Dissertation Title:  "Measuring Productivity and Quality in Mental Health Care"The treatment of mental illness has undergone significant organizational and technological change in the past decade. Spending continued to grow at rates above changes in general price in the 1990s. Are increased mental health expenditures due to price increases, quantity increases or some combination? What has been the effect of these recent changes on quality of care? This thesis addresses these questions. Mental health carve-outs separate the insurance risk of mental health care from that of general medical care and are increasingly common in both the public and private sector.

In the first chapter I focus on the quality of treatment for depression in one mental health carve-out. Comparing treatment patterns in the pre- and post-carve-out period, I find that although quality declined over the course of the time period studied, implementation of the carve-out was associated with an increased probability of receiving guideline-level treatment. Although broad trends in medical spending have received widespread attention from policymakers, very little attention has focused on the components of those changes.

In the second chapter, colleagues and I make use of results from clinical literature, and identify therapeutically similar treatment bundles that are then linked and weighted to construct price indexes for the treatment of major depression. In doing so, we construct medical price indexes that deal with prices of treatment episodes rather than discrete inputs, that are based on transaction prices, that take quality changes and expected outcomes into account, and that employ current, time-varying expenditure weights in the aggregation computations. We find that over the 1991-1995 time period this treatment price index has hardly changed, remaining at 1.00 or falling slightly to around 0.97. This index grows considerably less rapidly than the various official PPIs. A hedonic approach to price index measurement yields broadly similar results.

In the third chapter I reported on additional analysis that indicate this bias is due to either between-item-strata substitution (63 percent) or failure to account for quality change (37 percent). This suggests recent modifications to BLS methodology will have little effect on the bias in official price indexes in so far as treatment for depression is a representative condition.