Cornerstone AI to Present Research Findings at 42nd ISPE Annual Meeting: AI-Assisted Laboratory Data Cleaning Achieves 99.9% Accuracy, Surfaces 20% More Records Than Manual Workflows

Peer-reviewed pharmacoepidemiology study concludes AI-assisted standardization outperforms manual data cleaning on accuracy, completeness, and reproducibility in a cohort of 19,415 patients with metabolic dysfunction-associated steatohepatitis.

‍
NEW YORK - Cornerstone AI, the expert-guided AI company for healthcare and life sciences data, today announced that company representatives will present its latest research at the 42nd International Society for Pharmacoepidemiology (ISPE) Annual Meeting. The conference will be held at the Allianz MiCo in Milan, Italy, from August 29 to September 2, 2026.

‍
The study, conducted in collaboration with researchers at Gilead Sciences and titled "AI-ssisted real-world lab data cleaning to improve quality and efficiency of real-world evidence generation," evaluated whether an AI-assisted approach could improve the accuracy, completeness, and reproducibility of laboratory data cleaning relative to manual methods. The analysis examined approximately 4 million laboratory records from 19,415 patients with metabolic dysfunction-associated steatohepatitis (MASH), drawn from HealthVerity electronic health record data. Cornerstone AI's hybrid system, which combines rule-based logic with natural language processing and hierarchical classification, was benchmarked against manual cleaning workflows.

‍
Key Findings:
• The Cornerstone AI system standardized 99.8% of units and 96.3% of laboratory tests to LOINC.
• Frequency-weighted accuracy at the LOINC component level reached 99.9% for AI-assigned mappings, compared with 93.5% for manual workflows.
• AI-assisted mapping identified 20% more MASH-related laboratory records overall (801,874 vs. 667,857), with a mean increase of 38% per test across 14 prespecified MASH-related tests.
• Third-party review adjudicated discrepancies between AI-proposed and manually assigned mappings, confirming the accuracy differential.

The authors conclude that the AI-assisted approach substantially improved data usability and demonstrated accuracy and explainability aligned with regulatory guidance. This has direct implications for cohort definition and disease characterization in pharmacoepidemiologic research.
‍

One of the study’s authors, Mary Beth Ritchey, PhD, FISPE, Principal and Owner of Med Tech Epi, LLC and Associate Research Professor, Rutgers Center for Pharmacoepidemiology and Treatment Science noted:

"Accurate, complete data is foundational to credible pharmacoepidemiologic research and laboratory data cleaning has historically been one of the most burdensome steps in the workflow. This study demonstrates that an AI-assisted approach, when designed with the right scientific rigor, can produce both higher accuracy and greater completeness than manual methods, in substantially less time and using fewer resources."

Michael Elashoff, PhD, Co-Founder and Chief Scientific Officer, Cornerstone AI added:

"Pharmacoepidemiologic research is held to a high evidentiary bar, and rightly so. What this study shows is that AI-assisted data cleaning, done properly, is not a tradeoff against rigor. It is a path to greater accuracy, completeness, and reproducibility than manual workflows can achieve, with the audit trails and explainability that relevant and reliable evidence generation demands. We are grateful to the Gilead team for their leadership in subjecting this approach to peer review, and to Dr. Ritchey for her independent contributions to the study design and analysis."

About the Study
"AI-assisted real-world lab data cleaning to improve quality and efficiency of real-world evidence generation" will be presented at the 42nd ICPE in Milan. Authors include Carrie M. Nielson,
Fang Xia, Colin Anderson-Smits, and Alex Asiimwe of Gilead Sciences; PJ Allen, Andrew Howland, and Michael Elashoff of Cornerstone AI; and Mary E. Ritchey of Med Tech Epi, LLC and the Rutgers University Center for Pharmacoepidemiology and Treatment Science.
‍

About Cornerstone AI
Cornerstone AI provides expert-guided AI software that delivers analysis-ready data in days, not months. Unlike generic AI tools, Cornerstone leverages a proprietary library of data patterns and algorithms driven by healthcare subject matter experts to accelerate and scale data cleanup for healthcare and life sciences organizations. Customers gain measurable productivity, higher data quality, and faster decisions, supported by a trusted partner who increases the value of their data investments. Learn more at www.cornerstoneai.com.
‍

Media contact:
Amber Moore
amber@moorecom2.com
503-943-9381

‍