Reporting errors in one-week diarrhoea recall surveys: experience from a prospective study in rural Bangladesh.

To estimate inaccuracy in a diarrhoea recall survey mothers of pre-school children in Teknaf, Bangladesh were interviewed every week from July 1980 through June 1983. Because the likelihood of an episode starting on any given day of the week should be equal, we were able to quantify any deviation observed. Results show an average of 34% less diarrhoea episodes reported prior to a 48-hour recall period in any week. The amount of reporting error was (a) directly related to the length of the recall period, and (b) inversely related to the severity of diarrhoea as indicated by presence of fever and frequency of motions. This analysis reveals that weekly diarrhoea recall surveys in Bangladesh underestimate severe diarrhoea cases by 20-22% and less severe cases by 42-44%. The findings also indicate that morbidity surveys based on lengthy recall are likely to mislead health planners with regard to the magnitude of the problem and the volume of resources required to combat it.