Community Support

the Evaluation Center@HSRI Toolkit

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

II. DETERMINING THE BEST APPROACH TO ESTIMATING
COSTS: BALANCING LEVEL OF PRECISION
WITH LEVEL OF EFFORT

Every increment of precision requires additional resources but varies in its usefulness. Thus, the level of precision needed in unit cost estimation must be considered in light of how the information is to be used. This is similar to the management of our own finances which is governed by decisions we make every day about how and on what to spend our money. Sometimes it is sufficient to make "back-of-the-envelope" estimates. Other times we make precise calculations about all our current and anticipated expenditures. In doing economic evaluations, the questions to ask are: Do we need to know precisely the costs for each service used for each person and if not, what magnitude of error are we (or others) willing to live with? What is the tolerable range of confidence limits around the estimate?

For example, if cost estimates are going to be used to predict future expenditures for large populations (such as future expenditures after health care reform), then even low levels of error in per unit costs for commonly used services become large errors in the prediction of population expenditures. On the other hand, in a small study, if certain services rarely occur, (e.g., several visits by a few clients to a court clinic, in a sample of 200 clients) even if the estimate is off by as much as 20% of the cost per visit, the total per person costs in this sample will hardly be affected by this imprecision. The level of precision required should be decided before embarking on the work necessary to develop the per unit costs.

Three guidelines for making decisions about what level of effort will provide the right level of precision in the per unit cost determination are:

What level of error are you (the evaluator) willing to accept?
What level of error are the important audiences for your evaluation willing to accept? Can you confidently defend yourself against charges that your findings would be substantively different if you had been more precise?
What level of effort will the available time and resources permit?

Estimating the precision of a given estimate assumes some knowledge of just what resources are needed to produce a unit of service. Wolff has written that "how an estimate is derived determines its quality, reliability, accuracy, and comparability. An estimate ultimately depends on a litany of implicit and explicit assumptions, random and nonrandom data collection errors, exactness of the data sources, breadth and depth of omissions and oversights, and ability to measure true economic costs." (page 173, Wolff and Helminiak, 1993).

A major part of practical cost estimation is deciding what services to include when the evaluator wishes to estimate the per person cost of treatments within a multi-service health plan or service system. The criteria for deciding what services to include are related to the guidelines that we offered above:

What level of completeness are you willing to accept? Can you confidently defend yourself against charges that your findings would be substantively different if you had been more comprehensive?
What level of completeness is the audience for your evaluation willing to accept?
What level of effort will the available time and resources permit?

The advantage of setting the scope as broadly as possible is that it provides evidence of cost shifting from one provider to another or unanticipated cost consequences outside the scope of your study. For example, in our recent study of formerly homeless adults, we chose to include psychiatric and substance abuse treatment, housing, and criminal justice costs. It would have been ideal to include, in addition, the costs to family of helping to care for their family member but we made our decision based on our estimate of the magnitude of family costs, the available resources and the time we had to complete the study. This addition would have been costly, and few of the subjects had any contact with family members. Adding these expenditures to the study would not have changed our findings (we argued) and the extra time it would have taken to determine these costs would have been considerable.

Table of Contents | Previous | Next