Healthcare

The SAS Cohort Builder is a high-performance analytics platform designed to ingest, cleanse, and standardize massive volumes of real-world data from diverse sources like electronic health records and insurance claims. It empowers researchers with a code-free, drag-and-drop interface to define complex patient cohorts using advanced temporal logic while visualizing the impact of inclusion criteria in real time. Beyond simple selection, the system provides robust analytical tools—including prebuilt predictive models and SAS/R integration—to generate deep insights into treatment regimens, health costs, and gaps in care. Built for speed and agility, the platform processes millions of data rows in seconds, ensuring that feasibility studies and longitudinal analyses are both efficient and actionable.

Project Type

UX Design, Information Architecture, User Research

Team

Development, Quality Assurance, Product Management, Documentation, and Presales

My Role

Senior User Experience Designer

Company

SAS

Timeline

Multi-release 5 year project

Stakeholders

Product Management, Development, Quality Assurance, Documentation, Presales, Health and Life Sciences Division

Project Brief

Situation

This UX designer joined the development of an existing cohort builder to improve its usability, workflow, and information architecture. Because the original version relied on custom code rather than the standard company design system, I worked to align its functionality and aesthetic with the broader product suite.

Task

Over a series of releases, upgrade the usability, workflow, features and functionality of the application.

Action

Create a UI that would allow users to quickly assemble cohorts that could be saved and modified for later reuse, allow for rapid comparison of available data sources, as well as see the results of adding expressions to filter and preview results by demographics, diagnosis, procedures, lab results, prescriptions, and other criteria.

Challenges

During its evolution, the application faced three primary challenges:

Poor usability: The interface was sluggish and featured confusing labels, lacking a clear workflow for building cohort filters.
Disconnected data: The search function for healthcare codes was not integrated with a live, up-to-date health language database.
Data opacity: The tool lacked preliminary profiling, forcing users to select data sources without first understanding their specific characteristics or quality.

Result

Over a series of releases the Cohort Builder was successfully modified to create a powerful and quick tool for cohort generation. Workflow was dramatically improved, and the addition of an actively managed health language source kept diagnostic, procedural, and laboratory and diagnostic codes current for user selection during cohort creation. Time to assemble a cohort decreased by 50-70%, depending on the cohort complexity, inclusion/exclusion criteria, and the use of temporal logic.

Case Study

Context

The initial version of this application was created by the development team as an experimental interface to test the ability of an application to access large healthcare data sources and filter the data to create cohorts that would be easy to create and store for later reuse.

Problems

Problems with the first user interface shown above included poor usability, disconnected and opaque data sources, poor information architecture, confusing terminology and workflows, accessibility issues, and the use of non standard company colors and components.

Research

User research of the newly redesigned user interface took place in conjunction with the Product Management and Presales organizations utilizing subject matter experts in those groups to test the usability, workflow and terminology to be in line with customers expectations and comparable to competitor cohort creation tools

Process

Each new release added new and crucial functionality. New versions began to use company standard color palettes and components, eliminated accessibility issues, connected the tool to an actively managed health language code selection repository, enabled a drag and drop interface for inclusion and exclusion criteria including temporal relationships, and allowed users to quickly preview data sources before cohort assembly.

Conclusion

The new versions of the application were well received by the Product Management and Presales teams. Customer feedback was very positive, including direct feedback from customers at company annual conference product demonstrations. Many competitors have cohort builders as part of larger healthcare data tools, and the Cohort Builder functionality and interface was planned to become a part of a new larger application for healthcare data ingestion including new AI functionality for cohort creation.