CDM Seminar Series 2002-03


Reporting, Investigating and Analyzing Medical Error and Adverse Events:
GALEN Common Reference Model-Based Approach
Meghan Dierks, MD
Instructor, Harvard Medical School
October 28, 2002

Medicine is a complex, safety-critical and highly interactive system. Despite efforts to provide safe, effective care, adverse events still occur clinicians make diagnostic and therapeutic errors, system constraints impact the coordination and delivery of care and patients suffer unexpected complications and injuries. Ideally, we strive to learn from past errors and events and develop operational improvements by reviewing cases reports derived from a number of sources:
· Incident reports filed by staff when the event occurs
· Departmental or divisional Quality Assurance (QA) Audits
· Investigation of a ‘sentinel’ event (e.g.: death or suicide)
· Root Cause Analysis required by the Joint Commission on Accreditation Hospital
· Patient-authored complaints to the patient advocacy department
· Malpractice Claims pending or completed

Despite the rich source of data from these reports, they have not been as useful as we had hoped in identifying common factors or trends, developing causal models or hypotheses or providing feedback to hospitals and practitioners so that they can reduce the incidence or severity of future events. The following factors may contribute to underutilization of these data:
· Poor methodological support during the report construction and/or analysis
· Analyst subjectivity
· Poor support for error prediction
· Focus on accidents and not incidents or near misses
· Individual operator/system focus
· Difficulty reaching consensus on the contextual sources of latent failures

To address these issues, Aziz Boxwala and Meghan Dierks, working with the Harvard Risk Management Foundation, are developing an ontology to classify and analyze heterogeneous sources of data on medical adverse events. They are using a compositional formalism based on the GALEN medical language model to develop a subsumption hierarchy and sanctioning statements to represent logical relationships between events and contributing factors. Early results and application to claims, incident reports and Root Cause Analysis reports will be presented.


CDM Abstract Archive