Bulletin of the World Health Organization

Effects of mutual health organizations on use of priority health-care services in urban and rural Mali: a case–control study

Lynne Miller Franco a, François Pathé Diop b, Clara R Burgert c, Allison Gamble Kelley d, Marty Makinen a, Cheick Hamed Tidiane Simpara e

Introduction

In most African countries, including the low-income, landlocked Sahelian nation of Mali, poor and rural populations have low utilization and coverage rates for key preventive and primary curative interventions. Because of their poverty, these populations tend to suffer more health problems; because of their health problems, they tend to be poorer.1 There are many reasons for low utilization of priority health services in Africa, including poor physical and financial access to care, socioeconomic factors, cultural factors and perceptions about the quality of care.2–9

In western Africa, mutual health organizations (MHO) have sprung up with amazing speed.10,11 MHOs are voluntary organizations that provide health insurance services to their members and they are usually owned, designed and managed by the communities they serve. Member households pay an enrolment fee and then regular premiums to cover a membership-defined benefits package. After a waiting period, the MHOs reimburse providers of care for the services used by beneficiaries in the member households, with users making a small co-payment. MHOs are not for profit and are based on ethical principles of mutual aid and social solidarity.10 The rise in popularity of MHOs reflects a need in communities to address the difficulty of paying for health care when care is required. The government of Mali recognized the potential of MHOs in its 1997 10-year health and social sector development plan.12

Promoted as a solution to many health system problems, MHOs can provide additional avenues of resource mobilization and financial protection against devastating health-care expenditures, increase financial access to care, promote equity through risk-pooling as well as strengthen community solidarity and demand for quality care.13–15 At a minimum, MHOs should contribute to increased use of effective and needed health services16 and serve as a proxy for improved health.

Although there is enthusiasm and consensus on the worthiness of the principles and concepts behind the MHO movement,17 concerns exist about their ability to meet all expectations. There is still little robust evidence of MHO cost-effectiveness, of their ability to cover significant portions of the population, or of the sustainability or effectiveness in increasing access to care and financial protection. Recent broad reviews of the MHO literature16,18 found few studies that measure the effects of MHOs on health care utilization and even fewer that used econometric regression analysis.18

What is known from the few studies that have rigorously investigated the effects of MHOs is that: (i) there is an ever-growing demand for such financial protection mechanisms; (ii) MHOs seem able to enrol individuals from a variety of socioeconomic strata, although perhaps not the very poor;19–21 (iii) members tend to have lower out-of-pocket expenditures than do non-members;19,20,22 and (iv) members tend to use health services more when needed than do people not enrolled in MHOs.19,23,24 The literature also highlights that MHOs require technical support to attain functionality, that they still tend to be small, and that they will be only one of many mechanisms for financing the health sector.13,16,17

In this paper we examine the effects of a community-based MHO intervention on the use of curative, maternal and child health inventions; inclusiveness of MHO membership, and MHOs’ ability to provide financial protection in a rural and urban setting in Mali.

Methods

Setting

Four MHOs were developed by the Ministry of Health of Mali and the USAID-funded Partners for Health Reform project25 as part of a pilot programme to improve financial access to health services. A steering committee chaired by the Mali Ministry of Health selected four MHO pilot sites for the study: two in the rural district of Bla (BlaVille and Kemeni) and two in the urban commune of Sikasso (Wayerma and Bougoula). These sites were selected for their representativeness of the socioeconomic conditions faced by a large portion of Mali’s population.

The USAID-funded Partners for Health Reform and Partners for Health Reformplus projects provided funding and technical assistance for MHO development and evaluation design. To ensure the sustainability of the organizations, no direct financial support was provided for the ongoing operation of the MHOs. At the start of the MHO intervention, a baseline household survey revealed low levels of coverage for antenatal care (57%); assisted deliveries (26%); child immunizations (29%); and treatment of child diarrhoea with oral rehydration therapy (30%). Utilization of curative services ranged from 0.24–0.30 visits per person per year.26,27 In Bla district, roads are few and there is no ambulance service.

MHO intervention and study design

Table 1 presents descriptive information for the four pilot MHOs. Member households paid a once-off enrolment fee and a monthly or annual premium (based on the number of beneficiaries). On joining, members committed to make use of preventive services, such as immunizations, prenatal care and insecticide-treated mosquito nets. The MHOs signed agreements with local primary health-care centres and referral health centres (where available). When members or their beneficiaries needed curative or maternal care and were up to date on their premium payments, they paid a portion of charges (usually 20–25%) at the time of service, and the MHO covered the larger remaining portion.

