Community Support

the Evaluation Center@HSRI Toolkit

Estimating Per Unit Treatment Costs for Mental Health and
Substance Abuse Programs

VI. CARRYING OUT COST EVALUATIONS (STEP 4)

Conducting a cost evaluation involves more than just attaching per unit costs to service use data. This section discusses four important economic concepts related to cost evaluation: opportunity costs, capital costs, discount rates and social costs. We briefly describe them and offer explanations of why these concepts are important. Fuller explanations are available in textbooks on cost-effectiveness evaluation and we believe that it is important for evaluators doing cost studies to know these concepts. We recommend that anyone doing a cost evaluation be familiar with these terms and refer to economic texts and colleagues with background in economics for a more complete understanding. Two useful books on conducting cost evaluations are now available:

Gold MR, Seigel JE, Russell LB, and Weinstein WC. Cost-Effectiveness in Health and Medicine 1996, New York: Oxford University Press, and

Hargreaves WA, Shumway M, Hu T, Cuffel B. Cost-Outcome Methods for Mental Health 1998, San Diego: Academic Press.

1. Guideline #1: The most precise estimate of a per unit cost is the opportunity cost.

It is important to note that the practical approach described in this toolkit will give us only an approximation of the "true" cost of a unit of service as economists think of it. From an economist's point of view, what is known as "opportunity cost" is the standard for estimating the true monetary costs of any resource. Opportunity cost is the economic concept of taking into account the value of a resource in its next best use.

The opportunity cost of a unit of service would be based on the value of the resources directly involved in producing the service in their next best use. For example, the use of public land may not require any expenditures. Its true cost, however, is not zero because that land could be rented out or otherwise used to generate revenue. The accounting cost of the land is zero, but its opportunity cost is its value in its best alternative use-a measure of the value of the opportunity that is lost by using it for the program being evaluated. When measuring the opportunity costs of a program or project, economists start with the market price. If this price does exist, they ask whether the market price is a reasonable measure of the opportunity cost.

If a market price for a resource does not exist, they estimate what is called a shadow price. To determine the shadow price, we would need to ask: what is the value in its best alternative use-the use to which it might otherwise be put? One method of measuring a shadow price is to conduct a survey asking persons what they would be willing to pay for a particular service. Alternatively, if we do not trust the results of such surveys (talk is cheap), we might look for evidence of what others have been willing to pay. For example, in the case determining the cost of family burden, how have other people faced with family caregiving responsibilities in fact responded? Did they quit their jobs, or reduce their hours? Did they hire a caregiver? The cost they have thus actually chosen to incur in response to their caregiving responsibilities is an alternative measure of the opportunity cost.

In mental health care, the "true" cost of delivering a particular unit of service may be very different from what providers charge for that treatment or what an insurance company pays for that treatment. For example, providers in a multiservice agency often price one service so that it subsidizes another service. Additionally, different types of persons can also be charged different amounts. Hospitals, for example, provide care that is "free" to some consumers, but of course not free to the hospital. Free and discounted care must be paid for by someone, and such costs are usually passed on to those who can afford to pay out-of-pocket or who have insurance.

When measuring the opportunity costs of a program or project, economists start with the market price. If it does exist, they ask whether the market price is a reasonable measure of the opportunity cost. If it does not exist, they estimate what is called a shadow price. In a perfect market, charges and costs would be equivalent, but this is certainly not true in health care. We urge the readers to avoid using charges for mental health services (the market price) which are distorted. Market prices can deviate from opportunity costs for a variety of reasons. The most significant example of this is associated with the practice in multiservice programs that price one service so that it subsidizes another service.

2. Guideline #2: All per unit cost estimates should include capital costs.

Whether the building in which a program takes place is rented or owned by the agency, the facility capital cost must be included in the per unit cost for any service. Capital costs must also be allocated across direct services. Typically, facilities have two types of capital costs associated with them: the value of the facility and land (which may show up as an expenditure in the accounts as depreciation for the building over some period of time) and interest associated with the repayment of a mortgage loan. When the property is privately owned, there are several ways to determine value: (1) cost accounting-- if it includes the depreciation and the debt service (typical of hospital costs reports), (2) the actual cost of the land and the construction-- in a new building and (3) the market rent paid by the agency or program-- which can be decomposed to estimate the amortized value.

Amortization of the value of a building means spreading this value over the presumed life of the building. Typically, buildings are amortized over 20 years⁸. For example, if a building was valued at $1,000,000 and we were to simply divide the total value by 20, we would fail to take into account the opportunity cost of the capital invested in the building. Thus, amortizing the value over twenty years means thinking what might be a reasonable rate of return on the investment. If we use 5% as the return on investment, then the annual amortized value for one year will be $52,500. Then, using the square feet of the building occupied by the particular service of interest, determine the share of this annual value that is associated with this service. Finally, add the annual value to the expenditure for the program for that year.

