Characteristics of responders and non-responders in an infant feeding study

Few studies have fully investigated and described the characteristics of non-responders in infant breast feeding studies. Examination of the characteristics of non-responders enhances this understanding and provides information on the representativeness of the sample actually investigated.

As part of a larger local longitudinal study, couples whose babies were due to be born during the three months of the study in the autumn of 1995, at two local maternity hospitals, were approached by midwives to participate in the study of infant feeding patterns and determinants, including interviews, questionnaires and postal survey. From the 648 eligible mothers, 91 (14 per cent) were not approached, 233 (35.9 per cent) consented but were not interviewed, 256 (39.5 per cent) consented and were interviewed and 68 (10.5 per cent) refused to participate (non-responders) in the longitudinal study. To make a comparison between the 'responders' (i.e. the 233 who consented but were not interviewed and the 256 who consented and were interviewed) and non-responders (the 68 who refused to consent), appropriate data were collected on their socio-demographic characteristics, their feeding intentions and their feeding behaviour.

Univariate analysis revealed a number of significant differences between responders and non-responders. However, log-linear analysis of these differences indicated that the main factors that differentiated between responders and non-responders were social class, smoking habit and actual feeding behaviour. When comparing the feeding intention at the time of 'booking', the non-responders compared with the responders were less likely to intend to breastfeed and more likely to be uncertain about their feeding intentions. Following delivery non-responders compared with responders were more likely to bottle feed their babies. By the time of discharge, of the non-responders, 83.6 per cent were bottle feeding in comparison with 47.3 per cent of responders who were also bottle feeding.

In this study non-responders were found to be more similar to bottle-feeding responders than to breastfeeding responders. Furthermore, non-responders were more likely to be smokers, from lower social class and to bottle feed. This study also showed that the non-responders were more undecided about their feeding intention at the time of the booking visit. Invariably a higher percentage of mothers who were undecided at the booking visit chose to bottle feed their babies at discharge. These results highlight the impact of failing to include information from non-responders when conducting research which examines patterns of infant breastfeeding and attitudes towards it.