Date post: | 30-May-2018 |
Category: |
Documents |
Upload: | bill-hannegan |
View: | 218 times |
Download: | 0 times |
of 33
8/14/2019 SSRN-id1359506
1/33
NBER WORKING PAPER SERIES
CHANGES IN U.S. HOSPITALIZATION AND MORTALITY RATES FOLLOWING
SMOKING BANS
Kanaka D. Shetty
Thomas DeLeire
Chapin White
Jayanta Bhattacharya
Working Paper 14790
http://www.nber.org/papers/w14790
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
March 2009
We thank seminar participants at the Research in Progress Seminar at Stanford Medical School for
their insights. We thank Dr. Alan Garber, Dr. Douglas Owens, and Dr. Mayer Brezis for their helpful
comments. Finally, we thank Dr. Catherine Su and Dr. Priya Pillutla for helpful comments on earlier
versions of this manuscript. The views in this paper are those of the authors and should not be interpreted
as those of the Congressional Budget Office. Dr. Shetty was supported by a U.S. Veterans Affairs'
Fellowship in Ambulatory Care Practice and Research. Dr. Bhattacharya thanks the U.S. National
Institute on Aging for partial funding. The authors have no relationships (financial or otherwise) with
any company making products relevant to this study. The views expressed herein are those of the author(s)
and do not necessarily reflect the views of the National Bureau of Economic Research.
2009 by Kanaka D. Shetty, Thomas DeLeire, Chapin White, and Jayanta Bhattacharya. All rights
reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission
provided that full credit, including notice, is given to the source.
8/14/2019 SSRN-id1359506
2/33
Changes in U.S. Hospitalization and Mortality Rates Following Smoking Bans
Kanaka D. Shetty, Thomas DeLeire, Chapin White, and Jayanta Bhattacharya
NBER Working Paper No. 14790
March 2009
JEL No. I1,I18
ABSTRACT
U.S. state and local governments are increasingly restricting smoking in public places. This paper
analyzes nationally representative databases, including the Nationwide Inpatient Sample, to compare
short-term changes in mortality and hospitalization rates in smoking-restricted regions with control
regions. In contrast with smaller regional studies, we find that workplace bans are not associated with
statistically significant short-term declines in mortality or hospital admissions for myocardial infarction
or other diseases. An analysis simulating smaller studies using subsamples reveals that large short-term
increases in myocardial infarction incidence following a workplace ban are as common as the large
decreases reported in the published literature.
Kanaka D. Shetty
RAND Corporation
1776 Main Street
Santa Monica, CA 90401
Thomas DeLeire
La Follette School of Public Affairs
University of Wisconsin - Madison1225 Observatory Drive
Madison, WI 53706
Chapin White
Congressional Budget Office
2nd and D Streets
Washington, DC 20015
Jayanta Bhattacharya
117 Encina Commons
Center for Primary Careand Outcomes Research
Stanford University
Stanford, CA 94305-6019
and NBER
8/14/2019 SSRN-id1359506
3/33
4
1Introduction
Stateandlocalgovernmentshaveincreasinglybannedsmokinginpublicplaces
(includingworkplaces,restaurantsandbars)asameansoflimitingnonsmoker
exposureandofdiscouragingsmoking(CentersforDiseaseControlandPrevention
2007).Severalrecentstudiesinthemedicalliteratureusingasmallnumberofregions
suggestthatsmokingbansleadtoashortterm8to40%decreaseintheannual
incidenceofacutemyocardialinfarction(AMI)(Sargent,Shepardetal.2004;Bartecchi,
Alseveretal.2006;Cesaroni,Forastiereetal.2008).Despitethesefindings,itisunclear
howwelltheresultswouldtranslatetotypicalU.S.communities.Weexaminewhether
governmentalsmokingrestrictionsaffecthospitalizationandmortalityratesinalarge
sampleofU.S.communities.
WecalculatedeathandhospitalizationratesforAMIandotherdiseasesusing
MedicareProviderAnalysisandReview(MEDPAR)files,nationaldeathrecords
(otherwiseknownasthemultiplecauseofdeathfiles,hereafterMCD),and
hospitalizationdatafromtheHealthcareCostandUtilizationProjectsNationwide
InpatientSample(NIS).Wecompareratesbeforeandafterimplementationofthese
bansrelativetocommunitiesthatdidnotimplementbans.Weusethevariationin
implementationdatesacrossthecountryandfixedeffectsmodelstocontrolfor
unobservablefactorsincludingimprovingpreventionandtreatmentofcardiovascular
disease,decreasingsmokingrates,andsmokingrestrictionsenactedbyprivate
8/14/2019 SSRN-id1359506
4/33
5
businesses.Wefindthatsmokingrestrictionsareunlikelytosubstantiallyaffectshort
termmortalityandhospitalizationratesinboththeelderly,workingage,andchild
populations.Wefindsomeevidencethatsmokingbanscouldreducemortalityinthe
elderlybuttheresultsarenotstatisticallysignificant(1.4%,95%confidenceinterval:
3.0to0.2%).
Allpreviouspublishedstudiesonthehealtheffectsofsmokingbanssharea
commonmethodology:theycomparetheoutcomesinasinglecommunitythathas
passedasmokingbanwithoutcomesinasmallsetofnearbycommunitiesthathave
notpassedbans. Amajorcontributionofthispaperisthatwesimulatetheresultsfrom
allpossiblesmallscalestudiesusingsubsamplesfromthenationaldata. Wefindthat
largeshorttermincreasesinAMIincidencefollowingasmokingbanareascommonas
thelargedecreasesreportedinthepublishedliterature.
2Background
Inthissection,wediscusshowenvironmentaltobaccosmoke(alsoknownas
secondhandsmoke)isrelatedtohealthoutcomes,thehistoryandeffectsofsmoking
bansintheU.S.andinternationally,andtheimplicationofpreviousstudiesforU.S.
smokingpolicy.