When the facility is publicly owned, determining the value raises special problems. There is unlikely to be any debt service associated with a particular public building, even if the state or local level of government floated bonds to cover the capital expenditures. Furthermore, there is no resale of public facilities that allows us to value capital periodically. In this situation, it is still possible to estimate the capital costs by determining the value of the building using indirect methods.

In a study some years ago (Cannon, McGuire and Dickey, 1985), the capital costs of a public community mental health center were estimated by using four different sources of valuation to determine the upper and lower bounds of valuation. These valuations were first converted into per square foot rental prices and these then converted to rental rates using four different rates of return (3% to 9%). The sources of information about the value of the facility came from four informants: a neighboring hospital that had just purchased "comparable" property adjacent to the building under study, the state appraiser who provided a written estimate, a real estate appraiser who provided an estimated purchase price for the building and land as commercial real estate (the opportunity cost), and a local architectural firm that had just received written bids for replacement of the building. When these estimated valuations were converted to rents, they ranged from $24.92 to $117.64 per square foot. The difference between the lowest and the highest estimate was very large indeed. These large differences illustrate the care with which such valuations must proceed.

One well-known cost-effectiveness study used a relatively high discount rate (see below) and found that the capital costs accounted for about 30% of the inpatient per diem. A small adjustment downward in the discount rate would have changed the principle findings of the study (Rosenheck, Frisman and Neale, 1994). If capital costs are not available and you calculate them yourself, we recommend reporting the results from a sensitivity analyses so that the reader can judge for herself whether the rate actually used seems appropriate (Cannon, et al, 1985).

3. Guideline #3: Projected expenditures need to be discounted.

The concept of discounting is important when cost evaluations include projections into the future, as studies with policy implications often do. To determine the costs of the program over the first five years, you cannot simply sum the annual costs because a cost of $1 incurred in a later year is equivalent to a cost of less than $1 incurred today. The reason is easy to see: if you will need to spend $1 next year, you can put, say 95 cents in the bank today, receive 5 cents interest over the next year, and have the dollar you need one year from now. Thus the future costs need to be discounted to make them comparable to current costs.

In order to discount future costs, one must choose a discount rate, which is the mirror image of an interest rate. In the example above, we assumed that 95 cents would earn five cents in interest over one year. This implies an interest rate of 100 x 5/95, or just over 5 percent. Why choose that particular rate?

In costing certain programs, we generally assume that costs will be incurred with certainty. The discount rate to use is therefore the amount of interest one would earn on a very safe investment⁹, such as Treasury bills. This interest rate is at the low end of interest rates available on investments , because the investment is virtually risk-free (the US Government is not expected to default on its loans). If one used a higher interest rate, then one would in effect be assuming that the costs might not have to be incurred at all. The chance of losing part or all of the initial investment that arises when one invests in, say, common stocks, would reflect the chance of not having to incur the costs. We prefer to conservatively assume the costs will be incurred, and so use the low rate of interest on safe investments.

But while we know the rate of interest on safe investments today, we cannot know that rate for future years. Because of this uncertainty, the choice of discount rate is somewhat arbitrary. A simple and popular way to deal with this uncertainty is to do a sensitivity analysis. In a sensitivity analysis, we see how the results change when we change our assumptions. Suppose we start (arbitrarily) with the average rate of indicator of future rates, 5%. Then we see what happens to the total discounted cost if we use, instead, a discount rate of 3% or 7%. If we report the range of results that come from using discount rates ranging from 3% to 7%, the reader's level of confidence in the results should increase.

You may wonder how many years in the future to go. This is a subjective decision; it depends on the anticipated duration of the program. One final note: if some assets purchased at the beginning of the program, such as buildings, have some value at the end of the program, that estimated resale value must also be discounted and subtracted from the total discounted cost.

For example, a five year program for housing a population of homeless mentally ill is expected to incur costs of $300,000 the first year, $100,000 during years 2 - 4, and a net cost (after resale of the house) of $10,000 in year 5. Using a discount rate of 5%, the present value of the cost (PVC) is:

PVC = $300,000 + $100,000 x (1 - .05) + $100,000 x (1 - .05)² + $100,000 x (1 - .05)³ + $10,000 x (1 - .05)⁴

PVC = $579,132.56

This means that, in today's dollars, the cost of the program over five years is $579,132.56. If we are comparing two programs, we need to be able to make the comparison equivalent in today's dollars to take into account that programs may have different time-spans.

4. Guideline #4: Cost evaluations should include social costs.

Social costs refer to resources used to produce services, like police and social services, beyond those mental health services under study. It also includes "costs" of not working because of a psychiatric disability or the time spent by family members who may provide housing or other care to a consumer. This approach leads to the broadest possible scope and implies that changes in the resource use in one area will effect resources used in other, perhaps peripheral areas.

Ideally, effort should be made to take into account all the costs (and the benefits) that result from a particular change in policy. Thus, a study of the social costs of managed care might include the costs of police services and the increased costs of worker compensation claims if access to hospitalization is limited to the most disturbed mentally ill. Jails might have to replace some of the hospital functions or nursing staff may suffer more accidents treating individuals with more serious behavioral problems.

