The RAND Corporation is a nonprofit institution that helps improve... decisionmaking through research and analysis.

CHILDREN AND FAMILIES EDUCATION AND THE ARTS The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. ENERGY AND ENVIRONMENT HEALTH AND HEALTH CARE INFRASTRUCTURE AND TRANSPORTATION This electronic document was made available from www.rand.org as a public service of the RAND Corporation. INTERNATIONAL AFFAIRS LAW AND BUSINESS Skip all front matter: Jump to Page 16 NATIONAL SECURITY POPULATION AND AGING PUBLIC SAFETY SCIENCE AND TECHNOLOGY TERRORISM AND HOMELAND SECURITY Support RAND Browse Reports & Bookstore Make a charitable contribution For More Information Visit RAND at www.rand.org Explore the Pardee RAND Graduate School View document details Limited Electronic Distribution Rights This document and trademark(s) contained herein are protected by law as indicated in a notice appearing later in this work. This electronic representation of RAND intellectual property is provided for noncommercial use only. Unauthorized posting of RAND electronic documents to a non-RAND website is prohibited. RAND electronic documents are protected under copyright law. Permission is required from RAND to reproduce, or reuse in another form, any of our research documents for commercial use. For information on reprint and linking permissions, please see RAND Permissions. This product is part of the Pardee RAND Graduate School (PRGS) dissertation series. PRGS dissertations are produced by graduate fellows of the Pardee RAND Graduate School, the world’s leading producer of Ph.D.’s in policy analysis. The dissertation has been supervised, reviewed, and approved by the graduate fellow’s faculty committee. Food Security, Livelihoods, and Antiretroviral Therapy for HIV Evidence for Policy in Resource-Limited Settings Kartika Palar This document was submitted as a dissertation in May 2012 in partial fulfillment of the requirements of the doctoral degree in public policy analysis at the Pardee RAND Graduate School. The faculty committee that supervised and approved the dissertation consisted of Kathryn Pitkin Derose (Chair), Homero Martinez, and Krishna Kumar. Sheri Weiser was the outside reader for the dissertation. PARDEE RAND GRADUATE SCHOOL The Pardee RAND Graduate School dissertation series reproduces dissertations that have been approved by the student’s dissertation committee. The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND’s publications do not necessarily reflect the opinions of its research clients and sponsors. R® is a registered trademark. All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from RAND. Published 2012 by the RAND Corporation 1776 Main Street, P.O. Box 2138, Santa Monica, CA 90407-2138 1200 South Hayes Street, Arlington, VA 22202-5050 4570 Fifth Avenue, Suite 600, Pittsburgh, PA 15213-2665 RAND URL: http://www.rand.org To order RAND documents or to obtain additional information, contact Distribution Services: Telephone: (310) 451-7002; Fax: (310) 451-6915; Email: order@rand.org TableofContents Acknowledgements Funding Acronyms Overview…………………………………………………………………………………………………………………….. 1 I.Effectoffoodassistanceonfoodsecurityandnutritionalstatusamongpatients receivingantiretroviraltherapyforHIVinHonduras Abstract………………………………………………………………………………………………………………… 6 Introduction………………………………………………………………………………………………………….. 8 Methods………………………………………………………………………………………………………………… 12 Results…………………………………………………………………………………………………………………… 23 Discussion……………………………………………………………………………………………………………… 37 Conclusion……………………………………………………………………………………………………………… 45 Appendix………………………………………………………………………………………………………………… 47 II.LivelihoodexperiencesofpeoplereceivingintegratedHIVtreatmentandfood assistanceinBolivia:Lessonsforsustainableinterventions Abstract………………………………………………………………………………………………………………… 52 Introduction………………………………………………………………………………………………………….. 54 Methods………………………………………………………………………………………………………………… 57 Results…………………………………………………………………………………………………………………… 62 Discussion……………………………………………………………………………………………………………… 70 Conclusion……………………………………………………………………………………………………………… 78 III.Roleofantiretroviraltherapyinimprovingfoodsecurityamongpatients initiatingHIVtreatmentandcareinUganda Abstract…………………………………………………………………………………………………………………. 79 Introduction………………………………………………………………………………………………………….. 81 Methods…………………………………………………………………………………………………………………. 84 Results…………………………………………………………………………………………………………………… 90 Discussion……………………………………………………………………………………………………………… 100 Conclusion……………………………………………………………………………………………………………… 104 References…………………………………………………………………………………………………………………. 106 iii Acknowledgements Firstandforemost,IthankmycommitteechairKatieDerose,whoopenedmany opportunitiesformeovertheyears,gavemethefreedomtopursuethemindependently,and helpedmetodeveloptheskillsIneededtotakefulladvantageofthem.Ithankmycommittee memberHomeroMartinezforinvolvingmeinthedatacollectionprocessthatledtomy dissertationpaperonHonduras,andforprovidingsuchacompellingexampleofleadershipin buildingpartnershipsinthefield.IthankmycommitteememberKrishnaKumarforhissupport, intellectualinsight,andmentorshipovermyyearsatRAND.Myoutsidereader,SheriWeiserat UCSF,wentaboveandbeyondherofficialdutiestoprovidecriticalandongoingmentorshiponmy dissertationandcareer,andIthankherdeeply.IowemuchgratitudetoGlennWagnerfor providingthedataandadvisingmydissertationpaperonUganda,andSebastianLinnemayrand BonnieGhosh‐Dastidarfortheirwillingnessandpatienceinmentoringmeinthefinerpointsof longitudinaldataanalysis.IalsothankSebastianLinnemayrforteachingmehowtowadethrough thedelugeofstatisticsfrommydataandfocusontellingagoodstorywiththem.Thanksto AlexandriaSmithforproblemsolvingdatamanagementissueswithbothskillandgoodhumor. Thisdissertationwouldnothavebeenpossiblewithoutthesupportofourmanypartnersin thefield.GivenmyleadroleintheBoliviastudy,Iamparticularlyindebttomycollaboratorsthere. First,IamfilledwithappreciationforAlexisMartin,mycounterpartfromtheWorldFoodProgram, RegionalOfficeforLatinAmericaandtheCaribbean(WFP‐LAC)andmymaincollaboratoronthe Boliviastudy.Ourcross‐cuttingresearch‐practitionerpartnershipoverthelastfouryearswasa highlightofmyPhDexperienceandhelpedkeepmyresearchgroundedinthe‘realworld’.Ialso v deeplythankHugoFariasatWFP‐LACforsupportingthefieldteaminBolivia,andJayneAdamsat WFP‐LACforopeningtheopportunitytoexplorelivelihoodsandHIVinBoliviainthefirstplace.In thefield,IthankMarthaBanzer,OliviaLoayza,andWillanMontaño,whoconductedtheinterviews anddatacollectionprocesswithsuchprofessionalismanddedication;clinicnutritionistsIsela Patón,GonzaloRamírez,andXimenaRojasfortheirhardworkinparticipantrecruitmentanddata collection;Dra.CarolaValenciafromtheNationalAIDSProgramforhersupportofourstudy; finally,IthankthemanystaffattheWFPCountryOfficeinvolvedintheimplementationofthis research,withspecialthankstoVitóriaGinja,SergioTorresandXimenaLoza.Thisresearchwould nothavebeenpossiblewithouttheparticipationoftheAsociaciónUnNuevoCamino(ASUNCAMI), partoftheBolivianNetworkofPeopleLivingwithHIV/AIDS(REDBOL)andthenumerousleaders inthecommunityofpeoplelivingwithHIVinLaPaz,CochabambaandSantaCruzwhooffered invaluablefeedbackandassistance.InHonduras,IthankBlancaRamírez,MonicaHeinemann, LourdesJimenez,AngelicaMorales,DinaRodriguez,andMarthaSuazofortheirdedicationinthe field,responsivenesstomymanyquestionsaboutthedataanddatacollectionprocess,andwarmth andopennessduringmyfieldvisits.Mostimportantly,Iofferdeepappreciationtotheparticipants inBolivia,HondurasandUgandawhogavesogenerouslyoftheirtimeandpersonalinformation. Andlastbutnotleast,Ithankmyfamilyandfriendsfortheirunfailingsupportandbeliefin methroughoutthisprocess. vi Funding Thisdissertationwasmadepossiblebyseveralsourcesoffunding: IoffermydeepestgratitudetoFredPardee,andthePardeeRANDGraduateSchool,forthe PardeeDissertationAwardwhichsupportedmuchofmytimebuildingpartnershipsandpreparing forfieldworkinBolivia,aswellasanalysisofdatafromUganda. ManythankstotheWorldFoodProgram(RegionalOfficeforLatinAmericaandthe Caribbean),whofundeddatacollectionformystudyinBoliviathroughagrantfromtheOPECFund forInternationalDevelopment. DatacollectioninHonduraswasfundedbytheNationalInstituteofMentalHealth (R34MH084675;PI:HomeroMartinez).DatacollectioninUgandawasfundedbytheRockefeller Foundation(HE007;PI:GlennWagner). vii Acronyms AIDS Acquiredimmunedeficiencysyndrome ART Antiretroviraltherapy ARV(s) Antiretroviraldrug(s)ormedications(s) BMI Bodymassindex ELCSA FA EscalaLatinoamericanadeSeguridad Alimentaria(LatinAmericanFoodSecurity Scale) Foodassistance HIV Humanimmunodeficiencyvirus OW Overweightorobese PLHIV PeoplelivingwithHIV UN UnitedNations WFP WorldFoodProgram WFP‐LAC WorldFoodProgram– RegionalOfficeforLatin AmericaandtheCaribbean WorldHealthOrganization WHO ix Overview DonorfundingforHIVinlowandmiddleincomecountriesincreased6‐foldsince2002, reaching$6.9billionindisbursementsin2010(Katesetal.,2011).Thesefundsrepresentamassive investmentinpreventingandtreatingHIV.By2010,almost7millionpeoplelivingwithHIV (PLHIV)werereceivingtreatmentwithantiretroviraltherapy(ART)indevelopingcountries, transformingHIVfromadeathsentenceintoamanageablechronicdiseaseforthoseabletoaccess it(UNAIDS,2011). Inthiscontext,animportantpolicychallengeisensuringthatgovernments,healthsystems andorganizationsworkinginsupportofPLHIVprovideservicesinsuchawaythatpeopleonART canfullybenefitfromtreatmentoverthelongterm.Thesebenefitsincludeimprovementsin individualhealth,whicharequitesignificant(Bartlettetal.,2001;Murphyetal.,2001),aswellas resumedeconomicproductivity,whichliteraturenowsuggestscanalsobesubstantial (Thirumurthyetal.,2011;Thirumurthyetal.,2008a;Thirumurthyetal.,2008b).Inaddition,the potentialforpositiveexternalitiesfromsustainedlifetimeARTunderscorestheimportanceof supportingpoliciesthatpromotegoodadherenceandminimizetreatmentinterruptionand attrition.TheseexternalitiesincludereducedHIVtransmission(i.e.“treatmentasprevention)and reduceddevelopmentofdrugresistantHIVstrains.MaximizingandsustainingthegainsfromART willrequirenotonlyidentifyingbarrierstoARToutcomes,butalsothepoliciesandinterventions thatbestreducethesebarriersandfacilitategoodadherence,treatmentretention,andultimately, positivehealthoutcomes. FoodinsecurityandpoornutritionalstatushavebeenidentifiedasbarrierstoART adherence,treatmentretention,andHIVoutcomesinresource‐limitedsettings.Evidenceabounds forthenegativeeffectsoffoodinsecurityonARTadherenceandtreatmentretention,with 1 implicationsforpoorCD4count,viralsuppression,morbidityandmortality(Anemaetal.,2009; Frankeetal.,2011;Weiseretal.,2009b;Weiseretal.,2009c;Weiseretal.,2012).These relationshipslikelyoperatethroughacombinationofbiologic,nutritionalandbehavioralpathways. Forexample,foodinsecuritymaycreateorexacerbatepoornutritionalstatus(e.g.lowBMI)which couldleadtopoorclinicaloutcomes(Weiseretal.,2009b).Ontheotherhand,foodinsecuritymay compromiseARTadherenceiflackoffoodisanissue,sincemanyantiretroviralmedicationsmust betakenwithfood(Deribeetal.,2008).FoodinsecuritycanalsoreduceARTaccess,adherenceand retentionifitleadstotrade‐offsbetweentreatment(whichinvolvesbothdirectandindirectcosts, suchasfees,transport,andlostworktime)andotherbasicindividualandhouseholdneeds(Martin etal.,2011b). Facedwiththesechallenges,ARTprogramsareincreasinglyintegratinginterventionsto supportthefoodsecurityofpatients,includingthroughdirectfoodassistance,nutritionalsupport, andlivelihoodsprograms(Byronetal.,2008;Fregaetal.,2010;J.Koetheetal.,2009;Tirivayietal., 2011a).Inthiscontext,mydissertationbroadlyasks:Whatpoliciesandinterventionswillbest reducefoodinsecurityandpoornutritionasbarrierstoARToutcomes?Iapproachthisquestion throughthreepapersthatexploredifferentaspectsofpoliciesaffectingfoodinsecurityandART, across2continents(LatinAmericaandAfrica)and3countries(Honduras,Bolivia,andUganda). Inmyfirstpaper,IfocusonPLHIVreceivingARTwhowerepartofanutritioneducation andfoodassistancepilotinterventioninHonduras,sponsoredbytheWorldFoodProgram(WFP) andformallyevaluatedthroughanNIH‐fundedstudy.Thereislittlepublishedevidencetoguide programsandpolicymakersconsideringintegratingfoodsecurityinterventionswithART, includingwhetherdirectfoodassistanceactuallyimprovesfoodsecurityandnutritionalstatus. Thisisparticularlysoinsettingswherehighprevalenceofhouseholdfoodinsecurity,overweight andobesitycoexistamongPLHIV.Thus,myresearchquestionsforthispaperwere:1)whatisthe 2 effectoffoodassistanceonhouseholdfoodinsecurityforpeopleonART?and2)whatistheeffect offoodassistanceonBMI(and,specifically,arethereadverseeffectsonoverweightandobese participants)?Toanswermyresearchquestions,Iemploymultivariatelongitudinalregressionwith individualfixedeffects.Ifindthatfoodassistanceplusnutritioneducationimproveshousehold foodsecurityamongARTrecipientsaboveandbeyondnutritioneducation‐only,anddoesnothave adverseeffectsonoverweightorobeseparticipantsovera12‐monthperiod.Trendsin improvementinfoodsecurityandBMIamongthenutritioneducation‐onlygroupsuggestthat nutritioneducationmayalsohavepositiveeffectsonthewell‐beingofPLHIV.However,Icouldnot formallytesttheeffectofnutritioneducation,giventheabsenceofacontrolgroupreceivingno intervention.TogetherwithliteratureidentifyingfoodinsecurityasabarriertoadherenceandHIV outcomes,ourresultssuggestthatfoodassistancemayimprovetheseoutcomesviaimprovedfood security.However,implementationissuesaroundfoodassistanceshouldbecarefullyconsidered, alongwithpotentialalternativeinterventions,toensuresustainabilityinresource‐limitedsettings. Inmysecondpaper,IagainfocusonPLHIVreceivingARTwhowerepartofaWFP‐ sponsoredfoodassistancepilotprogram,thistimeinBolivia.UnlikeinHonduras,therewasno formalstudycomponentbuiltintothepilottoevaluatethefoodassistanceintervention.However, WFPwasinterestedinexploringtransitionstrategiesfromfoodassistance–inparticular,the potentialforlivelihoodinterventionstoprovidetransitionfromfoodassistanceandpromotemore sustainablefoodsecurityinthelongtermforpeopleonART.Livelihoodinterventionstoimprove foodsecurityandsustainableHIVtreatmentoutcomesareincreasinglypromotedforpeopleliving withHIVreceivingART.Yet,anin‐depthunderstandingofhowfoodinsecurePLHIVexperience theirownlivelihoodsinrelationtoHIVtreatment(intheabsenceofexternalprograms)islacking, especiallyinurbansettingsindevelopingcountries.Thus,inthisstudy,Iaimtoexploreand describetheinterconnectionbetweenlivelihoodexperiencesandARTinthreecitiesinBolivia,in ordertoidentifymajorbarriersandopportunitiesforlivelihood‐relatedpoliciesandinterventions 3 inthecontextofART.Closed‐endedquestionnairesandqualitativeinterviewswereconductedwith participantsofthefoodassistancepilot,capturingquantitativedataondemographics,household composition,socio‐economicsituation,includingworkstatus,andfoodinsecurity,andqualitative dataonwork‐relatedbarrierstoARTadherence,HIV‐relatedbarrierstowork,rangeofeconomic activitiesconducted,andeconomiccopingstrategies.Analyzingthesedata,Ifindthatstudy participantshavecomplexeconomiclivesoftencharacterizedbymultipleeconomicactivities, includingbothformalandinformallabor.TheystruggletomanageARTtreatmentandlivelihoods simultaneously,andfacebarrierstothisdualmanagementthatrangedfromtheinterpersonalto thestructural.Inparticular,issuesoflackofdisclosureofHIVstatus,stigmaanddiscrimination, arehighlysalient.Inaddition,healthsystemissuessuchaslimitedclinichoursordrugshortages exacerbatethestruggletobalanceeconomicactivitieswithHIVtreatment.Improvedpolicy‐level effortstoenforceexistinganti‐discriminationlaws,reduceHIV‐relatedstigma,andexpandhealth servicesaccessibilitycouldmitigatemanyofthebarriersdiscussedbyourparticipantsandreduce theneedforseparatelivelihoodinterventions. Inmyfinalpaper,IturntoUgandatoexploreadifferentquestion:howdoesARTaffectfood security?WeknowthatfoodsecurityaffectsART,butisthereactuallyabidirectionalrelationship betweenthetwo?FewstudieshaveexaminedifandhowARTaffectsfoodinsecurity,althoughthe scientificliteraturesuggeststheremaybeabenefitviaimprovedhealthandabilitytowork.Using datafroma12‐monthprospectivecohortstudy,Iemploymultivariatelongitudinallogistic regressiontoinvestigatewhetherARTdecreasesfoodinsecuritycomparedtoHIVcarewithout ARTamongasampleoftreatment‐naïvepatientsinitiatingclinicalcareinUganda,andtoexplore thepotentialpathwaysthroughwhichARTmayaffectfoodinsecurity,includingimprovedmental health,physicalhealth,andworkstatus.Ifindthatfoodinsecuritydecreasessignificantlyforboth theARTandnon‐ARTgroupsovertime,withtheARTgroupexperiencinggreaterreductionsbythe endofthestudy.ARTremainsasignificantpredictorofreductioninfoodinsecurityovertimeafter 4 controllingforbaselinedifferencesinthemultivariatelongitudinalregressionmodel. Improvementsinworkandmentalhealthstatusaremoststronglyassociatedwithdecreasedfood insecurityovertimeandweakenedtheARTcoefficientsignificantlywhenaddedtothemodel. Takentogetherwiththewell‐knownbenefitsoffoodsecurityonARTadherence,treatment retentionandclinicaloutcomesinresource‐poorsettings,ourresultssuggestan“upwardspiral”of improvedfunctioningandproductivitycouldresultfrompositivefeedbackbetweenfoodsecurity andART.Policymakerscouldleveragethispositivecyclebystrengtheningmentalhealthsupport andpromotingsustainablefoodsecurityinterventionsaspartofHIVtreatmentprograms. Takentogether,mythreepapersprovideevidencethatfoodassistance,livelihood interventions,andARTallhavearoletoplayinimprovingtheeconomicandnutritionalwell‐being ofpeoplelivingwithHIVindevelopingcountries,butthattheyarelikelytoworkbestwhenwell‐ targeted(tothosewhoneedthemmost,atthepointintimetheyneedthemmost),andintegrated withbothcomprehensivecare(includingmentalhealthsupport)andsocialsafetynets.In particular,myresultsindicatethatintegratingART,foodassistance,nutritionalsupport,and livelihoodsprogramsinanefficient,sustainablemannercouldeffectivelycreateapositivefeedback loopbetweenfoodsecurityandART.Policymakerscouldleveragethis“upwardspiral”inwell‐ beingtocounteractthe“viciouscycle”ofHIVandfoodinsecuritythathastakensuchatollin resource‐limitedsettings(Bukusubaetal.,2007;Crushetal.,2011).Thiscannotonlyimprovethe livesofPLHIVaroundtheworld,buthelprealizethegainsofdonorandrecipientcountrieswho investedbillionsofdollarsandsignificanthumancapitalinfulfillingthepromiseofARTtosaveand transformlives. 5 I.Effectoffoodassistanceonfoodsecurityandnutritionalstatusamong patientsreceivingantiretroviraltherapyforHIVinHonduras ABSTRACT Background:ThedeleteriouseffectsoffoodinsecurityandundernutritiononHIV treatmentoutcomesandantiretroviraltherapy(ART)adherencearenowwellrecognizedin resource‐limitedsettings.InterventionstoaddressfoodsecurityforpeoplelivingwithHIV(PLHIV) arethereforebeingplannedandimplementedinregionsacrosstheworld.However,thereislittle publishedevidencetoguideprogramsandpolicymakersconsideringintegratingfoodsecurity interventionswithART,includingwhetherdirectfoodassistanceactuallyimprovesfoodsecurity andnutritionalstatus.Thisisparticularlysoinsettingswherehighprevalenceofhouseholdfood insecurity,overweightandobesitycoexistamongPLHIV Methods:Thispaperusesdatafroma12‐monthpilotinterventionstudyconductedfrom 2009‐2010in3citiesinHondurasamongPLHIVreceivingART.Thegoalofthepilotwasto investigatetheroleoffoodassistanceandnutritioneducationinimprovingfoodsecurity, nutritionalstatus,healthoutcomes,andultimately,ARTadherence.Inthispaper,wefocusonfood securityandbodymassindex(BMI)outcomes.Weemploymultivariatelongitudinalregression withindividualfixedeffectstodeterminewhetherfoodassistanceplusnutritioneducation improvedfoodsecurityasmeasuredbythevalidatedLatinAmericanandCaribbeanFoodSecurity Scale,comparedtonutritioneducationalone,overthreeassessments.Wethenusethesame regressionapproachtoexamineBMI,modifiedtoadditionallycaptureeffectsforparticipantswho wereoverweightorobeseatbaseline. Results:Thesampleincluded400participants,including203receivingfoodassistanceplus nutritioneducationand197receivingeducation‐only.Wefindthatfoodassistanceplusnutrition 6 educationimprovedthehouseholdfoodsecurityscoreby2.7points(p<0.01)(slightlylessthan onestandarddeviationofthemeanbaselinescore)aboveandbeyondthenutritioneducation‐only group,whosescoreimprovedby1.7points(p<0.01).Effectswerestrongerwhenthesamplewas limitedtowomen.Inaddition,wefoundthatfoodassistancewasnotassociatedwithadverse effectsonnutritionalstatusforparticipantswhowereoverweightorobeseatbaseline.Regardless ofstudygroup,wefoundasmalloveralltrendofimprovementinBMIforparticipantswhowere eitherunderweight(b=0.534;p<0.01)oroverweightorobese(b=‐0.316;p<0.05)atbaseline. However,withoutacontrolgroupreceivingnointervention,wecannottestwhetherthesetrends werecausallyduetothenutritioneducationprovided. Conclusions:FoodassistanceimproveshouseholdfoodsecurityamongasampleofART recipientsinLatinAmericaanddoesnothaveadverseeffectsonoverweightorobeseparticipants overa12‐monthperiod.Althoughtheabsenceofacontrolreceivingnointerventionlimitedour abilitytotesttheeffectofnutritioneducation,trendsindicatingimprovementinfoodsecurityand BMIamongthenutritioneducationgroupsuggestthatnutritioneducationmayalsohavepositive effectsonthewell‐beingofPLHIV,pointingtotheneedforfurtherinvestigation.Togetherwith literatureidentifyingfoodinsecurityasabarriertoadherenceandHIVoutcomes,ourresults suggestthatfoodassistancemayimprovetheseoutcomesviaimprovedfoodsecurity.However, implementationissuesaroundfoodassistanceshouldbecarefullyconsidered,alongwithpotential alternativeinterventions,toensuresustainabilityinresource‐limitedsettings. 7 INTRODUCTION ThedeleteriouseffectsoffoodinsecurityandmalnutritiononarangeofHIVantiretroviral therapy(ART)outcomes,includingmorbidity,mortality,adherenceandretentionincare,arenow wellrecognizedinresource‐limitedsettings(Anemaetal.,2009;Castlemanetal.,2004;Deribeet al.,2008;Marcellinetal.,2008;Normenetal.,2005;Oguntibejuetal.,2007;Weiseretal.,2009b; Weiseretal.,2009c;Weiseretal.,2012).Yet,evidencetoinformhowbesttoimproveandsustain foodsecurityandnutritionsoastopromoteoptimalHIVtreatmentoutcomesremains underdeveloped,particularlyforpopulationswithhighprevalenceofbothfoodinsecurityand overweightorobesity. Foodsecuritycanbedefinedasphysicalandeconomicaccesstoadequatefoodforall householdmembers,withoutriskoflosingsuchaccess(Haeringetal.,2009);foodinsecurityoccurs whenthereislimitedoruncertainavailabilityofnutritionallyadequateandsafefoods,orinability toacquirethesefoodsinsociallyacceptableways(Radimeretal.,1992).Meanwhile,malnutritionis theconditionofhavinginadequatevitamins,mineralsandnutrientstomaintainhealthytissueand organfunction.Malnutritionismostoftenassociatedwithundernutrition,butcanalsoaffectpeople whoareoverweightandobese.Whilethecoexistenceoffoodinsecurityandoverweight/obesity maybecounterintuitive,ithasbeenincreasinglydocumented–particularlyamongwomen–in bothresource‐richandresource‐limitedsettings(Alaimoetal.,2001a;Dinouretal.,2007; Tanumihardjoetal.,2007;Townsendetal.,2001),includinginLatinAmerica(Uauyetal.,2001). However,thisissuehasnotbeendirectlyexploredamongpeoplelivingwithHIV(PLHIV). Overthelast10years,theWorldHealthOrganization(WHO)andotherinternational organizationshaveissuedrecommendationsthatnutritionalassessment,counselingandsupport beastandardpartofcomprehensivecareforHIV(FANTA,2004;WorldBank,2007;WorldHealth Organization,2008),withspecificguidelinesforhigh‐riskpopulations(e.g.pregnantwomen, 8 patientswithHIVwasting,etc).Meanwhile,healthcareproviders,NGOsandinternational organizations–particularlytheUnitedNations(UN)WorldFoodProgramandtheUNFoodand AgricultureOrganization–haveincreasinglydevelopedandimplementeddiverseinterventionsto addressfoodinsecurityandmalnutritionforpeoplelivingwithHIV(PLHIV),rangingfrom nutritionalcounselingandeducation(Almeidaetal.,2011;Kayeetal.,2011;Torresetal.,2008), therapeuticmicro‐andmacronutrientsupplementation(J.Koetheetal.,2009;Rawatetal.,2010; Swaminathanetal.,2010;vanOosterhoutetal.,2010),householdfoodassistance(Byronetal., 2008;Cantrelletal.,2008;Iversetal.,2010),andlivelihoodsinterventions(Panditetal.,2010; Yageretal.,2011). ThehighprevalenceofundernutritioninplaceswithlargeandsevereHIVepidemics– primarilysub‐SaharanAfrica–hasledtoanimportantandgrowingbodyofresearchevaluating interventionstohelppeoplelivingwithHIVtogainandmaintainweightaspartoftreatmentand care(J.Koetheetal.,2009).Studiesrepeatedlyfindlowbodymassindex(BMI)(ameasureof weight‐for‐height)tobeastrong,independentpredictorofearlymortalityforpeopleonART(Liu etal.,2011;Mohetal.,2007;Weiseretal.,2009b;Zachariahetal.,2006),andevidenceindicates thatundernutritionalsoaffectsARToutcomesbycompromisingviralsuppressionand immunologicresponse(J.R.Koetheetal.,2010a;J.R.Koetheetal.,2010b).Interventionsinthis contexttendtoeitheraimtodirectlyraisethecaloricintakeofunderweightpeopleonART(i.e. therapeuticfeedingapproach)(Bahwereetal.,2009;MNdekhaetal.,2009a;MJNdekhaetal., 2009b;vanOosterhoutetal.,2010),ortoaddressfoodsecurityatthehouseholdlevel(i.e. traditionalfoodassistanceapproach)withanimplicitfocusonalleviatingundernutrition(Byronet al.,2008;Iversetal.,2010).Resultsfromthisbodyofinterventionstudiesprovidepreliminary evidencesupportingthepositiveeffectsofsupplementalfeedingonnutritionalstatusandART outcomesofunderweightPLHIV,particularlyready‐to‐usetherapeuticfeeding(RUTF)(J.Koetheet al.,2009;Tirivayietal.,2011a). 