8/14/2019 SSRN-id1359506
5/33
6
2.1Environmentaltobaccosmokeandhealthoutcomes
Inarecentreview,theU.S.SurgeonGeneralreportsthatnumerous
epidemiologicandlaboratorystudieshavelinkedenvironmentaltobaccosmoke(ETS)
exposuretoincreasedratesofcardiovasculardisease,respiratoryillnessandlung
cancer(GlantzandParmley1991;He,Vupputurietal.1999;BarnoyaandGlantz2005;
U.S.DepartmentofHealthandHumanServices.2006).Laboratorystudiessupportthe
notionthatsmallquantitiesofinhaledcigarettesmokecaninducesimilarbiochemical
responsesinnonsmokersasinchronicsmokers.Sucheffectscouldpredisposenon
smokerstogreatlyelevatedriskofAMIandstroke.Epidemiologicstudiestypically
comparedoutcomesnonsmokingspousesofsmokersandnonsmokers.Although
somemetaanalysesdisputedthesefindings,mostagreedthatchronicETSexposure
increasedriskofAMIby20to30%.Althoughnoamountofsecondhandsmokeislikely
tobebeneficial,theabovestudiesdonotaddressthepotentialeffectsofintermittent
exposuretosecondhandsmoke,asmightbecausedbyexposuretocigarettesmokein
publicplaces.
SecondhandsmokeinpublicplaceshasbeenmoststronglylinkedtoAMI
amongallpotentialadversehealthoutcomes(OngandGlantz2004;Sargent,Shepardet
al.2004;Bartecchi,Alseveretal.2006).Thereissomebiologicaljustificationforthisin
themedicalliterature(U.S.DepartmentofHealthandHumanServices.2006).Thefull
8/14/2019 SSRN-id1359506
6/33
7
effectsofeliminatingtobaccosmokemaytakeyearstooccurbecausesomeaspectsof
coronaryarterydisease(suchasnarrowingofthecoronaryarteries)developslowly
overtime.However,aheartattackoccursduetosuddenclotformationindiseased
arteries;inlaboratorysettings,exposuretoevensmallquantitiesoftobaccosmokecan
inducebiochemicalstatesthatpredisposetoheartattacks.Therefore,asmokingban
couldplausiblyreduceAMIincidenceandmortalityasearlyasthefirstyearafteraban
ifiteliminatesevenrelativelyminorexposure.Asaresultmanypriorstudiesexamined
AMIratesinsingleregionsinthe6to18monthsfollowingasmokingban.Inaddition
toincreasedriskofAMI,thoseexposedtosecondhandtobaccosmokemaysuffer
higherratesofasthma,chronicobstructivepulmonarydisease,infections,cancerand
otherdiseases.
2.2PublicbansonsmokinginU.S.publicplaces
AsevidenceonETSaccumulated,manyU.S.employersbeganrestricting
smokingintheworkplace;theproportionofcoveredworkersincreasedfrom25%to
70%between1986and1993(Farkas,Gilpinetal.1999;Farrelly,Evansetal.1999;
CentersforDiseaseControlandPrevention2000).Followingtheseprivaterestrictions,
severalcommunitiesinCaliforniabannedsmokinginworkplaces,restaurants,andbars
intheearly1990s.Manyotherstatesandmunicipalitiesfollowed(American
Nonsmokers RightsFoundation2007).Inadditiontotheselocalpolicies,several
8/14/2019 SSRN-id1359506
7/33
8
prominentpoliticiansadvocatedanationalpolicybanningsmokinginpublicplaces
(OnTheIssues.org2007).
Althoughthesebanshaveprovenpopular,thescientificliteraturetodatehas
onlyexaminedtheimpositionofasmokingbaninafewspecificU.S.regionsand
severalEuropeancountries.AMIratesdecreasedapproximately40%inHelena,
Montanaand27%inPueblo,Colorado(relativetosurroundingcommunities)following
theimpositionofbroadrestrictions(Sargent,Shepardetal.2004;Bartecchi,Alseveret
al.2006).AlargerstudycomparedratesofAMIandacutestrokeadmissionsinNew
YorkStatebeforeandaftercomprehensivesmokingbans,whichwerelargely
implementedinMarchandJuly2003(Juster,Loomisetal.2007).Thisstudyestimated
thatthelawsreducedAMIadmissionsby8%,althoughitdidnotcomparethese
changesinAMIandacutestrokeadmissionstochangesthatmayhaveoccurredin
nearbystatesthatdidnotimplementsmokingbansoverthisperiod.
Examiningadifferentmechanism,AdamsandCotti(2008)findincreasedratesof
vehiculardeathsfollowingtheenactmentofsmokingbans. Theyattributethisincrease
tosmokersdrivingoutoftheirnativeareatofindaplacetosmokeinpublic. Another
possibilityisthatthesebansleadsmokerstosmokemoreinvehicles,whichcouldbea
distractionwhiledriving.AdamsandCottiusenationaldataandhencemeasuresthe
averageeffectofabanacrossallcommunitiesintheU.S.;unlikethepreviouslycited
studies,theydonotmeasuretheeffectinindividualcommunities.
8/14/2019 SSRN-id1359506
8/33
9
2.3Internationalsmokingbans
TworelativelylargestudiesoftheeffectofsmokingbansonAMIincidencein
RomeandthePiedmontregionofItalyconcludedthatsmokingbansreducedAMI
incidenceby7to11%inyoungerpopulations(BaroneAdesi,Vizzinietal.2006;
Cesaroni,Forastiereetal.2008).Theauthorsdemonstratedthattheprevalenceof
smoking,cigaretteconsumptionpersmoker,andnonsmokerexposuredroppedin
Italyafterthenationalban(BaroneAdesi,Vizzinietal.2006;Gallus,Zuccaroetal.2007;
Cesaroni,Forastiereetal.2008).1
AstudyoftheScottishpublicsmokingbanfoundlargereductionsinAMIrates
aswell(Pell,Hawetal.2008).TheScottishgovernmentimplementedacomprehensive
baninMarch2006.Theauthorsmeasuredadmissionsto9hospitals(whichtreatover3
millionpeople)aswellasdeathrecordsforthegeneralpopulation.Theymeasured
actualexposuretocigarettesmokeamongnonsmokers(formersmokersandnever
smokers)andcurrentsmokers.Inpriorstudiesofthegeneralpopulation,serum
cotininelevelsdeclinedfrom0.43to0.25nanograms/dLinnonsmokersandfrom167
to103nanograms/dLincurrentsmokers.Theauthorsthennotedstatisticallysignificant
declinesinadmissionsforacutecoronarysyndromesinallgroups 14%insmokers,
19%informersmokersand21%inthosewhoneversmoked.ThedeclineintheAMI
1 The authors used cotinine levels, obtained from blood or salivary samples, to measure cigarette smoke exposure in
non-smokers reliably.
