Online Program

Impact of Prior Authorization on Quality, Cost and Outcomes Among Adults with Mental Illness in Medicaid
*Alyce Sophia Adams, Kaiser Permanente Division of Research 
Stephen B. Soumerai, Harvard Medical School and Harvard Pilgrim Health Care Institute 

Keywords: Medicaid, mental health, coverage policy

Rising pharmaceutical costs, particularly for psychotropic medications, have prompted many Medicaid and Medicare Part D prescription drug programs to institute prior authorization before dispensing of specific drugs to discourage the use of more costly medications. Yet, little is known about the effectiveness and safety of these policies, particularly for the mentally ill. We recently conducted a series of evaluations of Medicaid policies which required prior authorization before dispensing of select non-preferred psychotropic medications to patients who were not previously on therapy. We hypothesized that the policy may have resulted in lower rates of psychotropic medication use and adverse health consequences if: 1) newly treated patients did not respond as well to the preferred medication; 2) patients already established on therapy were forced to switch to a new medication; and/or 3) administrative hassles associated with these policies discouraged physicians from writing or patients from filling new prescriptions. Using an interrupted time series with comparison series design, we evaluated the intended and unintended impacts of the policy. We used time series models to assess population-level effects and patient-level Cox proportional hazards models and generalized estimating equations to directly link changes in prescribing due to the policy with individual-level behavior changes and outcomes. Among newly treated patients with severe mental illness (i.e., schizophrenia, bipolar disorder) the policy was associated with higher rates of disruptions in therapy. For the cohort of patients with bipolar disorder, there was also a statistically significant decline in the proportion of patients initiating therapy over time. Among patients with severe mental illness, there was a small cost savings associated with PA. Among newly treated patients with depression, we found no impact of the policy on disruptions in antidepressant therapy or adverse psychiatric events. We did, however, observe small declines in the proportion of patients initiating antidepressants each month and a short term increase in switching among those using antidepressant pre-policy. Our findings suggest that for patients with severe mental illness, the increased disruptions in therapy may not be worth the cost savings of PA programs. Use of PA to reduce antidepressant spending may require ongoing evaluation, particularly during the initial stages of policy implementation.