9 Whilethemajorityofstudiesonsupplementalfeedinghavefocusedonreversing malnutritionamongunderweightPLHIV,veryfewstudieshaveaddressedfoodsecurity interventionsamongpeoplereceivingART.Inparticular,thereislittledirectevidenceaboutthe roleoffoodassistancetoaddressfoodinsecurityamongpeoplereceivingARTinsettingswhere overweightandobesitycoincidewithhighlevelsofhouseholdfoodinsecurity.Thisinformationis sorelyneededforARTprogramconsideringnutritionalinterventionsinsuchsettings. Studyaimsandhypotheses Weinvestigatewhetherfoodassistanceplusnutritioneducation1)improvesfoodsecurity morethaneducationalone,and2)affectsnutritionalstatusmorethaneducationalone,with particularinterestinwhetherithasadverseeffectsontheBMIofoverweightorobeseparticipants. Providingfoodassistanceincreasestheamountoffoodavailabletoahouseholdandisthusvery likelytoincreaseaccesstofood–andconsequentlyfoodsecurity–foritsmembers,particularly giventheimportanceoffoodavailabilityfortheindividual(s)livingwithHIVinthehousehold. However,therearevariousreasonswhyfoodassistancemaynotimprovefoodsecurity.First,itis possibletousefoodassistanceinwaysthatdonotnecessarilyimproveimmediatefoodsecurity, suchassellingfoodforextraincometopurchasenon‐foodgoodsorservices,givingfoodawayto family,friends,orcommunitymembers(whichmayneverthelessimprovelongrunfoodsecurityas partofreciprocityarrangements),ortakinginextradependentstothehousehold.Furthermore, economictheorysuggeststhatfoodassistancemaynotimprovenetfoodsecurityifitsimply “crowdsout”eitherindividuallaborsupplyorotherin‐kindtransfersfromfamilyorfriends (Barrett,2006;Tirivayietal.,2011a).Nevertheless,studiesontheeffectivenessoffoodassistance programsinbothdevelopedanddevelopingcountriesgenerally–butnotuniversally–findsome improvementoffoodsecurityasaresultofaid(Barrett,2002,2006;Mykerezietal.,2010;Ratcliffe etal.,2011;Yenetal.,2008).Weproposeandtestthefollowinghypothesisforourstudy population: 10 H1:Providinghouseholdfoodassistanceplusnutritioneducationwillimprove householdfoodsecurityovertime,comparedtonutritioneducationalone(i.e.the beststandardofcare). Inaddition,providingfoodsupporttopeoplewithdiversebaselinenutritionalstatus (includingunderweightaswellasoverweightandobesePLHIV)mayresultindifferentialeffectson nutritionalstatus,someofthemadverse.Forexample,researchonthefoodstampprograminthe UnitedStates(servinglow‐incomeindividualsandfamiliesinneed)hasraisedconcernsthatin additiontoimprovingfoodsecurity,foodassistancemayalsoleadtoincreasedoverweightand obesity,particularlyforwomen(N.I.Larsonetal.,2011;Wilde,2007).Adversehealtheffects associatedwithoverweightandobesitysuchasmetabolicsyndrome,diabetesorcardiovascular diseasecouldbeparticularlyundesirableforPLHIVreceivingART,eveniffoodsecurityimproves. Thisisbecausetheyareparticularlyvulnerabletometabolicabnormalitiesandcentralfat accumulation(Alvarezetal.,2010;Friis‐Mølleretal.,2003),whichmaycompromiseimmune responsetotreatment(Crum‐Cianfloneetal.,2010).Wetestthefollowinghypothesisforourstudy population: H2:Providinghouseholdfoodassistanceplusnutritioneducationwillincreasebody massindexovertime,includingofoverweightandobeseparticipants,comparedto nutritioneducationalone(i.e.thebeststandardofcare). WhilestudiesofindividualfoodsupplementationformalnourishedpeoplewithHIVsuggest thatfoodassistancecanhelpimproveBMI(Tirivayietal.,2011a),thereareseveralreasonswhy householdfoodassistancemaynothaveaneteffectonindividualfoodconsumptionorBMI. Evidencefrombothdevelopedanddevelopingcountriessuggeststhatfoodassistancemayinstead increasethefoodconsumptionofotherhouseholdmembers,especiallychildren(Quisumbing, 2003;Roseetal.,1998),substitutefornormallypurchasedfoodsandthusfreeupresourcesinthe householdbudgettopurchaseotherfoodsornon‐foodgoods(Reutlingeretal.,1984),orstabilize foodconsumptionovertime(Barrett,2002). 11 Giventherangeofwaysahouseholdmayutilizefoodassistancedescribedinthissection, weexpectthattakingintoaccountmeasuresofmaterialresources,laborsupply,household composition,andhealthstatuswillbeimportantintestingourhypotheses.Wealsonotethatthe literaturesuggeststhattherelationshipbetweenfoodassistanceandouroutcomesmaydiffer alongkeydemographicdifferences,particularlygender. METHODS BackgroundofResearchCollaboration ThisstudyinvolvedpartnershipamongtheUNWFPRegionalOfficeforLatinAmericaand theCaribbean,theWFPCountryOfficeforHonduras,andtheRANDCorporation,anonprofit researchorganizationbasedintheUnitedStates.In2008‐2009,RANDandWFPbegan implementingjointactivitiesinHondurasbyconductingformativeresearchonthedietaryhabits andnutritionalstatusofpeoplelivingwithHIVreceivingART.Thedatafromthisphaseofthestudy wasusedtodesigncontextandneeds‐specificnutritioneducationmethodologiesforuseinthe pilotfoodassistanceinterventionsforadultswithHIVinHondurasduring2010. Studydesignandsample ThispaperusesdatafromalargerRAND/WFPpilotinterventionstudydesignedtoassess theeffectoffoodassistanceplusnutritioneducationonARTadherenceandotherhealthand nutrition‐relatedoutcomesofpeoplewithHIVreceivingARTinHonduras,comparedtonutrition educationalone(resultsonadherencefromthelargerstudywillbepublishedseparately).Atthe timeofthestudy,nutritionalassessmentandeducationwererecommendedasthe‘bestpractice’to provideadequatemacroandmicronutrientintakeforPLHIVaccordingtointernationalguidelines (WorldBank,2007;WorldHealthOrganization,2004),arecommendationadoptedbythe Honduras’NationalAIDSPlan(Martinetal.,2011a)butnotyetofferedatallHIVhealthcare 12 providersinthecountry,includingourstudysites.Asapilotstudy,itwasthusconsideredtobean ethicalandpracticalimperativetoprovidenutritioneducationtoallparticipantsinthestudy, ratherthanuseacontrolgroupwithnonutritionalintervention.Thisdoesnotprecludeusfrom drawingconclusionsabouttheeffectivenessoftheintervention,however.Rather,weassessthe addedeffectoffoodassistanceaboveandbeyondnutritionaleducation. TheinterventionwasbasedinfourHIVcarecenters(CentrosdeAtenciónIntegral,orCAI), thatwerematchedonsizeoftheHIVpopulationandlocation(tominimizedifferencesinaccessto foodandsocio‐economicdifferences),selectingtwolargehospitalsinthecapitalcityTegucigalpa andtwosmallerhospitalsincitieslocatedintheCaribbeancoastregion.TheCAIsarerunbythe MinistryofHealthundertheNationalAIDSProgram,whichparticipatedcloselyinthestudy.The studyhiredfourprofessionalnutritionists–oneforeachsite–andacoordinatorbasedatWFP.The nutritionistscarriedoutrecruitmentintothestudy,conductednutritioneducation,assistedwith distributionoffoodassistance,andcarriedoutallstudyassessments. Assignmenttothefoodassistancestudygroupwasattheclinicratherthanpatientlevel. Giventhegeneralizedfoodinsecurityinthestudyregionsandthesmallsizeoftheparticipating HIVclinics,itwasconsideredunethicaltorandomlyofferfoodtosomeindividualsandwithholdit fromotherswhoqualifiedwithinthesamehospital.Instead,werandomizedoneofthetwo matchedhospitalswithinthesameregiontothefoodassistanceplusnutritioneducationgroup usingacointosstominimizeselectionbias(e.g.clinicself‐selectionintofoodassistanceor investigatorassignmentbasedonperceivedneed).Thecointosswasattendedbymembersofthe NationalAIDSProgram,representativesoftheparticipatingclinics,andrepresentativesofthe AssociationofPeopleLivingwithHIV/AIDSinHonduras(ASONAPSIDAH).Attheconclusionofthe study,theclinicsassignedtonutritioneducation‐onlythenreceivedthefoodassistance. 13 Onceclinicswereassignedtostudygroups,participantsattendingoneofthefourCAIwere recruitedconsecutivelyintothestudybetweenDecember2009andOctober2010,duringtheir regularclinicvisit.Inclusioncriteriawerebeingalocalresidentofthecommunityformorethan oneyear,18yearsoldorabove,receivingART,havingundernutrition(definedasbeing underweight,i.e.havingBMI≤18.5)and/orhouseholdfoodinsecurity,and,ifreceivingARTforat least6months,indicationsofsuboptimaladherence(i.e.missedclinicappointments,delayed pickingupmedications,orreportedstoppingtakingpills).Exclusioncriteriaincludedbeingunable tospeakandunderstandSpanish,orhavingplanstomoveinthenextyear.Inaddition,pregnant womenwereexcludedfromthedatacollectionportionoftheintervention,toavoidthe confoundinginfluenceofpregnancyonchangeonnutritionaloutcomes;however,theystill receivedtheprograminterventionsiftheymettheinclusioncriteria. Participantsinthefoodbasketgroupreceivedasupplementaryfoodration,whichthe participantwasresponsibleforpickingupeverymonthatafixeddateandtimefromtheclinicor othercommunitylocation.ThecontentsofthefoodassistancefollowedWFP’spoliciesandincluded 1000gramsofmaize,240gramsofrice,370gramsofbeans,500gramsoffortifiedcorn‐soyblend (CSB),and90gramsofvegetableoilperpersonperday,standardizedforahouseholdoffivepeople for30days.Providingahouseholdfoodbasketratherthananindividualfoodrationwasintended toavoiddiversionofthefoodmeantfortherecipienttootherusessuchassharingwiththefamily. Thefooddistributionprocesswasmanagedbythestudynutritionistswithlogisticalsupportfrom WFPandtheparticipationoftheAssociationofPeopleLivingwithHIV/AIDSinHonduras (ASONAPSIDAH).Familymemberswerepermittedtopickupthefoodrationinlieuofthe participantifneeded. Inordertoassureproperuseofthesefoods,aswellastoimprovetheoverallqualityofthe diet,anutritioneducationcomponentwasdevelopedbasedoncomprehensivereviewof 14 nutritionalguidelinesforPLHIVpublishedaroundtheworld,andadaptedtothelocalcontextbased onformativeresearchconductedbyRANDandtheWFPfromMarchtoOctober2009onthemacro andmicronutrientintake,foodconsumptionhabitsandnutritionalstatusofthetargetpopulation, aswellasculturalacceptability,andlocalfoodavailability.Nutritionaleducationconsistedof monthly20‐minuteone‐on‐onenutritionalcounselingsessionsbasedontheparticipant’sschedule, andmonthly1‐hourgroupsessionsatafixedtime.Nutritionalcounselingconsistedofthe nutritionistsdeliveringnutritionmessagesusingcolorful,graphicmaterials(developedspecifically forthelocalcontext),reinforcedbyverificationquestionsandtake‐homepamphlets.Thegroup sessions,alsoledbythenutritionists,werehighlyparticipatory,basedoninteractiveactivitiesand games,andsometimesincludedcookingactivitiesordemonstrations.Allnutritioneducation activitieswereaccessibletoparticipantswithlowliteracy. Follow‐upassessmentsconsistedofmonthlyappointmentswiththenutritionistsfor12 months.Everymonth,thenutritionistwouldconductthenutritionalcounselingsession,take anthropometricmeasures(height,weight,bodyfat,waistcircumference,mid‐upper‐arm circumference),andassessdietaryintake(foodfrequencyand24‐hourdietaryrecall).Atbaseline, 6‐monthsand12‐months,thenutritionistwouldadministeramorecompletequestionnaire includinginformationonhouseholdcomposition,socio‐economicstatus,nutritionalknowledge, foodsecurity,mentalhealth,stigma,HIVknowledge,andARTadherenceself‐efficacy.Participants wereprovidedwithamonetaryincentivetocovertransportationcostsandinrecognitionoftheir participation,equivalentto~$5USDinlocalcurrencypaidatbaseline,6and12months(~$15 total). ThestudywasapprovedbyRAND’sHumanSubjectsProtectionCommittee,aswellas Honduras’NationalBioethicsCommittee,basedoutoftheNationalAutonomousUniversityof 15 Honduras.Writtenconsentwasobtainedfromallparticipants.ASONAPSIDAHcollaboratedinall aspectsoftheprograminclosecollaborationwiththestudynutritionists. Measures Dependentvariables Foodinsecurity:FoodinsecuritywasassessedusingtheLatinAmericanandCaribbeanFood SecurityScale(ELCSA)avalidated15‐itemscaledevelopedspecificallytoassessfoodinsecurityin theLatinAmericanandCaribbeanregions(Melgar‐Quiñonezetal.,2010).Thescalecaptures experiencesofhouseholdfoodsecurityoverthelast3months,includingfoodquantityand sufficiency(e.g.“Inthelast3months,wasthereeveratimethatyouoranotheradultinyour householddidn’teatbreakfast,lunchordinnerbecausetherewasn’tenoughmoney?”),foodquality andsafety(e.g.“Inthelast3months,wasthereeveratimetherewasn’tenoughmoneyforasafe, variedandnutritiousdiet?”,andanxietyaboutfoodsupplies(e.g.“Inthelast3months,wasthere everatimethatyouworriedthatfoodwouldrunoutbecausetherewasn’tenoughmoney?”).The scaledifferentiatesbetweenhouseholdswithandwithoutchildren,wherethefirst8questionsare askedtoallparticipants,andanadditional7questionsareaskedtoparticipantswithchildren.All questionsreceive“yes”or“no”answers.Rawscoreswerethentabulatedasthesumofaffirmative answers,withhigherscoresindicatinghigherlevelsoffoodinsecurity.Classificationoffood insecuritywasbasedonvaluesofrawscores:foodsecurity(0,allhouseholds(HH)),lightfood insecurity(1‐3,HHw/ochildren;1‐5HHw/children),moderatefoodinsecurity(4‐6,HHw/o children;6‐10HHw/children),andseverefoodinsecurity(7‐8,HHw/ochildren;11‐15,HHw/ children).Therefore,higherscoresindicatehigherfoodinsecurity,andlowerscoresindicatelower foodinsecurity(orbetterfoodsecurity).Tocreateacontinuousfoodinsecurityscoreforall participants(0‐15),scoresofparticipantswithoutchildrenwerestandardizedtothe15‐point scoringsystem. 16 Bodymassindex(BMI):WefocusonBMIinthispaperasthemostbasic,acceptedapproachto assessingnutritionalstatusacrossadultindividuals(Gibson,2005).Weightandheight measurementsweretakenbyprofessionalnutritionists,whowerepreviouslytrainedand standardizedaccordingtoacceptedmethods(Habicht,1974).Weightandheightmeasurements wereusedtoderivebodymassindex(BMI)usingtheequationweight(kg)/height(m)2.Body weight(kg)wasmeasuredonadigital,calibratedscalewithaprecisionof100g,whilethe participantworeaclinicalrobeandnoshoes.Aslidingmetallicmeasuringtapewithaprecisionof 0.1cmwasusedtomeasureheight(cm),withtheparticipantstandingerectwithoutshoesnexttoa verticalwall.BMIwasusedtoclassifythenutritionalstatusofparticipantsaccordingto internationalstandarddefinitions:underweight(BMI<18.5),normal(18.5≥BMI>25),overweight (25≥BMI>30),andobese(BMI≥30)(WorldHealthOrganization).Abinaryvariableequalto1if theparticipantwasoverweightorobese,and0otherwise,wasconstructedforuseinanalysis.In thispaper,werefertothecombinedoverweight/obesegroupas“OW”.AlthoughwefocusonBMI asthenutritionalstatusoutcomeinthispaper,additionalanthropometricmeasureswerealso collectedinastandardizedmannerandanalyzedinsensitivityanalyses,includingbodyfatpercent usingbioelectricalimpedance,andbodycircumferences(waistandmid‐upper‐arm). Keycovariates HIV‐relatedhealth:HIV‐relatedhealthiscloselyrelatedtoBMI,particularlyunderweightstatus, andhasalsobeentiedtofoodinsecurityinresource‐limitedsettings(Liuetal.,2011;Mohetal., 2007;Wangetal.,2011;Weiseretal.,2009b;Zachariahetal.,2006).HIV‐relatedhealthwas assessedusingdataabstractedfromclinicrecords,includingthemostrecentCD4count(cells/µL), dateofHIVdiagnosisandARTinitiation,andabinaryvariableindicatingwhetherthepersonwas symptomatic(i.e.presenceofopportunisticinfectionsand/orAIDSdiagnosis).Medicationrecords wereabstractedtoidentifypatientstakingproteaseinhibitorsaspartoftheirARVregimen,which 17 havebeenlinkedtoincreasedcentralweightgaininsomestudies(Friis‐Mølleretal.,2003).The amountoftimereceivingARTwascalculatedbysubtractingthedateofARTinitiationfromthedate ofbaselineinterview(alldateswerecodedasthenumberofdayssinceJanuary1,1960),anda binaryvariableforbeingintheearlystagesofARTwasconstructed,using<100daysasacutoffto capturetheperiodafterinitiatingtreatmentwhenhealthresponsetoARTislargest(andissmallto nonexistentthereafter)(Thirumurthyetal.,2008b;Wools‐Kaloustianetal.,2006). Socio‐economiccharacteristics:Socio‐economicstatusiscloselytied–althoughnotsynonymous with–theconceptoffoodsecurity(Maxwell,1996;Maxwelletal.,1992),andmayalsoaffectBMI throughaffectingresourcesavailableforfoodandhealthcare(Campbell,1991;Sauerbornetal., 1996).Weusetwomeasurestoapproximatechangesinsocio‐economicsituation.Currentwork statuswasabinaryvariabledefinedashavingworkedinthelastmonth,basedonself‐report. Materialsupportwasabinaryvariableindicatingiftheparticipantwascurrentlyreceiving economicsupportfromarelative,friendorothersource(notinstitutional),basedonself‐report.In additionwemeasureeducationasabinaryvariableindicatingwhethertheparticipanthas completedatleastprimaryschool. Demographiccharacteristics:Weassesseddemographiccharacteristics(gender,race/ethnicity, age,householdcomposition)inordertofacilitategroupcomparisonsoncharacteristicsthatmay affectfoodinsecurityand/orBMI(Anemaetal.,2009). Analysis Analyseswerebasedoncomparisonsoffoodinsecurityandnutritionalstatusacrossthe studygroupsatbaselineandovertime.Wefirstusedbivariatestatistics(Chi‐squaretest,two samplet‐test)tocomparethebaselinecharacteristicsofthefoodassistanceplusnutrition educationgrouptotheeducation‐onlygroup.Wealsoconductedcomparisonsacrossgender.To 18 examinechangeovertime,weexploredtrendsintheoutcomevariablesbytestingforstatistically significantdifferencesfrom0to6and12monthsbyinterventiongroup(pairedt‐test). Wethenconductedmultivariatelongitudinallinearregressiontoinvestigatetheeffectof foodassistanceplusnutritionaleducationonfoodinsecurityandBMI,comparedtonutrition educationalone.Weidentifytheresponsetofoodassistancebyexaminingchangesinfood insecurityandnutritionalstatusbetweeninterviewassessmentsacrossstudygroups.Ourkey identifyingassumptionisthatdatafromtheeducation‐onlygroupcanbeusedtocontrolfortrends inthefoodassistancegroupduetonutritioneducationand/orsecularfactorssuchaschangesin theeconomyorclimate.Sincebothgroupsreceivednutritioneducation,theeffectweidentifyisthe addedeffectoffoodassistanceaboveandbeyondnutritionaleducation. Anexaminationofbivariatestatisticscomparingtheinterventiongroupsatbaselinereveal thattheydifferonobservablecharacteristicslikelytoaffectfoodinsecurityandBMIlevels,asmight beexpectedintheabsenceofindividualrandomization.Ofequalormoreconcern,however,maybe thatparticipantscouldalsodifferalongunobservablecharacteristics,suchaspreferencesand abilities.Onestrategyfordealingwithselectiononunobservablesinidentifyingcausaleffectin observationaldataisto“differenceout”timeinvariantcharacteristicsandtoincludeonly covariatesthatchangeovertimewhichwebelievetoaffectouroutcomes. Toimplementthisapproach,weusedanindividualfixedeffectsmodel,whichmeasuresthe averagechangebetweenassessmentsinouroutcomesasafunctionofthechangeinour explanatoryvariables.Thisapproachcausessomelossinefficiencycomparedtomodelswith individualrandomeffects,butismoreconservativebecauseitallowstheindividual‐specifictime‐ invarianteffectstobecorrelatedwiththeregressors.Keyassumptionsofthefixedeffectsmodel includethatalltime‐varyingfactorsaffectingtherelationshipbetweentheinterventionandthe 19 outcomeareincludedascovariates,andthattime‐invariantfactorsmayaffectthelevelbutnot changeintheoutcome.WeestimateEquation1forthefoodinsecurityoutcome: (1) FI it   i   1 ( MONTH 6 t )   2 ( FAi * MONTH 6 t )   3 ( MONTH 12 t )   4 ( FAi * MONTH 12 t )  X it    1   INTMONTH t   it , 12 where, FI it isthecontinuousfoodinsecurityscoreforindividualiinintimet(interview roundsatbaseline,6and12months),  i isafixedeffectforindividualithatcapturestheeffectsof time‐invariantvariablessuchasdemographicsandeducation,aswellasunobservablessuchas preferencesandabilities, MONTH 6 t and MONTH12 t indicatetheinterviewassessmentthatthe observationisfrom(withthebaselineassessmentastheomittedindicator), FAi isanindicator variableequaltooneifindividualiisafoodassistancerecipient, X it  isavectoroftime‐varying covariates,and INTMONTH t consistsof12monthofinterviewindicatorstocontrolformonthly fluctuationsinfoodavailabilityorpricesinthecommunity(withonemonthomitted).Ourprimary explanatoryvariablesofinterestwerethebinaryindicatorsrepresentingthe6and12month assessments(wherethebaselineassessmentwastheomittedvariable),andthecrossproductterm interactingthefoodassistancegroupbyeachtimeindicator.In X it  ,wecontrolledfortime‐ varyingcovariates(a)whosechangewesuspectedwouldbeassociatedwithchangeinour outcomes,basedontheliterature,and(b)thatdifferedbetweentheinterventiongroupsatbaseline tocontrolforthesedifferences.Forthefoodinsecurityoutcome,theseincludedbeingHIV symptomatic(inlieuofCD4count,whichwasonlyavailableatbaselineatthetimeofanalysis), householdsize,havingworkedinthelastmonth,andreceivingmaterialsupportfromfriendsor family(Anemaetal.,2009;Bukusubaetal.,2007;Tsaietal.,2011). 20 TheprimarydependentvariableforthenutritionalstatusregressionwasBMI(kg/m2).For thisregression,weaddedseveraltermstotheindividualfixedeffectsmodelinEquation1to capturehowbeingoverweightorobeseatbaselinemayhavemodifiedtheeffectoffoodassistance overtime.ForBMI,weestimate: BMI it   i   1 ( MONTH 6 t )   2 (OWi * MONTH 6 t )   3 ( FAi * MONTH 6 t ) (2)   4 (OWi * FAi * MONTH 6 t )   5 ( MONTH 12 t )   6 (OWi * MONTH 12 t )   7 ( FAi * MONTH 12 t )   8 (OWi * FAi * MONTH 12 t )  X it    1   INTMONTH t   it , 12 where OWi representswhetherindividualiwasoverweightorobeseatbaseline.Inadditiontothe keyexplanatoryvariablesnotedinthefoodinsecurityregression,theinteractionsofthetime indicatorswithOWstatusatbaseline,andthetripleinteractionsforbeingOWatbaselinewithboth thefoodassistancegroupandthetimeindicators,werealsoofprimeinterest. Covariatesin X it  wereequivalenttothefoodinsecurityregression,butalsoincludedfoodinsecurityscore.1 Allanalysesincludedattritionweightstoaccountfordropoutfromthestudy,whichwere derivedvialogisticregressionusingcompletionstatusastheoutcomeandbaselinemeasures associatedcompletionstatusandassignmenttothefoodassistancestudygroupastheindependent variables.AllstatisticalanalyseswereconductedinSTATA/IC11.1(StataCorp:CollegeStation, Texas). Sensitivityanalysis Weconductedseveralsensitivityanalysestotesttherobustnessofourresults,particularly togroupdifferencesatbaselinethatwebelievedmightmodifytheeffectoftheintervention.First, weomittedpeoplewhowereintheearlystagesofreceivingART(<~3months)atbaseline(n= 1 Inaddition,time‐invariantvariableswhichwereinteractedwiththebinarytimevariableswereincludedas stand‐alonecovariates(interventiongroupstatusandbeingOWatbaseline),knowingthatbydesignthey wouldfalloutoftheregression. 21 36),sincetheinitialmonthsonARTtendtobeaccompaniedbydramatichealthimprovements, whichmayaffectbothfoodinsecurityandanthropometricoutcomes.Second,weomitted householdswithoutchildrenatbaseline(n=73),sincefood‐relateddecision‐makingand distributionmaybefundamentallydifferentinhouseholdswithandwithoutchildren.Third,forthe BMIregressiononly,weomittedpeoplewhoseARVschemesincludedproteaseinhibitorsat baseline,whichsomestudieshavefoundtobeassociatedwithcentralweightgain(Friis‐Mølleret al.,2003). Inadditiontorestrictingthepopulationinvariousways,weexploredtwoalternate empiricalspecifications1)apopulation‐averaged(PA)modelusingthegeneralizedestimating equations(GEE)approachtoanalysisofrepeatedmeasurementdata,and2)anindividualrandom‐ effects(RE)model.Bothofthesealternatepanelmodelsallowforpossibilitythattimeinvariant characteristics(e.g.studysite,education,etc.)andbaselinecharacteristics(e.g.baselinefood security,baselineCD4,etc.)wereassociatedwithhowouroutcomeschangedovertimeinresponse totheinterventions.FortheregressiononBMI,wewereparticularlyconcernedthatheterogeneity inthestyleandeffectivenessofthenutritionistprovidingtheeducationcomponentateachsite couldaffectnutritionalstatus.Fortheregressiononfoodinsecurity,wewereconcernedthatthe significantdifferencesinbaselinefoodinsecuritybetweenthestudygroupsmayaffecttheirchange overtime.Inthealternatemodels,weincludedthesamecovariatesastheindividualfixedeffects model,butinadditioncontrolledforstudysite,gender,race/ethnicity,educationstatus,baseline versionsoftheoutcome,andbaselineCD4,andmodeledtimeasanordinalvariablerepresenting thethreeassessments.Finally,wealsoexploredhowanalyzingthebinaryvariablefor“severefood insecurity”wouldperformasanalternateoutcomeincomparisontofoodinsecurityscore,using thepopulation‐averagedregressionmodel. 22 RESULTS Samplecharacteristics Thesampleconsistedof400participants,including203receivingfoodassistanceand nutritioneducationand197receivingeducation‐only.Eighty‐eightpercentofthefoodassistance groupand76%oftheeducation‐onlygroupcompletedthe12‐monthassessment.Thosewhowere HIVsymptomaticatbaselinewerelesslikelytocompletethestudy(regardlessofintervention group),whileparticipantsinthefoodbasketgroupweremorelikelytocompletethestudy. Baselinecharacteristicsofthetotalsamplebyinterventiongroupandgenderaregivenin Table1.AveragetimesinceHIVdiagnosisandARTinitiationwas5.3and3.7years,respectively, with9%ofparticipantsreceivingARTforlessthan100days,and7%takingproteaseinhibitorsas partoftheirARVscheme.Participantsinthefoodassistanceinterventiongroupweremorelikelyto befemaleandhavechilddependentsinthehousehold,butlesslikelytoself‐identifyas afrodescendent,havecompletedprimaryschool,haveworkedinthelastmonth,andbereceiving economicsupportfromfamilyorfriends.ThefoodassistancegrouphadhigheraverageCD4counts atbaseline,indicatingbetterimmunehealth,butalsohadhigherprobabilityofbeingsymptomatic. Womenweremorelikelytohavechilddependentsinthehousehold,lesslikelytobe workingandlesslikelytobereceivingeconomicsupportfromfamilyorfriendscomparedtomen. Onaverage,womenhadbeenreceivingARTforlongerthanmen,despitesimilarmeantimesince HIVdiagnosis. 