8/14/2019 SSRN-id1359506
9/33
10
ratewasnoticeablylargerthanthe4%declineinneighboringEnglandduringthesame
timeperiod(whichlackedacomprehensiveban).
2.5ImplicationsofpriorworkforU.S.smokingpolicy
Theaforementionedstudieslinkingsmokingbanstoimpressivepublichealth
gainssuggestthatwidespreadpublicsmokingbanswoulddemonstrablyimproveU.S.
publichealth.However,theinternationalexperiencealsomaynottranslatebecause
nonsmokersexposuretosecondhandsmokeandsmokingprevalenceinItalyandin
Scotlandweremuchhigherandprivatesmokingrestrictionswereweakerthaninthe
U.S.
RestrictingtheanalysistoU.S.studiesdoesnoteliminatequestionsabout
generalizability.PriorU.S.studiesweresmallinscale,havingexaminedonlyafew
regions;itispossiblethatthoseregionsarenotrepresentativeoftypicalU.S.
communities.Althoughdifferenceindifferenceanalysescancontrolforunobserved
factors,asimplepairwisecomparisonusinganatypicalpairofcommunitieswillyield
resultsthatmaynotberepresentative.Bycontrast,asimulationstudyfoundthat
extendingsmokingrestrictionsfrom70%to100%ofU.S.workplaceswouldprevent
roughly1,500myocardialinfarctionsinthefirstyear(OngandGlantz2004).Although
thismaybeaclinicallyrelevantimprovement,itrepresentsamuchsmallerreduction
8/14/2019 SSRN-id1359506
10/33
11
(
8/14/2019 SSRN-id1359506
11/33
12
similarityimpliesexposuretosecondhandsmokepresentslargeheathrisksatlow
levelsandnoadditionalhealthrisksathigherlevels,whichseemsunlikely.2
WeaddresstheseissuesbyanalyzingtheimpactofU.S.publicsmoking
restrictionsonhealthoutcomesinalarge,heterogeneousgroupofU.S.communities.By
analyzingmuchlargerpopulationsinadiversegroupofU.S.hospitalsandcounties
andbyaccountingforunderlyingseculartrendsandregionspecificcharacteristics,we
mitigatethepossibilityofselectionbias.
3.Data
Toconstructthedatasetsusedforouranalysis,wemergedataonthetimingand
locationofsmokingbanstothreelargenationwidedatasourcesonhealthoutcomes.
3.1Dataonsmokingbans
WeuseordinancedatafromtheAmericanNonsmokersRightsFoundationto
identifystates,counties,andmunicipalitiesthatimplementedrestrictionsonsmoking
between1990and2004.WeadapttheclassificationschemefromtheAmerican
NonsmokersRightsFoundationtoidentifythosebansthatrestrictsmokinginall
workplacesexceptbarsandrestaurantsasworkplacebans.Althoughnotincludedin
2Similar risk reductions were found in European and U.S. studies though baseline smokingprevalence, secondhand smoke exposure, and efficacy varied, which argues against a plateauin risk.
8/14/2019 SSRN-id1359506
12/33
13
theformerlist,Californiaadoptedanearlycompletebanonworkplacesmokingin
1995.WeclassifiedCaliforniaassmokingrestrictedstartingin1995.Wealsocreatea
datasetofbansofanysite workplaces,barsorrestaurants.Weclassifyeach3digitzip
code,city,andcountyintheU.S.bysmokingbanstatusanddateofimplementation.
3.2Datasourcesforhealthoutcomes
WeanalyzehealthoutcomesusingtheMultipleCauseofDeath(MCD)database
(19892004),Medicareclaims(19972004),andtheNationwideInpatientSurvey(NIS),
collected19932004bytheHealthcareCostandUtilizationProject,whichissponsored
bytheAgencyforHealthcareResearchandQuality(AgencyforHealthcareResearch
andQuality2006).TheMCDdatabaseidentifiestheunderlyingcauseforeachdeathin
theU.S.OursourceforMedicareclaimsaretheMedicareProviderAnalysisand
Review(MEDPAR)files,whichincludeallfeeforserviceMedicarebeneficiariesinthe
U.S.TheNISisanationallyrepresentative20%sampleofalldischargesfromU.S.
communityhospitals(whichincludesallnonfederalacutecarehospitals).Excluded
hospitalsincludeVeteransAffairshospitalsandlongtermrehabilitationhospitals.
Weidentifiedmortalityandhospitalizations duetoAMIandallcausedeaths
andhospitalizationsbecausebothmightplausiblyimproveintheshortrun.In
addition,broaddiseasemeasureslikeAMIandallcauseeventsarelesslikelymiscoded
inadministrativedata,reducingmeasurementerror(Petersen,Wrightetal.1999;
8/14/2019 SSRN-id1359506
13/33
14
TirschwellandLongstreth2002).Wealsoconsidercasesofasthmaandchronic
obstructivepulmonarydisease;whilethesearechronicdiseases,secondhandsmoke
possiblytriggersacuteexacerbations(U.S.DepartmentofHealthandHumanServices.
2006).Wealsoidentifyhipfracturehospitalizationsbecausethesecanactasanegative
controlbecausetheirincidencewouldbeunlikelytochangequicklyafterasmoking
ban(Hoidrup,Prescottetal.2000). Wedontexpecttheincidenceofhipfracturetobe
affectedbysmokingbans. Ifweweretofindanassociationbetweensmokingbansand
theincidenceofhipfracture,wewouldquestionthevalidityoftheempiricalstrategy.
