Frequently asked questions

You can. You need a PHI-safe endpoint, and to handle a whole pile of technical challenges, including importing and exporting large datasets, prompt engineering for multiple complex variables, and understanding when and why the LLM is hallucinating answers.

Brim is designed to make it easy to work with LLMs and integrate into your existing chart abstraction process, without adding additional technical headaches.

‍

Maybe you’re fascinated by the details of RAG, context windows, and techniques to combat hallucinations – we sure are!

But Brim’s tools are designed to make Large Language Models work for you, not the other way around. Brim handles interfacing with the LLM so you and your team can focus on the decisions involved in chart review and the important analysis it enables, no prompt engineering required.

‍

Brim Analytics takes data security seriously. The Brim system is SOC 2 Type II and HIPAA compliant and designed from the ground up for the security, privacy and governance requirements for healthcare data.

Brim includes an administration management system that allows project owners to restrict access to specific users, manage permissions per project, and log all access events for HIPAA compliance. The system also is containerized to be distributed and deployed across varied environments, such as on-premise, in AWS or in Azure, depending on healthcare governance requirements. And we allow using your institutionally-approved Large Language Model for added data control.

‍

Costs depend on number of users, volume of abstraction tasks, complexity of the clinical variables, infrastructure configuration, support level, and deployment model. We work with enterprise customers to define custom pricing tailored to your needs.

To start exploring your use case, schedule a demo or email us.

Brim works across many clinical domains, including oncology, cardiology, neurology, pediatrics, and more. It is particularly helpful in settings where manual chart abstraction is frequent like registries, retrospective studies, clinical trials, quality improvement, and observational cohorts. If you tell us your specialty, we can confirm fit.

Once you engage, setup often involves environment provisioning, software deployment, initial training/validation, and configuring an initial project. The time involved for environment provisioning depends on your organization, but once that is ready, deployment can be done in as little as an hour. The Brim team typically stays engaged after that, ensuring that your team receives the training they need and you see value from your first project.

Accuracy depends on the complexity of the abstraction task, the quality of source data, and initial tuning. In pilot studies, we've seen agreement with human reviewers of 96%+, and sensitivity of 100% in some cases. We work with your team to calibrate and validate performance.

Yes. Brim can be deployed in federated or distributed architectures, respecting data sovereignty and sharing rules, so multiple sites can run Brim locally, use the same validated variables for abstraction, and aggregate abstraction outputs without sharing raw patient data.

We offer onboarding training, user documentation, workshops, ongoing support, and optional professional services for customizing abstraction models and pipelines.

Yes. Brim logs the raw evidence associated with every variable and provides reasoning behind the determination. You can inspect the chain of logic and provenance for each abstraction decision.

Yes. One key Brim advantage is flexibility in model choice. You can plug in your institutionally approved Large Language Model or use a third-party model, giving you control over cost, data, and performance.

Contact us to define a small scope for a pilot project. We’ll help you onboard, validate, and evaluate performance. After the pilot hits your desired metrics, you can expand.

Brim is EHR-agnostic; we support connecting with Brim via CSV upload, REDCap integration, and our RESTful API. This makes it easy to securely connect your system via Epic Clarity, Databricks, or your other data system. Contact us if you have specific integration questions or needs.