23 Table1:Demographic,healthandsocio‐economiccharacteristicsatbaseline Intervention Group Gender Food assistance + Nutrition education Nutrition education only Men 74% *** 62% *** ‐‐ ‐‐ 69% Afrodescendent 4% *** 19% *** 8%* 13%* 12% Primary school or more 49% ** 58% ** 53% 56% 54% Age in years [SD] 40 [0.70] 41 [0.68] 41 [0.59]* 40 [0.89]* HH w/ children < 18 y.o 87%*** 77%*** 67%*** 88%*** 82% HH size (incl. participant) 5.1 [2.5] 4.7 [2.5] 4.3 [2.4] 5.1 [2.5] 4.9 [2.5] 317 [13.04]** 274 [13.7] ** 228 [15.4]*** 329 [11.4]*** 297 [9.5] Years since HIV diagnosis [SD] 5.1 [0.28] 5.4 [0.29] 5.2 [0.23] 5.3 [0.42] 5.3 [0.20] Years receiving ART [SD] 3.6 [0.18] 3.8 [0.19] 3.3 [0.15]** 3.9 [0.23]** 3.7 [0.13] Receiving ART < 100 days 7% 11% 15% ** 6% ** 9% Takes protease inhibitors 8% 6% 5% 8% 7% 12% ** 6% ** 10% 9% 9% Worked in last month 33% ** 44% ** 45% ** 35% ** 38% Receives material support 28% * 35% * 39% ** 28% ** 32% 203 197 123 277 400 Demographics Female HIV‐related health CD4 count (cells/µl) [SD] Currently symptomatic Socio‐economic status Number of observations Women All *** p < 0.01; ** p < 0.05; * p < 0.01 24 40 [0.49] Baselinefoodinsecurityandnutritionalstatus Overall,therewasahighprevalenceofseverefoodinsecurityamongstudyparticipants [65%],withanaveragefoodinsecurityscoreof11.4(outof15possiblepoints,with15being indicatingthehighestleveloffoodinsecurity)(Table2).Examiningnutritionalstatus,meanBMI was24m/kg2,with58%ofparticipantsclassifiedinthenormalrange,while31%totalwereeither overweight[23%]orobese[8%].Elevenpercentofparticipantswereunderweight.Whilewefound significantprevalenceofbothfoodinsecurityandOWinthestudypopulationatbaseline,Figure1 demonstratesthatthesetraitsoverlappedformanyindividualsinourstudy.Lookingattheupper rightareaofthescatterplotplottingthefoodinsecurityscoreagainstBMI,weseethatasignificant amountofsevereandmoderatefoodinsecurityexistedamongOWparticipantsatbaseline. Meanwhile,lookingattheupperleftquadrantofthescatterplot,weseethathigherfoodinsecurity wasnotvisiblyconcentratedamongpeoplewithlowerBMI,althoughthereisaslightand statisticallysignificantinversecorrelationbetweenfoodinsecurityscoreandBMI[corr=‐0.12;p< 0.01]. Higherprevalenceofseverefoodinsecurityatbaselinewasfoundamongthefood assistancegroup[72%]comparedtotheeducation‐onlygroup[58%;p<0.01],andtherewerealso significantdifferencesinmeanfoodinsecurityscore.However,comparisonsbystudygroupdidnot revealsalientdifferencesinBMI. Comparingmenandwomen,therewerenosignificantbaselinedifferencesinthe prevalenceofseverefoodinsecurityorthemeanfoodinsecurityscorebygender.However,women weresignificantlymorelikelythanmentobeoverweightorobeseandhavelargerwaist circumferences,despitehavingloweroverallweightcomparedtomen. 25 Table2:Nutritioncharacteristicsatbaseline Intervention Group Nutrition education only Food assistance + Nutrition education Household food insecurity 1 Men Women All Mean standard. score [SD] 12.0 [3.3]*** 10.7 [4.4]*** 11.2 [4.1] 11.5 [3.8] Severe insecurity 72% *** 58% *** 61% 67% 65% Moderate insecurity 22% 26% 28% 23% 24% Light insecurity 5% *** 15% *** 9% 10% 10% No insecurity 0% * 1% * 2% 0% 1% Mean BMI [SD] 23 [4.3] 24 [4.7] 22 [3.3] *** 24 [4.8]*** Underweight 12% 9% 15% ** 9% ** 11% Normal 58% 58% 70% *** 52% *** 58% Overweight 23% 24% 13% *** 28% *** 23% Obese 7% 9% 2% *** 11% *** Weight (kg) [SD] 57.0 [11.4] *** 60.0 [12.8] *** 61.0 [9.5]*** 57.2 [13.1]*** 8% 58.4 [12.2] Number of observations 203 197 123 277 400 Body mass index (BMI) Gender 11.4 [3.9] 24 [4.5] *** p < 0.01; ** p < 0.05; * p < 0.01 1 Range 0 – 15 26 15 Figure1:CorrelationbetweenfoodinsecurityandBMIatbaseline 10 Moderate food insecurity 5 Food insecurity score Severe food insecurity 0 Light or no food insecurity Underweight 10 Overweight and obese 20 30 BMI (kg/m2) Note:Correlation=‐0.12(p<0.01) 27 40 50 Longitudinalresultsonfoodinsecurity Figure2presentsunadjustedtrendsinfoodinsecurityscoreoverthreeassessments(0,6 and12months)byinterventiongroup.Foodinsecurityscoredecreasedforbothgroupsoverthe12 monthsofthestudy,withthemostdramaticimprovementoccurringbetweenbaselineand6 months.Inthefoodassistancepluseducationgroup,themeanfoodinsecurityscoredecreased significantlyfrombaseline[mean=12.0]toMonth6[mean=6.6;p<0.01],andincreasedslightly atMonth12whileremainingsignificantlybelowbaselinelevels[mean=7.4,p<0.05].Theoverall trendfrom0to12monthsforthefoodassistancegroupwassignificant[p<0.01].Inthe education‐onlygroup,thefoodinsecurityscoredecreasedbyasmalleramountfrombaseline [mean=10.4]toMonth6[mean=8.6;p<0.01],andalsoincreasedslightlyatMonth12while remainingsignificantlybelowbaselinelevels[mean=9.3,p<0.05].Theoveralltrendfrom0to12 monthsforthenutritioneducationgroupwassignificant[p<0.01].Differencesinfoodinsecurity scorebetweenstudygroupsweresignificantateachassessment[p<0.01].Notably,whilethefood assistancepluseducationgroupbeganthestudywithhigherfoodinsecuritythantheeducation‐ onlygroup,theyendedthestudywithlowerlevelsoffoodinsecurity. 28 6 Food insecurity score 8 10 12 Figure2:Trendinfoodinsecurityscoreover12monthfollow‐up,bystudygroup 0 6 Month Nutritional education only 95% CI 12 Food assistance + education Note: Lower scores represent reduced food insecurity (i.e. improvement) 29 Totesttheunadjustedtrendsweobservedabove,weusedlinearregressionwithindividual fixedeffectstoestimatetheeffectoffoodassistanceplusnutritioneducationonfoodinsecurity score,controllingforarangeofcovariates.Thismodeluseswithin‐subjectschangetoestimate effects,andthereforeallcoefficientsrelatechangeintheexplanatoryvariabletochangeinfood insecurity.Itaccountsfortime‐invariantheterogeneityacrossindividualsthatcouldleadtoself‐ selection.Inthemultivariatelongitudinalregressionmodeloffoodinsecurity(Table3),wewere primarilyinterestedinthecoefficientsonthetimevariables(‘Montht’),andtheirinteractionwith thevariableforfoodassistance(‘MonthtXFA’).Thecoefficientson‘Month6’[b=‐1.891;p<0.01] and‘Month6XFA’[b=‐3.150;p<0.01],werebothnegativeandhighlysignificant,indicatingthat thefoodinsecurityscoredecreasedforbothstudygroupsbyMonth6,withthefoodassistance groupexperiencingalargerdecreasecomparedtotheeducation‐onlygroup.Movingtothenext assessment,thecoefficientsfor‘Month12’[b=‐1.750;p<0.01]andtheinteractionof‘Month12X FA’[b=‐2.740;p<0.01]remainednegativeandhighlysignificant,withnofurtherdecreaseinthe foodinsecurityscorecomparedtoMonth6. ThecoefficientspresentedabovecanbeinterpretedtomeanthatbyMonth12,thefood insecurityscoredecreasedbyanaverageof1.750pointsfortheeducation‐onlygroup,whilethe foodassistancepluseducationgroupexperiencedanadditionaldecreaseof2.740points,foratotal improvementof4.49points.Thesizeoftheeffectfortheadditionaldecreaseof2.740(i.e.theeffect offoodassistance)isslightlylessthanonestandarddeviationofthefoodinsecurityscoreforthe foodassistancegroupatbaseline,implyingthatby12monthsalmost68%ofthefoodassistance grouphadshiftedbelowthemeanbaselinescoreduetofoodassistance.Theresultsdidnotchange significantlyafterconductingtwosensitivityanalyses,thefirstthatdroppedpeopleinearlystages ofreceivingART,andthesecondthatdroppedhouseholdswithoutchildren(seeAppendix,Table A1).Inaddition,changingthespecificationofthemodeltoaccountfortime‐invariantand/or baselinedifferences,includingbaselinefoodinsecurityscore,didnotchangethedirectionor 30 significanceofeffects(seeAppendix,TableA3).Finally,resultsfromthepopulation‐averaged logisticregressionusingthebinaryoutcomeforseverefoodinsecuritymirrorthedirectionand significanceoftheresultsonfoodinsecurityscore(SeeAppendix,TableA3). Examiningtheregressionsonfoodinsecuritybygender(Table3)indicatethatbyMonth 12,bothwomenandmenhadexperiencedasignificantdecreaseinfoodinsecurity,inboth interventiongroups.Thecoefficienton‘Month12XFA’suggeststhatwomen[b=‐3.244;p<0.01] experiencedgreaterimprovementinhouseholdfoodsecuritycomparedtothefullpopulation regression[b=‐2.740;p<0.01].Inaddition,foodinsecurityscoredecreasedforwomenmost stronglyinthefirstsixmonths[‘Month6XFA’,b=‐4.368;p<0.001],similartothefullpopulation regression.Meanwhile,menalsoexperiencedanimprovement[b=‐2.036;p<0.10]duetofood assistancebytheendofthestudy,althoughthiseffectwasonlymarginallysignificant.Wedidnot detectasignificanteffectoffoodassistanceformenat6months.Poweranalysisofthesplitgender samplerevealedthatwehadhighpower(>80%)todetectchangesinfoodinsecurityscoreatthe 5%levelforwomen.However,wehadverylowpowertodetectchangesinfoodinsecurityscore formenacrossinterventiongroupsatthe5%levelandthusdonotmakecomparisonsonpoint estimatesacrossthetwogendergroups. 31 Table3:Longitudinallinearregressionresultsonfoodinsecurityscore,bygender Food insecurity score Month 6 Month 6 X Food assistance Month 12 Month 12 X Food assistance HIV symptomatic Household size Worked in the last month Material support from family or friends Constant Observations Number of person ID All Women Men -1.891*** (0.473) -3.150*** (0.606) -1.750*** (0.433) -2.740*** (0.581) -0.290 (0.633) 0.267* (0.136) -0.769** (0.352) 0.245 (0.391) 9.596*** (0.952) -0.570 (0.514) -4.368*** (0.658) -0.969* (0.504) -3.244*** (0.652) -0.535 (0.717) 0.379** (0.160) -1.178*** (0.399) 0.249 (0.400) 9.726*** (1.111) -4.569*** (0.974) -0.488 (1.236) -3.077*** (0.826) -2.036* (1.224) 0.115 (1.147) 0.166 (0.224) 0.357 (0.699) -0.037 (1.092) 8.475*** (1.642) 971 382 692 267 279 115 *** p<0.01, ** p<0.05, * p<0.1 Notes: a) Regressions include individual fixed effects and month-of-interview indicator variables. Robust standard errors in parentheses 32 Longitudinalresultsonnutritionalstatus UnadjustedtrendsinBMIwereexaminedafterstratifyingonnutritionalstatus classificationatbaseline,combiningoverweightandobeseparticipantsintoonegroup(designated as“OW”)(Figure3).ParticipantswhowereunderweightorinthenormalBMIrangeatbaseline increasedtheiraverageBMIoverthefirst6monthsofthestudy,withbothinterventiongroups experiencingincreasesofsimilarmagnitude(allincreasessignificantatp<0.01).Nofurther significantchangesinBMIoccurredatMonth12forthesegroups(Figures4aand4b).Theoverall increasefrom0to12monthswassignificantforbothunderweightandnormalBMIrangegroups [p<0.01].ThosewhowereOWatbaselinehadslightchangesinBMIthatdifferedbyintervention group(Figure3c).IntheOWfoodassistancegroup,BMIincreasedslightlyfrombaseline[mean= 28.5]toMonth6[mean=29.0;p<0.01],andthendecreasedslightlyagainatMonth12[mean= 28.7,p<0.10].However,whiletherewasslightmovementbetweenbaselineandMonth6,and Month6andMonth12,therewasnostatisticallysignificantdifferenceinBMIbetweenbaselineand Month12.IntheOWeducation‐onlygroup,BMIdecreasedbyfrombaseline[mean=28.9]toMonth 6[mean=28.6;p<0.05],withnostatisticallysignificantadditionalchangeatMonth12.Theoverall trendindecreasedfrom0to12monthswassignificant[p<0.05]. 33 Figure3:TrendinBMIover12monthfollow‐upbystudygroup,stratifiedbybaseline nutritionalstatus b)Normalweightatbaseline(n=230) 16 21.5 17 BMI (kg/m2) 18 19 BMI (kg/m2) 22 22.5 20 23 21 a) Underweightatbaseline(n=43) 0 6 Month Nutritional education only 95% CI 0 12 6 Month Nutritional education only 95% CI Food assistance + education 28 BMI (kg/m2) 28.5 29 29.5 30 27.5 0 6 Month Nutritional education only 95% CI 12 Food assistance + education 34 Food assistance + education c)Overweightorobeseatbaseline(n=127) 12 Table4showsresultsfromtheindividualfixedeffectsregressionmodelofBMI.Asinthe foodinsecurityregression,wewereprimarilyinterestedinthecoefficientsonthetimevariables (‘Montht’)andtheirinteractionwiththefoodassistancegroup(‘MonthtXFA’).Inaddition,we wereinterestedinthetripleinteractiontermsincludingOWstatusatbaseline(‘MonthtXFAX OW’).Whilethecoefficienton‘Month6’[b=0.649;p<0.01]waspositiveandsignificant,the coefficienton‘Month6XFA’wasnotsignificant,indicatingthatBMIincreasedbythesameaverage amountfornon‐OWparticipantsinbothinterventiongroups.However,both‘Month6XOW’[b=‐ 0.906;p<0.01]and‘Month6XFAXOW’[b=0.690;p<0.10]weresignificant,indicatingthatat Month6,BMIdecreasedintheOWeducation‐onlygroupby‐0.363points(∆BMIMonth6,education,OW= 0.649–0.906=‐0.216),butthatthoseintheOWfoodassistancegroupactuallyexperiencedaslight increaseinBMIof0.474points(∆BMIMonth6,FA,OW=‐0.216+0.690=0.474).However,thiseffectfor theOWfoodassistancegroupat6monthswasonlymarginallysignificantandofverysmall magnitudeinpracticalterms,comparedtothestandarddeviationofBMIatbaseline,whichwas quitelarge[SD=4.5]. ByMonth12,thecoefficientson‘Month12’[b=0.534;p<0.01]and‘Month12XOW’[b=‐ 0.850;p<0.01]werestillsignificant,indicatingthatnon‐OWparticipantsexperiencedanincrease inBMIby0.534points,whiletheOWparticipantsexperiencedadecreaseintheirBMIby0.316 points(∆BMIMonth12,OW=0.534–0.850=‐0.316),regardlessofintervention.Meanwhilethe coefficienton‘Month12XFAXOW’wasnolongersignificant.Theseresultsindicatethatnonet changeinBMIoccurredforthefoodassistancegroupoverthecourseofthestudy,includingthose OWatbaseline. 35 Table4:LongitudinallinearregressionresultsonBMI(kg/m2),bygender BMI Month 6 Month 6 X OW Month 6 X FA Month 6 X FAX OW Month 12 Month 12 X OW Month 12 X FA Month 12 X FA X OW HIV symptomatic Food insecurity score Household size Worked in the last month Material support from family or friends Constant Observations Number of person ID *** p<0.01, ** p<0.05, * p<0.1 All Women Men 0.649*** (0.165) -0.906*** (0.260) 0.027 (0.223) 0.690* (0.394) 0.534*** (0.166) -0.850*** (0.317) 0.079 (0.230) 0.278 (0.482) -0.915*** (0.273) -0.024* (0.013) -0.011 (0.037) 0.012 (0.111) 0.091 (0.145) 24.218*** (0.302) 0.361* (0.208) -0.543* (0.298) 0.266 (0.275) 0.412 (0.436) 0.217 (0.211) -0.370 (0.369) 0.311 (0.300) 0.008 (0.557) -0.901*** (0.329) -0.015 (0.017) -0.028 (0.057) 0.002 (0.137) 0.160 (0.166) 24.958*** (0.411) 0.748*** (0.254) -1.579*** (0.548) -0.078 (0.363) 0.616 (0.871) 0.859*** (0.276) -2.194*** (0.619) -0.177 (0.369) 0.716 (0.997) -1.033** (0.471) -0.040** (0.020) 0.033 (0.052) 0.136 (0.199) -0.248 (0.332) 22.586*** (0.399) 969 382 690 267 279 115 Notes: a) Regressions include individual fixed effects and month-of-interview indicator variables. Robust standard errors in parentheses 36 Inallthreesensitivityanalyses,whichconsecutivelydroppedfromthesample1)peoplein theearlystagesofART,2)householdswithoutchildren,and3)peopletakingproteaseinhibitorsas partoftheirmedicationscheme,themarginallysignificanteffectontheOWfoodassistancegroup at6monthsdisappearedcompletely(SeeAppendix,TableA2).Inaddition,changingthe specificationofthemodeltoeitherpopulation‐averagedorindividualrandom‐effectsstillidentified thetrendsofimprovedBMIregardlessofwhichinterventiontheyreceived,andfoundno additionaleffectoffoodassistanceover12months(SeeAppendix,TableA4).Finally,usingbodyfat percentageasanalternatemeasureofnutritionalstatusdidnotidentifyaneffectoffoodassistance (SeeAppendix,TableA5). Examiningtheregressionsonfoodinsecuritybygender(Table4)indicatedthatatMonth6, menwhowerenotOWatbaselineexperiencedasignificantincreaseinBMI[b=0.748;p<0.01], whilemenwhowereOWatbaselineexperiencedasignificantdecrease[b=‐1.579;p<0.01; (∆BMIMonth6,education,OW=0.748–1.579=‐0.831],regardlessofinterventiongroup.AtMonth12, thistrendwaspersistentandmorepronounced.Trendsforwomenbyoverweightstatusmovedin thesamedirectionasthoseformen,butwithsmallermagnitude.Notrendwasevidentforwomen byMonth12.Aswiththefullpopulationregression,wedidnotdetectasignificanteffectoffood assistanceforanygroupateitherMonth6or12.However,wehadlowpowertodetecttrendsin BMIbygender(<50%),includingtheeffectoffoodassistance,andthusdonotmakecomparisons onpointestimatesacrossmenandwomen, DISCUSSION Thisstudycontributestoagrowingliteratureontheeffectsoffoodassistanceintegrated withARTonthehealthandwelfareofPLHIVandtheirhouseholds(Tirivayietal.,2011a).Wefind strongevidencethatfoodassistanceplusnutritioneducationimprovesfoodsecurityforbeneficiary householdscomparedtonutritioneducationalone.Evidenceofapositiveeffectoffoodassistance 37 onhouseholdfoodsecuritysuggeststhatatleastsome,butperhapsnotall,oftheintendedbenefits ofincreasedaccesstofoodarereachinghouseholdmembersandarenotbeingdivertedtoother purposes(e.g.sellingfoodtopayforothernon‐fooditems),beinglosttospillovertoextended familyandcommunitymembers,orcrowdingoutallothersourcesofotherexternalsupport,ascan beaconcernwithin‐kindtransfersincontextsofgeneralizedeconomicinsecurity(Marchione, 2005). Improvingfoodsecuritymayplayanimportantroleinalleviatingsufferingandimproving thementalandphysicalhealthofPLHIV.Foodinsufficiencyhasbeenlinkedtodepressionina rangeofsettingsandhealthconditions(Alaimoetal.,2002;Coleetal.,2011;Heflinetal.,2005; Siefertetal.,2004),particularlyamongwomen.Increasingly,studiesfindthatfoodinsecurityis closelyassociatedwithdepressionamongPLHIV(Tsaietal.,2012;D.Y.Wuetal.,2008)andthat alleviatingdepressioncanimproveHIVoutcomes(Tsaietal.,2010).Foodinsecurityalsopredicts morbidityandmortalityofpeopleonARTindependentlyofBMI(Weiseretal.,2009b;Weiseretal., 2009c;Weiseretal.,2012),suggestingthatreducingfoodinsecuritycanimproveHIVoutcomesfor peopleregardlessofnutritionalstatus.Whilewedidcollectinformationonmorbidity(e.g. hospitalizations,opportunisticinfections,etc.)andmentalhealthofPLHIVinourstudy,these outcomeswillbeanalyzedinfuturepapers.Furthermore,wehadinsufficientstatisticalpowerto examinemortalityinourstudy. TogetherwithevidencethatfoodinsecuritymaycompromiseARTadherence,ourresults supportthepropositionthatfoodassistancemaybesuccessfulatimprovingadherenceandHIV outcomesviaapathwayofimprovedfoodsecurity(SeeFigure4).Foodinsecurityisassociated withworseimmunologicstatusatARTinitiation(i.e.CD4count)(Normenetal.,2005;Weiseretal., 2009a)andpoormorbidityandmortalityonART(Weiseretal.,2009b;Weiseretal.,2009c;Weiser etal.,2012).FoodinsecurityhasbeendocumentedtoadverselyaffectARTadherenceinarangeof 38 resource‐‐limitedsettiings,includin ngLatinAmeerica(Deribeeetal.,2008;;Frankeetall.,2011;Marrcellin etal.,200 08;Martinettal.,2011b).Somestudiesattributeth hiseffecttorrecommendaationsthatm many antiretroviralmedicaationsbetakeenwithfood.Wheretherreisinadequatefoodsupply,peoplem may maydothemh harm.Inadd dition,peopleewithresourrceconstrain nts skipdoseesaltogetherforfearitm maybefo orcedtomak ketrade‐offsbetweentheedirectandi ndirectcostssoftreatmen nt(e.g.fees, transportt,lostworkttime),andoth herbasicneeeds.Whereth hereisinadeequatefoodiinthehouseh hold, apersonmaychoosetoskipcliniccorpharmaccyvisitsinorrdertoeatorrtoensureth hatfamily dinsecurityaasabarrierttoadherencee,our membersseat.Togetheerwiththeliiteratureidentifyingfood findingsssuggestthatprovidingsu upplementalfoodtopatieentsatahouseholdlevelcouldimpro ove adherencceintheshorrttermbyallleviatingtheneedtomakkethesedifficultandharm mfultradeofffs betweenfoodandheaalth. pathwayslin nkingfoodan ndnutrition ninterventio onswithART Toutcomes Figure4::Potentialp 39 Wefindsomeevidencethatfoodassistancebenefitedwomen’shouseholdfoodsecuritytoa greaterdegreethansuggestedbythefullpopulationregressions,whichcouldhaveparticular implicationsforadherenceinterventionstargetedatwomen.Inaddition,despitereducedpower availabletoanalyzetheeffectoffoodassistanceonfoodinsecurityamongmen,westillfinda positiveeffectformenafter12months.Althoughitappearsthatwomenmayhavebenefitedfrom foodassistancemorethanmenbasedonacomparisonofpointestimates,lowpowerlimitsany comparisonsacrossgender.Ratherweconsiderthesepotentialdifferencesacrossgenderan importantavenueforfutureexploration. Toourknowledge,nopublishedstudiesinresource‐limitedsettingshaveinvestigated nutritionalinterventionsintegratedwithARTthataddressfoodinsecurityofPLHIVinsettingswith highprevalenceofoverweightandobesity.ExcessweightcanexacerbateART‐relatedmetabolic syndrome,whichincludescentralfataccumulation,insulinresistance,lipidabnormalities,and hypertension(Alvarezetal.,2010;Friis‐Mølleretal.,2003),sometimesasideeffectofART.People whotakecertainclassesofantiretroviralmedications,specificallyproteaseinhibitors,maybeat particularrisk.Inturn,metabolicsyndromeisassociatedwithincreasedriskofcardiovascular diseaseandtype2diabetesmellitus(Albertietal.,2005;Grundyetal.,2005).Despiteevidence suggestingthatobesitymayhaveprovidedaprotective‘survival’effectforPLHIVbeforehighly effectivecombinationARTbecameavailable(Shor‐Posneretal.,2000),currentstudiesofpeople receivingARTsuggestthatobesityisassociatedwithpoorerimmunologicresponsetotreatment (Crum‐Cianfloneetal.,2010),andthatnotbeingunderweightyetbeingfoodinsecurecarriesa higherriskofmortalityontreatmentthanbeingunderweightandfoodsecure(Weiseretal., 2009b).Thesestudiesindicatethatmaintainingexcessweightisnotlikelytobeaprotective strategyforPLHIVreceivingART,particularlywherefoodinsecurityexists.Whileagrowing epidemiologicalliteratureinLatinAmericahasidentifiedoverweightandobesitytobeasalient issueforPLHIV,thisliteraturedoesnotspecificallyaddressfoodinsecurity,norexploretheeffects 40 offood‐relatedinterventionsonnutritionaloutcomes,foodinsecurityandadherenceinthecontext ofoverweightandobesity(Alvarezetal.,2010;Jaimeetal.,2006;Leiteetal.,2010;Marizetal., 2011). Inthiscontext,itisencouraging–butnotnecessarilysurprising–thatwedonotfindfood assistancetoadverselyaffectBMIforPLHIVwhoareoverweightorobeseandreceivingART, especiallywomen,atleastintheshortterm.Recentreviewsofthelinkbetweenfoodinsecurity interventionsandnutritionalstatusamongresource‐limitedindividualswithintheUnitedStates havefoundgrowingevidencethatweightgainisassociatedwithparticipationinfoodassistance programsamongwomen(butnotmen)(N.I.Larsonetal.,2011;Wilde,2007).However,aclear causalchainbetweenfoodstampsandweightgainhasyettobeempiricallyestablished. Furthermore,theriskofweightgainmayincreasewiththedurationofparticipationinfood assistanceprograms(Zagorskyetal.,2009),implyingthateveniffoodassistancedoesincrease weightandBMI,itmaynotdosoovertherelativelyshortperiodoftimemeasuredinourstudy. Meanwhile,thereisaslighttrendofreduced(i.e.improved)BMIamongoverweightor obeseparticipantsinourstudy,regardlessofwhichinterventionstheyreceived.Theseeffects appeartobeparticularlystronginmen,althoughreducedpowerforsplitgenderanalysisofBMI outcomeslimitsourabilitytocompareestimatesacrossgender.ItispossibleimprovementsinBMI mayberelatedtoreceivingnutritioneducation,viaimproveddietarycounselingandhabits. However,withoutacomparablecontrolgroupofPLHIVnotreceivinganyintervention,wecannot separatetheeffectsofthenutritioneducationfromseculartrendsunrelatedtoourinterventions. Futureanalysiscouldalsoconsiderfoodintakeanddietarydiversityasoutcomes,bothofwhich couldhaveimprovedoverthecourseofthestudy,andperhapsexplainsomeofthechangesinfood consumptionandutilizationexperiencedattheindividualandhouseholdlevelunderlyingobserved improvementsinfoodinsecurityandBMI.Inaddition,evidenceofincreasedBMIforparticipants 41 startingthestudyinthenormalBMIrangeinbothstudygroups(asportrayedinFigure3band suggestedbytheregressionresultsinTable4)couldmeritfurtherinvestigation.Althoughmean BMIforthosestartinginthenormalrangedidnotexceedacceptablelevelsoverthecourseofthe study,wedidnotexaminetheincidenceofoverweightandobesityasanoutcome.Futurestudies couldconsidertheeffectoffoodassistanceontheincidenceorprevalenceofoverweightand obesity,conditioningoninitialnutritionalstatus. Significantissuesremainindeterminingthemostoptimalandsustainablewaytoaddress foodinsecurityinthecontextofHIVtreatment(Fregaetal.,2010).Amongthemostpressingissues aredeterminingtheappropriatecompositionoffoodassistance,aswellasthemostrelevant criteria,timing,andstrategyfortransitioningoffofdirectfoodaid.Theseareissuesthatmanyfood assistanceprogramsintegratedwithARTarefacing,withlittleguidancefromthepublished literature(Yageretal.,2011).Thisisparticularlysoforprogramswherefoodaidisintendedto addressadherencebyalleviatinghouseholdfoodinsecurityratherthanacuteundernutritionat ARTinitiation.Foodinsecurityinresource‐limitedsettingstendstobeapervasiveandchronic realitythat,whileintensifiedbyHIV,ispersistentatboththeindividualandcommunitylevel (Bukusubaetal.,2007;Crushetal.,2011).Wherefoodinsecuritycoincideswithunderweight statusatARTinitiation,itmaybemorestraightforwardtophaseoutfoodassistanceasthe beneficiarygainsweightandentersanormalBMIrange,ratherthanconsiderhowmuchfood securityis‘enough’.Inaddition,regardlessofnutritionalstatus,improvingfoodsecurityinthefirst yearoftreatmentmaystrengthenoverallARTadherenceandoutcomes,amplifyingthepositive effectofARTonlaborsupply(Thirumurthyetal.,2011;Thirumurthyetal.,2008b),andpromoting evengreaterimprovementsinfoodsecurityaseconomicwell‐beingimproves(Palar,2012). However,thecriteriaforphasingoutfoodassistancebecomeincreasinglyinconclusivewherefood insecuritycoincideswithhigherBMIsorisgivenduringthecourseoftreatmentratherthanatthe onset.Facedwithlimitedbudgets,programleadersmaysimplychooseanarbitraryendpointto 42 phaseoutfoodassistancebasedonprogrammaticfeasibility.Furthermore,withoutaddressingthe underlyingreasonsforfoodinsecurity,thereislittlereasontobelievethatfoodassistancealone willimproveadherencebeyondthedurationoftheprogram.Forthesereasons,theWorldFood ProgramandotherorganizationsengagedwithART‐integratedfoodassistanceareincreasingly turningtolivelihoodapproachestosupportfoodsecurity(Fregaetal.,2010;Kadiyalaetal.,2009; Yageretal.,2011). Findingsfromourstudypointtothepositiveeffectsoffoodassistancebeingconcentrated inthefirst6monthsoftheintervention.Whilethismaysuggestthatshortertermfoodassistance coupledwithHIVtreatmentandcarecouldbesufficienttoachieveimprovedfoodsecurity(e.g.a6 monthprogram),wecannotbecertainthatfoodsecuritydidnotcontinuetoimproveafter6 months,giventhatournextdatapointdidnotoccuruntilmonth12.Moreimportantly,itispossible thatanticipatingtheendoffoodassistancechangedhouseholdfoodbehaviorinthelatterpartof theintervention,orincreasedanxietyanduncertaintyabouthouseholdfoodsupply,possibly counteractingsomeoftheimpactsoffoodassistanceonfoodsecurity.Forexample,ifhouseholds conservedfoodsuppliesratherthanconsumingthem,thiscouldhavedecreasedfoodsufficiencyin theshortterm.Suchbehaviorwouldbeconsistentwithfoodinsecurityasa‘managedprocess’ wherebyhouseholdstradeoffbetweencurrentandfutureconsumption(Coatesetal.,2006; Corbett,1988).Meanwhile,regardlessofwhetherfoodconsumptionwasactuallyconstrainedafter 6months,worryoverthefoodsupply–whichiscapturedinourfoodinsecuritymeasure–could havealsohaltedimprovementinfoodsecurity.Therefore,wedonottakeourfindingstosuggest thattransitionoffoffoodassistanceshouldstartat6months;rather,theyindicatethatmore researchisneededtounderstandhowhouseholdsareutilizinganddynamicallymakingfood‐ relateddecisionsinthecontextofexpectationsaroundfoodassistanceandongoingHIVtreatment. 