health financing: a new role for microfinance institutions?

HEALTH FINANCING: A NEW ROLE FORMICROFINANCE INSTITUTIONS?

SHEILA LEATHERMAN1, KIMBERLEY GEISSLER1*, BOBBI GRAY2 and MEGAN GASH2

1Department of Health Policy and Management, Gillings School of Global Public Health, Universityof North Carolina at Chapel Hill, Chapel Hill, NC, USA

2Freedom from Hunger, Davis, CA, USA

Abstract: An innovative and scalable approach, individual or small group health financing bymicrofinance institutions can expand existing options for the poor. We examined productivity losses,direct health costs and health financing methods among microfinance clients in Bolivia, Beninand Burkina Faso. Health costs and lost productivity for households were substantial. Microfinanceinstitutions offer advantages in developing health financing options: global reach, expertise in loansand savings and their mission to facilitate household financial stability. Health financing products,including savings and loans, hold considerable potential to benefit clients but require careful designto optimise value and minimise risk. Copyright © 2012 John Wiley & Sons, Ltd.

Keywords: microfinance; health financing; health costs; global health

1 INTRODUCTION

Poor health and the inability to access health care are key factors both leading to, and resultingfrom, poverty (Narayan and Patesch, 2000). Facilitating access to health-related services isimportant because of the disproportionate burden of disease among the poor (Matherset al., 2008), the burden of health costs and associated risk of further impoverishment(Russell, 2004; McIntyre et al., 2006) and the effect of poor health on productivity, whichjeopardises progress towards economic empowerment (Russell, 2004; McIntyre et al., 2006).There is a clear need to expand health financing options for the poor in developing coun-

tries as affordability represents a major barrier to healthcare access. Existing microfinanceinstitutions (MFIs) are beginning to offer new types of financial products to address thiscritical need. Among innovative and scalable approaches to health financing for the poor,

*Correspondence to: Kimberley Geissler, Department of Health Policy and Management, Gillings School ofGlobal Public Health, University of North Carolina at Chapel Hill, 1101 McGavran-Greenberg Hall, CB 7411,Chapel Hill, NC 27599-7411, USA.E-mail: [email protected]

Copyright © 2012 John Wiley & Sons, Ltd.

Journal of International DevelopmentJ. Int. Dev. (2012)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/jid.2829

MFIs represent a unique opportunity. More than 3500 MFIs globally provide credit andother financial services to more than 155 million households—collectively about one-halfbillion people—in support of income generation and microenterprise. Estimates show atleast 34 million of these households are very poor (Leatherman and Dunford, 2010). MFIscan provide a crucial role by allowing households to save and borrow specifically forhealth costs. Health financing can eliminate some barriers to treatment seeking and helpmodulate health crises and their associated costs (Leatherman et al., 2012).The financial and time costs of illness and treatment are a large burden, representing a

simultaneous assault of high treatment costs and reduced income because of reduction ofproductive labor (Russell, 2004; McIntyre et al., 2006). The size and frequency of these costshave important impacts on financing health expenditures; health costs often come in peaksof intense cost burdens over a few days or weeks instead of being evenly distributed overtime. Households’ ability to cope is influenced by the type, severity and duration of illness(Russell, 2004). Generally, households pay the costs of illness by using current income andsavings, by reducing consumption, borrowing money or selling assets (Whitehead et al.,2001; Russell, 2004; McIntyre et al., 2006). Borrowing money and selling assets resultin additional costs, such as very high interest rates or the devaluation of large assets(e.g. sale of livestock at low prices) accompanied by the sacrifice of future income streams(Leive and Xu, 2008; Kruk et al., 2009). These financing and coping mechanisms may havelasting effects on the ability of households to withstand future shocks, creating a cycle ofeconomic vulnerability and poverty (Whitehead et al., 2001; Leive and Xu, 2008).To address the burden of health costs among the poor, Freedom from Hunger, a

US-based international development organisation, partnered with MFIs to offer integratedfinancial and health-related services in a 4-year demonstration in five countries, known asthe Microfinance and Health Protection (MAHP) initiative, funded by The Bill andMelinda Gates Foundation. The portfolio of services included health education, commu-nity health workers, health savings, health loans, health microinsurance, linkages to healthproviders and the sale of health products (e.g. bed nets and oral rehydration solution). Theinitiative tested whether the integration of health financing for microfinance clients in theform of health savings and health loans could deliver positive health, social and financialimpacts for clients and be financially sustainable for MFIs.In this article, we present data from household level surveys and qualitative interviews.

We document the productivity losses and high health costs experienced by MFI clients inBenin, Bolivia and Burkina Faso; describe issues and barriers in treatment-seekingbehavior; and examine the potential for health financing products offered by MFIs tohelp facilitate timely access to necessary medical care.

