Full Terms amp Conditions of access and use can be found athttpwwwtandfonlinecomactionjournalInformationjournalCode=caic20

AIDS CarePsychological and Socio-medical Aspects of AIDSHIV

ISSN 0954-0121 (Print) 1360-0451 (Online) Journal homepage httpwwwtandfonlinecomloicaic20

Costs of accessing HIV testing services amongrural Malawi communities

Linda Sande Hendramoorthy Maheswaran Collin Mangenah LawrenceMwenge Pitchaya Indravudh Phillip Mkandawire Nurilign Ahmed MarcdrsquoElbee Cheryl Johnson Karin Hatzold Elizabeth L Corbett MelissaNeuman amp Fern Terris-Prestholt

To cite this article Linda Sande Hendramoorthy Maheswaran Collin Mangenah LawrenceMwenge Pitchaya Indravudh Phillip Mkandawire Nurilign Ahmed Marc drsquoElbee Cheryl JohnsonKarin Hatzold Elizabeth L Corbett Melissa Neuman amp Fern Terris-Prestholt (2018) Costs ofaccessing HIV testing services among rural Malawi communities AIDS Care 30sup3 27-36 DOI1010800954012120181479032

To link to this article httpsdoiorg1010800954012120181479032

copy 2018 The Author(s) Published by InformaUK Limited trading as Taylor amp FrancisGroup

Published online 08 Jul 2018

Submit your article to this journal

Article views 245

View Crossmark data

HOUSEHOLD ECONOMIC STRENGTHENING

Costs of accessing HIV testing services among rural Malawi communitiesLinda Sandeab Hendramoorthy Maheswaranc Collin Mangenahd Lawrence Mwengee Pitchaya IndravudhabPhillip Mkandawireg Nurilign Ahmedb Marc drsquoElbeeb Cheryl Johnsonh Karin Hatzoldi Elizabeth L CorbettajMelissa Neumank and Fern Terris-Prestholtb

aMalawi-Liverpool-Wellcome Trust Clinical Research Programme Blantyre Malawi bFaculty of Public Health amp Policy London School of Hygieneamp Tropical Medicine London UK cInstitute of Psychology Health and Society University of Liverpool Liverpool UK dThe Centre for SexualHealth and HIV AIDS Research (CeSHHAR) Harare Zimbabwe eZambart Lusaka Zambia gPopulation Services International Lilongwe MalawihDepartment of HIVAIDS World Health Organisation Geneva Switzerland iPopulation Services International Harare Zimbabwe jFaculty ofInfectious and Tropical Diseases London School of Hygiene amp Tropical Medicine London UK kFaculty of Epidemiology and Population HealthLondon School of Hygiene amp Tropical Medicine London UK

ABSTRACTHIV testing is free in Malawi but users may still incur costs that can deter or delay them accessingthese services We sought to identify and quantify these costs among HIV testing service clients inMalawi We asked residents of communities participating in a cluster randomised trial investigatingthe impact of HIV self-testing about their past HIV testing experiences and the direct non-medicaland indirect costs incurred to access HIV testing We recruited 749 participants whose most recentHIV test was within the past 12 months The mean total cost to access testing was US$245 (95CIUS$211ndashUS$270) Men incurred higher costs (US$381 95CI US$291ndashUS$450) than women (US$183 95CI US$161ndashUS$200) Results from a two-part multivariable regression analysis suggestthat age testing location time taken to test visiting a facility specifically for an HIV test and districtof residence significantly affected the odds of incurring costs to testing In addition gender wealthage education and district of residence were associated with significant user costs

Abbreviations AIDS Acquired Immune Deficiency Syndrome ANC Antenatal Care ARTAntiRetroviral Therapy CBDA Community-Based Distribution Agent CBHTS Community-BasedHIV Testing Services CRT Cluster Randomized trial GLM Generalised Linear Model HIV HumanImmunodeficiency Virus HIVST HIV Self-Testing HTC HIV Testing and Counselling IHSIntegrated Household Survey OLS Ordinary Least Squares PCA Principal Component AnalysisPITC Provider Initiated Testing and Counselling PLHIV People Living with HIV STAR Self-TestingAfRica TB Tuberculosis TPM Two-Part Model UNAIDS The Joint United Nations Programme onHIVAIDS VCT Voluntary Counselling and Testing

ARTICLE HISTORYReceived 4 March 2018Accepted 17 May 2018

KEYWORDSHIV HIV testing andcounseling total costs

Introduction

Eastern and Southern Africa account for the highestnumbers of people living with HIV (PLHIV) newlyinfected with HIV and dying from HIV (UNAIDS2017) HIV testing is an essential gateway to HIV pre-vention treatment care and support services sincereceipt of an HIV diagnosis empowers individuals tomake informed decisions about follow on services inthe cascade (World Health Organization 2015 WorldHealth Organization amp UNAIDS 2017) The global enti-ties involved in AIDS eradication have adopted ambi-tious treatment targets by 2020 90 of all PLHIV willknow their HIV status 90 of all people with diagnosedHIV infection will receive sustained antiretroviraltherapy (ART) and 90 of all people receiving ART

will have viral suppression (UNAIDS 2014a) Ensuringthat 90 of PLHIV are aware of their status will supportenrolment in HIV care and achievement of these globaltreatment goals (UNAIDS 2014a)

However despite impressive efforts in scaling-upavailability of HIV testing and treatment services in theregion including freely available HIV testing at nearlyall healthcare settings testing uptake remains inadequateto reach the global goals (Church et al 2017) Malawihas been leading the way in scaling-up HIV services(Lowrance et al 2008 UNAIDS 2014b) but an esti-mated 35 of men and 18 of women have never testedfor HIV and 60 of young people aged 15ndash19 years havenever tested (CDC amp GoM 2017) Uptake of HIV testingalso remains low amongst poorer individuals and those

copy 2018 The Author(s) Published by Informa UK Limited trading as Taylor amp Francis GroupThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (httpcreativecommonsorglicensesby40) which permits unrestricted usedistribution and reproduction in any medium provided the original work is properly cited

CONTACT Linda Sande lindasandelshtmacuk Faculty of Public Health amp Policy London School of Hygiene and Tropical Medicine Keppel StreetLondon WC1E 7HT UK

AIDS CARE2018 VOL 30 NO S3 27ndash36httpsdoiorg1010800954012120181479032

with less formal education (Kim Skordis-Worrall Hagh-parast-Bidgoli amp Pulkki-Braumlnnstroumlm 2016)

Previous studies in sub-Saharan Africa have citedlocation distance waiting time costs confidentiality con-cerns low perceived risk and infrequent contact with thehealth-care system as barriers to accessing HIV testing(Angotti et al 2009 Morin et al 2006 Musheke et al2013 Sharma Ying Tarr amp Barnabas 2015) Individualsoften incur substantial access costs when utilising publicsector HIV testing and treatment services even whenthey are provided free at point of use (Chimbindi et al2015 Lubega et al 2013 Maheswaran et al 2016Pinto Lettow Rachlis Chan amp Sodhi 2013)

In urban settings HIV testers incur costs close totwice their daily earning incomes (Maheswaran et al2016) These costs are likely to be higher in more ruralsettings however little is known about these costs andwhether these vary by different population groups ortesting modalities which limits efforts to minimise oroffset testing costs to increase uptake Awareness ofcosts incurred by rural HIV testers is particularly impor-tant since 84 of the Malawi population is rural with57 of the rural population classified as poor comparedto 17 of the urban population (International MonetaryFund 2017 World Bank 2014) The poor in developingcountries like Malawi are even less likely than the betteroff to receive effective health care with existing costsbarriers proposed as one of the deterrents of this lowuse (OrsquoDonnell 2007 Russel 2004)

The World Health Organisation (WHO) guidelineshave highlighted the need for strategic approaches todeliver HIV testing services (HTS) (World HealthOrganisation 2016) HIV self-testing (HIVST) and com-munity-based HIV testing are proposed as having thepotential of increasing testing uptake especially formen key populations and young people who wouldnot normally access HIV testing services (Malawi Minis-try of Health 2016 World Health Organisation 2016)Young people for instance have previously demon-strated an aversion to price due to their limited accessto resources (Indravudh et al 2017 Sibanda Maringwaet al 2017) Research on these costs is essential to appro-priately targeting these sub-populations lagging behindin access to testing

In this study we sought to examine (1) the costsborne by users of HIV testing services in rural Malawi(2) whether certain population subgroups incur highercosts and (3) whether costs differ based on the modeof testing To the best of our knowledge this is thefirst study to identify and quantify specific costs ofHIV testing in a rural setting Other studies in theregion have explored determinants of testing (Camlinet al 2016 Helleringer Kohler Frimpong amp

Mkandawire 2009 Leacutepine Terris-Prestholt amp Vicker-man 2014) costs of providing HIV services (Mahes-waran et al 2016 Mangenah Mwenge et al 2017Mwenge et al 2017 Sharma et al 2015) and costsof accessing tuberculosis (TB) treatment (KempMann Simwaka Salaniponi amp Squire 2007) andART (Bergmann Wanyenze amp Stockman 2017 Chim-bindi et al 2015 Pinto et al 2013 Rosen KetlhapileSanne amp DeSilva 2007) The few that have exploredcosts associated with HIV testing have either focusedon urban settings (Maheswaran et al 2016) or exam-ined costs without considering lost income (Bergmannet al 2017) The results of this study will inform thedesign of future HIV testing services and interventionsaimed at overcoming financial barriers to testing

Methods

Study setting and design

HIV testing in Malawi is freely provided Individualsmay voluntarily access HIV testing at a health facilitymay be advised to test by a health professional [provi-der-initiated testing and counseling (PITC)] may beoffered testing as part of routine antenatal care (ANC)(accessed by both the pregnant women and their accom-panying male partners) or TB care (also a form of PITC)or may have access to community-based HIV testing ser-vices (CBHTS) including through testing campaigns andoutreach home-based or door-to-door testing work-place testing mobile testing and testing through edu-cational institutions

We undertook a baseline household survey as part of acluster-randomised trial (CRT) investigating the impactof community-based distribution of HIVST in ruralMalawi (ClinicalTrialsgov Identifier NCT02718274)The CRT was conducted in rural villages of BlantyreMachinga Mwanza and Neno in Southern MalawiThe CRT comprised a population of approximately62500 residents with 22 clusters defined by the servicecatchment area of public primary health facilities withactive ART clinics The HIV prevalence in the four dis-tricts was approximately 11 (National Statistics Officeamp ICF Macro 2017)

Within each cluster villages were selected forinclusion in the baseline survey based on locationpopulation size road accessibility and presence ofpre-existing reproductive health community-based dis-tribution agents Households in these evaluation vil-lages were randomly sampled for a baselinehousehold survey which was conducted between Mayand August 2016 The sampling of the survey ensuredinclusion of at least 250 adults per cluster with the

28 L SANDE ET AL

sample size calculated based on the primary outcome ofthe trial All household members aged 16 years or olderwere eligible to participate in the survey Details on thesample size calculation for the main trial can be foundin the trial protocol available at httphivstarlshtmacuk

Research assistants visited selected households andadministered an electronic face-to-face questionnaireto all household members aged above 16 years whoagreed to participate The main questionnaire includedquestions about sociodemographics and HIV testing his-tory Due to time and resource constraints an extendedquestionnaire was administered to a random 20 subsetof participants responding to the main questionnaireThe extended questionnaire included questions on thecosts of HIV testing as well as other questions on healthcare utilisation and stigma

