Elsevier

Journal of Health Economics

Volume 32, Issue 6, December 2013, Pages 1043-1056
Journal of Health Economics

Does health insurance improve health?: Evidence from a randomized community-based insurance rollout in rural Burkina Faso

https://doi.org/10.1016/j.jhealeco.2013.08.003Get rights and content

Highlights

  • In 2004, a community-based insurance (CBI) scheme was introduced in Nouna health district, Burkina Faso.

  • We use six years of panel data to evaluate the welfare and health impact of the insurance scheme exploring the random rollout of the program as exogenous source of variation.

  • The average reductions in health expenditure among insurance enrollees are small.

  • The insurance scheme induced a temporary increase in mortality among older community members.

Abstract

From 2004 to 2006, a community-based health insurance (CBI) scheme was rolled out in Nouna District, Burkina Faso, with the objective of improving access to health services and population health. We explore the random timing of the insurance rollout generated by the stepped wedge cluster-randomized design to evaluate the welfare and health impact of the insurance program. Our results suggest that the insurance had limited effects on average out-of-pocket expenditures in the target areas, but substantially reduced the likelihood of catastrophic health expenditure. The introduction of the insurance scheme did not have any effect on health outcomes for children and young adults, but appears to have increased mortality among individuals aged 65 and older. The negative health effects of the program appear to be primarily driven by the adverse provider incentives generated by the scheme and the resulting decline in the quality of care received by patients.

Introduction

Over the last few years, few subjects have taken as prominent a spot in national politics in both developed and developing countries as health insurance. Lack of health insurance is not only presumed to expose individuals and families to large and unexpected negative financial shocks (Bärnighausen et al., 2007, Criel et al., 2005b, Frenk et al., 2006, Gertler and Gruber, 2002, McIntyre et al., 2006, Meessen et al., 2003, World Bank, 2010), but may also provide strong incentives for not seeking professional health care, particularly in settings where access to credit is limited (Gottret and Schieber, 2006, WHO, 2000).

With limited public tax resources and often severe institutional capacity constraints, national health insurance schemes remain rare in low-income countries. In the absence of national health insurance, a variety of micro-insurance schemes have emerged over the past 15 years. Given their generally small scale and high degree of flexibility, these community-based insurance (CBI) schemes are comparatively easy to set up, and have been widely implemented in sub-Saharan Africa and South-East Asia (Asenso-Okyere et al., 1997, Atim, 1999, De Allegri et al., 2006a, Devadasan et al., 2006, Schneider and Hanson, 2007, Wang et al., 2009).

While the primary objective of CBI schemes is generally to protect individuals against large financial shocks, the increased popularity of the schemes over the past decade is also partially driven by efforts to make health care more affordable in settings where limited resources and credit constraints may impede seeking professional care in case of ill health (Kawabata et al., 2002, Robyn et al., 2012d). A large literature reviewed in Ekman (2004) has analyzed the effects of health insurance on health seeking behavior1; while most studies find insurance to increase health service utilization (Chankova et al., 2008, Franco et al., 2008, Gnawali et al., 2009, Rao et al., 2009, Robyn et al., 2012d, Schneider and Hanson, 2006, Wagstaff, 2007, Zhou et al., 2009), this does not appear to be true universally (Chankova et al., 2008, Robyn et al., 2012c, Smith and Sulzbach, 2008). Several studies on China's New Cooperative Medical Scheme (NCMS) have identified positive effects of insurance enrollment on utilization, but no or negative (increasing) effects for average out-of-pocket and catastrophic health expenditure (Wagstaff and Lindelow, 2008, Wagstaff et al., 2009).2

While several studies have investigated the impact of health insurance on health outcomes, no study to our knowledge has directly analyzed the health impact of community-based health insurance schemes in low-income settings. Beyond the influential RAND experiment which showed negative effects of increased insurance coverage on health outcomes for some of the more vulnerable subgroups, but no effects on average health outcomes (Brook et al., 1983, Manning et al., 1984, Newhouse, 2004), the majority of well-identified empirical evidence of the impact of health insurance on health outcomes is based on the Seguro Popular program in Mexico. Two separate studies have investigated the health impact of the PROGRESA program (Barros, 2009, King et al., 2009), both finding no impact of the insurance scheme on health.

In this paper, we analyze a comprehensive community-based insurance scheme, which was rolled out using stepped wedge cluster randomization in Nouna District, Burkina Faso, between 2004 and 2006. In an effort to make health care more affordable and protect local communities from large health expenditure shocks, the Ministry of Health and Nouna Health Research Center (Centre de Recherche en Santé de NounaCRSN) decided to implement a community based insurance scheme in cooperation with the University of Heidelberg in Nouna health district. In order to allow for a proper evaluation, the rollout of the program followed a stepped wedge cluster-randomized design, enrolling randomly selected communities in three phases between 2004 and 2006. As discussed in further detail in Hussey and Hughes (2007), the main idea of the stepped wedge design is to embed randomizations into existing program rollout or scale-up efforts, sequentially adding a randomly selected number of communities to a given program. While this design comes at the cost of a longer time frame needed for evaluations, the main advantages are that the program rollout can be done more slowly, and that interventions which are expected to benefit the treated group can eventually be provided to all subjects or areas (Mdege et al., 2011).