2 METHODS

In Burkina Faso and Bolivia, extensive household surveys were conducted with randomlyselected microfinance clients of Le Réseau des Caisses Populaires du Burkina (RCPB,n = 192 households) and Crédito con Educación Rural (CRECER, n = 266 households),respectively. In Benin, surveys were conducted with randomly selected female communitymembers (n = 3623 households) in villages served by the MFI Promotion et l’Appui auDéveloppement de Micro-Entreprises (PADME). The household surveys analysed inBurkina Faso (October–December 2009) and Benin (February, November 2009) werecollected post-intervention. The data from Bolivia (January–February 2007) derive from

S. Leatherman et al.

Copyright © 2012 John Wiley & Sons, Ltd. J. Int. Dev. (2012)DOI: 10.1002/jid

the baseline survey because full data from the endline survey were not available. Foreach of the three countries, we combined data on health costs for household illness fromtreatment and comparison areas to focus on understanding health and health financingneeds of these populations and to increase our sample size for estimates of health costs,which were not significantly different after the intervention. For the analysis of healthfinancing products, we limit the analyses to a subset of clients that were eligible for theintervention in the treatment areas only. For analysis of preventive care use, we comparetreatment and control areas.The surveys included questions about socioeconomic status, incidence of illness episodes,

costs of seeking treatment, financing methods for health costs and time spent incapacitatedby poor health. Additionally, we collected data regarding the impact of poor health onhouseholds and information regarding preferences for health financing options.Health costs were collected for anyone reporting that she or a family member was sick

and sought treatment in the 30 days prior to the survey. To determine the impact of illnesson productivity, days of work missed were recorded for a household member’s own illnessor for the time spent in caregiving for another household member. In Benin, questionsabout food and lodging costs, travel time and costs (including drugs and labs), and numberof days sick were only asked for children in the household; therefore, reported values ofthese measures are likely to underestimate costs and time for the entire household. Someof the questions regarding health costs were asked for all household members; from this,we know that a small proportion of adults were sick relative to children.The seasonality of illness is important in analysing health costs and financing. The sur-

vey in Bolivia was conducted in January and February, reported as low-frequency monthsfor illnesses during market research previously conducted there (Alcon et al., 2006). Thus,our data may underestimate the frequency and annual cost of seeking treatment. In BurkinaFaso, market research found the period October to December, when the survey wasconducted, to be after high season for malaria; however, stomach pains and other illnessesare common during these months. Other periods, particularly during the rainy season, werementioned as times with high healthcare expenditures, indicating our annualised estimateis likely an average period of health expenditures (Traore et al., 2006). The survey inBenin was conducted in two waves, before and after the rainy season, meaning seasonalityissues exist within the data set. However, roughly equal numbers of villages reportedhealth costs before and after the rainy season. Thus, if differences exist between periods,they likely average out within the data set (Gray and Ekoue-Kouvahey, 2010). To accountfor the fact that health costs may vary significantly throughout the year depending onseason, illness types and availability of funds for treatment, we calculated a range ofannualised health costs based on monthly health costs gathered in the survey.Additional details about sampling, survey implementation, qualitative data collection

and household income calculations can be found in the Technical Appendix.

3 RESULTS

Important findings emerge regarding self-report of illness, experience of large losses ofproductive time, exposure to high direct health costs and threats to household resourcesfor health expenditures. These data provide a useful foundation for examining health finan-cing needs, challenges and impact.

Health Financing: A New Role for Microfinance Institutions?


3.1 Participation

Descriptive statistics for respondents are given in Table 1.

3.2 Reports of Illness and Productivity Losses

Wide variation occurred in the percentages of households that reported a householdmember was sick in the previous 30 days, ranging from 58 per cent of respondents inBurkina Faso to just 17–18 per cent in Benin and Bolivia (Table 2). Non-use of medicalcare among those reporting illness was common, ranging from 7 per cent in Bolivia to 40per cent in Benin. Non-use of medical care among the poor may have many reasonsincluding lack of awareness or information, inability to pay, access barriers or simplythe unavailability of adequate health services. Similar estimates reported in the literatureand based on national surveys range from 20 to 86 per cent (Xu et al., 2007).In addition to information requested regarding direct health costs, all respondents who

said someone in the household was ill were also asked to report the number of days a house-hold member was sick in the last 30 days, the number of days missed from work because ofillness or caring for a sick household member, and the time spent seeking treatment(Table 2).Most striking was the number of days reported for people in the household being sick.

The average number of days sick for households reporting a member was ill during theperiod ranged from 6.9 days in Benin to 10.8 days in Bolivia. This includes all householdmembers (e.g. either 1 person sick for 9 days or 3 people sick for 3 days each results in9 days of sick time for the household).

Table 1. Respondent characteristics

Burkina Faso Bolivia Benin

GNI per capita (2008 US$)* $480 $1460 $700HH Size =Total Fertility Rate (2010 est) + 2 8 5 7Estimated GNI per household (2008 US$) $3941 $7402 $5180Respondent characteristicsFemale (% of respondents) 44% 100% 100%Number of total respondents 192 266 3,623Average age (in years) 39 40 31

(10) (13) (7)Number of respondents reporting age 188 266 3,579Reported annual income (US$)† $2315 $3415 $2932

($4914) ($2834) ($5089)Number of respondents reporting income 141 265 2293

Notes:Standard errors in parentheses.*Downloaded from the World Bank data collection, indicator is for GNI per capita, Atlas method (current US$).Total fertility rate is from CIA World Factbook.

