The use of clustering to analyze symptom-based case definitions for acute gastrointestinal illness

Gastrointestinal illness is an important public health issue. To better estimate the true level of morbidity associated with gastrointestinal illness in the community, several countries have conducted population-based studies. Unfortunately, comparing the results of such studies is complicated because the symptom-based case definitions used vary, despite the fact the studies are often aimed at evaluating the same phenomenon. This potential problem, although widely noted in the literature, has not been formally explored. The research presented here demonstrates the impact of using different symptom-based case definitions on the observed epidemiology of acute gastrointestinal illness by applying previously published case definitions to a common, population-based data set and then using clustering (k-means and SOM) to create a data-driven view of the cases.