43 Ourstudywassubjecttoseverallimitations.First,theinterventionswerenotrandomly assignedbyindividual,butbyasmallnumberofclinicsites.Givenourlackoflargenumber randomization,itisnottoosurprisingthatthefoodassistancegroupwassystematicallydifferent fromthenutritioneducationgroup.Inparticular,itwasworseoffthanalongseveralkey dimensionsatbaseline,includingoneofourkeyoutcomes(foodinsecurity),pointingtopotential selectionbias.Weattemptedtominimizetheeffectsofselectionontime‐invariantcharacteristics (bothobservablesandunobservables)byusingamodelwithindividualfixedeffectstoidentifythe effectoftheintervention.Asarobustnesscheck,wethentestedourmodelassumptionthat differencesintime‐invariantcharacteristics(e.g.clinicsite,baselinefoodinsecurity,etc.)didnot affectchangeinouroutcomevariablesbyimplementingalternatemodelspecificationscontrolling forthesedifferences.Resultsofthealternatemodelsdidnotnullifyorreversetheoveralldirection orsignificanceofresults.However,thereremainsapossibilityofomittedvariablesbiasinour regressions.Whilethefixedeffectsregressionallowstime‐invariantcomponentoftheerrortobe correlatedwiththeregressors,itrequiresthatthetime‐varyingcomponentbeuncorrelated(i.e. thatalltime‐varyingcharacteristicsthatinfluencetheeffectoftheinterventiononfoodinsecurity areincludedascontrolsintheregression).Forexample,wedidnothaveagoodmeasureofwealth orincome,whichmayhavechangedovertimeasthefoodbasketaddedin‐kind‘income’tothe household,andwhichiscloselytiedwithfoodinsecurity.Controllingforchangeinworkstatusand economicsupportfromfamilyorfriendsmaybeareasonableproxytooverallchangeinincomeor wealth,butonlyiftheyweresensitivetochangesinrecipienthouseholdwell‐beingandifthe binarymeasureswereabletocapturethesechanges.Inaddition,ourvariablerepresentingchange inHIV‐relatedhealth(thebinary‘HIVsymptomatic’variable)maynothaveadequatelymeasured physicalhealthstatus;however,itwasouronlyobjectiveHIV‐relatedhealthmeasurethatwas availablelongitudinallyatthetimeofanalysis.Finally,therewassomeevidenceofattritionbias.In particular,thosewhowereHIVsymptomaticatbaselinewerelesslikelytocompletethestudy 44 (regardlessofinterventiongroup),perhapsreflectinggreaterratesofhospitalizationand potentiallydeath.Inaddition,participantsinthefoodassistancegroupweremorelikelyto completethestudy,perhapsreflectingthegreaterneedinthisgroup.Weattemptedtominimize thisbiasbyincludingattritionweightsintheanalysis.Finally,onecaveattothegeneralizabilityof ourresultsisthatwewerepotentiallysubjecttounobservedself‐selectionbypatientswhohad betteradherence,giventhatrecruitmentwasconductedduringregularclinicvisits.Thus,patients withbetterclinicattendancemayhavebeenmorelikelytobeselectedintooursample. CONCLUSION FoodassistancecombinedwithnutritioneducationprovidedinthecontextofARTmay improvethehouseholdfoodsecurityofPLHIVandprovideanimportantpathtoadherenceacross theBMIspectrum.PositivetrendsinBMIforbothunderweightandoverweightorobese participantssuggestthatnutritioneducationmaybeworkingtoimprovenutritionalstatusfor PLHIV,andmeritsfurtherrigorousevaluation.Whilewefoundnoadditionaleffectoffood assistanceonnutritionalstatus,wewerelimitedbyrelativelysmallsamplesizes,particularlyinthe non‐normalBMIgroups.Therefore,researchshouldcontinuetoassesswhetherfoodassistance positivelyincreasesweightamongmalnourishedindividualslivingwithHIV,oradverselyincreases weightfornormal,overweightorobesefoodbeneficiaries,wherevertheyareincludedin interventions.Atthesametime,policyandprogramleadersshouldcarefullyconsiderwhethera combinationoftargetedfoodassistancetopeoplereceivingARTwhoarecriticallyunderweight, livelihoodsupportforallfoodinsecureARTpatients,andHIV‐specificnutritionaleducationby trainednutritionistsmaybetteraddress1)thelong‐termfoodandnutrition‐relatedchallengesto adherencethangeneralfoodassistanceintegratedwithART,and2)programmaticsustainabilityof interventions,givenlimitedfinancialresources.Morerigorousresearchisneededtodetermine howeachofthesestrategiesimprovesthehealthandwelfareofPLHIV,andaffectsARTprogram costs,particularlylivelihoodsstrategiesandnutritioneducation.Itisclear,however,isthatfood 45 insecuritymustbeactivelyaddressedinordertofullyrealizethepotentialofARTtoimprovethe livesofPLHIVinresource‐limitedsettings. 46 APPENDIX TableA1:Foodinsecuritysensitivityanalyses Food Insecurity Score Month 6 Month 6 X FA Month 12 Month 12 X FA HIV symptomatic Household size Worked in the last month Material support Constant Observations Number of person ID *** p<0.01, ** p<0.05, * p<0.1 Original SA1: Drop early stage ART recipients SA2: Drop households w/o kids -1.891*** (0.473) -3.150*** (0.606) -1.750*** (0.433) -2.740*** (0.581) -0.290 (0.633) 0.267* (0.136) -0.769** (0.352) 0.245 (0.391) 9.596*** (0.952) -1.588*** (0.493) -3.372*** (0.619) -1.824*** (0.459) -2.586*** (0.607) -0.571 (0.621) 0.266* (0.136) -0.751** (0.369) 0.254 (0.396) 9.576*** (0.952) -1.398*** (0.518) -3.495*** (0.652) -1.470*** (0.473) -2.867*** (0.617) -0.182 (0.727) 0.361** (0.148) -0.783** (0.373) 0.289 (0.394) 9.093*** (1.071) 971 382 892 348 807 310 Notes: a) Results of longitudinal linear regression (same model specification as Table 3) b) Regressions include individual fixed effects and month-of-interview indicator variables. c) Robust standard errors in parentheses 47 TableA2:Foodinsecurity–Alternatemodelspecifications Food assistance group Time Time X FA CAI2 CAI3 CAI4 Baseline food insecurity score Constant Observations Number of ID Severe food insecurity (a) Population averaged (b) Population averaged (c) Individual random effects -0.275 (0.390) -0.666*** (0.170) -0.742*** (0.226) 1.501*** (0.237) 0.000 (0.000) 0.508 (0.413) 0.494*** (0.052) -4.362*** (0.698) -0.380 (0.476) -1.218*** (0.213) -1.086*** (0.290) 1.626*** (0.393) 0.000 (0.000) 0.246 (0.534) 0.745*** (0.038) 3.262*** (0.723) -0.467 (0.732) -1.224*** (0.239) -1.038*** (0.317) 1.653*** (0.395) 0.000 (0.000) 0.195 (0.510) 0.737*** (0.034) 3.441*** (0.865) 837 326 837 326 844 330 Continuous food insecurity *** p<0.01, ** p<0.05, * p<0.1 Notes: a) Regression (a) gives results from the logistic regression in log odds to facilitate easy comparison of sign (but not magnitude) with the linear models. b) “Time” is an ordinal variable for the 3 assessments (models would not converge with indicators for 6 and 12 months) c) CAI1 is the omitted study site d) All regressions include gender, race, education, baseline CD4 count, and the following timeupdated variables: household size, work status, material support, HIV symptomatic status, and month of interview indicators. e) Robust standard errors in parentheses 48 TableA3:BMIsensitivityanalyses BMI SA1: Drop early stage ART recipients SA2: Drop households w/o kids SA3: Drop if taking protease inhibitors 0.649*** (0.165) -0.906*** (0.260) 0.027 (0.223) 0.690* (0.394) 0.534*** (0.166) -0.850*** (0.317) 0.079 (0.230) 0.278 (0.482) -0.915*** (0.273) -0.024* (0.013) -0.011 (0.037) 0.012 (0.111) 0.091 (0.145) 24.218*** (0.302) 0.437*** (0.163) -0.689*** (0.247) 0.179 (0.222) 0.542 (0.387) 0.431** (0.172) -0.590** (0.290) 0.123 (0.228) 0.019 (0.459) -0.794*** (0.296) -0.017 (0.013) -0.013 (0.037) -0.003 (0.115) 0.083 (0.137) 24.299*** (0.296) 0.421** (0.179) -0.697** (0.282) 0.156 (0.235) 0.611 (0.416) 0.436** (0.193) -0.778** (0.351) 0.133 (0.252) 0.121 (0.536) -0.995*** (0.300) -0.022 (0.015) -0.046 (0.044) 0.090 (0.120) 0.182 (0.157) 24.489*** (0.341) 0.662*** (0.173) -0.838*** (0.266) 0.052 (0.231) 0.434 (0.421) 0.521*** (0.175) -0.826** (0.327) 0.107 (0.238) 0.168 (0.516) -0.983*** (0.288) -0.024* (0.013) -0.009 (0.038) 0.042 (0.117) 0.086 (0.153) 24.061*** (0.313) 969 382 890 348 806 310 906 355 Original Month 6 Month 6 X OW Month 6 X FA Month 6 X FAX OW Month 12 Month 12 X OW Month 12 X FA Month 12 X FA X OW HIV symptomatic Food insecurity score Household size Worked in the last month Material support Constant Observations Number of person ID *** p<0.01, ** p<0.05, * p<0.1 Notes: a) Results of longitudinal linear regression (same model specification as Table 4) b) Regressions include individual fixed effects and month-of-interview indicator variables. c) Robust standard errors in parentheses 49 TableA4:BMI–Alternatemodelspecifications BMI Food assistance group OW at baseline FA X OW Time Time X OW Time X FA Time X FA X OW CAI2 CAI3 CAI4 BMI at baseline Constant Observations Number of ID (a) Population averaged (b) Individual random effects -0.238 (0.204) 0.746** (0.308) 0.015 (0.298) 0.221** (0.091) -0.333* (0.178) 0.037 (0.120) 0.113 (0.262) 0.316 (0.202) 0.000 (0.000) -0.095 (0.209) 0.929*** (0.022) 2.001*** (0.527) -0.276 (0.278) 0.699** (0.350) 0.049 (0.441) 0.210** (0.088) -0.279* (0.153) 0.051 (0.115) 0.087 (0.205) 0.328** (0.154) 0.000 (0.000) -0.100 (0.204) 0.925*** (0.019) 2.048*** (0.495) 840 326 861 339 *** p<0.01, ** p<0.05, * p<0.1 Notes: a) “Time” is an ordinal variable for the 3 assessments (models would not converge with indicators for 6 and 12 months) b) CAI -1 is the omitted study site c) All regressions include gender, race, education, baseline CD4 count, indicator for taking protease inhibitors at baseline, and the following time-updated variables: household size, work status, material support, HIV symptomatic status, and month of interview indicators. d) Robust standard errors in parentheses 50 TableA5:Bodyfatpercentageasanalternativenutritionalstatusmeasure Body fat % Month 6 Month 6 X OW Month 6 X FA Month 6 X FAX OW Month 12 Month 12 X OW Month 12 X FA Month 12 X FA X OW HIV symptomatic Food insecurity score Household size Worked in the last month Material support from family or friends Constant 0.998** (0.410) -1.221** (0.528) -0.099 (0.515) 0.931 (0.725) 0.460 (0.387) -0.609 (0.534) 0.411 (0.495) 0.109 (0.779) -1.396*** (0.535) 0.012 (0.027) 0.043 (0.080) -0.129 (0.251) 0.124 (0.304) 26.950*** (0.632) 985 Observations Number of person ID 382 *** p<0.01, ** p<0.05, * p<0.1 Notes: a) Regressions include individual fixed effects and month-of-interview indicator variables..Robust standard errors in parentheses. Waist and mid-upper arm circumference not included in regressions because of significant missing data. 51 II.LivelihoodexperiencesofpeoplereceivingintegratedHIVtreatment andfoodassistanceinBolivia:Lessonsforsustainableinterventions ABSTRACT Introduction:LivelihoodinterventionstoimprovefoodsecurityandsustainableHIV treatmentoutcomesareincreasinglypromotedforpeoplelivingwithHIV(PLHIV)receiving antiretroviraltherapy(ART).Yet,anin‐depthunderstandingofhowPLHIVexperiencetheirown livelihoodsinrelationtoHIVtreatmentintheabsenceofexternalprogramsisstillneededtobetter informthedevelopmentofappropriateinterventionsandrelatedpolicies,particularlyinurban settings. Methods:Weusedamixedmethodsapproachtoinvestigatethelivelihoodexperiencesof peoplereceivingARTinthreecitiesinBoliviawhowerepartofaclinic‐based,foodassistancepilot project(n=211).Closed‐endedquestionnairesandqualitativeinterviewswereconductedwithall participants.Thequestionnairecaptureddataondemographics,householdcomposition,socio‐ economicsituation,includingworkstatus,andfoodinsecurity.Thesemi‐structuredqualitative interviewthenexploredlivelihoodandHIVtreatmentexperiencesinmoredepth,includingwork‐ relatedbarrierstoARTadherence,HIV‐relatedbarrierstowork,rangeofeconomicactivities conducted,andeconomiccopingstrategies.Extensivequalitativecodingwasperformedtoidentify prominentthemesthatemergedfromthesemi‐structuredinterviews,usingtwocodersto maximizethevalidityofthethemesandreliability.Quantitativedatawereanalyzedusing univariateandbivariatestatistics.Datadisplaymatriceswereusedtoexplorepatternsand compareresponsesacrossthequantitativeandqualitativedata. Results:Studyparticipantsreportedcomplexeconomiclivesoftencharacterizedby multipleeconomicactivities,includingbothformalandinformallabor.Theystruggledtomanage 52 ARTtreatmentandlivelihoodssimultaneously,andfacedbarrierstothisdualmanagementthat rangedfromtheinterpersonaltothestructural.Inparticular,issuesoflackofdisclosureofHIV status,stigmaanddiscrimination,werehighlysalientforstudyparticipants,manifestingthrough conflictaroundrequestingtimeoffforclinicvisits,managingresentmentfromco‐workersabout timeoffandtakingmedicationsatworkinsecretorunderotherpretenses.Inaddition,health systemissuessuchaslimitedclinichoursordrugshortagesexacerbatedthestruggletobalance economicactivitieswithHIVtreatment. Conclusions:Effectivelivelihoodprogramsshouldtakeintoaccountthecurrentsetof economicactivities,skills,andbarriersexperiencedbybeneficiariesandprovidearangeof opportunitiesthatcomplementtheseexperiences.Improvedpolicy‐leveleffortstoenforceexisting anti‐discriminationlaws,reduceHIV‐relatedstigma,andexpandhealthservicesaccessibilitycould mitigatemanyofthebarriersdiscussedbyourparticipantsandreducetheneedforseparate livelihoodinterventions.GiventhemultiplelayersofdisadvantagefacedbyPLHIV,comprehensive HIVcareandsupportpackagesmustintegratehealth,socialandeconomiccomponentsthatare supportedbystrongnationalHIVpolicyandlinktonationalsocialprotectionandsocialsafetynet programs. 53 INTRODUCTION Livelihoodprogramsandpoliciesareincreasinglypromotedtosupporttheeconomicwell‐ beingandfoodsecurityofpeoplelivingwithHIV(PLHIV),improveantiretroviraltreatment(ART) adherenceandoutcomes,andserveasasustainabletransitionfromfoodassistanceinHIV treatmentsettings(Fregaetal.,2010;Kadiyalaetal.,2009;Yageretal.,2011).However,anin‐ depthunderstandingofhowfoodinsecurePLHIVexperiencetheirownlivelihoodsinrelationto HIVtreatmentisstillneededtoinformappropriateinterventionsandpolicies.Whilestudies suggestthathealthimprovementsaccompanyingARTmayleadtorenewedproductivecapacity andincreasedlaboursupplyinresource‐limitedsettings,themultidimensionalwaysinwhich peopleonARTexperiencetheirlivelihoods,andhowtheseinturnaffectstheirtreatmentdecisions, arenotwell‐understood,particularlyforurbansettings.Fewstudiesinlow‐incomecountries explorein‐depthhowPLHIVco‐manageARTandwork,thebarrierstheyfaceinthequestto integratetheireconomicliveswiththeexpectationoflifetimetreatment,andhowtousethis informationtodeveloplivelihoodsinterventionsinthecontextofART. Livelihoodscanbedefinedasthe“thecapabilities,assets(includingbothmaterialand socialresources)andactivitiesrequiredforameansofliving”(Chambersetal.,1991),whilefood securitycanbedefinedas“physicalandeconomicaccesstoadequatefoodforallhousehold members,withoutriskoflosingsuchaccess”(Haeringetal.,2009).Proponentsofalivelihoods approachtosupportingfoodsecurityprioritizethelong‐termwell‐beingofPLHIValongmultiple dimensions,includingeconomicproductivity,health,andsustainedaccesstofoodandnutrition. In2008,WFP’sRegionalOfficeforLatinAmericaandtheCaribbean(WFP‐LAC)began implementingastrategytosupportregionalgovernments’capacitytointegratefoodand nutritionalinterventionswithHIVtreatmentandcare,includingattentiontolivelihoods,in responsetopolicygapsaddressingnutritionandfoodsecurityforPLHIVinLAC.Amongotherfood 54 andnutrition‐relatedactivities,WFPutilizesfoodassistancetosupporttreatmentandcare,and mitigatetheimpactoftheepidemiconvulnerablehouseholdsinmorethan50countriesworldwide 2,includingcountriesinLatinAmericaandtheCaribbean(LAC).Recognizingthatfoodsecurity interventionsforPLHIVmustsupportsustainablehealthoutcomes,WFPalsoencouragesfood‐ basedinterventionsworldwidetoincorporatelivelihoodsstrategiesthatcontributetothelong‐ termfoodsecurity,nutritionalrecoveryandARTadherenceofitsbeneficiaries.However,WFP’s food‐basedinterventionsarerelativelynewinLatinAmerica,andlivelihoodsinterventionshave yettobecomprehensivelydevelopedandimplemented. Asmallnumberofstudiesexaminingadherencemoregenerallyhaveidentified unemploymentandfearoflostworktimeasbarrierstoadherence(Hardonetal.,2007;Rachliset al.,2011).However,littleisstillknownabouthowlivelihoodsaffectadherence,withevenless knownabouthowARTinturnaffectslivelihoods.AmongmajorissuesaffectingPLHIV,socialissues suchasstigmaanddiscrimination,andstructuralissuessuchaspovertyandhealthcareaccess,are relatedtobothHIVtreatmentexperiencesaswellaslivelihoods.Forexample,fearofdisclosurehas beenidentifiedasabarriertoadherenceindevelopingcountries(Millsetal.,2006;Nachegaetal., 2004),whilefearofdisclosureanddiscriminationintheworkplaceagainstpeoplewithHIVhave alsobeenwidelydocumented(Mahajanetal.,2008;Spragueetal.,2011).Insufficienteconomic resourcesalsoposesabarriertoadherence,suchaslackoffood,transportorhousing(Anemaetal., 2009;Hardonetal.,2007;Kageeetal.,2011),oftencoincidingwithlivelihoodinsecurity(Rachliset al.,2011),aswellasstructuralpovertyandinequality(Stillwaggon,2006). Furthermore,fewpublishedstudiesexplicitlyinvestigatethelivelihoodexperiencesof peopleonARTwhoarefoodinsecurewithaneyetointerventiontargetinganddevelopment. Qualitativestudiesinsub‐SaharanAfricahavelookedattheimportanceofworkasapartof 2 See http://www.unaids.org/en/aboutunaids/unaidscosponsors/wfp/ 55 “comingbackfromthedead”forpeoplestartingART(Russelletal.,2009;Russelletal.,2007), whilequantitativestudiesinsub‐SaharanAfricaandIndiahaveprimarilyinvestigatedtheimpactof ARTonmeasuresoflabourforceparticipationandeconomicproductivity(d’Addaetal.,2009;B. Larsonetal.,2008;Rosenetal.,2010;Thirumurthyetal.,2011;Thirumurthyetal.,2008b).We identifiedonlyonepublishedstudy,setinZambiaandKenya,thatexplicitlyexploredhow livelihoodsstrategiesplayaroleinachievingfoodandnutritionsecurityforpeoplereceivingART, withimplicationsforinterventions(Samuelsetal.,2011).Althoughtheliteratureissparseinall geographicareas,mostexistingdataonHIVandlivelihoodscomesfromruralsettingsinsub‐ SaharanAfrica,whicharecharacterizedbyadifferentsetoflivelihoodsandfoodsecurityissues thaninurbansettings(Crushetal.,2011)andinLatinAmerica. ThroughoutLAC,widespreadinequalities,discriminationandpovertyaresignificantfactors shapingtheHIVepidemic(Smallman,2007).Concentrationsofextremepovertyinruralareasfuel seasonallabourmigrationandattendantHIVrisks,andhavealsocontributedtorapidurbanization andincreasingnumbersofpeoplelivinginextremepovertyincitieswhereconditionsareripefor therapidspreadofHIV(Stillwaggon,2006).WhileLACgovernmentshavemadestrongprogress towardsuniversalaccesstoARToverthelastdecade,accesstocomprehensiveHIVcare–which includesattentiontofoodandlivelihoodsecurity–remainslowinresource‐limitedsettings throughouttheregion(Martinetal.,2011a;UNAIDS,2008;WorldBank,2007). Inthisstudy,weinvestigatethelivelihoodexperiencesoffoodinsecurepeoplereceiving ARTinBoliviawhowerepartofaclinic‐basedpilotprojectofferingfoodassistanceandnutritional education,sponsoredbytheWFP‐LAC.Thenutrition‐basedpilotdidnotincludeastructured livelihoodscomponentatitsinception,althoughtheWFPwasinterestedinaddingoneinthefuture asatransitionstrategyfromfoodassistance.Amixedmethodsstudywasthusimplementedto explorethecurrentwork,economic,andHIVtreatmentexperiencesoffoodpilotparticipantsin 56 ordertoinformthecreationoffuturelivelihoodsinterventions,giventheoverallabsenceof evidenceonthesubjectforpeoplelivingwithHIVinurbansettingsandinLatinAmerica.Using datafromthisstudy,weaimtoexploreanddescribetheinterconnectionbetweenlivelihood experiencesandHIVtreatment,andidentifymajorbarriersandopportunitiesforlivelihoods interventionsinthecontextofART.Ourgoalistobothinformlivelihoodinterventionsforpeople onART,includingthosetransitioningfromfoodassistance,aswellastocontributetothebroader scientificevidenceonthelivelihoodexperiencesofpeopleonARTinresource‐limitedsettingswho nowlivewithHIVasachroniccondition. METHODS BackgroundofResearchCollaboration ThisstudyinvolvedcollaborationbetweentheUNWFPRegionalOfficeforLACandthe RANDCorporation,anonprofitresearchorganizationbasedintheUnitedStates.In2008‐2009, WFP/RANDbeganimplementingjointactivitiesinBoliviaandHondurasbyconductingqualitative, formativeresearchonthedietaryhabitsandnutritionalstatusofpeoplelivingwithHIVreceiving ART.Thedatafromthisresearchwasusedtodesigncontextandneeds‐specificfoodbasketsand nutritionalcounselingmethodologiesforuseinpilotfoodassistanceandnutritionaleducation interventionsforpeoplewithHIVinBoliviaandHondurasduring2010and2011. DataCollection WFP/RANDimplementedamixedmethodsstudyoflivelihoodandeconomiccoping experienceswithasampleoffoodinsecureARTpatientsinBoliviaparticipatinginthefood assistancepilotsponsoredbyWFP‐LAC.Thecoreresearchteamconsistedoftwoleadresearchers fromtheUnitedStates(onefromRANDandtheotheraconsultantforWFP),bothfluentinSpanish andwithsubstantialexperienceinLatinAmericaandinHIV,aBolivianprojectmanager,andthree 57 Bolivianinterviewers.AnotherRANDresearcherwhowasalsofluentinSpanishandexperiencedin LatinAmericaandHIVprovidedinputthroughoutthestudy.TheBolivianmembersoftheresearch teamallhadextensiveexperienceworkingwithPLHIVaswellasresearch‐basedinterviewing.The leadresearcherstrainedtheprojectmanagerandinterviewersontheresearchmethodsandstudy instruments.Allstagesoftheresearchprocess,includingtheinterviews,wereconductedin Spanish. ThesampleframeforthelivelihoodsstudywastheuniversalsetofadultsreceivingART recruitedintothefoodassistancepilotduringitsfirstsixweeks(November–December2010). Studyexclusioncriteriawerebeinghospitalizedorbedriddenduetoillness,notspeakingSpanish, orbeingunderage18.Duetoexternalcircumstances,theinterviewsdidnottakeplaceuntil severalmonthslater(April2011),andbythistime,someofthepeopleoriginallyrecruitedintothe studycouldnotbelocatedandwerereplacedwithpeoplewhohadstartedthefoodpilotmore recently.Replacements(n=38)werechosenpurposivelytomirrorthebasicdemographic characteristicsofthemissingparticipants.Asanexploratorystudy,ourgoalwasnottoachievea representativesample;rather,wesoughtalargeenoughsampletogeneratetherangeofsalient livelihoodsissuesexperiencedbyourstudypopulationinrelationtoHIVtreatmentandtoreach saturationofideas.Byrecruitingtheuniversalsetofadultsinthefirstmonthofthefoodpilot,we aimedtoreduceselectionissuesinoursample. Face‐to‐faceinterviewsconsistedofaclosed‐endedquestionnaireandasemi‐structured qualitativeinterview(N=211),whichtogethertookapproximatelyonehour.Theentireinterview wasadministeredorally,toenablepatientsofallliteracylevelstoparticipate. Theclosed‐endedquestionnairewasadaptedfromaSpanish‐languagequestionnaire previouslydeveloped,validatedandusedbyRAND/WFPinBoliviaandHonduras.Itincluded questionsondemographics,householdcomposition,socio‐economicsituation,includingwork 58 status,andfoodinsecurity(ELCSAscale)(Melgar‐Quiñonezetal.,2010).Thequestionnaire measuredworkstatususingaseriesofquestionsaskingiftheparticipant1)hadworkedinthelast 6months,2)hadworkedinthelastmonth,and3)wascurrentlyworking. Thequalitativeinterviewexploredlivelihoodexperiencesinmoredepth,inordertoexpand fromtheclosed‐endedquestionnaireandfacilitatecomparisonbetweenquantitativeand qualitativeresponses.Forexample,protocoltopicsexaminedinthispaperincludetherangeof economicactivitiesperformedbyparticipantsingeneralandinthelastweek,addingrichnessto thebinaryquestionsonworkstatus.Inaddition,theprotocolincludedquestionsregardingwhat participantswishedtochangeabouttheirlivelihoods;barrierstheyexperiencedtochanging livelihoods;howlivingwithHIVandreceivingtreatmentaffectedtheirlivelihoods;andhowtheir livelihoodsaffectedtheirtreatmentregimen.Withinthebroadprotocoltopics,interviewersused probesandclarifyingquestionstodrawoutrichnesswithinparticipantstories. Theprotocolforthesemi‐structuredinterviewwasdevelopedinEnglishandprofessionally translatedintoSpanish.MembersofAsociaciónUnNuevoCamino(ASUNCAMI),anNGOmemberof Bolivia’snationalnetworkofPLHIV(REDBOL),reviewedthetranslatedprotocolforlanguageand culturalappropriatenessfortheBoliviancontext.Afterthisreview,theresearchteamworkedwith ASUNCAMItopre‐testtheprotocolbyadministering10interviewstoASUNCAMImembersaspart ofavalidationexercise.Thegroupthendiscussedissues,identifiedareaswhereclarificationor reframingwasneeded,andmadesuggestionsforchangestothestudymaterials.Thisvalidation exerciseledtosubsequentrevisionsandafinalversionoftheinstrument. Participationinthestudywascompletelyvoluntaryandnotaconditionforreceivingfood assistance.Upontheadviceoflocalpartners,RANDprovidedasmallmonetaryincentiveof10 Bolivianos(~$1.75)toeachrespondentasatokenofappreciation.Informedconsentwasobtained fromallparticipants.Participantresponseswereidentifiedbyauniquecodeandtheiridentities 59 remainedanonymoustotheU.S.