Extracting Qualitative Patterns from Electronic Records (EQsPERt)
Using principles of qualitative temporal reasoning, EQsPERt models longitudinal patient records as graphs of temporally-connected events. EQsPERt feeds these graphs as input to learning frameworks to detect and predict patterns of similarities between patients with similar phenotypes.

The current uses cases are:

  • mining the Case Register Interactive Search (CRIS) database for patterns of comorbidities and drug-drug interactions from the records of psychiatric patients experiencing adverse drug reactions,
  • detecting patterns leading to mortality in septic patients admitted to a critical care unit
  • learning data-inspired severity scores in critical care patients.

Part of: