalattackrates byagegroup,thenationalInfluenzaLikeIllness(ILI)rates (ILIvisitsper100,000),andtheILIratesofmajorUScities. AsshowninEquation2,theinfluenzaattackrateincounty i atagegroup g ^ Attackratei ; g isequaltothenationalattackrateattheagegroup Nationattackrateg g adjusted byaratiobetweenthecountyILIrate ^ CountyILIi and thenationalILIrate( NationalILI) .Simiarly,thestandard devationof ^ Attackratei ; gistheadjustedstandarddevation of Nationattackrateg. ^ Attackratei ; g ^ CountyILIiNationILI NationattackrategStddev ^ Attackratei ; g ^ CountyILIiNationILI StddevNationattackrateg 8 > > > > < > > > > : 2 Thenationalattackratesbyagegroup Nationattackrateg andtheassociatedstandarddeviationswereadopted fromthesurveillancedataandestablishedliterature[4], withdetailslistedinSupplementaryfile1:TableS1.The nationalILIrate NationILI wastheaverageofweekly naitonalILIratesfrom2003to2010,publishedbythe GoogleFluTrends[10].TogainarepresentativeILI rateforseasonalinfluenza,thedataofSeason2009 – 10 waseliminatedbeforeaveraging,duetotheH1N1flu pandemic.TheFluTrendsdatacouldbeareasonable proxytoinfluenzaactivitie sbecauserecentliterature hasreporteditscapabilityofpredictinginfluenzaactivity[11,12].Thecorrelationbetweenthedatafromthe GoogleFluTrendsandCDCVirusSurveillancecan achieveupto82%[13]. ThecountyspecificILIrate ^ CountyILIi(inEquation2) wasinterpolatedusingtheILIratesofmajorcitiesnearto thecountycentroid.TheILIratesof117majorUScities werealsoaveragedfromtheweeklyILIratesfromGoogle FluTrends.SimilartothenationalILIrate,thedataofSeason2009 – 10wasremovedbeforeaveraging.Theordinary kriging,asophisticatedgeostatisticalmethod,wasemployed forinterpolation,becauseithasbeencommonlyusedfor predictingILIrates[14,15].Inthisresearch,amajoradvantageofkrigingliesinitsabilitytominimizethestandardMao etal.InternationalJournalofHealthGeographics 2012, 11 :16Page2of8 http://www.ij-healthgeographics.com/content/11/1/16

PAGE 3

errorofanestimateandquan tifythisstandarderror [16,17].Thekriginginterploationwasaccomplishedbyfour steps.First,the117majorcitiesweregeoreferencedtotheir geographiclocations.TheirILIrateswereusedtocompute semivariancesbetweencitiesatdifferentseparations,and consituteasamplesemivariog ram(squaresinFigure1a). Second,asphericalmodelwasfittedtothesamplesemivariogram(curvesinFigure1a),usingaleastsquaremethod. Thesill,nugget,andmajorrangeofthesphericalmodel wereestimatedtobe34.21km,46,692,and3405.5kmrespectively.Third,thesearchingneighborwassettothe nearest12to20citiesaroundthecentriodofeachcounty. Thefittedspehricalmodelwasusedtoassignweightsto neigbhorcities,andthekrigedcountyILIwastheweighted sumofILIratesatneighborcities.Lastly,across-validation wasconductedandgaveameanerrorof Š 0.05(closeto zero)andanaveragestandarderrorof310.1(Figure1b). XThekrigedcountyILIrates(Figure2a)andassociatedstandarderrors(Figure2b)wereusedtogetherto simulatethe ^ CountyILIi.Basedupontheestimatesof Nationattackrateg NationILI ,and ^ CountyILIi,the countyattackrates ^ Attackratei ; gwerecacluatedforeach agegroup,andthenmultipli edbytheage-grouppopulationstoobtainthenumberofinfluenzacasesbyage group.RisksandhealthoutcomesbyagegroupInfluenzacasesmayprogresstodifferentoutcomes, whichcostsdistinctly.Torefinethecostestimation,this