Prescription Drug Cost Sharing:  Associations With Medication and Medical Utilization and Spending and Health FREE

Medical practice in the United States has changed dramatically in the last several decades, including an increase in use of prescription drugs. More and better-quality drugs are available to prevent and manage chronic illness, and these drugs reduce mortality, forestall complications, and make patients more productive.¹ Thus, access to outpatient drugs is now a cornerstone of an efficient health care system.

But with recent increases in pharmacy spending, pharmacy benefit managers and health plans have adopted benefit changes designed to reduce pharmaceutical use or steer patients to less-expensive alternatives. The rapid proliferation of mail-order pharmacies, mandatory generic substitution, coinsurance plans, and multitiered formularies has transformed the benefit landscape. In this review, we analyze how the salient cost-sharing features of prescription drug benefits may affect access to prescription drugs and synthesize what is known about how these features may affect medical spending and health outcomes.

Most beneficiaries are now covered by incentive-based formularies in which drugs are assigned to one of several tiers based on their cost to the health plan, the number of close substitutes, and other factors.² For example, generics, preferred brands, and nonpreferred brands might have co-payments of $5, $15, and $35, respectively. In contrast, plans may require beneficiaries to pay coinsurance—ie, a percentage of the total cost of the dispensed prescription. The purpose of tiered co-payments and coinsurance is to give beneficiaries an incentive to use generic or low-cost brand-name medications and to encourage manufacturers to offer price discounts in exchange for inclusion of their brand-name products in a preferred tier. By 2005, most workers with employer-sponsored coverage (74%) were enrolled in plans with 3 or more tiers, nearly 3 times the rate in 2000 (27%).³

Some plans also impose benefit caps that limit either the coverage amount or the number of covered prescriptions. For example, the standard Medicare Part D benefit offers beneficiaries coverage of up to $2400 in spending in 2007, at which point coverage stops until beneficiaries reach a catastrophic cap ($5451). Once the catastrophic cap is reached, coverage resumes with minimal cost sharing. Prior to the introduction of Part D, benefit caps—without this catastrophic limit—were a standard feature of Medicare + Choice plans (now known as Medicare Advantage) and some retiree plans. As of 2002, 94% of Medicare + Choice plans that covered branded drugs had an annual dollar cap ranging from $750 to $2000 per year.⁴ Analogous policies used by state Medicaid programs place limits on the number of prescriptions dispensed per patient per month. Because benefit caps represent an extreme version of cost sharing—patients who reach them must pay all additional pharmacy costs out of pocket—and their central role in Part D, we include them in our review.

Additional cost-saving measures include prior authorization (requiring permission before certain drugs can be dispensed), step therapy (requiring use of lower-cost medications before providing coverage for more expensive alternatives), closed formularies, mandatory generic substitution, and reference pricing (a cap on the amount a plan will pay for a prescription within a specific therapeutic class). There is a growing literature on each of these cost-containment measures.

We conducted electronic searches of PubMed for studies published in English between 1985 and 2006. The primary search was based on combinations of 2 sets of key words. The first set included various terms for drug cost sharing: cost-sharing, incentive-based, copay, coinsurance, tiered benefit, benefit cap, patient charge/fee, user charge/fee, prescription charge/fee, step therapy, reference pricing, prior authorization, formulary, formulary restriction, formulary limit, closed formulary, open formulary, and generic only. The second set included drug spending, drug cost, prescription drug, medication, and pharmacy benefit. Articles that contained at least 1 term were included. We performed another search specifically for Medicaid-related drug cost sharing measures by combining one of the terms access restriction, drug/prescription/reimbursement limit, or preferred drug list with Medicaid and with one of the terms spending, use, or cost. We excluded issue briefs, comments, letters, editorials, essays, articles without author names, and reviews. This process yielded 923 studies. We then screened these studies based on titles, abstracts, and, in a few cases, the full text, as described in the Figure.

A study was included in this review if (1) the article was published in a peer-reviewed journal; (2) it examined the effects of cost sharing (co-payments, tiers, coinsurance, reference pricing, formulary restrictions, or benefit caps) on at least 1 of the relevant outcomes (prescription drug utilization or spending, medical utilization or spending, or health outcomes); and (3) it analyzed primary or secondary data (to exclude simulations).

