|
An Evaluation Of Oregon’s Evidence-Based Practitioner-Managed Prescription Drug Plan
This paper describes Oregon’s implementation of its publicly developed, evidence-based, Practitioner-Managed Prescription Drug Plan (PMPDP). Oregon’s PMPDP was initially self-enforced with a dispense as written (DAW) exception process, followed by an educational prior authorization (soft PA) method, and finally no active enforcement. Market-share trends indicate that the educational prior authorization process was most effective at increasing the use of preferred agents. Pharmacy costs decreased 9.1 percent and 17.7 percent after implementation of the DAW and soft PA policies, respectively. Data from nonenforced PMPDP classes showed no change; this suggests the need for effective methods to encourage PMPDP compliance.
SHARP DECLINES IN REVENUES combined with rapidly rising health care spending have forced many states to adopt far-reaching measures to sustain their Medicaid programs.1 Prescription drugs are a key driver of health care costs for state Medicaid programs and a frequent target for cost-controlling programs. In response to these pressures, in August 2001 Oregon enacted Senate Bill (SB) 819, which mandated development of the Practitioner-Managed Prescription Drug Plan (PMPDP) for the fee-for-service (FFS) Medicaid program.2 The intent of SB 819 was to provide Oregonians "the most effective drugs in the most cost-effective manner" by developing a process that systematically identified the most effective drugs within a class. Preferred drug lists (PDLs) are common among state Medicaid programs; however, Oregon’s approach to both establishing and implementing its list was unique. Here we describe the methods used to develop and implement the PMPDP, examine prescription drug use after it took effect, and estimate cost savings attributable to this policy. PMPDP list development process. Establishment of the Oregon PMPDP was predicated on a several-step process outlined in the original legislation. First, SB 819 asserted that Oregon Medicaid enrollees should have access to the most effective prescription drugs for their clinical conditions. Effectiveness was to be determined by health practitioners, who are informed by the latest peer-reviewed research. Finally, cost was to be considered only after the relative clinical effectiveness was established and then managed through market competition. Comparative effectiveness was evaluated using a systematic process; employing established methods of evidence-based medicine; and open to input from the public, stakeholders, and the pharmaceutical industry.3 The use of an evidence-based approach for policy development is not novel; however, the scope and transparency of Oregon’s process makes it unique among both public and private health care payers.4 The evidence-based review process has been reviewed in detail elsewhere.5 The first step in the process developed key effectiveness and safety questions. To build community consensus about what effectiveness and safety outcomes were important, volunteer committees of local health care practitioners and members of the public were convened to develop the questions. Clinicians were restricted from participating in the subcommittees if they had major financial conflicts of interest. Key questions varied by drug class but generally focused on the comparative effectiveness, drug safety, and differential effects among subpopulations (for example, by sex, age, or race). Draft questions were posted to a Web site to allow public review, comment, and testimony.6 Once finalized, key questions were submitted to an Evidence-based Practice Center (EPC) to conduct the literature search and systematic evidence review. For each drug class, the EPC conducted a systematic literature search identifying the key peer-reviewed studies relevant to each specific key question. Comprehensive reports generated by the EPC were used by the committees, in conjunction with public comment and testimony, to determine which drugs were most effective and safe within the class as dictated by the key questions. Committee reports were submitted to the Oregon Department of Human Services for price determination. If the practitioner committee determined that the evidence did not support clinically meaningful differences between drugs within a class, then all comparatively effective drugs, competitively priced to the least expensive (benchmark) agent, were selected as preferred drugs on the PMPDP. A standing update committee was also established to meet regularly and evaluate new evidence or new drugs in the reviewed classes based on a preliminary EPC-generated update report. If the update committee believed that new conclusions would be drawn, the drug-specific practitioner committees were reconvened with the standing update committee for a review of the new evidence. Implementation. The first four reviewed drug classes were implemented in August (proton pump inhibitors, or PPIs, and long-acting opioids) and September (nonsteroidal anti-inflammatory drugs, or NSAIDs, and statins) 2002. Initially, the PMPDP was practitioner-enforced, and prescribers were given a simple way to prescribe off the list. If a prescriber determined that a patient required a nonpreferred PMPDP medication, the prescriber had the option of writing "dispense as written" (DAW exception) on the prescription to notify the pharmacist to override the policy. Dispensing pharmacists played a key education role during this period, because they were responsible for notifying prescribers of the need for an override. Because the initial changes in use did not meet anticipated budget targets, the state implemented an administrative rule in May 2003 requiring prescribers to contact the state’s pharmacy benefit administrator to hear an educational message about the comparative evidence. The "soft" prior authorization (PA) did not require the prescriber to use a preferred agent or submit justification for the nonpreferred selection. Only the first four classes were subject to these enforcement policies (DAW exemption, soft PA). In October 2003, legislation was passed that prohibited any active enforcement of the PMPDP, thereby revoking the soft PA administrative rule.7 All classes added subsequent to October 2003 have not been subject to any enforcement. Under this voluntary process, the PMPDP was expanded in May 2003 to include estrogens and in November 2003 to include overactive bladder drugs, oral hypoglycemics, skeletal muscle relaxants, and triptans.
Prescription drug claims from January 2001 through June 2004 were used to evaluate the impact of the various enforcement methods on the first four drug classes (PPIs, NSAIDs, long-acting opioids, and statins). There was a thirty-four-day supply limit on dispensed drugs during the evaluation period. Usage over the study period was divided into four distinct time periods: pre-policy period (1 January 2001 through 31 July 2002 or 31 August 2002 for PPIs and long-acting opioids, or NSAIDs and statins, respectively), DAW exception period (1 August 2002 or 1 September 2002, depending on class, through 30 April 2003), soft PA period (1 May 2003 through 30 September 2003), and voluntary period (1 October 2003 through 30 June 2004). The proportion of prescriptions dispensed as preferred PMPDP agents (PMPDP market share) was quantified and averaged over each period. Changes in market share were quantified as the relative change of one period over the previous. The list was dynamic because the preferred PMPDP agents changed depending on pricing variation or the emergence of new evidence (such as new outcome evidence for atorvastatin within statin class). To better account for these list changes, the proportion of prescriptions dispensed as the PMPDP benchmark agent, which did not change during the evaluation period, was evaluated similarly. Market-share trend analyses were also conducted on the drug classes that were not subject to direct enforcement (for example, triptans, oral hypoglycemics, overactive-bladder drugs, and skeletal-muscle relaxants) eight months before and after they were added to the PMPDP. Estrogen use was not evaluated because of the concomitant changes in prescribing occurring as a result of the recent negative findings about hormone replacement therapy.8 A cost analysis was performed for the NSAIDs, long-acting opioids, PPIs, and statins. Cost trends were estimated using the ingredient cost per enrolled member per month and adjusted for federal and supplemental rebates. A segmented ordinary least squares linear regression analysis on class-specific and total cost per member per month was performed using the time segments described above.9 Coefficients reflecting a one-time immediate change are presented as the percentage difference of the expected trend if the policy had not been implemented. Changes in trend are reported as the absolute increase or decrease in slope. Statistical analyses were performed with SAS version 9.1. Program savings were estimated taking the difference between a linear projection of total cost per member per month and the observed cost per member per month adjusted for the cost of administering the soft PA during that time. In January 2003, in the midst of a major budget shortfall, the OHP reorganized its Medicaid program into two distinct benefit packages that differed in benefit structure: Oregon Health Plan (OHP) Plus and OHP Standard.10 Another component of the restructuring involved instituting various cost-sharing mechanisms such as copayments for drugs and services, which varied dramatically between the OHP Plus and OHP Standard populations. Most notably, OHP Standard enrollees were required to pay copays between $3 for generics and $15 for brand-name prescription drugs. Clients enrolled in OHP Plus were asked, but not required, to pay copays of between $2 for generics and $3 per brand-name prescription. To minimize the confounding effects of this policy, we excluded all OHP Standard benefit recipients from our analysis.
