Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

Point-in-time estimates suggest that spending per obese person is more than one-third higher than spending per normal-weight person in the United States. It does not necessarily follow that rising obesity prevalence can explain a large portion of spending growth, however. (By analogy, per-capita spending among the elderly is much higher than among the nonelderly, but aging itself explains only a modest amount of spending growth.) For 1987 and 2001, we estimate spending levels and population shares by body-weight status (distinguishing underweight, normal weight, overweight, obese, and morbidly obese persons based on standard body-mass index, or BMI, criteria) in order to estimate the effect of changes in the population BMI distribution on spending.

Changes in spending can result from either shifts in the population's BMI distribution or from changes in mean spending by BMI status. To estimate the contribution of body weight changes to spending growth, the appropriate counterfactual analysis would allow the BMI distribution (prevalence) to change while holding all other factors constant. Alternative methods yield different answers. A 'Paasche' type of approach is to freeze relative spending by BMI category at 1987 levels, then estimate what mean per-capita spending would have been in 2001 if nothing changed other than the BMI distribution. A 'Laspeyres' approach is to allow spending by BMI category to reach 2001 levels, then undo the effect of rising obesity by applying the 1987 BMI breakdown.

Since the estimated 'obesity effect' is highly sensitive to the choice of approach, the method used should be explicit. The appropriate choice depends on context. The Paasche approach, which is appropriate for a discussion concerning sources of long-term cost growth ('how much of spending growth has resulted from the obesity epidemic?'), yields an estimate of only about 4 percent. In contrast, the Laspeyres approach, which is preferable for a discussion on what could happen to spending in the future if we returned to a past BMI distribution, yields an estimated obesity contribution of 15 percent.

Estimates such as these may be biased by endogeneity of obesity prevalence to technological changes in medicine. Much of the change in obesity prevalence is due to reliably-exogenous factors such as diet and exercise behavior, but part of it may result from improvements in the medical management of chronic illnesses that have enabled some patients to live longer. Simple calculations like the ones above could therefore overstate the impact of obesity on spending growth. A similar point can be made about changes in the prevalence of chronic illnesses.