With the evolution of forecasting models, analysts have more tools at their disposal to accurately forecast drugs and diagnostics.
Patient flow analysis is one of the most powerful of these tools.
It enables analysts to model the unique flow of patients through a specific disease area and capture the nuances of that market’s treatment process. (For more on the fundamentals of patient flow, see “Patient flow analysis and forecasting”.)
“Especially in a Recession, when companies need to understand more about where the areas of risk are and where black swans could lurk, it’s through some of this dynamic flow,” says Nic Talbot-Watt, managing director at Black Swan Analysis.
As more analysts embrace the technique, however, drawbacks are coming to light.
Patient flow analysis, for instance, introduces greater complexity into a model and thus a higher potential for error if analysts are inexperienced.
The technique is also unnecessary for forecasting 80 percent of drugs and diagnostics, for which static models will suffice.
Knowing when not to use patient flow analysis, therefore, is as valuable as understanding how to do it.
“You could use a patient flow model for everything, but you wouldn’t want to,” says Talbot-Watt. “It’s kind of like using a sledgehammer to knock in a picture hook. The idea is that you only roll out the big guns when you definitely need them.”
What, then, are the 20 percent of cases for which patient flow analysis is essential?
Talbot-Watt and her Black Swan team have created a four-dimensional compass to help analysts determine when patient flow analysis is appropriate.
Each dimension should be considered in keeping with the others, Talbot-Watt says, “so you arrive at a balanced perspective on whether or not you need a patient flow model.”
The first dimension considers the current level of treatment effectiveness. Are the current treatments out there doing what’s needed for the disease?
“The more effective the treatment is, the more static your model probably should be,” says Talbot-Watt. “Less effective generally means more dynamic.”
For instance, statins are quite effective, so when introducing a new statin to market, a static model will generally suffice.
Cholinesterase inhibitors for Alzheimer’s disease, meanwhile, are not particularly effective. Their rate of uptake, therefore, is closely linked and sensitive to the diagnosis rate of the patient population. If that diagnosis rate were to change and you had a static model, it would lead to a significant error in your forecast.
The second dimension is “market density.” If many products are clustered on the market, patient flow analysis is likely unnecessary.
“If there’s a very sparse market, however, or there’s no cohesion in terms of the classes used to treat the indication, you might be looking at a more dynamic model,” says Talbot-Watt.
The third dimension is “disease knowledge level.”
The less that’s known about a disease, the more volatility there’s likely to be in the treatments that surround it. More volatility, in turn, points to patient flow analysis.
The fourth dimension is “homogeneity of disease,” which analyzes the amount of disease variety. Divergent disease pathways generally necessitate a more dynamic model.
“Often this goes hand-in-hand with treatment effectiveness,” says Talbot-Watt. “When you have a homogenous disease, the treatments tend to be very effective, whereas if you’ve got a very heterogeneous disease, the treatments tend to have less efficacy.”
If your product passes through the above dimensions and comes out pointing toward patient flow analysis, Talbot-Watt offers the following advice.
First, the math at the core of patient flow analysis is actually quite simple. All of it derives from the basic equation Prevalence = Incidence X Duration.
However, as you begin to wade through logs and logs of data that indicate how long it takes patients to transition from one stage of a disease to another, and all the factors that could influence that transition, your world quickly gets complicated.
Stay patient. If numbers aren’t adding up—as they often don’t—look for simple mistakes you may have made along the way, like multiplying by the wrong figure.
“It’s generally something very simple but it takes a while to track that down,” says Talbot-Watt. “You have to have a lot of perseverance to go through it and keep going though it until you get to the solution that you need.
“People at that point often get disheartened and go back to how they were doing it before.”
Talbot-Watt also says it’s important to approach problems from the medical perspective of the disease, rather than simply by the numbers.
Analysts ideally suited for patient flow analysis have a nuanced understanding of medicine and can move beyond number crunching into analysis of how patients follow various treatment pathways.
“You need somebody who is very robust in Excel but also someone who can think a bit more laterally than your average business analyst or forecaster,” says Talbot-Watt.
Finally, be careful in the way forecasts that include patient flow analysis are communicated to the end user, says Talbot-Watt.
Upper level managers have seen a lot of forecasts in their day and have expectations for how they’ll look.
Forecasts built from patient flow analysis will often challenge those expectations, as they fluctuate due to dynamics in the market that static models simply gloss over.
“Marketing managers and brand managers are not used to seeing projections fluctuate in a dynamic way,” says Talbot-Watt.
Those communicating the forecast need to exhibit sensitivity to the newness of their approach coupled with an unwavering confidence in the method they’ve used.
“They have to trust it’s a rock-solid tool whose output they can stand by,” says Talbot-Watt.