Clusters of community-dwelling individuals empirically derived from stool diaries correspond with clinically meaningful outcomes

Background: Functional gastrointestinal disorders (FGIDs) are diagnosed according to expert consensus criteria based on recall of symptoms over periods of 3 months or longer. Whether the expert opinion concords with underlying disease process and whether individual recall is accurate are both in doubt. This study aimed to identify naturally occurring clusters of individuals with respect to symptom pattern, evaluate their significance, compare cluster profiles with expert opinion and evaluate their temporal stability. Methods: As part of a random population study of FGID-related symptoms, we first explored the use of prospective stool and symptom diaries combined with empirical grouping of individuals into clusters using nonhierarchical cluster analysis. Results: The analysis identified two clusters of individuals, one of which was characterized by elevated scores on all domains of symptoms (26% of the sample) and one that was low to average on all domains (74% of the sample). Cluster membership was found to be stable over a long interval. Clusters were found to differ on most domains of quality-of-life (d = 0.46-0.74), self-rated health (d = -0.42) and depression (d = -0.42) but not anxiety. Prevalence of clinically diagnosed irritable bowel syndrome (IBS) was higher in the more impacted cluster (33%) compared with the healthy cluster (13%; P < 0.0001). Conclusion: A naturalistic classification of individuals challenges consensus criteria in showing that some IBS individuals have a symptom experience not unlike health. The proposed approach has demonstrated temporal stability over time, unlike consensus criteria. A naturalistic disease classification system may have practical advantages over consensus criteria when supported by a decision-analytic system.