The Nouna CBI program is of notice for at least three reasons: first, because the low levels of income and literacy and high levels of mortality in the area make the underlying population representative of a large number of rural communities in the subcontinent; second, because a comprehensive epidemiological and demographic data collection system has been set up in the area; and third, because the sequence of the rollout was randomized across communities to allow for a proper empirical evaluation.

Exploring the exogenous temporal variations in insurance availability generated by the randomized rollout of the program, we use six rounds of household survey data collected in the study area between 2003 and 2008 to estimate the causal effect of insurance on treatment-seeking behavior and expenditure using both reduced form and two-stage-least squares estimation in a first step. In a second step, we use the vital registration data collected at the population level through the Health and Demographic Surveillance System (HDSS)3 to evaluate the aggregate health impact of the insurance scheme.

Overall, the CBI program appears to have yielded mixed results: on average, insurance enrollment appears to have reduced the likelihood of catastrophic health expenditure, and was also associated with small but not statistically significant reductions in average out-of-pocket expenditure. On the other hand, the rollout of the program appears to have led to short-term increases in mortality among individuals aged 65 and older, while it did not affect health for children and working-age adults.

Based on the quantitative and qualitative evidence collected as part of the insurance rollout, two main factors appear to be the main mechanisms underlying the limited uptake and lacking health impact of the scheme: a lack of understanding of the benefit package by the targeted population and limited attractiveness of the level of coverage offered (Robyn et al., 2011a), and negative incentives for health facility staff generated by the CBI provider payment system (Robyn et al., 2012b).

While the insurance covered a large number of health services in principle, actual coverage appears to have been poorly communicated, so that insured households often did not benefit from the scheme even if sick. Most of the negative effects appear to have been triggered by unintended negative incentives for health centers and health center staff generated by the specific capitation payment mechanism chosen. Based on the notion that per patient lump sum transfers will incentivize health facilities to engage in prevention, improve efficiency of service delivery and minimize the cost of treatment, the use of fixed payment reimbursement schemes (salary and capitation) has been promoted as the most effective means of containing treatment costs (Roberts et al., 2008). In the case of the Nouna CBI scheme, these incentives appear to have had strong negative repercussions: by paying health centers a flat per capita payment for each person enrolled in their catchment area rather than service- or delivery-based payments, the insurance scheme removed the out-of-pocket service fees previously captured by health facilities. These service fees were not only used to finance day-do-day facility operations and maintenance, but also constituted direct performance incentives for health workers, who under the traditional payment system received quarterly bonuses (known as ristournes) based on their facilities' total service fee revenue (approximately 20–22% of total service fees were distributed among workers as financial bonuses). The lack of financial incentives for providers to treat insured patients at the facility level resulted in rather poor services received by insured patients, and may have prevented individuals to seek care rather than encouraging it (Robyn et al., 2011a, Robyn et al., 2012c).

The results presented in this paper further highlight the difficulties involved with the optimal design of insurance schemes in developing countries. The health services research literature has devoted much attention to the difficulty in adjusting capitation payment adequately to reflect the diversity in disease severity among patients and to mitigate facilities’ incentives to select or dump patients (Giacomini et al., 1995, Newhouse, 1994, Robinson, 2001). Further research will be needed to determine incentive structures that induce improved health system performance in developing country settings.

The remainder of this paper is organized as follows: in Section 2 we provide background information on the Nouna CBI scheme and study population; in Section 3 we introduce the household survey data, and present the main household level results. In Section 4, we introduce the vital registration data and show population-level results. We conclude with a short summary and conclusion in Section 5.

Section snippets

Study setting

Nouna health district is located in northwestern Burkina Faso. The area is predominantly rural, with the majority of the population engaged in small-scale farming (Sauerborn et al., 1996a, Sauerborn et al., 1996b). Primary subsistence crops include corn, millet and sorghum, while the main cash crops are sesame, cotton, groundnuts and fonio. More than 80% of the population within the Nouna HDSS is illiterate (Souarès et al., 2010). Several ethnic groups are present in the district, including the

The HDSS Household Survey

The Nouna HDSS Household Surveys were designed as a household panel study, first implemented in 2000, and restructured in 2003 to more closely fit the monitoring needs around the CBI launch in 2004 (De Allegri et al., 2008). Households were randomly selected from the population in the Nouna HDSS database. In this paper, we use survey data from 2003 to 2008, which offers one year of baseline data prior to the intervention, and two years of data after the final rollout phase within the HDSS area.

