Price Indexes for the Treatment of Depression
"From 1991-5, when broad BLS indexes indicated a 15-25 percent price increase in overall health care costs, their indexes suggest a 20-30 percent price decrease in the cost of treating major depression."
Recent upticks in the Bureau of Labor Statistics (BLS) health care price indexes have rekindled worries about rising health care costs. In Price Indexes for the Treatment of Depression (NBER Working Paper No. 6417) , Richard Frank, Ernst Berndt, and Susan Busch compare major BLS health care price indexes with newly calculated indexes for the cost of outpatient treatment of an episode of major depression. In the first of a series of analyses, they employ a definition of depression that corresponds to treatment in clinical research. They show that from 1991-5, when broad BLS indexes indicated a 15-25 percent price increase in overall health care costs, their indexes suggest a 20-30 percent price decrease in the cost of treating major depression.
Current health care policy seeks to control aggregate spending. But if treatment costs have declined, increases in spending result from increases in the quantity of care, and controlling overall spending would reduce the volume of care delivered. The authors conclude that if further work supports their conclusions and shows that the cost of treating episodes of disease has indeed fallen, relying on the broad BLS indexes to formulate health care policy could mislead policy decisions.
For practical reasons, the BLS medical care price indexes collect data on the prices of health care inputs, such as the average cost of a single hospital visit, office visit, or drug. In contrast, Frank and his coauthors calculate price indexes for an episode of treatment of a particular illness. By focusing on the cost of treating a particular condition, the authors can construct price indexes that reflect price changes brought about by changes in organizational form, technique, and technology.
The data used to construct the indexes come from medical claims filed by 428,168 people covered by four large self-insured employers from 1991 through 1995. Treatments were defined as episodes of outpatient care using methods that had been tested in clinical trials and that conformed to the American Psychiatric Association and the Agency for Health Care Policy and Research treatment guidelines for major depression. All treatments conforming to the guidelines were assumed to produce results of equal quality. Five years of data on each of the five types of treatments were used to construct the indexes. The authors produced 7 different types of price index for both the total payments made to providers and for consumers' out-of-pocket costs.
Depending on the treatment used, average 1991 total payments to providers varied from $254 to $924. Out-of-pocket patient costs varied from $11 to $151. By 1995, provider costs ranged from $117 to $646 and average out-of-pocket consumer costs varied from $21 to $95. All of the indexes constructed from these data indicated price declines of roughly 30 percent for payments to providers and 22 to 30 percent for consumer out-of-pocket costs. Similar results were obtained when the definition of an episode of treatment was relaxed to include claims that were merely close to the guideline criteria.
The authors caution that several important issues need to be addressed before such indexes can be implemented by public statistical agencies. In addition to the problem of accounting for quality differences across different treatments with similar effectiveness in clinical trials, retrospective insurance claim data are typically available only with a nine month to two year time lag. Finally, most insured consumers of health care prepay the non-deductible and non-copayment components of their health care costs by buying insurance. How to incorporate that into a "readily interpretable" medical consumer price index remains a problem with, as yet, no practical solution.
-- Linda Gorman
The Digest is not copyrighted and may be reproduced freely with appropriate attribution of source.