We developed a way to store, assemble, and visualize medical data algorithms for Outcomes Insights.
As a consulting firm specializing in biomedical data analysis, Outcomes Insights sought to expand the capabilities of their services through a custom web application. At its core, the platform would help healthcare researchers translate natural language questions (How many diabetes patients had outpatient surgery in the year leading up to their diagnosis?) into something machine-readable.
Successful translation of these questions into algorithms would open the next door: Assembling data sources, settings, and algorithms into a study cohort in record time.
- UX Design
- UI Design
- Data Visualization
- Ruby on Rails
Prototype to Production
Outcomes Insights had a handful of prototypes and a stellar founding customer, and it was time to pull all of their ideas together into a polished version one.
Complex Question, Simple Chart
Converting medical research questions into SQL is as tricky as it sounds. A successful interface would handle most of those complexities behind the scenes and empower more potential users.
For the non-intuitive portions of the application, a toolbox of maps, documentation, and defaults would go a long way towards helping users (new and old) out.
We kicked things off with a crash course on ConceptQL, Outcomes Insights’ high-level language used to define algorithm SQL. Catching up with their standard was our trailhead for translating over to a guided interface with the same features.
Started From the Bottom
The first piece of the puzzle sat on the bottom of the stack: The algorithm builder, a flowchart-esque combination of algorithm possibilities. With that successfully solidified, we were freed to move up the feature chain towards directories and study builders.
The Visualization Gamut
Algorithms with (potentially) hundreds of nodes made most of this phase center around prototypes. Interactive trials were moved into the browser as quickly as possible for team feedback.
Jigsaw like the saw, not the puzzle. We cleaned up the core concept and underlying colors to help the product stand on its own beside the consultancy.
Keeping the Interface Clean
The lion’s share of project resources went toward development, but that didn’t detract from multiple interface iterations as new features spun up. Interactive timelines and documentation helped communicate difficult concepts.
D3 and Ember Trees
Pairing Ember, D3, and JSON, the algorithm builder was part flowchart, part map. The SVG foundation made zooming and panning seamless, and allowed the completed project to be exported as an image.
A Repository of Algorithms
Above the algorithm builder sat Jigsaw's algorithm repository: As the name implies, the underlying algorithms were organized into a directory to support metadata, filtering, and study selection.
Combining into Cohorts
The top of the pyramid was reserved for the Study Builder. Here, researchers combined algorithms and myriad settings to define a medical cohort.
Three Ideas, One Application
For the first time, Outcomes Insights’ custom take on assembling medical study data could be defined end-to-end from within the product.
Beyond being usable at launch, the platform supported and received continued development around new features and configuration options.
Across the Finish Line
The completed product came with an improved means of demoing to potential customers, soon adding Amgen and the National Cancer Institute to the software.
“The entire Envy organization are excellent partners, and we look forward to working with them in [the future].”Mark Danese