The HDSS data

As described in Section 2, the Nouna Health and Demographic Surveillance Site (HDSS) has been collecting complete vital registration data on a six-month basis for the entire study area. Table 9 shows the aggregate population and mortality rates for the 41 villages as well as the 7 Nouna sectors over the period 2003–2008. Due to continued high fertility rates11 and declining mortality rates, total population in the area

Summary and discussion

In this paper, we have used the randomized rollout of a community-based insurance scheme in Burkina Faso to identify the causal effect of health insurance on population level mortality as well as household level health and welfare outcomes. While we find that the insurance did have some positive effects on health expenditure in general, and catastrophic health expenditure in particular, we find no health improvements for children and working age results, and, more surprisingly, negative health

Acknowledgments

The authors would like to thank in particular Mr. Cheik Bagagnan and Mr. Alphonse Zakané for their data management support, as well as the entire team at Nouna Health Research Center for their support in the field. This work was supported by the German Research Foundation (DFG) collaborative research grant SFB 544 ‘Control of tropical infectious diseases’, Project D2.

References (72)

  • D.P. Gnawali et al.

    The effect of community-based health insurance on the utilization of modern health care services: evidence from Burkina Faso

    Health Policy

    (2009)
  • M.A. Hussey et al.

    Design and analysis of stepped wedge cluster randomized trials

    Contemporary Clinical Trials

    (2007)
  • G. King et al.

    Public policy for the poor? A randomised assessment of the Mexican universal health insurance programme

    Lancet

    (2009)
  • D. McIntyre et al.

    What are the economic consequences for households of illness and of paying for health care in low- and middle-income country contexts?

    Social Science & Medicine

    (2006)
  • N.D. Mdege et al.

    Systematic review of stepped wedge cluster randomized trials shows that design is particularly used to evaluate interventions during routine implementation

    Journal of Clinical Epidemiology

    (2011)
  • P.J. Robyn et al.

    Health insurance and health-seeking behavior: evidence from a randomized community-based insurance rollout in rural Burkina Faso

    Social Science & Medicine

    (2012)
  • R. Sauerborn et al.

    Household strategies to cope with the economic costs of illness

    Social Science & Medicine

    (1996)
  • R. Sauerborn et al.

    Seasonal variations of household costs of illness in Burkina Faso

    Social Science & Medicine

    (1996)
  • K.V. Smith et al.

    Community-based health insurance and access to maternal health services: evidence from three West African countries

    Social Science & Medicine

    (2008)
  • A. Wagstaff et al.

    Can insurance increase financial risk? The curious case of health insurance in China

    Journal of Health Economics

    (2008)
  • A. Wagstaff et al.

    Extending health insurance to the rural population: an impact evaluation of China's new cooperative medical scheme

    Journal of Health Economics

    (2009)
  • H. Alderman et al.

    Do the poor insure? A synthesis of the literature on risk sharing institutions in developing countries

  • T. Bärnighausen et al.

    Willingness to pay for social health insurance among informal sector workers in Wuhan, China: a contingent valuation study

    BMC Health Services Research

    (2007)
  • R. Barros

    Wealthier But Not Much Healthier: Effects of a Health Insurance Program for the Poor in Mexico. SIEPR Discussion Paper 2009; 09-002

    (2009)
  • R. Basaza et al.

    Low enrollment in Ugandan Community Health Insurance schemes: underlying causes and policy implications

    BMC Health Services Research

    (2007)
  • H. Becher et al.

    Patterns of malaria, cause-specific and all cause mortality in a malaria endemic area of West Africa

    American Journal of Tropical Medicine and Hygiene

    (2008)
  • R. Brook et al.

    The effect of coinsurance on the health of adults. Results from the RAND Health Insurance Experiment

    New England Journal of Medicine

    (1983)
  • S. Chankova et al.

    Impact of mutual health organizations: evidence from West Africa

    Health Policy and Planning

    (2008)
  • B. Criel et al.

    Difficulties in partnerships between health professionals and Mutual Health Organisations: the case of Maliando in Guinea-Conakry

    Tropical Medicine & International Health

    (2005)
  • B. Criel et al.

    Cryopreservation and abiotic stress tolerance in potato: a proteomic approach

    Communications in Agricultural and Applied Biological Sciences

    (2005)
  • M. De Allegri et al.

    Understanding enrolment in community health insurance in sub-Saharan Africa: population-based case-control study in rural Burkina Faso

    Bulletin of the World Health Organization

    (2006)
  • M. De Allegri et al.

    Step-wedge cluster-randomised community-based trials: an application to the study of the impact of community health insurance

    Health Research Policy and Systems

    (2008)
  • B. Ekman

    Community-based health insurance in low-income countries: a systematic review of the evidence

    Health Policy and Planning

    (2004)
  • M. Fafchamps

    Risk Sharing and Quasi-credit

    Journal of International Trade and Economic Development

    (1999)
  • D. Filmer et al.

    Estimating wealth effects without expenditure data – or tears: an application to educational enrollments in states of India

    Demography

    (2001)
  • L.M. Franco et al.

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

    Bulletin of the World Health Organization

    (2008)
  • Cited by (63)

    • The dynamics of poverty targeting

      2023, Journal of Development Economics
    View all citing articles on Scopus
    View full text