ACT: What are the greatest benefits to utilizing artificial intelligence (AI) in data management?
Lacroix: Coming from a data management background, there’s of course, across the trial lifecycle, lots of opportunities where AI can be applied. We hear a lot about patient recruitment. We hear about predicting the appropriate patients, the appropriate patients that will achieve the outcomes we’re looking for. We hear about drug discovery and accelerating drug discovery and more patient-centric approaches, but where my experience lies is more in the data management area, and for us, for my group specifically, that is the biggest area that we’ve been focused on, and where we started at eClinical is that as a data organization, we had access to the data, so that made sense for us to start with clinical data review. As data managers, we are largely responsible for processing that data—reviewing, data cleaning, coding—to ensure the data is of the of the highest quality for analysis. Traditional data review models don’t lend themselves to managing and processing these high volumes of data sources. In traditional approaches, data managers would have to sift through rows and rows and thousands and thousands of data points with a lot of overhead for manual programming, QC, and execution of edit checks. This is just not going to scale. It doesn’t scale with what we’re seeing today, so we really do need to look at more innovative approaches to be able to manage the complexity, and it’s just extremely inefficient.
Using AI and ML (machine learning) for data cleaning and interrogation is no longer an option and I think data management teams that aren’t traveling towards the use of AI and ML are going to be left behind, and are going to find themselves in a difficult position. We have so much more to manage as data managers that we need to look at the areas where AI can assist us and reduce that manual effort and burden and allow us to focus again on what’s critical. So, being able to utilize artificial intelligence for our data cleaning and data processing in our experience, and what we’re using at eClinical is it truly reduces the overall effort of looking at large volumes of data, stripping away what is not important, what the model deems as acceptable, and allowing us to focus only on what’s absolutely critical and anomalous. This really does reduce the time and effort. It allows us also to identify risks sooner in the process. When I think about looking at lab data, vitals data—these data sets that continually tend to be your largest, voluminous data sets, and trying to surface things that are concerning or trends, risks, shifts without intelligence is near impossible, or if it’s achievable, it takes a lot of individuals looking at it and a lot of time and effort to get there. The other piece that AI is assisting us with is manual reviews. They are very subjective, and the human is making all of the decisions as to what is deemed in general. We have rules around risk and safety risks and concerns, but there’s a human and that’s making a lot of the judgment calls, and people think differently. There’s inconsistency, so AI is really adding that consistency, reducing that kind of human subjectiveness, but yet still allowing the human to stay in the loop as the decision maker to determine what action needs to be taken.
I’m probably going on too much about data management, but this is where we’re experiencing the most benefit and the most benefit for our customers, and that we’re able to identify those risks and trends much sooner in the process. We’re able to allocate resources differently because we’re not having to have so many individuals executing on these data reviews and really allow us to work much more efficiently for our customers.