Chief Complaint based Surveillance vs. Whole Record Surveillance Evaluation

Chief Complaint based Surveillance vs. Whole Record Surveillance Evaluation Peter Elkin, MD, Steven Brown, MD, Andrew Balas, MD, PhD, Zelalem Temesgen, MD, Dietlind Roedler, MD, David Froehling, MD, Mark Liebow, MD, Brett Trusko, PhD, Katie Skeen-Morris, MPH, S. Trent Rosenbloom, MD, Gail Welsh, MD, Greg Poland, MD Background: Chief complaints (CC) make up the linchpin of the clinical content being sent by Biosense. CC are often incomplete or blank, are sometimes filled with administrative data (e.g. the name of clinic seeing the patient) and typically does not have missing data that can be found in the remainder of the document. Practices are rapidly transitioning to electronic health records and this will provide the opportunity to include more data as the substrate for Biosurveillance. It remains unclear as to the level of advantage that is provided by the addition of this additional content. Methods: 28,802 Patients were tested for Influenza from 2000 – 2006. Of these patients 27,125 Patients Tested Negative and 1,677 tested positive. The entire record was parsed using the Multi-threaded Clinical Vocabulary Server (MCVS) coding the output into the SNOMED CT reference terminology. The elements of the case definition were matched with the Influenza and Control records by section of the record. Odds Ratios for Influenza were established by element of the case definition, by section of the clinical record. Results: This produces a Receiver Operator Characteristics (ROC) curve with a C-Statistic of 0.822. The same statistical method was repeated only considering the elements when found in the chief complaint section of the record. Whole record surveillance was found to be superior to chief complaint surveillance alone for predicting Influenza (C-Statistic = 0.822 as compared with 0.714, p<0.0001; Delong’s Method). Conclusion: Whole record surveillance is superior to chief complaint surveillance alone.