Some of the increase in cost sharing, however, derives from explicit changes in law. The passage of the Medicare Prescription Drug, Improvement, and Modernization Act (PL 108–173) (MMA) resulted, most famously, in the initiation of prescription drug coverage for Medicare beneficiaries. However, the MMA also created the option for pretax health savings accounts tied to high-deductible health plans. In such plans, an unspent balance in an account could be rolled over from one year to the next (3). Conforming high-deductible health plans must have deductibles of between $1,000 and $5,000 a year for an individual and between $2,000 and $10,000 a year for a family.

And, of course, the Medicare drug benefit has a complex set of rules for cost sharing: an annual deductible of $275, then 25% copayment until drug expenditures reach $2,510 a year, then no coverage (the so-called “donut hole”) until expenditures reach $5,726, before reaching 95% coverage (4). For example, a person with rheumatoid arthritis (RA) who incurs $5,726 a year in prescription drug expenses can expect to pay $4,050 out-of-pocket ($275 for the deductible, $559 for the 25% copayment level, and $3,216 for the donut hole). Another person with RA (e.g., one who is taking an anti–tumor necrosis factor [anti-TNF] agent necessitating $20,000 a year in prescription drug expenses) can expect to pay relatively little more out-of-pocket: $4,764 (the same amounts for the deductible and 25% copayment levels, the full $3,216 for the donut hole, but just another $714 for the remaining $14,274 of expenditures). From the perspective of society, the anti-TNF agent would have to be highly effective to justify the increment of more than $14,000 in expenditures between the 2 persons with RA; paradoxically, however, from the perspective of the patient, the increment of costs he or she would experience is relatively small.

In this issue of Arthritis & Rheumatism, Lurie and colleagues (5) have shown us the extent to which cost sharing had already proliferated for persons with arthritis just prior to the passage of the MMA. Their estimates derive from the Medical Expenditures Panel Survey (MEPS), an annual tally of health costs from the federal government’s Agency for Health Care Research and Quality (AHCRQ) (6). In real terms, between 1998 and 2004, median total out-of-pocket expenditures increased from $757 to $1,154 (52%), and median prescription out-of-pocket expenditures increased from $415 to $714 (72%). The increases at the 90th percentile were even larger: 71% and 77% for total and prescription out-of-pocket expenditures, respectively.

After tallying the rise in out-of-pocket expenditures actually experienced by MEPS respondents, Lurie et al projected the impact of the MMA on out-of-pocket expenditures under very specific assumptions. The assumptions were that the increase recorded between 1998 and 2004 would have continued in the absence of the MMA, that Medicare beneficiaries would choose the standard drug benefit available in Medicare unless they already had benefits more extensive than those offered through the MMA, that no major changes would occur in private drug insurance plans at the same time as the implementation of Medicare prescription benefits, that all expenditures for drugs would be through a Medicare formulary, and that no change in physicians’ prescribing behaviors would occur. Under this set of assumptions, they projected a slight decrease in total and prescription out-of-pocket expenditures at the median and 75th and 90th percentiles as a result of the addition of Medicare coverage. However, their projected decrease in such expenditures would not be sufficient to offset the increases that had occurred between 1998 and 2004.

We know that several of these assumptions may not be true. For example, there is ample evidence that physicians change their prescribing behavior in response to explicit changes in rules governing reimbursement and with knowledge of their patients’ ability to pay through insurance coverage and level of affluence (7–11). We also know that the pharmaceutical manufacturers will do whatever they can to upset the assumptions that patient demand for prescriptions will only increase on the pre-MMA trend line and that physicians’ behavior will be unchanged.