researchclassifiedinfluenzacasesofeachagegroupinto twotypesofrisks,andthenfourhealthoutcomes.The twotypesofrisksrefertonon-highandhighrisks ofdevelopingseriouscomplications.Aninfluenzacase wasdefinedtobehighriskifoneormoremedicalconditionswereconsistentwithhig h-riskconditionsidentified bytheAdvisoryCommitteeonImmunizationPractices (ACIP),thusmorelikelytodevelopsevereoutcomes[18]. Foreachage-riskgroup,influenzacaseswerefurtherseparatedintofourhealthoutcomes ,including:self-care(not medicallyattended),outpatientvisit,hospitalization,and death.Thelikelihoodsofbecomingahigh-riskcaseandthe probabilitiesofdevelopingeachhealthoutcomewereestimatedbyMolinarietal.[4]basedontheliterature,medical records,andreports.MeanvaluesandstatisticaldistributionsoftheseparametersareshowninAdditionalfile1: TableS1byageandbyriskgroup.Finally,eachinfluenza caseinacountywasassignedoneof40categories(5age groups2typesofrisks4healthoutcomes).Eachcategorywasassociatedwithadirectcostandanindirectcost discussedbelow.DirectcostsbycountyThedirectcostscomefromth emedicalexpenditureinresponsetoinfluenza(e.g.,hospitalizations,outpatientvisits, anddrugpurchases),andvaryoverthe40age-riskoutcomecategories.Molinarietal.[4]hadestimatedthe nationalaveragemedicalcost(anddistribution)foreachof the40categoriesaccordingtoaproprietarydatabasethat containshealthinsuranceclaimsdatafrom4million insuredpersons[19].Detailsaboutthecostspercategory aregiveninAdditionalfile1: TableS2,andtheseestimates wereusedtoparameterizethecountymodelafteraninflationadjustment.SincetheworkofMolinarietal.wasdone in2003,thisresearchinflatedthemedicalcostsfrom2003 to2010(theyearofcensusdata)basedontheconsumer Figure1 ThekrigingmodeltoestimatecountyILIrates:(a)SamplesemivariogramofcityILIrates(redsquares)andthefitted sphericalmodel(bluecurves): 46692*Nugget+342120*Spherical(3405500,1599800,328.8) ;(b)ScatterplotofreportedILIratesvs. krigingpredictedILIrates. Thebluestraightlinesummarizesthetrendas: y=0.698x+465.73 (FigureswerecreatedwithESRIArcGIS10.0). Mao etal.InternationalJournalofHealthGeographics 2012, 11 :16Page3of8 http://www.ij-healthgeographics.com/content/11/1/16

PAGE 4

priceindex(CPI)ofthesetwoyears.TheCPIratiobetween year2010and2003wassetto1.185asreportedbytheBureauofLaborStatistics[20].Therefore,thisresearchestimatedtheeconomicimpactsifatypicalseasonalinfluenza hitUSin2010oryearslateron.Bysimulatingtheinflated directcostsforeachcase,thedirectcostsofinfluenzatoa countywerethesumofcostsfromallcases.IndirectcostsbycountyTheindirectcostsincludethelossofproductivitydueto work/schoolabsenteeismanddeath.Thelossofproductivityduetowork/schoolabsenteeismwascalculatedby multiplyingthelengthofdaysabsentfromworkbythe monetaryvaluelostperday[21].Thenationalaverage lengthofwork/schoolabsenteeism,alongwithitsdistribution,hadbeenpreviouslyestimatedbyMolinarietal.[4] forthe40age-risk-outcomecategories(Additionalfile1: TableS2).Thisresearchassumedthatthelengthofabsenteeisminacountyfollowsthenationwidedistribution, whilethemonetaryvaluelostperdayvariesbetween counties.Foreachcounty,themonetaryvaluelostperday (Figure2c)wasequaltotheaveragewageperjob[22] dividedbythetotalworkingdaysperyear(260days). Forinfluenzacasesendingwithdeaths,theproductivity losswasestimatedasthepresentvalueoflostearnings (PVLE),theprojectedearningsuntilretirementbasedon aperson ’ scurrentsalary[23].Thenationalaverageof PVLEhadbeenpreviouslyextrapolatedbyagegroupin theworkofMolinarietal.