†The exchange rate information is from OANDA online. The surveys collected information in the currency of thecountry. For Benin and Burkina Faso, the currency is converted from CFA francs to US dollars as of 1 December2009 at a rate of 428 CFA francs to US$1. For Bolivia, the currency is converted from Bolivianos to US$ as of1 January 2007 at an exchange rate of 8 Bolivianos to US$1. For comparisons, all costs and income throughoutthe article are given in US$. No accommodation is made for inflation.



Although the time household members were sick was substantial, perhaps more damagingto household finances were the large number of days of work missed while ill or caregiving.Households with a sick household member in Burkina Faso missed almost 4 days of work inthe prior month; in Bolivia, 26 per cent of households missed 6 days or more of work in theprior month because of illness.In general, time seeking treatment was a small but substantive part of overall time lost to

illness. The total time spent between leaving and returning home from seeking treatmentranged from a low of 2.55 hours in Benin to a high of 14 hours in Burkina Faso.

3.3 Direct Health Costs

Direct health costs are significant, particularly relative to household income (Table 3).Direct costs include medical consultations, medications and travel (transportation, lodgingand food) when household members were away from home. When using annualised healthcosts (calculated as health costs over the last 30 days * 12) and annual income, we foundvery significant spending on direct health costs as a percentage of income, ranging from 22

Table 2. Reports of illness and productivity losses


Reports of illnessIn the last 30 days, respondent or other member of household has hada health problem (% of households)

58% 17% 18%

Sought treatment outside of home for illness (% of households)* 86% 93% 60%Number of respondents reporting information on illness and treatment 111 44 640

Productivity lossesMean number of days people in HH were sick in the past 30 days 10.1 10.8 6.9

(14.2) (10.6) (7.0)Number of respondents reporting number of days sick† 111 44 599Number of days of work missed because of illness 4.0 3.3Categorical responses (9.3) (5.9)0 days 59%1–3 days 16%4–5 days 0%6 days or more 25%

Number of respondents reporting days of work missed† 111 44 640Time for health servicesTime to arrive to health provider (in minutes) 35 35 24

(44) (48) (27)Time getting drugs or lab tests (in minutes) 50 59 10

(89) (107) (30)Time spent between leaving home and getting back afterconsultation (in hours)

14 4.02 2.55(40) (8) (8)

Number of respondents reporting times relating to treatment 95 41 385

Notes:Standard deviations are in parentheses.*If anyone in the household sought treatment outside of home for illness, then she is included in this percentage.For example, if two household members were sick and only one sought treatment, this is given a value of “Yes”for whether the household sought treatment outside of the home.

†In Benin, the number of days people in household were sick in the past 30 days was only asked for childrenunderage 10. The number of days of work missed due to illness or caregiving was asked of adults and children.



per cent in Bolivia to 67 per cent in Burkina Faso. To account for the fact that health costsvary throughout the year because of seasonality, we conducted a simulation to calculate arange of annualised health costs. If health costs as reported for this 30-day period rangedfrom 25–75 per cent of this period for the other 11months of the year, then direct healthcosts as a percent of annual income would range from 10–25 per cent in Benin, 7–17per cent in Bolivia and 21–52 per cent in Burkina Faso.Catastrophic spending was significant. We define catastrophic spending as having direct

annual health costs (30-day costs for 12months) greater than 10 per cent of annual householdincome. In fact, 65 per cent of households in Burkina Faso, 46 per cent of households inBenin and 17 per cent of households in Bolivia experienced catastrophic levels of spendingon health costs. The definition of catastrophic health spending in the literature ranges from10 per cent of annual income or expenditure (Russell, 2004; McIntyre et al., 2006) to 40per cent of annual income after accounting for subsistence needs (Xu et al., 2003).The majority of health costs incurred by microfinance members and their families in each

country were for medication (Figure 1). Qualitative interviews showed that self-treatmentwas often the first line of defense against illness and frequently involved purchase anduse of medications. The consultation fees play only a small part in total direct treatmentcosts in all three countries, despite differences in their health systems and in healthcaresources (Table 4). Use of hospitals is most common in Bolivia, whereas the majority

Table 3. Direct health costs


Health costs for those seeking treatmentTotal direct health costs (US$) $82.60 $46.31 $21.71

($244.35) ($131.78) ($37.42)Consultation (US$) $1.44 $1.01 $3.22

($3.63) ($1.36) ($13.69)Medicine (US$) $71.84 $37.30 $14.05

($228.06) ($122.21) ($26.09)Transportation (US$) $3.22 $2.32 $1.37

($12.58) ($6.57) ($2.80)Lodging (US$) $3.86 $3.98 $1.17

($24.87) ($16.87) ($5.62)Food and drink while away from home (US$) $2.24 $1.70 $1.90

($8.33) ($6.37) ($5.44)Number of households with health costs information 95 41 385

Health costs as percentage of incomeTotal health costs as a percentage of income{ 67% 22% 32%Number of respondents reporting health costs and incomeinformation†

74 41 260

Notes:Standard deviations in parentheses.†Three outliers have been removed from the calculation for Burkina Faso. These three households reportedhealth costs for incidents in the last 30 days that were larger than their reported annual household income.Two of these had somewhat high expenses (slightly higher than average) but low incomes, whereas the thirdhad very high expenses. These outliers have not been removed from the health costs or income averages. Withthese outliers included, there are 77 respondents in Burkina Faso with health costs and income information,average total health costs as a % of income (30 day*12)/income is 190%, and average total health costs as a% of income (30 day/income) is 16%. [The simulation for the range of costs noted in the article also excludesthese three observations.]