Using a case–control design, we sought to answer three major research questions:

The intervention (case) group consisted of households joining one of the four MHOs. Controls fell into two categories: those who were living in areas where there was a functioning MHO but who did not join, and those who lived in areas where there was no MHO.

To evaluate the impact of the MHO intervention, we collected data from two sources: a household survey conducted in September and October 2004 and a review of MHO registers. The household survey collected household and individual data through interviews with the head of the household on socioeconomic variables, self-reported distance to the nearest health facility, utilization of priority health services, reasons for non-utilization and MHO membership.25 Questionnaires were pilot tested in an area outside the study sites. MHO registers provided data on membership, premium payment and services covered from January 2003 (when the MHOs became operational) to October 2004. Both sources used the same set of unique household identifiers, allowing the data sets to be linked.

Household survey sampling was conducted separately for members and non-members. All MHO member households in BlaVille, Kemeni and Bougoula study sites were included in the sample. However, because the MHO in Wayerma was much larger than those at the other three sites, we randomly selected 350 households from this site to allow statistically any existing significant differences to be detected among three groups: (i) members joining before April 2004, (ii) members joining after April 2004 and paying premiums for September 2004, and (iii) members joining after April 2004 but not up to date on premium payments. MHO registers provided a full list of member households.

Sampling of non-member households was based on a random selection of enumeration areas (census-defined population clusters), an updated mapping of all households in the selected enumeration areas, and systematic selection of individual households based on a random number.

Because rural Mali’s economy is largely non-cash and most household production is consumed, household wealth was measured by an approximation of consumption.28 Information on consumption was systematically collected from each household and included data on food (purchased and self-grown), transportation, lodging, utilities (water, electricity, combustibles, etc.), school fees, health, and clothing. All estimations were annualized and summarized for the household, and then converted to an adjusted overall per capita figure by dividing the total value of household consumption by the number of members of the household, weighting adults (aged > 14 years) at a value of 1 and children (aged ≤ 14 years) at a value of 0.75. The mean per capita income for all sampled households was US dollars (US$) 358 (US$ 231 in rural Bla district and US$ 510 in urban Sikasso), well within the range of The World Bank’s 2004 estimate of US$ 390. All consumption rates were converted into US dollars at the October 2004 rate of 527 CFA francs to 1 US$. Five equally-sized socioeconomic quintiles were developed, based on the adjusted per capita consumption figures: poor, middle poor, middle, middle rich and rich quintiles.

Households may pay the enrolment fee but later fail to make premium payments, causing their MHO coverage to lapse. Thus, additional groupings were used in the analyses: active household membership in an MHO – households having paid premiums at least once in the 6 months before the survey; and eligibility for MHO coverage – being registered as a beneficiary in an MHO household that paid premiums in the month(s) when services were used.

All survey instruments and confidentiality and data security protocols were reviewed by Abt Associates Inc.’s institutional review board and the Mali Study Steering Committee.

Statistical analyses

Data entry was conducted with use of MSACCESS data entry screens (Microsoft, Redmond, WA, United States of America). Data manipulation and analysis were performed with Intercooled STATA 8.0 (StataCorp. LP, College Station, TX, USA). Household data were weighted by the inverse of the probability of selection at the household level, and weights were incorporated into all subsequent analyses. Non-MHO households were weighted based on the probability of the enumeration section being selected and of a household being selected in that enumeration area. The base sampling weight for MHO households was 1.0 but was adjusted for non-response, and in Wayerma it was also adjusted for sampling.

Multivariate statistical analysis used STATA’s survey logit regression function to ascertain whether being an MHO beneficiary was a predictor of higher rates of health service utilization using the following formula:

Model: ln[Prob(individual used care) / Prob(individual did not use care)] = α1 + βX.

The formula for establishing MHO household and individual enrolment determinants was:

Model: ln[Prob(being enrolled) / Prob(not being enrolled)] = α1 + βX

if living where there was an MHO.

We used a multivariate linear regression to examine whether MHO membership translated into lower out-of-pocket payments for health services, both at the household and the individual level, using the formula

Model: ln[Y + 1] = α + βX

Results

Table 1 shows MHO packages from the four study sites with a comparison of fees, benefits, membership and coverage. Table 2 presents sample sizes for all groups and for priority-health-service target populations and Table 3 summarizes the characteristics of the households surveyed in the sample.