Assessing costs and location of HIV testing

Participants who reported testing within the previous 12months were asked the location of testing includingwhether facility- or community-based if their mostrecent test was accessed separately from other health ser-vices or as part of antenatal care ANC or PITC total timetaken to access HIV testing and the direct non-medicaland indirect costs they incurred The 12 months recallperiod is in line with other studies on health care useandor out-of-pocket expenditure (van Doorslaer ampMasseria 2004 Heijink Xu Saksana amp Evans 2011)and a similar recall period is used to collect householdnon-food expenditures in the Malawi integrated house-hold survey which is a major socio-economic survey con-ducted by the Malawi National Statistical Office It isworth noting that there is no general answer to the ques-tion of optimal recall period with the choice dependenton the primary objective of the data collection (ClarkeFiebig amp Gerdtham 2008)

We derived a list of potential costs based on the litera-ture and previous work undertaken in Malawi to informdevelopment of the study questionnaire (Kemp et al2007 Maheswaran et al 2016 Pinto et al 2013) Weasked participants how much they had paid for theround trip to the testing facility (transport cost) and ifthey had paid any consultation or service fees (consul-tation cost) related to testing (sometimes incurred at pri-vate facilities) excluding any fees for other services theyaccessed at the same time Participants were also asked ifthey spent money on any food and drink items (foodcosts) while accessing testing and if so how muchthey spent Additionally we asked participants aboutany costs they might have incurred by paying a caretakerto watch their children for the time they sought testing

(child care costs) and about any other costs theymight have incurred as they sought testing (othercosts) We further asked participants to approximatethe amount of money they would have earned duringthe entire time they took to access testing (lost income)

Other covariates

Participants were also asked questions on socio-demo-graphics (age gender and education) the number ofchildren they have and ownership of eight householdassets1 We estimated household wealth using the princi-pal component analysis (PCA) method with householdassets as a proxy for wealth (Filmer amp Pritchett 2001)and we further classified wealth into quintiles Table 1further summarises all the covariates

Ethical approvals were obtained from the College ofMedicine Research Ethics Committee in Malawi andthe Research Ethics Committee of the London Schoolof Hygiene and Tropical Medicine We obtained writteninformed consent from all participants in the extendedquestionnaire before their interview

Statistical methods

All analysis was undertaken in STATA version 140(Stata Corporation Texas USA) Costs were estimatedin 2016 Malawi Kwacha (MWK) and converted to2016 US dollars at an exchange rate of MWK 72989US$ (Reserve Bank of Malawi 2017)

Cost data were categorised into direct non-medicalcosts and indirect costs Direct non-medical costsincluded those directly incurred by participants andindirect costs refer to productivity and income lossesdue to accessing testing services We include data forthe entire sample who had complete cost data and pre-sent it using means with 95 confidence intervals Toassess the burden imposed on participants we comparedtheir total direct non-medical and indirect costs with thenational poverty line of US$120day The poverty linewas adopted from the Third Malawi Integrated House-hold Survey (IHS) of 2011 converted to US$ at the aver-age 2011 exchange rate of MWK16284US$ (NationalStatistics Office 2012 World Bank 2018) and adjustedfor inflation using the national gross domestic product(GDP) deflator for 2011 of 14 (World Bank 2018)

To determine the significant predictors of costs weestimated a multivariable two-part model (TPM) Indi-vidual-level user cost data pose estimation challengessince individual-level medical expenditures or costs oftreatment typically feature a spike at zero and arestrongly skewed with a heavy right-hand tail (Jones2010) There is no unique way to deal with these

AIDS CARE 29

sample size calculated based on the primary outcome ofthe trial All household members aged 16 years or olderwere eligible to participate in the survey Details on thesample size calculation for the main trial can be foundin the trial protocol available at httphivstarlshtmacuk

Research assistants visited selected households andadministered an electronic face-to-face questionnaireto all household members aged above 16 years whoagreed to participate The main questionnaire includedquestions about sociodemographics and HIV testing his-tory Due to time and resource constraints an extendedquestionnaire was administered to a random 20 subsetof participants responding to the main questionnaireThe extended questionnaire included questions on thecosts of HIV testing as well as other questions on healthcare utilisation and stigma

Assessing costs and location of HIV testing

Participants who reported testing within the previous 12months were asked the location of testing includingwhether facility- or community-based if their mostrecent test was accessed separately from other health ser-vices or as part of antenatal care ANC or PITC total timetaken to access HIV testing and the direct non-medicaland indirect costs they incurred The 12 months recallperiod is in line with other studies on health care useandor out-of-pocket expenditure (van Doorslaer ampMasseria 2004 Heijink Xu Saksana amp Evans 2011)and a similar recall period is used to collect householdnon-food expenditures in the Malawi integrated house-hold survey which is a major socio-economic survey con-ducted by the Malawi National Statistical Office It isworth noting that there is no general answer to the ques-tion of optimal recall period with the choice dependenton the primary objective of the data collection (ClarkeFiebig amp Gerdtham 2008)

We derived a list of potential costs based on the litera-ture and previous work undertaken in Malawi to informdevelopment of the study questionnaire (Kemp et al2007 Maheswaran et al 2016 Pinto et al 2013) Weasked participants how much they had paid for theround trip to the testing facility (transport cost) and ifthey had paid any consultation or service fees (consul-tation cost) related to testing (sometimes incurred at pri-vate facilities) excluding any fees for other services theyaccessed at the same time Participants were also asked ifthey spent money on any food and drink items (foodcosts) while accessing testing and if so how muchthey spent Additionally we asked participants aboutany costs they might have incurred by paying a caretakerto watch their children for the time they sought testing

(child care costs) and about any other costs theymight have incurred as they sought testing (othercosts) We further asked participants to approximatethe amount of money they would have earned duringthe entire time they took to access testing (lost income)

Other covariates

Participants were also asked questions on socio-demo-graphics (age gender and education) the number ofchildren they have and ownership of eight householdassets1 We estimated household wealth using the princi-pal component analysis (PCA) method with householdassets as a proxy for wealth (Filmer amp Pritchett 2001)and we further classified wealth into quintiles Table 1further summarises all the covariates

Ethical approvals were obtained from the College ofMedicine Research Ethics Committee in Malawi andthe Research Ethics Committee of the London Schoolof Hygiene and Tropical Medicine We obtained writteninformed consent from all participants in the extendedquestionnaire before their interview

Statistical methods

All analysis was undertaken in STATA version 140(Stata Corporation Texas USA) Costs were estimatedin 2016 Malawi Kwacha (MWK) and converted to2016 US dollars at an exchange rate of MWK 72989US$ (Reserve Bank of Malawi 2017)

Cost data were categorised into direct non-medicalcosts and indirect costs Direct non-medical costsincluded those directly incurred by participants andindirect costs refer to productivity and income lossesdue to accessing testing services We include data forthe entire sample who had complete cost data and pre-sent it using means with 95 confidence intervals Toassess the burden imposed on participants we comparedtheir total direct non-medical and indirect costs with thenational poverty line of US$120day The poverty linewas adopted from the Third Malawi Integrated House-hold Survey (IHS) of 2011 converted to US$ at the aver-age 2011 exchange rate of MWK16284US$ (NationalStatistics Office 2012 World Bank 2018) and adjustedfor inflation using the national gross domestic product(GDP) deflator for 2011 of 14 (World Bank 2018)

To determine the significant predictors of costs weestimated a multivariable two-part model (TPM) Indi-vidual-level user cost data pose estimation challengessince individual-level medical expenditures or costs oftreatment typically feature a spike at zero and arestrongly skewed with a heavy right-hand tail (Jones2010) There is no unique way to deal with these

AIDS CARE 29

estimation challenges associated with cost data with lit-erature recommending that the choice of appropriateestimation approach should be determined by theresearch questions and the characteristics of the data(Buntin amp Zaslavsky 2004 Diehr Yanez Ash Horn-brook amp Lin 1999 Gregori et al 2011 Griswold Parmi-giani Potosky amp Lipscomb 2004) The commonproposed estimation approaches are the log-transformedOLS Tobit model TPM and generalised linear models(GLM) with a log-link function (Buntin amp Zaslavsky2004 Gregori et al 2011 Griswold et al 2004 Jones2010 Nichols 2010)

A Tobit regression model and a TPM were better fitfor our data as they are both able to handle excess zer-oes and positive distribution associated with cost data(Jones 2010) GLM and log-transformed ordinaryleast squares (OLS) on the other hand do not takeinto account the excess zeroes in the data and therefore

generates biased estimates We therefore estimated alog-transformed Tobit and a TPM with a logit modelfor the first part and log-transformed OLS regressionfor the second part Given our main objective a TPMis the appropriate estimation approach as it can dis-tinguish the probability of incurring costs for testingand assess significant cost drivers for those whoincurred costs

To account for the clustering of the data by district afixed effect approach was used We then applied a likeli-hood ratio test to identify the most parsimonious modelbetween the restricted and unrestricted TPMmodels Wefurther identified the most appropriate functional formfor age (testing for non-linearity) using the likelihood-ratio test and did not find significant justification forthis quadratic relationship

We explored socio-demographic and socio-economicvariables and accessibility of testing centres as

Table 1 Descriptive statisticsVariable Regression Inclusion Expected Direction

Gender IndicatorMen (reference group)Women

Men are expected to incur higher costs than women to reflect their higherearning potential relative to women

Age (Years) Indicator16ndash19 Years 20ndash24 Years 25ndash39 Years 40ndash64

Years 65+ YearsFinancial productivity is expected to increase with age starting from age 20 henceraising the opportunity cost to testing up to age 65

Education IndicatorNo Formal education (reference group)Incomplete Primary educationSome Secondary EducationComplete Secondary Education or higher

Education as a proxy for earning potential implying that the higher the level ofeducation the higher the cost for testing

Number of Children Continuous The participantrsquos number ofchildren

Number of children is positively associated with any child care costs a participantmight have incurred while accessing testing hence increasing the total costsincurred

Test Location IndicatorFacility-Based Testing (reference group)Community HTCOther Place

Community-based HTC reduces logistic barriers hence lowers the opportunitycost of testingOther place testing depends on where the person tested for example if at hometesting eg self-testing then lower costs than facility-based testing

Amount of Time Takento Receive Testing

Continuous Time taken (including travel) inhours to access HIV testing

The more time taken away from work to seek testing the higher the cost oftesting through lost income

Reason for visitingTesting Centre

Indicator

Had other reasons for visiting a testing centreaside from HIV testing (reference group)

Visited a testing centre specifically for an HIVtest

Visiting a testing centre for other reasons aside from HIV testing has potential ofeconomies of scope hence reduced total costs

Wealth Index IndicatorHouseholds are ranked into wealth quintiles

with the poorest as the reference groupWealth is a proxy for ability to pay the higher the wealth quintile the higher theparticipantrsquos expenditure to access testing

District of Residence IndicatorBlantyre District (Reference Group)Machinga DistrictMwanza DistrictNeno District

There should not be difference in costs of testing by district

30 L SANDE ET AL

determinants of total costs

ln (Total Costsi + 1) = fDistrict GenderWealthhhAge categories EducationNumber of Children

TimeTaken (Hours) Reason for visiting testing centre

[ ]

