The high and rising cost of health care in the U.S. raises the important question of whether there are positive returns to increased spending in terms of patient health. Past research based on international comparisons between the U.S. and other industrialized nations, and based on comparisons between high- and low-spending regions within the U.S., suggests that these returns may be low. However, a fundamental barrier to estimating the returns to medical spending is that the relationship between spending and health outcomes is confounded by the influence patient health. In this project we seek to overcome this barrier using two complementary strategies that rely on settings where provider assignment is more convincingly separated from patient health. The first is to exploit the fact that in areas served by multiple ambulances, hospital assignment for emergencies is to some extent randomly determined by the referral patterns of ambulance units. The second strategy exploits contiguous areas on opposite sides of hospital service area boundaries. This strategy is motivated by the observation that there is marked variation in treatment intensity within narrowly-defined geographic areas served by different hospitals, and that similar emergent patients on either side of service area boundaries may be treated very differently.
In Aim 1 we will use these strategies to document differences in hospital spending incurred by similar patients from the same geographic area who are taken to different hospitals based on ambulance referral patterns and based on their geographic proximity to different hospitals.
In Aim 2 we will extend these strategies to consider instrumental variables models of patient mortality based on exogenous variation in hospital assignment.
In Aim 3 we will further decompose sources of spending and outcome differences to pin down what aspects of higher spending are driving any returns in terms of improved health overall and for specific conditions and demographic characteristics. Finally, in Aim 4 we will consider more broadly the effects of spending using longitudinal sample of Medicare beneficiaries over a six-year period. The primary contribution of our project is that by using multiple approaches and a variety of populations and outcome measures, we both produce a cross-validated set of findings and extend the previous literature on the returns to spending.
The very high cost of health care in the U.S. relative to other nations raises the question of whether health care spending in the U.S. is productive in terms of improving health. This question is difficult to answer because providers generally direct the greatest amount of resources towards individuals for whom the returns may be greatest - those in poor health. By using multiple approaches and a variety of populations and outcome measures, we will produce a cross-validated set of findings and extend the previous literature on the returns to spending.