Are Copays Penny-Wise and Pound-Foolish?

Copays for doctor visits are an “ill-advised cost-containment strategy,” says Dr. Amal Trivedi of Brown University, one of the authors of a recent study. “They led to more health-care spending and, most of all, worsened health for seniors.” The NEJM study that Dr. Trivedi is referring to received extensive publicity, probably because of the perception that people who have to pay for their own care skimp on basic care and end up raising overall costs.

Yet the penny-wise and pound-foolish claim has been repeatedly contradicted by a literature that began with the 1982 RAND Health Insurance Experiment. A recent addition is the May 2009 monograph on consumer-driven health plans from the American Academy of Actuaries.

The RAND experiment showed that cost sharing reduced utilization with no adverse effects on health. The American Academy of Actuaries monograph concluded that “[the] primary indications are that properly designed [consumer-directed health] plans can produce significant (even substantial) savings without adversely affecting member health status. To the knowledge of the work group, no data-based study has emerged that presents a contrary view.”

A close look at the underpinnings of Professor Trivedi’s study suggests that it does nothing to refute these findings. Since the Trivedi paper does not measure changes in individual health, it cannot be cited as information on how raising copays affects individual health. Since it does not measure overall individual health care costs, it cannot be cited as containing information on the effect that higher copays have on overall health care costs. In fact, it does not appear to measure any health care costs at all.

What it does measure is changes in three inpatient hospitalization measures in the years before and after 18 Medicare Advantage health plans raised physician visit copays. Those 18 “case” plans were chosen from a total of 172 plans operating between 2001 and 2006. They were chosen because they were the only plans that raised copayments for primary and specialty physician visits in one year while keeping prescription drug copays constant. Whether they are representative plans is anyone’s guess.

To calculate the difference that increased copays made, the before and after hospitalization rates for the “case” plans that raised copays were matched with 18 handpicked “control” plans that didn’t raise copays over the same two year period. Readers were solemnly assured that the “case” plans were similar to the handpicked “control” plans because they were matched on the basis of census region, some unspecified “model type,” and “tax status.”

The authors provide little comparative information for the case and control plans. The use of Census regions to match plans means that it is impossible to rule out the possibility that a “case” plan in rural Idaho was matched with a “control” plan in urban Los Angeles. Given that existing data suggest that disease burden and hospital use varies geographically, that Medicare fee-for-service heart failure hospitalization rates range from 7 to 61 per 1,000 among US counties, and that rural Medicare beneficiaries are more likely to have knee or hip replacement surgeries than urban ones, the data given do not preclude the possibility that differences in the health of the case and control plan groups, not differences in copays, that are driving the results.

There is also no discussion of differences in plan premiums, network adequacy, plan benefits, prescription drug coverage, total annual costs, or the ability to see out-of-plan physicians, all factors that have been shown to have been important to people weighing cost versus the quality of care when they were choosing to enroll in a particular Medicare Advantage health plan during the time period covered.

Finally, the plans that were compared had different people in them from one year to the next. It is well known that income and other socioeconomic status variables are important predictors of health. Nevertheless, this study controls for only the most basic demographic information about plan members–their sex, categorical age, and race (black, white, and other). It includes only two measures of socioeconomic status: the 2000 Census measures of the proportions of people 65 or older who were below the federal poverty level, and the proportion of those 65 and older who had attended college, and does so by matching the Census measures to the plan member’s residential zip code.

Unfortunately, the 2000 Census zip code tabulation areas did not necessarily match U.S. postal service zip code areas when the 2000 Census was taken, and, given that the U.S. postal services changes thousands of zip code areas every year, may or may not provide a reliable measure of the income of people living in a plan member’s residential area some years later. Existing work on using Census-based aggregate variables to proxy for socioeconomic variation in zip code areas suggests that there is 89 percent as much variation in income within zip codes as in the general population. The correlation between whether an individual in the 1985 wave of the Panel Survey of Income Dynamics was a high school graduate and the proportion of high school graduates in his 1980 Census zip code was a disappointing 0.26.

Print