Weassembleeachdatasetinasimilarfashion:wefirstclassifydeathsand
hospitaladmissionsaccordingtotheirprimarydiagnoses(AMI,asthma,etc.);wethen
sumalloutcomesineachregion(hospitalcatchmentarea,county,orzipcode);finally,
wemergeinformationonsmokingordinancesforthatregion.
Workplacesmokingbanscouldhavedifferentialeffectsbyage.Theelderlyand
childrenmaybemorevulnerabletothediseasesexacerbatedbyETS,andcouldstandto
gainmorebenefitthanatypicalworkingadult.Ontheotherhand,childrenandthe
elderlyareprimarilyexposedtoworkplaceETSascustomers,whichwouldreduce
theirbenefitfromasmokingban.Toaccountforthesedifferences,wefurtherstratify
outcomesintothreeagegroupsperregion:children(017years),workingageadults
(1864years),andtheelderly(65+years).
8/14/2019 SSRN-id1359506
14/33
15
Fromeachdataset,weexcludedeathsandhospitalizations whereweareunable
todeterminewhetherthepersonlivedinanareawhereasmokingrestrictionwas
implemented(3540%oftheNISandMCDdata).Inaddition,weexcludethefollowing
fromouranalysisofNIShospitals:transferpatients,hospitalsincludedinasingle
surveyyear,hospitalsthatmergedduring19932004,hospitalsdevotedtoacute
rehabilitation(becausefewAMIsareadmitted)andsmallhospitals(
8/14/2019 SSRN-id1359506
15/33
16
example,manyregionsexperiencedunobservedincreasesinprivatesmoking
restrictions,reductionsinsmokingprevalence,orimprovedmedicaltreatmentthat
couldhavecausedchangesinoutcomes.Tomitigatethesepotentialconfounding
factors,wecomparetrendsinregionswheresmokingbanswereimplementedtothose
incontrolregionswheresmokingrestrictionswerenotimposed.Inparticular,we
estimateregionfixedeffectsmodels. Foreachoutcome(e.g,AMIdeathor
hospitalizations), weusethefollowingregressionmodel:
(1) ititstiit SmokingBanOutcome +++=
Here itOutcome representsthenumberofdeathsorhospitaladmissionsinregion
i(1N)andtimet(1T), itOutcome isanindicatorforeachyear,and it istheerror
term.Weinclude i ,anindicatorforeachregion(county,3digitzipcode,orhospital),
tocontrolforidiosyncraticdifferencesbetweenregions. s isthecoefficientofinterest,
representingthebreakinthetimetrendinducedbyasmokingban,aftercontrollingfor
seculartrends.Inpresentingfinalresults,wepresentthemeanpercentagechangein
outcomes %)100(
swhere is = =
N
i
T
t itOutcome
TN 1 11
.Weuseblockbootstrap
clusteredattheregionalleveltocalculateallstandarderrors.Ofnote,theeffectsof
privaterestrictionsenactedpriortoagovernmentbanareincludedinprebantrends
andareexcludedfromthefinalestimate.Asaresult,thisstrategyplausiblyidentifies
theshorttermeffectsofgovernmentrestrictionsalone.Duetodatalimitations,wealter
8/14/2019 SSRN-id1359506
16/33
17
ourapproachinanalyzinghospitaladmissionsinNIShospitals;wecomparechangesin
admissionsinthefirstyearfollowingasmokingrestrictiontochangesincontrol
hospitalswithoutsmokingrestrictions.3
4.2Sensitivityanalyses.
Inourmainregressionmodel(1)wedonotincludecovariatesotherthantime
andregionspecificindicators.However,itispossiblethatfactorsmightchangeover
timedifferentiallybetweenregions.Ifso,thefixedeffectsmodelwouldyieldbiased
estimates.Wethereforetestmodelspecificationsofthefollowingform:
(2) ititstiit SmokingBanOutcome ++++= X
Thetermsandresultsarethesameasin(1)exceptfortheadditionofX,avectorof
countylevelcharacteristicstakenfromthe2005AreaResourceFile.Thevariables
includepopulationsize,numberofphysiciansandhospitalbedspercounty,household
income,andpercentofpopulationinlaborforce.Thesedatawerelinkedbycountyand
year,whereavailable,tocountiesfromtheMCDaswellashospitalsintheNISdata.
Wedonotincludethesevariables(whichcouldpotentiallyaddexplanatorypower)in
3In our analysis of the NIS data, our primary unit of analysis is the hospitalscatchment area, which includes the hospitals home city but whose full extent is unobservable.We use the number of admissions to a hospital in a particular month as a measure of theadmission rate within its catchment area. Catchment areas tend to remain stable over time, sothe approximation error is likely to be small. It is possible that we misclassified someadmissions. However, many bans were enforced in the county or state of origin, which wouldinclude the entire catchment area. In addition, we only considered patients with seriousillnesses who tend to be taken to the nearest hospital, which further reduces bias.
8/14/2019 SSRN-id1359506
17/33
18
allmodelspecificationsbecausedataforX isoftenmissingforseveralyears,which
substantiallylimitstheirsamplesizes.Weperformsensitivityanalysesbycomparing
theresults
using
model
specification
(2)
to
the
results
using
model
specification
(1)
for
thesamesetofincludedregions.
