Insights from a Machine Learning Engineer: eIQ Review Use Case Spotlight | Part 2

As discussed in Part 1 of this blog series, managing the influx of clinical trial data has become a challenge for data management teams as today’s trials are becoming increasingly complex and data intensive. Regardless of how much the industry continues to change, ensuring study integrity and patient safety will remain core priorities across clinical teams. Leveraging innovative technologies will not only ensure that these objectives are achieved but will also provide opportunities to streamline processes and increase productivity across the clinical trial life cycle.

For data management teams, the integration of innovative technologies such as AI and ML can address the limitations of traditional approaches that are simply not scalable enough to support the pace at which clinical trials are evolving. Highlighted below are two examples of the AI/ML-enabled use cases deployed within elluminate^® IQ Review (eIQ Review) to automate clinical data review processes, ensure data quality, and increase efficiency across data management teams.

Domain Classification (DMCL)

Across clinical trials, submission to the FDA requires all data to be clean and recorded in the correct forms. Accurate data recording on the correct forms/domains is essential for maintaining the integrity of clinical trials, meeting regulatory requirements, facilitating analysis, and ensuring patient safety. With the adoption of SDTM v3.2, procedures now fall under a separate form than medical history and adverse events. This entails a data reviewer to check all records across many fields of the MH, AE, and PR forms to confirm correct data entry. Traditionally, manual data review – which is a highly labor-intensive and time-consuming process – is performed to guarantee this.

Domain Classification (DMCL) within eIQ Review identifies records and fields that do not belong in the entered form (i.e. a procedure entered in the medical history form) to accelerate timelines and minimize errors during the data review process. The DMCL model leverages the strength of a large language model to automatically flag incorrect data entries at both a record and field level which are shown to users within eIQ Review. With a near 98% reduction in records to review, DMCL significantly reduces data monitors’ workload, freeing up their time for more complex tasks.

Univariate Outlier Detection (UOD)

Exception listing reports are created for every study within a clinical trial. This involves defining data review objectives, rules to accomplish the objectives, and then coding/programming these rules. Reviews of these reports are prone to manual error and require a significant number of resources. Additionally, given the variety of the different types of exceptions and the datapoints within the exceptions, this review process requires a high level of domain expertise.

Univariate Outlier Detection (UOD) within eIQ Review automates the identification of anomalous data points on any numerical data set to reduce the time and resources spent on data review, increase consistency, and improve data quality. The UOD model uses a collection of unsupervised outlier detection algorithms to flag anomalous datapoints and surface them to the end user. In eIQ Review, users can configure the numerical datasets for UOD, allowing customized outputs targeting specific exception listings. As new data comes in, UOD can quickly find and update the anomalous data points, making the review process that much easier.

As clinical trial data becomes more abundant, complex, and digitized, the ways in which that data is managed must also evolve accordingly. eIQ Review is a simple way to begin incorporating AI/ML in your trials while delivering significant value. By leveraging these capabilities embedded within the elluminate Clinical Data Cloud, you can significantly reduce time spent on clinical data review while ensuring data quality at scale. You can learn about eClinical Solutions’ approach to ideating and building AI/ML uses cases aimed to mitigate the challenges of modern data management, increase productivity across clinical teams, and reduce cycle times here: Practical Approach in Ideating and Building AI/ML Use Cases

Authors

Dalton Macres

Machine Learning Engineer, eClinical Solutions

Alison Cronin

Director of Product Marketing, eClinical Solutions