By Benjamin Ross

January 9, 2018 | TriNetX, a Cambridge, MA-based global health research network founded in 2013, has recently unveiled their new Natural Language Processing (NLP) service for general availability. The NLP service uses algorithms to extract critical information from clinical notes and reports in order for CROs and healthcare organizations to assess study feasibility, protocol design, site selection, and subsequent identification of patients for clinical trials.

David Fusari, CTO and co-founder of TriNetX, told Clinical Informatics News that, in the 22 years since being in healthcare, he’s seen a lot of patient information locked in files, whether that be a primary care doctor’s file or records a patient may collect from previous clinical trials.

For example, Fusari said, information like left ventricular ejection fraction—a measurement of how well the heart is pumping out blood—might be embedded in a cardiology report and not listed in an electronic medical record (EMR), hindering the researcher’s ability to accurately identify patients for a clinical trial.

The TriNetX NLP service is based on technology from Averbis, a text-mining and machine-learning company headquartered in Germany. TriNetX reached out to Averbis in 2016 and began a partnership in early 2017. “TriNetX chose Averbis as its NLP partner because of Averbis’ experience applying NLP within healthcare, the accuracy of their solution across various data domains including oncology, and the ability to work with multiple languages,” according to a TriNetX press release.

Fusari explained that the NLP service understands a wide range of clinical terminologies, including the recognition of negated terms.

TriNetX’s NLP service reviews structured data such as demographics, diagnoses, procedures, medications, labs, genomics, and deep oncology data, as well as data taken from clinical documents like discharge summaries, radiology reports, and pathology reports. This information is crucial when attempting to identify candidates for clinical trials, Fusari said.

Once NLP has extracted the desired information from the clinical notes and reports, that data is made available through TriNetX Live, the company’s cloud-accessible health research platform. Healthcare organizations who are members of the TriNetX networkcan also import the structured data into their own data warehouse for their own research purposes. According to Fusari, TriNetX covers almost 67 organizations, each one of them storing data. That’s close to 100 million patients worldwide being covered.

The purpose of the NLP service is to lead pharmaceutical companies, CROs and healthcare groups to the right patients for their clinical trials, minimizing the time and resources typically used during that process. According to a TriNetX press release, “Members access TriNetX Live to analyze patient populations and perform ‘what-if’ analyses in real-time. Users of the platform are presented with aggregate views, but each data point in the TriNetX network can be traced to healthcare organizations who have the ability to identify individual patients, allowing clinical researchers to develop virtual patient cohorts that can then be re-identified for potential recruitment into a clinical trial.”

TriNetX has 10 global locations, and is planning to expand in the coming years and months. Fusari said the idea is to bring together data from both structured and unstructured content that really drives a comprehensive view of the patient. He went on to say that the question that CROs and sponsors ask themselves most often when designing a clinical trial is, “Do the patients we want for this trial exist in the real world?” TriNetX hopes to provide a definitive answer to that important question.

Editor’s Note: TriNetX will be a premier sponsor and exhibitor at the Summit for Clinical Ops Executives—SCOPE—in Orlando, Fla. from February 12-15, 2018.