‐basedresearchers.EthicalapprovalwasobtainedfromRAND’s HumanSubjectsProtectionCommittee,andfromtheBoliviannationalinstitutionalresearchboard, ComitéNacionaldeBioética,ComisióndeÉticadelaInvestigación(CEI).Inaddition,theWFP‐LAC, WFP‐Bolivia,andmembersoftheNationalAIDSProgramviewedandagreedonallmaterials. Analysis TheinterviewsweretranscribedinSpanishandanalyzedintheoriginallanguage.Extensive codingofinterviewtranscriptswasperformedtoidentifyprominentthemesthatemergedfrom participantinterviewsusingAtlas.ti,aqualitativetextmanagementsoftware.Twocoders(thetwo leadresearchers)wereusedtomaximizethevalidityofthethemesandreliability(Milesetal., 1994).Asafirststep,acodebookofoverarchingthemeswasdeveloped(Weber,1990)basedon themajortopicsinourinterviewprotocol.Second,contentcodingprocedureswereusedto identifythepresenceofthesethemes(Altheide,1996;Krippendorff,2004;Weber,1990)in combinationwithinductiveapproachestoidentifynewthemesthatwerethenaddedtothe codebook(Milesetal.,1994;Straussetal.,1998).Tofacilitateahighlevelofagreementbetween coders,transcriptsweredoublecodedatpre‐determinedintervals(every20transcripts).After codingwascompleted,thecodersproducedasummaryofcodingissuesandanalyticinsightsfor eachsetofcodes,andvalidatedtheindependentcodingworkoftheothercoder.TheBolivian membersoftheresearchteamreviewedquotesusedinthepaperforaccuracyincapturinglocal meaning.ThebilingualmembersoftheresearchteamwhowerealsonativeEnglishspeakers translatedthequotesfromSpanishtoEnglish. Datafromtheclosedendedquestionnaireswereanalyzedusingunivariateandbivariate statistics.Samplecharacteristicsweredescribedusingpercentagesormeans.Differencesbetween groupsweretestedusingindependentsampleT‐testsforcontinuousvariablesorChi‐squaredtests (orFisher’sexacttestwhereappropriate)forcategoricalvariables. 60 Finally,theinterviewtranscriptdataweresummarizedandarrayedintodatadisplay matrices(RyanandBernard2000,2003)sidebysidewithdatafromtheclosedended questionnairestoidentifypatternsandsalienceofthemes,andfacilitatecomparisonbetweenthe quantitativeandqualitativedata. 61 RESULTS Socio‐economiccharacteristicsandeconomicactivitiescomprisinglivelihoods Table1summarizesthedemographicandsocio‐economiccharacteristicsofthestudy population.Mostparticipantswerewomen,betweentheagesof25‐44,andhadcompletedprimary school.Almosthalfofallparticipantsreportednotworkinginthepastmonth(41%).HIVdiagnosis wasreportedtohaveresultedinachangeofworkstatus(49%)andadecreaseinincome(59%). Womenwerelesslikelytohavecompletedprimaryschool(54%vs.84%,p<0.01)andtohave workedinthelastmonth(51%vs.73%,p<0.01)comparedtomen.Halfofallparticipant householdswereseverelyfoodinsecure,affectingahigherproportionofwomenthanmen(58%vs. 32%,p<0.01). Table1:Demographicandsocio‐economiccharacteristicsofstudypopulation(n=211) Female Agegroup 18‐24 25‐44 45‐64 65+ Householdswithchildren<age18 Meanhouseholdsizenotincl.respondent Primaryschoolormore Workedinthelastmonth WorkchangedasresultofHIVdiagnosis IncomeworsesinceHIVdiagnosis Severefoodsecurity(ELCSAscale) No.ofobservations *p<0.10**p<0.05***p<0.01 62 All Women Men 65% ‐‐ ‐‐ 15% 65% 19% 1% 73% 3.2 64% 59% 49% 59% 49% 211 9% 67% 21% 1% 80%*** 3.4** 54%*** 51%*** 50% 60% 58%*** 137 18% 64% 16% 1% 61%*** 2.9** 84%*** 73%*** 46% 56% 32%*** 74 Whenaskedtoself‐definetheiroccupation,participantsreportedarangeoflivelihoods, whichappearedtoencompassbothpaidandunpaidwork(thoughcompensationwasnotexplicitly referredtointhequestion)(Table2).Topoccupationsreportedbywomenincludedbeinga “housewife”(36%),commercialenterprise(18%),andservices(16%).Commonwomen’sservice activitiesincludeddomesticemployee,sewing,washingclothes,orchildcare.Topoccupationsfor menalsoincludedservices(16%),althoughofdifferenttypesthanwomen,suchasfoodanddrink service,gardening,andtransportation.Manuallaborwasthenextmostcommonoccupation reportedbymen(14%),followedbyhousework(8%)andcommercialenterprise(8%). Inthequalitativedata,participantsreportedmorefrequentandcomplexeconomicactivity thanindicatedinthequantitativeresults.Whilediscussingtheiractivitiesinthelastweek,almost allparticipantsdiscussedengaginginatleastoneeconomicactivityduringthelast7days,in contrasttoresultsfromthequantitativedatawhichsuggestedhalfofthesamplehadnotworkedin thelastmonth.Onethirdofparticipants–roughlythesameproportionsformenandwomen‐ reportedtakingontwoormoreeconomicactivitiesinthelastweek.However,amuchhigher proportionofwomenthanmenreportedtakingonmorethanthreeeconomicactivities.Notably, thistrendwassimilaramongwomenreporting“housewife”astheiroccupationinthequantitative data,indicatingthatbeingahousewifedidnotprecludeeconomicactivity.Womenweremuch morelikelythanmentoreportpiecingtogethervarioussmallworkopportunitiesinordertoearn income.Forexample,onewomaninLaPazdescribedheroverallsetofeconomicactivities: SometimesIwashclothes,orsomebodyhasasmalljobformethatIknow how to do, sometimes cooking…sometimes cleaning…it really depends on whateverpeopletellmetheyneed. 63 Table2:Topoccupationsofstudyparticipants,%(n) All Women Men Housework1 26% 36% 8% Services2 16% 16% 16% Commerce3 14% 18% 8% Manuallabour4 6% 2% 14% Health5 4% 4% 5% Education 3% 1% 5% Arts/Entertainment6 2% 1% 4% Industry/Manufacturing7 2% 0% 5% 1Referstohouseworkforone’sownhousehold,notdomestichouseholdworkforothers.Womentendedto reportthisoccupationasbeingan“amadecasa”,orhousewife.Mentendedtoreportthisoccupationas “laboresdecasa”,orhousework. 2Includesworkasadomesticemployee,foodanddrinkservice,gardening,sewing,washingclothes, childcare,transportation,etc. 3Includesbothentrepreneurialoremployer‐basedcommerce 4Includesconstruction,recycling,laborer‐for‐hire,etc 5Includesmedical/nursingpositions,aswellasbeinganHIVpeercounselor 6Includesartisans,painters,andactors 7Includesindustrialmechanic,garmentmanufacturer,factoryworker,etc. Womenalsoreportedhavingspecificpeopleorbusinessesfromthecommunitytheyrelied onforextraworkforwhentimesweretight,especiallywashingclothesordishes.These arrangementswerenotalwaysforcash–theyofteninvolvedworkinexchangeforfood,asone womaninSantaCruzdescribed: IfIseethere’snofoodinthehouse,Igotohelpoutatthemarket,tothefood stand, to peel vegetables, wash dishes, and then they give me soup and I bring it home so my family can eat… I’m constantly trying to figure out wherethefoodiscomingfrom. Notably, very few people talked about engaging in agriculture as part of their economicactivities.Thosewhodidmentionagriculturewouldepisodicallytraveloutsideof 64 thecitytofamilyfarmsduringpeakplantingandharvesttimes,ratherthanengageinurban farming. Similarly, very few people mentioned using kitchen gardens as a strategy to augmentfoodstoresoralleviatefoodshortages. Dualmanagementoftreatmentandlivelihoods Withinthecontextofthelivelihoodsreportedbyourstudyparticipants,dualmanagement ofHIVtreatmentandlivelihoodsemergedasasalienttheme.By“dualmanagement”wereferto bothhowpeoplemanagedtheirlivelihoodsinlightofHIVtreatmentdemands,andhowpeople managedtheirARTtreatmentregimensgiventhestructureoftheirlivelihoods.Withindual management,wefoundtwomainsub‐themes:negotiatingtimeofffromworkandstayinghealthy atwork,bothstronglyrelatedtoissuesofdisclosureofHIVstatus,andHIV‐relatedstigmaand discrimination. Gettingpermission:“Timeoffforthedoctor?Youlookfine!” Acommonthemediscussedbyparticipantswastheissueofmanagingworkscheduleand HIVstatusdisclosureinthecontextofARTmedicationregimensandmedicalappointment schedules.MostpeoplewithoutsideemploymentreportedthattheyhadnotdisclosedtheirHIV statustotheirsupervisors,nortocoworkers,withnonotabledifferenceindisclosurebygenderor occupation.Fearofdiscrimination,particularlyfearofbeingfired,wasthemostcommonreason givenfornon‐disclosure;however,participantsalsoreportedinternalizedstigma,suchasfeelings ofshameorperceptionsoflowself‐worth,asanadditionalreasonfornon‐disclosure. LackofHIVdisclosureatworkintroducedinternalorinterpersonalconflictintothe workplaceformanyparticipantswhentheyhadtotaketimeofftoattenddoctorappointmentsor pickupmedication.Timeofffromworkcouldonlybesecuredbyasking‘permission’,whichwas notalwaysgiven.Meanwhile,lackofdisclosurepreventedtheparticipantfromexplainingthe importanceoftherequest.Thus,askingforpermissionwasoftenexpressedasastrategicaction– 65 whentoaskforit,howoften,andwhattodoifpermissionwasnotgiven.Insomecases,people inventedotherillnessesorreasonswhyavisittothedoctorwasnecessary;however,this sometimespromptedemployerstodemanddoctor’snotesormedicalhistories,forcingachoice betweenkeepingtheirHIVstatusconfidential(andrisklosingtheirjobfrommissingtoomuch work)ordisclosing(andrisklosingtheirjobbecauseofdiscrimination).Onemanemployedasa technicianinLaPaztalkedabouttheconsequencesofrefusingtosharehismedicalhistorywithhis boss: Ofcourse,[mybosses]wantedtoaskmeformymedicalrecords,sayingthat Igotothedoctorallthetime.Theydidn’twanttogivemepermissiontogo. ButItoldmyboss,‘Idon’thavetogiveyoumymedicalrecords’,becauseit’s true.ButnowI’minabadsituationatworkbecauseofthis. Inadditiontotroublewithbosses,ourparticipantsalsoreportedtroublewithcolleagues, includingjealousyattheperceivedbenefitofadditionaltimeoff.OnewomaninCochabamba discussedhavingadifficultexperiencemanagingco‐workers,bosses,andhertreatmentschedule: Atmywork,theydon’tgivemepermission[togotothedoctor],andalso there’sawoman[coworker]whoisalwayssaying“whydoesshegetso muchtimeoff,what’sthatabout?”Everyonenotices,everyoneasksmewhyI gotothedoctor,butIdon’ttellthemanything.Theysay“You,whatcanyou possiblyhave,sinceyoulookjustfine?”It’sdifficult,havingtoexplain.But evenworsewouldbetotellthem[aboutmyHIVstatus]becauseI’dgetfired inaninstant. Severalpeoplefacingthissituationchosejobswithlessconsistenthoursandlessstabilityin orderavoiddealingwiththeconstantthreatofpunishmentorharassmentforaskingfortoomuch ‘permission’.Forexample,onemaninLaPazreportedchoosingtoworkasadaylaborer,despiteits inconsistency,inordertogainflexibilityaroundhistreatmentschedule: Ido[daylaborwork]inordertocomehere[totheclinic],formytreatment. Idon’tlookforastablejob,becausethenIwouldn’tbeabletobeabsentas much. I have to stay with the work I have, because it gives me more flexibilitytocome,pickupmymedicines,domylabtests,orwhateverIneed 66 todo.Inmyjobsnow,IcanworkwhenIchoose,butinaregularjobIcan’t– ifIevenmissonedaytheyletmegoandI’dbewithoutworkatall. Giventhedifficultyofdualmanagementoflivelihoodsandtreatment,some peopleriskedtheirlivelihoodsbychoosingtoskipworkondayswhentheyknewthey hadtoattendtheclinic.Stillothersriskedtheirhealthbydecidingnottokeeptheir appointmentsorpickuptheirmedsattheallottedtimesinordertoavoidproblemsat work. Stayinghealthyatwork:“Mymedicationsarethemostimportantthing.” Medication‐takingbehaviorwasalsosubjecttoworkplaceissuesrelatedtodisclosurefor participantsinourstudy.SincemostpeoplehadnotdisclosedtheirHIVstatustotheirbossesorco‐ workers,theyhadtoeithertakeARVsinsecret(e.g.thebathroom)orlieandsayARVmedications were‘vitamins’orpillsforacommonillness.Attimes,thedailypill‐takingwasnotedbyemployers andthepersonwasaskedtoprovidemedicalrecordstoassuretheemployerthattheemployee washealthyandcapableofworking. Veryfewpeoplereportedfailingtotaketheirpillsattheprescribedtimeinordertoavoid conflictatwork–rather,peopleimplementedcreativestrategiestoavoidbeingpressuredto discloseorhavetheirregimendetected.Afewpeopledidnotethattheywereabletotweaktheir medicationschedule,inconsultationwiththeirdoctor,tofacilitatepill‐takingrightbeforeorright afterwork,inordertoavoidtakingmedicationsatworkentirely.OnewomaninElAltowho workedinchildcaresharedherexperienceofworkingthroughthedaydespitepainfulmedication sideeffects: I have side effects…nausea, headache, and there are times that my body hurts so much that it leaves me paralyzed for a while and I can’t stand up quicklybecauseithurtsmyfeet.ButluckilyI’mwithoutsupervisionmostof thetimeatmywork,sonoonenoticesanything.Ijustsufferthroughit,stay silent.Theydon’tnotice,andIdon’ttellanyonethatI’minpain. 67 Overwhelmingly, participants in our study reported taking extraordinary measures to maintain adherence and livelihoods despite the barriers they confrontedintheirworkplaces. Structuralbarrierstolivelihoods Gettingtreated:“SometimesIcan’tgetmymedicines” Participantsalsoreporteddifficultieswiththedualmanagementoflivelihoodsand treatmentinrelationtocharacteristicsofthehealthcaresystem.Inparticular,thelimited schedulesandgeographiclocationoftheclinicswhereparticipantsreceivedtreatmentposed significantbarriers.Manyparticipantsnotedthattheclinicsclosedtooearlyatnightoropenedtoo lateinthemorningforthemtoscheduledoctorappointmentsorpickupmedicationswithout conflictingwithworkschedules.Thisincludedbothpeopleworkingasemployeesforothers,who tendedtohavemorefixedschedules,aswellasthoseworkingforthemselves,suchasmarket sellers,whotendedtohavemoreflexibleschedules.Forthelatter,althoughtheydidn’thavethe issuewithaskingforpermissiontoleavework,theirprimesellinghourswereoftenthesamehours neededtoattendtheclinic–thus,takingtimeawaytoattendthecliniccompromisedtheir economicsecurity.Finally,therewasonlyonepublicHIVclinicineachcity,thelocationofwhich wasnotalwayseasilyaccessiblebyallstudyparticipants. Inaddition,somepeoplereportedissueswithlimitedclinicstaffingandresources,which causedthemtowastevaluableworktimewithoutgainingthehealthbenefitsoftreatment.For instance,participantsnotedthatattimesthepharmacywasunstaffedduringopenhoursorwasout ofspecificmedicationsneededbytheclient.Inthesecases,theyhadtoreturnanotherday, requiringyetanotherroundofpermissionorabsencefromimportantlivelihoodactivities,as describedbyonewomanfromSantaCruz: 68 Sometimes when I come there aren’t medicines…like the last two times I camethereweren’tmedicines,andtheydidn’ttellmebefore,sothenIhad toreturnanothertime,yetagain. Earningalivingthroughwork:“There’sjustnojobs…andpriceskeeprising” Participantsinourstudyalsodiscussedbroadereconomicbarrierstolivelihoods,suchas lackofoverallworkopportunities,lackofspecificworkopportunitiesthatareflexibletoHIV‐ relatedneeds,anddifficultiesnegotiatingmarketdynamics. Thelackofoverallworkopportunitiesandwidespreadunemploymentwereissuesofgreat concerntomanyofourstudyparticipants.EvenastheytalkedaboutthespecificwaysthatHIVand ARTaffectedtheirlivelihoods,theyalsorecognizedlackofopportunitiesandunemployment, particularlyforthosewithlowskillsoreducation–assystemicproblemsaffectingpeopleintheir communitiesregardlessofhealthstatus.Youngeradultswithlessworkexperience,peoplewith lesseducation,womenwithyoungchildrenandnochildcare,andolderadultspastprimeworking agereportedparticulardifficultyfindingwork. AppropriateworkwasevenscarcerwhenHIV‐relatedneedsweretakenintoaccount.In particular,participantsreportedmanyproblemsmaintainingjobsinvolvingstrenuouslabor (includingmanuallabororserviceworksuchaswashingclothes,etc.)duetoillnessepisodesor ARVsideeffects.However,locatinglessstrenuouswork–suchasofficejobs–oftenrequiredalevel ofeducation,qualifications,andpersonalconnectionsthatmanyofourparticipantslackedasa resultofsocio‐economicdisadvantage.Someparticipantsalsonotedthatwhiletheircurrentjob didnotalwaysprovideagoodincome,itwasbetterthantheuncertaintyoffindingabetterjob.One maninLaPaznotedthatwhilehe’dliketofindbetter,lessstrenuouswork,searchingforabetter jobwouldbedestabilizingtohishealth: “Ithinkthatlookingforanotherjobwillbehardandcouldharmme–I’llfeel moretired,Imayforgetmymedications,Ifeellikeitwilldoworseforme.“ 69 Studyparticipantswhoworkedprimarilyinentrepreneurialactivities–e.g.artisans, vendors,etc.–alsoreportedstructuraldifficultiesinearningenoughincometosustainthemselves andtheirfamilies.Entrepreneursreportedhavingahardtimefindingmarketsfortheirgoodsthat werebothlucrativeandgeographicallyaccessible.Geographicallyaccessiblemarketstendedto havefiercecompetition,whilegeographicallyfarmarketshaslesscompetitionbutposedother challenges(transportation,abilitytoreturnhomeintimeformeals,attendingclinicvisits,etc).One mansellinggarmentsinSantaCruzdescribedtheintersectingissuesconfrontingsmallsellersin Bolivia: Today,thebiggestproblemiscompetition.Becausethereisalwayssomeonewho willofferthegoodscheaperthanyou.IfIsellat‐shirt,forexample,for50 Bolivianos,someoneelsewilljustsellitfor40.Yetthecostofmaterialsisalways rising–thepricesofinputsjustkeeprisingandwecan’tkeepup. Likethisman,mostentrepreneursinourstudyalsoreportedstrugglingwithsystemic inflation,whichbothraisedthecostofinputsandmadeitdifficultforthemtoselltheirproductsas thepurchasingpoweroftheircustomersdeclined. DISCUSSION ThisisoneofthefirststudiestoexplorehowfoodinsecurepeoplereceivingARTdually managetheirtreatmentregimenswiththeirlivelihoodsinanurban,resource‐limitedcontext,and, toourknowledge,thefirsttoexplorethisissueinLatinAmerica.Wefoundthatourstudy participantshavecomplexeconomiclivesoftencharacterizedbymultipleeconomicactivities, includingbothformalandinformal.Theystruggledtomanagetreatmentandlivelihoods simultaneously,andfacedbarrierstothisdualmanagementthatrangefromtheinterpersonalto thestructural.AnimportantmessagefromourstudyisthatPLHIVoftenfacemultiplelayersof disadvantagewhichmuchbeaddressedbyfoodsecurityandlivelihoodsinterventions.The structuralbarrierscitedbyourstudyparticipants,suchaslackofeconomicopportunities,overall 70 unemployment,andrisingprices,whilenotuniquetoPLHIV,wereintensifiedbyHIV‐specific challengessuchasreducedphysicalstrength,complexmedicationschedules,stigmaand discrimination,andcostsrelatedtotreatment.OftenHIV‐relatedchallengeswerecompoundedby socioeconomicorstructuralchallenges,andviceversa,suggestingthatpeoplereceivingARTin resourcelimitedsettingsrequireinterventionsthatrecognizeandaddressthesemultiplelayersof disadvantageincoordination,ratherthaninisolation. Wefoundimportantdifferencesbetweenthelivelihoodexperiencesofwomenandmen livingwithHIV,especiallyasrevealedinourqualitativedataanalysis.Inparticular,women– includinghousewivesandthosewhoindicatednotworkinginthelastmonthinthequantitative data–reportedveryhighratesofinformaleconomicactivityinthequalitativedata,piecing togethersmall,inconsistentjobs,suchaswashingdishesorclothes.Thesecontrastingfindings fromquantitativevs.qualitativedataforwomenlivingwithHIVareconsistentwithwell‐ establishedresearchdocumentingthedeficiencyofusingstandardlaborstatisticstounderstand women’s‘work’indevelopingcountries(Chen,2001;Donahoe,1999).Programdesignersusing simplebinarymeasuresofemploymenttoassessandthenaddresswork‐relatedneedsintheir targetpopulationmayruntheriskofduplicatingorreplacingtheirparticipants’currenteconomic activitieswithoutsolvingtheactualproblem(e.g.lowfinancialreturnofcurrenteconomicactivity, difficultyfindingclientsormarkets,etc).Wherefeasible,detaileddataonhowmuchtime individualsspendineconomic,householdandhealthcareseekingactivitiesmaybeabettersource ofinformationabouttheexperiencesandneedsofPLHIV,particularlywomen(Appsetal.,2003; d’Addaetal.,2009;Justeretal.,1991).Wheresuchtimeallocationstudiesarenotfeasibleorhave notbeenconducted,wedemonstratethataugmentingbluntmeasuresofemploymentwith qualitativeprobingabouteconomicactivitiesmaygreatlyenhanceunderstandingoflivelihood experiencesforinterventionpurposes. 71 Ourfindingsaddtoagrowingbodyofliteratureexploringtherelationshipbetweenwork, adherenceandstigmainresource‐limitedsettings.Studyparticipantsidentifiednegotiating permissionfortimeofftoattendtheclinic,aswellastakingdailyARVsatwork,asmajorongoing sourcesofconflictwithbossesorcoworkers,towhomparticipantshadalmostneverdisclosed theirHIVstatusbecauseofstigmaandfearofdiscrimination(e.g.gettingfired).Toavoidconflict relatedtoARTschedulesinthecontextofnon‐disclosureandfearoftermination,many participantsinourstudychosetotrade‐offmorelucrativeormorestableworkformoreflexible butlowerpayingandlessstablejobsinordertoaccommodatehealthcareneeds.Althoughnon‐ disclosureofHIVstatushasbeentiedtofearsofstigmaanddiscrimination(Mahajanetal.,2008), thereisstilllimitedpublishedworkexamininghownon‐disclosuretobossesorcoworkersaffects careandtreatmentexperiences,includingadherence,inlowincomecountries(Rachlisetal.,2011). OnestudyonARTadherenceinBotswanafoundthatamongindividualsciting‘frequencyof requiredclinicvisits’asabarriertoadherence,‘can’tleavework’wasthemostcommonlystated reason,althoughtherewasnofurtherexplanationofwhytheindividualfeltunabletoleave(Weiser etal.,2003).Meanwhile,unwillingnesstoaskforpermissionfromemployerstoattendclinicvisits wasidentifiedasabarriertoadherenceamongoutpatientsinBenin(Erahetal.,2008).Morehas beenwrittenabouttheinterconnectionbetweenemployment,disclosuredecisions,andART adherenceinhigh‐incomecountries(Fesko,2001;Glennetal.,2003;Torres‐Madrizetal.,2010; Worthingtonetal.,2012)–however,thesestudiesaresituatedinlegal,healthcareand organizationalcontextsthatareverydifferentthanthoseinlowincomecountries,makingthis literatureoflimitedrelevancetoguideprogramandpolicyforPLHIVinthosecontexts. EmploymentdiscriminationbasedonHIVstatus–includingforceddisclosure,exclusion withintheworkplace,andtermination–isparticularlysevereinLatinAmericaandtheCaribbean (Spragueetal.,2011).NationallawinBoliviaprohibitsHIV‐relateddiscriminationbyemployers, protectingagainstterminationbasedonHIVstatus(PlurinationalStateofBolivia,2007).However, 72 ourdataindicatethattheseprotectionsareeitherlargelyunknowntoourparticipants,or enforcementisnottrusted.Policyoradvocacyeffortsaimedatreducingstigmaanddiscrimination asabarriertolivelihoodsandadherenceforPLHIVaregreatlyneededtoaddressthechallenges identifiedbyourstudypopulation,asisresearchtoinformandevaluatesuchefforts.Policymakers haveastrongroletoplayinimprovinglivelihoodsandtreatmentoutcomesforPLHIVby communicatingandenforcingantidiscriminationlaws.Meanwhile,advocacyorganizationsand NGOsplayanimportantroleinpromotingworkplace,community,andnationalstigmareduction. Broaderbasedstigmareductioneffortsareimportantforthewell‐beingofallPLHIV,but particularlyforthosewhoworkintheinformalsectorandareunlikelytodirectlybenefitfrom formalworkplaceinterventions. Improvementsinthestructureofhealthcarecouldalsoplayakeyroleinalleviatingissues relatedtobothlivelihoodsandARTaccessandadherence.Healthcarebarriersstemmingfrom limitedclinicresourcesaffectedourparticipant’sabilitytogetthecaretheyneededevenwhenthey wereabletogettimeoffwork,andsometimesintensifiedproblemsatwork,affectingproductivity andeconomicwell‐being.Expandingclinicschedules(includingstaffing),improving communicationwithpatients,anddevelopinginitiativestoensuresustainablesupplychainsof essentialmedicinescouldhelptoaddresstheseproblems.Previousresearchinsub‐SaharanAfrica hasfoundthatadaptingpatientappointmentstotheirworkschedulescanreducedefaultfromcare (Pearsonetal.,2006);anotherstudysuggesteda24‐hourclinicasonesolutiontoimprovepatient accessandadherencetoART(Kageeetal.,2011).Ourresultssuggestthatsuchchangestohealth facilitiescouldnotonlyhaveanimportanteffectonadherence,butontheabilityofPLHIVto managetheirworkschedulesandrelationshipswithemployersvis‐à‐visART.Thus,froma livelihoodsperspective,improvingthehumanandfinancialresourcesavailabletooverburdened healthfacilitiesremainsanimportantpolicygoal. 73 Researchershavebeguntodescribeandevaluateintegratedhealthandlivelihoods programsinthescientificliterature,butmuchisstillunknownwithregardstooptimalprogram andpolicydesign(Yageretal.,2011).LivelihoodinterventionsforPLHIVrelevanttofoodinsecure, urbanpopulationshavebeendocumentedinresource‐limitedsettings,andincludeskillstraining, linkingtobasiceducationopportunities,group‐basedeconomicactivityamongPLHIV(e.g. restaurant),accesstofinancialinstruments(credit,savings,etc),subsidizingbusinessinputsand materials,identifyingmarketsandmarketingstrategies,jobplacement,reducingstigmaand discriminationintheworkplace,andurbanagricultureorkitchengardens(S.Gillespieetal.,2005; Holmesetal.,2011;Roopnaraineetal.,2011;Samuelsetal.,2011).Inonemixedmethodsstudyin ZambiaandKenya,theauthorssuggestthatleveragingexistinglivelihoodnetworks,providing skillstrainingandfacilitatingassetaccumulationarethemostpromisingapproachestosupportthe foodsecurityofpeopleonART(Samuelsetal.,2011).Keyassetstotargetincludefacilitating ownershipofproperty/land,providingaccesstoagarden/urbanfarm,andpromotingsavings. Otherstudiesfocusonprogrammaticaspectsoflivelihoodsinterventions–i.e.howtobestinvolve communitypartners,theroleofgroup‐basedvs.individuallivelihoodactivities,andconsiderations fortransitionstrategiesfromfoodassistance(Kadiyalaetal.,2009;Roopnaraineetal.,2011;Yager etal.,2011). OurstudyaddstothisliteraturebydescribinghowfoodinsecurePLHIVinanurbanLatin AmericancontextmanagetheirlivelihoodsinthecontextofARTandviceversa,includingthe primarybarrierstheyfacedandtheirresponsetothesebarriers.Thisiscrucialinformationforthe developmentofeffective,well‐targetedlivelihoodsinterventions,whichshouldbebasedoncareful assessmentsofthecurrentsetofeconomicactivities,skillsandbarriersexperiencedbythetarget populationinordertomeetparticipantneeds.Eachlivelihoodinterventionorpolicyhaspotential advantages,disadvantages,likelytargetgroup,andspecificrolesforpoliciesandprograms.In Table3,weexplorethevarietyoftheseimplicationsbasedonourstudyresultsandourreviewof 74 theliteraturepresentedinthispaper.Forexample,ourstudysuggeststhataprogramfocusedon “skillstraining”isunlikelytohelppeoplewhoarealreadyskilledbutwhoavoidmorestablework opportunitiesinordertocomplywithARTtreatmentscheduleswithouthavingtodiscloseHIV statusintheworkplace.Meanwhile,trainingorfinancingPLHIVinclinedtowardsentrepreneurship tostartnew(orimprovecurrent)enterprisescouldhelpthemtogainmoreflexibilitytomanage theirtreatmentandlivelihoods.However,thisapproachwillrunintosustainabilityissuessuchas findingappropriatemarketsandbeingabletoaffordinputsinthelongrunwhichmustbetaken intoconsiderationupfront.Therelativescarcityofparticipantsinourstudywhoreported engaginginurbanfarmingorgardeningimplythatinvestinginfood‐relatedlivelihoodsstrategies maybeamoresustainablewaytosupportlivelihoodsandfoodsecurityinthecontextoftreatment. Supportingaccesstourbangardensorfarms,aswellaslandownership,mightbeaproductive strategyforsomepeople,particularlywomenwhoexperiencehigherhouseholdfoodinsecurityand lessstableemployment.Finally,broaderinterventionstoreducestigmaandimprovehealthcare accesscouldremovekeybarrierstoproductivelivelihoodsforpeoplereceivingART,reducingthe needfortargetedlivelihoodinterventionsforsomeindividuals. 