This is the old United Nations University website. Visit the new site at http://unu.edu
Accuracy of mothers' responses to questions about breast-feeding practices
Enrique Ríos, Linda Neuhauser, Sheldon Margen, and Vijaya Melnick
Abstract
Women giving birth to infants at 59 health facilities in three regions of Mexico were surveyed. Five hundred forty-six mothers were questioned on admission to the hospital about their plans for infant feeding when they returned home; 485 were questioned again prior to discharge; and 160 were interviewed a third time at home two weeks after discharge about their actual feeding practices. The results of a formal structured questionnaire asking about plans to use any of six specific milk and non-milk substances that are fed by bottle were compared with those of a questionnaire methodology in which the interviewers probed the answers. The study identified a third problem in dealing with self-reported information, in addition to the previously identified problems of memory failure and recall bias - which did not arise here - that of differences between respondents' perception of their behaviour and their actual behaviour. By supplementing a formal questionnaire with more probing and less structured questions, these biases can be lessened and more accurate data obtained.
Editor's note
The strengths of this paper lie in the way the information was obtained by probing, unstructured questions to supplement those of the formal questionnaire and in the relatively large sample size. Investigators depending only on formal questionnaires should consider what the opportunity for probing questions can add to the validity of their results.
The recognition that interview questionnaires may be biased if they do not allow for probing underlies the rapid assessment procedures (RAP) methodology that has been promoted by the United Nations University and UNICEF.* The Food and Nutrition Bulletin is interested in receiving additional articles illustrating the value of qualitative approaches to data collection at the household level either alone or as a complement to quantitative approaches.
Introduction
In recent years, a number of studies have documented the decline of breast-feeding in less developed countries [1] Concern over this decline is justified because of the advantages provided to the infant by breast milk: a lower risk of exposure to environmental contaminants. immunological protection against infection, reduced incidence of allergic diseases, and psychological bonding between the infant and its mother An inverse relationship between breast-feeding and infant morbidity and mortality from diarrhoeal diseases is evident in less-developed countries [2-5]. In view of these associations. the type of feeding practiced by mothers in developing countries is critical to the welfare of their children.
Some studies [6, 7], however, have questioned the validity of reports showing a decline in breast feeding in less developed countries. These studies argue that the reports that show a decline are methodologically flawed because they use non-representative or noncomparable samples or make implicit assumptions about past breast-feeding practices. Less attention has been focused on the accuracy of the information given by respondents.
Most of the studies documenting infant feeding practices follow a cross-sectional design in which an interviewer administers a structured questionnaire to obtain self-reported information. Consequently it is important to design questionnaires that obtain responses about infant feeding practices with the required level of accuracy. There is some published information [8, 9] on how answers to questions about why mothers do not choose to breast-feed can be misleading when the questionnaires do not allow for the inclusion of specific answers.
In this paper, we compare mothers' answers about their plans for and practice of infant feeding using two different questioning strategies: self-reported answers are compared with those obtained by more detailed questioning about types of feeding practices. We use data from a descriptive study on infant feeding practices in Mexico.
Materials and methods
A stratified random sample of 59 health facilities was selected from three regions of Mexico: Mexico City, Sonora, and Chiapas. The facilities were stratified by administrative classification, size specialties, and geographic location. Teams of nurses conducted continuous (24-hour) surveys of all mothers admitted to the facilities for delivery (except in a few very-large hospitals, where random samples were selected) during a four-day period. Further details of this prospective study are provided elsewhere [10].
Five hundred forty-six mothers were interviewed on admission to the hospital about their plans for feeding their infants when they returned home. Immediately prior to their discharge, 485 of the mothers were re-interviewed about their feeding plans. To check the accuracy of the mothers' answers, the interviewers probed their answers by asking them about plans to use any of six specific milk and nonmilk substances that are fed by bottle.
A random sample of 160 of the mothers interviewed at discharge were selected for a follow-up interview at home two weeks after discharge. At that interview they were asked about their current infant feeding practices. Mothers who initially said they were breast-feeding only were asked: "Even though you are feeding your baby primarily by breast, do you on some occasions give him/her a bottle?" Then, for the mothers who answered yes, the interviewer used a list of 16 milk and non-milk liquids to find out exactly what the mother was feeding her infant and how often. This list was also used to probe the practices of the mothers who reported mixed feeding or exclusive bottle-feeding.