{Calculated as (30 day cost * 12)/annual income.



of households seek care from an ambulatory health center in Benin and from a dispensary(i.e. a public medical center) in Burkina Faso.

3.4 Design and use of Financing Products for Health Costs

As part of the MAHP initiative, RCPB and CRECER offered three innovative healthfinancing products: health loans, health savings and health loans linked to savings. Theyoffered these products in addition to implementing other health-related activities such ashealth education, community health workers, holding regional health fairs, and formalising

Figure 1. Breakdown of health costs

Table 4. Source of health care used


Source of treatment (% of respondents)*Hospital 29% 49% 27%Health center 20% 34% 65%Dispensary 33% 0% 4%Mobile/outreach clinic 0% 2% 2%Private hospital, health center or clinic 1% 5% 3%Traditional healer — — 4%Traditional birth attendant 0% — 3%Religious healer 0% — 1%Quack 0% — 2%Pharmacy or western drug seller 0% 0% 7%Traditional drug seller 1% 2% 7%Relative or friend (outside the household) 1% 0% 0%Other provider 20% 7% 1%

Number reporting information on where sought treatment 95 41 383

*If anyone in the household reported using this source, she is included. Therefore, percentages may add to greaterthan 100%.



referrals and linkages to providers. Market research with MFI clients strongly influenced thedesign of the health programs and financing products. PADME did not offer health finan-cing, concentrating instead on initiating health education while maintaining group-basedmicroenterprise loans.CRECER offered a health loan, designed for large health expenses with a lower interest

rate and longer repayment period than microenterprise loans, which could be accessed withproof of health expenses. Often, the loans were paid by the MFI to providers on the clients’behalf, ensuring that the loan went directly to covering health expenses.In Burkina Faso, the health savings product was a voluntary savings account where

clients deposited a set minimum amount (at least $1) each month in an account dedicatedto use for health expenses. During the first 6months, the client could not access the funds;after this waiting period, clients could withdraw savings with proof of health expenses. Anactive savings account also entitles clients to apply for a health loan in cases of health costsexceeding their savings balance. These health loans have a lower interest rate and moreflexible repayment terms than business loans (Reinsch and Ruaz, 2010).As shown in Table 5, there are rather small numbers of clients who used health loans to

finance health expenses during the survey period. In Bolivia, low usage may be caused bytwo reasons: the lack of availability of health loans at the time of the baseline survey in2007 and anecdotal evidence that a second type of loan made available by CRECER, outsidethe scope of MAHP, may have been used for health expenses. However, by December 2009,about 1 per cent of eligible clients had received a health loan to access health care; backof the envelope calculations using the baseline data on treatment-seeking behavior, wefind take-up rates of approximately 6 per cent of households with health costs. Thispercentage is obtained from 26 000 eligible clients, of which 17 per cent report illnessin their household. As 93 per cent of those households reporting illness sought treatment,

Table 5. Sources of financing for health costs


Ability to pay for health servicesAble to pay for all expenses of treatment outside the home(% of respondents)*

93% 93% 83%

Number of respondents reporting information on ability to pay 95 40 357Health financingSource of funds for treatment (% of respondents)†

Business loan 6% 0% 5%Personal savings 53% 11% 1%Health savings or loan 3% 3% 0%Earnings 24% 55% 91%Other{ 9% 21% 1%Informal Loan (or loan from other MFI) 0% 11% 0%Family and friends 9% 0% 2%

Number of respondents reporting information on sources of funds 94 38 363

Notes:*This is given a value of “Yes” if the household answered that they were able to pay for all expenses of treatmentfor all household members who were reported ill during the period.

†If anyone in the household reported using this source, she is included. Therefore, percentages may add to greaterthan 100%.

{In Bolivia, three households said their health costs were covered by “SUMI”, the national health insuranceprogram for poor pregnant women and children under 5; two households mentioned that the costs were coveredby a form of old age pension.



we estimate that approximately 4111 households with an eligible client member hadhealth costs in which 256 health loans were used, resulting in a 6 per cent take-up ratefor households with health costs.In Burkina Faso, the take-up on these financing products was very low for the first

2 years but grew rapidly in the third year of the MAHP initiative because of geographicexpansion of the program, intensive staff training regarding promotion and use, and thetemporary use of marketers earning a fee when an account was opened. The take-up ofhealth savings accounts was higher among individual clients in urban areas than it was withvillage banking (i.e. group lending model) clients in rural areas (Reinsch and Ruaz, 2010).This was in part because of two factors: individual health savings and loan accountsrequired a form of personal identification, which many village banking clients do nothave; and village banks normally manage group savings and loans—requiring significantproduct adaptations to make individual products available. The 12 099 households withhealth savings accounts by December 2009 made up 3 per cent of eligible RCPB clients(Reinsch and Ruaz, 2010). Because a health savings account had to be active to qualify fora loan, there is no way to separate those who desired to save for future health expenses fromthose households who opened an account as a form of insurance to be eligible for the loan ifthe need arose. The take-up on loan products was very low (<0.1 per cent of eligible clients),but this may be because clients had earmarked savings that prevented them from needing aloan. These issues underscore the complexity of developing and implementing new financialproducts for health.Although take-up rates appear low, qualitative interview data in both Bolivia and Burkina