4.3Simulatingsmallsampleresults
Wecompleteouranalysisbyusingsubsamplesofthenationaldatatosimulatea
completesetofpairwisecomparisonstudiesofthesortavailableinthepublished
literature,includingBartecchietal.(2005)andSargentetal.(2004).Thesestudies
comparedatreatmentunitwhereabanwaspassedagainstacontrolunitwithnoban
onthebasisofthechangeinanoutcomevariable(heartattackadmissionrates,for
instance)inashortperiod(618months)afterthebanwaspassedinthetreatment
region.Wesimulatetherangeofsucheffectsbyfirstcalculatingthepercentchangein
admissionsineachhospitallocatedinaregionwithworkplacesmokingrestrictions
betweentheyearbeforeandtheyearfollowingaworkplacesmokingban;wealso
calculatethesamestatisticsforallcontrolhospitalsfromthesametimeperiod.Wethen
subtractthechangeinoutcomesineachcontemporaneouscontrolhospitalfromeach
interventionhospital.Theresultingdatasetconsistsoftheuniverseofpossible(19,406)
pairwisecomparisonsofonesmokingrestrictedhospitalwithonecontrolhospital.We
conductasimilaranalysisofheartattackmortality,exceptinthiscase,weuseMCD
8/14/2019 SSRN-id1359506
18/33
19
mortalitydataandourunitofanalysisisthecounty.Thefinaldatasetconsistsof23,938
pairwisecomparisonsandrepresentstheuniverseofsuchcomparisonsintheMCD
data.
5.Results
5.1Smokingrestrictionsovertime
ThepercentageofU.S.regionsthatimposedsmokingrestrictionsincreased
dramaticallybetween1988and2004;theprevalenceofsmokingrestrictionsrose
sharplyin1995and20032004andmoregraduallyinotheryears(seeFigure1).
Notably,thisnationaltrendfollowstheactionsofprivateemployers.By1993,most
workplacesisolatedcigarettesmoke,whichwoulddiminishtheimpactoflegislative
bans(Farrelly,
Evans
et
al.
1999;
Centers
for
Disease
Control
and
Prevention
2000).
5.2Mainestimates
Workplacesmokingrestrictionsareunrelatedtochangesinallcausemortalityor
mortalityduetootherAMIinallagegroups.Restrictionsonsmokingofanysortare
associatedwithreducedallcausemortalityamongtheelderly(1.4%,95%CI: 3.0to
0.2%)buttheresultisonlysignificantatthe10%level(p=0.06)(seeTable2).
WefindnostatisticallysignificantreductioninadmissionsduetoAMIamong
workingageadults(4.2%,95%CI: 10.2to1.7%,p=0.165)oramongtheelderly(2.0%,
8/14/2019 SSRN-id1359506
19/33
20
95%CI: 3.7to7.7%,p=0.48)followingtheenactmentofaworkplacesmoking
restriction(seeTable3).Wesimilarlyfindnoevidenceofreductioninadmissionsfor
otherdiseasesinanyagegroup,thoughsmokingrestrictionsofallsortsareassociated
withstatisticallyinsignificantincreasesinasthma(11.4%, 95%CI: 2.4to25.3%,p=0.11)
andtotaladmissions(3.7,95%CI: 2.1to9.5%,p=0.21)amongchildren.Amongboth
theelderlyandworkingageadults,wefindnostatisticallysignificanteffectofsmoking
bansonhipfractureadmissions,ournegativecontrol.Thissupportsthehypothesisthat
unobservedcharacteristicsofregionsdonotconfoundourresults.
5.3Sensitivityanalyses
Solelyusingregionspecificindicatorstocapturetrendspotentiallyomits
importantconfoundingvariablesbutcomprehensivedataisnotavailableforallregions
andyears.Wetestforbiasbyaddingvariablesfordemographicsandmedicalresource
availabilitytotheregressionmodel.(Weobtainthesedatafromthe2005AreaResource
File,whichcontainsinformationforallcountiesandsomeyears.)Weexaminethe
importanceoftheseomittedvariablesbycomparingresultswithandwithout
additionalvariablesforthesametimeperiods.
Table4showstheeffectofanysmokingbanontotaldeathratesinallagegroups
usingdifferentsetsofcontrolvariables.Theestimatesdonotchangedramaticallywith
inclusionofdifferentvariables,butdochangeasthestudysamplechanges. Thesample
8/14/2019 SSRN-id1359506
20/33
21
isreducedfrom24,884countyquartersincolumn(1)to7,308countyquartersin
column(4).Regressionestimatesusingthesamesamplesbutdifferentregression
modelsaresimilar. Forexample,basedonalimitedsampleof15,512countyquarters
across9yearsincolumn(3),theestimatedassociationbetweenanysmokingbanand
totaldeathrateswas+0.330%(withastandarderrorof0.794%)withoutadditional
controlsand+0.328%(withastandarderrorof0.804%)withcontrolsforhospitalbeds
percapitaandphysicianspercapita.WeshowsimilarresultsinTables5and6.These
resultssuggestthatordinarydemographicchangesareaccountedforusingthe
differenceindifferenceidentificationstrategy.
Inalloftheresultsreportedinthetables,wecalculatestandarderrorsthatallow
forclusteringatthearealevel(county,hospitalarea,orzipcode). Ifinaddition,we
weretocorrectformultiplecomparisonsusingHochbergsmethod(Hochberg1988),
ourstandarderrorsincreasefurther.
5.5Pairwisecomparisonssimulatingsmallsampleresults
Figures2and3plotallpossiblepairwisecomparisonsofchangesinAMI
incidenceafteraworkplacesmokingbantochangesinrandomlyselectedcontrol
regions.Figure2showsthatthemeanmeasuredeffectofworkplacesmokingbanson
heartattackadmissionsisclosetozero,but10%orgreaterdeclinesand10%ofgreater
increasesinAMIadmissionsarecommon.Figure3showssimilarresultsfora
8/14/2019 SSRN-id1359506
21/33
22
comparisonofAMImortalityinsmokingrestrictedcountiesfromtheyearaftera
workplacebanwithratesincountieswithoutaban. Theresultsofthissimulation
analysisshowsthatresultsfrompriorsmallsamplestudies,whichfoundverylarge
decreasedinAMIadmissionsandmortalityfollowingtheenactmentofsmokingbans,
arefeasible. However,resultswiththeoppositesignandofsimilarmagnitudearealso
feasibleandshouldbeequallycommon.
6.Conclusions
WefindnoevidencethatlegislatedU.S.smokingbanswereassociatedwith
shorttermreductionsinhospitaladmissionsforacutemyocardialinfarctionorother
diseasesintheelderly,childrenorworkingageadults.Wefindsomeevidencethat
smokingbansareassociatedwithareducedallcausemortalityrateamongtheelderly
(1.4%)butonlyatthe10%significancelevel.
