Economic arguments in favour of user charges are based on the concept of allocative efficiency, which deems resources to be efficiently allocated when people are willing to pay for a commodity at a price that reflects the marginal cost of producing the commodity 17. This has two ramifications. First, only those who are willing to pay should have access to a particular commodity. Second, providing a harmful or ineffective commodity to those willing to pay for it is efficient, whereas providing an effective and beneficial commodity to those unable to pay for it is inefficient. For example, if the presence of health insurance means that health care is free at the point of use, the consumption of health care will not reflect the marginal costs of its production, leading to welfare loss since scarce resources might be better spent on producing and consuming other commodities 18. User charges redress this loss by reinstating price: those willing to pay the price may use health care, those unable to pay must do without. From an economic perspective, any reduction in the use of health care following the introduction of user charges contributes to allocative efficiency, regardless of the distributional or health consequences.

We might ask what relevance allocative efficiency has for policy making in health care. If it is to be understood as a normative concept, then we must assume either that the distributional and health consequences are of no importance or, if they are important, that all individuals in a given society share the same level of income, the same tastes and preferences, and the same risk of ill health, etc. 1719. But neither assumption reflects reality. Policy makers in all OECD countries show demonstrable concern for population health and equity of access to health care, albeit to varying degrees, while the people living in these countries experience different levels of income and health. We therefore prefer to use a definition of efficiency that explicitly refers to the external criterion of health improvement. Under this definition, an efficient allocation of health care resources would be one that maximises health gain, where health gain is measured in a standardised manner (for example, through quality-adjusted life years) 20. For equity, we consider two dimensions. Equity in finance requires richer people to pay more for health care, as a proportion of their income, than poorer people 21. Equity of access to health care implies access to health care based on need rather than ability to pay. Because equity of access is difficult to measure, most studies employ equal use of health care as a proxy for equal access, as we do in our concluding discussion.

Economic theory would consider any reduction in use attributed to user charges as an improvement in allocative efficiency, regardless of the value or effectiveness of the health services foregone. However, many non-economists assume that, faced with user charges, rational consumers will forego the health services of least benefit to them (certainly those that are potentially harmful and perhaps those that are less effective). In this way, they argue, reduced use will not adversely affect health, but will help to contain costs and make health care more effective. Does this assumption hold? If patients do not have sufficient information to make rational choices, they may forego or delay useful treatment, perhaps damaging their health and leading to greater expenditure at a later date. Conversely, patients may turn to free (but more resource-intensive) forms of health care to avoid paying charges. The result might be higher rather than lower health care costs.

User charges can raise revenue for the health system if they are set low enough not to deter significant amounts of use. This policy argument is more prevalent in low-income countries, where public resources for health care may be severely limited or non-existent. Under such circumstances, drawing on private resources to ensure an adequate supply of drugs, for example, could lead to health improvement, particularly if poorer people are exempt from user charges. In high-income countries it is hard to see how private finance could be more efficient (in contributing to health improvement) than public finance. Unless user charges exempt high users of health care, they are really a form of tax on people in poor health.

Which of these arguments in favour of user charges seems most convincing in the case of prescription drug charges? Applying the allocative efficiency argument to a form of health care that requires a doctor's prescription only underlines its irrelevance to health policy. The focus on patient use seems misplaced when it is doctors who make the decision to prescribe drugs. If charges are not applied across all health services, the use of substitutes for prescription drugs (often emergency care) may increase costs. This leaves the revenue-raising argument, which, as we have noted, is barely plausible in high-income countries. Is it possible, then, that the real reason third party payers impose prescription drug charges is to contain their own budgets by shifting costs to patients? And is this why so many countries exempt particular groups of people from paying prescription drug charges, in the hope that cost shifting will not adversely affect health? By studying policy outcomes, we may be able to provide an indirect answer to these and other questions.

In this section we consider two types of expenditure: total prescription drug expenditure and patients' out-of-pocket expenditure. If patients are not particularly sensitive to changes in the price of prescription drugs, introducing or increasing user charges will have little effect on total prescription drug expenditure but will increase out-of-pocket spending on prescription drugs. Conversely, if patients are sensitive to changes in price, the impact on total expenditure will be greater, while the impact on out-of-pocket expenditure may be smaller as patients lower their use of prescription drugs.

Sixty-three papers examined the impact of cost sharing on total or out-of-pocket prescription drug expenditure using aggregate and non-aggregate data (see Tables 3 and 4). Aggregate data is defined as data collected at the macro-economic level so that individual- or household-specific information is not identifiable. The co-payment levels studied ranged from $0.50 to $35; co-insurance rates ranged from 0% to 95%. Most studies found that higher cost sharing lowered total prescription drug expenditure. Expenditure reductions ranged from a non-significant 0.04% of total prescription drug expenditure when moving from a two-tier to a three-tier formulary 23 to 58% of expenditure on ACE inhibitors when moving from a one-tier to a two-tier formulary 24. Variation in the magnitude of expenditure reductions was influenced by contextual factors such as the size of the increase in user charges, the type of drugs associated with user charges, and the population groups subject to user charges. Some studies calculated expenditure elasticities to measure the extent of the change in total prescription drug expenditure in response to changes in the level of cost sharing. Most expenditure elasticity estimates ranged from -0.29 to -0.06 (that is, a 10% increase in cost sharing would result in a 0.6% to 2.9% decrease in total expenditure). The largest expenditure elasticity estimate (-1.07), a clear outlier, was from a study that focused on older people in the United States without any form of protection from user charges (for example, through employer-sponsored additional coverage or Medicaid, the publicly financed health insurance program for the poor), which may explain why this group was relatively sensitive to price 25. Studies examining the impact of reference pricing found that it lowered prescription drug expenditure in the short term (generally in the first one or two years), but had little effect beyond this period.

