About CLARK

CLARK (Clinical Annotation Research Kit) is a natural language processing and machine learning software application specifically designed to be user friendly, freely sharable, and able to answer a variety of translational research questions.

CLARK takes free-text clinical notes and structured clinical data (demographics, labs, vitals, and meds) as input and classifies those notes and the associated patients based on features (i.e., words and phrases) defined by the user.

CLARK was developed through a collaboration between NC TraCS and CoVar Applied Technologies.

Who can use CLARK?

CLARK is freely available for download in the TraCS ShareHub. The latest version is CLARK 2, which offers a new interface and the ability to analyze data based on both structured and unstructured data (notes). CLARK 1 is also available for download and includes the ability to analyze data based on unstructured data.

To use CLARK, you must have a set of clinical data for CLARK to process. UNC researchers interested in CLARK may submit a consult request to learn how they can get a CLARK-ready dataset.

How can CLARK support my research?

CLARK is designed to supplement human effort and judgment in order to reduce time spent doing chart review, produce more robust computable phenotypes, and move studies to recruitment and/or data analysis faster. It can help to identify patient cohorts for a study when inclusion/exclusion criteria are framed as a classification problem.

CLARK Instructions

CLARK 2.0 Documentation
CLARK 1.0 Documentation
CLARK Conceptual Guide - Guidance for applying CLARK to your research and interpreting results

Webinars

Using CLARK 2.0 software for clinical data and notes | Nov 16, 2021

CLARK 2.0 is a user friendly machine-learning classifier designed to answer a variety of translational research questions. This version of CLARK offers a new interface with additional features including the ability to analyze data based on both structured and unstructured data.

In this session, we:
- Discuss use cases ideal for CLARK
- Discuss the FHIR data formats required by CLARK
- Walk through the process for requesting data from TraCS and CDWH
- Provide an overview of how to write regular expressions, which are a core component of using CLARK
- Demonstrate using the new CLARK interface