75 Table3:ImplicationsoflivelihoodexperiencesforintegratedHIVandlivelihoodsinterventions Policyor Intervention Skillstraining Linktobasic education opportunities Group‐based(HIV) economicactivity (e.g.restaurant) Improveaccessto financial instruments(credit, savings,etc) Provide/subsidize businessinputsand materials Advantages Disadvantages Whoislikelytobenefit Roleforprogram/policy Iftargetedwell,canhelpto createcurrentopportunities, includingtoprocuremore profitable,lessstrenuousor moreflexibleemployment Canbeunproductiveandcan wastetime/resourcesifthereis noclearmarkettousenewskills; takestimeawayfromcurrent livelihoods Peoplewhodesirespecific skillstoboostorchangetheir livelihoods;peoplewhohave lowskillsorlowliteracy Increasesfundamentalskills (e.g.literacy)tosupport economicopportunityand empowerment;boostsfeeling ofself‐worth Couldcreatelessstrenuous andmoreflexibleemployment toeaseconflictswith treatmentschedules;could circumventHIV‐related stigmaintheworkplace Providesaccesstokeyassets thatcouldboostlivelihoods (e.g.property,land,inputs, etc). Takestimeawayfromcurrent livelihoodactivities;important forhumandevelopmentand overalleconomicwell‐being Peoplewhoneverfinished primaryschool;peoplewith lowliteracy Supportsexisting entrepreneurstolowertheir costsofbusiness;helps beginningentrepreneursto getofftheground Subsidyorinputtargetedata specificsectororproductmay promptpeopleintoalivelihood areathatisn’tlikelytobe successful;doesn’taddress managingtreatmentwiththe demandsofentrepreneurship Directlyprovidetraining, orhelpplacepeoplein appropriateprograms; supportthemtofind opportunitiesthatwilluse theirnewskills Helpplacepeoplein appropriateprograms; supportthemtofind opportunitiesthatwilluse theirnewskills Supportformationof group;supportbusiness planing;potentially provideand/orsubsidize property,business resources,orinputs Linktocurrent microfinanceinstitutions orprovideownservices. Beawareofwhat communityfinancial servicesalreadyexist,and whatotherNGOshave donebefore. Targetsubsidyorinput provisiontothoseableto useit;provideenough flexibilityforthebenefitto beuseful; Canrequirefinancial contributionbyindividual withoutindividualcontrol; sustainabilitycanbeanissue; difficulttoscaleup PLHIVwhoseeksolidarity withotherPLHIV;peoplewho havedisclosedtheirHIV status;peopleforwhom stigmaatworkhasbeena strongbarriertolivelihoods Repaymentcanbestressfuland Forcredit,peoplewhohave uncertain,especiallyincontextof reasonableabilitytopayback HIV;dependingonlender, credit,butnotenough interestratemightbe collateraltoobtainabank prohibitive/harmful loan;Forsavings,people withoutcurrentsavings 76 Currentorpotential entrepreneurs Policyor Intervention Identifyingmarkets andmarketing strategies Jobplacement Decrease stigma/discriminati onintheworkplace Urbanagriculture/ kitchengardens Advantages Disadvantages Whoislikelytobenefit Supportsentrepreneurstobe moresuccessfulatearning incomethroughtheirexisting activities Alleviateslivelihoodsbarriers duetolackofcontactsor exposuretotheright employmentmarkets;ability toplacepeopleinjobswith structuresthateasehealth caredemands Easesdisclosureasabarrier toworkandproductivity; facilitatescooperationwith treatmentschedulewith employers Requiresacloseunderstanding ofthelocaleconomyandculture Currententrepreneursor Conductmarket peoplebeingtrainedwith assessments;train skillstobecomeentrepreneurs participants (e.g.makingcrafts) Can’tdomuchagainstsystemic unemploymentorlackofjobsin community;availabilityof appropriatejobsmaybeslim; maynotaddressHIV‐related needslikeflexibility,lessphysical intensity Difficulttodo;long‐termcultural changerequiredtohave widespreadeffect Peoplewithmoreeducationor Usecommunityandcivil skills;peoplewhohadtoleave societynetworkstohelp ajobafterHIVdiagnosisbut placePLHIVintojobs arereadytoreturntowork Directlyaddressesfood sufficiencyandnutrition;can promotebothcashflow(sale ofproducts)and consumption;potentially scalable Needforinputsandland;must workaroundclimate;mustwork aroundlanduselaws;takestime awayfromotherwork 77 Peoplewhoalreadyhave formalemploymentorwhoare likelytoworkintheformal sector Peoplewithavailablelandor spaceforagarden;people withoutacurrentfulltime livelihood,primarilywomen Roleforprogram/policy Enforceantidiscrimination laws;holdworkshopsor interventionstosensitize employerswithout disclosingHIVstatusof employees;providework‐ relatedsupportgroupsor workshopsforPLHIVto addressstigmaand discrimination Facilitateaccesstoland useorlandownership; providetrainingonurban gardening;subsidize inputs Ourstudyhasseverallimitations.Oneprimarydrawbackisthatwewereunableto interviewpeoplebeforetheystartedreceivingfoodassistance,soitispossiblethattheadditionof thefoodbaskettotheirhouseholdsaffectedtheirreportsoflaborparticipationandworkdecisions severalmonthslater.However,studiesontheimpactoffoodassistanceonlaborsupplyinother settingsareinconclusiveastowhetheraneffectexists,andinwhatdirectionitgoes(Tirivayietal., 2011a,b).Inaddition,wedidnothavereliableclinicaldataonhealthstatusandtimeonARTforall participants.Therefore,wewerenotabletoexploreissuesdirectlyrelatedtoatransitionprocess– i.e.adjustingtolifeon,ART,changesinhealth,andtheroleoflivelihoodswithinthat.Itislikelythat peopleindifferentstagesoftreatmentmayrequiredifferentkindsoflivelihoodsupportor interventions,astheyregaintheirhealthandadapttolivingwithHIVasachroniccondition. CONCLUSION Itisessentialthatpolicymakerstaskedwithcreatingandenforcingworkplaceandsocial protectionpoliciesforPLHIV,aswellaspromotingARTaccessandadherence,understandhow PLHIVmanagetheirtreatmentwithregardstotheirlivelihoodsandsocioeconomicsituations,the barrierstheyface,andthestrategiestheyusetoovercomethem.Policyinitiativesshouldaddress specificbarriersidentifiedbyPLHIVandimprovetheopportunitiestheyhavetothriveinall aspectsoftheirlives,includingintheirlivelihoods.ComprehensiveHIVcareandsupportpackages mustintegratehealth,socialandeconomiccomponentsthataresupportedbystrongnationalHIV policyandlinktonationalsocialprotectionandsocialsafetynetprograms.Inthehighlyunequal contextofLAC(UnitedNationsDevelopmentProgramme,2010),wherethemostvulnerable populationsfacesevereeconomicandfoodinsecurity,itisevenmorecrucialthatlivelihoodsbe addressedincoordinationwithhealthandsocialprogramming. 78 III.Roleofantiretroviraltherapyinimprovingfoodsecurity amongpatientsinitiatingHIVtreatmentandcareinUganda ABSTRACT Background:AlthoughthephysicalhealthbenefitsofHIVantiretroviraltherapy (ART)arewelldocumented,thedirectandindirecteconomicandnutritionalbenefitsof ARTarestillbeingestablished.FewstudieshaveexaminedifandhowARTaffectsfood insecurity,althoughthescientificliteraturesuggeststheremaybeabenefitviaimproved healthandabilitytowork. Methods:Usingdatafroma12‐monthprospectivecohortstudy,weemploy multivariatelongitudinallogisticregressiontoinvestigatewhetherARTdecreasesfood insecuritycomparedtoHIVcarewithoutARTamongasampleof602treatment‐naïve patientsinitiatingclinicalcareinUganda.Wethenuseastagedregressionapproachto explorethepotentialpathwaysthroughwhichARTmayaffectfoodinsecurity,including improvedmentalhealth,physicalhealth,andworkstatus. Results:WefindthatfoodinsecuritydecreasedsignificantlyforboththeARTand non‐ARTgroupsovertime,withtheARTgroupexperiencinggreaterreductionsbytheend ofthestudy.ARTremainedasignificantpredictorofreductioninfoodinsecurityovertime aftercontrollingforbaselinedifferencesinthemultivariatelongitudinalregressionmodel (OR=0.642;p<0.01).Improvementsinworkandmentalhealthstatusweremoststrongly associatedwithdecreasedfoodinsecurityovertimeandweakenedtheARTcoefficient significantlywhenaddedtothemodel. 79 Conclusion:Takentogetherwiththewell‐knownbenefitsoffoodsecurityonART adherence,treatmentretentionandclinicaloutcomesinresource‐poorsettings,ourresults suggestan“upwardspiral”ofimprovedfunctioningandproductivitycouldresultfrom positivefeedbackbetweenfoodsecurityandART.Policymakerscouldleveragethis positivecyclebystrengtheningmentalhealthsupportandpromotingsustainablefood securityinterventionsaspartofHIVtreatmentprograms. 80 INTRODUCTION ThephysicalhealthbenefitsofHIVantiretroviraltherapy(ART)arewell documented(Bartlettetal.,2001;Murphyetal.,2001).However,thedirectandindirect economicandnutritionalbenefitsofARTarestillbeingestablished(Beardetal.,2009). AbundantstudieshaveillustratedthenegativeeffectthatHIVcanhaveontheeconomic well‐beingandqualityoflifeofindividualsandhouseholdsviadecreasedphysicaland mentalhealth,includingreducedincome(McIntyreetal.,2006;Russell,2004),reduced productivecapacity(B.Larsonetal.,2008),problemsofabsenteeism(Rosenetal.,2004), andconsequentialjobloss(Rosenetal.,2004;Russell,2004).Todealwithlostincome, individualsandhouseholdsmayturntocopingstrategies(suchasincurringdebt,selling assets,orexhaustingsavings)thatunderminebothcurrentwell‐beingandfuture livelihoods(McIntyreetal.,2006).Foodinsecuritycandeepenasincome,productivityand assetsdecreaseduetonegativeconsequencesofHIV(Crushetal.,2011;Marstonetal., 2004).SeveralrecentstudiesinUgandahavedocumentedfoodinsecurityasaserious consequenceforHIV‐affectedhouseholds(Bukusubaetal.,2007;Tsaietal.,2011;Weiseret al.,2010).Foodinsecuritycanbedefinedaslimitedoruncertainavailabilityofnutritionally adequateandsafefoods,orinabilitytoacquirethesefoodsinsociallyacceptableways (Radimeretal.,1992);foodsecurityisachievedwhenthereisphysicalandeconomicaccess toadequatefoodforallhouseholdmembers,withoutriskoflosingsuchaccess(Haeringet al.,2009). FoodinsecurityisassociatedwithworseimmunologicstatusatARTinitiation(i.e. CD4count)(Normenetal.,2005;Weiseretal.,2009a),poorARTclinicaloutcomes(e.g. virologicalsuppression,morbidityandmortality)(Weiseretal.,2009b;Weiseretal.,2009c; Weiseretal.,2012),andisakeybarriertoARTaccess,adherenceandtreatmentretention 81 inresource‐poorsettings(Anemaetal.,2009;Deribeetal.,2008;Frankeetal.,2011; Marcellinetal.,2008;Martinetal.,2011b;Weiseretal.,2012).Lessisknownabout whetherandthroughwhatmechanismsARTaffectsfoodinsecurity,althoughthefew studiesexaminingnutrition‐relatedARToutcomessuggesttheremaybefoodsecurity benefitstoART.OnestudyinKenyafoundpreliminaryevidencethatchildren’snutrition improvedinhouseholdswhereadultswerereceivingART(Zivinetal.,2009),whileastudy inIndiafoundthatconsumptionofkeylocalfoodgroupsincreasedovertimeforpatientson ART(Thirumurthyetal.,2008a).Meanwhile,narrativesfromarecentqualitativestudyon foodinsecurityandARTadherenceinruralUgandasuggestedARTmayalsoreducefood insecuritybyrestoringeconomicproductivitytopatients(Weiseretal.,2010). Conceptualframework Basedontheliterature,wedevelopedaconceptualframeworktoexplainthe possiblepathwaysbywhichARTmayindirectlyaffectfoodsecurity(Figure1).Usingthis framework,wehypothesizethatARTwillinfluencefoodsecurityviatheprimarypathways ofimprovedmentalhealth,improvedphysicalhealth,andanincreasedabilitytoworkand conductdailyactivitiesasaconsequenceoftheseimprovements. Foodsecurity,bydefinition,iscloselyrelatedtotheabilitytoconducteconomic activities(work,ordailyactivitiesrelatedtofulfillbasicneeds),whichfacilitatesthe acquisitionofasufficientquantityandqualityfoodtomeetnutritionalneeds.Anincreasing bodyofscientificstudies,conductedmainlyinsub‐SaharanAfrica,suggeststhat improvementsinphysicalandmentalhealthduetoARTcanleadtoimprovedeconomic well‐beingviaincreasedlaborproductivityandabilitytoconductdailyactivitiesforbasic needs(Beardetal.,2009;Bocketal.,2008;d’Addaetal.,2009;Jelsmaetal.,2005;Rosenet al.,2010).ARThasbeenassociatedwithimprovedworkperformance,increasedwork 82 hourss,increasedo oddsofbeinggeconomicalllyactive,as wellasaretturntoworkamongthose previouslyemployedindevelo opingcountrries(B.Larsoonetal.,2008 8;Thirumurthyetal., hyetal.,2008b).Inaddition,workab bsenteeismap 2011;Thirumurth ppearstosiggnificantly Larsonetal.,,2008).Beyo ondthe decreeasewithinittiationofART(Eholieetaal.,2003;B.L formaallabormark ket,time‐useeevidencefro omAfricaind dicatesthatA ARTmayalsohelppeoplle withHIVtoreturn ntoproductiive(butunpaaid)activitieesrelatedtoffoodprocureementand hasgatheringgfirewoodan ndwater(d’A Addaetal.,2 2009) prepaaration,such Figurre1:Concep ptualframew workofpath hwaysbetw weenARTan ndfoodsecu urity Groupdiffferencesink keycharacterristicsassoci atedwithfoo odinsecurityycouldaffectt itsrellationshipw withARTandpotentialpathwaysinou urframework k.Theseincllude differrencesineconomicresou urces,suchassfinancialorrassetwealth h(Alaimoetal.,2001b; Leeetal.,2001;M Misselhorn,2005;Normen netal.,2005 5;Rose,1999 9),materialsupport withinsocialnetw works(i.e.giiftsoftangiblleresources,,withorwith houtexpectaationof 83 reciprocity)(Kaschula,2011;Tsaietal.,2011),demographiccharacteristics(gender,age, education,householdsize,urbanvs.rurallocation,headofhouseholdstatus)(Anemaetal., 2009;Crushetal.,2011;Leeetal.,2001;Roseetal.,2002),andHIV‐relatedhealthstatus (CD4count,AIDSdiagnosis)(Normenetal.,2005;Weiseretal.,2009a).Clearly,the provisionoffoodsupplementationtiedtoHIVcareandtreatmentwouldalsobehighly likelytoaffectthefoodinsecurityofpatientsreceivingit,althoughthescopeandmagnitude ofthebenefitsofsuchprogramsarefarfromconclusive(Tirivayietal.,2011a). WeusedatafromaprospectivecohortstudyinUgandatoexaminehowthefood insecurityofatreatment‐naïvepopulationchangesoverthefirstyearofHIVcare.Wethen investigatetheaddedeffectsofARTbycomparingthefoodsecurityofpatientsonARTwith thatofpatientsnotyetonARTovertime,andexplorethepotentialpathwaysthrough whichARTmayaffectfoodinsecurity. METHODS StudyDesign&Sample TheJointClinicalResearchCenter(JCRC)/RANDprospectivecohortstudy(January 2008–November2009)wasdesignedtodeterminetheeffectofARTonmultiplehealth outcomes,includingphysicalandmentalhealth,aswellassocio‐economicoutcomes.New patientsattwoclinicsinUgandawereconsecutivelyrecruitedintothestudyiftheyfulfilled thefollowingcriteria:(1)adultsover18;(2)justenteredcareandbeenassessedforART eligibility;(3)CD4<400cells/mm3ifnotyeteligibleforART.TheprimarycriterionforART eligibilitywashavingeitheraCD4count<250cells/mm3orWHOstage3or4disease (AIDSdiagnosis);thepatientmustalsohavedisclosedtheirHIVstatustosomeonecloseto them,whocouldthenserveasthepatient’sself‐identified“treatmentsupporter”(a 84 commoncriterionforARTprescriptionacrosssub‐SaharanAfrica).Thetwoclinicswere bothoperatedbyJCRC,oneinKampalawhichisthecapitalandtheonlylargeurbanareain Uganda,andoneinKakirawhichisaruraltownabout100kilometersawayfromKampala. Importantly,HIVcareandtreatmentattheJCRCclinicsdidnotincludefood supplementation,eitherdirectlythroughtheclinicorthroughanexternalorganization. Participantswereinterviewedthreetimes(baseline,6and12months)andpaid 5000UgandanShillings($2.50)foreachassessment.Writteninformedconsentwas obtainedfromallparticipants.TheinterviewprotocolwasapprovedbytheInstitutional ReviewBoardatRANDandJCRCinUganda. Measures Allquestionnaire‐basedmeasuresintheprotocolwerecollectedinface‐to‐face interviewsadministeredinLuganda,thenativelanguageoftheparticipants.Theprotocol wastranslatedintoLugandausingstandardtranslationandback‐translationprocedures. Clinicaldatawereabstractedfromclinicalchartsusingstandardizedforms.Forallscale‐ basedmeasures,higherscoresrepresentgreaterlevelsoftheconstruct. Foodinsecurity.Foodinsecuritywasassessedusinga5‐itemscaleadaptedbythestudy teamfromtheU.S.HouseholdFoodSecuritySurvey(USHFSS)Module6‐ItemShortForm (Bickeletal.,2000).Theadaptedscaleassessedindividualfoodsufficiencyandaffordability inthecontextofhouseholdresourcesbyaskinghowoftentheindividualhadtocuttheir mealsize,eatless,feelhungry,skipameal,ornoteatforawholedaybecausethe householddidnothaveenoughmoneyforfoodinthelast30days.Possibleresponseswere “alot”,“sometimes”,“never”,and“don’tknow”.Rawfoodinsecurityscores[0‐5]were calculatedbysummingaffirmativeresponses(“alot”or“sometimes”)andthencategorized 85 into3levelsusingUSDAclassifications:“marginalfoodsecurityorbetter”(0‐1);“lowfood security”(2‐3);and“verylowfoodsecurity”(4‐5)(USDA,2008).Thescalereliability coefficientfortheadaptedscale(Cronbach’salpha)atbaselinewas0.92,indicatinghigh internalconsistency.Foranalysis,weconstructedabinaryvariablerepresentingsevere foodinsecurity,indicatingwhetherornotthepatienthad‘verylowfoodsecurity’. Pathwayvariables Mentalhealth.Depressionwasassessedusingthe9‐itemPatientHealthQuestionnaire (PHQ‐9),whoseitemscorrespondtosymptomsofmajordepressionfromtheDSM‐IV(e.g. feelingdownorhopeless,troublesleeping,etc).Therangeofpossiblescoresis0–27,with higherscoreindicatinghigherdepression(clinicaldepressionisindicatedwithascoreof10 orhigher).ThePHQ‐9hasbeenusedsuccessfullywithHIVclientsinotherpartsofsub‐ SaharanAfrica(Adewuyaetal.,2006;Monahanetal.,2009). Physicalhealth.Physicalhealthwasassessedusingthe2‐itemsub‐scaleforrolefunctioning oftheMedicalOutcomesStudyHIVHealthSurvey(MOS‐HIV)(A.Wuetal.,1997),whichhas beenspecificallyadaptedforUganda(Mastetal.,2004).Therolefunctioningsub‐scale assesseswhetherhealthhaslimitedtheparticipantfromworkingatajoboraroundthe house,andwastheMOS‐HIVphysicalhealthmeasureclosesttotheconstructofphysical healthwewereinterestedin.Rawscoresrangefrom0–2. Workstatus.Currentworkstatuswasabinaryvariabledefinedashavingworked(other thanhousework)inthelast7daysbasedonself‐reportasmeasuredbymodulesofthe WorldBankLivingStandardMeasurementSurveys(Groshetal.,2000). 86 Keycovariates. Assetwealth.Weusedthemethodofprincipalcomponents(Filmeretal.,2001)tocreatean assetindexbasedonrelativelyliquidformsofwealth(cellphone,TV,radio,motorized vehicle,andlivestock)thatmightbeexpectedtochangewithcurrenteconomic circumstances(Linnemayretal.,2011).Thismethodcreatesasingleindexfromthese multipleindicatorssuchthatitexplainsthehighestproportionofobservedvariance betweenindividualsrelativetootherpossibleindexes. Materialsupport.Materialsupportwasassessedusingtwoquestionsaskingifthe participanthadreceivedfoodorfinancialsupportfromanysourceinthelastmonth. Responseswerecombinedtocreateabinaryvariableindicatingmaterialsupportequalto oneiftheparticipanthadreceivedsupportofeitherkind. HIV‐relatedhealth.HIV‐relatedhealthwasassessedusingcontinuousCD4count(cells/µL) andWHOHIVdiseasestage.AIDSdiagnosiswasdeterminedasabinaryvariableequalto oneforpatientsatWHOHIVdiseasestage3or4.Thesemeasureswereabstracted manuallyfromthepatient’sclinicalchart. Demographics.Demographiccharacteristics(gender,age,education,headofhousehold, householdsize)wereassessedbasedonmodulesoftheWorldBankLivingStandard MeasurementSurveys(Groshetal.,2000).Ourmeasureofhouseholdsizeexcludedthe participant. DataAnalysis AnalyseswerebasedoncomparisonsoffoodinsecurityacrosstheARTandnon‐ ARTgroupsatbaselineandovertime.Wefirstusedbivariatestatistics(Chi‐squaretest,two 87 samplet‐test)tocomparethebaselinecharacteristicsoftheARTandnon‐ARTgroups,as wellasacrossfoodinsecuritygroups.Toexaminechangeovertime,wevisuallyexplored trendsintheoutcomeandpathwayvariables,andtestedforstatisticallysignificant differencesoverthethreeassessmentsbyARTstatus(pairedt‐test,McNemar’stest). Ourprimaryanalysiswasanintention‐to‐treat(ITT)approachthatincludedall participants.Weconductedmultivariatelongitudinallogisticregressiontoinvestigatethe effectofARTonfoodinsecurityoverthethreeassessments,wherethedependentvariable wasthebinaryvariableindicatingseverefoodinsecurityandthemainindependent variableswereARTstatusatbaseline,time(ordinalvariablerepresentingthethree assessments),andaninteractiontermofARTstatusbytime.Wethenusedastaged regressionapproachtoexplorethepotentialexplanatoryroleofthepathwayvariables identifiedinourconceptualframework.Inthefirststep,weanalyzedtheregressionmodel forfoodinsecurityusingthemainindependentvariableslistedabove.Insubsequentsteps, wereexaminedthemodelswhileaddinginthepathwayvariablesonebyone(changein depression(inverse),workstatus,androlefunctioningfrom0to12months).Changein depressionwasenteredintotheregressionasaninversechangesothatanincreaseinall pathwayvariableswouldindicateimprovement.Ourfinalmodelincludedallhypothesized pathwayvariablestogether.Weimplementedtheregressionusingthegeneralized estimatingequation(GEE)methodforanalysisofcorrelatedrepeatedmeasurements(Zeger etal.,1988),usingsemi‐robuststandarderrorsandassumingabinomialdistributionfor thebinaryfoodinsecurityoutcome. Inallregressions,wecontrolledforbaselinecharacteristicsidentifiedinour conceptualframework(female,Kampala,CD4count,foodsecurityscore,workstatus,role functioningscore,depressionscore,assetindexscore,andreceiptofmaterialsupport),as 88 wellasasetofindicatorvariablesforthemonthofinterviewtotakethepossibilityof seasonalfoodinsecurityintoaccount.Forparsimony,weexcludedage,householdsize,and headofhouseholdstatusfromtheregressions,sincetheyneitherdifferentiatedtheART groupsorfoodsecuritygroupsinbivariateanalysis(Table1). Allanalysesincludedattritionweightstoaccountfordropoutfromthestudy,which werederivedvialogisticregressionusingcompletionstatusastheoutcomeandbaseline measuresassociatedwithARTandcompletionstatusastheindependentvariables.All statisticalanalyseswereconductedinSTATA/IC11.1(StataCorp:CollegeStation,Texas). Sensitivityanalyses Weconductedtwosensitivityanalysestochecktherobustnessofourlongitudinal regressionresults.Insensitivityanalysis1,wechangedourITTanalysistoan“astreated analysis”byexcluding50patientsassignedtothenon‐ARTgroupatbaselinewhostarted ARTduringthecourseofthestudy.Insensitivityanalysis2,weaddressedtheissueof comparabilitybetweentheARTandnon‐ARTgroupsbyrestrictingthenon‐ARTgroupto thosepatientswithCD4count(<250)oranAIDSdiagnosis,whichwouldnormallyqualify themforARTbutwhereARThadbeendeferredforothermedicalorpsychosocialreasons. Inaddition,were‐ranouroriginalmodelswithalternatemeasuresofphysical healthstatus,includingthephysicalhealthfunctioningandoverallhealthsub‐scalesofthe MOS‐HIV,toexplorewhetherchoiceofmeasureaffectedourresults. 