Utilization of priority health services

Appendix A (available at: http://www.urc-chs.com/mali-article.html) presents the results of logit regression on the utilization of modern paying health services: fever treatment (all ages), diarrhoea treatment in children younger than 5 years, prenatal care and delivery in a modern health facility. The regressions control for individual, household and community characteristics. Results (significant at the P<0.10 or better) indicate that, compared with non-members and lapsed members, up-to-date MHO members were 1.7 times more likely to seek treatment for fever in a modern facility; three times more likely to seek modern and/or oral rehydration therapy for diarrhoea in their children under 5 years; and twice as likely to make at least four prenatal visits during pregnancy.

Among control variables, distance to the health facility was a significant negative predictor for health-care seeking: those living more than 2 km away were half as likely to seek fever treatment and two-thirds to four-fifths less likely to deliver in a modern facility than were people who lived within 2 km of a health facility; those living 6–10 km from a health facility were two-thirds less likely to complete at least four prenatal visits. The diarrhoea treatment variable includes home treatment with oral rehydration therapy, which may explain why distance here was not a significant predictor. Household wealth quintiles did not show any consistent pattern of influence on use of services.

Appendix A also shows the results of logit regression on utilization of preventive services provided free of charge by health facilities (diphtheria-tetanus-pertussis 3 immunization before the first birthday among children aged 12–23 months; vitamin A supplementation in children 6–59 months as reported on a card or by a caretaker if no card available) and use of insecticide-treated mosquito nets (which are promoted but not subsidized by the MHO) by children younger than 5 years and by pregnant women. Although MHO membership did not appear to influence the use of child vaccinations or vitamin A supplementation, it was a significant predictor of treated-mosquito-net use in both children and women during pregnancy. Having access to an MHO was a significant predictor for treated-mosquito-net use in pregnant women, but not in children. Again, household wealth quintiles showed no consistent association with the use of insecticide-treated mosquito nets.

Inclusion of the poor and key target populations

Appendix B (available at: http://www.urc-chs.com/mali-article.html) presents the results of logit regression on overall household and individual enrolment in an MHO. While enrolment for all categories (household, individual, women 15–49 years and children under 5 years) was significantly higher in the rich household wealth quintile, enrolment rates did not differ between the poor, middle poor, middle or middle rich households. A key predictor of enrolment for all categories was distance to a health facility, except for children under 5 years. Household size had a significantly positive association with enrolment across all categories, as did education levels of the household head and female/caretaker. Households headed by a female were five times more likely to be enrolled in an MHO; four times more likely to enrol women of reproductive age; and eight times as likely to enrol children.

Ethnicity was also associated with enrolment: the majority ethnic group (Bambara) was significantly less likely to enrol across all categories. Finally, some adverse selection appears to be present: households with a household head who reported being in less than excellent health and households with chronically ill and/or handicapped individuals were more likely to enrol.

MHOs, financial protection and affordability

Appendix C (available at: http://www.urc-chs.com/mali-article.html) presents the results of linear regressions on overall household health expenditures, annual household health-care expenditures as a percentage of total household cash consumption, and out-of-pocket expenditures for fever treatments. Being an active MHO member was associated with lower household health expenditures as a percentage of overall cash consumption and lower out-of-pocket payments for fever treatments. Positive predictors for all household health expenditure measures included a high education level for the household head and higher household wealth quintiles. Health expenditure tended to be lower in urban areas than in rural ones; data from the study do not provide any explanation for this finding, but it may be due to competition and a wide choice of options in urban areas.

Table 4 presents two other financial protection measures for active MHO members and the overall population. The ratio of mean-to-median expenditures expresses a measure of financial risk: when the ratio is high, some households in the group are spending considerably higher amounts than others. Whether examining expenditures as an absolute value or relative to cash consumption, MHO members spend more, but have less financial risk, as their mean-to-median ratios are lower, especially in Bla district (BlaVille and Kemeni) with its largely rural population.

Table 5 presents estimates of annual household MHO expenditure, including premiums and co-payments for care, based on MHO register data. At US$ 29 to US$ 54 per household, estimated MHO spending is 1.7% to 3.0% of annual income at Mali’s poverty line (US$ 295 per capita or US$ 1765 per household29). Examining these estimated household MHO expenses in light of cash income shows that even if MHO households enrolled all their members (which many currently do not), MHO-related spending would come to between 2% and 8% of cash income, and this expanded MHO spending still falls between mean and median household cash expenditures on health as a percentage of total cash income (for the whole study population – MHO members and non-members).

