Identifying Clusters of Rare and Novel Words in Emergency Department Chief Complaints

Abstract

How do you find something when you don't know exactly what you are looking for? This is a common challenge in surveillance. Here, we present a method that supplements current syndromic surveillance efforts by relaxing some of the requirements to predefine syndromes or symptoms of interest. It looks for words in free text fields of healthcare encounters, such as emergency department chief complaints, which have either never occurred before, or which appear much more often in the current interval than they had in the past. It also applies constraints on how closely those encounters occur in time for further specificity.