Bias in the Reporied Incidence of Smoking by Children

Abstract

McKennell A C [Social Science Faculty, The University, Southampton]. Bias in the reported incidence of smoking by children. International Journal of Epidemiology 1980, 9: 167–177. Factors which bias self-reports of smoking behaviour were studied using a balanced sample of over 4000 boys and girls aged 11 to 16 years drawn from 48 British secondary schools. It was found that, for boys, the reported incidence of those smoking one cigarette or more per week tends to increase when the questionnaire is self-administered rather than completed by an interviewer, when the answers are obtained in school rather than in home interviews and, in school, when children were interviewed together in classrooms rather than individually. These effects did not reach statistical significance level for girls and were most pronounced for the younger boys. Among the latter there was nearly a five-fold difference between the incidence reported in classroom and home-oral interviews. For younger boys and girls it was found too that the inclusion of a persistently probing question about trying even one cigarette increased the numbers admitting to regular smoking, while an emphasis on the confidentiality of the interview produced a decrease, not an increase, in admissions. The presence or absence of parents was not found to influence the reports of smoking obtained in the home interviews. It was concluded that prevalence estimates from studies of children's smoking are highly contingent, especially for younger boys, on the method by which the data are obtained. Comparability would be aided if reports of study design would include more detail on those aspects which have been shown to produce bias. The correlations reported in many studies between the level of smoking and other variables should be independent of biases affecting estimates of the level itself. But care needs to be exercised in interpreting correlations with variables, such as parental strictness, which could themselves be a cause of response bias. The relative strictness of parents versus teachers was found to be associated with the bias in response between the home and school interview conditions. At the same time a negative correlation was found for older children between the amount of smoking reported and parental punishment for smoking, and this correlation replicated in all interview conditions.