Among the 923 studies, 111 met these criteria. An additional 21 studies were added based on reference lists, resulting in 132 studies for final analysis. Sixty-five studies examined co-payments, tiers, or coinsurance^{5- 4243- 69}; 11 examined benefit caps^4,43,70- 78; 41 examined formulary restrictions^79- 119; and 16 examined reference pricing.^120- 135 (One study examined both co-payments and benefit caps.⁴³)

Because the majority of these studies analyzed observational data, understanding how the associations between cost sharing and the outcomes of interest were measured is important. We classified study designs as follows:

The literature on cost sharing is much more diffuse than many medical interventions, which benefit from clear delineation of primary and secondary clinical end points. For example, some articles examine pharmaceutical spending, while others observe utilization. And, among the latter, utilization is measured in at least 5 different ways: medication possession ratio, proportion of days covered, cumulative multiple-refill gap, number of prescriptions, and aggregate days supplied. This problem is further exacerbated by the wide range of “treatments”—eg, adding a second or third tier, raising co-payments, requiring coinsurance—and treated populations with very different diseases. The result is tremendous heterogeneity in benefit changes, the way results are reported, and for which affected populations. Thus, many of the conclusions of this review are necessarily qualitative.

Details from the 132 studies in this review, including study sample, study design, drug benefit variation, outcomes, and key findings, are available online (http://www.jama.com).

The evidence from the 65^6,69 studies that examined the relationship between 3 of the most important features of drug benefits—co-payments, tiering, and coinsurance—and pharmacy utilization and costs is summarized in eTable 1A, eTable 1B, eTable 1C, and eTable 1D.

Most of the evidence comes from observational studies, with the exception of 2 studies of the RAND Health Insurance Experiment (HIE). The HIE randomized 2750 families to different levels of cost sharing ranging from free care to 95% coinsurance. The HIE demonstrated that individuals subject to higher coinsurance rates reduced their demand for care and that the cost-sharing response for prescription drugs was similar to the response for all ambulatory care.^66,68 However, data from the HIE are nearly 3 decades old. In addition, the health insurance packages in the HIE varied the prescription drug benefit at the same time as other benefits. Thus, it is unclear whether the higher drug expenditures among patients with more generous coverage were due to lower out-of-pocket costs for drugs or lower cost sharing for office visits and other medical services that are the usual pathways for receiving prescriptions.

All of the remaining studies are observational. Key features of the best studies include large sample sizes, variation in benefit design both across plans and over time, and attempts to control for other factors known to affect pharmaceutical use.^{19- 20,22- 24,27,36,38,42} Of particular value are studies that used data from multiple plans and controlled for medical benefits, especially when they may have been changing in concert with the pharmacy benefit.^{8- 9,26,34,40,49- 50,52} For example, changing office visit co-payments affects how frequently patients see a physician and, hence, the number of prescriptions they receive. Poor features include analyses that do not control for other factors that might be changing, including observations before and after a benefit change with no control group. These designs include (but are not limited to) many international studies in which co-payments changed for the entire population.^{5,7,12,15- 16,29- 30,35,37,41,44,57,59,61,64- 65,67}

Some of the studies found relatively small changes in utilization in response to higher cost sharing,^{17,51,55,60,62,69} but these focused on small changes in co-payments. In some studies, the control groups had very different characteristics,^{28,32,39,45- 46,53,67} patients may have self-selected into treatment groups on the basis of medication choice,¹³ or the source of co-payment variation was not clear.²⁵ Given the evidence that there is differential response by condition or class of medication, studies that restrict attention to a specific patient population or conduct subgroup analysis can yield additional insight.

The effects of cost sharing can be summarized using the price elasticity of demand. This measure represents the percentage change in drug spending that would be associated with a 1% increase in cost sharing. When we excluded the studies that involved very small cost-sharing changes or did not have an adequate control group, we found elasticities ranging from −0.2 to −0.6, indicating that cost sharing increases of 10% (through either higher co-payments or coinsurance) would be associated with a 2% to 6% decline in prescription drug use or expenditures.

Eleven of the 65 studies in our review explicitly looked at changes in coinsurance rates,^{28,32,35,41- 42,44,48,54- 55,66,68} with 2 of these coming from the HIE and 4 from a benefit change in Quebec in 1996. Overall, increasing coinsurance is at the low end of our range of −0.2 to −0.6, but these associations are attenuated by the simultaneous imposition of out-of-pocket maximums in most of these studies.

