“Zeroes and ones will take us there!”
– Jesus Jones (1993)
I’ve always loved the data part of strategy.
But I’ve always had to be honest about its flaws.
In clinical research, we use data to assess how and where patients been recruited in the past,
To drive assumptions how recruitment will happen in the future.
Some like to use the mean or average for this data, others the median or midpoint.
[And there’s still the mode but I only accept Depeche]
When it comes to the “right” way, I’ve found that every scenario is unique
But to always look at the data from multiple angles.
For example, I was putting together a proposal using data from a handful of similar trials.*
One particular trial was showing a really high average enrollment rate.
With that, I could develop a strategy with very short timelines, and save the client some cost.
But a few of the other trials showed much slower enrollment rates…
Making me think “something is amiss.”
[Really, I just wanted to use the term amiss]
Sure enough, the median rate showed enrollment rate was much lower.
The difference between one patient per month and one patient every 3 – 4 months.
What caused this?
One investigator was the outlier, enrolling 7 times the number of patients.
[A terrific outlier, but outlier nonetheless]
This one site expanded the average rate well above the median, and could potentially overestimate how the new trial would enroll.
In the end, after discussing the level of risk the client was willing to take, I used a rate much closer to the median. A longer timeline, but one that took into account the full range of trials rather than just the one.
My takeaway? In this context both rates are valuable.
The average shows where I could leverage higher enrollers,
The median shows how a wider set of sites will perform overall.
Having both, I could make and present a well-informed decision!
*Similar has many layers of course. Several other factors are involved like study design, the drug itself, ongoing competition, excitement etc., but that’s for another post.