In a new IHFE paper, Jack Hadley and James Reschovsky conclude,
Using an instrumental variable method, our analysis suggests that increased medical care use is associated with statistically significant and quantitatively meaningful reductions in mortality rates and rates of avoidable hospitalizations among fee-for-service Medicare beneficiaries who are predicted, based on their medical conditions, to have high medical care use. These results are inconsistent with a large body of area-level analyses (Fisher et al. 2009), but are similar to the results of two other recent analyses that applied instrumental variable methods to data sets of individual Medicare beneficiaries’ medical spending and health outcomes. Hadley et al. (2011) found that a 10% increase in Medicare beneficiaries’ medical spending significantly increased the value of a health index by 1.2–2.2% and survival rates by 1.2–1.7%. Kaestner and Silber (2010) estimated that a 10% increase in spending reduced mortality rates by 3–6% for cohorts of Medicare patients hospitalized for various medical or surgical conditions. Since hospitalization is one of the key determinants of being a high cost patient, the magnitude of their estimated effect is quite comparable to our estimate that a 10% increase in medical care use is associated with a 8.4% decrease in the mortality rate, along with a 3.8% decrease in the rate of avoidable hospitalizations among high-cost beneficiaries. Moreover, it appears that people with chronic conditions account for more of the reduction in avoidable hospitalizations.
However, […] these results do not necessarily imply that there are no inefficiencies in the Medicare program. Even with a positive marginal impact of spending on health, efficiency can still be increased by shifting the health production function upwards. What is needed is much more precise information about where those inefficiencies occur and how they are related to practice organization, financial incentives, market structure, and information gaps. Comparative effectiveness research focused on specific treatments for specific diseases is a step in that direction. In fact, considerable effort is underway to apply IV methods to comparative effectiveness research that uses observational data (Pizer 2009). [Bold added.]
Two points:
- There is a temptation by many to interpret the existence of seemingly inconsistent results documented in the first paragraph as evidence of a conflict. In fact, there is no conflict. Not only are we not on the flat of the curve, we’re not even on the production possibility frontier. Consistent with that fact, it’s well documented that different health systems have different production functions (i.e., some convert dollars to health much more efficiently than others). In such a setting, finding different correlations (even of differing in sign) between spending and outcomes is not surprising or necessarily inconsistent. (See Joynt and Jha.)
- Observational study-based comparative effectiveness holds a lot of promise for improving the efficiency of health care delivery at lower cost (and without some of the ethical issues) of randomized trials. Though no substitute for RTCs, they’re worthy weapons in the arsenal. For more on Steve Pizer’s paper, cited by Hadley and Reschovsky in the second paragraph above, see this prior post.