Market share. The average monthly enrollment in the OHP FFS program during the months of our study was 129,021; it remained relatively stable throughout. The average fluctuation from month to month was 1.8 percent. During the evaluation period, the PMPDP and how it was enforced changed several times, particularly for the first four classes reviewed (Exhibit 1  ). In addition to the two policy changes, two major PMPDP preferred list changes occurred during the evaluation period. In April 2004, Zocor, Lescol, Lescol XL, Lipitor, and Altocor were added to the preferred statin list. Duragesic was removed from the long-acting opioid list in November 2003.
To adjust for these drug changes, we performed another analysis of percentage of prescriptions dispensed as the PMPDP benchmark (lowest-cost agent) (Exhibit 2 ). The benchmark agents for the initial four classes remained constant during the evaluation period. The average monthly market-share figures for each policy period for all preferred and benchmark drugs are shown in Exhibit 3.
The statin class was the most profoundly affected by the PMPDP implementation. After the DAW exception was activated, market share of preferred agents increased 28 percent overall. Market share for preferred statins doubled, and preferred NSAID use increased the least. After the soft PA was employed, the use of preferred products in aggregate increased 43 percent, from 58 percent to 82 percent. Again, the use of preferred statins demonstrated the greatest increase, followed by NSAIDs. Preferred PPI market share increased the least. When the soft PA policy was retracted, overall PMPDP preferred market share decreased overall 17 percent. NSAIDs showed the largest decrease. The 37 percent reduction in preferred long-acting opioid use was attributable to the removal of Duragesic as a preferred agent. When we restricted the market-share analysis to just the benchmark drug, generic long-acting morphine use dropped only 4 percent. Generally speaking, the benchmark drug market-share trends followed a pattern similar to that of the overall preferred PMPDP market share.
Overactive-bladder drugs, skeletal-muscle relaxants, oral hypoglycemics, and triptans were added to the PMPDP in November 2003. Aside from being posted on the Medicaid Web site, no formal policies were implemented to encourage their use. No discernible changes in the use of preferred drugs were observed after these classes were added to the PMPDP (Exhibit 4
Cost analysis. Exhibit 5 shows the results of the segmented regression models. The aggregate model demonstrates that implementation of the DAW exception was associated with a 9.1 percent immediate per member per month cost decrease, as well a significant reduction in overall trend. The soft PA period was associated with a 17.7 percent immediate decrease, along with a significant increase in slope of $0.28. Within the specific classes, our model showed that the DAW period was associated with statistically significant abrupt declines of 8.8–14.6 percent. Significant immediate decreases of 11.0–42.0 percent were detected after establishment of the soft PA. Elimination of the soft PA was not accompanied by any statistically significant change in aggregate cost per member per month. However, immediate 46.6 percent and 9.0 percent increases in costs were observed in the NSAID and statin class, respectively, after the soft PA was repealed.
Exhibit 6 shows observed and predicted spending for the first four PMPDP classes before and after the two policies were implemented. Overall, we estimate that the DAW exception was responsible for an average monthly reduction of $1.41 per member per month (95 percent confidence interval: $0.80, $2.78), or $191,932 in total (95 percent CI: $108,456, $275,409) in monthly savings. The PA policy was projected to reduce spending by $3.58 per member per month (95 percent CI: $2.78, $4.37) or a total of $444,660 (95 percent CI: $363,205, $570,870) per month. During the entire period, the DAW policy and soft PA saved an estimated $1,727,392 (95 percent CI: $976,102, $2,478,682) and $2,223,300 (95 percent CI: $1,816,027, $2,854,353) in pharmacy costs, respectively.