Faso clearly indicated that clients find health savings or loan products appealing. In inter-views in Bolivia with clients who had taken a health loan, 37 per cent of clients said thehealth loan fully covered their medical expenses; 80 per cent said the unmet costs werefor medicines. These loans were used for cosmetic dental procedures (29 per cent),surgery (27 per cent), preventive or curative dental care (17 per cent), gall bladder surgery(10 per cent), broken bones (10 per cent) and other health expenses. The majority of thoseinterviewed with health loans where the loan did not fully cover expenses covered the unmetcosts by borrowing money from a family member; the remainder used business earnings,savings at home or with other family members, using their microenterprise loan fromCRECER or taking out a loan from another institution. Without receiving a health loan,almost equal proportions of these clients said they would have borrowed money from friendsor family (27 per cent), postponed treatment (24 per cent) or sold an asset (24 per cent). Theremainder would have taken credit from another institution (15 per cent), not sought treatmentat all (12 per cent), opted for lower quality treatment (5 per cent) or other reasons.Qualitative interviews in Burkina Faso found that 85 per cent of those interviewed who

had a health savings account said that they preferred the health savings account to borrow-ing money from friends or family for medical expenses. Forty percent had withdrawn fromtheir health savings one or more times. Clients valued the sense of security in havingmoney available in the event of a health problem and also noted the benefit of privacyand confidentiality in not having to explain medical problems to families and friends whenseeking funds for health care. RCPB clients also said the health savings product reducesthe temptation to spend this earmarked money on things other than health costs. Manyindicated a health savings account allowed them to use their business loans and othersavings more effectively. Seventy-two percent of these clients with a health savingsaccount said they were more confident than in the past about their ability to pay for futuremedical expenses.



4 DISCUSSION

This study focuses on the health risks and costs that affect the poor worldwide, focusingprimarily on MFI clients in Burkina Faso, Bolivia and Benin.1 As shown by the compar-ison between estimated GNI per household2 and reported annual household income, ourrespondents were poorer than average households in each country (OANDA, 2010; TheWorld Bank, 2010). The high level of variability in direct costs is likely in part becauseof differences in health services available for treatment in each region. Direct health costsas a percentage of income for these MFI clients are considerably higher than direct costs asa percent of income found in systematic literature reviews of health costs; in these studies,the average yearly costs as a percentage of annual household income ranged from 2.5–16per cent (Russell, 2004; McIntyre et al., 2006).3 Our findings that pharmaceuticals are alarge percentage of direct health costs correspond with previous studies’ finding thatpharmaceuticals account for 30–50 per cent of total healthcare expenditures in developingcountries (Whitehead et al., 2001).Our results regarding the number of days lost when a member of the household is ill

speak to the double burden of illness in developing countries, which suffer high rates ofboth communicable and chronic diseases (Anderson, 2009). Although we were not ableto calculate financial costs of lost productivity, this second onus of lost earnings at a timewhen treatment costs are also high represents a significant financial threat to households.Qualitative interviews with clients who had access to health savings and health loan pro-ducts showed how the clients viewed and used these accounts for their health needs. With-out these options from the MFI, clients reported they would have been forced to borrowfrom other sources, sell assets and postpone or avoid treatment. This corresponds withthe literature, which suggests the poor in developing countries ‘patch together’ funds fromdifferent sources to pay for large expenses (Collins et al., 2009).Although the findings from Benin, Bolivia and Burkina Faso are similar to existing

literature in showing substantial direct health costs compromising access to necessaryhealth care, possible solutions are different. MFI clients, who are among the hundreds ofmillions of the poor worldwide, already have relationships with MFIs that represent thecapacity to provide health financing worldwide. MFIs are a potentially important sourceof health financing because of several qualities: an already established relationship withmillions globally, expertise in the administration of loans and savings, a desire to improveclients’ financial stability, and in many cases, an institutional social mission. The motivationor incentive for MFIs to expand their traditional financing products to include health mayrely on either their social mission or business self-interest in improving clients’ financialstability. One of the most frequently cited reasons for client default and dropout is health—the illness of anMFI client herself or a family member (Reinsch et al., 2010). Microenterpriseloans are already being used for the ‘non-productive’ (i.e. non-income generating) use ofhealthcare expenses; MAHP research found that 11–48 per cent of clients used a portion oftheir microenterprise loans for health expenses (Reinsch et al., 2010).We found no evidence upon which to judge the take-up rates compared with other health

financing options from MFIs or compared with an ‘optimal rate’. Sufficient data to

1Survey participants in Benin were not all PADME clients.2Calculated as the [total fertility rate nationwide (2010) + 2 (heads of household)]* per capita GNI (2008).3Terminal HIV/AIDS cases were considerably higher, particularly taking into account funeral costs, which areoften as large or larger than costs of illness.