Wealsoshowthatthereiswideyeartoyearvariationinmyocardialinfarction
deathandadmissionrateseveninlargeregionssuchascountiesandhospital
catchmentareas.Comparisonsofsmallsamples(whichrepresentsubsamplesofour
dataandaresimilartothesamplesusedinthepreviouspublishedliterature)might
haveledtoatypicalfindings.Itisalsopossiblethatcomparisonsshowingincreasesin
cardiovasculareventsafterasmokingbanwerenotsubmittedforpublicationbecause
theresultswereconsideredimplausible.Hence,thetruedistributionfromsingle
8/14/2019 SSRN-id1359506
22/33
23
regionswouldincludebothincreasesanddecreasesineventsandameanclosetozero,
whilethepublishedrecordwouldshowonlydecreasesinevents.Thus,publicationbias
couldplausiblyexplainwhydramaticshorttermpublichealthimprovementswere
seeninpriorstudiesofsmokingbans.
Ourstudyfocusesonlyonthehealtheffectsofsmokingbans. Futureresearch
shouldestimatenonhealthrelatedbenefitsofthesebanstononsmokers.Priortoa
smokingban,nonsmokersatriskforrespiratorysymptomsorcardiovascularevents
mighthaveavoidedbusinesseswithhighETSlevels.Afteraban,nonsmokerscould
gaincomfortableaccesstothesebusinesses,butbasedonourfindingsinthisstudy,this
benefitwouldnotalsoresultinreducedhospitalizationordeathrates.Ourstudy
designplausiblyidentifiesonlyshorttermbenefitsofsmokingbans(ashasthestudy
designsusedbypreviousstudies).Wecannotanalyzewhethersmokingbansimprove
longtermtrendsforchroniccardiovasculardiseaseorlungcancer.Inaddition,
smokingbansmayinducesmokerstoquitordiscouragenonsmokersfromstarting
smoking.Thesepotentiallongtermbenefitswillnotbeapparentinstudyofshortterm
outcomesandwouldbenefitfromfurtherstudy.
8/14/2019 SSRN-id1359506
23/33
24
REFERENCE
Adams,S.andC.Cotti.(2008).DrunkDrivingAfterthePassageofSmokingBansin
Bars,JournalofPublicEconomics92(56):12881305.Adda, J. and F. Cornaglia. (2006, 18 December 2007). "The Effect of Bans and Taxes on Passive
Smoking." CeMMAP working paper series, from
http://www.ucl.ac.uk/~uctpjea/passive.pdf.
Agency for Healthcare Research and Quality. (2006). "INTRODUCTION TO THE HCUPNATIONWIDE INPATIENT SAMPLE (NIS), 2004." Retrieved 18 December 2007,
from http://www.hcup-us.ahrq.gov/db/nation/nis/NIS_Introduction_2004.pdf.
American Nonsmokers' Rights Foundation. (2007). "Percent of U.S. State/Commonwealth
Populations Covered by 100% Smokefree Air Laws." from http://www.no-smoke.org/pdf/EffectivePopulationList.pdf.
Barnoya, J. and S. A. Glantz (2005). "Cardiovascular effects of secondhand smoke: nearly as
large as smoking." Circulation 111(20): 2684-98.Barone-Adesi, F., L. Vizzini, et al. (2006). "Short-term effects of Italian smoking regulation on
rates of hospital admission for acute myocardial infarction." Eur Heart J 27(20): 2468-72.Bartecchi, C., R. N. Alsever, et al. (2006). "Reduction in the incidence of acute myocardial
infarction associated with a citywide smoking ordinance." Circulation 114(14): 1490-6.
Centers for Disease Control and Prevention (2000). "State-specific prevalence of current
cigarette smoking among adults and the proportion of adults who work in a smoke-freeenvironment--United States, 1999." MMWR Morb Mortal Wkly Rep 49(43): 978-82.
Centers for Disease Control and Prevention (2007). "Reduced secondhand smoke exposure after
implementation of a comprehensive statewide smoking ban--New York, June 26, 2003-
June 30, 2004." MMWR Morb Mortal Wkly Rep 56(28): 705-8.Cesaroni, G., F. Forastiere, et al. (2008). "Effect of the Italian smoking ban on population rates
of acute coronary events." Circulation 117(9): 1183-8.
Evans, W., M. Farrelly, et al. (1999). "Do Workplace Smoking bans Reduce Smoking?"American Economic Review 89(4): 728-747.
Farkas, A. J., E. A. Gilpin, et al. (1999). "The effects of household and workplace smoking
restrictions on quitting behaviours." Tob Control 8(3): 261-5.Farrelly, M. C., W. N. Evans, et al. (1999). "The impact of workplace smoking bans: results from
a national survey." Tob Control 8(3): 272-7.
Gallus, S., P. Zuccaro, et al. (2007). "Smoking in Italy 2005-2006: effects of a comprehensive
National Tobacco Regulation." Prev Med 45(2-3): 198-201.Glantz, S. A. and W. W. Parmley (1991). "Passive smoking and heart disease. Epidemiology,
physiology, and biochemistry." Circulation 83(1): 1-12.
He, J., S. Vupputuri, et al. (1999). "Passive smoking and the risk of coronary heart disease--a
meta-analysis of epidemiologic studies." N Engl J Med 340(12): 920-6.Hochberg, Y. (1988). "A sharper Bonferroni procedure for multiple tests of significance."
Biometrika 75(4): 800-2.Hoidrup, S., E. Prescott, et al. (2000). "Tobacco smoking and risk of hip fracture in men and
women." Int J Epidemiol 29(2): 253-9.
Juster, H. R., B. R. Loomis, et al. (2007). "Declines in hospital admissions for acute myocardial
infarction in New York state after implementation of a comprehensive smoking ban." AmJ Public Health 97(11): 2035-9.