This analysis, which focused on the implementation of the PMPDP, demonstrated that the use of both a DAW exception and a soft PA method were effective at reducing spending and shifting market share within the four initially implemented drug classes in the PMPDP. The DAW exception policy shifted market share toward preferred agents by 28 percent. The market share of preferred agents rose another 43 percent after the soft PA took effect. Revocation of these policies in October 2003 for the initial classes has contributed to a reversal of previously established market-share gains. NSAIDs and long-acting opioids experienced the largest reversal of preferred drug market share after the soft PA was eliminated. Negligible market-share changes occurred in classes that were not subject to any active enforcement policy. Cost savings were estimated to occur in the initial PMPDP drug classes under both enforcement strategies. Total per member per month costs were reduced significantly, as both an immediate reduction and a decrease in the month-to-month trend, for both the DAW policy and soft PA. These decreases translate into approximately $4 million in pharmacy program savings over the life of both policies. It is estimated that almost 70 percent of state Medicaid programs use PDLs to assist in managing drug spending.11 However, similar to other drug management strategies, little is known about these policies’ impact on economic or health outcomes.12 The general lack of information about PDLs is further compounded by the considerable heterogeneity with respect to how PDLs are synthesized and enforced in various state Medicaid programs. Although nearly all states use clinical evidence to help decide which drugs will be included on their lists, Oregon is unique in the explicit process it uses to evaluate and compare this evidence. Both the DAW exception and educational soft PA enforcement methods were also unique among Medicaid programs.
Study limitations.
Several important limitations of our study should be considered. One potential confounding factor was the implementation of a patient copay policy in January of 2003. Prescription drug cost-sharing policies have been demonstrated to reduce use and potentially shift market share. Although we attempted to control for this by excluding clients who faced the highest copayment schedule, it is impossible to completely eliminate its effect. Trends described in Exhibits 1 Concerns about the process. Many have expressed concerns and skepticism about the process in which Oregon has translated clinical evidence into policy. Challenges to the process include developing questions that adequately capture important outcomes, identifying and evaluating unpublished data, the insufficient quantity of high-quality research, lack of direct comparative studies between drugs, how to appropriately integrate evidence from observational research, and how to fairly deal with potential conflicts of interest at several levels. Also, the reviews focused only on comparing agents within a class and did not compare agents from different classes that could be used for the same condition. Finally, while the EPC-generated reviews made no claims of therapeutic equivalence when evidence was lacking to suggest a difference between drugs within classes, the state made its preferred drug selections based on this assumption. Some have argued that agents should not be selected for the PDL when evidence is lacking to establish whether or not a true difference exists.13 OREGON HAS ADOPTED a unique system for translating the principles of evidence-based medicine into policy decisions regarding the use of prescription drugs. However, translating this research into meaningful changes in practice has lagged somewhat because of lack of enforcement mechanisms aimed at encouraging preferred drug use. Our data suggest that for an evidence-based PDL to have an impact on drug use or costs, effective methods of promoting evidence-based choices need also be employed. Data from nonenforced classes suggest that having an evidence-based process to determine preferred agents alone is not sufficient for changing prescribing behavior. The state is now using educational techniques such as faxed prescription-change forms and provider detailing to disseminate research generated from the comparative reviews.
Daniel Hartung is assistant director of research and program evaluation in the Drug Use Research and Management Program, College of Pharmacy, Oregon State University, in Portland. Kathy Ketchum is coordinator of Medicaid-related programs. Dean Haxby (haxbyd{at}ohsu.edu) is an associate professor of pharmacy and director of the Drug Use Research and Management Program. This work was supported by the Oregon State University College of Pharmacy, which receives funding from the Oregon Office of Medical Assistance Programs.
| ||||||||||||||||||||||||||||||||||||||||||||||||
Top
Evaluation Methods
Study Results