understand the take-up rate for health financing in large scale practice are not readily avail-able from the study findings or from the published literature. A recent paper by Dupas andRobinson (2011) found very high take-up for health savings in a field experiment lookingat different forms of health savings within Rotating Savings and Credit Associations(ROSCA). For their most basic product for health savings, they had take-up of almost75 per cent after 1 year and 40 per cent after 3 years. The primary differences from ourstudy of health savings products are that ROSCAs that were not interested in the healthsavings products (13 per cent) were already excluded from the sample, the savingsmechanisms and education were provided by the researchers within the groups at the timeof the intervention and all interventions relied on the group to implement and track. Mostof the groups mentioned that they had not saved for health previously because they had notthought of it (Dupas and Robinson, 2011). This is contrary to the health savings and loansin the MAHP project, which were formal individual savings or loan accounts. Althoughthe take-up in their experiment was significantly higher, it relied on hired workers totravel to the field to promote savings; the MAHP project relied on existing humanresource infrastructure to deliver the health education and product promotion. Activereach-out and promotion of the health savings and loans was a deficiency in MAHPand is being increased in long-term scale-up of the programs.Our study suffers from several limitations, including differences in data collection surveys

and methods across countries, differences in the products offered and populations covered,the limited ability to quantify the cost of lost productivity, and the inability to speakdirectly to the impact of health loans and savings on client financial or health outcomes.The differences in data collection across countries lead to a limited ability to compare theresults, although the results are in line with other literature in the field. The variety ofhealth products offered to the populations covered by the MAHP project was a resultof careful market research evaluating client demand and assessments of the ability of theMFI to carry out specific interventions successfully. Although this leads to less comparableresearch results from the three countries, having the products be suited to the particular MFIclient base to improve sustainability was an important objective of the MAHP project. Asinterest in health financing products grows within the microfinance field, it will becomeincreasingly important to evaluate their impact on health and financial outcomes.Our results regarding health financing products demonstrate that these products are

sought and used by clients, although effective implementation and use of these productsis admittedly challenging. Careful market research, product design and impact evaluationis necessary to ensure financing options correspond well with client health needs. Forexample, loans covering only consultation costs are unlikely to have high demand or usagerates; our data show definitively that medications are a large part of health expendituresand should be considered for all health financing products. Sinha and Batjiji (2010)described the benefits of microfinance for financing health care for currently excludedpopulations, classifying strengths and weaknesses of methods including loans, savingsand health microinsurance. Categorising healthcare needs according to cost and likelihoodof occurrence helps determine which financial tool(s) provided by MFIs are optimal foreach type of health event; credit and savings may be favorable for lower cost frequentneeds such as prevention and acute illness, whereas microinsurance provides potentiallygreater protection against losses for higher cost events. However, microinsurance presentsconsiderable complexity for MFIs to design and implement (Sinha and Batjiji, 2010);therefore, understanding the role of easier to implement health financing products suchas loans and savings is important.



Health financing may work best when accompanied by a cohesive package of healthservices influencing health knowledge and behaviors and providing greater access topreventive and primary services. For example, CRECER facilitated visits by healthcareproviders to outlying client communities; at these ‘health day’ events, clients couldaccess diagnostic and primary healthcare services for a small fee. Nearly one quarter(24 per cent) of ‘health day’ participants had never been seen by a doctor before. Theevidence on preventive care from Burkina Faso showed that MAHP clients weresignificantly more likely to have received some type of preventive care than non-MAHPclients. Research shows that earlier access to care and better adherence to prescribedtreatments can result in reduction in the use of more expensive health servicesrequired after an illness is exacerbated because of non-treatment (Whitehead et al.,2001; Sabate, 2003).MFIs are rarely mentioned as potential actors in improving access to healthcare globally;

however, evidence is emerging that they can play an important role in reducing financialbarriers to medical treatment. Microfinance has been successful in providing opportu-nities to the poor where conventional financing has not been available and can poten-tially address an unmet need for health financing for the poor (Sinha and Batjiji,2010). However, offering health financing products is not restricted to MFIs; othertypes of development and commercial organisations may be motivated to offer healthfinancing and can learn from the MFI community.

FUNDING SOURCES

The Microfinance and Health Protection Initiative and all of the related research studieswere funded by the Bill and Melinda Gates Foundation. The funders played no part inthe design, collection, or analysis of the data.

ACKNOWLEDGEMENTS

Special thanks to Marcia Metcalfe, director of Microfinance and Health at Freedom fromHunger, for her comments on earlier drafts.

ETHICAL APPROVAL

The randomized controlled trial conducted with PADME in Benin was approved by theInstitutional Review Board (IRB) of Yale University. The quantitative data collection atCRECER was approved by the IRB of the University of Colorado, Denver, and theUniversidad Mayor de San Andres, in Bolivia. Data collected with RCPB was notapproved by an official review but the same ethical considerations were applied as thosesubmitted through official review: survey participants were made aware of their right torefusal to participate, their right and protection of privacy and that their participation wouldnot influence their relationship with the participating microfinance institution or othercommunity organizations. In addition, data collected with RCPB clients were collectedby the same research institute as the PADME data. No adverse events were reported or



observed throughout the course of any of the studies mentioned in this report. Dataanalysis for this manuscript was approved by the Public Health and Nursing IRB of theUniversity of North Carolina at Chapel Hill.