89 RESULTS Samplecharacteristics Thesampleconsistedof602participants,including300ARTand302non‐ART patients,distributedevenlybetweenKampalaandKakira.Retentioninthestudywasvery high–92%oftheARTgroupand94%ofthenon‐ARTgroupcompletedthe12‐month assessment.BaselinecharacteristicsofthetotalsamplebyARTstatusandfoodinsecurity statusaregiveninTable1.Atbaseline,theARTgrouphadworseHIV‐relatedhealth(lower CD4count,moreAIDSdiagnosis),lowereducation,lowerassetwealth,lowerhealth functioning,andhigherdepressioncomparedtothenon‐ARTgroup.TheARTgroupwas lesslikelytobeworkingthanthenon‐ARTgroup.Age,gender,andhousehold characteristicsdidnotdifferbetweentheARTandnon‐ARTgroups. Foodinsecurity Atbaseline,50%ofparticipantshadseverefoodinsecurity,withameanfood insecurityscoreof2.8.Inbivariateanalysis,theARTgrouphadahigherprevalenceof severefoodinsecurity[54%]comparedtothenon‐ARTgroup[46%][p<0.05],although theirmeanfoodinsecurityscoredidnotdiffer(Table1).Meanwhile,participantswith severefoodinsecurityatbaselineweremorelikelytobefemale,inKakira,havelower educationalattainment,andhavefewerassets.Theywerealsolesslikelytoreportworking inthelast7days,hadlowerhealthfunctioning,andexperiencedhigherdepression(Table 1). Examiningunadjustedchangeovertime,theprevalenceofseverefoodinsecurity decreasedforboththeARTandnon‐ARTgroupoverthefirst12monthsoftreatment, althoughthistrendwasmorepronouncedfortheARTgroup(Figure2).IntheARTgroup, 90 foodinsecuritydecreasedsignificantlyfrombaseline[53%]toMonth6[37%;p<0.001], andagainatMonth12[13%;p<0.001].Thenon‐ARTgroupexperiencedasimilar reductionfrombaseline[46%]toMonth6[33%;p<0.001],andagainatMonth12[18%;p <0.001].AtMonth12,theprevalenceoffoodinsecuritywaslowerfortheARTgroup comparedtothenon‐ARTgroup[p<0.05]. .1 Severe food insecurity .2 .3 .4 .5 .6 Figure2.Trendinprevalenceofseverefoodinsecurity,byARTstatus 0 6 Month Non-ART 95% CI 91 12 ART Table1:BaselinesamplecharacteristicsbyARTandfoodinsecuritystatus All Non-ART ART Not severely food insecure Severely insecure Food insecurity raw score, mean (SD) 2.8 (2.08) 2.7 (2.04) 2.9 (2.13) 0.9 (1.18) *** 4.8 (0.47) *** Severe food insecurity 50% 46%** 54%** Kampala 50% 50% Age, mean (SD) 36 (8.5) Female Food Security Status - - 50% 57% *** 43% *** 36 (8.6) 36 (8.3) 36 (8.6) 36 (8.3) 69% 70% 67% 64% ** 73% ** Head of household 66% 67% 66% 66% 67% More than primary education 46% 49%* 43%* 55% *** 36% *** Household size, mean (SD) 3.4 (2.6) 3.4 (2.7) 3.5 (2.5) 3.5 (2.7) 3.4 (2.5) 175 (117) 301 (73) *** 126 (83) *** 217 (119) 210 (116) 45% 30% *** 60% *** 43% 47% 1.4 (0.92) 1.6 (0.81)*** 1.1 (0.98)*** 1.3 (0.91) 1.4 (0.93) 5.2 (3.93) 4.4 (3.64) *** 6.1 (4.02) *** 4.8 (3.87) *** 5.6 (3.94) *** Receives material support 43% 42% 45% 45% 41% Worked in last 7 days 58% 69% *** 47% *** 65% *** 51% *** Asset index score, mean (SD) - 0.19 (0.99) - 0.12 (0.97) ** - 0.26 (1.0) ** -0.52 (1.03) *** 0.14 (0.81) *** 602 302 300 300 302 Demographics Physical health CD4 count, mean (SD) AIDS diagnosis Role functioning score, mean (SD) Mental health Depression score, mean (SD) Socioeconomic characteristics No. of observations *** p < 0.01 ; ** p < 0.05 ; * p < 0.10 Notes: (a) Chi-square and independent t-tests used to compare group differences. Significant indicators compared the ART and non-ART groups, and the severely and not severely food insecure groups 92 Inthemultivariatelongitudinalregressionmodeloffoodinsecurityover12months, thesignificantoddsratiosonboth‘Time’[OR=0.352;p<0.001]andtheinteractionof‘ART XTime’[OR=0.642;p<0.01]indicatethatbothtimeandARTweresignificantpredictors ofdecreasedoffoodinsecurityaftercontrollingforbaselinedifferences(Table2,Column 1).WecaninterprettheORon‘ARTXTime’tomeanthatcomparedtothenon‐ARTgroup, theARTgrouphadalmost36%loweroddsoffoodinsecurityateachassessment, suggestinganadditionaleffectofARTonfoodinsecurityaboveandbeyondthenon‐ART group.Inaddition,giventheoverallsignificantreductioninfoodinsecurityinboththeART andnon‐ARTgroups,togetherwiththefactthatallparticipantsbeganreceivingHIVcareat studyinitiation,weareinclinedtoviewtimetrendforthenon‐ARTgroupasindicatingthe effectsofHIVcareratherthanaseculardecreaseinfoodinsecurity.However,withouta controlgroupofPLHIVnotreceivingHIVcare,wecannotformallytestthishypothesis. 93 Table2:Longitudinallogisticregressionresultsonfoodinsecurity (1) ART group Odds of severe food insecurity (95% CI) (2) (3) (4) (5) 0.725 0.769 0.943 0.801 0.957 (0.471 - 1.117) (0.487 - 1.213) (0.602 - 1.479) (0.523 - 1.225) (0.617 - 1.484) 0.352*** 0.357*** 0.351*** 0.351*** 0.345*** (0.276 - 0.450) (0.280 - 0.455) (0.274 - 0.449) (0.274 - 0.449) (0.269 - 0.443) 0.642** 0.682* 0.684* 0.689* 0.691* (0.467 - 0.882) (0.495 - 0.940) (0.494 - 0.946) (0.500 - 0.951) (0.499 - 0.957) Time ART X Time ∆ Role functioning 0.864 1.377 (0.669 - 1.114) (0.924 - 2.052) ∆ Depression 0.902*** 0.889*** (0.862 - 0.944) (0.835 - 0.946) ∆ Work status Constant 0.438*** 0.505** (0.285 - 0.673) (0.325 - 0.784) 0.700 0.985 0.584 1.284 0.500 (0.305 - 1.607) (0.405 - 2.391) (0.247 - 1.384) (0.528 - 3.123) (0.163 - 1.538) 1,739 1,697 1,697 1,697 1,697 602 573 573 573 573 Observations Number of person-identifiers *** p<0.001, ** p<0.01, * p<0.05 Note: All regressions control for the following baseline variables: Kampala, female, CD4 count, food insecurity, work status, role functioning score, depression score, asset index score, receipt of material support, and a set of indicators for month of baseline interview (March – Sept., with February as the omitted month). 94 WhileoursamplesizelimitstheanalysisofhowgendermodifiestheeffectofARTonfood insecurity,regressionresultsinTable3suggestthatmendrivetheoveralleffectsonfoodinsecurity thatweseeinourstudypopulation[OR=0.485;p<0.05]comparedtowomen[b=notsignificant at0.05level]. Table3:Bygender:Longitudinallogisticregressionresultsonfoodinsecurity Odds of severe food insecurity (95% CI) Women Men ART group Time 0.670 1.104 (0.418 - 1.074) (0.415 - 2.939) 0.368*** 0.296*** (0.281 - 0.483) (0.171 - 0.511) 0.703 0.485* (0.489 - 1.009) (0.244 - 0.964) 1.210 0.532 (0.481 - 3.046) (0.083 - 3.419) 1,193 546 416 199 ART X Time Constant Observations Number of person-identifiers *** p<0.001, ** p<0.01, * p<0.05 Note: Does not include pathway variables. All regressions control for the following baseline variables: Kampala, female, CD4 count, food insecurity, work status, role functioning score, depression score, asset index score, receipt of material support, and a set of indicators for month of baseline interview (March – Sept., with February as the omitted month). Oursensitivityanalysesrevealsimilarresultsasourprimaryregressionanalysis.Usingthe sameregressionmodel,ourfirsttwosensitivityanalysesappeartostrengthentherelativebenefit ofARTonfoodinsecuritycomparedtothenon‐ARTgroup(Table4).Theseincludeour“astreated” analysis(Table4,Column1)andouranalysiscomparingthesickestofthenon‐ARTgrouptothe ARTgroup(Table4,Column2). 95 Table4:Sensitivityanalyses Odds of severe food insecurity (95% CI) ART group Time ART X Time ∆ Role functioning ∆ Depression ∆ Work status Constant Observations Number of person-identifiers Original regression (Table 2, Column 5) SA1: Drop switchers SA2: Use only sickest in nonART group 0.957 (0.617 - 1.484) 0.345*** (0.269 - 0.443) 0.691* (0.499 - 0.957) 1.377 (0.924 - 2.052) 0.889*** (0.835 - 0.946) 0.505** (0.325 - 0.784) -0.797 (0.594) 0.976 (0.607 - 1.570) 0.361*** (0.273 - 0.478) 0.685* (0.484 - 0.970) 1.317 (0.891 - 1.947) 0.896*** (0.842 - 0.953) 0.497** (0.322 - 0.767) 0.607 (0.187 - 1.972) 0.962 (0.598 - 1.548) 0.352*** (0.258 - 0.480) 0.659* (0.453 - 0.958) 1.448 (0.900 - 2.331) 0.888*** (0.829 - 0.952) 0.440*** (0.273 - 0.709) 0.679 (0.196 - 2.353) 1,697 1,549 1,341 573 523 452 *** p<0.001, ** p<0.01, * p<0.05 Note: All regressions control for the following baseline variables: Kampala, female, CD4 count, food insecurity, work status, role functioning score, depression score, asset index score, receipt of material support, and a set of indicators for month of baseline interview (March – Sept., with February as the omitted month). 96 Roleofpathwayvariables Workstatus.WorkstatusimprovedsignificantlyfortheARTgroupfrombaseline [49%]toMonth6[72%;p<0.01]andagainatMonth12[81%;p<0.01],whilethenon‐ ARTgroupexperiencedasmallerincreasefrombaseline[69%]toMonth6[75%;p<0.01] andagainatMonth12[80%;p<0.01].WhiletheARTgroupwaslesslikelytobecurrently workingatbaselinethanthenon‐ARTgroup,therewasnostatisticallysignificantdifference betweenthegroupsattheendofthestudy(seeFigure3.a). Whenchangeinworkstatuswasaddedtotheregressionmodelforfoodinsecurity, improvedworkstatuswasasignificantpredictorofdecreasedoddsoffoodinsecurity[OR= 0.438;p<0.001](Table2,Column4),whiletheARTbytimevariableweakenedin magnitudeandsignificance[OR=0.689;p<0.05].Intheregressionincludingallpathway variables(Column5),thecoefficientonchangeinworkstatusweakenedslightlybut remainedasignificantpredictoroffoodinsecurity[OR=0.505;p<0.01].Inpracticalterms, theseresultsmeanthatimprovedworkstatuswasassociatedwith~50%loweroddsof foodinsecurity. 97 Figure3:Trendsinhypothesizedpathwayvariables .4 .5 Currently working .6 .7 .8 .9 (a) Work status 0 6 Month 12 Non-ART 95% CI ART 0 2 Depression score 4 6 8 (b) Depression 0 6 Month 12 Non-ART 95% CI ART (c) Role functioning 1 Role functioning score 1.2 1.4 1.6 1.8 2 0 6 Month Non-ART 95% CI 12 ART 98 Depression.IntheARTgroup,depressiondecreasedsignificantlyfrombaseline [mean=5.9]toMonth6[mean=2.4;p<0.01]andagainatMonth12[mean=1.4;p<0.01]. Inthenon‐ARTgroup,depressiondecreasedfrombaseline[mean=4.2]toMonth6[mean= 2.4]butdidnotchangefurtheratMonth12(seeFigure3.b.).AtMonth12,depressionwas lowerfortheARTgroupcomparedtothenon‐ARTgroup[p<0.01]. Whenchangeindepressionwasaddedtotheregressionmodelforfoodinsecurity, oneunitofdecreaseddepressionwasasignificantpredictorofdecreasedoddsoffood insecurity[OR=0.902;p<0.001](Table2,Column3).MeanwhiletheARTbytimevariable weakenedinmagnitudeandsignificance[OR=0.684;p<0.05].Intheregressionincluding allpathwayvariables(Column5),thecoefficientonchangeindepressionremaineda significantpredictoroftheoddsoffoodinsecurity[OR=0.889;p<0.001].Inpractical terms,theseresultssuggestthataoneunitdecreaseindepressionoverthecourseofthe studywasassociatedwithroughly11%loweroddsofseverefoodinsecurity. Physicalhealth.RolefunctioningimprovedfortheARTgroupfrombaseline[mean= 1.1]toMonth6[mean=1.9;p<0.01]butdidnotchangefurtheratmonth12,whilethe non‐ARTgroupexperiencedasmallerincreasefrombaseline[mean=1.6]toMonth6 [mean=1.9;p<0.01]andaveryslightdecreaseatMonth12[mean=1.8;p<0.01](see Figure3.c).AtMonth12,theARTgrouphadhigherrolefunctioningcomparedtothenon‐ ARTgroup[p<0.01].However,whenchangeinrolefunctioningwasaddedtothe regressionmodelforfoodinsecurity,itwasnotasignificantpredictoroftheoddsoffood insecurity.Insensitivityanalysiswherealternatemeasuresofphysicalhealthwereused (includingphysicalhealthfunctioningandoverallhealth),physicalhealthwasstillnota significantpredictoroffoodinsecurity. 99 DISCUSSION Inthisstudy,wefindthatfoodsecurityimprovesovertimeforpatientsentering careandtreatmentregardlessofARTstatus,butthatARTisasignificantpredictorof improvedfoodsecurityaboveandbeyondHIVcarewithoutART.Thesefindingscontribute togrowingevidenceoftheeconomicandnutritionalbenefitsofARTandextendtherange ofthesepotentialbenefitstoincludefoodsecurity.Weprovidesomeofthefirstrobust evidencethatARThelpsalleviatethefoodinsecurityofadultswithHIVreceivingART,even intheabsenceoftreatment‐relatedfoodassistanceorlivelihoodsinterventionswhichare becomingincreasinglycommon(Tirivayietal.,2011a;Yageretal.,2011). Toourknowledge,nopublishedstudieshaveexplicitlyexaminedchangesinfood insecurityasanoutcomeofHIVtreatmentandcare.Previousworklookedatchildren’s nutritionasaproxyforincomeeffectsofARTinKenya.ResearchersinKenyainvestigated changesinthenutritionalstatusofveryyoungchildreninhouseholdswithatleastone adultreceivingART;theauthorspostulatedthatanimprovementinchildnutritioncould signalanincomeeffectastreatmentimprovedhealthandeconomicproductivity(Zivinet al.,2009).Resultsindicatedthatchildreninearly‐stageARVhouseholds(receivingART≤ 100days)mayexperienceimprovednutritionasaresultofART,comparedtoarandom sampleofhouseholdswhoseHIVstatuswasnotknown;however,noeffectwasdetectedfor later‐stageARVhouseholds(receivingART>100days).Whiletheseresultsaregenerallyin linewithourfindingsofimprovedfoodsecurity,theyapplytodifferencesbetweenART‐ enrolledhouseholdsandhouseholdswhichmayormaynotcontainanHIV‐positive member,andwhicharenotseekingtreatment. Ourresultsalsosuggestthatgreaterabilitytoworkandreducedsymptomsof depressionmaybetheprimarypathwaysthroughwhichARTimprovesfoodinsecurity. 100 TheseresultsareconsistentwithgrowingevidencethatARTimprovesemploymentstatus andworkproductivityforPLHIV(B.Larsonetal.,2008;Thirumurthyetal.,2008b;G. Wagneretal.,2009),andthatmentalhealthmayplayakeyrolegivenevidencethat depressionislinkedtoHIVinlow‐resourcesettingswithnegativeimplicationsfor productivityandworkstatus(Collinsetal.,2006;Majetal.,1994;Rabkin,2008;G.J. Wagneretal.,2010).Identifyingimprovedworkandmentalhealthaspossiblepathways betweenARTandreducedfoodinsecurityreinforcesthegrowingrecognitionofthe importanceofmentalhealthsupport(Collinsetal.,2006;Yunetal.,2005)andlivelihoods programs(SGillespieetal.,2001;Panditetal.,2010;Roopnaraineetal.,2011)aspartof comprehensivetreatmentandcareforpeoplewithHIV,includingART. TakentogetherwiththeliteraturethatfoodsecurityimprovesARToutcomes,our findingsprovideempiricalsupportforthebidirectionalrelationshipbetweenARTandfood insecurity.EvidenceaboundsforthenegativeeffectsoffoodinsecurityonARTadherence andtreatmentretention,withimplicationsforpoorCD4count,viralsuppression,morbidity andmortality(Anemaetal.,2009;Frankeetal.,2011;Weiseretal.,2009b;Weiseretal., 2009c;Weiseretal.,2012).Theserelationshipslikelyoperatethroughacombinationof biologic,nutritionalandbehavioralpathways.Forexample,foodinsecuritymaycreateor exacerbatepoornutritionalstatus(e.g.lowBMI)whichcouldleadtopoorclinicaloutcomes (Weiseretal.,2009b).Ontheotherhand,foodinsecuritymaycompromiseARTadherence iflackoffoodisanissue,sincemanyantiretroviralmedicationsmustbetakenwithfood (Deribeetal.,2008).FoodinsecuritycanalsoreduceARTaccess,adherenceandretentionif itleadstotrade‐offsbetweentreatment(whichinvolvesbothdirectandindirectcosts,such asfees,transport,andlostworktime)andotherbasicindividualandhouseholdneeds (Martinetal.,2011b).Facedwiththesechallenges,ARTprogramsareincreasingly integratinginterventionstosupportthefoodsecurityofpatients,includingthroughdirect 101 foodassistance,nutritionalsupport,andlivelihoodsprograms(Byronetal.,2008;Fregaet al.,2010;J.Koetheetal.,2009;Tirivayietal.,2011a). ThebidirectionalrelationshipbetweenfoodinsecurityandARThasimplicationsfor programsandpoliciesconsideringfoodsecurityinterventionstiedtoHIVtreatment.In particular,thisrelationshipimpliesfoodsecurityandARTmayworkinapositivefeedback cyclethatARTprogramscouldleveragetoproducean“upwardspiral”inwell‐beingby promotingmentalhealthsupport(e.g.counseling,treatmentfordepression),livelihoods interventions(e.g.incomegenerationprojects,microfinance,employmentservices,etc)and foodsupplementationaspartofHIVtreatmentprograms.Livelihoodsprogramsandfood supplementationinterventionsarecloselyrelated:whilefoodsupplementationisone approachoftenusedtosupporttheshort‐termnutritionandfoodsecurityofHIVpatients (Cantrelletal.,2008;Iversetal.,2010;vanOosterhoutetal.,2010),thetransitionfrom short‐termfood‐basedapproachesintolonger‐termlivelihoodsapproachesisincreasingly emphasized(Yageretal.,2011).Furthermore,althoughARTmayimprovefoodsecurity,we findthatthemagnitudeofthisbenefitonitsownismodest,andthatitmayprimarilyaccrue tomen.Thus,ARTalonecannotbeexpectedtoresolvetheproblemoffoodinsecurityfor PLHIV,especiallyforwomen.Rather,inconjunctionwithotherkeyinterventions,ARTmay playaroleinmovingPLHIVtowardsgreaterwell‐being. Inaddition,thefindingthatfoodsecurityimprovedforthenon‐ARTgroupaswell astheARTgroupsuggeststhatenteringtreatmentandcare–evenpriortoART–mayhave positiveeffectsonwell‐being.ThefactthatallparticipantsbeganreceivingHIVcareat studyinitiation,togetherwithouraccountingforseasonalitythroughmonth‐of‐interview controls,suggeststhatthistimetrendmayindicatetheeffectsofHIVcareratherthana seculardecreaseinfoodinsecurity.Itcouldstandtoreasonthatenteringcare(including 102 receivingtreatmentforopportunisticinfections,clinicalmonitoring,etc)couldalsoimprove foodsecurityviasimilarpathwaysasART(Booysenetal.,2007).Thisissuggestedbythe strongchangeswitnessedindepressionandworkstatuseveninthenon‐ARTgroup. However,withoutacontrolgroupofPLHIVnotreceivingHIVcare,wecannotformallytest thishypothesis. Thisstudyissubjecttoseverallimitations.First,ARTwasnotrandomlyassigned,as thiswouldhavebeenunethicalgiventhewidespreadaccesstoARTinUgandaatthetimeof studyenrollment.Therefore,weconstructedoursampletobeascomparableaspossibleby restrictingthecontrolgrouptothosewhowerealmost,thoughnotyet,eligibleforART.We thenincludedkeygroupdifferencesascovariatesinourregressionmodels.Mostimportant wastocontrolforbaselineCD4count,whichwastheprimaryselectioncriteriaandwhich reflectsthelevelofHIV‐relatedhealthoftheindividual.Oneconcernwasthatourresults couldbedrivenprimarilybythefactthattheARTsamplewassickeratbaseline.However, therobustnessofourresultstosensitivityanalysiswhichlimitedthenon‐ARTgrouptoits sickestmemberscomparablewiththeARTgroupsuggeststhatourfindingsmaybevalid. Nevertheless,givenournon‐randomizeddesign,wecannotruleoutregressiontothemean asanexplanationforourresults.Inaddition,ourresultsshouldnotbegeneralizedtothe generalpopulationofPLHIV,sincewedonothaveacontrolgroupofPLHIVnotincare.Our lackofacomparisongroupofPLHIVnotincarealsomeanswecannotaccountfornatural changesinourkeyoutcomeandpathwayvariablesthatmaynotberelatedtoHIVcare. Anotherlimitationisthatweassessfoodinsecurityusinganindividualratherthana householdmeasure.Combinedwithlackofdataonintra‐householdfooddistribution,this makesitdifficulttoknowifwetrulyobservechangesinfoodinsecurityorsimplychanges infooddistributionpatternswithinthehousehold(e.g.givingthepersonontreatment morefood).However,thewordingofthequestiondoesaskforreductionsinindividual 103 consumptionduetolackofhouseholdresources,whichshouldpromptananswerreflecting overallfoodinsecurityratherthandistribution. WhileourstudydemonstratesanassociationbetweenARTandimprovedfood security,andprovidespreliminaryevidencethatimprovingworkandmentalhealth outcomesmaybekeytoachievingthisbenefit,significantworkremainstoidentifyand describethedynamicsofthepathwaysconnectingARTandimprovementsinfood insecurity.Inparticular,ourreviewoftheliteratureandourpreliminaryresultsongender suggestthattheunderlyingprocessesaffectingthepathwaysbetweenARTandfood insecuritymaydifferbetweenmenandwomen,especiallywithregardstohoweconomic activitiesrelatedtofoodprocurement(whetherlabormarketordomesticactivities) respondtoART(d’Addaetal.,2009;B.A.Larsonetal.,2009;Thirumurthyetal.,2008a). Thisshouldbeexploredinfuturestudies. CONCLUSION ThepotentialforARTtoimprovethefoodsecurityofpeoplelivingwithHIV strengthensboththepolicyandpublichealthcaseforsustainingandscaling‐uptreatment forallthoseinneedofit.InadditiontoevidencethateconomicreturnstoARTmayoutpace costsduetoincreasedproductivityforlow‐incomecountries(Reschetal.,2011),andthat ARTplaysapivotalroleinpreventionofHIVtransmission(Anglemyeretal.,2011;Granich etal.,2009),evidencesupportingitssocialandeconomicbenefitsenhancesthealready powerfulrationaleformaintaining,andincreasing,investmentsinARTinresource‐poor settings.ThisismoreimportantthaneverastheWHOnowrecommendsARTinitiationat earlierCD4thresholds,effectivelyincreasingthedemandforART(Konde‐Luleetal.,2011). AnuancedunderstandingofthebidirectionalrelationshipbetweenARTandfoodsecurity, inwhichemploymentandmentalhealthplaykeyroles,shouldinformdecision‐makersas 104 theyconsiderthedevelopmentofinterventionstohaltthe“viciouscycle”ofHIVandpoor socialandeconomicoutcomes,includingfoodinsecurity.Well‐integratedandimplemented interventionsinthecontextofcomprehensivecarehavethepotentialtoproducean “upwardspiral”wherefoodsecurityandARTcanmutuallyreinforceeachotherforthe benefitofallthoseintreatment. 105 REFERENCES Adewuya,A.O.,Ola,B.A.,&Afolabi,O.O.(2006).Validityofthepatienthealthquestionnaire(PHQ‐ 9)asascreeningtoolfordepressionamongstNigerianuniversitystudents.Journalof AffectiveDisorders,96(1‐2),89‐93. Alaimo,K.,Olson,C.M.,&Frongillo,E.A.(2002).Familyfoodinsufficiency,butnotlowfamily income,ispositivelyassociatedwithdysthymiaandsuicidesymptomsinadolescents. JournalofNutrition,132(4),719‐725. Alaimo,K.,Olson,C.M.,&FrongilloJr,E.A.(2001a).Lowfamilyincomeandfoodinsufficiencyin relationtooverweightinUSchildren:isthereaparadox?ArchivesofPediatricsand AdolescentMedicine,155(10),1161. Alaimo,K.,Olson,C.M.,FrongilloJr,E.A.,&Briefel,R.R.(2001b).Foodinsufficiency,familyincome, andhealthinUSpreschoolandschool‐agedchildren.AmericanJournalofPublicHealth, 91(5),781. Alberti,K.G.,Zimmet,P.,&Shaw,J.(2005).Themetabolicsyndrome‐‐anewworldwidedefinition. Lancet,366(9491),1059. Almeida,L.B.,Segurado,A.C.,Duran,A.C.F.,&Jaime,P.C.(2011).Impactofanutritional counselingprogramonpreventionofHAART‐relatedmetabolicandmorphologic abnormalities.AIDSCare,23(6),755‐763. Altheide,D.(1996).Qualitativemediaanalysis.ThousandOaks,CA:SagePublications. Alvarez,C.,Salazar,R.,Galindez,J.,Rangel,F.,CastaÃ±eda,M.,Lopardo,G.,etal.(2010).Metabolic syndromeinHIV‐infectedpatientsreceivingantiretroviraltherapyinLatinAmerica. BrazilianJournalofInfectiousDiseases,14(3),256‐263. Anema,A.,Vogenthaler,N.,Frongillo,E.,Kadiyala,S.,&Weiser,S.(2009).Foodinsecurityand HIV/AIDS:currentknowledge,gaps,andresearchpriorities.CurrentHIV/AIDSReports,6(4), 224‐231. Anglemyer,A.,Rutherford,G.W.,Egger,M.,&Siegfried,N.(2011).Antiretroviraltherapyfor preventionofHIVtransmissioninHIV‐discordantcouples.CochraneDatabaseofSystematic Reviews,5. Apps,P.,Reduction,W.B.P.,&Division,E.M.G.(2003).Gender,timeuseandmodelsofthe household:IZA. Bahwere,P.,Sadler,K.,&Collins,S.(2009).Acceptabilityandeffectivenessofchickpeasesame‐ basedready‐to‐usetherapeuticfoodinmalnourishedHIV‐positiveadults.PatientPrefer Adherence,3,67‐75. 106 Barrett,C.B.(2002).Foodsecurityandfoodassistanceprograms.HandbookofAgricultural Economics,2(B),2103‐2190. Barrett,C.B.(2006).Foodaid'sintendedanduntendedconsequences:BackgroundpaperforFAO 2006.CornellUniversity. Bartlett,J.A.,DeMasi,R.,Quinn,J.,Moxham,C.,&Rousseau,F.(2001).Overviewoftheeffectiveness oftriplecombinationtherapyinantiretroviral‐naiveHIV‐1infectedadults.AIDS,15(11), 1369. Beard,J.,Feeley,F.,&Rosen,S.(2009).Economicandqualityoflifeoutcomesofantiretroviral therapyforHIV/AIDSindevelopingcountries:asystematicliteraturereview.AIDSCare, 21(11),1343‐1356. Bickel,G.,Nord,M.,Price,C.,Hamilton,W.,&Cook,J.(2000).GuidetoMeasuringHouseholdFood Security,Revised2000.Alexandria,VA:U.S.DepartmentofAgriculture,FoodandNutrition Service. Bock,N.,Lee,C.W.,Bechange,S.,Moore,J.,Khana,K.,&Were,W.(2008).Changesindepressionand dailyfunctioningwithinitiationofARTinHIV‐infectedAdultsinRuralUganda..Conference onRetrovirusesandOpportunisticInfectionsBoston,MA. Booysen,F.L.R.,VanRensburg,H.C.J.,Bachmann,M.,Louwagie,G.,&Fairall,L.(2007).Theheartin HAART:qualityoflifeofpatientsenrolledinthepublicsectorantiretroviraltreatment programmeinthefreestateprovinceofSouthAfrica.SocialIndicatorsResearch,81(2),283‐ 329. Bukusuba,J.,Kikafunda,J.,&Whitehead,R.(2007).Foodsecuritystatusinhouseholdsofpeople livingwithHIV/AIDS(PLWHA)inaUgandanurbansetting.BritishJournalofNutrition, 98(1),211‐217. Byron,E.,Gillespie,S.,&Nangami,M.(2008).Integratingnutritionsecuritywithtreatmentof peoplelivingwithHIV:lessonsfromKenya.FoodandNutritionBulletin,29(2),87‐97. Campbell,C.C.(1991).Foodinsecurity:Anutritionaloutcomeorapredictorvariable?Journalof Nutrition,121(3),408. Cantrell,R.,Sinkala,M.,Megazinni,K.,Lawson‐Marriott,S.,Washington,S.,Chi,B.,etal.(2008).A pilotstudyoffoodsupplementationtoimproveadherencetoantiretroviraltherapyamong food‐insecureadultsinLusaka,Zambia.JAIDSJournalofAcquiredImmuneDeficiency Syndromes,49(2),190‐195. Castleman,T.,Seumo‐Fosso,E.,&Cogill,B.(2004).FoodandNutritionImplicationsofAntiretroviral TherapyinResourceLimitedSettings.FoodandNutritionTechnicalAssistance(FANTA). Washington,DC:AcademyforEducationalDevelopment. 107 Chambers,R.,&Conway,G.(1991).Sustainablerurallivelihoods:practicalconceptsforthe21st century.IDSDiscussionPaper.Sussex,UK:InstituteofDevelopmentStudies. Chen,M.A.(2001).Womenintheinformalsector:Aglobalpicture,theglobalmovement.SAIS Review,21(1),71‐82. Coates,J.,Frongillo,E.A.,Rogers,B.L.,Webb,P.,Wilde,P.E.,&Houser,R.(2006).Commonalitiesin theexperienceofhouseholdfoodinsecurityacrosscultures:whataremeasuresmissing? JournalofNutrition,136(5),1438S. Cole,S.M.,&Tembo,G.(2011).Theeffectoffoodinsecurityonmentalhealth:panelevidencefrom ruralZambia.SocialScience&Medicine,73(7),1071‐1079. Collins,P.Y.,Holman,A.R.,Freeman,M.C.,&Patel,V.(2006).Whatistherelevanceofmentalhealth toHIV/AIDScareandtreatmentprogramsindevelopingcountries?Asystematicreview. AIDS,20(12),1571. Corbett,J.(1988).Famineandhouseholdcopingstrategies.WorldDevelopment,16(9),1099‐1112. Crum‐Cianflone,N.F.,Roediger,M.,Eberly,L.E.,Vyas,K.,Landrum,M.L.,Ganesan,A.,etal.(2010). ObesityamongHIV‐infectedpersons:impactofweightonCD4cellcount.AIDS,24(7),1069. Crush,J.,Drimie,S.,Frayne,B.,&Caesar,M.(2011).TheHIVandurbanfoodsecuritynexusinAfrica. FoodSecurity,3(3),347‐362. d’Adda,G.,Goldstein,M.,Zivin,J.G.,Nangami,M.,&Thirumurthy,H.(2009).ARVtreatmentand timeallocationtohouseholdtasks:EvidencefromKenya.AfricanDevelopmentReview, 21(1),180‐208. Deribe,K.,Hailekiros,F.,Biadgilign,S.,Amberbir,A.,&Beyene,B.(2008).Defaultersfrom antiretroviraltreatmentinJimmaUniversitySpecializedHospital,SouthwestEthiopia. TropicalMedicine&InternationalHealth,13(3),328‐333. Dinour,L.M.,Bergen,D.,&Yeh,M.C.(2007).Thefoodinsecurity‐obesityparadox:areviewofthe literatureandtherolefoodstampsmayplay.JournaloftheAmericanDieteticAssociation, 107(11),1952‐1961. Donahoe,D.A.(1999).Measuringwomen'sworkindevelopingcountries.Populationand DevelopmentReview,25(3),543‐576. Eholie,S.P.,&Nolan,M.(2003).Antiretroviraltreatmentcanbecost‐savingforindustryandlife‐ savingforworkers:acasestudyfromCoted'Ivoire'sprivatesector.InJ.P.Moatti,T. Barnett,B.Coriat,Y.Souteyrand,J.Dumoulin&Y.A.Flori(Eds.),EconomicsofAIDSand accesstoHIV/AIDScareindevelopingcountries:issuesandchallenges.Paris,France: AgenceNationaledeRecherchessurleSida. 108 Erah,P.O.,&Arute,J.E.(2008).AdherenceofHIV/AIDSpatientstoantiretroviraltherapyina tertiaryhealthfacilityinBeninCity.AfricanJournalofPharmacyandPharmacology,2(7), 145‐152. FANTA.(2004).HIV/AIDS:Aguidefornutritionalcareandsupport.Washington,DC:Academyfor EducationalDevelopment. Fesko,S.L.(2001).DisclosureofHIVstatusintheworkplace:Considerationsandstrategies.Health andSocialWork,26(4),235‐244. Filmer,D.,&Pritchett,L.H.(2001).Estimatingwealtheffectswithoutexpendituredata—ortears: anapplicationtoeducationalenrollmentsinstatesofIndia.Demography,38(1),17. Franke,M.,Murray,M.,Muñoz,M.,Hernández‐Díaz,S.,Sebastián,J.,Atwood,S.,etal.