Megan Elizabeth Ross MPH, Presenter: Nothing to Disclose

Debra Sue Dyer MD, Abstract Co-Author: Nothing to Disclose

David Augustine Lynch MD, Abstract Co-Author: Research support, Siemens AG Scientific Advisor, Perceptive Informatics, Inc Consultant, Actelion Ltd Consultant, InterMune, Inc

Elizabeth Kern MD,MS, Abstract Co-Author: Nothing to Disclose

As we enter the lung cancer screening era, serial evaluation of indeterminate small lung nodules by repeated chest CT scans is of growing importance.  Patients and providers need systematic methods to track the need for follow-up scans. We evaluated two text-based systems for tracking patients with lung nodules.

Systems were built in to the radiology dictation system as macros, to be added to the CT report at the discretion of the radiologist. System 1 presented a choice of 1 of 5 phrases, with blank spaces to be filled in by the radiologist. System 2 presented a choice of 1 or more of 8 specific phrases, including a cryptic code and characterization of the nodule (solid versus ground glass).Computer queries were compared to manual audit (reference standard) of 389 (System 1) and 256 (system 2) CT reports. Ambiguous cases were adjudicated by experienced thoracic radiologists. System 1 was compared to System 2.

The sensitivity of system 1 and system 2 to ascertain cases by computer query, using word search for the phrase or code, was 63% and 92% respectively, p<.0001. The specificity of each system was 100%. The capability of system 1 to accurately specify the time frame for follow up was 94%, while that of system 2 was 100%. System 2 identified that 9% of patients with indeterminate lung nodules had ground glass nodules, a feature not possible with System 1.Radiologists identified a higher percent of cases using System 2. The major barrier to case identification was failure of the radiologist to insert the phrase. Ad lib editing of the phrases by the radiologist in system 1 was a barrier to ascertainment of follow-up time. System 2 had the added advantage of coding the character of the nodule. Specificity in both systems was excellent, a quality essential for patient-level tracking and outreach intervention.

A system using cryptic phrases to identify and track patients with lung nodules was well accepted by radiologists, had high sensitivity and specificity, and yielded information on the character of the lung nodule as well as follow-up recommendations.  

A lung nodule registry created through data mining of cryptic phrases from radiology reports is feasible and suitable to track patients needing follow-up exams for lung nodules.

Ross, M, Dyer, D, Lynch, D, Kern, E, Creation of a Lung Nodule Registry: Comparison of Natural Language and Pre-specified Phrases for Computerized Tracking of Patients with Indeterminate Lung Nodules.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12024370.html