[4],andwereinflatedto2010 inthisresearch(Additionalfile1:TableS2).Insucha way,theindirectcostofeachinfluenzacasewasevaluated andthesumofallcasesgavetheindirectcostsofinfluenza inacounty.TotaleconomiccostsbycountyUpontheevaluationofinfluenzacasenumber,associated directandindirectcosts,thetotaleconomiccostsforeach countywerevaluedusingEquation1.Toquantifytheuncertaintiesinestimation,thisresearchemployedan individual-basedstochasticapproach.Eachindividualina countyisadiscretemodelingunitwithproperties,suchas theagegroup,attackrate,healthoutcome,directandindirectcosts.UsingMonte-Carlosimulation,theseproperties wereassignedrandomvaluesfromtheiradjustedprobability distributionsaccordingto(Additionalfile1:TableS1and S2)andkrigingestimates.Thesimulationwasrunby1,000 realizationstoestablish95%confidenceintervalsforestimatesofinterest.Averagedfrom1,000realizations,thetotal economiccostsforeachcountyweremappedinFigure3a, andtheeconomiccostspercapitawerealsocomputedby proratingthetotalcoststothecountypopulation (Figure3b).Duetothewordlimits,thecountyestimates andassociated95%confidenceintervalswerepresentedin Additionalfile2andpublishedonaninteractiveonline (a)Kriged influenza like illness (ILI) visits(b)Standard errors of Kriged ILI visits(c)Daily average wage(d)Percentage of vulnerable population Figure2 SpatialheterogeneitybetweenUScounties,intermsof(a)Krigingestimatedinfluenzalikeillness(ILI)visits(per100,000 persons),(b)StandarderrorsofILIestimation,(c)Dailyaveragewageasanindicatorofincomelevel,(d)Percentageofvulnerable population(age<5and>50years). Mao etal.InternationalJournalofHealthGeographics 2012, 11 :16Page4of8 http://www.ij-healthgeographics.com/content/11/1/16

PAGE 5

mappingsystem(http://heep.geog.ufl.edu/flucost).Tocapturethespatialclustersofeconomicimpacts,aMoran ’ s I statisticwasappliedthecountycostmap.NetreturnsofvaccinationagainstinfluenzaVaccinationiswidelysuggestedtobeamajorstrategy forreducingtheimpactsofinfluenza.Theuseofvaccine wasassumedtoavertoutcomesofinfluenzacases,and thusledtosavingsofcoststhatneedtotreattheseoutcomes.Thenetreturnfromvaccinationisanimportant economicmeasureincost-benefitanalysisofsuchintervention.FollowingthemethodofMeltzer,etal.[3],the totalnetreturnwascalculatedbyEquation3:TotalNetReturn P NetReturnage ; riskgroup XSavingsfromoutcomes avertedinpopulationage : riskgroup Vaccinatedpopoulationage ; riskgroup Costofvaccineperperson 3 Thesavingsfromavertedoutcomesinapopulation wasdeterminedbytheeffectivenessofvaccines (reportedinAdditionalfile1:TableS3).Toconsiderthe matchbetweenvaccineandinfluenzavirus,aprobability ofgoodmatch80%[24]wasincorporatedintothesimulationmodel.Thevaccineeffectivenessofagoodmatch wasassumedtobetwiceashighasthatofapoormatch. Thevaccinatedpopulationwasafunctionofvaccination coverage,i.e.,theproportionofpeoplebeinginoculated. Thecostofvaccineperpersonwasestimatedtobe$21 