8/14/2019 SSRN-id1359506
24/33
25
Longo, D. R., R. C. Brownson, et al. (1996). "Hospital smoking bans and employee smokingbehavior: Results of a national survey." Jama 275(16): 1252-7.
Metzger, K. B., F. Mostashari, et al. (2005). "Use of pharmacy data to evaluate smoking
regulations' impact on sales of nicotine replacement therapies in New York City." Am JPublic Health 95(6): 1050-5.
Ong, M. K. and S. A. Glantz (2004). "Cardiovascular health and economic effects of smoke-freeworkplaces." Am J Med 117(1): 32-8.OnTheIssues.org. (2007). "2007 Democratic primary debate at Dartmouth College." Retrieved
18 December 2007, from
http://www.ontheissues.org/Archive/2007_Dem_primary_Dartmouth_Drugs.htm .
Pell, J. P., S. Haw, et al. (2008). "Smoke-free legislation and hospitalizations for acute coronarysyndrome." N Engl J Med 359(5): 482-91.
Petersen, L. A., S. Wright, et al. (1999). "Positive predictive value of the diagnosis of acute
myocardial infarction in an administrative database." J Gen Intern Med14(9): 555-8.Sargent, R. P., R. M. Shepard, et al. (2004). "Reduced incidence of admissions for myocardial
infarction associated with public smoking ban: before and after study." Bmj 328(7446):
977-80.Thun, M., J. Henley, et al. (1999). "Epidemiologic studies of fatal and nonfatal cardiovascular
disease and ETS exposure from spousal smoking." Environ Health Perspect 107 Suppl 6:
841-6.Tirschwell, D. L. and W. T. Longstreth, Jr. (2002). "Validating administrative data in stroke
research." Stroke 33(10): 2465-70.
U.S. Department of Health and Human Services. (2006). The Health Consequences of
Involuntary Exposure to Tobacco Smoke: A Report of the Surgeon General. USDHS.Atlanta, GA.
8/14/2019 SSRN-id1359506
25/33
26
Figure1
0
10
20
30
40
Percentageofregionswithbans
1988 1992 1996 2000 2004Year
Any smoking restriction
Workplace restriction
Regions are 3 digit US postal (zip) codesSource: Authors' analysis of data from American Non-smoker's Rights Foundation and California Department of Industrial Relations
Figure 1. Prevalence of smoking bans over time
Regionsare3digitU.S.postal(zip)codes
Source:AuthorsanalysisofdatafromAmericanNonsmokersRightsFoundationand
CaliforniaDepartment
of
Industrial
Relations
8/14/2019 SSRN-id1359506
26/33
27
Figure2.
0
2
4
6
8
10
Percent
-200 -150 -100 -50 0 50 100 150 200Relative difference in acute myocardial infarction incidence (%)
Percent change
Normal density curve
Mean change
Source: HCUP. Hospitals with smoking restrictions: 77. Control hospitals: 631 Pairwise comparisons: 19406
Figure 2. Changes in AMI admissions relative to all control hospitals
Source:AuthorsanalysisofdatafromNationwideInpatientSample.Hospitalswith
smokingrestrictions:52.Controlhospitals:336.Pairwisecomparisons:8,529.Relative
differencemaybelessthan 100%ifalargepercentageincreaseinacontrolhospitalis
subtractedfromalargedecreaseinasmokingrestrictedhospital.
8/14/2019 SSRN-id1359506
27/33
28
Figure3.
0
2
4
6
8
1
0
Percent
-75 -50 -25 0 25 50 75Relative difference in acute myocardial infarction incidence (%)
Percent change
Normal density curve
Mean change
Source: MCD. Counties with smoking restrictions: 89. Control counties: 462 Pairwise comparisons: 23938
Figure 3. Changes in AMI deaths relative to all control counties
Source:AuthorsanalysisofdatafromMultiplecauseofdeathdatabase.Countieswith
smokingrestrictions:89.Controlcounties:462.Pairwisecomparisons:23,905.Relative
differencemaybelessthan 100%ifalargepercentageincreaseinacontrolcountyis
subtractedfromalargedecreaseinasmokingrestrictedcounty.
8/14/2019 SSRN-id1359506
28/33
29
Table1.Datasourcecharacteristics
Datasource
NationwideInpatientSample:
Hospitals 874
States
27
Years 19932004
Alladmissions 21,820,484
Admissionsbydisease
Acutemyocardialinfarction 217023
CombinedasthmaandCOPD 433674
MultipleCauseofDeath:
Counties 468
States 50
Years 19892004
Alldeaths 24,610,532
Acutemyocardialinfarction 2,042,812
Medicarepatients:
Regions(3digitzipcodes) 868
States 51
Years 19972004
IncludedMedicarepopulation(personyears) 275,303,008
Alldeaths 13,106,175
Alladmissions
Admissionsby
disease
72,542,544
Acutemyocardialinfarction 2,382,386
CombinedasthmaandCOPD 2,984,382
Hipfracture 3,381,690
Chronicobstructivepulmonarydisease
Source:AuthorsanalysisofdatafromNationwideInpatientSample,MultipleCauseof
Deathfiles,and100%MedicareProviderAnalysisandReviewfiles.
Chronicobstructivepulmonarydisease
Source:AuthorsanalysisofdatafromNationwideInpatientSample,nationaldeath
statistics,and100%MedicareProviderAnalysisandReviewfiles.
8/14/2019 SSRN-id1359506
29/33
30
Table2.Mortalityandsmokingrestrictions*
Disease %changeinmortality(95%CI) Pvalue
Workplacesmokingrestrictions:
Alldeaths
(0
17
years
old)
4(10.2
to
2.2)
0.204
AMI (1864yearsold) 4.4(11.6to2.7) 0.224
Alldeaths(1864yearsold) 1.1(2.9to0.8) 0.247
Alldeaths(age65+) 0.8(1.2to2.8) 0.413
AMI(allages) 1.5(4.8to1.8) 0.374
Alldeaths(allages) 0.3(1.6to0.9) 0.624
Anysmokingrestrictions:
Alldeaths(017yearsold) 2(6.5to2.6) 0.400
AMI(1864yearsold) 3.5(10to3.1) 0.299
Alldeaths(1864yearsold) 0.4(2.1to1.2) 0.601
Alldeaths(age65+) 1.4(3.0to0.2) 0.062
AMI(allages) 1.1(4.1to1.9) 0.470
Alldeaths(allages) 0.5(0.6to1.6) 0.396
*AuthorsanalysisofdatafromMultipleCauseofDeathfiles,years19932004exceptin
age65+,whicharefrom100%MedicareProviderAnalysisandReviewfiles,19972004.
