Updates from AfHEA

On March 12, 2009, in Ghana, health insurance, by Karen Grepin

Today is the final day of the inaugural African Health Economics Conference, which was held this week in Accra, Ghana. The conference was well organized, well attended, and provided a lot of interesting discussion both about research findings but also about the role of health economists in influencing health policy in Africa. Health economists from nearly every country in Africa were in attendance.

Among the papers presented, one of my favorites was a paper presented by Evelyn Amsah on behalf of her research team that reported on a randomized evaluation of the effect of the new National Health Insurance in Ghana on the utilization of services. There were a few things I really liked about this study: the creative use of field research methods, the intelligent research design, as well as the main substantive findings of the paper.

In many research projects randomization is carried out by drawing up a list of names, or households, or villages, and then using a computer is used to randomly assign treatment to the relevant units. However, the researchers felt that in their study population such methods were generally not well received as people did not believe that it was actually the computer that was selecting randomly but rather it was a person who was actually selecting among people. So instead, the communities to be treated were brought together in town hall meetings and pieces of paper were placed in barrels and then heads of households came forward and selected a piece of paper. About half of the participants were assigned to treatment, which was to receive health insurance coverage for the entire household, with the understanding of the participants that the rest of the population would receive the treatment in the following year. This method allowed for more trust and acceptance of the trial among the communities. It also provided a fair way to allow for randomization at the household level, rather than the village level.

They also asked the households to keep track of health care episodes and treatment outcomes using really interesting visual diaries. Pictures of children with fevers and other conditions were drawn. Households were asked to tick off a box following each occurrence of ones of these events.

A year after the intervention, the researchers found that there were no differences in the reported number of fevers among the treatment and intervention groups, which was a good sign that their randomization worked. The treatment households reported much higher utilization of formal and less informal care, as we would expect, however this effect was ONLY observed among households living within 5 km of a health facility. It seems that the new Health Insurance is only effective when you have good physical access to health services. This finding is very relevant in terms of thinking about how effective this major policy will be in this country.

Hopefully the full findings of this study will soon be published for all to learn from.

Share on Facebook
 

2 Responses to “Updates from AfHEA”

  1. Atheendar says:

    Karen, thanks for reporting on such an interesting paper! I had one question about the design: during the town hall randomization process (very well-played!) was treatment assignment made public knowledge? If so, those that “got lucky” may have felt socially obligated to help out their unlucky peers over the course of their study period (I have in mind informal insurance arrangements). If this is true, there would be spillovers from treatment to control, biasing downward the true program effects.

    Come to think of it, this might happen anyway, if treatment assignment becomes observable sometime after the initial randomization…

  2. Karen Grepin says:

    Good point. I think yes, people know who else gets coverage. I am not sure how the insurance cards work and how much ability they would have to help their neighbors out. I guess this is one of the draw backs of randomization at the individual rather than village or district levels, but that approach would have its own drawbacks….

Leave a Reply



Analytics Plugin created by Web Hosting