Several studies suggest that consumer sensitivity to cost sharing depends on a drug's therapeutic class and that increased cost sharing may decrease “nonessential” drug use (eg, antihistamines) more than “essential” drug use such as antihypertensives and oral hypoglycemics. However, the empirical evidence in this area is mixed. Harris et al⁶² found substantially larger reductions in the use of discretionary medications than essential medications in response to a modest increase in co-payments. More recently, Goldman et al²⁶ found that doubling co-payments was associated with reduced use of 8 classes of medication by 25% (antidiabetics) to 45% (anti-inflammatories). Patients were less likely to reduce use of these drugs if they were receiving ongoing care from a physician for the disorder, ranging from 8% (antidepressants) to 31% (antihistamines). Landsman et al²⁰ found similar price responses across 9 therapeutic classes. Several other studies found modest but inconsistent effects of higher co-payments on use of essential and nonessential drug classes.^{33,47,50,55,69}

Information from the 11 studies^4,43,70,78 that examined the association between benefit caps, including caps that limit the number of prescriptions and caps on an annual pharmacy benefit, and drug use and drug costs is summarized in eTable 2.

Soumerai et al⁷⁷ compared Medicaid patients in New Hampshire—for whom the program had imposed a 3-drug limit per patient per month—with those of New Jersey, where no such cap was introduced. They found a 35% reduction in drug use relative to the control group. For patients taking psychotropic medications, they found that the cap was associated with a 15% to 49% reduction in the use of these drugs.⁷⁶

The most salient evidence on the impact of benefit caps comes from an analysis of medical and pharmacy claims from a single closed-network health maintenance organization.⁷⁰ Members whose benefits were capped at $1000 had 31% lower pharmacy costs than comparable enrollees not subject to a cap. One survey of Medicare beneficiaries suggested that elderly individuals who experience gaps in coverage report using fewer medications, are more likely to switch to generics or lower-cost medications, and rely more on drug samples from their physicians.⁷¹ Two other studies found that patients exceeding the cap were 2 to 3 times more likely to discontinue a medication⁷³ and unenroll from the plan.¹³⁶

Information from the 16 studies^121,135 examining reference pricing, wherein insurers cap the amount they will pay for a prescription within a specific therapeutic class, is summarized in eTable 3.

Few health plans in the United States have adopted reference pricing so far. However, it is widely used in parts of Canada and Europe. In general, almost all of the studies found large increases in use of drugs priced at or below the reference price and sharp declines in use of higher-cost drugs that require some patient cost sharing. In a series of studies of reference pricing in British Columbia, Schneeweiss et al^{123,128- 129} found that an increase in co-payments for the most expensive angiotensin-converting enzyme inhibitors (drugs priced above the reference price) was not associated with stopping treatment for hypertension or higher health care utilization. They found similar associations with use of calcium channel blockers¹²⁵ and proton pump inhibitors.¹²¹ The only potential concern raised by these studies was that low-income patients were more likely than high-income patients to stop hypertensive therapy (odds ratio, 1.65; 95% confidence interval, 1.43-1.89).¹²⁸ Grootendorst and colleagues^122,131 examined similar policies for nonsteroidal anti-inflammatory drugs (NSAIDs) and nitrates. They found that most of the savings to British Columbia's Pharmacare program could be explained by the substitution of low-cost drugs and higher patient cost sharing for restricted medications. The remaining studies listed in eTable 3 found that reference pricing was only weakly associated with overall use within the restricted drug class and uncorrelated with medical service use.

Evidence from the 41 studies^79,119 examining the association between prior authorization or formulary restriction and drug and medical utilization and spending is summarized in eTable 4A, eTable 4B, and eTable 4C.

Increasingly, public and private health plans are imposing prior authorization and/or fail-first requirements on nonpreferred prescription drugs. These programs require use of older or less expensive medications before covering newer therapies. For example, a plan may require a patient to try at least 1 generic NSAID before paying for a cyclooxygenase 2 inhibitor. The main concern about these cost-containment policies is that patients may switch to less-effective medications or become nonadherent and, as a result, experience adverse health effects. Several studies support such concerns. Two studies found that Medicaid beneficiaries taking a restricted statin medication filled fewer prescriptions and were more likely to be nonadherent than unrestricted patients.^79,84 Similar associations were found for antihypertensive medications.⁹⁵ Another study found that a preferred drug list for cardiovascular medications was associated with an increase in outpatient visits in the first 6 months of implementation, but these differences did not persist over time.⁹³

Most other studies, in contrast, suggest that the outcomes associated with prior authorization and step therapy requirements are modest, although these policies can have strong associations with use of restricted medications. For example, Tennessee Medicaid's expenditures for NSAIDs declined by 53% following implementation of prior authorization and fail-first requirements for brand-name NSAIDs.¹¹³ The reduction in spending was associated with higher use of generic NSAIDs and a 19% decline in overall NSAID use. The findings of Kotzan et al^114- 115 were similar. More generally, prior authorization programs are associated with lower drug spending within the restricted class but are uncorrelated with medical care utilization and spending.^90- 91,99 Two studies on step therapy^82,111 also reported decreased drug spending without adverse effects on drug utilization or physician concerns.

