We gathered data on state Medicaid prior authorization policies for biologic DMARDs used to treat RA and other rheumatic diseases. Of the 50 states studied, we identified 32 states that had implemented, or planned to implement, such policies. There was significant heterogeneity in the drugs that were included in the prior authorization policies and states also varied widely in the criteria required, both in terms of the amount of detail requested and in terms of the clarity of how authorization was determined.
Complete Medicaid utilization data allowed us to perform quantitative data analyses for adalimumab and etanercept. Combined use of these two drugs in Medicaid increased sharply between 1999 (etanercept alone) and 2005, both in absolute terms and as a proportion of all DMARDs. We found that states with prior authorization programs in place at the beginning of the study period had relatively low use of the targeted biologic DMARDs initially, with a sharp increase over the years studied. Many of the states with the highest levels of use of etanercept in 1999 went on to implement prior authorization policies between 2000 and 2004 and had a smaller increase in use of biologic DMARDs compared to the other states. States with the highest levels of use in 2005 were those that were in the process of implementing prior authorization requirements.
Previous studies have evaluated prior authorization for a variety of medication classes. Evaluation of Medicaid policies for selective cyclooxygenase-2 inhibitors (coxibs) found wide variability in the degree to which prior authorization criteria adhered to clinical evidence.4
Likewise, studies of psychiatric medications found considerable heterogeneity in how states approached prior authorization for these medications.7, 8
The state-to-state variability that we observed in prior authorization policies for biologic DMARDs is consistent with these prior findings.
Earlier research has also examined the quantitative impact of Medicaid prior authorization policies. A relatively consistent reduction in use of the targeted medications has been observed in studies of coxibs and other anti-inflammatories6, 9, 10
and angiotensin-receptor blockers.5
In both of those cases, the drug classes studied have straightforward substitution patterns that can be anticipated and built into policies (using non-selective NSAIDs instead of coxibs; using angiotensin-converting enzyme inhibitors instead of angiotensin-receptor blockers). Management of rheumatic diseases is more difficult and the medication choices are more complicated. Patients may receive a biologic agent together with a synthetic DMARD, after failing one or more non-biologic DMARDs, or without trying a non-biologic DMARD and these subtleties cannot easily be measured in our data. In addition, the utilization data for biologic DMARDs do not include important and frequently used agents such as infliximab. Some of the non-biologic DMARDs may be used for non-rheumatologic indications, and their inclusion in the denominator of our outcome measure could make the measure less precise. The variability that we see in the drug utilization data may both limit our ability to measure a precise impact of prior authorization policy and also reflects the difficulty of applying and implementing these kinds of policies to complicated conditions such as rheumatic diseases. Nevertheless, interpretation of our results does provide some insight into these policies.
Our data suggest that states implement prior authorization for biologic DMARDs when the use of and spending on these agents reaches a high level, as can be seen in the last two sections of . States with the highest levels of etanercept use in 1999 implemented prior authorization policies for etanercept and adalimumab in the following years and these states did have a lower rate of increase than other groups of states. The seven states with the highest levels of adalimumab and etanercept use in 2005 were in the process of implementing prior authorization policies at that time. On the other hand, it is notable that the two states with prior authorization in place as of early 1999 had the lowest levels of biologic DMARD use in the first year, but showed a sharp increase during the study period. By 2005, the level of biologic DMARD use in these two states was similar to the level in the nine states that implemented prior authorization between 2000 and 2004. Results for these eleven states suggest that the initial impact of prior authorization may blunt the growth in the use of targeted agents, but the rapid growth in the two states with early prior authorization raises question about the sustainability of these effects. The interrupted time series analyses provide additional support for this interpretation, since we observed an initial decrease in use of biologic DMARDs when prior authorization was implemented that was offset by a subsequent increase in use of biologic DMARDs over time. Most of the time-series results were not statistically significant, so we must interpret them with caution.
There are limitations that must be considered in interpreting this study. We obtained data on prior authorization policies for biologic DMARDs by contacting state Medicaid agencies directly. It is possible that the information provided was incomplete, and it may not have reflected changes in the policies over time or subtle aspects of the policies not included in the written documents. Local officials may have discretion in approving individual requests when the criteria are ambiguous, or the actual implementation of the policy may differ from the written rules, and these discretionary elements would not be captured in our data. States may use other policy tools besides prior authorization, such as drug utilization review and dispensing limits, and the impact of these interventions could alter our results. The data available for research contain the actual amount paid by state Medicaid programs, while information on pricing policy for Medicaid programs, any discounts negotiated by programs, and the amount rebated to states by manufacturers are not publicly available. These additional factors may affect spending more than interventions such as prior authorization, but we cannot determine that from our analyses.
The data for the quantitative analyses did not include infused medications and our inability to evaluate these agents, especially infliximab, may limit the generalizability of these findings. If etanercept or adalimumab are administered in doctor’s office the billing may be similar to that for infused medications and may not be fully captured. The drug use data are aggregated at the state level and thus cannot capture clinical details of individual drug use patterns or prior authorization decisions. The non-biologic agents that we used as a comparator group can be used to treat a variety of inflammatory and rheumatic diseases, and changes in the prescribing patterns for these medications may impair our ability to measure true changes in the use of biologic DMARDs. Biologic and non-biologic DMARDs may be used together in treating rheumatic disease, so prior authorization policies do not necessarily target a simple trade-off between biologic and non-biologic DMARDs. Since we only have data on overall drug use, and not on population numbers or characteristics, we used total volume of DMARD use as the denominator of our outcome measure to adjust for changes in the Medicaid population, but this technique is imprecise. This limitation may cause us to underestimate the impact of prior authorization policy on biologic DMARD prescribing for specific conditions. If prior authorization policies are developed in response to increasing drug spending, then the presence of a policy may be endogenous, introducing bias into our effect estimates. These aggregate data cannot provide insight into the impact of these policies on clinical endpoints, just on drug use. Future research using patient-level data would be helpful for illuminating the effect of these policies on patterns of care and patient outcomes.
Our results have implications for prescription drug reimbursement policy, both for Medicaid and for other programs. In terms of Medicaid programs, although the clinical decisions about use of biologic DMARDs for inflammatory diseases are unquestionably complex, the heterogeneity across states in authorization criteria reveals limitations in policy development. It is not clear how state agencies determine which clinical factors are included in the prior authorization rules or how closely these rules adhere to clinical evidence. This finding may be a particular concern with the transition of many patients to the Medicare part D drug benefit in 2006. If the heterogeneity that we observe across Medicaid programs is also found across various Medicare part D plans, clinicians treating patients with RA and other inflammatory diseases are likely to face many different sets of criteria when prescribing biologic DMARDs, further complicating care for these already vulnerable patients. The transition to Medicare Part D may also complicate the attempts of state Medicaid programs to control prescribing of targeted medications.
In conclusion, we found wide variation in state Medicaid prior authorization policies for biologic DMARDs. Quantitative data analyses suggested that these policies may have a sentinel effect on use of biologic DMARDs, but we could not demonstrate a consistently measurable effect as has been seen for some other drug classes. Policy-makers must weight the costs imposed by these policies in terms of professional time and patient delays of therapy against potential savings on these expensive medications. Further examination of these policy approaches including their impact on patterns of care and patient outcomes will be critical to foster the development of more rational policies in the future, for Medicaid and for all drug insurance programs.