OFFICES IN Washington D.C. & New Delhi

Center For Disease Dynamics, Economics & Policy

Guest Blog: The Indian Health Insurance Experiment

Guest Blog: The Indian Health Insurance Experiment

By Anup Malani and Phoebe Holtzman

This post will discuss the US and Indian experiences with healthcare financing reform and describe the Indian Health Insurance Experiment, a study that an international team of researchers, led by Ramanan Laxminarayan (CDDEP) and Anup Malani (U. Chicago) and including Cynthia Kinnan (Northwestern), Alessandra Voena (U. Chicago), Anuj Shah (U. Chicago), Gabriella Conti (UCL) and Kosuke Imai (Princeton) has just begun with support from the Department for International Development and the University of Chicago.

Anup likes to ask students and scholars of the American health care system: What study of the US healthcare system would you design if you were transported back to 1965 or 1970, soon after the adoption of Medicare and Medicaid? Usually their eyes light up when they see the potential to pre-emptively correct problems in the modern US health care system. 

Well, this is the position in which Indian policymakers currently find themselves. India has just adopted its first national public insurance program. The program hasn t yet been fully rolled out and the country is planning an uncharted future. The Indian Health Insurance Experiment aims to produce reliable evidence on the effects of different health insurance reforms on the health care system to aid policymakers in implementing the best possible program.

In 2010, the US adopted a bold expansion of its public health insurance system, the Affordable Care Act (ACA). The ACA committed roughly $1.4 trillion over 10 years in equal parts to expand Medicaid and to provide tax credits to reduce health insurance premium payments for households under 400% of the federal poverty line (roughly $90,000 for a family of 4). This is expected to provide insurance to 20-30 million who are presently uninsured.

Although policy research expenditures lag far behind actual health expenditures, the US research community has significantly invested in rigorously researching the impacts of the ACA. Most notable is the Oregon Health Insurance Experiment. Prior to the ACA, Oregon sought to expand its Medicaid program but lacked funds to provide insurance for everyone who fit its new eligibility criteria. So it provided Medicaid to a random subset of the newly eligible population. Health economists such as Kate Baicker and Amy Finkelstein, among others, used this truly natural experiment to study how expanding Medicaid coverage impacts utilization, health and financial status. Although the results are sobering (additional utilization, especially of emergency departments, insignificant increases in objective health outcomes, reduced out of pocket expenditures), the study serves as a role model for how research should inform policy. It carries on the legacy that began with the RAND Health Insurance Experiment in the 1980s, and continued with, e.g., the Seguro Popular experiment in Mexico in the early 2000s.

Just two years before the US adopted the ACA, India adopted its first national public health insurance scheme, Rashtriya Swasthya Bima Yojna (RSBY). India clearly has a far less developed health care and financing system country than the US. Whereas roughly 80% of the US was insured prior to the ACA, less than 11% were insured prior to RSBY; and while the US has an out-of-pocket payment rate of 18%, more than 75% of medical expenses are financed out-of-pocket in India. Unsurprisingly, nearly 63 million people are forced into poverty each year because of medical expenses. In response, the central government adopted a scheme that would provide hospital insurance to individuals with Below Poverty Line cards, roughly the bottom quartile of the population (in terms of asset distribution) nearly 300 million people! That is an order of magnitude larger than the ACA health insurance expansion. In five years, the program has already provided coverage to 150 million lives, nearly 50% more than the number of lives covered by Medicaid and Medicare combined. This is partly because RSBY coverage is sparser than Medicaid coverage, hospital utilization is lower in India, and health care prices are lower in India. The cost of the existing RSBY expansion is on the order of $25 billion over 10 years nearly an order of magnitude less than the cost of the ACA expansion. Even if the coverage and thus costs were to expand fourfold (e.g., including outpatient care) RSBY would still be less than one-tenth the price of the ACA.

Just as the ACA demands rigorous evidence on its effects, so too does RSBY. In the case of RSBY, this evidence would influence the durability of RSBY and its future expansion. Like the ACA, RSBY has been criticized based on implementation and costs. Yet policymakers are also debating whether and how to expand RSBY eligibility to the middle-income populations and coverage to include outpatient care and medicines. A solid base of evidence on the program could help resolve these debates and point to a way forward as India considers the government s role in health care financing.