byMeltzeretal.in1999[3]andinflatedto$27.50in 2010.Thiscostincludesthevaccineprice,itsdistributionandadministrationfees(health-careworkertime, supplies),patienttravel,timelostfromworkandother activities;andcostofsideeffects. Byexplicitlyconsideringthecountydifferences,this researchwascapableofinvestigatingfourcounty-based vaccinationstrategies.Thefirststrategy,referredtoas the ‘ randomstrategy ’ ,randomlyvaccinatedUSpopulationtoapredefinedcoverage(inpercentage),regardless oftheagegroupandthecountytheylivein.Thesecond strategyvaccinatedthesameamountofpeople,but prioritizedthosewholiveinthecountiesofhighinfluenzaattackrates,thuscalled ‘ High-Attack-First ’ strategy. Thethirdstrategyissimilartothesecond,butprioritizingthosewholiveinthecountieswithhighproportion ofvulnerablepopulation,namedasa ‘ HighVulnerability-First ’ strategy.AccordingtotherecommendationbyCDC[25],thepopulationunder5yearsand over50yearswasdefinedasthemostvulnerablepopulation(Figure2d).Thelaststrategyaimstovaccinate peoplewholiveinthecountieswithhighincomelevel, andthuscalled ‘ High-Income-First ’ strategy.Ofthefour strategies,eachwassimulatedbasedonEquation3ata rangeofvaccinationcoveragefrom10%to90%witha 10%increment.Theaverageinfluenzacasenumberand thenetreturn($)wereestimatedforeachstrategycoveragecombinationafter1,000simulationruns (Figure4).Thefourstrategieswerecomparedtoidentify theoptimalonethatproducesthefewestinfluenzacases andhighestnetreturn.ResultsanddiscussionAtthenationallevel,theseasonalinfluenzaresultedin 25.34millioncasesayear(95%CI:24.83 – 25.86million),8.1%ofthetotalpopulationin2010(Table1). Theannualeconomiccostswereestimatedtobe $29.12billion(95%CI:$28.44 – $29.87billion),approximately0.2%ofthegrossdomesticproductofUS in2010.About65%oftheeconomiccostscomesfrom theindirectcost,i.e.,lossofproductivityduetowork absenteeismanddeath,whiletherestof35%isfrom thedirectmedicalcost. Theeconomicimpactsofinfluenzavarieddramatically acrossUScounties(Table2).KalawaoCounty,Hawaii, hadthesmallestnumberofinfluenzacases(7cases,95% (a) Economic costs by county(b) Economic cost per capita by county Figure3 SpatialheterogeneitybetweenUScounties,intermsof(a)Totaleconomiccostsofinfluenza,and(b)Economiccostper capita. Thedetailedcountyestimatesandassociated95%confidenceintervalswereavailableathttp://heep.geog.ulf.edu/flucostandAdditional file2. Mao etal.InternationalJournalofHealthGeographics 2012, 11 :16Page5of8 http://www.ij-healthgeographics.com/content/11/1/16

PAGE 6

CI:2 – 13)andthelowesteconomiccosts($13,883,95% CI:$281 – $97,037),becauseitspopulationwasonly90by 2010.Ontheotherhand,LosAngelesCounty,California, isthemostpopulouscountyinUS(9,818,605personsin 2010).Notsurprisingly,thiscountypossessedthelargest numberofcases(868,587,95%CI:610,091 – 1,118,763) andthehighesteconomiccosts($957.5million,95%CI: $664.7 – 1248.2million).Duetothewidevariabilityamong counties,themedianvaluewouldbeamorereasonable statistictoindicatethecentralitythanthemeanvalue.On average,aUScountywouldhave2009influenzacasesper yearbasedonUScensus2010.Theeconomiccostsfora countyaveraged$2.47million,andaquarterofcounties (abovethe75%percentile)mayexperienceaneconomic lossgreaterthan$6.42million. Thegeographicdistributionofeconomiccostsisof