Human Data Science: A Multidisciplinary Approach

Developing a strategy for selecting investigator sites for a global clinical trial is incredibly complex. The goal sounds simple: Find the most qualified sites to enroll the right patients, faster. But there’s a multitude of factors to consider. First, you need a deep scientific understanding of the molecule as well as a clinical understanding of the protocol, the treatment pathway, and the patient population – the kind of knowledge that only clinicians possess. At the same time, you need detailed knowledge of the data sets available to find patients, assess site performance, and forecast enrollment – the kind of insight that only IQVIA data scientists have.

My role is typical of a human data scientist in that I serve as a go-between, or translator, between the world of the clinician and the world of the data specialist. My daily challenge is to see that the very specific and complex requirements of a clinical protocol are understood by our data analysts so that the data can be applied properly, given the nuances from data source to data source, from country to country.

So, I form a bridge of sorts between those with domain expertise in the therapy area, in our data assets, in technology, and in advanced analytics. Michael Kleinrock made this point in an earlier article, and I agree: It’s a group effort. No single person can do this alone because no one is an expert in everything. Everyone on the team has his or her own specialty, and then beyond that we each must have knowledge that’s a mile wide and an inch deep. But that’s the way it should be – how else are we going to come up with new ways of doing things if we don’t work with new groups of people?

See us in action: Site Selection

Site identification is such a good example of why human data science is so critical to improving how things are done; and how human data scientists rise to the challenge! From the very beginning of the site strategy, we are pushing to look at the full picture; this isn’t about number crunching, or algorithms in isolation. We gather clinical, therapeutic, regulatory, and logistical experts, because we know that we need to understand the clinical, the analytical and logistical aspects of the study in detail.

Only then do we look at how the data can support the trial strategy, and this is where the human data scientist shines, because I need to be able to translate the very specific clinical language of the protocol into data language and understand how this data acts as the main currency for site ID strategy. What variables are important? How should the query be structured? Without a nuanced understanding of what our data assets look like, we’d be lost. There are subtle differences in the data from one country to the next, and they make all the difference in the validity of the recommendations we produce; Even a variable as simple as patient counts may not be at all straightforward, due to the complex, underlying methodology that produced it.

Once we have covered this ground, we can work with our other colleagues to confidently input all the right variables into machine learning (ML) algorithms that will help us build a better, more informed, more objective site strategy that supports the overall study goals.

Human Data Scientists: The Unicorns of the Industry

The ideal candidate for most jobs in human data science just doesn’t exist in reality – so when we’re hiring, we say we’re looking for a unicorn. There are just too many areas of expertise required, and every one of us has our strengths and weaknesses. That’s why human data science is a multi-disciplinary field. There are, though, some commonalities that I think set the human data scientist apart from other data scientists:

• Curiosity. You have to know the right questions to ask. For example, do we really understand the overall landscape of the drug market in the specific indication? What are the treatment patterns? Are we using the right data sources, or do we need to approach the problem in a different way?

• Being a life-long learner. The learning curve in this field is steep. I’m still learning something new every day about our data sources and the nuances within them. Plus, we’re always evolving and adapting. What I learn one week may change the next.

• Ingenuity/creativity. The answers are rarely obvious, especially as each protocol is unique. Plus, every approach has a caveat. So, we have to be ready to apply different principles to the problem at hand and to be agile in our thinking.

• A desire to be transformational. Human data scientists can’t be order takers who treat assignments as transactional. Each project requires exploration, with the goal of transforming the sponsor’s approach through a data-driven strategy. If we were to merely hand off the data, it would be helpful, but we wouldn’t be living up to our goal of transforming the trial process.

• Passion. At the end of the day, our mission is to help companies usher drugs to market faster for patients’ benefit. We’re not just “geeking out.” Our work has meaning.

For human data scientists, the underlying goal is to ensure that healthcare decision-makers use data (the right data) as the foundation of their decisions. I believe in this so strongly that when I realized that IQVIA was a pioneer in human data science, I said, “Sign me up!”

This is one profile in an ongoing series from IQVIA on the “life of a human data scientist.”  Read more installments designed to explain this emerging discipline and how it is poised to address healthcare’s biggest questions and support its toughest decisions.

To learn more, please visit our human data science page.