Nine main categories of infant feeding practices were identified, comprising one category of exclusive breast-feeding, four categories of exclusive bottle feeding, and four categories of mixed feeding. Exclusive breast-feeding was defined as breast-feeding only, with no other liquid or solid given to the infant.
Results
When the mothers were asked? on admission to and at discharge from the hospital, how they planned to feed their infants at home, they tended to classify themselves into one of three broad categories: breast-feeding, bottle-feeding with formula, and mixed feeding. Table 1 presents a comparison of the answers on admission, before and after probing, showing how they changed. The most important change occurred in the category of exclusive breast feeding: after probing, this category was reduced by 18% due to the identification of mothers who actually were planning to use mixed feeding. The same information is presented in table 2 for the answers given at discharge. Again the most remarkable change was the reduction in the breast-feeding category, 23%.
Both on admission and at discharge, mothers who were planning to feed their babies with non-milk liquids (such as tea, juice, and cereal water) in addition to breast milk considered themselves as planning to breast-feed exclusively. If the mothers who did not initially report planning to feed their babies with other liquids besides breast milk had been classified as planning to breast-feed exclusively the misclassification (overestimation ) errors would have been 18% of all mothers (or 25% of those who said they planned to breast-feed exclusively) on admission, and 23% (35%) at discharge.
TABLE 1. Comparison of mothers' initial reports of plans for infant feeding and responses obtained after probing on admission to hospital (percentages; N= 546)
Initial report |
After probing |
|
Breast-feeding only | 72 |
54 |
Bottle-feeding | ||
with formula | 5 |
4 |
with formula and non-milk | <1 |
|
with milk | <1 |
|
Breast and bottle | ||
with formula | 23 |
22 |
with non-milk | 19 |
|
with formula, milk, and non-milk | < 1 |
|
Total | 100 |
100 |
"Non-milk" = liquids such as tea, juice, or water.
Another example of misclassification of the mothers by type of feeding comes from the follow-up questionnaire on current infant feeding practices at two weeks after discharge. The addition of two probing questions dramatically reduced reporting errors for mothers who classified themselves as belonging to one of the three main feeding categories. The mothers' actual feeding practices were markedly different from the patterns reported initially, as shown in table 3. Although 4() mothers initially reported exclusive breast-feeding, 36 of them in fact were also bottle-feeding (12 with formula. 24 with a non-milk product); only 4 mothers were actually breast-feeding exclusively. Other types of misclassification involved an initial failure to report non-milk liquids, such as tea.
The actual infant feeding practices at two weeks were: 3% breast-feeding only, 19% bottle-feeding only, 51% combining breast-feeding with bottle feeding using formula. 23% combining breast feeding with bottle-feeding using a non-milk product, and 4% combining breast-feeding with bottle-feeding using a non-formula milk. Only 4% of the mothers used non-formula milks at two weeks these bottles were prepared with powdered or evaporated milk. Ninety-five per cent of the mothers gave their infants non-milk products - most often tea (95% ), sugar syrup (61%), water (52%), rice or cereal water (13%) and honey (7%) A number of medicines were also given by bottle.
Although the most dramatic frequency change occurred in the exclusive breast-feeding category. 17% of the mothers initially failed to report using the bottle with formula plus nonmilk product (juice, tea, or water). Combining breast-feeding with a non milk product was under-reported by 14% of the mothers. This last category is critical since most of these women were misclassified in the exclusive breast-feeding category.
Discussion
Two potential problems have been recognized in relying on self-reported information: memory failure and recall bias [11] Neither of these problems arose in the data presented here, since the questions had to do with planned and current practices and memory or recall was not involved. However, the study identified a third problem: that of differences between respondents' perception of their behaviour and their actual behaviour - in this case specifically the mothers' perceptions of the type of feeding they were giving their babies and their actual feeding practices.
The results demonstrate that the mothers tended to overestimate their breast-feeding practice. If the mothers" answers at admission or discharge had been accepted at face value, the conclusion would have been that 72% and 66% respectively of all the mothers were intending to breast feed their babies exclusively, when in fact only 54% on admission and 43% at discharge were planning to do so. In other words, only 75% of the mothers who said on admission that they planned to breast-feed exclusively. and only 665% of those who said so at discharge, were actually planning to do so.