interest(Figure3a).Notethateachcolorshaderepresents differencesordersofmagnitude.Thedistributionofeconomiccostsstronglycoincideswiththepopulationdistribution,inthatalargepopulationoftenhasmore influenzacasesandrequiresmoremoneytoalleviatethe disease.Ingeneral,countieswithhighcostswereconcentratedinthecoastalareas,suchasthePacificregion (Washington,Oregon,California)andtheMiddleAtlantic region(NewYork,Pennsylvania,Connecticut,New Jersey).TheinlandareasintheMid-WestandMountain regions,suchasMontana,Idaho,Minnesota,and Oklahoma,hadrelativelylowercosts.TheMoran ’ s I indicatedthatthecountyeconomiccostsweresignificantlyclusteredoverspace(Moran ’ s I =0.28, Z score= 27.95, p -value<0.05).Twospatialclustersofhighcosts canbeeasilyidentified:onewascenteredatLosAngeles County,CaliforniaalongtheWestCoast,whiletheother encompassedtheBoston-NewYorkCitymetropolitan areasattheEastCoast.Otherscatteredmetropolitan areasalsoshowedhighlevelsofcosts,suchasMiami (Florida),Seattle(Washington),andHouston(Texas),etc. Itisalsointerestingtoexaminewhatiftheeconomic costswereproratedtoeverypersoninacounty (Figure3b).Theeconomiccostpercapitabycounty exhibitedadistinctspatialpattern.Theannualcostper capitarangedfrom$32.5(ZiebachCounty,SouthDakota) to$272.4(LlanoCounty,Texas).Onespatialclusterof highcostpercapitawastheSouthCentralregion(Texas, Oklahoma,Arkansas,andLouisiana),whereeveryresident neededtospendmorethan$150tocombatinfluenza. Theotherclustercanbefoundinthemid-Atlanticand southAtlanticregions,withapersonalcostbetween $100~150percapita.ResidentsintheMountainregion borethelightestburden(under$50percapita)exceptfor southerncountiesinArizonaandNewMexico.This spatialpatterncanbeexplainedbythefactthattheeconomiccostpercapitaisindependentofcountypopulation size.Countieswithhighcostpercapitawereassociated withhighlevelsofinfluenzaattackrates.Apersonina highattack-ratecountyismorelikelytobeinfectedand develophigh-riskcomplications,andthuscostsmorethan apersoninalowattackratecounty. Withregardingtothevaccinationstrategies,the ‘ High-Attack-First ’ strategysignificantlyoutperformsany otherstrategyduetothelowestinfluenzacasesand highestnetreturn(Figure4).Inotherwords,itwouldbe anoptimalstrategytofirstvaccinatingpeoplelivingin theWest-South-Centralregion,includingTexas,Oklahoma,Arkansas,andNewMexico(darkredregionsin Figure2a).Apossiblereasonisthatthisstrategydirectly prohibitedthetransmissionofinfluenza,whileother strategiesaddressinfluenzaindirectlythroughpopulationvulnerabilityandincomelevel.Theeffectof ‘ HighTable1Estimatedannualinfluenzaimpactsontheentire USusingpopulationandcostsof2010TotalLower95%CIUpper95%CI Numberofcases (Million) 25.3424.8325.86 Directcosts (Millions$) 10,262.9810,046.6010,482.80 Indirectcosts (Millions$) 18,853.6618,321.8119,439.80 Totaleconomiccosts (Millions$) 29,116.6528,441.2429,870.84 Figure4 Comparisonofcost-benefitsbetweenfourcounty-basedstrategies:(a)Totalinfluenzacasesand(b)Netreturns($)asa functionofvaccinationcoveragefrom10%to90%ofthepopulation. Mao etal.InternationalJournalofHealthGeographics 2012, 11 :16Page6of8 http://www.ij-healthgeographics.com/content/11/1/16

PAGE 7

