Unlocking unstructured clinical data
Background
Today’s healthcare environment generates terabytes of clinical data, which is now accessible to clinicians and researchers. Most of this healthcare data, estimates range from 85 to 95 percent, is in unstructured formats such as free text. This huge volume of data, spread across different systems, presents a unique opportunity to generate insights into daily clinical practice and the outcomes of different therapeutic strategies. Clinical researchers who want to reKuse this data for hypothesis generation or the discovery of a set of patients for a clinical trial, for example, need a way to mine it to discover if it is of interest or not, even if they don’t know exactly what they’re looking for. In effect they need the data to speak to them, to tell them what is hidden inside and may be important. New text analysis technology exists to enable such discovery.
InterSystems iKnow text analysis
iKnow is a bottomKup text analysis technology that shows users what is important in bodies of free text by identifying multiKword concepts and the relationships between them, without requiring any predefined knowledge about the text’s subject. On top of this indexing functionality, a number of analysis capabilities are built including support for matching against existing ontologies, intelligent browsing and text categorization.
Use Cases to be demonstrated
We will demonstrate how to leverage unstructured data and normalize clinical data to support the discovery and analysis of dynamic populations/cohorts. The technology has been used in Cancer Registries and a number of research projects, both academic and with public and private organizations (e.g. PASTEL at University Hospital Antwerpen (UZA , leveraging i2b2 capabilities at University Hospital Brussels (UZB).