Brim's Approach to Chart Abstraction: the Four C's

Our Expertise in Chart Abstraction Helps Us See Differently

‍

Structured data is often the cornerstone of medical research and improved patient care. Yet, extracting that data from unstructured medical notes is anything but simple.

Dr. Dan Fabbri, founder of Brim Analytics, has firsthand experience addressing this challenge. Having worked on hundreds of clinical research projects, including NICU discharge prediction, drug surveillance, and cancer treatment response, he knows how challenging it is to transform unstructured medical notes into actionable insights - and how vital.

This need extends beyond research into critical areas such as clinical trial recruitment and registry data collection. While other teams are also working to improve chart abstraction, Brim Analytics’ deep experience in the field provides a unique and informed perspective.

At Brim, we believe effective chart abstraction should be Complete, Consistent, Cost-Effective, and Collaborative.

‍

Complete: Look at All the Data

‍

Completeness means having the capability to review the entire patient history and abstract any number of fields. Traditional manual chart abstraction is limited by the time it takes for human reviewers to read charts and locate information.

Brim makes completeness achievable through:

• Scalable LLM-powered reading: LLMs can process large volumes of medical records efficiently, suggesting structured data points for human reviewers to validate.
• Flexibility: Users can define or adjust abstraction fields dynamically, accommodating new insights as research progresses.

This approach ensures no detail is overlooked, even in large scale projects.

‍

Consistent: Extract the Same Way Every Time

‍

Consistency is crucial in chart abstraction, yet human error and variation between reviewers can make this difficult. Challenges include:

• Tedious, repetitive tasks: Monotony can lead to mistakes.
• Semantic complexity: Making decisions based on concepts like cancer progression isn’t always straightforward.
• Protocol shifts: Changes mid-project can create inconsistencies.

Brim addresses these challenges by:

• Centralizing the abstraction protocol to guide decision-making.
• Using an LLM to standardize extractions, driving consistency across reviewers and even multiple healthcare sites.
• Supporting seamless protocol updates mid-stream, enabling nimble and efficient research.

With Brim, the data is extracted the same way, every time.

‍

Cost-Effective: Fit the Research Budget

‍

Chart abstraction is traditionally labor-intensive, often costing upwards of $100 per hour for skilled human abstractors. Even when leveraging LLMs, processing millions of tokens can become expensive, especially for large projects or long medical histories.

Brim significantly reduces these costs by:

• Transforming the role of abstractors from manually coding data into reviewing and validating results.
• Optimizing LLM usage to extract information cost-effectively without compromising quality.

This allows research teams to focus on critical thinking rather than tedious tasks, maximizing both time and budget.

‍

Collaborative: Support Real, Multi-Site Systems

‍

Healthcare registries and consortiums often span multiple sites, each with unique processes and teams. Maintaining consistency in abstraction protocols and practices across sites can be a logistical nightmare.

Brim simplifies collaboration through:

• Streamlined sharing: Abstraction variables and protocols can be easily shared and replicated across sites.
• Standardized implementation: Brim ensures every site adheres to the same processes.

This collaborative approach enables large-scale research projects to operate seamlessly across healthcare systems.

‍

Conclusion

‍

At Brim Analytics, we’re reimagining chart abstraction through the lens of the Four C’s: Complete, Consistent, Cost-Effective, and Collaborative. These goals guide our innovative solutions, helping researchers unlock structured data efficiently and effectively.

By leveraging advanced AI tools and deep expertise, Brim empowers healthcare organizations to turn unstructured medical records into actionable insights that drive better research, care, and outcomes.

‍