The estimates of out-of-pocket expenditures reported by Lurie and colleagues stop at the 90th percentile, way below the level logic tells us that persons taking biologic agents for RA would experience. Indeed, in a study of several chronic diseases including RA, Gold-man and colleagues (12) reported that, at the 90th percentile, persons with RA experienced almost $9,000 in out-of-pocket payments; they also reported that those at the 99th percentile had >$50,000 a year in such payments. So, as they say in the ads, your experience may differ from what is typical for persons with arthritis. Even in the study by Lurie et al, however, the upper ten percentiles of persons age ≥65 years with arthritis constitute between ~950,000 and 1.3 million individuals in the years covered by their study, and, of course, the study omits those of working ages, some of whom have poor or no coverage. But even if one questions the projections by Lurie and colleagues of out-of-pocket expenditures post-MMA, the increases pre-MMA are substantial and, the authors claim, not reversible by the MMA.

There are 2 principal rationales for increased cost sharing, namely, that it reduces health care utilization and, thus, total expenditures, and that it fosters prudent buying by creating an incentive for the patient to equilibrate costs and benefits more thoroughly. There is now strong empirical backing for the first rationale. Starting with the RAND Health Insurance Experiment more than a quarter of a century ago (13) and subsequently continuing with a series of observational studies reviewed by Yegian (14), researchers have shown that increased cost sharing reduces health care utilization and costs. There is even some evidence in the rheumatic diseases. In a pair of studies of a natural experiment in British Columbia in which the cost sharing for medications for elderly persons with RA was changed, Li et al (15) and Anis et al (16) showed that increased cost sharing reduced medication expenditures, but with the paradoxical effect that outpatient costs—not controlled at the same time—increased. In their study of several chronic conditions, Goldman and colleagues (12) reported that a doubling of cost sharing resulted in a 21% decrease in specialty drug spending for persons with RA (read: biologic agents), as well as a 16% decrease for other RA-related spending. These responses to cost sharing were far greater than those for kidney disease, multiple sclerosis, and cancer, the other diseases studied, perhaps because expenditures for RA are considered more discretionary by patients, physicians, or health plans.

Cost sharing may not save money appropriately. Prudent buying in the context of electronic equipment may mean purchasing the store brand or last year’s model. In the health context, prudent buying means that one would choose to purchase more of those treatments that return good value and fewer of those that do not. However, good value can clearly differ among individuals; patients may choose to relieve pain rather than cure disease, even if you or I think they should do otherwise. And even if we think that rational decision making dictates a certain course of action, well, there’s all that emerging evidence from behavioral economics that suggests that even those trained to think “rationally” make the same errors of judgment that the rest of us do, for example, by being more averse to losses than to gains of equal expected value and prevalence (17–19).

In the RAND Health Insurance Experiment (HIE), the investigators defined packages of services for which there was consensus about the appropriateness of care and reported that those with higher levels of cost sharing reduced appropriate and inappropriate services indiscriminately (20). There was also some evidence in the HIE that cost sharing was inequitable in so far as the poor and elderly were less likely to get care deemed appropriate for hypertension, the only condition for which there was sufficient statistical power despite the magnitude of the HIE. More recently, Trivedi and colleagues reported that when higher cost sharing is imposed on women, fewer undergo mammography even though it is clearly indicated (21). Moreover, the effect of cost sharing was heightened among women living in areas with high concentrations of the poor and poorly educated. In reviewing the literature, Yegian reported actual and projected impacts of trends in cost sharing on those with severe chronic disease (14), while Tamblyn et al (22) observed actual reductions in use of essential medications by the elderly and welfare recipients as a result of increases in cost sharing, with a subsequent rise in adverse events and emergency department visits as a result of the reductions.

Given the current status of federal research agencies in general as well as that of the AHCRQ, the agency most likely to conduct such studies in particular, pivotal trials of the health impacts of cost sharing are unlikely to occur in the near future. But from the important study by Lurie and colleagues, we know that cost sharing for persons with arthritis is increasing (whether or not one agrees with their assumptions about the immediate post-MMA period). We know from other studies that persons with arthritis will respond to the economic incentives targeted on them by reducing their use of services for which cost sharing is imposed (and, perhaps, by increasing their use of services without additional cost sharing, if that is possible).