To reduce the skewness in the cost data we modelled thecosts using a log transformationWe log transformed usercosts as ln (Total Costsi + 1) as suggested by the literature(McCuneGrace ampUrban 2002) Table 1 summarises thea priori direction of association of the determinants

Results

Participantsrsquo characteristics

A total of 5551 participants were recruited into the base-line survey and 1388 responded to the extended ques-tionnaire Seven hundred and forty-nine (14)participants reported having had at least one HIV testin the previous 12 months making them eligible forthis sub-study Baseline characteristics of these 749 par-ticipants are presented in Table 2 In brief 32 of theparticipants were men 33 of the participants wereaged 16ndash24 years and 18 had no formal educationMost of the participants (83) reported facility-basedtesting as their most recent testing approach Amongthose who tested in a facility more participants (76)accessed testing through PITC In addition menreported spending an average of 29 h and womenreported spending an average of 35 h to access testingservices

Direct non-medical and indirect costs

Direct non-medical and indirect costs stratified by gen-der and cost-category are summarised in Table 3Twenty percent of the participants incurred zero costsfor testing The median cost for participants whoincurred costs was US$206 The mean total cost per par-ticipant was US$245 (95CI US$211ndashUS$270) withlost income accounting for 83 of the total costs Menincurred higher mean total costs than women US$381(95CI US$291ndashUS$450) versus US$183 (95CIUS$161ndashUS$200)

Cost determinants

The logit component of the TPM demonstrated thatage testing location time taken to acquire a test visit-ing a facility specifically for an HIV test and district ofresidence significantly affected the odds of incurringcosts for testing The odds of incurring testing costsare 18 higher for participants aged between 25ndash39years than participants aged between 16ndash19 years In

addition participants who tested within their commu-nities (mobile testing) had 61 lower odds of incurringcosts than participants who tested at facilities Eachadditional hour spent seeking testing increased theodds of incurring costs by 48 Participants who vis-ited a testing site specifically for an HIV test had48 higher odds of incurring costs for testing thanthose who accessed testing in addition to other healthcare services And finally residence in Mwanza districtwas associated with 95 higher odds of incurring costswhen compared to residence in Blantyre district(Tables 4 and 5)

Table 2 Participant characteristics (n = 749)aMen (n = 237

32)Women (n = 512

68)

N Percentage N Percentage

Age (Years) 16ndash19 23 98 52 10220ndash24 35 148 135 26425ndash39 96 407 205 4040ndash64 63 267 102 19965+ 19 81 18 35

Education No formal Edu 19 80 112 219Primary Edu 160 675 331 647Some SecondaryEdu

38 160 57 111

CompleteSecondary orHigher Edu

20 84 12 23

WealthIndexbc

Lowest Quintile 64 270 227 4432nd LowestQuintile

40 169 57 111

Middle Quintile 28 118 69 1352nd HighestQuintile

45 190 70 137

Highest Quintile 60 253 89 174Test Location HospitalClinic

Health Centre148 625 295 576

ANC Clinic 17 72 106 207VCT Centre 24 101 31 61CommunityMobile HTC

47 198 74 145

Other TestingPlace

1 042 6 11

Number ofChildren

Mean (min-max) 3 (0ndash12) 3 (0ndash13)

Reason forfacility visit

HIV Test 168 709 283 553HIV Test + OtherServices

69 291 229 447

Time Taken le1 h 73 308 104 2031ndash3 h 83 350 181 3543ndash6 h 66 279 182 356gt6 h 15 63 45 88

District Blantyre 62 262 147 287Machinga 70 295 172 336Mwanza 30 127 51 10Neno 75 317 142 277

a3 Participants had incomplete databWealth index estimated through undertaking principal component analysisof responses to asset ownership and housing environment

cAssets selected in the baseline data did not do well in differentiating thepoorest from one another

AIDS CARE 31

On the other hand the log-transformed OLS com-ponent of the TPM demonstrated that gender agewealth education and district of residence was associatedwith significant user costs Holding everything else con-stant men on average incurred 52 higher costs for test-ing than women

Older age groups incurred significantly higher coststhan the 16ndash19 age group Participants aged between20ndash24 years 25ndash39 years incurred 61 and 96 highercosts respectively than participants aged between 16ndash19 years Participants aged between 40ndash64 years and

65+ years on average incurred more than double and74 higher costs respectively than participants agedbetween 16ndash19 years There was no difference in averagetesting costs among participants with lower than com-plete secondary education and those without any formaleducation However participants with complete second-ary education or higher on average incurred 63 highercosts than those with no formal education Finally par-ticipants in Mwanza district incurred on average 43higher costs than participants resident in Blantyredistrict

Table 3 Direct non-medical and indirect costs by gender and cost category

Men (US$) Women (US$) Total Sample (US$)

Cost CategoryMean

(95 CI) of MenMean

(95 CI) of

WomenMean95 CI of Total Sample

Directnon-medical costs

Transport 025(015ndash036)

66 016(011ndash022)

87 019(014ndash024)

78

Consultation 003(000ndash005)

08 003(001ndash004)

16 003(001ndash004)

12

Food 018(014ndash022)

47 013(010ndash015)

71 014(012ndash017)

57

Other 005(002ndash009)

13 002(001ndash004)

11 003(002ndash005)

12

IndirectCosts

Child Care 006(002ndash011)

16 001(000ndash003)

06 003(001ndash005)

12

Lost Incomea 324(245ndash403)

850 148(131ndash165)

809 203(175ndash231)

829

Total direct non-medical and indirectcost

381(291ndash450)

100 183(161ndash200)

100 245(211ndash270)

100

aLost Income had a median cost of US$137 US$206 for men and US$096 for women

Table 4 Multivariable analysis of log-transformed Tobit regression model (Dependent Variable total direct non-medical and indirectcosts)

Determinants (Reference Category) Coefficient 95 CI P-value

Gender (Male)Female minus0323 (minus)0457ndash(minus)0189 0000

Wealth (Lowest Quintile)2nd Lowest Quintile minus0049 (minus)0239ndash0141 0613Middle Quintile 0169 (minus)0024ndash0362 00862nd Highest Quintile 0003 (minus)0176ndash0182 0975Highest Quintile 0175 0007ndash0343 0041

Age (Years) (16ndash19)20ndash24 0411 0178ndash0643 000125ndash39 0640 0406ndash0873 000040ndash64 0685 0395ndash0974 000065+ 0195 (minus)0169ndash056 0293

Education No Formal EduPrimary Edu 0013 (minus)0151ndash0177 0877Incomplete Secondary Edu 0253 0017ndash0489 0036Complete Secondary or Higher 0530 0198ndash0863 0002

Children No of Children 0000 (minus)0033ndash0034 0982Testing Location Facility

Community minus0396 (minus)0571ndash(minus)0220 0000Other minus0175 (minus)0858ndash0508 0614

Time Taken Time (Hours) 0049 0023ndash0077 0000Reason for visiting HIV Test + Other

HIV Test 0079 (minus)0045ndash0204 0211District Blantyre

Machinga 0059 (minus)0097ndash0214 0460Mwanza 0350 0139ndash0560 0001Neno minus0007 minus0164ndash0149 0927Constant 0208 (minus)0113ndash0529 0164Observations 746a

Note p lt 001 p lt 005 p lt 01a3 observations had incomplete data

32 L SANDE ET AL

Discussion

This study examined the costs borne by users whenaccessing HIV testing services in rural villages ofSouthern Malawi Our findings indicate that the averagecost of accessing HIV testing in rural Malawi is less thanthat reported in urban areas of the country (US$309 pertest) (Maheswaran et al 2016) yet rural testers incurcosts that are equivalent to twice the daily minimumincome required for their basic needs (national povertyline at US$120 a day) (National Statistics Office 2012)In a country where at least 51 of the population livebelow the national poverty line and 71 live below theinternational poverty line of US$190 a day (NationalStatistics Office 2012 World Bank 2014) these costsare likely to be prohibitive for a large proportion of thepopulation

Our study also demonstrated that there are signifi-cant average cost differences between men (US$381)

and women (US$183) Historically there has beenlow uptake of HIV testing and poor linkage into careamongst men relative to women particularly in sub-Saharan Africa (Camlin et al 2016) It is likely thatthese high costs have contributed to the lower uptakeSeeking testing imposes both a direct non-medicalcost but also the lost opportunity cost of hours awayfrom productive activities (Angotti et al 2009 Ganesh2015 Musheke et al 2013 Wolff et al 2005) Ourfindings show that these opportunity costs comprise asignificant proportion (83) of the total testing costsin this population For most the prospect of learningtheir HIV status may not be a sufficient incentive tobear these costs (Angotti et al 2009) unless they arealready sick This is further evidenced by the large pro-portion of men in our sample who accessed testingthrough PITC (70) and very few who voluntarilyattended facilities for the sole purpose of learningtheir HIV status (10) suggesting that most men inrural Malawi access testing as an add-on to other healthcare services rather than seeking out testingindependently

The large proportion of total costs associated with lostincome was driven by long travel times and long waitingtimes at testing facilities On average participants spentthree hours to access HIV testing services with menspending less time (29 h) than women (35 h) Similarlong wait times (34 h) were observed among adults uti-lising public sector HIV and TB services in South Africa(Chimbindi et al 2015) Taking measures to improveefficiency at HIV testing facilities such as increasingstaffing for this service could reduce waiting times andtherefore reduce the time taken from employment andother activities

Delivering HIV testing closer to peoplersquos homes or attimes convenient to users may also mitigate financialbarriers to testing We found that community-basedtesting is associated with a lower probability of incur-ring costs than facility-based testing therefore decentra-lising testing services beyond static facilities may benecessary to increase uptake The popularity especiallyamong men of community-based HIV testing andHIVST models has been previously demonstrated(Angotti et al 2009 Choko et al 2015 Morin et al2006 Mwenge et al 2017 Sebapathy Van den BerghFidler Hayes amp Ford 2012 Sharma et al 2015World Health Organization 2015) HIVST and otherhome-based testing can be advantageous in that theysubstantially reduce or completely eliminate costsborne by users when testing (Maheswaran et al 2016Sharma et al 2015)

Financial and non-financial incentives also offer analternative to reducing or offsetting testing costs and

Table 5 Multivariable analysis of Two-Part Model on total directnon-medical and indirect costs with first part (logit) and secondpart (Log-transformed OLS)

Determinants (Reference Category)

Two-Part Model

logitLog-transformed

OLS

Gender (Male)Female minus0221 minus0517

Wealth (Lowest Quintile)2nd Lowest Quintile minus0196 minus00113Middle Quintile minus0108 03982nd Highest Quintile minus0168 00644Highest Quintile 0342 0161

Age (Years) (16ndash19)20ndash24 0468 061025ndash39 0777 096440ndash64 0674 103165+ minus0323 0736

Education (No Formal Edu)Primary Edu 0177 minus00569IncompleteSecondary Edu

0430 0248

Complete SecondaryEdu

0951 0628

Number ofChildren

No of Children 00604 minus00164

TestingLocation

(Facility)Community testing minus0946 minus0204Other minus0820 00617

Time Taken Time (Hours) 0203 00161(00530) (00197)

Reason forvisiting

(HIV Test + Other)HIV Test 0393 00374

District (Blantyre)Machinga 0253 00857Mwanza 0666 0434Neno minus0190 00594Constant minus00902 minus0118Observations 746a 746a

Pseudo R2 0116Adjusted R2 01579Log Likelihood minus33504519 minus84703399Note p lt 001 p lt 005 p lt 01a3 observations had incomplete data

