Open Clinical Analysis with MITRE Part 1
Open Clinical Analysis
There is tremendous innovation going on in the world of health analysis and health applications. From companies like ours that are using data science to learn more about patients to consumer facing apps that help manage disease or engage patients with providers. Even with all of this innovation, it can be difficult to see how all these new capabilities should knit together and communicate. There is no healthcare equivalent of the ubiquitous ACH transactions used in banking, at least not yet. ;)
This is challenging for the developer or data scientist because it is often difficult to simulate a health care data environment. What are the transactions involved? What are common metrics of data quality? Can I get simulated data to use in unit testing or load testing?
I want to call attention to the great work going on at the MITRE CMS Alliance to Modernize Healthcare. Perhaps more than any other group, MITRE has pushed forward in establishing not just standards (like FHIR and C-CDa) but real software that uses those standards effectively.
MITRE Patient Toolkit from MITRE.org
With the team at Algorex, I got to visit MITRE, two weeks ago, as part of their tech Tuesdays. On a tour of their labs, the developers took time to show off some of the amazing health technologies they are working on for various stakeholders or just for research. These include open-source solutions to all of the questions listed above.
I have encountered other MITRE projects in the past including popHealth, which is now under the maintenance of Open Source Electronic Health Record Alliance. However, their current slate of projects better adopt the open-source ethos that libraries should “do one thing and do it well”.
- Need realistic clinical data for testing? Synthea
- Need reference services for using FHIR standards? goFHIR
- Need a engine that can automatically run CMS’ quality measures? Cypress and others
- Need a framework for evaluating patient matching systems? Patient Matching
- Need a proven logical model for querying clinical data? Clinical Query Language
Yea, MITRE has completely open-source packages for all of this and more. Many of these packages have now been transitioned to community maintainers outside of MITRE which is great.
Ok so a lot of praise. I should caveat. MITRE is a research organization and some of these projects are proof-of-concept or use standards that are less than ideal. But the point is that you can look at them for proven ideas. Even if you want to be critical of implementation details that’s kind of the point, they give the health development community something to compare against something innovate past.
Tour of MITRE Stack
Ok, so using MITRE can we stand up a complete population health stack with synthetic clinical data, database, calculated measures, and a front-end to visualize. The answer is yes. I have put directions in part 2 of this series. It maybe wasn’t the easiest software install but it was 100% free and open-source.
When open-source software contains this capability, it helps move the entire health ecosystem forward. Being able to run a quality measure should be commodity low-cost functionality not highly custom siloed software. MITRE is helping make that a reality. Its part of a trend in smoothing over the most challenging obstacles in healthcare IT. If you are building a health app, MITRE might give you clinical quality and decision support functionality or test data. A company like Algorex can provide utilization and risk stratification libraries. Furthermore, Datica can provide a safe HIPAA compliant cloud architecture. These massive barriers to innovative health applications are getting torn down.
Back to my challenge: synthetic patients, clinical database, running measures, visualized. Take a look at the results here
Here is a screenshot of the CMS IVD Measure running on MITRE’s stack.
with patient drill-down
No its not ground breaking functionality in pophealth, but its an amazing foundation to power future innovation. The directions for the install are in part 2