Louisville-based EdjAnalytics has developed a predictive model that helps federal regulators more effectively select drug makers for inspections, which should improve public health.
The model helps the Food and Drug Administration predict where its inspections in the drug supply chain are mostly likely to generate a finding that requires some kind of corrective action. The idea is to help the agency use its limited resources more effectively so that it can increase the chance that it inspections reveal a situation that needs to be addressed to improve the safety of drug production and, ultimately, public health.
Meanwhile, leaders of EdjAnalytics expect that they can leverage the experience they’ve gained from their collaboration with the FDA to make further inroads into the life sciences sector.
EdjAnalytics Chief Operating Officer Susan Olson told Insider that she couldn’t provide details about data sources, methodologies or implementation strategies because of confidentiality. However, she said that in general, the model involved Edj compiling and analyzing data sets about matters including drug evaluations and research processes.
The company said in a news release that it “trained machine learning models on historical inspection outcomes to create a predictive model. Features related to facilities, their associated products and their historical inspection outcomes were used as model inputs. Model development was guided by predictive error, implementation feasibility and the interpretability of findings to stakeholders.”
Olson said the work for the FDA was a big deal for Edj, as it was paid with a $400,000 federal grant and involved about a quarter of the company’s workforce, including data scientists, analysts, product managers and executives.
“We’ve learned a lot,” she said Monday by phone from California, where she attended a biotech conference to give a presentation about machine learning.
Olson said Edj leaders are exploring how to deploy the company’s new capabilities in the life sciences sector by providing predictive services for pharmaceutical companies.
For example, she said, adverse side effects of many prescription drugs can be predicted for a broad population — but not for individuals. Edj leaders hope to determine an individual patient’s risk for adverse events based on an analysis of their electronic health record and other factors, such as income level and the neighborhood in which they live.
EdjAnalytics has its origins in astrophysics and high-level backgammon and helps government agencies and businesses make better decisions through data analysis and predictive modeling. Its sports subsidiary has gained national attention because its analytics power helped the Philadelphia Eagles win the Super Bowl in 2018.