To this end, in 2012, we reached an agreement with RSBY and the state of Karnataka to use RSBY insurance as a vehicle to study the benefits of public health insurance and how different types of coverage expansions may affect insurance uptake. We also obtained support from DFID and the University of Chicago to financially support the study. We have assembled an Indian-led team with international collaboration. The study began in March 2013 with the help of the Centre for Microfinance, housed at the Institute for Financial Management and Research in Chennai. Baseline is nearly complete and we will soon begin treatment assignment and insurance enrollment. Like RAND, the Seguro Popular study and the Oregon study, our study will be randomized, involves large populations, and looks at multiple outcomes.

Our first study is a social experiment that aims to estimate the benefits of different policies that expand eligibility for RSBY. In broad strokes, the study enrolls roughly 12,000 households to receive RSBY insurance with different levels of support (free insurance or premium subsidies). It follows them for 2.5 years, observing their health and financial outcomes. The study also conducts a number of corollary surveys to tease out demand for insurance, capture spillovers from insurance, and analyze uptake of insurance and utilization of care.

The sample. The sample for the study is roughly 12,000 households, including nearly 60,000 covered lives, in two districts of Karnataka: Gulbarga and Mysore. The sample size was selected to ensure we had adequate power to estimate the effect of RSBY on hospital utilization, even after 10% attrition and assuming we did not match households before randomly assigning them to treatment arms. In addition, the sample size was fixed in order to estimate different treatment effects each year (or different treatment effects across the two districts).

 Although the two sample districts are in the same state, they have different populations and cultures. The two districts are geographically diverse. Gulbarga is in the center of the country and Mysore is in the far south, giving the study external validity for two regions of the country. These districts are not the richest in Karnataka, which is one of the more developed states of India, and so they give the study external validity for other less well developed states.

 The study population includes individuals who are not currently eligible for RSBY or other public insurance programs, i.e., individuals with above poverty line (APL) rather than below poverty line (BPL) ration cards. One reason is that we need a treatment na ve population; individuals already eligible for RSBY may get RSBY at a minimal price even if they are allocated to a control group. Enrolling such populations would understate the impact of insurance given to the treatment group or compromise the power of the study. A limitation is that the study may not tell us as much as we would like about the impact of the existing RSBY eligibility rules. The limitation will be partly mitigated because we track the income and assets of sample households and can project the impact of insurance as income or assets fall. Moreover, our anecdotal experience is that many individuals who had APL cards just months ago have received BPL cards, often in advance of state elections.

Treatment arms. Sample households are randomized into groups that mimic different policy options. One arm gets free RSBY insurance, which serves as a proxy for a Medicaid-like scheme. Another gets a cash transfer and the opportunity to buy RSBY, which is a proxy for a premium support program.

Randomization. Households are randomized using a matched randomization algorithm. First, households are only matched if they are in the same village (exact match on village). Second, they are matched to their nearest neighbors on other variables that proxy for income and health (based on data from a census exercise we conducted prior to the start of the study, akin to a pre-baseline baseline). Third, once households are grouped into sets in this manner, households within sets are randomly allocated to different arms. This matching allows us to increase the power of our study.

Outcomes. Our primary outcomes are financial status and health. To see why both are important, think of hospitalization as a binary decision (you get hospitalized or not), and group all sample households into two groups. One group includes households which, had they fallen ill prior to receiving RSBY, would not have sought hospital care. For these households, RSBY increases utilization because it reduces the price of care at the moment of consumption. This in turn may improve health. A second group includes households which, had they fallen ill prior to receiving RSBY, would have sought care anyway. For these households, RSBY reduces the price of care without affecting utilization. This will manifest as an income effect. Since care is a roughly continuous variable, we could see both effects income effects and substitution effects on the margin. But that just highlights the need to measure both outcomes. We conduct a careful accounting of assets, income and consumption at baseline and track changes in each subsequent survey. Cynthia Kinnan will help us use this data to understand the impact of insurance on financial status. We also measure both self-reported and objective measures of health, including medical history, medication usage, and blood pressure, lung capacity, body fat etc. We are also hoping to incorporate other biomarkers into the study, including hair cortisol and dried blood spots; Gabriella Conti is helping that effort.

We examine a number of secondary outcomes. One is cognitive capacity. With the help of Anuj Shah, we examine the impact of health shocks on cognitive capacity, determine whether the cognitive load from health shocks is mediated by stress, and identify whether load is amplified by the financial costs of treatment and, if so, whether insurance can mitigate it. Another outcome we explore is the household allocation of insurance and its measurable counterpart, medical care, in a study led by Alessandra Voena. Prior literature suggests a large gender skew in the allocation of cash transfers to households: males get a disproportionate share. An interesting question is whether that is also true for health insurance or health care. An argument suggesting it might not be true in this context is that women may have better knowledge of health needs and a greater willingness to act on it.