The resources and time required for estimating social costs can be considerable, depending on the method used. In addition, estimating social costs requires considerable expertise. We recommend that the reader turn to an economist to determine morbidity and mortality costs, (e.g., the cost of iatrogenic illness or death as a result of treatment).

Some mental health services, particularly community based ones, involve labor or supplies that are provided by families or other donors. These inputs have costs, although families or donors may not require payment. These are sometimes referred to as "contributed costs." When resources permit, evaluators may wish to estimate these contributed costs. These costs can be examined separately or combined with the costs of other program inputs.

5. Guideline #5: Cost variation should be explained.

If you are familiar with regression analytic techniques, then it is possible to use this approach to understand statistically what accounts for variation in the total annual per client cost to provide services. There are groups of factors that might affect the cost of care (in addition to the amount of service provided): personal and clinical characteristics of the individuals, the characteristics of the providers of care, the health care environment and the specific program or treatment intervention (e.g., managed care). Examples of individual client socio-demographic characteristics are age, sex, ethnicity and income level. Provider variables can include characteristics of individuals or facilities. Whether or not a hospital is a teaching hospital, has been shown to be a factor in cost. Some studies have examined training and experience of professionals. Variables that might be grouped under health care environment include the number of physicians or hospital beds per capita. The intervention, which is the independent variable, is a dummy variable in the multivariate estimation model (1 = received the intervention or 0 = did not receive the intervention).

Take for example the introduction of managed care. Clients are randomly assigned to a managed care system or the usual conventional community-based service system. One evaluation question is whether or not the managed care intervention costs more or less, on average, than conventional care. Once the per unit costs are determined, they are used to calculate the cost of services used by each individual. Then these costs can be summed (as in table 9) for each person. The regression analyses will estimate the contribution that the managed care intervention makes to variation in per client annual treatment costs after other factors known to influence costs are controlled.

6. Guideline #6: Cost effectiveness studies are more helpful to policymakers than a study of costs alone.

Evaluators are likely to be asked to compare changes in service costs relative to changes in service outcomes. One simple approach to doing so is to compute the change in service (and social) costs and the change in outcomes after an intervention. If the per person costs are lower and the outcomes (i.e., effectiveness) are the same or higher, then no more analysis is required because it is assumed that lower costs and better outcomes are desired. However, if costs go up and outcomes improve, then cost-effectiveness ratios can be calculated to help determine the "best" treatment relative to the usual and customary care (the typical comparison). A good example of this type of study was carried out by Lave et al (1998). They suggest that rather than using the simple approach described above, more precise estimates can be had from using regression models that help to determine how costs and outcomes vary across treatment groups. Their method is quite sophisticated, but bears attention in an era when improved medical technology often leads to better outcomes but at increased costs.

No effort is made to convert the outcomes into dollars. Converting any type of outcome to dollars is difficult and for mental health treatment outcomes, it may be very difficult. Nevertheless, cost-effectiveness ratios can be obtained which provide the dollar amount needed to raise an "outcome score" by one unit. A more complete discussion of these analytic techniques can be found in texts such as Cost-Effectiveness in Health and Medicine, Edited by Gold, Siegel, Russell and Weinstein; Oxford Press, 1996 or in Cost-Outcome Methods for Mental Health by Hargreaves, Shumway, Hu and Cuffel; Academic Press, 1998. These two texts also discuss, more briefly, cost-benefit analyses which compares changes in the costs of service in relation to changes in an outcome measure when these changes are translated into monetary values, such as wages earned or taxes paid.

7. Guideline #7: Cost studies have limitations.

It is the convention to use the term "cost estimates", rather than simply "costs." This is a useful reminder of the many limitations in developing precise "true" costs. One limitation is our inability to truly capture all the resources used in providing a treatment program. For example, volunteers may be an important component in a particular service, but the cost of a volunteer's hour may be very difficult to calculate because the time spent may not be documented or the value of the contribution may be difficult to assess.

Another example is the practice of using consumers to provide services to other consumers, but the consumer may not be paid in cash, but might have some other form of compensation, such as free room and board. In this case, the value of the room and board per day would be the starting point of calculating the per hour cost of the service provided by that consumer. Benefits, especially intangible ones, should be used to balance costs, but these are even more difficult to estimate, even though they may be very real. Other limitations to keep in mind are whether the sample is large enough to provide reasonably stable estimates, or whether cost estimates are truly generalizable to the population sampled.

Despite these limitations, we encourage evaluators to find ways to add to the cost literature, literature that would benefit from more well done cost studies. We hope that this toolkit format improves access to this kind of work, even though we must acknowledge that it opens the door to many questions best answered by more detailed discussions which are likely to be found in accounting and economics textbooks. Perhaps this short "introductory" course will lead to further study in these matters. Despite these shortcomings, we hope that we have given you a jump-start into this important and interesting field of work.

Table of Contents | Previous | Next