AIDS CARE 33

promoting uptake Small non-monetary incentives areassociated with significantly increased community test-ing and HIV case diagnosis (Sibanda Tumushimeet al 2017) It is worth noting that although smallfinancial incentives have been effective in increasinghealth care uptake (Choko et al 2017 MangenahSibanda et al 2017 Pettifor MacPhail Nguyen ampRosenberg 2012) different amounts of incentiveshave different levels of effectiveness Incentives thatcover transport and opportunity costs are generallyassociated with better testing and linkage to care thanincentives equivalent to transport reimbursement only(Choko et al 2017)

Study limitations and strengths

Our study used retrospective interviews to collect expen-diture data for participantsrsquo most recent HIV test Thisapproach introduces potential for recall bias We limitedthis recall bias by recruiting participants with an HIV testwithin a period of 12 months preceding the interview Inaddition there is potential for downward bias of the test-ing costs because individuals with prohibitively highexpected costs will not have tested Our follow-upresearch will explore more advanced statistical modelsto reduce this downward bias

Despite these limitations our study adds valuableinformation to the literature on access to HIV testingUnlike previous studies we included lost income as acost to testing which enabled us to determine the fulleconomic burden of testing on users in a rural setting

Conclusion

Though HIV testing services are ldquofreerdquo in Malawiusers incur costs to access these services in ruralparts of the country that are double the national pov-erty line In these contexts men incur higher costs toaccess HIV testing services than women with lostincome as the largest cost component Increasinguptake of testing services especially for men will likelyrequire bringing testing services closer to the commu-nities improving efficiency of facility-based testing andpotentially introducing financial or non-financialincentives as a way to motivate uptake and offset thetotal costs associated with this portion of the HIVcascade

Note

1 Asset index Electricity radio working television setmobile phone landline telephone refrigerator and bedwith mattress

Acknowledgements

We acknowledge the participants who generously agreed torespond to the questionnaires the entire Population ServicesInternational-Malawi research and implementation teamsresearchers and data teams from the London School ofHygiene and Tropical Medicine and the Malawi-LiverpoolWellcome Trust Clinical Research Programme The Datawill be deposited to London School of Hygiene amp TropicalMedicinersquos digital repository (httpdatacompasslshtmacuk)

Disclosure statement

No potential conflict of interest was reported by the authors

Funding

This research is under Self-Testing AfRica (STAR) Project isfunded by UNITAID [grant number PO8477-0-600] ELCis funded by a Wellcome Trust Senior Research Fellowshipin Clinical Science [grant number WT200901Z16Z]

References

Angotti N Bula A Gaydosh L Zeev Kimchi E ThorntonR L amp Yeatman S E (2009) Increasing the acceptabilityof HIV counseling and testing with three CrsquosConvenience confidentiality and credibility Social Scienceamp Medicine 68(12) 2263ndash2270 doi101016jsocscimed200902041

Bergmann J N Wanyenze R K amp Stockman J K (2017)The cost of accessing infant HIV medications and healthservices in Uganda AIDS Care 29(11) 1426ndash1432

Buntin M B amp Zaslavsky A M (2004) Too much ado abouttwo-part models and transformation Comparing methodsof modeling Medicare expenditures Journal of HealthEconomics 23(3) 525ndash542 doi101016jjhealeco200310005

Camlin C S Ssemmondo E Chamie G El Ayadi A MKwarisiima D Sang Nhellip Collaboration S (2016) Menldquomissingrdquo from population-based HIV testing Insightsfrom qualitative research AIDS Care 28(Suppl 3) 67ndash73doi1010800954012120161164806

CDC amp GoM (2017) Malawi population-based HIV impactassessment (MPHIA) 2015-16 First report Population-based HIV impact assessment Lilongwe Ministry of Health

Chimbindi N Bor J Newell M Tanser F Baltusen RHontelez Jhellip Baumlrnighausen T (2015) Time and moneyThe true costs of health care utilization for patients receiv-ing ldquofreerdquo HIVTB care and treatment in rural KwaZulu-Natal Journal of Acquired Immune Deficiency Syndromes(1999) 70(2) e52ndashe60

Choko A T Lepine A Maheswaran H Kumwenda MDesmond N Corbett E L amp Fielding K (2017)Improving linkage to treatment and prevention after self-testing among male partners of antenatal care attendeesIn International AIDS society Paris

Choko A T MacPherson P Webb E L Willey B AFeasy H Sambakunsi Rhellip Corbett E L (2015)Uptake accuracy safety and linkage into care over two

34 L SANDE ET AL

determinants of total costs

ln (Total Costsi + 1) = fDistrict GenderWealthhhAge categories EducationNumber of Children

TimeTaken (Hours) Reason for visiting testing centre

[ ]

To reduce the skewness in the cost data we modelled thecosts using a log transformationWe log transformed usercosts as ln (Total Costsi + 1) as suggested by the literature(McCuneGrace ampUrban 2002) Table 1 summarises thea priori direction of association of the determinants

Results

Participantsrsquo characteristics

A total of 5551 participants were recruited into the base-line survey and 1388 responded to the extended ques-tionnaire Seven hundred and forty-nine (14)participants reported having had at least one HIV testin the previous 12 months making them eligible forthis sub-study Baseline characteristics of these 749 par-ticipants are presented in Table 2 In brief 32 of theparticipants were men 33 of the participants wereaged 16ndash24 years and 18 had no formal educationMost of the participants (83) reported facility-basedtesting as their most recent testing approach Amongthose who tested in a facility more participants (76)accessed testing through PITC In addition menreported spending an average of 29 h and womenreported spending an average of 35 h to access testingservices

Direct non-medical and indirect costs

Direct non-medical and indirect costs stratified by gen-der and cost-category are summarised in Table 3Twenty percent of the participants incurred zero costsfor testing The median cost for participants whoincurred costs was US$206 The mean total cost per par-ticipant was US$245 (95CI US$211ndashUS$270) withlost income accounting for 83 of the total costs Menincurred higher mean total costs than women US$381(95CI US$291ndashUS$450) versus US$183 (95CIUS$161ndashUS$200)

Cost determinants

The logit component of the TPM demonstrated thatage testing location time taken to acquire a test visit-ing a facility specifically for an HIV test and district ofresidence significantly affected the odds of incurringcosts for testing The odds of incurring testing costsare 18 higher for participants aged between 25ndash39years than participants aged between 16ndash19 years In

addition participants who tested within their commu-nities (mobile testing) had 61 lower odds of incurringcosts than participants who tested at facilities Eachadditional hour spent seeking testing increased theodds of incurring costs by 48 Participants who vis-ited a testing site specifically for an HIV test had48 higher odds of incurring costs for testing thanthose who accessed testing in addition to other healthcare services And finally residence in Mwanza districtwas associated with 95 higher odds of incurring costswhen compared to residence in Blantyre district(Tables 4 and 5)

Table 2 Participant characteristics (n = 749)aMen (n = 237

32)Women (n = 512

68)

N Percentage N Percentage

Age (Years) 16ndash19 23 98 52 10220ndash24 35 148 135 26425ndash39 96 407 205 4040ndash64 63 267 102 19965+ 19 81 18 35

Education No formal Edu 19 80 112 219Primary Edu 160 675 331 647Some SecondaryEdu

38 160 57 111

CompleteSecondary orHigher Edu

20 84 12 23

WealthIndexbc

Lowest Quintile 64 270 227 4432nd LowestQuintile

40 169 57 111

Middle Quintile 28 118 69 1352nd HighestQuintile

45 190 70 137

Highest Quintile 60 253 89 174Test Location HospitalClinic

Health Centre148 625 295 576

ANC Clinic 17 72 106 207VCT Centre 24 101 31 61CommunityMobile HTC

47 198 74 145

Other TestingPlace

1 042 6 11

Number ofChildren

Mean (min-max) 3 (0ndash12) 3 (0ndash13)

Reason forfacility visit

HIV Test 168 709 283 553HIV Test + OtherServices

69 291 229 447

Time Taken le1 h 73 308 104 2031ndash3 h 83 350 181 3543ndash6 h 66 279 182 356gt6 h 15 63 45 88

District Blantyre 62 262 147 287Machinga 70 295 172 336Mwanza 30 127 51 10Neno 75 317 142 277

a3 Participants had incomplete databWealth index estimated through undertaking principal component analysisof responses to asset ownership and housing environment

cAssets selected in the baseline data did not do well in differentiating thepoorest from one another

AIDS CARE 31

On the other hand the log-transformed OLS com-ponent of the TPM demonstrated that gender agewealth education and district of residence was associatedwith significant user costs Holding everything else con-stant men on average incurred 52 higher costs for test-ing than women

Older age groups incurred significantly higher coststhan the 16ndash19 age group Participants aged between20ndash24 years 25ndash39 years incurred 61 and 96 highercosts respectively than participants aged between 16ndash19 years Participants aged between 40ndash64 years and

65+ years on average incurred more than double and74 higher costs respectively than participants agedbetween 16ndash19 years There was no difference in averagetesting costs among participants with lower than com-plete secondary education and those without any formaleducation However participants with complete second-ary education or higher on average incurred 63 highercosts than those with no formal education Finally par-ticipants in Mwanza district incurred on average 43higher costs than participants resident in Blantyredistrict

Table 3 Direct non-medical and indirect costs by gender and cost category

Men (US$) Women (US$) Total Sample (US$)

Cost CategoryMean

(95 CI) of MenMean

(95 CI) of

WomenMean95 CI of Total Sample

Directnon-medical costs

Transport 025(015ndash036)

66 016(011ndash022)

87 019(014ndash024)

78

Consultation 003(000ndash005)

08 003(001ndash004)

16 003(001ndash004)

12

Food 018(014ndash022)

47 013(010ndash015)

71 014(012ndash017)

57

Other 005(002ndash009)

13 002(001ndash004)

11 003(002ndash005)

12

IndirectCosts

Child Care 006(002ndash011)

16 001(000ndash003)

06 003(001ndash005)

12

Lost Incomea 324(245ndash403)

850 148(131ndash165)

809 203(175ndash231)

829

Total direct non-medical and indirectcost

381(291ndash450)

100 183(161ndash200)

100 245(211ndash270)

100

aLost Income had a median cost of US$137 US$206 for men and US$096 for women

Table 4 Multivariable analysis of log-transformed Tobit regression model (Dependent Variable total direct non-medical and indirectcosts)

Determinants (Reference Category) Coefficient 95 CI P-value

Gender (Male)Female minus0323 (minus)0457ndash(minus)0189 0000

Wealth (Lowest Quintile)2nd Lowest Quintile minus0049 (minus)0239ndash0141 0613Middle Quintile 0169 (minus)0024ndash0362 00862nd Highest Quintile 0003 (minus)0176ndash0182 0975Highest Quintile 0175 0007ndash0343 0041

Age (Years) (16ndash19)20ndash24 0411 0178ndash0643 000125ndash39 0640 0406ndash0873 000040ndash64 0685 0395ndash0974 000065+ 0195 (minus)0169ndash056 0293

Education No Formal EduPrimary Edu 0013 (minus)0151ndash0177 0877Incomplete Secondary Edu 0253 0017ndash0489 0036Complete Secondary or Higher 0530 0198ndash0863 0002

Children No of Children 0000 (minus)0033ndash0034 0982Testing Location Facility

Community minus0396 (minus)0571ndash(minus)0220 0000Other minus0175 (minus)0858ndash0508 0614