Attack-First ’ vaccinationis,thus,morestraightforward thanotherstrategies.Therandom, ‘ High-VulnerabilityFirst ’ and ‘ High-Income-First ’ strategiesreducethe influenzacasestoasimilarextent,butthe ‘ HighVulnerability-First ’ strategyreturnsmoremonetarybenefits,andisthereforethesecondcost-effectivestrategy. Thisisbecausethe ‘ High-Vulnerability-First ’ strategy greatlydecreasesthepeoplewhomayotherwisedevelop severeoutcomes,suchashospitalizationsanddeath,and thuscurtailsthemajorsourceofinfluenzacosts.Inthis sense,itisalsoawisestrategytoprioritizevaccinestothe NorthWestregionandNorthMidwestregion(including Idaho,Wyoming,NorthDakota,SouthDakota,MinnesotaandWisconsin),wheretheproportionofvulnerable populationisprettyhigh(darkredregionsinFigure2c). Thereareseveralissuesthatrequirespecialattentionas aresultofthisstudy.First,a lthoughmanyparameterswere estimatedforindividualcoun ties,afewparameterswere stillassumedtobenationwidehomogeneous,suchasthe lengthofabsenteeismandthepresentvalueoflost earnings.Themajorreasonisthelackofrelevantdatafor eachofUScounties.Thesehomogenousparametersmay reducethevariabilityofcosts betweencountiesandleadto asmoothermapthanthereality.Thismayalsomislead ourunderstandingsoncountieswithextremelylowand highcosts.Second,theuseofGoogleFluTrendsdata wouldintroducebiasesintotheestimation,eventhough itscorrelationtotheviralsurveillancecanreach82%.A finerscaleinfluenzaILIdat aatthecountylevelwouldbe helpfultoimproveestimation.Theauthorshadtriedto directlyobtaincountyfluinfe ctiondata,insteadofusing Googlefludata.Unfortunat ely,amongthe51USstates, only6haveexplicitlypublishedtheircounty-levelfluattack rates.TheUSCDConlyreleasesregionalfludata,with eachregioncoveringseveralst ates.Therefore,theGoogle fludatacouldbethebestdatathatispubliclyavailableand containssufficientdetails.Third,theinfluenzaattackrates, directandindirectcostswereassumedtobehomogeneous withinacounty,regardlessofurbanandruralareas.Butin reality,theseparametersmayvarybetweenurbanandrural areasbecauseofdifferentpopulationdensity,landusepatterns,accessibilitytohealth care,etc.Itremainsunclear whetherornotthemodelingofurban – ruraldifference wouldsignificantlychangethe currentestimation,butthis questionwarrantsafuturestudy.ConclusionsThecontributionsofthisresearchhavetwofolds.First, thecurrentestimationofinfluenzaimpactsusesnationwidehomogenousparameters,whichflattenthespatial variationsamongUScounties.Thisresearchisthefirst attempttocalculateandmapthecounty-leveleconomic impactsofseasonalinfluenzaintheUS,therebyfilling thecurrentgapthatonlynationalestimatesareavailable.Theeconomiccostsofinfluenzarangefrom$13.9 thousandto$957.5millionamongUScounties,varying overdemographics,economy,andepidemics.Thereare twospatialclustersofhighcosts:onecenteredatLos AngelesCounty,CaliforniaalongtheWestCoast,while theotherembracingtheBoston-NewYorkCitymetropolitanareasontheEastCoast.Secondly,beforethisresearch,moststudiesinvestigatevaccinationstrategies onlyatanationallevel,butfewhaveconsideredthe countydifferences,i.e.,distributingvaccineresources basedoncountycharacteristics.Thisresearchhas exploredfourcounty-basedstrategiesandsuggestedthat vaccinationprioritizingcountieswithhighattackrates wouldproducethegreatestcost-benefits.Thisresearch