In general, the literature also found that having any form of insurance coverage (as opposed to none) increased total prescription drug expenditure. However, a Canadian study found that the provision of free prescription drugs for individuals aged 65 and over did not increase total prescription drug expenditure 26. This result may be related to the fact that the author was only able to control for the unhealthiest respondents in the first year of the sample. Alternatively, additional insurance may have had no effect as people approaching this age were already using prescription drugs for chronic conditions. All except one 27 of the studies that examined the impact of user charges on patients' out-of-pocket expenditure on prescription drugs found that user charges increased patients' costs, while having additional voluntary health insurance coverage lowered their costs (see Table 5).

Thus, there is some evidence to suggest that cost sharing leads to slightly lower total expenditure or lower expenditure growth on prescription drugs and higher out-of-pocket expenditure for patients. The finding that patients are relatively insensitive to changes in the price of prescription drugs has important implications. Few of the studies we reviewed were of sufficient duration to permit assessment of long-term expenditure control. However, the reference price studies that took a slightly longer perspective found that reference pricing had little effect on expenditure beyond the first year or two, which suggests that user charges may not be relied upon to reduce pharmaceutical budgets in the long term. It also indicates that, rather than substantially lowering total expenditure on prescription drugs, user charges shift some prescription drug costs from third party payers to patients.

If user charges lower total expenditure on prescription drugs, they might also lower total expenditure on health care. Conversely, they could lead to a squeezed balloon effect, causing expenditure to rise in other parts of the pharmaceutical sector or health system. To assess the impact of prescription drug charges on total health care expenditure, we consider 23 papers that examined the relationship between prescription drug charges and the use of other forms of health care (a proxy indicator for expenditure) such as over the counter (OTC) drugs, doctor and outpatient visits, and inpatient and emergency care (see Table 6).

As the use of prescription drugs requires a doctor's prescription, in most cases we would expect prescription drug charges to result in lower use of doctors (to the extent that patients are sensitive to changes in price). Several studies found that user charges, reimbursement limits, and reference pricing did indeed lead to a reduction in doctor visits; three studies found that prescription drug charges had no effect on doctor visits; and one study found that they increased doctor visits. However, two of the three studies finding no relationship between prescription drug charges and doctor visits examined situations in which user charges were designed to encourage the use of lower-cost drugs through multi-tier formularies 28 or differential charges for generics and brand-name medications 29, rather than to lower the use of prescription drugs. In the third of the three studies, the insignificant effect on doctor visits may be explained by the fact that everyone in the study sample had experienced a heart attack, while those with lower incomes were afforded greater protection from prescription drug charges 30. Consequently, this group was less likely to be sensitive to changes in price and perhaps more likely to see the doctor for reasons other than to obtain a prescription. The positive result came from a sample of older people in Canada with rheumatoid arthritis 31. As health insurance in Canada fully covers doctor visits, it is not surprising that some patients would substitute physician care for prescription drugs.

Studies show mixed results for OTC drugs. In the RAND experiment, higher co-insurance rates lowered the probability of purchasing an OTC drug but, after controlling for this, cost sharing had no impact on OTC expenditure 32. Additional insurance coverage led to higher use of prescription drugs compared to OTC drugs in another US study 33, but had the opposite effect in a French study 34. The French result probably differed because additional health insurance in France covers more than just prescription drugs; and as doctors often recommend the use of OTC drugs, increased OTC drug use may have been prompted by increased doctor visits. Having a limit on the number of free prescriptions an individual is allowed per month (a policy most often associated with Medicaid in the United States) positively influenced the quantity of OTC drugs used 35.

The results for outpatient, inpatient, and emergency care are much more consistent. As expected, user charges designed to encourage patients to choose lower-cost drugs had no significant effect on the use of inpatient or emergency care 2836. Otherwise, with the exception of four studies, there was generally a positive relationship between cost sharing and outpatient, inpatient, and emergency care. Studies also found that prescription limits increased the frequency of partial hospitalisation 37 and nursing home admissions 38 and the use of emergency mental health services 37. Two of the studies that found no effect were based on chronically ill patients 3038. Soumerai et al. 38 also suggest that their insignificant result for inpatient admissions might be due to the fact that the outcome variable they used (time to first hospital admission) would not highlight repeat hospital visits.

These findings reveal two things. First, prescription drug charges are unlikely to lower total health care expenditure and may in fact increase spending overall. Although a decline in the use of services that complement prescription drugs (doctor visits) may lead to cost savings, any savings are likely to be outweighed by increased use of the highly resource-intensive services that substitute for prescription drugs (inpatient, emergency and long-term care). Second, the design of a cost sharing policy can mitigate this potentially explosive effect on total health care expenditure. Policies that give patients incentives to switch to lower-cost drugs and policies that protect low-income groups may prevent inefficient patterns of health care use which, while more accessible to patients, are more costly to the health system.

In this section we return to the question of total prescription drug expenditure and consider whether expenditure reductions resulting from user charges are achieved through reductions in the price of prescription drugs or reductions in the quantity of prescription drugs consumed. From a policy perspective, lowering expenditure through reductions in price would be preferable because, as the previous sections have shown, reductions in quantity can have unwelcome consequences for total health care costs and for equity (unless user charges exclusively reduce the use of unnecessary or ineffective prescription drugs).