[Image courtesy of Adobe Stock]
COTA Healthcare recently unveiled what it calls a major breakthrough in real-world oncology data (RWD). At the heart of this achievement is a generative AI (GenAI) platform that the company says makes large-scale curation of cancer data both accurate and profitable, marking a significant shift from previous industry attempts to automate the labor-intensive RWD abstraction process.
“The careful, expert, labor-intensive process that has defined RWD data abstraction at the core of drug development and treatment is not profitable,” noted COTA Healthcare CEO Miruna Sasu.
According to Sasu, earlier attempts to automate data abstraction often failed owing to reliance on technology that wasn’t sufficiently reliable for expert abstractors to trust. She notes that COTA itself went through multiple cycles of re-engineering processes with technology that ultimately came up short. When foundational large language models (LLMs) emerged in 2023, however, COTA began experimenting with ways to automate core abstraction tasks and front-end queries.
“It soon became apparent that using expert abstractors in the loop—but no longer at the front of the process—would be critical,” Sasu explains. She describes a hybrid model in which leading-edge AI does much of the initial data interpretation. Meanwhile, human oncologists and data scientists validate for accuracy.
Miruna Sasu
COTA says this new approach has dramatically shifted the operational and financial dynamics of oncology RWD. Sasu points to a steep reduction in cost per record, a faster turnaround time, and accuracy that meets or surpasses previous benchmarks.
“We believe COTA is the first in oncology RWD to develop a profitable data abstraction model. In 2024, COTA reduced cost per record by 23% year over year, which is expected to drive positive EBITDA and cash flow in 2025 while maintaining 98% accuracy—far outpacing traditional methods that take hours or even days,” Sasu said. “The speed, scale, and accuracy are continuing to improve.”
Study findings and RWD significance
As academic and industry researchers increasingly rely on RWD to inform drug development and clinical decisions, COTA’s soon-to-be-published findings could be pivotal. According to Sasu, the organization’s research suggests that GenAI-assisted data abstraction can achieve near-instant queries. The firm notes that the method matches or exceedes the 93% accuracy benchmark set by human abstractors.
The potential impact is twofold: it saves significant time—reducing tasks that once took days or weeks to just minutes or seconds—and it also democratizes access to insights. Historically, Sasu notes, only data scientists with specialized coding skills could interpret unstructured EHR data at scale. COTA’s platform, dubbed “CAILIN,” aims to open real-time data exploration to broader R&D teams, including researchers who lack formal analytics training.
Business metrics and implications for life sciences
COTA reports a 25% decrease in cost per record (CpR) compared to the previous year and a 32% drop in abstraction handling time (AHT). The company’s data inventory also grew by 350%, while it reduced its contracted abstractor workforce by 40%—all signals, Sasu suggests, of a successful transition to a more automated, LLM-driven model.
In biopharma, the technique could mean faster, more precise identification of patient cohorts, as well as near-instant insights on treatment patterns and biomarker expression. Previously, any question—such as the number of patients on a specific third-line therapy for a certain cancer subtype—could take weeks to answer while waiting for data scientists to code and query the data.
According to Sasu, CAILIN’s search-engine-like interface drastically shortens that timeline and lowers barriers to entry for researchers across clinical development, medical affairs, and health economics and outcomes research (HEOR). Users with the proper permissions simply type their query into a platform interface and receive accurate results in seconds.
Filed Under: machine learning and AI, Oncology