Surveys. Our primary instrument for measuring outcomes is an annual survey conducted on all study households. This annual survey is conducted prior to treatment allocation and then once at midline (1 year into the study) and then again at endline (2 years into the study). In addition, we will conduct a serious of sickness surveys in which we repeatedly call a random subset of sample households and ask them if they had a sickness or accident in the last month, whether they utilize medical care, and how they paid for that care. This survey will improve recall of health events and, importantly, increase our power by examining how insurance affects utilization conditional on sickness.

Corollary studies. Our study also embeds a number of corollary studies that improve our understanding of the value of insurance in a low-income country such as India. First, for roughly 150 households at baseline and then a large fraction of the sample at endline, we will ask households about their willingness to pay for insurance. The surveys are incented and are intended to tease out a demand curve for insurance so the government can determine uptake at different premiums and premium subsidy levels.

Second, our study includes a design tweak intended to determine whether public insurance has spillover effects on informal insurance and local credit markets. A concern with providing public insurance is that individuals who participate in that formal insurance may stop participating in informal village insurance schemes, harming neighbors; alternatively, if providing public insurance doesn t stop an individual from participating in an informal insurance pool, it may reduce demands on that pool, benefiting neighbors. Similar issues arise with local credit markets. To determine whether there are such spillovers, we worked with Kosuke Imai to design a two-stage randomization design. The proportion of households within each treatment arm varies from village to village. Stage one of randomization is the random allocation of villages to these varied proportions. In stage two we randomly assign households within a village to arms based on the village-level allocation. Thus some villages will randomly have more households given free RSBY insurance than other villages. This will allow us to study the effect of increasing the provision of free formal insurance on households not given free formal insurance.

Third, with the help of Vani Kulkarni at Yale and Stefan Ecks at U. Edinburgh, we will conduct an ethnographic study to determine how individuals value health and western health care and whether they understand the concept of health insurance. RSBY insurance pays for hospitals that provide western healthcare. An important puzzle is why there is not a higher uptake of RSBY given that it is practically free. One possibility is that individuals do not value healthcare or value self- or traditional-care over western care. This is hard to tease out in traditional economic surveys, but can be targeted with open ended interviews and participant observation employed by anthropologists.

We are constantly looking for ideas and funding to supplement our existing study with new outcomes and corollary studies. For example, we are working with Sarojini Rao, a graduate student, and Shantveer Patil of the RSBY office in Karnataka to add a hospital survey in order to determine why hospitals participate in RSBY and what the hurdles are to utilization.

The study above, which is already in the field, is intended to determine the impact of expanding eligibility for RSBY. We just completed our baseline survey and we begin treatment assignment this month. At the same time we are planning and seeking funding for a series of follow-up studies that ask alternate policy questions and would begin later this year. One would examine the benefit of expanding RSBY coverage to include physician care and medicines. The design of that study would be similar to the existing study, but would compare hospital only insurance to expanded coverage insurance. Another would examine the relative merits of expanding insurance coverage to the merits of universal financial access for the purpose of improving access to medical care.

While we have embarked on a large and ambitious project, we think the policy questions that motivate the study warrant both scale and rigor. When we return to the question we posed at the beginning of this post ( What study of the US healthcare system would you design if you were transported back to 1965 or 1970? ) it is clear that there is a great deal at stake. Imagine how much we could improve health and welfare in India if we had reliable evidence on the impact of different policy reforms on table. Just imagine that.


Anup Malani (PhD, JD) is a professor at the Univeristy of Chicago Law School and the University of Chicago’s Pritzker School of Medicine. He is also a University Scholar at Resources for the Future, a Research Associate at the National Bureau of Economic Research, a Senior Fellow at the University of Southern California’s Schaeffer Center and an editor at the Journal of Law and Economics. Dr. Malani is the prinicipal investigator on the Indian Health Insurance Experiment. 

Phoebe Holtzman is a co-founder of the International Innovation Corps, a program that jump-starts innovative development projects in India, and a Research Professional at the Coase-Sandor Institute for Law and Economics at the University of Chicago Law School. Phoebe manages the Indian Health Insurance Experiment in Karnataka, India. She has a AB in Anthropology from the University of Chicago.


Image via Gates Foundation/Flickr