TABLE 2. Comparison of mothers initial reports of plans for infant feeding and responses obtained after probing at discharge from hospital (percentages: N=485)
Initial report |
After probing |
|
Breast-feeding only | 66 |
43 |
Bottle-feeding | ||
with formula | 6 |
5 |
with formula and non-milk | 1 |
|
with milk | 1 |
|
Breast and bottle | ||
with formula | 28 |
27 |
with non-milk | 23 |
|
with formula. milk, and non-milk | 1 |
|
Total | 100 |
100 |
TABLE 3. Initially reported versus actual infant feeding practices two weeks after discharge from hospital
Reported |
Actual |
|||
No. |
% |
No |
% |
|
Breast-feeding only | 40 |
25 |
4 |
3 |
Bottle-feeding | ||||
with formula | 25 |
16 |
2 |
1 |
with formula and | ||||
non-milk | 2 |
1 |
28 |
18 |
Breast and bottle | ||||
with formula | 62 |
39 |
2 |
1 |
with formula and non-milk | 12 |
7 |
81 |
50 |
with non-milk | 15 |
9 |
37 |
23 |
with non-formula milka | 4 |
3 |
6 |
4 |
Total | 160 |
100 |
160 |
100 |
a .Evaporated or powdered milk.
At the two-week follow-up interview, if the mothers' initial reports of their feeding practices had been accepted at face value, exclusive breast-feeding would have accounted for 25% of the sample, when in fact only 3% were breast-feeding exclusively. Ninety per cent of the mothers who reported breast-feeding exclusively at two weeks were actually combining breast- and bottle-feeding In fact, 30% of these mothers were bottle-feeding with formula in addition to breast-feeding. Other types of misclassification errors involved mothers who neglected to mention bottle-feeding with nonmilk liquids.
These data show that much of the information gathered from interview questionnaires may be biased if there is no allowance for probing. This problem of misclassification of mothers' own perceptions about their type of feeding behaviour is critical not only when dealing with infant feeding trends at the national or regional level but also in studies that analyse the relationship between type of feeding and health outcomes.
To overcome the problem identified in this study, we recommend: (1) providing respondents with precise operational definitions of the type of feeding given to the infant, or (2) the use of careful probing questions when asking mothers about their infant feeding practices.
Acknowledgements
This study was funded by the Nestle Infant Formula Audit Commission. Dr. L. H. Allen and Dr. H. Martinez provided helpful comments on an earlier drain of this paper.
References
1. Popkin BM, Akin SJ, Flieger W, Wong EL. Breast feeding trends in the Philippines, 1973 and 1983. Am J Pub Health 1989;79(1):32-33.
2. Puffer RR. Serrano CV. Diarrheal diseases and breast feeding. In: Patterns of mortality in childhood: report of the InterAmerican Investigation of Mortality in Childhood. Washington, DC: Pan American Health Organization, 1973.
3. Urrutia JJ. Sosa R, Kennell JH et al. Prevalence of maternal and neonatal infections in a developing country: possible low-cost preventive measures. Ciba Found Symp 1980;77:171-86.
4. Lepage P, Munyakazi C. Hennart P. Breast-feeding and hospital mortality in children in Rwanda. Lancet 1981;2:409-1 1.
5. Brown HK, Black ER. Romaña LG et al. Infant feeding practices and their relationship with diarrhea! and other diseases in Huascar (Lima), Peru. Pediatrics 1989;83(1):31-41.
6. Millman S. Trends in breast-feeding in developing countries. Int Fam Plan Perspectives 1986;12(3):91-95.
7. Notzon F. International review: trends in infant feeding in developing countries. Pediatrics 1984;74(4,part 2) :648-66.
8. Simopoulos PA, Grave DG. Factors associated with the choice and duration of infant-feeding practice. Pediatrics 1984;74(4.part2):603-14.
9. Forman MR. Review on the factors associated with the choice and duration of infant feeding in less-developed countries. Pediatrics 1984;74(4,part2):667-94.
10. Margen S, Melnick V, Neuhauser L. Ríos E. Infant feeding practices in Mexico. Washington, DC: Nestle Infant Formula Audit Commission. 1991.
11. Raphael K. Recall bias: a proposal for assessment and control. Int J Epidem 1987: 16(2): 167-70.