Zheng H

A comprehensive electronic medical record (EMR) represents a rich source of information that can be harnessed for epidemic surveillance. At this time, however, we do not know how EMR-based data elements should be combined to improve the performance of surveillance systems. In a manual EMR review of over 15 000 outpatient encounters, we observed that two-thirds of the cases with an acute respiratory infection (ARI) were seen in the emergency room or other urgent care areas, but that these areas received only 15% of total outpatient visits. Because of this seemingly favorable signal-to-noise ratio, we hypothesized that an ARI surveillance system that focused on urgent visits would outperform one that monitored all outpatient visits.

A comprehensive electronic medical record (EMR) represents a rich source of information that can be harnessed for epidemic surveillance. At this time, however, we do not know how EMR-based data elements should be combined to improve the performance of surveillance systems. In a manual EMR review of over 15 000 outpatient encounters, we observed that two-thirds of the cases with an acute respiratory infection (ARI) were seen in the emergency room or other urgent care areas, but that these areas received only 15% of total outpatient visits.1 Because of this seemingly favorable signal-to-noise ratio, we hypothesized that an ARI surveillance system that focused on urgent visits would outperform one that monitored all outpatient visits.

As the electronic medical record (EMR) market matures, long-term time series of EMR-based surveillance data are becoming available. In this work, we hypothesized that statistical aberrancy-detection methods that incorporate seasonality and other long-term data trends reduce the time required to discover an influenza outbreak compared with methods that only consider the most recent past.