Health and the Factors Driving Medical Spending

David M. Cutler

NBER Retirement Research Center Paper No. NB 06-14
Issued in September 2008

With the rising cost of health care already a major concern for government, employers and individuals, one might reasonably ask what the future holds. Is the future burden of health care greater, lower, or just different? This paper reports on a multi‐faceted research agenda on the factors that influence health and health care costs over time, and what they suggest for the future (see Cutler 2006 and Cutler, Glaeser, and Rosen 2008). Projecting the future is substantially more complicated than looking at age‐specific health care spending today, and adding to that the component costs that might be associated with an older population demographic. It is at least as importantly about the evolution over time in health behaviors, chronic illness, medical innovation, education, disability, health care financing systems, the state of the economy and many other influences. Thus I have been engaged in research across a multitude of these inter‐related influences. I start by considering theoretically how to forecast medical spending. Traditional models of spending consider population demographics only. I highlight those models and their results in Section I. Section 2 then focuses on the evolution of health behaviors and risk factors, such as smoking, obesity, and control of hypertension and high cholesterol. These risk factor changes have had a mixed influence on health trends to date and, depending on how health behaviors continue to evolve in the future, could be fundamental to population health in the future. Section 3 considers the major advances in medical innovation in recent years, most of which have improved health, some of which have done so with expensive technological care, but some of which have contained costs with more effective disease and risk management. While I make no attempt to quantitatively integrate these various influences to project future health care costs, I conclude with some observations about a model that would do so.

PDF (298 K)

Acknowledgments