Choosing an appropriate measure of diarrhoea occurrence: examples from a community-based study in rural Kenya

Epidemiological studies which aim to identify protective or risk factors for diarrhoea may rely on any of several established measures of disease occurrence, including cumulative incidence, (CI), incidence density (ID), and point prevalence (P); each with its own strengths and limitations. Comparison of these measures was afforded by a community-based study in rural Kenya in which the incidence and prevalence of diarrhoea were measured simultaneously but by independent means in the same group of children. In a cohort study, CI and ID among 138 infants (aged 0-6 months) were 26.1% and 1.41 episodes per 100 infant-weeks, respectively. Among 111 toddlers (aged 18-29 months) CI and ID were 39.6% and 1.96 episodes per 100 toddler-weeks, respectively. In a cross-sectional study of these children, the Ps among infants and toddlers were 9.8 and 6.1%, respectively. Data on incident cases required more time and resources to obtain and evidenced more underreporting of diarrhoea episodes relative to the prevalence data. Other community-based studies have reported the ID of diarrhoea five times more often than any other measure, but appear to have calculated ID incorrectly. The infrequency with which CI and P are reported does not reflect their actual utility.