Background: Methodological issues in the design and interpretation of cross-sectional interview surveys of the prevalence of acute respiratory infections (ARI) were assessed among young children.
Methods: A cross-sectional survey was conducted in approximately 20,000 children in the north of Ghana. Approximately half were administered a questionnaire in which the initial questions about recent illnesses were direct questions about the presence or absence of three specific ARI-related symptoms (cough, rapid breathing, difficulty breathing), while the other half were administered a questionnaire which started with an open-ended question on whether the child was ill, designed to elicit spontaneous responses. A 2-week recall period was used in addition to point prevalence questions for half of the children in each group, while 4 weeks was used for the other half. The results were compared with those from a longitudinal morbidity surveillance system in an adjacent population of children. The repeatability of the responses to each of the symptoms/conditions was assessed in a subsample of the children.
Results and conclusions: The point and period prevalence rates of ARI symptoms or conditions based on spontaneously elicited responses were more likely to be valid than those based on prompted responses. Furthermore, using a 2-week recall period appeared to give more valid period prevalence rates than a 4-week recall period. The repeatability of the various ARI questions was not high (kappas 0.14 to 0.49), irrespective of the questionnaire design. Whether these findings will also be true in other populations needs to be assessed.