Ontology-driven search tool identifies MRI contraindications, medical history

An ontology-driven search tool is being deployed at Massachusetts General Hospital to rapidly search existing electronic medical records of patients admitted to the emergency department for important medical history and identify patients scheduled for MRI exams who've experienced prior contraindications. The objective: to improve patient care and patient safety.

Its use and how well it performs will be explained in these nearly back-to-back scientific papers by Dr. Arun Krishnaraj, an assistant radiologist and an instructor in radiology at Harvard Medical School.

The Queriable Patient Inference Dossier (QPID), a search engine that performs 74 automated structured and natural language searches customized for emergency department care, is used with the hospital's electronic health record to extract and assemble salient clinical information. Researchers conducted a study using the records of 500 consecutive patients admitted to the emergency department in 2010 to determine QPID's sensitivity and specificity.

QPID scanned each patient's medical record within 10 to 20 seconds. A manual review of 30 randomly selected patient records out of the 500 demonstrated 97% sensitivity and 99% specificity for laboratory results, and 98% sensitivity and 93% specificity for findings extrapolated from free-text records.

QPID is also being used to identify patients with contraindications to MRI exams. These are classified as absolute and relative.

The software was similarly tested using records of 83 patients, 23% of whom had documented contraindications. All the records were also manually reviewed. QPID showed 100% sensitivity for both levels of contraindications and 95.7% specificity for acute contraindications and 70.5% for relative ones. It scanned each patient record in less than 30 seconds.