Diagnosing The Prescribed Rx

It is one of the most common questions asked in healthcare marketing: how much influence do physicians and patients have over the decision to prescribe a treatment? Some disease states are clearly much more patient- driven and require patients to be activated to initiate a conversation with their physician about a particular treatment. And so a common objective of pharmaceutical market research is to measure key elements of the patient/physician conversation: that is, how likely a patient is to ask a physician for a brand, or how likely physicians are to prescribe a brand.

These objectives are central to a lot of pharmaceutical market research, in particular brand equity tracking, concept testing and patient flow research. For many years, we have collected patient datasets and/or physician datasets with these metrics and used them to make marketing decisions for each stakeholder.

But there has never been a way to have these datasets interact in a formal and rigorous framework to understand and measure the dynamics between these key stakeholders.

Ipsos Science Center is our analytic think tank, responsible for developing and bringing to market cutting edge analytic approaches. Often these approaches are based on principles of artificial intelligence algorithms in which market dynamics are replicated thousands of times to simulate how specific markets behave. One of the modelling approaches that Ipsos Science Center has pioneered is known as Multi-Agent Interaction Modelling (MAIA), a form of evolutionary game theory in which simulated market agents are endowed with behavioural properties and then allowed to interact with each other. This approach is well-suited to understanding purchase decisions in markets where multiple stakeholders influence a single purchase decision.

Clearly, the conversation between a patient and their physician is a prime example of such a multi-stakeholder purchase decision. Both parties may have some degree of influence on which treatment the patient ends up being prescribed.

To model this conversation, we have to articulate a formal model of how patients and physicians interact. Like all modelling, the simpler the model dynamics can be while still capturing the essence of the situation, the better. An illustration of these dynamics is as follows:

This interaction tree has the patient and physician agents interacting in a formal sequence of moves that result in one of five outcomes. Two of these outcomes result in a script being written and three of them result in no script being written.

The MAIA approach takes patient and physician datasets that ask each type of respondent their likelihood of doing each of these actions (asking for a script, granting a patient request, etc.) and randomly selects one physician and one patient and simulates the likely outcome of that individual interaction using their specific probabilities of each action. The simulation then replicates this process thousands of times, each time generating probabilities of outcomes. We then aggregate these simulations to create a description of the market.

The model can be made more realistic by imposing restrictions on which patients interact with which physicians. For example, we can create more interactions with those physicians who see more patients in a typical month. Or we can restrict patients and physicians interacting only if they are in the same geography (and hence face the same reimbursement realities).

The fundamental output of this analysis is an assessment of how likely each of the five outcomes are:

So we are now able to answer the following types of questions:

• How much relative influence do patients and physicians have on a prescription decision and therefore where should you put more marketing investment?
• To what extent is a patient education campaign increasing patient activation to ask for particular treatments and are physicians reacting positively to that activation?
• In concept testing, how do you compare situations where patients and physicians prefer different concepts? How do you determine which one concept is best to bring to market?
• Are there specific barriers (patient or physician) that are preventing greater adoption?

This approach can be extended to model more subtle dynamics as well. In the launch phase of a product, physicians' early clinical experiences of a new treatment can have a dramatic impact on speed of adoption. Using MAIA, we can formally determine how sensitive a new product is to this phenomenon and, based on clinical trial evidence, provide a more nuanced assessment of how early adopters might react. Or, we can also model peer-to-peer information sharing in disease states where physicians often interact with each other to determine how important word of mouth advocacy might be.

MAIA is providing a powerful new lens through which to study the interactions of patients and physicians and allowing us to formally model dynamics we've never been able to before. We are excited about the potential of this new approach and would welcome discussing how we can apply it to your strategic imperatives.

For more information, please contact David Scowcroft, SVP Ipsos Healthcare at [email protected] or (416) 847 9036.