Discussion

These four Malian MHOs sought to rearrange community financing provisions, building on community-based organizations to mitigate barriers associated with Bamako Initiative resource mobilization strategies. Further, they aimed to improve access to health-care services while protecting the income of the poor and strengthening their voice in the health sector. While confirming the effects of traditional determinants of health-care utilization (illness severity, education, income, and distance), our results support evidence that MHOs improve utilization, even for the poor, and help households to better manage their health-care expenditures. The results of this study corroborate findings from other MHO studies in Ghana,30,31 India,21 Rwanda,20,23 Senegal19,30,31 and Viet Nam.22

Our results show that MHOs have a positive effect on the utilization of many priority services. Up-to-date MHO members and beneficiaries, compared with controls, were 1.7 times more likely to have their fever treated in a modern health facility; three times more likely to use oral rehydration salts or seek modern care for their children under 5 years with diarrhoea; and twice as likely to make at least four prenatal visits during pregnancy. Sleeping under an insecticide-treated mosquito net was also twice as likely during pregnancy and in children under 5 years of age in the up-to-date group.

However, distance to health facilities remains a significant negative predictor of utilization of treatment for fever, prenatal services and assisted-delivery care, indicating that even 2 km can represent a geographic barrier to the seeking of health care. Geographic barriers related to preventive services for children, such as immunization, vitamin A supplementation and insecticide-treated bednets, seem to have been overcome, probably due to outreach activities. The distance barrier was especially strong for assisted deliveries, suggesting that the inclusion in the MHO package of transportation to health-care facilities for women about to give birth might be beneficial.

MHOs reached most parts of the population, and even though higher-income groups are more likely to enrol, MHOs do not exclude the poor. Analysis by household wealth quintiles showed that only membership of the richest quintile was a significant predictor of enrolment for households, individuals, and women of reproductive age, but no discrimination was seen among the other quintiles. Approximately half of the population in Sikasso commune and about 80% in Bla district fall below the poverty line, and MHO membership is drawn from a broad cross-section in both areas. While the very poor may have difficulty enrolling and paying premiums, they join as frequently as those in other quintiles, with the exception of the richest quintile. The outlay for 1 year of premiums plus co-payments for an entire household would average US$ 29–54 per year and represent approximately 2–3% of annual household income at the poverty line in Mali, and 2–8% of household cash consumption of MHO households. MHO membership reduced the variability of health-care spending and saved households money on care for fevers, although there was no reduction or savings for active members in terms of overall health spending.

In developing countries where health insurance coverage is generally limited to formal sector employees in urban areas, MHOs are a promising mechanism for reaching households in the rural and informal sector. This study has provided evidence of MHOs’ positive effects on the utilization of many priority health services, on reaching many poor people, and on providing some income protection, even though MHOs may not achieve complete coverage of the poorest of all. Our results also demonstrate the need to address not only financial but also geographical barriers to care. Since the proportion of those eligible who joined MHOs in the study areas was well below 100%, efforts are needed both to expand coverage with MHOs and find alternative methods to improve financial access to health care.

Further research may be needed to validate our findings in other settings and to evaluate strategies to increase access for the poorest. In particular, results related to equity in MHO membership and the specific effects on service use should be confirmed in other settings. MHOs remain one viable mechanism, among others, to increase financial access to – and equity in – the utilization of essential health services. However, a more concerted effort from governments is needed to develop coherent strategies for MHO development, to develop and sustain MHO support capacities through effective partnerships, and to continuously learn from the experiences of other MHOs with respect to strengthening these organizations and their ability to reach the key target populations of women, children and the poor. ■

Funding: This publication was made possible through support provided by the Office of Health, Infectious Diseases and Nutrition, Global Health Bureau, US Agency for International Development, under the terms of Project No. 936-5974, Contract No. HRN-C-00-95-00024.

Competing interests: None declared.

References

Affiliations

  • University Research Co., LLC, Bethesda, MD, United States of America (USA).
  • Abt Associates, Bethesda, MD, USA.
  • Emory University, Atlanta, GA, USA.
  • Independent Consultant, Geneva, Switzerland.
  • Results for Development Institute, Washington, DC, USA.
Share
Email Twitter Facebook Google Delicious LinkedIn
More...