Background: Promoters of new medications often argue that using newer drug can reduce use of non-drug medical services and therefore reduce total healthcare spending. This cost-offset argument is plausible both in theory and in practice, but rigorous research on specific drugs or drug categories is needed to make targeted and efficient policy and management decisions.
Aims of the study: I examined the drug-offset hypothesis for bipolar disorder, an important yet under-studied clinical condition where effective medication treatments can service as substitutes for non-drug medical treatments. I compared two first line long-term treatments, a new atypical antipsychotic medication, olanzapine, and a traditional mood stabilizer, lithium.
Methods: I used private sector insurance claims data collected from a nationally representative sample of U.S. health plans between January 1998 and December 2001. I first selected a cohort of patients with bipolar disorder who were continuously enrolled for at least two years. I then used a propensity-score method to match individuals taking each drug on observed variables that are known to affect medication choices. The central challenge for estimation is that drug treatments are not randomly assigned among patients with bipolar disorder. To identify a causal link between choice of drugs and non-drug medical spending, I employed three different advanced econometrics techniques to assess the robustness of findings; namely interrupted time series, differencing strategies, and an instrumental variables approach.
Results: I found that compared to similar lithium users, olanzapine users spent approximately $330 more on monthly average non-drug medical services during the first year after initiation of drug treatment. The higher spending for olanzapine users was accounted for by both higher rates of re-hospitalization and more outpatient visits. In addition, olanzapine cost $153 per month while lithium cost $16 per month. Including the direct cost of the drugs, compared to similar patients taking lithium, patients with bipolar disorder taking olanzapine spent $5,600 more annually on health care services.
Discussion: These findings do not support the hypothesis that new drugs "pay for themselves" by reducing the need for other health care services in the case of olanzpine for bipolar disorder. This does not mean that the new drug is not "cost-effective" because increased "benefits" associated with the drug in terms of the improved quality of life may be worth the increased costs. However the findings do indicate that "cost-offsets" must be measured and not taken for granted. Incorporating such drug-offset evidence into policy and business decisions can facilitate appropriate clinical practices and improve efficiency of resource allocation. The methods used in this study to test for cost-offsets can be applied to other clinical areas and drug classes.