Time Taken Time (Hours) 0049 0023ndash0077 0000Reason for visiting HIV Test + Other

HIV Test 0079 (minus)0045ndash0204 0211District Blantyre

Machinga 0059 (minus)0097ndash0214 0460Mwanza 0350 0139ndash0560 0001Neno minus0007 minus0164ndash0149 0927Constant 0208 (minus)0113ndash0529 0164Observations 746a

Note p lt 001 p lt 005 p lt 01a3 observations had incomplete data

32 L SANDE ET AL

Discussion

This study examined the costs borne by users whenaccessing HIV testing services in rural villages ofSouthern Malawi Our findings indicate that the averagecost of accessing HIV testing in rural Malawi is less thanthat reported in urban areas of the country (US$309 pertest) (Maheswaran et al 2016) yet rural testers incurcosts that are equivalent to twice the daily minimumincome required for their basic needs (national povertyline at US$120 a day) (National Statistics Office 2012)In a country where at least 51 of the population livebelow the national poverty line and 71 live below theinternational poverty line of US$190 a day (NationalStatistics Office 2012 World Bank 2014) these costsare likely to be prohibitive for a large proportion of thepopulation

Our study also demonstrated that there are signifi-cant average cost differences between men (US$381)

and women (US$183) Historically there has beenlow uptake of HIV testing and poor linkage into careamongst men relative to women particularly in sub-Saharan Africa (Camlin et al 2016) It is likely thatthese high costs have contributed to the lower uptakeSeeking testing imposes both a direct non-medicalcost but also the lost opportunity cost of hours awayfrom productive activities (Angotti et al 2009 Ganesh2015 Musheke et al 2013 Wolff et al 2005) Ourfindings show that these opportunity costs comprise asignificant proportion (83) of the total testing costsin this population For most the prospect of learningtheir HIV status may not be a sufficient incentive tobear these costs (Angotti et al 2009) unless they arealready sick This is further evidenced by the large pro-portion of men in our sample who accessed testingthrough PITC (70) and very few who voluntarilyattended facilities for the sole purpose of learningtheir HIV status (10) suggesting that most men inrural Malawi access testing as an add-on to other healthcare services rather than seeking out testingindependently

The large proportion of total costs associated with lostincome was driven by long travel times and long waitingtimes at testing facilities On average participants spentthree hours to access HIV testing services with menspending less time (29 h) than women (35 h) Similarlong wait times (34 h) were observed among adults uti-lising public sector HIV and TB services in South Africa(Chimbindi et al 2015) Taking measures to improveefficiency at HIV testing facilities such as increasingstaffing for this service could reduce waiting times andtherefore reduce the time taken from employment andother activities

Delivering HIV testing closer to peoplersquos homes or attimes convenient to users may also mitigate financialbarriers to testing We found that community-basedtesting is associated with a lower probability of incur-ring costs than facility-based testing therefore decentra-lising testing services beyond static facilities may benecessary to increase uptake The popularity especiallyamong men of community-based HIV testing andHIVST models has been previously demonstrated(Angotti et al 2009 Choko et al 2015 Morin et al2006 Mwenge et al 2017 Sebapathy Van den BerghFidler Hayes amp Ford 2012 Sharma et al 2015World Health Organization 2015) HIVST and otherhome-based testing can be advantageous in that theysubstantially reduce or completely eliminate costsborne by users when testing (Maheswaran et al 2016Sharma et al 2015)

Financial and non-financial incentives also offer analternative to reducing or offsetting testing costs and

Table 5 Multivariable analysis of Two-Part Model on total directnon-medical and indirect costs with first part (logit) and secondpart (Log-transformed OLS)

Determinants (Reference Category)

Two-Part Model

logitLog-transformed

OLS

Gender (Male)Female minus0221 minus0517

Wealth (Lowest Quintile)2nd Lowest Quintile minus0196 minus00113Middle Quintile minus0108 03982nd Highest Quintile minus0168 00644Highest Quintile 0342 0161

Age (Years) (16ndash19)20ndash24 0468 061025ndash39 0777 096440ndash64 0674 103165+ minus0323 0736

Education (No Formal Edu)Primary Edu 0177 minus00569IncompleteSecondary Edu

0430 0248

Complete SecondaryEdu

0951 0628

Number ofChildren

No of Children 00604 minus00164

TestingLocation

(Facility)Community testing minus0946 minus0204Other minus0820 00617

Time Taken Time (Hours) 0203 00161(00530) (00197)

Reason forvisiting

(HIV Test + Other)HIV Test 0393 00374

District (Blantyre)Machinga 0253 00857Mwanza 0666 0434Neno minus0190 00594Constant minus00902 minus0118Observations 746a 746a

Pseudo R2 0116Adjusted R2 01579Log Likelihood minus33504519 minus84703399Note p lt 001 p lt 005 p lt 01a3 observations had incomplete data

AIDS CARE 33

promoting uptake Small non-monetary incentives areassociated with significantly increased community test-ing and HIV case diagnosis (Sibanda Tumushimeet al 2017) It is worth noting that although smallfinancial incentives have been effective in increasinghealth care uptake (Choko et al 2017 MangenahSibanda et al 2017 Pettifor MacPhail Nguyen ampRosenberg 2012) different amounts of incentiveshave different levels of effectiveness Incentives thatcover transport and opportunity costs are generallyassociated with better testing and linkage to care thanincentives equivalent to transport reimbursement only(Choko et al 2017)

Study limitations and strengths

Our study used retrospective interviews to collect expen-diture data for participantsrsquo most recent HIV test Thisapproach introduces potential for recall bias We limitedthis recall bias by recruiting participants with an HIV testwithin a period of 12 months preceding the interview Inaddition there is potential for downward bias of the test-ing costs because individuals with prohibitively highexpected costs will not have tested Our follow-upresearch will explore more advanced statistical modelsto reduce this downward bias

Despite these limitations our study adds valuableinformation to the literature on access to HIV testingUnlike previous studies we included lost income as acost to testing which enabled us to determine the fulleconomic burden of testing on users in a rural setting

Conclusion

Though HIV testing services are ldquofreerdquo in Malawiusers incur costs to access these services in ruralparts of the country that are double the national pov-erty line In these contexts men incur higher costs toaccess HIV testing services than women with lostincome as the largest cost component Increasinguptake of testing services especially for men will likelyrequire bringing testing services closer to the commu-nities improving efficiency of facility-based testing andpotentially introducing financial or non-financialincentives as a way to motivate uptake and offset thetotal costs associated with this portion of the HIVcascade

Note

1 Asset index Electricity radio working television setmobile phone landline telephone refrigerator and bedwith mattress

Acknowledgements

We acknowledge the participants who generously agreed torespond to the questionnaires the entire Population ServicesInternational-Malawi research and implementation teamsresearchers and data teams from the London School ofHygiene and Tropical Medicine and the Malawi-LiverpoolWellcome Trust Clinical Research Programme The Datawill be deposited to London School of Hygiene amp TropicalMedicinersquos digital repository (httpdatacompasslshtmacuk)

Disclosure statement

No potential conflict of interest was reported by the authors

Funding

This research is under Self-Testing AfRica (STAR) Project isfunded by UNITAID [grant number PO8477-0-600] ELCis funded by a Wellcome Trust Senior Research Fellowshipin Clinical Science [grant number WT200901Z16Z]

References

Angotti N Bula A Gaydosh L Zeev Kimchi E ThorntonR L amp Yeatman S E (2009) Increasing the acceptabilityof HIV counseling and testing with three CrsquosConvenience confidentiality and credibility Social Scienceamp Medicine 68(12) 2263ndash2270 doi101016jsocscimed200902041

Bergmann J N Wanyenze R K amp Stockman J K (2017)The cost of accessing infant HIV medications and healthservices in Uganda AIDS Care 29(11) 1426ndash1432

Buntin M B amp Zaslavsky A M (2004) Too much ado abouttwo-part models and transformation Comparing methodsof modeling Medicare expenditures Journal of HealthEconomics 23(3) 525ndash542 doi101016jjhealeco200310005

Camlin C S Ssemmondo E Chamie G El Ayadi A MKwarisiima D Sang Nhellip Collaboration S (2016) Menldquomissingrdquo from population-based HIV testing Insightsfrom qualitative research AIDS Care 28(Suppl 3) 67ndash73doi1010800954012120161164806

CDC amp GoM (2017) Malawi population-based HIV impactassessment (MPHIA) 2015-16 First report Population-based HIV impact assessment Lilongwe Ministry of Health

Chimbindi N Bor J Newell M Tanser F Baltusen RHontelez Jhellip Baumlrnighausen T (2015) Time and moneyThe true costs of health care utilization for patients receiv-ing ldquofreerdquo HIVTB care and treatment in rural KwaZulu-Natal Journal of Acquired Immune Deficiency Syndromes(1999) 70(2) e52ndashe60

Choko A T Lepine A Maheswaran H Kumwenda MDesmond N Corbett E L amp Fielding K (2017)Improving linkage to treatment and prevention after self-testing among male partners of antenatal care attendeesIn International AIDS society Paris

Choko A T MacPherson P Webb E L Willey B AFeasy H Sambakunsi Rhellip Corbett E L (2015)Uptake accuracy safety and linkage into care over two

34 L SANDE ET AL

On the other hand the log-transformed OLS com-ponent of the TPM demonstrated that gender agewealth education and district of residence was associatedwith significant user costs Holding everything else con-stant men on average incurred 52 higher costs for test-ing than women

Older age groups incurred significantly higher coststhan the 16ndash19 age group Participants aged between20ndash24 years 25ndash39 years incurred 61 and 96 highercosts respectively than participants aged between 16ndash19 years Participants aged between 40ndash64 years and

65+ years on average incurred more than double and74 higher costs respectively than participants agedbetween 16ndash19 years There was no difference in averagetesting costs among participants with lower than com-plete secondary education and those without any formaleducation However participants with complete second-ary education or higher on average incurred 63 highercosts than those with no formal education Finally par-ticipants in Mwanza district incurred on average 43higher costs than participants resident in Blantyredistrict

Table 3 Direct non-medical and indirect costs by gender and cost category

Men (US$) Women (US$) Total Sample (US$)

Cost CategoryMean

(95 CI) of MenMean

(95 CI) of

WomenMean95 CI of Total Sample

Directnon-medical costs

Transport 025(015ndash036)

66 016(011ndash022)

87 019(014ndash024)

78

Consultation 003(000ndash005)

08 003(001ndash004)

16 003(001ndash004)

12

Food 018(014ndash022)

47 013(010ndash015)

71 014(012ndash017)

57

Other 005(002ndash009)

13 002(001ndash004)

11 003(002ndash005)

12

IndirectCosts

Child Care 006(002ndash011)

16 001(000ndash003)

06 003(001ndash005)

12

Lost Incomea 324(245ndash403)

850 148(131ndash165)

809 203(175ndash231)

829

Total direct non-medical and indirectcost

381(291ndash450)

100 183(161ndash200)

100 245(211ndash270)

100

aLost Income had a median cost of US$137 US$206 for men and US$096 for women

Table 4 Multivariable analysis of log-transformed Tobit regression model (Dependent Variable total direct non-medical and indirectcosts)