The outcomes associated with closed formularies or generic-only drug coverage may differ substantially from those of prior authorization requirements. One study found that a closed formulary was associated with lower rates of medication continuation among patients with chronic conditions.¹⁰⁶ Two other studies found that limiting coverage to generic drugs was associated with decreased medication use⁸⁷ and increased hospitalizations.⁹⁶ Another study¹⁰⁸ found that the degree of formulary restrictions was positively correlated with higher drug costs, more office visits, and high likelihood of hospitalization among patients with certain diseases.

The evidence clearly demonstrates that increased cost sharing is associated with lower pharmaceutical use. These effects can be quite large—even for long-term medications—suggesting that there are long-term health consequences. In fact, the direct evidence on the link between cost sharing and health is rather limited. Most studies examine important proxies for health (and medical spending), such as emergency department use and hospitalizations. The findings from studies focusing solely on chronically ill patients are unambiguous: for patients with congestive heart failure,⁶ lipid disorders,^8,10 diabetes,²¹ and schizophrenia,⁷⁶ greater use of inpatient and emergency medical services are associated with higher co-payments or cost sharing for prescription drugs or benefit caps. These findings are corroborated by the only article that studied clinical outcomes in a population with benefit caps.⁷⁰

In contrast, studies that observed the outcomes associated with cost sharing more broadly (on all drugs or a wide range of classes) were ambiguous in their findings. Some found that higher cost sharing is associated with adverse outcomes,¹³⁷ particularly among vulnerable populations such as elderly and poor patients.^48,77 But most found that when the population is not limited to those with certain chronic illnesses, the outcomes associated with prescription drug cost-containment policies are mostly benign. For example, studies by Fairman et al,³³ Motheral and Fairman,⁴⁷ Johnson et al,⁵⁴ and Smith and Kirking¹³⁸ found that increased co-payments were not associated with more outpatient visits, hospitalizations, or emergency department visits.

Although there is ample evidence that the demand for pharmaceuticals declines with higher co-payments, there is concern that low-income beneficiaries will be more responsive to cost sharing. Most evidence on this point comes from nonexperimental studies of state Medicaid programs that introduced very modest co-payments. Medicaid enrollees in South Carolina used significantly fewer drugs after the imposition of a $0.50 co-payment.⁶⁹ A more recent study found that elderly Medicaid recipients residing in states with co-payment provisions consumed fewer drugs and were less likely to fill any prescriptions during the year than those in states without co-payments.⁵¹ Survey data indicate that 1 in 4 Medicaid patients aged 18 to 64 years could not afford to fill at least 1 prescription in the past year compared with less than 1 in 10 among privately insured individuals.¹³⁹ A study of Medicare beneficiaries in Pennsylvania found that elderly individuals with annual incomes of more than $18 000 were 18% more likely to treat medical problems with prescription drugs than those with incomes of less than $6000.¹⁴⁰

We reviewed studies examining the association of co-payments and other salient benefit features with pharmaceutical utilization and spending, as well as their association with nonpharmaceutical services and health outcomes. The evidence summarized here suggests that for each 10% increase in cost sharing, overall prescription drug spending decreases by 2% to 6%, depending on class of drugs and patient condition. Benefit changes are not without consequences: for some chronic conditions, we found that higher cost sharing for prescription drugs was associated with greater use of expensive medical services.

It is interesting to compare these effects with other interventions designed to improve use of chronic medications. A 2002 review, for example, identified 33 interventions designed to improve patient adherence to prescribed medications.¹⁴¹ Even the most successful interventions did not result in large improvements in adherence and generally relied on complicated, labor-intensive regimens of uncertain effectiveness. Thus, pharmacy benefit design represents one of the most important public health tools for improving patient treatment and adherence.

Several key research issues remain unresolved. First, while greater cost sharing is clearly associated with reduced access, the precise mechanisms are not clear. Less pharmaceutical use could come about through 3 different behavioral pathways: reduced initiation of prescription drug treatment, worse adherence among existing users, or more frequent discontinuation of therapy (although the latter could be interpreted as an extreme example of poor adherence among users). Distinguishing among these hypotheses is important because it affects the advice and monitoring that physicians and plans should use to counteract any adverse consequences of plan design changes. We found evidence that all 3 pathways may be complicit when cost sharing rises, although adherence among existing users seems to be the primary mechanism. On the other hand, if one accepts the criteria in current national guidelines, then even small effects of cost sharing on the likelihood of initiating therapy could have dramatic health consequences. For example, Topol¹⁴² notes that 36 million people in the United States should be taking a statin but only 11 million are currently being treated.

