Why We Need Fast and Flexible Chart Abstraction in Medical Research

Chart Abstraction is the Key to Medical Knowledge

‍

Structured data is often the first requirement to unlock new medical knowledge and improve patient care. However, achieving this structure isn’t easy. Dr. Dan Fabbri, founder at Brim Analytics, has firsthand experience tackling this challenge through his work on hundreds of clinical research projects. These projects often involve transforming unstructured medical notes into structured datasets to enable research.

Examples of such impactful work include predicting NICU discharge outcomes (study), monitoring drug safety (study), and assessing cancer treatment responses (study). Beyond these examples, structured data is essential for initiatives like clinical trial recruitment and building registries.

Despite its importance, chart abstraction—the process of creating structured data from medical notes—remains a bottleneck for many researchers.

‍

Traditional Abstraction is Slow, Costly, and Inflexible

‍

The traditional approach to chart abstraction is a manual, labor-intensive process. It typically involves:

1. Sifting through medical notes to locate specific fields.
2. Manually entering this information into spreadsheets.
3. Repeating this process for hundreds of patients.

This approach has significant downsides.

It's slow and labor-intensive. Extracting and inputting data takes considerable time. A single patient may have hundreds of notes and a project may require a hundred or more extracted fields. This is many hours of work for a single patient.

It burdens medical research with high cost. Skilled personnel are usually required for chart abstraction, which often makes the process prohibitively expensive. Many research projects face staffing challenges or high costs.

Complex variables require managing subjectivity. Different reviewers may interpret fields inconsistently, requiring adjudication that adds more time and cost. The semantics of cancer progression, for example, is complex and two human reviewers often disagree.

Inflexibility limits research. Researchers must define their data fields upfront. Since many projects learn more about their data and protocol as the data is abstracted, this limits the potential for iterative discoveries.

Although tools like Natural Language Processing (NLP) can speed up the process, they introduce new challenges, such as technical expertise requirements and potential errors in data extraction. The result? Many researchers remain stuck in the old, inefficient methods of chart abstraction, delaying breakthroughs in areas like cancer research and clinical trials.

To truly accelerate research and improve care, we need a better solution.

‍

Brim is Designed to Democratize Chart Abstraction

‍

Enter Brim—a platform designed to revolutionize chart abstraction and make it accessible to anyone in medical research, clinical trial recruitment, registry creation, and beyond.

• Fast and Scalable: Brim leverages the power of large language models (LLMs) alongside an intuitive user interface to reduce chart abstraction time by up to 80%. Researchers can extract and structure data at scale quickly and affordably. Brim can handle complex variables and extract data from many notes with ease.
• Flexible: Brim enables iterative workflows. Researchers can extract data, refine protocols, and re-extract easily. This flexibility supports exploration and learning, turning data extraction into an adaptive and insightful process.
• Collaborative: Brim is built for teamwork. Features like permissions management and up-to-date protocols allow for smooth collaboration on complex or subjective data fields. The platform is user-friendly, requiring no technical expertise, so entire research teams can contribute.

By combining speed, flexibility, and collaboration, Brim addresses the major pain points of traditional chart abstraction.

‍

Better Chart Abstraction Will Unlock Research

‍

At Brim Analytics, we think of chart abstraction tools as the "picks and shovels" of modern medical research, enabling professionals to uncover the "gold" of valuable insights. We believe that better tools for chart abstraction will have a profound impact, including:

1. Accelerating research discoveries at lower costs: Faster chart abstraction means quicker clinical trials, new treatments, and improved patient care.
2. Enhancing medical education: Students can spend more time on analysis and less on tedious data entry, improving their training.
3. Expanding access to research: By lowering costs and technical barriers, Brim democratizes structured data, enabling more researchers to produce high-quality work.

‍

Conclusion

‍

Chart abstraction is essential for transforming unstructured medical data into actionable insights, but traditional methods are too slow, costly, and rigid to meet the demands of modern research. Brim Analytics is changing the game by making chart abstraction faster, more flexible, and accessible to all.

With Brim, we aim to empower medical researchers, students, and clinicians to spend less time on data extraction and more time on discovery. Together, we can unlock the potential of medical data, driving innovation and improving outcomes for patients worldwide.

‍