By Doug Willson, Senior Vice President, and Rick Nelson, Associate Vice President
In today’s competitive pharmaceutical marketplace, quantitative research with managed care decision makers has begun to play an increasingly important role in assessing early stage opportunities for drugs in development. Historically, quantitative research with payers was confined to later stages, focusing primarily on pricing decisions for drugs with established clinical performance. With many brands going generic in the next few years and increasingly crowded pipelines, payers will play an even greater role in the success of drugs that launch in the future. As a result, there has been renewed interest in factoring in the impact of payers in early stage forecasts.
Forecasts of managed care decisions are often developed using a choice modeling approach. In the typical managed care choice model, managed care decision makers are shown a series of market scenarios reflecting the launch of a new drug with a specific clinical profile, pharmacoeconomic claims, and cost information including wholesale pricing, discounts/rebates and other contracting parameters. For each scenario, respondents are asked where they would place the new drug on formulary and what restrictions (if any) they would require. These models are typically estimated using standard discrete choice modeling tools and are used to forecast formulary coverage on a national basis. The managed care model is frequently combined with related (“linked”) choice models for physicians and patients to create an integrated forecast.
At first glance, choice modeling with Managed Care Organizations (MCOs) appears to be a straightforward application of standard research methods. However, the performance of traditional approaches in real-world situations often leaves much to be desired. Samples of MCO decision makers are often small – typically ranging from 30 to 60 respondents – and must be configured to represent the diverse set of payers (e.g., commercial managed care, Medicare Part D, managed Medicaid), geographies, and plan sizes (i.e., number of covered lives). The complexity and heterogeneity of managed care formularies in the market today also causes serious problems for traditional aggregate choice modeling approaches.
This article provides a short primer for developing choice modeling studies with MCOs, to help ensure your next forecasting study runs smoothly. We review the basics of preliminary qualitative research, sample design, survey content, choice modeling, and analysis, and illustrate with examples from specific studies.
Almost all quantitative survey research will benefit significantly from a preliminary round of qualitative interviews. Qualitative research can be used to investigate and refine presentation of the new product profile, choice model attributes and levels, and other aspects of the quantitative questionnaire. It is important to realize that there are ongoing groundbreaking changes in how health care will be organized and paid for; how these changes play out will have enormous impact on the opportunities for new drugs launching in the next five years. Qualitative research is extremely valuable for investigating and ultimately developing and properly presenting the appropriate future market context for payers in the forecast. As an example, most branded atypical antipsychotics will be going generic in the United States in the next two to three years. For new branded drugs launching to treat schizophrenia, bipolar disorder and depression, how will this impact their opportunities? A large percentage of use for these drugs is currently off-label, presumably reflecting a high degree of unmet need for these patients. With genericization of current branded drugs, will payers start to manage these drugs as a class, requiring step therapy (prior use of generics) or prior authorization before they can be prescribed? What reference prices/discounts are appropriate for branded drugs in the future market? Answers to these types of questions help shape the market context for the choice exercises in the quantitative survey. For the qualitative research phase, it is extremely important to work with a moderator who is experienced with interviewing payers. While the clinical information presented to payers is similar to what is shown to physicians, payers are ultimately concerned about the relationship between price and value. Their context for decision making, language and hot buttons is different than for physicians, and it is crucial that the moderator recognize and appreciate these differences.
The U.S. managed care sector includes approximately 405 licensed health maintenance organizations (HMOs) and 925 preferred provider organizations (PPOs). Employers or the government are the ultimate financiers of health care – they contract with MCOs to offer health insurance to their employees. MCOs function like insurance companies, promising to pay health costs under the plan for enrollees, although in practice they act as intermediaries, purchasing insurance from providers. While there are a small number of managed care organizations, they offer a large number of plans. When surveying managed care decision makers, we typically ask them to focus on one plan (e.g., the largest) selected from the set of plans they manage.
As the universe of MCOs is relatively small (at least when compared with the universe of physicians), the typical managed care sample is also relatively small, ranging from a minimum of 30 to a maximum of 100 respondents (if multiple respondents from the same organization responding for different plans are allowed). Some research vendors offer online panels of MCO decision makers, so it is now possible to quickly recruit respondents for online choice modeling studies.