Second, increased cost sharing is associated with adverse medical events such as hospitalizations and worsening clinical outcomes over 1 to 2 years for patients with congestive heart failure, lipid disorders, diabetes, and schizophrenia. Additional evidence suggests that there may be adverse consequences for asthma as well. Because patients leave employers and plans with relative frequency, and plan benefit designs change rapidly, it is difficult to isolate the long-term health consequences of changes in cost sharing using existing study designs.

A key challenge in this type of analysis is that disease severity cannot be measured directly and that patients who are more severely ill tend to use more drugs and more of other services. If this tendency is not properly accounted for in the data analysis, estimates of the effects of prescription drug use on other costs will demonstrate little or no cost savings. This spurious correlation probably has limited past efforts in this area. This is especially problematic when patients have a choice of drug plans—a feature that introduces bias in the same way that patients self-select into treatment regimens. Some of this bias is mitigated because while many employers offer employees a choice of medical plans, the majority standardize 1 drug benefit regardless of medical plan choice. Ultimately, the long-term consequences of benefit changes remain elusive.

Third, if cost containment policies have adverse effects, those effects are likely to be magnified among low-income groups, whose high rate of chronic health problems and low incomes may result in more price-sensitive behavior. Survey data indicate that nearly 40% of chronically ill low-income persons with public insurance and 35% with private coverage report that they have been unable to fill at least 1 prescription because of cost concerns.¹⁴³ One of the severe limitations of analyses of claims data is that they do not include information on race, ethnicity, income, education, and wealth, and, when economic status is included using national survey data, there is substantial bias in its measurement.¹⁴⁴ Thus, while it is often claimed that low-income groups are most sensitive to cost sharing changes, there is little reliable evidence to support this contention.

Fourth, the introduction of Medicare Part D has initiated a bold experiment with benefit caps. While the effects of benefit caps clearly are consistent with those of other cost-sharing features, little is known about the dynamics of these changes. For example, if patients do discontinue therapy or take their medicine less frequently once they reach their benefit cap, how quickly—if ever—do they reinitiate drug therapy once coverage resumes in the next benefit year? Benefit caps also provide a counterpoint to consumer-directed health plans that encompass high-deductible catastrophic coverage. With these plans, patients must pay all the costs until a cap is reached, beyond which they pay nothing. Benefit caps, in contrast, provide coverage up to a specified limit. A comparison of these financing alternatives is needed, especially with regard to how they might affect those with chronic illness.

Fifth, there has been a dramatic increase in the use of specialty drugs (ie, agents targeting a gene or protein, which are typically injected or infused). They are often used to treat complex, chronic conditions such as anemia, cancer, growth hormone deficiency, and multiple sclerosis, but at prices that can be substantially higher than traditional medications. Historically, only a small percentage of plan members had these conditions, so the total population of specialty drug users was quite small. However, spending on specialty drugs is expected to increase substantially in the near future as new drugs enter the market for treatment of diabetes, osteoporosis, and rheumatoid arthritis—diseases that affect much larger populations. Many insurers are contemplating a variety of cost-sharing strategies to control their use and costs. There is some evidence that patients are less price-responsive for these products than for traditional oral agents,¹¹ perhaps because of relatively few alternative therapeutic options. Whatever the reason, this area may be the next frontier on which we observe dramatic changes in benefit design, and it will be important to assess the consequences for spending and health.

The majority of articles that we examined in this review were outcomes studies conducted using administrative data. The researchers typically isolated a plan or plans that changed their benefit design and analyzed the resulting patterns of prescription drug use and (less frequently) medical utilization, with the best designs including a control plan that did not change benefits during the same period. Such data are rich in sample size and measures of utilization, but they have limitations beyond the lack of important clinical detail. There is no information on socioeconomic status and one cannot control for key health-related behaviors such as diet, exercise, and smoking. Physician prescribing practices—especially whether a prescription was written but not filled—are unobserved. Furthermore, long-term follow-up is difficult because plan enrollment often changes over the course of the study.

In sum, the evidence suggests that patients are responsive to cost-sharing arrangements in prescription drug plans—even among chronically ill patients. For certain conditions, the evidence clearly indicates that more cost sharing is associated with increased use of other medical services, such as hospitalizations and emergency department visits. These findings make benefit design an important public health tool for improving population health. The challenge for public and private plans is to make patients more sensitive to the cost of treatment without encouraging them to forego cost-effective care. This requires knowing how patients respond to different incentives and cataloging the net benefits of alternative therapies, not only for health, but also for current and future health care costs, productivity, and patient utility.