Reliability of environmental, biological and health characteristics of infants from a developing country

Background. Although most epidemiological data are gathered by interviews, few studies ascertain their reliability. This study quantified inter-observer reliability of environmental, biological and health characteristics of underprivileged children, in a southern Brazilian state. Methods. Five health care professionals who had received extensive standardized interview and observation training interviewed mothers and observed the home environments of a random sample of 102 children. A second interview was conducted by a sixth health professional who had received separate but identical training without the initial interviewer knowing a follow-up interview would be conducted. Data from the two independent interviews and observations were compared for agreement using the kappa statistic. Results. There was excellent agreement (kappa > 0.75) for most of the household characteristics, such as type of walls, ceiling, floor and windows in spite of the diversity of buildings. Presence of the parents, number of people in the household, age of the mother and health assistance were reliable as well. However, information about skin colour of the mother and the number of cracks in the house only reached fair agreement. Conclusions. These results showed that data gathered by observation and interviews generate accurate information about environmental, biological and health care characteristics. Those based on written information, such as birth date, were highly concordant.