Arun Krishnaraj MD, MPH, Presenter: Research support, General Electric Company

Hani H. Abujudeh MD, MBA, Abstract Co-Author: Research grant, Bracco Group Consultant, RCG HealthCare Consulting

Mitchell A. Harris PhD, Abstract Co-Author: Nothing to Disclose

Neeraj Joshi MS, Abstract Co-Author: Nothing to Disclose

Abraham Lin BS, Abstract Co-Author: Nothing to Disclose

Sarita Nair MS, Abstract Co-Author: Nothing to Disclose

Garry Choy MD, Abstract Co-Author: Nothing to Disclose

Michael Ethan Zalis MD, Abstract Co-Author: Research grant, General Electric Company

Identification of patients with contraindications to entering the high field zone of MRI is challenging given rising volume and time pressures associated with more rapid imaging protocols. Moreover, current processes of screening are subjective and fraught with significant variation due to human error. In order to address these challenges and improve patient safety, we have developed an application based on an ontology driven search tool (Queriable Patient Inference Dossier, QPID) for the EHR which objectively identifies patients with contraindications to MRI. The QPID application then sends automatic advanced alerts to MRI technologists concerning these patients. In order to validate the accuracy of this tool we compared the results of QPID versus a manual review.

MRI contraindications were classified as ‘red’(absolute) vs. ‘yellow’(relative contraindication) according to published criteria for MRI safety labeling and department protocol. Search modules were developed on the QPID platform which identified ‘red’ and yellow’ concepts, terms, and phrases in the EHR corpus for each patient investigated. From the safety records of our institution, we created an enriched cohort of 83 patients who underwent MRI exam, 23% (19/83) of whom had documented ‘red’ or ‘yellow’ contraindications to MRI. QPID retrospectively reviewed the EHR data for each patient and generated ‘red’ or ‘yellow’ warnings accordingly. A manual review of the QPID results was subsequently conducted as the gold standard and results were categorized as True Positive (TP), False Positive (FP), True Negative (TN) and False Negative (FN), and tabulated summary statistics for each category of contraindication were calculated. The number of cases requiring QPID to search for longer than 30 seconds per patient was also noted.

For ‘red’ contraindications (n=14), QPID results were TP=14, FP-3, TN-66, FN-0 (sensitivity-100%, specificity-95.7%) and for yellow indications (n=5), QPID results were TP- 5,FP- 23,TN-55,FN-0 (sensitivity-100%, specificity-70.5%). Per patient search time for each ‘red’ or ‘yellow’ QPID module was uniformly less than 30 seconds.

QPID is highly sensitive and specific for detecting absolute MRI contraindications but has moderate specificity for relative contraindications.

Based on its performance, QPID may serve as a rapid screening tool for identifying patients with absolute contraindications to MRI.

Krishnaraj, A, Abujudeh, H, Harris, M, Joshi, N, Lin, A, Nair, S, Choy, G, Zalis, M, Automated Identification of Patients with Contraindications to MRI: Validation of an Ontology Driven Search Tool for the EHR.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11011384.html