addsacounty/spatialperspectiveintothedesignof healthinterventions,andshedsinsightonnewcosteffectivehealthpolicies. Itisarguedthatanyestimationmodelcanonlyproduce crudeapproximationstoreality.Thekeyofthisresearch isnottolookfortheabsolutenumericpredictions,butfor differencesinoutcomesbetweendifferentcountiesorbetweendifferentscenarios.Inthissense,thisresearchlaysa Table2AnnualinfluenzaimpactsonUScountiesusingpopulationandcostsof2010MinimumMaximumMeanMedianStandard deviation 25% percentile 75% percentile Numberof cases 7868,5888,061200927,9508295530 Directcosts (Millions$) 0.004327.113.260.8910.560.382.35 Indirectcosts (Millions$) 0.008630.386.001.5720.170.654.10 Totalcosts (Millions$) 0.014957.499.262.4730.721.026.42 Costs/Capita($) 32.51272.3697.5089.2732.7775.84113.16*Statisticsand95%confidenceintervalsbycountyareavailableat: http://heep.geog.ufl.edu/flucost/.Mao etal.InternationalJournalofHealthGeographics 2012, 11 :16Page7of8 http://www.ij-healthgeographics.com/content/11/1/16

PAGE 8

foundationforcounty-levelstudyofinfluenzaeconomics andinterventions,andcanbeeasilyexpandedtootherinfectiousdiseases.Thenumericalestimatesandmapspresentedherenotonlyinformgeneralhealthplanningfor theentireUS,butalsoofferdetailedguidanceforstateor countylevelinterventionstofightfutureinfluenza outbreaks.AdditionalfilesAdditionalfile1: Supplementaryfile1. AnnualEconomicImpactsof SeasonalInfluenzaandVaccinationonUSCounties:SpatialHeterogeneity andPatterns[3,4]. Additionalfile2: Supplementaryfile2. AnnualEconomicImpactsof SeasonalInfluenzaandVaccinationonUSCounties:SpatialHeterogeneity andPatterns. Competinginterests Theauthorsdeclarethattheyhavenocompetinginterests. Authors ’ contributions LMdesignedthework,performedallcodingandsimulation,anddraftedthe manuscript.YYcarriedoutpartofdatacollectionandanalyses.YLQandYY establishedtheGISwebsite.Allauthorsreadandapprovedthefinal manuscript. Authors ’ information LMandYLQareAssistantprofessorsofGeographyintheUniversityof Florida.YYisaPhDcandiateofGeographyintheUniversityofFlorida.The otherYYisaPhDcandiateofGeographyintheUniversityatBuffalo,State UniversityofNewYork. Acknowledgements Theauthorsarethankfulforthevaluablecommentsfromtheeditorandtwo reviewers.PublicationofthisarticlewaspartiallyfundedbytheUniversityof FloridaOpen-AccessPublishingFund. Authordetails1DepartmentofGeography,UniversityofFlorida,Gainesville,FL32611,USA.2DepartmentofGeography,UniversityatBuffalo,StateUniversityofNew York,Amherst,NY14261,USA. Received:17January2012Accepted:19April2012 Published:17May2012 References1.SchoenbaumSC: Economicimpactofinfluenza:Theindividual's perspective. AmJMed 1987, 82 (6):26 – 30. 2.USCongress: OfficeofTechnologyAssessment:Cost-effectivnessofInfluenza Vaccination .Washington,DC:GPO;1981. 3.MeltzerMI,CoxNJ,FukudaK: Theeconomicimpactofpandemic influenzaintheUnitedStates:prioritiesforintervention. EmergInfectDis 1999, 5 (5):659 – 671. 4.MolinariNAM,Ortega-SanchezIR,MessonnierML,ThompsonWW,WortleyPM, WeintraubE,BridgesCB: TheannualimpactofseasonalinfluenzaintheUS: measuringdiseaseburdenandcosts. Vaccine 2007, 25 (27):5086 – 5096. 5.FotheringhamAS,WongDWS: Themodifiablearealunitproblemin multivariatestatisticalanalysis. EnvironPlannA 1991, 23 (7):1025 – 1044. 