Particle's first hackathon was a success!

As part of our recent company onsite, we did our first-ever hackathon. The hackathon was cross-functional, with folks from engineering, data science, product, and solution engineering participating on various teams.  The participants were asked to work on projects that would either demonstrate the capabilities of Artificial Intelligence, or explore additional capabilities of our data.  
‍

A few project highlights worth mentioning: 

Enhanced Record Locator Service (RLS): We take pride in our ability to find patient records via our in-house Record Locator Service, and in bake-offs we typically perform better than our competitors. There are a lot of nuances in how Health Information Network directories are managed and mapped, which directly impacts how we search for patient records. One team applied advanced analysis and data science techniques to improve our ability to locate additional data. They were able to achieve improvements of up to 20% for some patients, and uncovered several ideas that we are excited to follow-up on.

‍Linking Data: Another team used sophisticated data science techniques to prototype a way to link related data more effectively, which can help reduce data quality issues.  

‍Extracting Value from PDFs: About 10% of the data on the Healthcare Information Networks is in PDF form, which means a lot of information is not easily available for structured processing. This data often includes important clinical information like hospital discharge summaries. Teams experimented with several approaches, including using Generative AI to parse out structured information from the PDFs. The results were very promising! 

‍SMART on FHIR: This is a technology standard that makes it easier for us to surface our data at the point of care.  One team created a prototype of an app that is integrated into OpenEHR and our API, demonstrating how our data can be surfaced directly within existing clinical workflows.  

‍Synthetic Clinical Data: One last team focused on improving the data available in our sandbox.  While we already have test data that captures the complexity of CCDAs, it wasn't designed to tell a narrative that's meaningful to clinicians. To address this, the team created synthetic patients with known clinical conditions, allowing us to showcase the full potential of the data we receive from the networks. One of our clinical experts developed datasets that would be relevant for providers. For example, a newly generated synthetic dataset showcased a patient with chronic heart failure. Within this patient’s data, we can see: 

• A chronic disease that is worsening over time
• A specialist visit
• Active medications and changes in medications over time
• Trends in labs and vitals
• Hospital discharge summary
• Open care gaps (overdue for a colonoscopy)

The team used generative AI tools to create CCDA files that we can now use in many different ways to illustrate the power of our data and platform capabilities, especially how the data shows up clinically.   

It was truly great to see the team accomplish so much in such a short time, and truly speaks to our culture of innovation and dedication to Particle’s mission.  There were many great projects, and about half of the projects will influence our roadmap in ways big and small!  It's also worth noting that many team members gained hands-on experience with Generative AI tools, which are becoming increasingly important in our work.

Thank you to everyone who participated, offered ideas, and cheered!  Can’t wait to do this again!