08/02/2024 Professor Harnett
In 2015 the UC Department of Biomedical Informatics was approached by a nascent company out of Cambridge, Massachusetts. The concept was to share with them de-identified clinical and demographic factors of the entire hospital’s patient population.
In 2012, BMI had built out a patient cohort identification tool called ‘Informatics for Integrating Biology & the Bedside’, commonly referred to as i2b2, a research project out of Harvard University. Since its release in the early 2000s, the open-source clinical data warehousing and analytics research platform scaled to hundreds of academic health centers.
The new company out of Cambridge, called TriNetX, had developed an impressive harmonization platform that could accurately characterize clinical attributes that could scale. While there were numerous standards in place even back then such as ICD-9, CPT, and others, there were still proprietary tools and partially embraced standards that made federated searching problematic.
TriNetX solved this problem and proceeded to onboard institutions with instances of i2b2 because they already demonstrated the need for a user-friendly, self-service research subject cohort analysis tool.
The win-win for the institutions, called the Healthcare Organizations, (HCOs), was that deploying the TriNetX platform not only provided a superior internal cohort identification tool by supplying data to the TriNetX data platform and utilizing the web-based tool, it also was available to pharmaceuticals and Contract Research Organizations (CROs) who desperately needed quality subjects for clinical trials.
This is the business model for TriNetX; aggregate data from multiple HCOs and sell subscriptions to pharmas and CROs so they could in a single interface, in sometimes minutes, enter specific inclusion/exclusion criteria in an intuitive interface and click a button to execute the query. A list appears of HCOs who had the patient populations they needed. The HCO with the most matching patients appeared on top, this also set the parameters for multisite studies, so they knew who to contact.
TriNetX brought industry-sponsored studies to UC Health and at the same time, provided a super cool, and wildly popular self-service tool for internal queries, hypothesis generation, analytics, and quickly identifying local patients who meet eligibility criteria.
Because TriNetX is de-identified using federal guidelines, researchers could not identify patients. But, with proper Institutional Review (IRB) approvals, they could request from BMI to reverse-engineer the synthetic IDs within TriNetX as BMI was the institutional Honest Broker. This service and subsequent process were very easy as there was a reverse mapping sequence derived from the extraction, transformation and loading (ETL) functions that fed the warehouse. For $194, investigators could narrow their queries to specific criteria, save and ‘star’ those queries and request the results to be reverse-engineered to get identifiers for things like recruitment letters or Epic MyChart messages.
Jump forward about four more years and UC BMI joined the TriNetX Research Network where Real World Evidence (RWE) analysis was possible. Instead of the limited population of a single local site, the Research Network with a denominator of over 130 million (compared to 1 - 2 million) is presented. This meets the federal regulations definition of “research”: “A systematic investigation, including development, testing, and evaluation, designed to develop or contribute to generalizable knowledge”. This is not patient recruitment, this is research on a given phenotype.
Now that we have solved patient recruitment problems, we are now looking into a more complex challenge of reliable time travel. The first meeting will be yesterday at noon.
[TriNetX reference used with permission]