(2011).Food insufficiencyisariskfactorforsuboptimalantiretroviraltherapyadherenceamongHIV‐ infectedadultsinurbanPeru.AIDSandBehavior,15(7),1483‐1489. Frega,R.,Duffy,F.,Rawat,R.,&Grede,N.(2010).FoodinsecurityinthecontextofHIV/AIDS:A frameworkforaneweraofprogramming.Food&NutritionBulletin,31(Supplement4), 292S‐312S. Friis‐Møller,N.,Weber,R.,Reiss,P.,Thiébaut,R.,Kirk,O.,Monforte,A.A.,etal.(2003). CardiovasculardiseaseriskfactorsinHIVpatients‐associationwithantiretroviraltherapy. ResultsfromtheDADstudy.AIDS,17(8),1179. Gibson,R.S.(2005).Principlesofnutritionalassessment:OxfordUniversityPress,USA. Gillespie,S.,Haddad,L.,&Jackson,R.(2001).HIV/AIDS,foodandnutritionsecurity:Impactsand actions.IFPRI(InternationalFoodPolicyResearchInstitute),WFP(WorldFoodProgramme). Gillespie,S.,&Kadiyala,S.(2005).HIV/AIDSandfoodandnutritionsecurity:Fromevidenceto action.Washington,D.C.:InternationalFoodPolicyResearchInstitute. Glenn,M.,Ford,J.,Moore,D.,&Hollar,D.(2003).EmploymentIssuesasRelatedbyIndividuals LivingwithHIVorAIDS.TheJournalofRehabilitation,69(1),30‐37. Granich,R.M.,Gilks,C.F.,Dye,C.,DeCock,K.M.,&Williams,B.G.(2009).UniversalvoluntaryHIV testingwithimmediateantiretroviraltherapyasastrategyforeliminationofHIV transmission:amathematicalmodel.TheLancet,373(9657),48‐57. Grosh,M.E.,&Glewwe,P.(2000).Designinghouseholdsurveyquestinnairesfordeveloping countries:lessonsfrom15yearsofthelivingstandardsmeasurementstudy.Washington, D.C.:WorldBank. Grundy,S.M.,Cleeman,J.I.,Daniels,S.R.,Donato,K.A.,Eckel,R.H.,Franklin,B.A.,etal.(2005). Diagnosisandmanagementofthemetabolicsyndrome.Circulation,112(17),2735‐2752. 109 Habicht,J.(1974).Estandarizacióndemetodosepidemiológicoscuantitativossobreelterreno; Standardizationofquantitativeepidemiologicalmethodsinthefield.BoletíndelaOficina SanitariaPanamericana(OSP),76(5),375‐384. Haering,S.A.,&Syed,S.B.(2009).CommunityfoodsecurityinUnitedStatescities:Asurveyofthe relevantscientificliterature.(p.2010):CenterforaLivableFuture,JohnsHopkins BloombergSchoolofPublicHealth. Hardon,A.,Akurut,D.,Comoro,C.,Ekezie,C.,Irunde,H.,Gerrits,T.,etal.(2007).Hunger,waiting timeandtransportcosts:timetoconfrontchallengestoARTadherenceinAfrica.AIDSCare, 19(5),658‐665. Heflin,C.M.,Siefert,K.,&Williams,D.R.(2005).Foodinsufficiencyandwomen'smentalhealth: findingsfroma3‐yearpanelofwelfarerecipients.SocialScience&Medicine,61(9),1971‐ 1982. Holmes,K.,Winskell,K.,Hennink,M.,&Chidiac,S.(2011).MicrofinanceandHIVmitigationamong peoplelivingwithHIVintheeraofanti‐retroviraltherapy:EmerginglessonsfromCôte d'Ivoire.GlobalPublicHealth,6(4),447‐461. Ivers,L.,Chang,Y.,Jerome,J.,&Freedberg,K.(2010).Foodassistanceisassociatedwithimproved bodymassindex,foodsecurityandattendanceatclinicinanHIVprogramincentralHaiti:a prospectiveobservationalcohortstudy.AIDSResearchandTherapy,7(33),1‐8. Jaime,P.C.,Florindo,A.A.,Latorre,M.R.D.O.,&Segurado,A.A.C.(2006).Centralobesityand dietaryintakeinHIV/AIDSpatients.RevistadeSaúdePública,40(4),634‐640. Jelsma,J.,Maclean,E.,Hughes,J.,Tinise,X.,&Darder,M.(2005).Aninvestigationintothehealth‐ relatedqualityoflifeofindividualslivingwithHIVwhoarereceivingHAART.AIDSCare, 17(5),9. Juster,F.T.,&Stafford,F.P.(1991).Theallocationoftime:Empiricalfindings,behavioralmodels, andproblemsofmeasurement.JournalofEconomicLiterature,29(2),471‐522. Kadiyala,S.,Rawata,R.,Roopnarainec,T.,Babiryed,F.,&Ochaid,R.(2009).Applyingaprogramme theoryframeworktoimprovelivelihoodinterventionsintegratedwithHIVcareand treatmentprogrammes.JournalofDevelopmentEffectiveness,1(4),470‐491. Kagee,A.,Remien,R.,Berkman,A.,Hoffman,S.,Campos,L.,&Swartz,L.(2011).Structuralbarriers toARTadherenceinSouthernAfrica:Challengesandpotentialwaysforward.GlobalPublic Health,6(1),83‐97. Kaschula,S.(2011).Usingpeopletocopewiththehunger:socialnetworksandfoodtransfers amongstHIV/AIDSafflictedhouseholdsinKwaZulu‐Natal,SouthAfrica.AIDSandBehavior, 1‐13. 110 Kates,J.,Wexler,A.,Lief,E.,Avila,C.,&Gobet,B.(2011).FinancingtheResponsetoAIDSinLow‐ andMiddle‐IncomeCountries:InternationalAssistancefromDonorGovernmentsin2010. MenloPark,CA:KaiserFamilyFoundation. Kaye,H.L.,&Moreno‐Leguizamon,C.J.(2011).Nutritioneducationandcounsellingasstrategic interventionstoimprovehealthoutcomesinadultoutpatientswithHIV:aliteraturereview. AfricanJournalofAIDSResearch,9(3),271‐283. Koethe,J.,Chi,B.,Megazzini,K.,Heimburger,D.,&Stringer,J.(2009).Macronutrient supplementationformalnourishedHIV‐infectedadults:areviewoftheevidencein resource‐adequateandresource‐constrainedsettings.ClinicalInfectiousDiseases,49(5), 787‐798. Koethe,J.R.,Limbada,M.I.,Giganti,M.J.,Nyirenda,C.K.,Mulenga,L.,Wester,C.W.,etal.(2010a). Earlyimmunologicresponseandsubsequentsurvivalamongmalnourishedadultsreceiving antiretroviraltherapyinUrbanZambia.AIDS,24(13),2117. Koethe,J.R.,Lukusa,A.,Giganti,M.J.,Chi,B.H.,Nyirenda,C.K.,Limbada,M.I.,etal.(2010b). Associationbetweenweightgainandclinicaloutcomesamongmalnourishedadults initiatingantiretroviraltherapyinLusaka,Zambia.JAIDSJournalofAcquiredImmune DeficiencySyndromes,53(4),507. Konde‐Lule,J.,Makumbi,F.,Pakker,N.,Muyinda,A.,Mubiru,M.,&Cobelens,F.(2011).Effectof changingantiretroviraltreatmenteligibilitycriteriaonpatientloadinKampala,Uganda. AIDSCare,23(1),35‐41. Krippendorff,K.(2004).ContentAnalysis:AnIntroductiontoitsMethodology.ThousandOaks,CA: SagePublications. Larson,B.,Fox,M.,Rosen,S.,Bii,M.,Sigei,C.,Shaffer,D.,etal.(2008).Earlyeffectsofantiretroviral therapyonworkperformance:preliminaryresultsfromacohortstudyofKenyan agriculturalworkers.AIDS,22(3),421. Larson,B.A.,Fox,M.P.,Rosen,S.,Bii,M.,Sigei,C.,Shaffer,D.,etal.(2009).Dothesocioeconomic impactsofantiretroviraltherapyvarybygender?AlongitudinalstudyofKenyan agriculturalworkeremploymentoutcomes.BMCPublicHealth,9,240. Larson,N.I.,&Story,M.T.(2011).FoodInsecurityandWeightStatusAmongUSChildrenand Families::AReviewoftheLiterature.AmericanJournalofPreventiveMedicine,40(2),166‐ 173. Lee,J.S.,&FrongilloJr,E.A.(2001).FactorsassociatedwithfoodinsecurityamongUSelderly persons.TheJournalsofGerontologySeriesB:PsychologicalSciencesandSocialSciences, 56(2),S94. 111 Leite,L.H.M.,&Sampaio,A.B.D.M.M.(2010).Progressiontooverweight,obesityandassociated factorsafterantiretroviraltherapyinitiationamongbrazilianpersonswithHIV/AIDS NutriciónHospitilaria,25(4). Linnemayr,S.,StearnsLawson,B.,Glick,P.,&Wagner,G.(2011).Economicstatusandcoping mechanismsofindividualsseekingHIVcareinUganda.JournalofAfricanEconomies,20(3). Liu,E.,Spiegelman,D.,Semu,H.,Hawkins,C.,Chalamilla,G.,Aveika,A.,etal.(2011).Nutritional statusandmortalityamongHIV‐infectedpatientsreceivingantiretroviraltherapyin Tanzania.JournalofInfectiousDiseases,204(2),282. Mahajan,A.P.,Sayles,J.N.,Patel,V.A.,Remien,R.H.,Ortiz,D.,Szekeres,G.,etal.(2008).Stigmain theHIV/AIDSepidemic:areviewoftheliteratureandrecommendationsfortheway forward.AIDS,22(Suppl2),S67. Maj,M.,Janssen,R.,Starace,F.,Zaudig,M.,Satz,P.,Sughondhabirom,B.,etal.(1994).WHO NeuropsychiatricAIDSstudy,cross‐sectionalphaseI:Studydesignandpsychiatricfindings. ArchivesofGeneralPsychiatry,51(1),39. Marcellin,F.,Boyer,S.,Protopopescu,C.,Dia,A.,Ongolo‐Zogo,P.,Koulla‐Shiro,S.,etal.(2008). Determinantsofunplannedantiretroviraltreatmentinterruptionsamongpeoplelivingwith HIVinYaoundé,Cameroon(EVALsurvey,ANRS12‐116).TropicalMedicine&International Health,13(12),1470‐1478. Marchione,T.J.(2005).Interactionswiththerecipientcommunityintargetedfoodandnutrition programs.JournalofNutrition,135(4),886‐889. Mariz,C.D.A.,deAlencarXimenes,R.A.,deMelo,H.R.L.,Bandeira,F.,eOliveira,T.G.B.,de Carvalho,É.H.,etal.(2011).BodymassindexinindividualswithHIVinfectionandfactors associatedwiththinnessandoverweight/obesityÍndicedemassacorporalemindivíduos cominfecçãopeloHIVefatoresassociadoscommagrezae.Cad.SaúdePública,27(10), 1997‐2008. Marston,B.,&DeCock,K.(2004).Multivitamins,nutrition,andantiretroviraltherapyforHIV diseaseinAfrica.TheNewEnglandJournalofMedicine,351(1),78. Martin,A.,&Palar,K.(2011a).HIV/AIDSPolicy,FoodandNutritioninLatinAmericaandthe Caribbean.PanamaCity:TheWorldFoodProgramme. Martin,A.,Palar,K.,Derose,K.,&Adams,J.(2011b).Foodinsecurityandnutritionalbarriersto antiretroviraltherapy:lessonsfromLatinAmericaandtheCaribbean.JournalofHIV/AIDS& SocialServices,10(2),194‐214. Mast,T.,Kigozi,G.,Wabwire‐Mangen,F.,Black,R.,Sewankambo,N.,Serwadda,D.,etal.(2004). MeasuringqualityoflifeamongHIV‐infectedwomenusingaculturallyadapted questionnaireinRakaidistrict,Uganda.AIDSCare,16(1),81‐94. 112 Maxwell,S.(1996).Foodsecurity:apost‐modernperspective.FoodPolicy,21(2),155‐170. Maxwell,S.,&Frankenberger,T.(1992).Householdfoodsecurity:Concepts,indicators, measurements.TechnicalReview.NewYork:UNICEF. McIntyre,D.,Thiede,M.,Dahlgren,G.,&Whitehead,M.(2006).Whataretheeconomic consequencesforhouseholdsofillnessandofpayingforhealthcareinlow‐andmiddle‐ incomecountrycontexts?SocialScience&Medicine,62(4),858‐865. Melgar‐Quiñonez,H.,AlvarezUribe,M.,FonsecaCenteno,Z.,Bermúdez,O.,PalmadeFulladolsa,P., Fulladolsa,A.,etal.(2010).CaracterísticasPsicométricasdelaEscaladeSeguridad AlimentariaELCSAaplicadaenColombia,GuatemalayMéxico.SegurançaAlimentare Nutricional,17(1),48‐60. Miles,M.,&Huberman,A.(1994).Qualitativedataanalysis:Anexpandedsourcebook.Thousand Oaks,CA:SagePublications. Mills,E.J.,Nachega,J.B.,Bangsberg,D.R.,Singh,S.,Rachlis,B.,Wu,P.,etal.(2006).Adherenceto HAART:asystematicreviewofdevelopedanddevelopingnationpatient‐reportedbarriers andfacilitators.PLoSmedicine,3(11),e438. Misselhorn,A.A.(2005).WhatdrivesfoodinsecurityinsouthernAfrica?Ameta‐analysisof householdeconomystudies.GlobalEnvironmentalChangePartA,15(1),33‐43. Moh,R.,Danel,C.,Messou,E.,Ouassa,T.,Gabillard,D.,Anzian,A.,etal.(2007).Incidenceand determinantsofmortalityandmorbidityfollowingearlyantiretroviraltherapyinitiationin HIV‐infectedadultsinWestAfrica.AIDS,21(18),2483. Monahan,P.O.,Shacham,E.,Reece,M.,Kroenke,K.,Ong’or,W.O.,Omollo,O.,etal.(2009). Validity/reliabilityofPHQ‐9andPHQ‐2depressionscalesamongadultslivingwith HIV/AIDSinwesternKenya.JournalofGeneralInternalMedicine,24(2),189‐197. Murphy,E.L.,Collier,A.C.,Kalish,L.A.,Assmann,S.F.,Para,M.F.,Flanigan,T.P.,etal.(2001). Highlyactiveantiretroviraltherapydecreasesmortalityandmorbidityinpatientswith advancedHIVdisease.AnnalsofInternalMedicine,135(1),17‐26. Mykerezi,E.,&Mills,B.(2010).TheimpactofFoodStampProgramparticipationonhouseholdfood insecurity.AmericanJournalofAgriculturalEconomics,92(5),1379‐1391. Nachega,J.,Stein,D.,Lehman,D.,Hlatshwayo,D.,Mothopeng,R.,Chaisson,R.,etal.(2004). AdherencetoantiretroviraltherapyinHIV‐infectedadultsinSoweto,SouthAfrica.AIDS Research&HumanRetroviruses,20(10),1053‐1056. Ndekha,M.,vanOosterhout,J.,Saloojee,H.,Pettifor,J.,&Manary,M.(2009a).Nutritionalstatusof Malawianadultsonantiretroviraltherapy1yearaftersupplementaryfeedinginthefirst3 monthsoftherapy.TropicalMedicineandInternationalHealth,14(9),1059‐1063. 113 Ndekha,M.,vanOosterhout,J.,Zijlstra,E.,Manary,M.,Saloojee,H.,&Manary,M.(2009b). Supplementaryfeedingwitheitherready‐to‐usefortifiedspreadorcorn‐soyblendin wastedadultsstartingantiretroviraltherapyinMalawi:randomised,investigatorblinded, controlledtrial.BMJ,338,b1867. Normen,L.,Chan,K.,Braitstein,P.,Anema,A.,Bondy,G.,Montaner,J.S.,etal.(2005).Food insecurityandhungerareprevalentamongHIV‐positiveindividualsinBritishColumbia, Canada.JournalofNutrition,135(4),820‐825. Oguntibeju,O.,vandenHeever,W.,&VanSchalkwyk,F.(2007).Theinterrelationshipbetween nutritionandtheimmunesysteminHIVinfection:areview.PakJBiolSci,10(24),4327‐ 4338. Palar,K.(2012).Roleofantiretroviraltherapyinimprovingfoodsecurityamongpatientsinitiating HIVtreatmentandcareinUganda.FoodInsecurity,LivelihoodsandHIVTreatment: EvidenceforPolicyinDevelopingCountries[Dissertation].SantaMonica,CA:PardeeRAND GraduateSchool. Pandit,J.,Sirotin,N.,Tittle,R.,Onjolo,E.,Bukusi,E.,&Cohen,C.(2010).ShambaMaisha:Apilot studyassessingimpactsofamicro‐irrigationinterventiononthehealthandeconomic wellbeingofHIVpatients.BMCPublicHealth,10(1),245. Pearson,C.R.,Micek,M.,Simoni,J.M.,Matediana,E.,Martin,D.P.,&Gloyd,S.(2006).Modified directlyobservedtherapytofacilitatehighlyactiveantiretroviraltherapyadherencein Beira,Mozambique:developmentandimplementation.JAIDSJournalofAcquiredImmune DeficiencySyndromes,43,S134. PlurinationalStateofBolivia.(2007).LeyNo.3729:LeyparalaprevenciondeVIH‐SIDA,proteccion delosderechoshumanos,yasistenciaintegralmultidisciplinariaparalaspersonasque vivenconelVIH‐SIDA.PlurinationalStateofBolivia. Quisumbing,A.R.(2003).FoodaidandchildnutritioninruralEthiopia.WorldDevelopment,31(7), 1309‐1324. Rabkin,J.G.(2008).HIVanddepression:2008reviewandupdate.CurrentHIV/AIDSReports,5(4), 163‐171. Rachlis,B.S.,Mills,E.J.,&Cole,D.C.(2011).Livelihoodsecurityandadherencetoantiretroviral therapyinlowandmiddleincomesettings:asystematicreview.PLoSONE,6(5),e18948. Radimer,K.L.,Olson,C.M.,Greene,J.C.,Campbell,C.C.,&Habicht,J.P.(1992).Understanding hungeranddevelopingindicatorstoassessitinwomenandchildren.JournalofNutrition Education,24. Ratcliffe,C.,McKernan,S.M.,&Zhang,S.(2011).HowMuchDoestheSupplementalNutrition AssistanceProgramReduceFoodInsecurity?AmericanJournalofAgriculturalEconomics, 93(4),1082‐1098. 114 Rawat,R.,Kadiyala,S.,&McNamara,P.E.(2010).Theimpactoffoodassistanceonweightgainand diseaseprogressionamongHIV‐infectedindividualsaccessingAIDScareandtreatment servicesinUganda.BMCPublicHealth,10,316. Resch,S.,Korenromp,E.,Stover,J.,Blakley,M.,Krubiner,C.,Thorien,K.,etal.(2011).Economic ReturnstoInvestmentinAIDSTreatmentinLowandMiddleIncomeCountries.PLoSONE, 6(10),e25310. Reutlinger,S.,&Katona‐Apte,J.(1984).Thenutritionalimpactoffoodaid:criteriafortheselection ofcost‐effectivefoods.FoodandNutritionBulletin,6(4),1‐10. Roopnaraine,T.,Rawat,R.,Babirye,F.,Ochai,R.,&Kadiyala,S.(2011).“Thegroup”inintegrated HIVandlivelihoodsprogramming:opportunityorchallenge?AIDSCare,24(5),649‐657. Rose,D.(1999).EconomicdeterminantsanddietaryconsequencesoffoodinsecurityintheUnited States.JournalofNutrition,129(2SSuppl),517S. Rose,D.,&Charlton,K.E.(2002).PrevalenceofhouseholdfoodpovertyinSouthAfrica:results fromalarge,nationallyrepresentativesurvey.PublicHealthNutrition,5(03),383‐389. Rose,D.,Habicht,J.P.,&Devaney,B.(1998).HouseholdparticipationintheFoodStampandWIC programsincreasesthenutrientintakesofpreschoolchildren.JournalofNutrition,128(3), 548‐555. Rosen,S.,Larson,B.,Brennan,A.,Long,L.,Fox,M.,Mongwenyana,C.,etal.(2010).Economic outcomesofpatientsreceivingantiretroviraltherapyforHIV/AIDSinSouthAfricaare sustainedthroughthreeyearsontreatment.PLoSONE,5(9),e12731. Rosen,S.,Vincent,J.,MacLeod,W.,Fox,M.,Thea,D.,&Simon,J.(2004).ThecostofHIV/AIDSto businessesinsouthernAfrica.AIDS,18(2),317. Russell,S.(2004).Theeconomicburdenofillnessforhouseholdsindevelopingcountries:areview ofstudiesfocusingonmalaria,tuberculosis,andhumanimmunodeficiencyvirus/acquired immunodeficiencysyndrome.AmericanJournalofTropicalMedicineandHygiene,71(2 Suppl),147‐155. Russell,S.,&Seeley,J.(2009).ThetransitiontolivingwithHIVasachronicconditioninrural Uganda:Workingtocreateorderandcontrolwhenonantiretroviraltherapy.SocialScience &Medicine,70(3),375‐382. Russell,S.,Seeley,J.,Ezati,E.,Wamai,N.,Were,W.,&Bunnell,R.(2007).Comingbackfromthe dead:livingwithHIVasachronicconditioninruralAfrica.HealthPolicyPlan,22(5),344‐ 347. Samuels,F.A.,&Rutenberg,N.(2011).“Healthregainsbutlivelihoodslag”:findingsfromastudy withpeopleonARTinZambiaandKenya.AIDSCare,23(6),748‐754. 115 Sauerborn,R.,Adams,A.,&Hien,M.(1996).Householdstrategiestocopewiththeeconomiccosts ofillness.SocialScience&Medicine,43(3),291‐301. Shor‐Posner,G.,Campa,A.,Zhang,G.,Persaud,N.,Miguez‐Burbano,M.J.,Quesada,J.,etal.(2000). Whenobesityisdesirable:alongitudinalstudyoftheMiamiHIV‐1‐infecteddrugabusers (MIDAS)cohort.JAIDSJournalofAcquiredImmuneDeficiencySyndromes,23(1),81. Siefert,K.,Heflin,C.M.,Corcoran,M.E.,&Williams,D.R.(2004).Foodinsufficiencyandphysical andmentalhealthinalongitudinalsurveyofwelfarerecipients.Journalofhealthandsocial behavior,45(2),171‐186. Smallman,S.(2007).TheAIDSPandemicinLatinAmerica:TheUniversityofNorthCarolinaPress. Sprague,L.,Simon,S.,&Sprague,C.(2011).EmploymentdiscriminationandHIVstigma:survey resultsfromcivilsocietyorganisationsandpeoplelivingwithHIVinAfrica.AfricanJournal ofAIDSResearch,10(sup1),311‐324. Stillwaggon,E.(2006).AIDSandtheEcologyofPoverty:OxfordUniversityPress,USA. Strauss,A.,&Corbin,J.(1998).Basicsofqualitativeresearch:Techniquesandproceduresfor developinggroundedtheory.ThousandOaks,CA:SagePublications. Swaminathan,S.,Padmapriyadarsini,C.,Yoojin,L.,Sukumar,B.,Iliayas,S.,Karthipriya,J.,etal. (2010).NutritionalsupplementationinHIV‐infectedindividualsinSouthIndia:a prospectiveinterventionalstudy.ClinicalInfectiousDiseases,51(1),51. Tanumihardjo,S.A.,Anderson,C.,Kaufer‐Horwitz,M.,Bode,L.,Emenaker,N.J.,Haqq,A.M.,etal. (2007).Poverty,obesity,andmalnutrition:aninternationalperspectiverecognizingthe paradox.JournaloftheAmericanDieteticAssociation,107(11),1966‐1972. Thirumurthy,H.,Jafri,A.,Srinivas,G.,Arumugam,V.,Saravanan,R.M.,Angappan,S.K.,etal.(2011). Two‐yearimpactsonemploymentandincomeamongadultsreceivingantiretroviral therapyinTamilNadu,India:acohortstudy.AIDS,25(2),239. Thirumurthy,H.,Saravanan,R.,Srinivas,G.,Jafri,A.,Sreevidya,J.,&Sahu,S.(2008a).Theimpactof antiretroviraltherapyonsocioeconomicoutcomesofHIV‐infectedpatientsinTamilNadu FamilyCareContinuumProgram,India.PaperpresentedattheXVIIInternationalAIDS Conference.MexicoCity,Mexico. Thirumurthy,H.,Zivin,J.G.,&Goldsten,M.(2008b).TheeconomicimpactofAIDSTreatment:Labor supplyinWesternKenya.JournalofHumanResources,43(3),511‐522. Tirivayi,N.,&Groot,W.(2011a).HealthandwelfareeffectsofintegratingAIDStreatmentwithfood assistanceinresourceconstrainedsettings:asystematicreviewoftheoryandevidence. SocialScience&Medicine,73(5),685‐692. Tirivayi,N.,&Groot,W.(2011b).LaboursupplyresponsestointegratingAIDStreatmentwithin‐ kindtransfers:evidencefromZambia.Masstricht,theNetherlands:MaastrichtUniversity. 116 Torres‐Madriz,G.,Lerner,D.,Ruthazer,R.,Rogers,W.H.,&Wilson,I.B.(2010).Work‐related barriersandfacilitatorstoantiretroviraltherapyadherenceinpersonslivingwithHIV infection.AIDSandBehavior,15(7),7. Torres,K.,Zive,M.,Scolari,R.,Olshefsky,A.,&Zúñiga,M.(2008).Acceptanceofanutrition curriculumforHIV‐positiveLatinosLivingontheU.S.‐Mexicoborder.Journalof TransculturalNursing,19(2),107‐113. Townsend,M.S.,Peerson,J.,Love,B.,Achterberg,C.,&Murphy,S.P.(2001).Foodinsecurityis positivelyrelatedtooverweightinwomen.JournalofNutrition,131(6),1738‐1745. Tsai,A.C.,Bangsberg,D.R.,Emenyonu,N.,Senkungu,J.K.,Martin,J.N.,&Weiser,S.D.(2011).The socialcontextoffoodinsecurityamongpersonslivingwithHIV/AIDSinruralUganda.Social Science&Medicine,73(12),1717‐1724. Tsai,A.C.,Bangsberg,D.R.,Frongillo,E.A.,Hunt,P.W.,Muzoora,C.,Martin,J.N.,etal.(2012).Food insecurity,depressionandthemodifyingroleofsocialsupportamongpeoplelivingwith HIV/AIDSinruralUganda.SocialScience&Medicine,74(12),2012‐2019. Tsai,A.C.,Weiser,S.D.,Petersen,M.L.,Ragland,K.,Kushel,M.B.,&Bangsberg,D.R.(2010).A marginalstructuralmodeltoestimatethecausaleffectofantidepressantmedication treatmentonviralsuppressionamonghomelessandmarginallyhousedpersonswithHIV. ArchivesofGeneralPsychiatry,67(12),1282. Uauy,R.,Albala,C.,&Kain,J.(2001).ObesitytrendsinLatinAmerica:transitingfromunder‐to overweight.JournalofNutrition,131(3),893S‐899S. UNAIDS.(2008).HIV,FoodSecurityandNutritionPolicyBrief.Geneva:JointUnitedNations ProgrammeonHIV/AIDS. UNAIDS.(2011).UNAIDSWorldAIDSDayReport2011.Geneva,Switzerland:UNAIDS. UnitedNationsDevelopmentProgramme.(2010).RegionalHumanDevelopmentReportforLatin AmericaandtheCaribbean2010.ActingontheFuture:BreakingtheIntergenerational TransmissionofInequality.NewYork,NY:UNDP. USDA.(2008).U.S.HouseholdFoodSecuritySurveyModule:Six‐ItemShortForm. http://www.ers.usda.gov/briefing/foodsecurity/surveytools/short2008.pdf. vanOosterhout,J.J.,Ndekha,M.D.,Moore,E.,Kumwenda,J.J.,Zijlstra,E.E.,&Manary,M.(2010). ThebenefitofsupplementaryfeedingforwastedMalawianadultsinitiatingART.AIDSCare, 22(6),737‐742. Wagner,G.,Ryan,G.,Huynh,A.,Kityo,C.,&Mugyenyi,P.(2009).Aqualitativeanalysisofthe economicimpactofHIVandantiretroviraltherapyonindividualsandhouseholdsin Uganda.AIDSPatientCareSTDS,23(9),793‐798. 117 Wagner,G.J.,Holloway,I.,Ghosh‐Dastidar,B.,Kityo,C.,&Mugyenyi,P.(2010).Understandingthe influenceofdepressiononself‐efficacy,workstatus,andcondomuseamongHIVclientsin Uganda.JournalofPsychosomaticResearch. Wang,E.A.,McGinnis,K.A.,Fiellin,D.A.,Goulet,J.L.,Bryant,K.,Gibert,C.L.,etal.(2011).Food InsecurityisassociatedwithpoorvirologicresponseamongHIV‐infectedpatientsreceiving antiretroviralmedications.JournalofGeneralInternalMedicine,26(9),6. Weber,R.(1990).BasicContentAnalysis.NewburyPark,CA:Sage. Weiser,S.,Bangsberg,D.,Kegeles,S.,Ragland,K.,Kushel,M.,&Frongillo,E.(2009a).Foodinsecurity amonghomelessandmarginallyhousedindividualslivingwithHIV/AIDSinSanFrancisco. AIDSBehav,13(5),841‐848. Weiser,S.,Fernandes,K.,Brandson,E.,Lima,V.,Anema,A.,Bangsberg,D.,etal.(2009b).The associationbetweenfoodinsecurityandmortalityamongHIV‐infectedindividualson HAART.JAIDSJournalofAcquiredImmuneDeficiencySyndromes,52(3),342‐349. Weiser,S.,Frongillo,E.A.,Ragland,K.,Hogg,R.S.,Riley,E.D.,&Bangsberg,D.R.(2009c).Food insecurityisassociatedwithincompleteHIVRNAsuppressionamonghomelessand marginallyhousedHIV‐infectedindividualsinSanFrancisco.JournalofGeneralInternal Medicine,24(1),14‐20. Weiser,S.,Tsai,A.C.,Gupta,R.,Frongillo,E.A.,Kawuma,A.,Senkugu,J.,etal.(2012).Food insecurityisassociatedwithmorbidityandpatternsofhealthcareutilizationamongHIV‐ infectedindividualsinruralUganda.AIDS,26(1),67‐75. Weiser,S.,Tuller,D.,Frongillo,E.,Senkungu,J.,Mukiibi,N.,Bangsberg,D.,etal.(2010).Food InsecurityasabarriertosustainedantiretroviraltherapyadherenceinUganda.PLoSONE, 5(4),357‐366. Weiser,S.,Wolfe,W.,Bangsberg,D.,Thior,I.,Gilbert,P.,Makhema,J.,etal.(2003).Barriersto antiretroviraladherenceforpatientslivingwithHIVinfectionandAIDSinBotswana.JAIDS JournalofAcquiredImmuneDeficiencySyndromes,34(3),281. Wilde,P.E.(2007).Measuringtheeffectoffoodstampsonfoodinsecurityandhunger:research andpolicyconsiderations.JournalofNutrition,137(2),307‐310. Wools‐Kaloustian,K.,Kimaiyo,S.,Diero,L.,Siika,A.,Sidle,J.,Yiannoutsos,C.T.,etal.(2006). Viabilityandeffectivenessoflarge‐scaleHIVtreatmentinitiativesinsub‐SaharanAfrica: experiencefromwesternKenya.AIDS,20(1),41. WorldBank.(2007).HIV/AIDS,NutritionandFoodSecuirty:WhatWeCanDo‐ASynthesisof InternationalGuidance.WashingtonDC:WorldBank. WorldHealthOrganization.BMIclassification. http://apps.who.int/bmi/index.jsp?introPage=intro_3.html.Geneva. 118 WorldHealthOrganization.(2004).Nutritioncounseling,careandsupportforHIV‐infectedwomen. GuidelinesonHIV‐relatedcare,treatmentandsupportforHIV‐infectedwomenandtheir childreninresource‐constrainedsettings.Geneva,Switzerland:WorldHealthOrganization. WorldHealthOrganization.(2008).Essentialpreventionandcareinterventionsforadultsand adolescentslivingwithHIVinresource‐limitedsettings.Geneva:WorldHealth Organization. Worthington,C.,O'Brien,K.,Zack,E.,McKee,E.,&Oliver,B.(2012).Enhancinglabourforce participationforpeoplelivingwithHIV:amulti‐perspectivesummaryoftheresearch evidence.AIDSandBehavior,16(1),231‐243. Wu,A.,Revicki,D.,Jacobson,D.,&Malitz,F.(1997).Evidenceforreliability,validityandusefulness oftheMedicalOutcomesStudyHIVHealthSurvey(MOS‐HIV).QualityofLifeResearch,6(6), 481‐493. Wu,D.Y.,Munoz,M.,Espiritu,B.,Zeladita,J.,Sanchez,E.,Callacna,M.,etal.(2008).Burdenof depressionamongimpoverishedHIV‐positivewomeninPeru.JAIDSJournalofAcquired ImmuneDeficiencySyndromes,48(4),500. Yager,J.E.,Kadiyala,S.,&Weiser,S.D.(2011).HIV/AIDS,foodsupplementationandlivelihood programsinUganda:awayforward?PLoSONE,6(10),e26117. Yen,S.T.,Andrews,M.,Chen,Z.,&Eastwood,D.B.(2008).FoodStampProgramparticipationand foodinsecurity:aninstrumentalvariablesapproach.AmericanJournalofAgricultural Economics,90(1),117‐132. Yun,L.W.H.,Maravi,M.,Kobayashi,J.S.,Barton,P.L.,&Davidson,A.J.(2005).Antidepressant treatmentimprovesadherencetoantiretroviraltherapyamongdepressedHIV‐infected patients.JAIDSJournalofAcquiredImmuneDeficiencySyndromes,38(4),432. Zachariah,R.,Fitzgerald,M.,Massaquoi,M.,Pasulani,O.,Arnould,L.,Makombe,S.,etal.(2006).Risk factorsforhighearlymortalityinpatientsonantiretroviraltreatmentinaruraldistrictof Malawi.AIDS,20(18),2355. Zagorsky,J.L.,&Smith,P.K.(2009).DoestheUSFoodStampProgramcontributetoadultweight gain?Economics&HumanBiology,7(2),246‐258. Zeger,S.L.,Liang,K.Y.,&Albert,P.S.(1988).Modelsforlongitudinaldata:ageneralizedestimating equationapproach.Biometrics,1049‐1060. Zivin,J.,Thirumurthy,H.,&Goldstein,M.(2009).AIDStreatmentandintrahouseholdresource allocation:Children'snutritionandschoolinginKenya.JournalofPublicEconomics. 119

The RAND Corporation is a nonprofit institution that helps improve... decisionmaking through research and analysis.

Related documents

Products

Support

The RAND Corporation is a nonprofit institution that helps improve... decisionmaking through research and analysis.

Related documents

Add this document to collection(s)

Add this document to saved

Suggest us how to improve StudyLib