Determinants (Reference Category) Coefficient 95 CI P-value

Gender (Male)Female minus0323 (minus)0457ndash(minus)0189 0000

Wealth (Lowest Quintile)2nd Lowest Quintile minus0049 (minus)0239ndash0141 0613Middle Quintile 0169 (minus)0024ndash0362 00862nd Highest Quintile 0003 (minus)0176ndash0182 0975Highest Quintile 0175 0007ndash0343 0041

Age (Years) (16ndash19)20ndash24 0411 0178ndash0643 000125ndash39 0640 0406ndash0873 000040ndash64 0685 0395ndash0974 000065+ 0195 (minus)0169ndash056 0293

Education No Formal EduPrimary Edu 0013 (minus)0151ndash0177 0877Incomplete Secondary Edu 0253 0017ndash0489 0036Complete Secondary or Higher 0530 0198ndash0863 0002

Children No of Children 0000 (minus)0033ndash0034 0982Testing Location Facility

Community minus0396 (minus)0571ndash(minus)0220 0000Other minus0175 (minus)0858ndash0508 0614

Time Taken Time (Hours) 0049 0023ndash0077 0000Reason for visiting HIV Test + Other

HIV Test 0079 (minus)0045ndash0204 0211District Blantyre

Machinga 0059 (minus)0097ndash0214 0460Mwanza 0350 0139ndash0560 0001Neno minus0007 minus0164ndash0149 0927Constant 0208 (minus)0113ndash0529 0164Observations 746a

Note p lt 001 p lt 005 p lt 01a3 observations had incomplete data

32 L SANDE ET AL

Discussion

This study examined the costs borne by users whenaccessing HIV testing services in rural villages ofSouthern Malawi Our findings indicate that the averagecost of accessing HIV testing in rural Malawi is less thanthat reported in urban areas of the country (US$309 pertest) (Maheswaran et al 2016) yet rural testers incurcosts that are equivalent to twice the daily minimumincome required for their basic needs (national povertyline at US$120 a day) (National Statistics Office 2012)In a country where at least 51 of the population livebelow the national poverty line and 71 live below theinternational poverty line of US$190 a day (NationalStatistics Office 2012 World Bank 2014) these costsare likely to be prohibitive for a large proportion of thepopulation

Our study also demonstrated that there are signifi-cant average cost differences between men (US$381)

and women (US$183) Historically there has beenlow uptake of HIV testing and poor linkage into careamongst men relative to women particularly in sub-Saharan Africa (Camlin et al 2016) It is likely thatthese high costs have contributed to the lower uptakeSeeking testing imposes both a direct non-medicalcost but also the lost opportunity cost of hours awayfrom productive activities (Angotti et al 2009 Ganesh2015 Musheke et al 2013 Wolff et al 2005) Ourfindings show that these opportunity costs comprise asignificant proportion (83) of the total testing costsin this population For most the prospect of learningtheir HIV status may not be a sufficient incentive tobear these costs (Angotti et al 2009) unless they arealready sick This is further evidenced by the large pro-portion of men in our sample who accessed testingthrough PITC (70) and very few who voluntarilyattended facilities for the sole purpose of learningtheir HIV status (10) suggesting that most men inrural Malawi access testing as an add-on to other healthcare services rather than seeking out testingindependently

The large proportion of total costs associated with lostincome was driven by long travel times and long waitingtimes at testing facilities On average participants spentthree hours to access HIV testing services with menspending less time (29 h) than women (35 h) Similarlong wait times (34 h) were observed among adults uti-lising public sector HIV and TB services in South Africa(Chimbindi et al 2015) Taking measures to improveefficiency at HIV testing facilities such as increasingstaffing for this service could reduce waiting times andtherefore reduce the time taken from employment andother activities

Delivering HIV testing closer to peoplersquos homes or attimes convenient to users may also mitigate financialbarriers to testing We found that community-basedtesting is associated with a lower probability of incur-ring costs than facility-based testing therefore decentra-lising testing services beyond static facilities may benecessary to increase uptake The popularity especiallyamong men of community-based HIV testing andHIVST models has been previously demonstrated(Angotti et al 2009 Choko et al 2015 Morin et al2006 Mwenge et al 2017 Sebapathy Van den BerghFidler Hayes amp Ford 2012 Sharma et al 2015World Health Organization 2015) HIVST and otherhome-based testing can be advantageous in that theysubstantially reduce or completely eliminate costsborne by users when testing (Maheswaran et al 2016Sharma et al 2015)

Financial and non-financial incentives also offer analternative to reducing or offsetting testing costs and

Table 5 Multivariable analysis of Two-Part Model on total directnon-medical and indirect costs with first part (logit) and secondpart (Log-transformed OLS)

Determinants (Reference Category)

Two-Part Model

logitLog-transformed

OLS

Gender (Male)Female minus0221 minus0517

Wealth (Lowest Quintile)2nd Lowest Quintile minus0196 minus00113Middle Quintile minus0108 03982nd Highest Quintile minus0168 00644Highest Quintile 0342 0161

Age (Years) (16ndash19)20ndash24 0468 061025ndash39 0777 096440ndash64 0674 103165+ minus0323 0736

Education (No Formal Edu)Primary Edu 0177 minus00569IncompleteSecondary Edu

0430 0248

Complete SecondaryEdu

0951 0628

Number ofChildren

No of Children 00604 minus00164

TestingLocation

(Facility)Community testing minus0946 minus0204Other minus0820 00617

Time Taken Time (Hours) 0203 00161(00530) (00197)

Reason forvisiting

(HIV Test + Other)HIV Test 0393 00374

District (Blantyre)Machinga 0253 00857Mwanza 0666 0434Neno minus0190 00594Constant minus00902 minus0118Observations 746a 746a

Pseudo R2 0116Adjusted R2 01579Log Likelihood minus33504519 minus84703399Note p lt 001 p lt 005 p lt 01a3 observations had incomplete data

AIDS CARE 33

promoting uptake Small non-monetary incentives areassociated with significantly increased community test-ing and HIV case diagnosis (Sibanda Tumushimeet al 2017) It is worth noting that although smallfinancial incentives have been effective in increasinghealth care uptake (Choko et al 2017 MangenahSibanda et al 2017 Pettifor MacPhail Nguyen ampRosenberg 2012) different amounts of incentiveshave different levels of effectiveness Incentives thatcover transport and opportunity costs are generallyassociated with better testing and linkage to care thanincentives equivalent to transport reimbursement only(Choko et al 2017)

Study limitations and strengths

Our study used retrospective interviews to collect expen-diture data for participantsrsquo most recent HIV test Thisapproach introduces potential for recall bias We limitedthis recall bias by recruiting participants with an HIV testwithin a period of 12 months preceding the interview Inaddition there is potential for downward bias of the test-ing costs because individuals with prohibitively highexpected costs will not have tested Our follow-upresearch will explore more advanced statistical modelsto reduce this downward bias

Despite these limitations our study adds valuableinformation to the literature on access to HIV testingUnlike previous studies we included lost income as acost to testing which enabled us to determine the fulleconomic burden of testing on users in a rural setting

Conclusion

Though HIV testing services are ldquofreerdquo in Malawiusers incur costs to access these services in ruralparts of the country that are double the national pov-erty line In these contexts men incur higher costs toaccess HIV testing services than women with lostincome as the largest cost component Increasinguptake of testing services especially for men will likelyrequire bringing testing services closer to the commu-nities improving efficiency of facility-based testing andpotentially introducing financial or non-financialincentives as a way to motivate uptake and offset thetotal costs associated with this portion of the HIVcascade

Note

1 Asset index Electricity radio working television setmobile phone landline telephone refrigerator and bedwith mattress

Acknowledgements

We acknowledge the participants who generously agreed torespond to the questionnaires the entire Population ServicesInternational-Malawi research and implementation teamsresearchers and data teams from the London School ofHygiene and Tropical Medicine and the Malawi-LiverpoolWellcome Trust Clinical Research Programme The Datawill be deposited to London School of Hygiene amp TropicalMedicinersquos digital repository (httpdatacompasslshtmacuk)

Disclosure statement

No potential conflict of interest was reported by the authors

Funding

This research is under Self-Testing AfRica (STAR) Project isfunded by UNITAID [grant number PO8477-0-600] ELCis funded by a Wellcome Trust Senior Research Fellowshipin Clinical Science [grant number WT200901Z16Z]

References

Angotti N Bula A Gaydosh L Zeev Kimchi E ThorntonR L amp Yeatman S E (2009) Increasing the acceptabilityof HIV counseling and testing with three CrsquosConvenience confidentiality and credibility Social Scienceamp Medicine 68(12) 2263ndash2270 doi101016jsocscimed200902041

Bergmann J N Wanyenze R K amp Stockman J K (2017)The cost of accessing infant HIV medications and healthservices in Uganda AIDS Care 29(11) 1426ndash1432

Buntin M B amp Zaslavsky A M (2004) Too much ado abouttwo-part models and transformation Comparing methodsof modeling Medicare expenditures Journal of HealthEconomics 23(3) 525ndash542 doi101016jjhealeco200310005

Camlin C S Ssemmondo E Chamie G El Ayadi A MKwarisiima D Sang Nhellip Collaboration S (2016) Menldquomissingrdquo from population-based HIV testing Insightsfrom qualitative research AIDS Care 28(Suppl 3) 67ndash73doi1010800954012120161164806

CDC amp GoM (2017) Malawi population-based HIV impactassessment (MPHIA) 2015-16 First report Population-based HIV impact assessment Lilongwe Ministry of Health

Chimbindi N Bor J Newell M Tanser F Baltusen RHontelez Jhellip Baumlrnighausen T (2015) Time and moneyThe true costs of health care utilization for patients receiv-ing ldquofreerdquo HIVTB care and treatment in rural KwaZulu-Natal Journal of Acquired Immune Deficiency Syndromes(1999) 70(2) e52ndashe60

Choko A T Lepine A Maheswaran H Kumwenda MDesmond N Corbett E L amp Fielding K (2017)Improving linkage to treatment and prevention after self-testing among male partners of antenatal care attendeesIn International AIDS society Paris

Choko A T MacPherson P Webb E L Willey B AFeasy H Sambakunsi Rhellip Corbett E L (2015)Uptake accuracy safety and linkage into care over two

34 L SANDE ET AL

Discussion

This study examined the costs borne by users whenaccessing HIV testing services in rural villages ofSouthern Malawi Our findings indicate that the averagecost of accessing HIV testing in rural Malawi is less thanthat reported in urban areas of the country (US$309 pertest) (Maheswaran et al 2016) yet rural testers incurcosts that are equivalent to twice the daily minimumincome required for their basic needs (national povertyline at US$120 a day) (National Statistics Office 2012)In a country where at least 51 of the population livebelow the national poverty line and 71 live below theinternational poverty line of US$190 a day (NationalStatistics Office 2012 World Bank 2014) these costsare likely to be prohibitive for a large proportion of thepopulation

Our study also demonstrated that there are signifi-cant average cost differences between men (US$381)

and women (US$183) Historically there has beenlow uptake of HIV testing and poor linkage into careamongst men relative to women particularly in sub-Saharan Africa (Camlin et al 2016) It is likely thatthese high costs have contributed to the lower uptakeSeeking testing imposes both a direct non-medicalcost but also the lost opportunity cost of hours awayfrom productive activities (Angotti et al 2009 Ganesh2015 Musheke et al 2013 Wolff et al 2005) Ourfindings show that these opportunity costs comprise asignificant proportion (83) of the total testing costsin this population For most the prospect of learningtheir HIV status may not be a sufficient incentive tobear these costs (Angotti et al 2009) unless they arealready sick This is further evidenced by the large pro-portion of men in our sample who accessed testingthrough PITC (70) and very few who voluntarilyattended facilities for the sole purpose of learningtheir HIV status (10) suggesting that most men inrural Malawi access testing as an add-on to other healthcare services rather than seeking out testingindependently

