It’s been said so many times because it’s true: History repeats itself. It’s also the basis of predictive analytics, a method that uses a variety of mathematical techniques to dig into historical data and arrive at data-driven predictions. And it could have serious implications for patient recruitment.
Long story short, predictive analytics allows you to optimize efforts to reach a particular audience. For instance, if a sponsor needs to recruit a certain number of patients for a study, we at Praxis plug in different media mixes to see how we can best reach that goal. We adjust for granular details, such as indication prevalence, therapeutic area, number of planned recruitment sites, desired media markets, media running time, and so on. Then we adjust again (and again and again) for all different scenarios to determine what the ideal course of action would be for a given study.
So how do we handle the predictive analytics process from start to finish? It goes something like this.
Collect the relevant historical data.
First we collect all the relevant data. This data can come from a wide range of sources with different structures, so this is when we get organized. There is a massive amount of information out there that we can add to our dataset, such as TV campaign start dates and flight durations, which we can correlate with performance to determine effectiveness.
Identify the key variables.
Once all the data is in one place and structured appropriately, we assess what should be included in our model. Keeping our objectives in mind, we select variables based on whether we believe they will influence the outcome we would like to predict. Variables can always be added or removed later, but we like to be inclusive during this step.
Create and refine the model.
With our key variables in hand, we use SPSS Statistics to generate a model that identifies and optimizes relationships between our variables and the outcome of interest. To ensure our model is accurate in its predictions, we test it against known data, such as the number of patients recruited in a past study. If results do not match well enough, or if we find that some of our variables do not contribute to the model, we can add or remove variables to increase the overall fit to the data.
Once we find the right model for the data, we plug in our desired parameters to see what the approximate results would be so we can then provide a realistic estimate of what to expect given the defined parameters – all within a margin of error, of course. At this point, we have a working model that can accurately predict the outcome of a media campaign. We can present our expected minimum and maximum numbers of patients recruited within any constraints, including from budget to time and any number of conditionals in between.
With predictive analytics, the Praxis team is able to stay as efficient as possible while always maximizing our clients’ spends, no matter the budget. If you’re interested in supporting your next recruitment effort with predictive analytics, get in touch.