The typical managed care interview covers:
- Types of plans and products
- Common benefit designs and co-pay amounts
- Formulary placement and perceptions of current agents
- Reactions to the new product profile
In the choice modeling portion of the interview, managed care decision makers are shown a sequence of scenarios describing different characteristics of the new drug, including:
- Clinical characteristics – efficacy, safety/tolerability, dosing/administration
- Pharmacoeconomic information – health-related quality-of-life, long-term medical cost savings, nonmedical costs savings
- Cost – average wholesale price, discounts, rebates
In the example exercise shown below, different combinations of clinical attributes, price(AWP), and managed care discount are displayed. The displayed net price is calculated using the AWP and the discount. For each scenario, managed care decision makers are asked to choose the tier and identify the restrictions they would recommend to the P&T committee.
While the clinical attributes presented to an MCO decision maker will be similar to those shown to a physician, the pharmacoeconomic and cost attributes will be different. Payers are ultimately concerned with the price-value trade-off; information on overall cost and cost effectiveness is very important for payers, less so for physicians. For later stage research where clinical performance of the new drug has been established, the choice exercise may focus on pharmacoeconomic claims and detailed pricing structure attributes with a fixed clinical profile. For earlier stage research, the pricing attributes may be coarser, focusing on net price relative to a fixed average wholesale price that is comparable to current branded products in the category.
The sequence of choice scenarios is developed using an experimental design. For each scenario, managed care decision makers are asked to identify the likely formula status (i.e., tier and restrictions) for the new drug. The experimental design and choice data allow us to develop a choice model to estimate the distribution of covered lives across formulary tiers for any configuration of attributes and levels in the design.
While the traditional discrete choice modeling tools might seem ideally suited for modeling formulary placement decisions, their performance in real-world situations leaves much to be desired. The primary difficulty rests with the complexity and heterogeneity of the managed care formularies in the market today. Different plans have different tier structures (both the number of tiers and their interpretation), and the set of restrictions employed also varies significantly across plans and therapeutic categories. Although conceptually similar, the structure of the dependent choice variable differs for each plan, creating havoc for aggregate choice modeling approaches.
In recent years, hierarchical Bayesian (HB) models have been the preferred method for estimating choice models in marketing research. While HB methods provide robust individual level estimates, one theoretical requirement is that the individual level models have identical structure across individuals, which would appear to invalidate their use in MCO samples with different formulary structures. However, it is possible to modify the classical HB framework to allow for a common utility structure (i.e., common clinical, pharmacoeconomic and cost attributes) while allowing the number (and interpretation) of tiers in the formulary to vary across individuals. As a result, the greater forecasting accuracy associated with HB methods is also available in MCO choice modeling.
Results of the modeling effort are typically captured in an Excel simulator that allows the user to forecast the formulary distribution for different clinical, pharmacoeconomic and pricing parameters. An example for a new drug in the Alzheimer’s disease market follows. The simulator (shown below) allows the user to estimate the impact of changes in these attributes on formulary placement. The second chart shows the percentage of covered lives with preferred status (Tier 2 unrestricted or higher) as a function of net price; as price increases, the percentage of lives with preferred status decreases from a high of over 80 percent to less than 20 percent.
The simulator can be used to evaluate a variety of strategic forecasting questions:
- Quantify the relative importance of different clinical, pharmacoeconomic and pricing attributes for achieving preferred formulary status
- Identify the specific pharmacoeconomic claims that will be most compelling for payers in the future market
- Assess the trade-off between pricing/discount strategy and value; if your product is only modestly differentiated from competitors, what pricing strategy will guarantee access?
- Evaluate specific contracting strategies; What proportion of payers will guarantee exclusive preferred access for a deep discount?
The managed care simulator can be used directly to inform managed care contracting strategy, or in combination with physician and patient choice model simulators, to develop an integrated forecast. In either situation, following the simple steps outlined in this article will help ensure your next managed care forecast is a success.
Forecasting Managed Care Decisions