REFERENCES

Alcon J, Baldizon S, Copari N, Claure MM, De la Cruz J, Mendoza E, Paredes M, FleischerProano L, Rueda I, Zalles W. 2006. Market Research: Demand for Health Protection Servicesprovided by CRECER - MAHP Initiative. Credito con Educacion Rural and Freedom fromHunger: La Paz.

Anderson GF. 2009. Missing In Action: International Aid Agencies in Poor Countries To FightChronic Disease. Health Affairs 28(1): 202–205.

Collins D, Morduch J, Rutherford S, Ruthven O. 2009. Dealing with Risk. Portfolios of the Poor:How the World’s Poor Live on $2 a Day. Princeton University Press: Princeton.

Dupas P, Robinson J. 2011. Why Don’t the Poor Save More? Evidence from Health SavingsExperiments. (Working Paper). Abdul Latif Jameel Poverty Action Lab: Cambridge.

Gray B, Ekoue-Kouvahey T. 2010. Microfinance and Health Protection Initiative ResearchSummary Report: PADME. Freedom from Hunger Research Paper No. 9. Freedom fromHunger: Davis.

Kruk ME, Goldmann E, Galea S. 2009. Borrowing and Selling to Pay for Health Care in Low- andMiddle-Income Countries. Health Affairs 28(4): 1056–1066.

Leatherman S, Dunford C. 2010. Linking health to microfinance to reach the poor. Bulletin of theWorld Health Organization 88(6): 470–471.

Leatherman S, Jones Christensen L, Holtz J. 2012. Innovations and Barriers in Health Microinsurance.In Protecting the Poor: A Microinsurance Compendium, Churchill C (ed.). International LabourOrganization: Geneva.

Leive A, & Xu K. 2008. Coping with out-of-pocket health payments: empirical evidence from 15African countries. Bulletin of the World Health Organization 86(11): 849–856.

Mathers C, Boerma T, Ma Fat D, World Health Organization. 2008. The global burden of disease:2004 update. World Health Organization: Geneva.

McIntyre D, Thiede M, Dahlgren G, Whitehead M. 2006. What are the economic consequences forhouseholds of illness and of paying for health care in low- and middle-income country contexts?Social Science & Medicine 62: 858–865.

Narayan D, Patesch P (eds). 2000. Voices of the Poor: From Many Lands. World Bank: Washington.OANDA. 2010. OANDA Currency Converter.Reinsch M, Dunford C, Metcalfe M. 2010. The Business Case for Adding Health Protection to

Microfinance. Freedom from Hunger Research Paper No. 10. Freedom from Hunger: Davis.Reinsch M, Ruaz F. 2010. Costs and Benefits of Providing Health Savings and Savings Loans:

RCPB’s Experience in Burkina Faso. Freedom from Hunger Research Paper No. 10E. Freedomfrom Hunger: Davis.

Russell S. 2004. The Economic Burden of Illness for Households in Developing Countries: A Reviewof Studies Focusing on Malaria, Tuberculosis, and Human Immunodeficiency Virus/AcquiredImmunodeficiency Syndrome. The American Journal of Tropical Medicine and Hygiene 71(Suppl 2): 147–155.

Sabate E. 2003. Adherence to Long-Term Therapies: Evidence for Action. World Health Organization:Geneva.



Sinha S, Batjiji R. 2010. Protecting health: thinking small. Bulletin of the World Health Organization88: 713–715.

The World Bank. 2010. GNI per capita, Atlas method (current US$). The World Bank: Washington.Traore B, Aouba F, Ouedraogo I, Messan F, Lankoande Z, Toe C. 2006. Staying healthy means

fighting poverty: The RCPB and Health Protection Options for the Poor in the North Region.Market Research under the MAHP Initiative. Freedom from Hunger: Davis.

Whitehead M, Dahlgren G, Evans T. 2001. Equity and health sector reforms: can low-incomecountries escape the medical poverty trap? The Lancet 358: 833–836.

Xu K, Evans DB, Carrin G, Aguilar-Rivera AM, Musgrove P, Evans T. 2007. Protecting HouseholdsFrom Catastrophic Health Spending. Health Affairs 26(4): 972–983.

Xu K, Evans DB, Kawabata K, Zeramdini R, Klavus J, Murray CJL. 2003. Household catastrophichealth expenditure: a multicountry analysis. The Lancet 362: 111–117.

TECHNICAL APPENDIX

Multiple methodologies, quantitative and qualitative, were utilised to assess feasibility andimpact of offering health services among microfinance clients. The surveys were specific tothe research objectives for each country such as health knowledge and behaviors as theyrelated to the interventions; however, all surveys in Benin, Bolivia and Burkina Faso includedquestions regarding socioeconomic status, incidence of illness episodes, the costs of seekingtreatment for illness, the financing methods used for health costs and the time spent incapaci-tated by poor health because of caregiving or treatment seeking for a family member in poorhealth. Additionally, we collected qualitative data regarding the impact of poor health onhouseholds and information about their preferences for health financing options.