The large proportion of total costs associated with lostincome was driven by long travel times and long waitingtimes at testing facilities On average participants spentthree hours to access HIV testing services with menspending less time (29 h) than women (35 h) Similarlong wait times (34 h) were observed among adults uti-lising public sector HIV and TB services in South Africa(Chimbindi et al 2015) Taking measures to improveefficiency at HIV testing facilities such as increasingstaffing for this service could reduce waiting times andtherefore reduce the time taken from employment andother activities

Delivering HIV testing closer to peoplersquos homes or attimes convenient to users may also mitigate financialbarriers to testing We found that community-basedtesting is associated with a lower probability of incur-ring costs than facility-based testing therefore decentra-lising testing services beyond static facilities may benecessary to increase uptake The popularity especiallyamong men of community-based HIV testing andHIVST models has been previously demonstrated(Angotti et al 2009 Choko et al 2015 Morin et al2006 Mwenge et al 2017 Sebapathy Van den BerghFidler Hayes amp Ford 2012 Sharma et al 2015World Health Organization 2015) HIVST and otherhome-based testing can be advantageous in that theysubstantially reduce or completely eliminate costsborne by users when testing (Maheswaran et al 2016Sharma et al 2015)

Financial and non-financial incentives also offer analternative to reducing or offsetting testing costs and

Table 5 Multivariable analysis of Two-Part Model on total directnon-medical and indirect costs with first part (logit) and secondpart (Log-transformed OLS)

Determinants (Reference Category)

Two-Part Model

logitLog-transformed

OLS

Gender (Male)Female minus0221 minus0517

Wealth (Lowest Quintile)2nd Lowest Quintile minus0196 minus00113Middle Quintile minus0108 03982nd Highest Quintile minus0168 00644Highest Quintile 0342 0161

Age (Years) (16ndash19)20ndash24 0468 061025ndash39 0777 096440ndash64 0674 103165+ minus0323 0736

Education (No Formal Edu)Primary Edu 0177 minus00569IncompleteSecondary Edu

0430 0248

Complete SecondaryEdu

0951 0628

Number ofChildren

No of Children 00604 minus00164

TestingLocation

(Facility)Community testing minus0946 minus0204Other minus0820 00617

Time Taken Time (Hours) 0203 00161(00530) (00197)

Reason forvisiting

(HIV Test + Other)HIV Test 0393 00374

District (Blantyre)Machinga 0253 00857Mwanza 0666 0434Neno minus0190 00594Constant minus00902 minus0118Observations 746a 746a

Pseudo R2 0116Adjusted R2 01579Log Likelihood minus33504519 minus84703399Note p lt 001 p lt 005 p lt 01a3 observations had incomplete data

AIDS CARE 33

promoting uptake Small non-monetary incentives areassociated with significantly increased community test-ing and HIV case diagnosis (Sibanda Tumushimeet al 2017) It is worth noting that although smallfinancial incentives have been effective in increasinghealth care uptake (Choko et al 2017 MangenahSibanda et al 2017 Pettifor MacPhail Nguyen ampRosenberg 2012) different amounts of incentiveshave different levels of effectiveness Incentives thatcover transport and opportunity costs are generallyassociated with better testing and linkage to care thanincentives equivalent to transport reimbursement only(Choko et al 2017)

Study limitations and strengths

Our study used retrospective interviews to collect expen-diture data for participantsrsquo most recent HIV test Thisapproach introduces potential for recall bias We limitedthis recall bias by recruiting participants with an HIV testwithin a period of 12 months preceding the interview Inaddition there is potential for downward bias of the test-ing costs because individuals with prohibitively highexpected costs will not have tested Our follow-upresearch will explore more advanced statistical modelsto reduce this downward bias

Despite these limitations our study adds valuableinformation to the literature on access to HIV testingUnlike previous studies we included lost income as acost to testing which enabled us to determine the fulleconomic burden of testing on users in a rural setting

Conclusion

Though HIV testing services are ldquofreerdquo in Malawiusers incur costs to access these services in ruralparts of the country that are double the national pov-erty line In these contexts men incur higher costs toaccess HIV testing services than women with lostincome as the largest cost component Increasinguptake of testing services especially for men will likelyrequire bringing testing services closer to the commu-nities improving efficiency of facility-based testing andpotentially introducing financial or non-financialincentives as a way to motivate uptake and offset thetotal costs associated with this portion of the HIVcascade

Note

1 Asset index Electricity radio working television setmobile phone landline telephone refrigerator and bedwith mattress

Acknowledgements

We acknowledge the participants who generously agreed torespond to the questionnaires the entire Population ServicesInternational-Malawi research and implementation teamsresearchers and data teams from the London School ofHygiene and Tropical Medicine and the Malawi-LiverpoolWellcome Trust Clinical Research Programme The Datawill be deposited to London School of Hygiene amp TropicalMedicinersquos digital repository (httpdatacompasslshtmacuk)

Disclosure statement

No potential conflict of interest was reported by the authors

Funding

This research is under Self-Testing AfRica (STAR) Project isfunded by UNITAID [grant number PO8477-0-600] ELCis funded by a Wellcome Trust Senior Research Fellowshipin Clinical Science [grant number WT200901Z16Z]

References

Angotti N Bula A Gaydosh L Zeev Kimchi E ThorntonR L amp Yeatman S E (2009) Increasing the acceptabilityof HIV counseling and testing with three CrsquosConvenience confidentiality and credibility Social Scienceamp Medicine 68(12) 2263ndash2270 doi101016jsocscimed200902041

Bergmann J N Wanyenze R K amp Stockman J K (2017)The cost of accessing infant HIV medications and healthservices in Uganda AIDS Care 29(11) 1426ndash1432

Buntin M B amp Zaslavsky A M (2004) Too much ado abouttwo-part models and transformation Comparing methodsof modeling Medicare expenditures Journal of HealthEconomics 23(3) 525ndash542 doi101016jjhealeco200310005

Camlin C S Ssemmondo E Chamie G El Ayadi A MKwarisiima D Sang Nhellip Collaboration S (2016) Menldquomissingrdquo from population-based HIV testing Insightsfrom qualitative research AIDS Care 28(Suppl 3) 67ndash73doi1010800954012120161164806

CDC amp GoM (2017) Malawi population-based HIV impactassessment (MPHIA) 2015-16 First report Population-based HIV impact assessment Lilongwe Ministry of Health

Chimbindi N Bor J Newell M Tanser F Baltusen RHontelez Jhellip Baumlrnighausen T (2015) Time and moneyThe true costs of health care utilization for patients receiv-ing ldquofreerdquo HIVTB care and treatment in rural KwaZulu-Natal Journal of Acquired Immune Deficiency Syndromes(1999) 70(2) e52ndashe60

Choko A T Lepine A Maheswaran H Kumwenda MDesmond N Corbett E L amp Fielding K (2017)Improving linkage to treatment and prevention after self-testing among male partners of antenatal care attendeesIn International AIDS society Paris

Choko A T MacPherson P Webb E L Willey B AFeasy H Sambakunsi Rhellip Corbett E L (2015)Uptake accuracy safety and linkage into care over two

34 L SANDE ET AL

promoting uptake Small non-monetary incentives areassociated with significantly increased community test-ing and HIV case diagnosis (Sibanda Tumushimeet al 2017) It is worth noting that although smallfinancial incentives have been effective in increasinghealth care uptake (Choko et al 2017 MangenahSibanda et al 2017 Pettifor MacPhail Nguyen ampRosenberg 2012) different amounts of incentiveshave different levels of effectiveness Incentives thatcover transport and opportunity costs are generallyassociated with better testing and linkage to care thanincentives equivalent to transport reimbursement only(Choko et al 2017)

Study limitations and strengths

Our study used retrospective interviews to collect expen-diture data for participantsrsquo most recent HIV test Thisapproach introduces potential for recall bias We limitedthis recall bias by recruiting participants with an HIV testwithin a period of 12 months preceding the interview Inaddition there is potential for downward bias of the test-ing costs because individuals with prohibitively highexpected costs will not have tested Our follow-upresearch will explore more advanced statistical modelsto reduce this downward bias

Despite these limitations our study adds valuableinformation to the literature on access to HIV testingUnlike previous studies we included lost income as acost to testing which enabled us to determine the fulleconomic burden of testing on users in a rural setting

Conclusion

Though HIV testing services are ldquofreerdquo in Malawiusers incur costs to access these services in ruralparts of the country that are double the national pov-erty line In these contexts men incur higher costs toaccess HIV testing services than women with lostincome as the largest cost component Increasinguptake of testing services especially for men will likelyrequire bringing testing services closer to the commu-nities improving efficiency of facility-based testing andpotentially introducing financial or non-financialincentives as a way to motivate uptake and offset thetotal costs associated with this portion of the HIVcascade

Note

1 Asset index Electricity radio working television setmobile phone landline telephone refrigerator and bedwith mattress

Acknowledgements

We acknowledge the participants who generously agreed torespond to the questionnaires the entire Population ServicesInternational-Malawi research and implementation teamsresearchers and data teams from the London School ofHygiene and Tropical Medicine and the Malawi-LiverpoolWellcome Trust Clinical Research Programme The Datawill be deposited to London School of Hygiene amp TropicalMedicinersquos digital repository (httpdatacompasslshtmacuk)

Disclosure statement

No potential conflict of interest was reported by the authors

Funding

This research is under Self-Testing AfRica (STAR) Project isfunded by UNITAID [grant number PO8477-0-600] ELCis funded by a Wellcome Trust Senior Research Fellowshipin Clinical Science [grant number WT200901Z16Z]

References

Angotti N Bula A Gaydosh L Zeev Kimchi E ThorntonR L amp Yeatman S E (2009) Increasing the acceptabilityof HIV counseling and testing with three CrsquosConvenience confidentiality and credibility Social Scienceamp Medicine 68(12) 2263ndash2270 doi101016jsocscimed200902041

Bergmann J N Wanyenze R K amp Stockman J K (2017)The cost of accessing infant HIV medications and healthservices in Uganda AIDS Care 29(11) 1426ndash1432

Buntin M B amp Zaslavsky A M (2004) Too much ado abouttwo-part models and transformation Comparing methodsof modeling Medicare expenditures Journal of HealthEconomics 23(3) 525ndash542 doi101016jjhealeco200310005

Camlin C S Ssemmondo E Chamie G El Ayadi A MKwarisiima D Sang Nhellip Collaboration S (2016) Menldquomissingrdquo from population-based HIV testing Insightsfrom qualitative research AIDS Care 28(Suppl 3) 67ndash73doi1010800954012120161164806

CDC amp GoM (2017) Malawi population-based HIV impactassessment (MPHIA) 2015-16 First report Population-based HIV impact assessment Lilongwe Ministry of Health

Chimbindi N Bor J Newell M Tanser F Baltusen RHontelez Jhellip Baumlrnighausen T (2015) Time and moneyThe true costs of health care utilization for patients receiv-ing ldquofreerdquo HIVTB care and treatment in rural KwaZulu-Natal Journal of Acquired Immune Deficiency Syndromes(1999) 70(2) e52ndashe60

