At GRAPH, we first identify those social and health policies that experts believe can significantly address risk factors for poor health and premature death. We then evaluate the cost-effectiveness of these policies. Because each village, community, county, state, and nation is unique, we build models that best reflect local circumstances. Factors like disease prevalence, physical infrastructure, or a community’s demographic mix can radically alter what the best policy mix might be for any locality. Ultimately, we provide a menu of policies, ranked according to the maximal health they produce locally. Finally, we are careful to not only provide menus of policy options, but also the ingredients needed for policymakers to understand what is that they are buying with their limited dollars.
GRAPH accomplishes this by building state-of-the-art policy models that use the best available data to estimate the cost-effectiveness of both medical and non-medical risk factors for the leading causes of disease. We then develop a “community diagnosis” instrument that allows our models to be populated with key inputs from localities that request our services. This instrument allows us to estimate the most cost-effective mix of social policies for a given locality, the burden of risk factors, the economic costs of disease, and the impact of each disease on health. There are a number of characteristics that make GRAPH unique:
Our models also explore the “side effects” of our policy recommendations. Even the most effective and best-tested policies can produce unpredictable benefits or harms. If they are big enough, side effects can change the cost-effectiveness of a policy.
We refine our models over time. Our approach will not always initially produce the best information, but it will help us understand what we need to know when the evidence is imperfect. Therefore, in conjunction with our policy models, we conduct randomized-controlled trials (RCTs) of what experts believe to be the best interventions.
We take a very broad approach, tackling policies that are believed to be most effective by public health experts rather than experimental data alone. We came to this approach after a complete review of experimentally-tested primary preventive interventions—and their health impacts. This project, which was funded by the Rockefeller foundation, found that most such experiments have been conducted within the medical arena, and most have very little impact on health as a whole. While other organizations study interventions that have been “proven” using replicated experiments employing experimental methodology, GRAPH recognizes that some of the most powerful policy “medicines” we have in our toolkit were never properly evaluated