6.CentersforDiseaseControlandPrevention: FluAidsoftware. In Centerfor DiseaeControlandPrevention ;2000.http://www.cdc.gov/flu/tools/fluaid/. 7.SzucsT: Thesocio-economicburdenofinfluenza. JAntimicrobChemother 1999, 44 (suppl2):11 – 15. 8. USCensus:Census2010SummaryFile1(SF1)100-PercentData. US CensusBureau;2010. 9. USCensus:TIGERW,TIGER/LineWandTIGERW-RelatedProducts. US CensusBureau;2010. 10.Google,Inc: GoogleFluTrend-UnitedStates: Google2011.http://www. google.org/flutrends/. 11.GinsbergJ,MohebbiMH,PatelRS,BrammerL,SmolinskiMS,BrilliantL: Detectinginfluenzaepidemicsusingsearchenginequerydata. Nature 2008, 457 (7232):1012 – 1014. 12.CarneiroHA,MylonakisE: Googletrends:aweb-basedtoolforreal-time surveillanceofdiseaseoutbreaks. ClinInfectDis 2009, 49 (10):1557 – 1564. 13.OrtizJ,ZhouH,ShayD,NeuzilK,GossC: DoesGoogleInfluenzaTracking CorrelateWithLaboratoryTestsPositiveForInfluenza. AmJRespirCrit CareMed 2010, 181 :A2626. 14.CarratF,ValleronAJ: Epidemiologicmappingusingthe “ kriging ” method: applicationtoaninfluenza-likeepidemicinFrance. AmJEpidemiol 1992, 135 (11):1293 – 1300. 15.SakaiT,SuzukiH,SasakiA,SaitoR,TanabeN,TaniguchiK: Geographicand temporaltrendsininfluenzalikeillness,Japan,1992 – 1999. EmergInfect Dis 2004, 10 (10):1822 – 1826. 16.OliverMA,WebsterR: Kriging:amethodofinterpolationforgeographical informationsystems. IntJGeogrInformSyst 1990, 4 (3):313 – 332. 17.IsaaksEH,SrivastavaRM: Appliedgeostatistics ,Volume2.NewYork:Oxford UniversityPress;1989. 18.FioreAE,ShayDK,HaberP,IskanderJ,UyekiT,MootreyG,BreseeJS,Cox NJ: Preventionandcontrolofinfluenza:Recommendationsofthe AdvisoryCommitteeonImmunizationPractices(ACIP). MorbidityMortality WklyRep 2007, 56 :1 – 54. 19.MarketScan: MedstatMarketScan :ThomsonReuters;2003. 20.USBureauofLaborStatistics: ConsumerPriceIndex BureauofLabor Statistics;2010. 21.AkazawaM,SindelarJL,PaltielAD: EconomicCostsofInfluenza-Related WorkAbsenteeism. ValueHealth 2003, 6 (2):107 – 115. 22.USBureauofEconomicAnalysis(BEA): LocalAreaPersonalIncome (Averagewageperjob) .U.S.DepartmentofCommerce2009.23.HaddixAC,TeutschSM,CorsoPS: Preventioneffectiveness:aguidetodecision analysisandeconomicevaluation .USA:OxfordUniversityPress2003. 24.NicholK,MargolisK,WuorenmaJ,VonSternbergT: Theefficacyandcost effectivenessofvaccinationagainstinfluenzaamongelderlypersons livinginthecommunity. NewEnglJMed 1994, 331 (12):778 – 784. 25.CentersforDiseaseControlandPrevention: KeyFactsaboutInfluenza (Flu)&FluVaccine .[http://www.cdc.gov/flu/keyfacts.htm].doi:10.1186/1476-072X-11-16 Citethisarticleas: Mao etal. : Annualeconomicimpactsofseasonal influenzaonUScounties:Spatialheterogeneityandpatterns. InternationalJournalofHealthGeographics 2012 11 :16. Submit your next manuscript to BioMed Central and take full advantage of: € Convenient online submission € Thorough peer review € No space constraints or color “gure charges € Immediate publication on acceptance € Inclusion in PubMed, CAS, Scopus and Google Scholar € Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit Mao etal.InternationalJournalofHealthGeographics 2012, 11 :16Page8of8 http://www.ij-healthgeographics.com/content/11/1/16