Sample Selection and Household Survey Implementation:

Trained fieldworkers from local research institutes in Burkina Faso, Bolivia and Beninconducted the surveys. We collected baseline and endline surveys in each of the three coun-tries. The household surveys used in this article in Burkina Faso (October–December 2009)and Benin (February, November 2009) were endline surveys, whereas the survey in Bolivia(January–February 2007) was the baseline survey. We used the baseline survey for analysisin Bolivia because full data from the endline survey were not available. We do not expecthealth costs and treatment seeking to be substantially different between baseline andendline data because in all three datasets, program and comparison areas were combinedto assess more general community-level behaviors and coping mechanisms, and it wasnot assumed that the interventions employed in any of the three countries would havea direct or significant impact on the direct costs incurred in seeking treatment.In Burkina Faso and Bolivia, surveys were conducted with microfinance clients of

Le Réseau des Caisses Populaires du Burkina (RCPB) and Crédito con Educación(CRECER), respectively. Survey participants from RCPB and CRECER were randomlyselected from a list of current clients in the program and comparison areas and were eitherinterviewed at the bank, at the village group meeting or at the client’s home. If a client wasnot present for the interview because of illness or other reasons, he or she was often replacedby a substitute. Respondents in Bolivia were women, whereas in Burkina Faso, men werealso included in the sample frame.



In Benin, surveys were conducted with randomly selected female community members invillages in the Plateau Region served by the MFI Promotion et l’Appui au Développementde Micro-Entreprises (PADME) as part of the randomised controlled trial of the intervention.One hundred twenty of the villages involved with the project were eligible for selection.Enumerators went to a central area of the village, and 30 women in each village wererandomly selected in a circular pattern from this point. Women were eligible to be inter-viewed if they were between the ages of 20 and 45, married and had children under theage of 10. Women were interviewed in their homes and did not have to be a PADME clientto be eligible.

Measures and Calculations

Different income measures were used in each country, and we adjusted survey responses toobtain an annual household income measure. In Bolivia, the survey respondent was asked toconsider all sources of income to the household and report monthly household earnings,which we multiplied by 12 to get an annual measure. In Burkina Faso, respondents wereasked to estimate their total annual household income, which we used unchanged. In Benin,respondents were asked to report sales (and profits) for a week in a normal month for eachenterprise they participated in; they were also asked what proportion of the householdincome they contributed.4 We calculated the income measure by assigning percentages toeach category of proportions that was asked (e.g. most of it, about half, etc.) and adjustingthe business profits to weekly household income, which was then multiplied by 52 to get anannual measure. Costs as a percent of annual income are calculated at the household level(i.e. total costs for household 1/household 1 income).Health costs information was collected for anyone who reported that she or a family

member was sick in the last 30 days and sought treatment from an outside source. Questionssought information about incidences of illness occurring within the 30 days prior to thesurvey. In Benin, detailed information for some measures was only collected for children.Questions about food and lodging costs while away for treatment, travel time and costs tothe provider, time for drugs and labs, and the number of days sick were only asked forchildren in the household. Therefore, reported values of these measures are likely to beunderestimates of these costs and time for the entire household. This is not expected to biasthe results significantly because of the small proportion of adults that were sick relative tochildren in Bolivia and Burkina Faso.To determine the direct impact of illness on productivity, respondents were asked how

many days of work were missed (either at home or outside the home) because of their ownillness, a parent missing work to care for a child, or another household member missing workbecause of his/her own illness. If a respondent reported that the household member was sickin the last 30 days, but no days of missed work were reported, this was assumed to be zero.

Qualitative Interviews

To gain additional information about specific groups and responses to the MAHP products,qualitative interviews were conducted. These interviews were conducted with different

4Other sources of income were also collected, but we did not include these in our income measure.



respondents from the household surveys. In Bolivia, interviews were conducted withCRECER clients (n = 89) including clients who obtained a health loan (n = 41), clientswho knew about the health loans but did not obtain one (n = 27) and clients who did notknow about and did not obtain a health loan (n = 21). In Burkina Faso, a random sampleof RCPB clients was interviewed who had health savings accounts (n = 39); additionalclients who had also utilised a health loan were interviewed (n = 8). Most clients withhealth savings accounts had had them for at least a year.The qualitative interviews were recorded with the permission of the interviewees, tran-

scribed in the native language of the respondent and then translated into English foranalysis.

Calculations at household level

To summarise data from the person level to the household level, for continuous variables,all costs or time for each member of the household were summed to get the householdtotal. For categorical variables such as source of care or type of funds used, a new variablewas created that was equal to one if anyone in the household used this source, and zerootherwise. Thus, the percentages for these variables may add to more than 100 per cent,if people in the household used different sources of care or if individuals used multiplesources of care.5 For costs information, costs are calculated for all respondents that soughtoutside treatment. For example, lodging costs were only collected for those who had tospend time away from home because of their illness and/or treatment; these costs wereassumed to be zero for households who did not report these costs. For the graphs, totalcosts are the sum of all costs reported in each household for all members, the breakdownof travel is done at the household level (i.e. it is travel for respondent 1/total householdcosts for respondent 1).

5Multiple sources of care were only reported in Benin.



health financing: a new role for microfinance institutions?

Documents