Choko A T Lepine A Maheswaran H Kumwenda MDesmond N Corbett E L amp Fielding K (2017)Improving linkage to treatment and prevention after self-testing among male partners of antenatal care attendeesIn International AIDS society Paris

Choko A T MacPherson P Webb E L Willey B AFeasy H Sambakunsi Rhellip Corbett E L (2015)Uptake accuracy safety and linkage into care over two

34 L SANDE ET AL

years of promoting annual self-testing for HIV in BlantyreMalawi a community-based prospective study PLoSMedicine 12(9) e1001873

Church K Machiyama K Todd J Njamwea BMwangome M Hosegood Vhellip Crampin A (2017)Identifying gaps in HIV service delivery across the diagno-sis-to-treatment cascade Findings from health facility sur-veys in six sub-Saharan countries Journal of theInternational AIDS Society 20(1) 21188

Clarke P M Fiebig D G amp Gerdtham U G (2008)Optimal recall length in survey design Journal of HealthEconomics 27(5) 1275ndash1284 doi101016jjhealeco200805012

Diehr P Yanez D Ash A Hornbrook M amp Lin D Y(1999) Methods for analyzing health care utilization andcosts Annual Review of Public Health 20 125ndash144doi101146annurevpublhealth201125

Filmer D amp Pritchett L H (2001) Estimating wealth effectswithout expenditure datamdashOR tears an application to edu-cational enrollments in States of India Demography 38(1)115ndash132

Ganesh L (2015) Impact of indirect cost on access to health-care utilization International Journal of Medical Science andPublic Health 4(9) 1255ndash1259

Gregori D Petrinco M Bo S Desideri A Merletti F ampPagano E (2011) Regression models for analyzing costsand their determinants in health care An introductoryreview International Journal for Quality in Health Care23(3) 331ndash341 doi101093intqhcmzr010

Griswold M Parmigiani G Potosky A amp Lipscomb J(2004) Analyzing health care costs a comparison of stat-istical methods motivated by Medicare colorectal cancercharges Biostatistics (Oxford England) 1(1) 1ndash23

Heijink R Xu K Saksana P amp Evans D (2011) Validityand comparability of out-of- pocket health expenditurefrom household surveys A review of the literature and cur-rent survey instruments World Health OrganizationDiscussion Paper 012011 1ndash30

Helleringer S Kohler H Frimpong J A amp Mkandawire J(2009) Increasing uptake of HIV testing and counselingamong the poorest in sub-saharan countries throughhome-based service provision Journal of AcquiredImmune Deficiency Syndromes (1999) 51(2) 185ndash193

Indravudh P P Sibanda E L DrsquoElbeacutee M Kumwenda MK Ringwald B Maringwa Ghellip Taegtmeyer M (2017)ldquoI will choose when to test where i want to testrdquo investi-gating young peoplersquos preferences for HIV self-testing inMalawi and Zimbabwe AIDS 31 S203ndashS212 doi101097QAD0000000000001516

International Monetary Fund (2017)Malawi economic devel-opment document Retrieved from httpswwwimforg~mediaFilesPublicationsCR2017cr17184ashx

Jones A M (2010) Models for health care York TheUniversity of York

Kemp J R Mann G Simwaka B N Salaniponi F M L ampSquire S B (2007) Can Malawirsquos poor afford free tubercu-losis services Patient and household costs associated with atuberculosis diagnosis in Lilongwe Bulletin of the WorldHealth Organization 85(8) 580ndash585

Kim S W Skordis-Worrall J Haghparast-Bidgoli H ampPulkki-Braumlnnstroumlm A M (2016) Socio-economic inequityin HIV testing in Malawi Global Health Action 9(1) 31730

Leacutepine A Terris-Prestholt F amp Vickerman P (2014)Determinants of HIV testing among Nigerian couples Amultilevel modelling approach Health Policy andPlanning 30(5) 579ndash592

Lowrance D W Makombe S Harries A D Shiraishi RW Hochgesang M Aberle-Grasse Jhellip Kamoto K(2008) A public health approach to rapid scale-up of anti-retroviral treatment in Malawi during 2004ndash2006 JAIDSJournal of Acquired Immune Deficiency Syndromes 49(3)287ndash293

Lubega M Musenze I A Joshua G Dhafa G Badaza RBakwesegha C J amp Reynolds S J (2013) Sex inequalityhigh transport costs and exposed clinic location Reasonsfor loss to follow-up of clients under prevention ofmother-to-child HIV transmission in Eastern Uganda - aqualitative study Patient Preference and Adherence 7447ndash454 doi102147PPAS19327

Maheswaran H Petrou S MacPherson P Choko A TKumwenda F Lalloo D Ghellip Corbett E L (2016) Costand quality of life analysis of HIV self-testing and facility-based HIV testing and counselling in Blantyre MalawiBMCMedicine 14(34) 493 doi101186s12916-016-0577-7

Malawi Ministry of Health (2016)Malawi HIV testing servicesguidelines

Mangenah C Mwenge L Sande L Sibanda E ChiwawaP Chingwenah Thellip Terris-Prestholt F (2017) Thecosts of community based HIV self-test (HIVST) kit distri-bution Results from 3 districts in Zimbabwe INTERESTconference Lilongwe Malawi

Mangenah C Sibanda E Hatzold K Maringwa GMugurungi O Terris-Prestholt F amp Cowan F M(2017) Economic evaluation of non-financial incentives toincrease couples HIV testing and counselling in ZimbabweInternational AIDS Society Paris Retrieved from httpswwwyoutubecomwatchv=xnEP3egV3xU

McCune B Grace J B amp Urban D L (2002) Data trans-formations Analysis of Ecological Communities 67ndash79

Morin S F Khumalo-Sakutumwa G Charlebois E DRouth J Fritz K Lane Thellip Coates T J (2006)Removing barriers to knowing HIV status same-day mobileHIV testing in Zimbabwe Journal of Acquired ImmuneDefficiency Syndrome 41(2) 218ndash224

Musheke M Ntalasha H Gari S McKenzie O Bond VMartin-Hilber A amp Merten S (2013) A systematic reviewof qualitative findings on factors enabling and deterringuptake of HIV testing in Sub-Saharan Africa BMC PublicHealth 13 442 doi1011861471-2458-13-220

Mwenge L Sande L Mangenah C Ahmed N Kanema SdrsquoElbeacutee Mhellip Johnson C C (2017) Costs of facility-basedHIV testing in Malawi Zambia and Zimbabwe PloS One 12(10) e0185740

National Statistics Office (2012) Third integrated householdsurvey 2010-2011 Household socio-economic characteristicssreport (Vol 3) National Statistics Office Retrieved fromhttpwwwnsomalawimwimagesstoriesdata_on_lineeconomicsihsIHS3IHS3_Reportpdf

National Statistics Office amp ICF Macro (2017)Malawi demo-graphic and health survey 2015ndash16 Rockvile MA NationalStatistics Office

Nichols A (2010) Regression for nonnegative skewed depen-dent variables Retrieved from httpswwwstatacommeetingboston10boston10_nicholspdf

AIDS CARE 35

OrsquoDonnell O (2007) Access to health care in developingcountries breaking down demand side barriers Cadernosde Sauacutede Puacuteblica 23(12) 2820ndash2834 doiS0102-311X2007001200003 [pii]

Pettifor A MacPhail C Nguyen N amp Rosenberg M(2012) Can money prevent the spread of HIV A reviewof cash payments for HIV prevention AIDS and Behavior16(7) 1729ndash1738 doi101007s10461-012-0240-z

Pinto A D Lettow M Rachlis B Chan A K amp Sodhi S K(2013) Patient costs associated with accessing HIVAIDScare in Malawi Journal of the International AIDS Society16(1) 18055

Reserve Bank of Malawi (2017) Exchange rates Reserve Bankof Malawi Retrieved from httpwwwrbmmwStatisticsMajorRates

Rosen S Ketlhapile M Sanne I amp DeSilva M B (2007)Cost to patients of obtaining treatment for HIVAIDS inSouth Africa South African Medical Journal 97(7) 524ndash529

Russel S (2004) The economic burden of illness for house-holds in developing countries A review of studies focusingon malaria tuberculosis and human immunodeficiencyvirusacquired immunodeficiency syndrome AmericanJournal of Tropical Medicine and Hygiene 71 147ndash155doi712_suppl147 [pii]

Sebapathy K Van den Bergh R Fidler S Hayes R amp FordN (2012) Uptake of home-based voluntary HIV testing inSub-Saharan Africa A systematic review and meta-analysisPLoS Medicine 9(12) doi101371journalpmed1001351

Sharma M Ying R Tarr G amp Barnabas R (2015)Systematic review and meta-analysis of community andfacility-based HIV testing to address linkage to care gapsin sub-Saharan Africa Nature 528 S77ndashS85 doi101038nature16044 Retrieved from httpswwwnaturecomarticlesnature16044supplementary-information

Sibanda E Maringwa G Ruhode N Madanhire CTumushime M Watadzaushe Chellip Terris-Prestholt F(2017) Preferences for models of HIV self-test kit distri-bution results from a qualitative study and choice exper-iment in a rural Zimbabwean community Retrieved fromhttphivstorgevidencepreferences-for-models-of-hiv-

self-test-kit-distribution-results-from-a-qualitative-study-and-choice-experiment-in-a-rural-zimbabwean-community

Sibanda E L Tumushime M Mufuka J Mavedzenge SN Gudukeya S Bautista-Arredondo Shellip Padian N(2017) Effect of non-monetary incentives on uptake ofcouplesrsquo counselling and testing among clients attendingmobile HIV services in rural Zimbabwe A cluster-ran-domised trial The Lancet Global Health 5(9) e907ndashe915

UNAIDS (2014a) 90-90-90 An ambitious treatment target tohelp end the AIDS epidemic Geneva Joint United NationsProgramme on HIVAIDS

UNAIDS (2014b) The gap report Geneva Joint UnitedNations Programme on HIVAIDS

UNAIDS (2017) UNAIDS data 2017 Geneva Joint UnitedNations Programme on HIVAIDS

van Doorslaer E amp Masseria C (2004) Income-relatedinequality in the use of medical care in 21 OECD countriesOECD Health Working Papers 5(14) 1ndash89 doi101787687501760705

Wolff B Nyanzi B Katongole G Sssesanga DRuberantwari A amp Whitworth J (2005) Evaluation of ahome-based voluntary counselling and testing interventionin rural Uganda Health Policy and Planning 20(2) 109ndash116 doi101093heapolczi013

World Bank (2014) The World Bank Data Retrieved fromhttpsdataworldbankorgindicatorSPRURTOTLZS

World Bank (2018) Malawi Data Retrieved from httpsdataworldbankorgcountrymalawiview=chart

World Health Organisation (2016) Guidelines on HIV self-testing and partner notification Supplement to consolidatedguidelines of HIV testing services WHO Retrieved fromhttpappswhointirisbitstream1066525165519789241549868-engpdfua=1

World Health Organization (2015) Consolidated Guidelineson HIV Testing Services 5Cs consent confidentiality coun-selling correct results and connection 2015

World Health Organization amp UNAIDS (2017) Statement onHIV testing services Retrieved from httpwwwwhointhivtopicsvcthts-new-opportunitiesen

36 L SANDE ET AL