The CC text field is a rich source of information, but its current use for syndromic surveillance is limited to a fixed set of syndromes that are routine, suspected, expected, or discovered by chance. In addition to syndromes that are routinely monitored by the NYC Department of Health and Mental Hygiene (e.g., diarrhea, respiratory), additional syndromes are occasionally monitored when requested by outside sources or when expected to increase during emergencies. During Hurricane Sandy, we discovered by manual inspection of data for a few EDs an increase in certain words in the CC field (e.g., 'METHADONE', 'DIALYSIS', and 'OXYGEN') that led to the creation of a 'needs medication' syndrome. Current syndromic surveillance systems cannot detect unanticipated events that are not defined a priori by keywords. We describe a simple data-driven method that routinely scans the CC field for increases in word frequency that might trigger further investigation and/or temporary monitoring.
Objective
To detect sudden increases in word frequency in the Emergency Department (ED) syndromic chief complaint (CC) text field.