TriNetX

TriNetX is a self-service tool that allows researchers to determine approximately how many patients in the Clinical Research Enterprise Warehouse meet certain criteria. Researchers use an intuitive user interface for selecting query criteria from UC Health’s clinical (Epic) repository.

TriNetX can be used to answer questions such as “How many patients do we have that have a diagnosis of atrial fibrillation who are taking anticoagulants between the age of 38 and 89 who are also on Warfarin?” Search criteria can include patient demographics, ICD coded diagnoses, laboratory test results and medications.

TriNetX is covered by a UC-IRB protocol and HIPAA waiver although the data is fully de-identified. To accomplish this, a series of methods are used to “cloak” data such as date grouping. This means the queries return an approximate count of the number of patients matching the search criteria, rather than the actual count. All queries you create are stored under your profile. When you are ready for a deeper dive, the Center for Health Informatics (CHI) — as the University’s Honest Broker — can access the warehouse and extract actual numbers and identified data based on IRB approvals.

Researchers can access TriNetX 24/7 to extract data and evaluate cohorts. Whether you are looking for particular clinical relationships or trying to identify if there are enough potential patients available for recruitment, TriNetX offers a flexible and dynamic tool for initial analysis.

To best use this tool, users need to understand where the data came from and in what context. The best example of this is the source of the diagnosis codes. Currently, the ICD codes come from the problem list, encounter and billing codes. Because of the complexity of Epic and clinical workflow, source of diagnosis codes vary. By including all three data elements, we find patients with a diagnosis code that could have come from all or any of those three areas. This means if a patient was ever diagnosis with “pancreatitis” they will be counted although they may no longer have pancreatitis. The metadata (data description section) explains the source and description of the data elements.

Note: The data goes back to the go-live of Epic in 2012. However, some data was back loaded from the legacy systems (Centricity and LastWord) so some of the data is older.

Issues and Known Bugs (persistently important)

TriNetX was created by a third-party company to facilitate patient cohort queries by the pharmaceutical industry. Because there are scores of other healthcare organizations in the TriNetX network, TriNetX provides a federated query for pharmas to simultaneously search across all the member organizations to identify sites for clinical studies. UC is part of this network to attract clinical research from industrial sponsors. An added benefit is we can use TriNetX to search our own patient population using this new tool with features beyond what i2b2 offered.

User Interface

To access TriNetX, you need a computer with an updated modern web browser and your assigned credentials from the CHI. After you are assigned credentials, be sure to watch the video tutorials before you begin (at least the first three).

We recommend these training videos (all videos 2–7 minutes long)

• Introduction to TriNetX

• General Navigation

• Terminology and Adding Terms

• ICD-10

• Working with Terms

• Qualifying Population

• Temporal Events

• Patient and Site Count

• Labs

Note: If you are interested in newly incident cases, the date function works well for acute (non-chronic) illnesses such as pneumonia, less so with chronic illnesses such as diabetes or complex diseases. The reason for this is the incidence of such diseases is much harder to index in the system. For example, a new patient may come into the system with diabetes diagnosed years ago, but diabetes is entered into the encounter record or problem list at the time of visit, this clearly does not tell us when the patient was originally diagnosed. From a prevalence standpoint, we can more accurately count how many patients simply have diabetes.

Getting Help

If you feel the query is too complex or are not sure what you want, contact the CHI for assistance.

Data Description

The TriNetX database includes clinical data from the Epic Electronic Health Record (EHR) for patients seen at UC Health and affiliated entities.

Epic: Dates of Encounter

Outpatient: July 10, 2012
Inpatient: October 29, 2012

Hospitals

• University of Cincinnati Medical Hospital

• West Chester Medical Center

• University Pointe Surgical Hospital

• Deaconess

• The Drake Center

Clinics

• UCH Primary Care

• UCP clinics

• University of Cincinnati Medical Hospital clinics

• West Chester Medical Center

• The Drake Center

Patients

Included

Patients seen at the participating UC Health hospitals and outpatient sites

Excluded

Former Health Alliance hospitals (patient data) no longer part of UC Health

Data from the Lindner Center of Hope

UC Students (FERPA)

Data before Epic go-live

Refresh Frequency

TriNetX will be refreshed bi-monthly. Expect the query results you find to be current to within 60 days. (Plans are to increase frequency to monthly)

HIPAA Compliance and De-identification

To be considered not electronic protected health information (ePHI) the UC implementation of TriNetX is de-identified. The UC instance of TriNetX employs a de-identified data set that excludes all of the eighteen HIPAA identifiers. This de-identification includes the following trifling modifications to the underlying data, called obfuscation:

• All ages over 89 are indicated as ≥90.

• Results will show ±x for additional obfuscation.

• Counts of less than 10 will show as <10.

The detailed, absolute data can be extracted through the Honest Broker, CHI.

The following data elements are included in the current UC release of TriNetX.

To use TriNetX

• Click: Access TriNetX.

• Read the User Agreement that is displayed and click “I Agree”.

• A User account will be created and credentials emailed to you from the System Administrator.

• After you log in, click on the Help menu in the upper right and watch the short training videos (at least: Introduction to TriNetX, General Navigation and Terminology and Adding Terms).

• Begin using the TriNetX interface to create your queries. Queries time out after 5 minutes (very few should ever take that long).