AMIindicatesdeathsfromacutemyocardialinfarction.
8/14/2019 SSRN-id1359506
30/33
31
Table3.Hospitaladmissionsandsmokingrestrictions
Disease %changeinadmissions
(95%CI)
Pvalue
Workplacesmokingrestrictions:
Alladmissions
(age
017)
3.7
(2.1
to
9.5)
0.211
Asthma(age017) 11.4(2.4to25.3) 0.106
AMI(age1864) 4.2(10.2to1.7) 0.165
Alladmissions(age1864) 2.2(0.4to4.8) 0.101
Asthma(age1864) 3.4(6.2to13.1) 0.485
COPD(age1864) 1.7(8.6to12) 0.746
Hipfracture(age1864) 5.1(15.6to5.4) 0.340
AMI(age65+) 2(3.7to7.7) 0.477
Alladmissions(age65+) 1.8(3.1to6.7) 0.477
Asthma(age65+) 0.8(13.5to15.1) 0.909
COPD(age65+) 3.4(3.6to10.5) 0.343
Hipfracture(age65+) 0.1(2.4to2.6) 0.946
Anysmokingrestrictions:
Alladmissions(age017) 4.6(0.9to10.1) 0.101
Asthma(age017) 13.7(2to29.3) 0.087
AMI(age1864) 4.7(10.3to01) 0.104
Alladmissions(age1864) 2.3(0.3to4.9) 0.083
Asthma(age1864) 3.5(5to11.9) 0.422
COPD(age1864) 2.1(11.8to7.7) 0.676
Hipfracture
(age
18
64)
4.7
(14.6
to
5.1)
0.348
AMI(age65+) 5.1(0.4to10.6) 0.068
Alladmissions(age65+) 1.7(2.0to5.4) 0.364
Asthma(age65+) 6.8(4.78to18.4) 0.246
COPD(age65+) 2.7(2.4to7.8) 0.303
Hipfracture(age65+) 1.1(1.4to3.6) 0.385
Source:AuthorsanalysisofdatafromNationwideInpatientSample(199320004)except
age65+,whicharefrom100%MedicareProviderAnalysisandReviewfiles,years1997
2004.
Acutemyocardialinfarction/Ischemicheartdisease
Chronicobstructivepulmonarydisease
8/14/2019 SSRN-id1359506
31/33
32
Table4.Comparisonofregressionmodels:
Allcausemortalityfollowingsmokingbanofanysort(MCDcounties)Variablesincluded (1) (2) (3) (4)
Anybans 0.495 0.324 0.328 1.335
(0.583) (0.573) (0.794) (0.547)
Hospitalbeds/person 0.0031
(0.0013)
Countypopulation 0.00005 0.00006 0.00005 0.00009
(0.00002) (0.00002) (0.00002) (0.00003)
Physicians/person 0.0003 0.0002 0.0007
(0.0007) (0.0008) (0.0010)
Percentpopulationinlaborforce 4.5
(6.6)
#Yearsinsample 15 13 9 5
#observations 24884 21580 15512 7308
Resultsfrom
original
model
0.495 0.322 0.330 1.343
(0.583) (0.573) (0.804) (0.546)
*Eachcoefficientrepresentsthe%changeinoutcomesduetoa1unitchangeinthe
explanatoryvariable. Standarderrorsareclusteredatthearealevelandarereported
inparentheses.
8/14/2019 SSRN-id1359506
32/33
33
Table5.Comparisonofregressionmodels:
Myocardialinfarctionadmissionratesinworkingageadults(NIShospitals)
followingworkplacebansVariablesincluded (1) (2) (3) (4)
Workplacebans 4.2 4.1 13.3 9.2
(3.1) (3.0) (5.8) (5.9)
Hospitalbeds/person 0.0054
(0.0029)
Countypopulation 0.00004 0.00004 0.00003 0.00023
(0.00003) (0.00003) (0.00002) (0.00008)
Physicians/person 0.0024 0.0020 0.0027
(0.0025) (0.0028) (0.0077)
Percentpopulationinlaborforce 310.3
(324.9)
#Yearsinsample 12 12 9 5
#observations
7476 7476 4932 1502
Resultsfromoriginalmodelon
thissample
4.2 4.2 13.3 9.2
(3.1) (3.1) (5.8) (5.7)
*Eachcoefficientrepresentsthe%changeinoutcomesduetoa1unitchangeinthe
explanatoryvariable. Standarderrorsareclusteredatthearealevelandarereported
inparentheses.
8/14/2019 SSRN-id1359506
33/33
Table6.Comparisonofregressionmodels:
Allcauseadmissionratesinworkingageadults(NIShospitals)
Variablesincluded (1) (2) (3) (4)
Workplacebans 2.20 2.12 0.35 1.52
(1.34) (1.33) (1.73) (2.10)
Hospitalbeds/person 0.000297
(0.000998)
Countypopulation 0.00004 0.00004 0.00003 0.00013
(0.00002) (0.00002) (0.00002) (0.00004)
Physicians/person 0.00126 0.00119 0.00185
(0.000993) (0.00109) (0.00299)
Percentpopulationinlaborforce 8.2
61.8)
#Yearsinsample 12 12 9 5
#observations
7476 7476 4932 1502
Resultsfromoriginalmodel 2.20 2.20 0.26 1.40
(1.34) (1.34) (1.73) (2.08)
*Eachcoefficientrepresentsthe%changeinoutcomesduetoa1unitchangeinthe
explanatoryvariable. Standarderrorsareclusteredatthearealevelandarereported
inparentheses.