Contents - Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
Participation rates of 65%-85% are very high. Other large-scale supplementation trials have reported rates of 50%  to 60% ; in food-distribution programmes, participation rates have often been lower than 10% .
Although absolute non-participation rates were generally similar between villages, clear differences existed between the supplement types in terms of percentage of days attending. That the nature of the food can affect compliance rates was observed in a programme in Pakistan as well; only 18% of children receiving a whey-soy drink mix actually consumed the supplement because it was considered unacceptable . In the present study, atole, a thick, sweet gruel, may have been perceived by mothers as more appropriate for infants and young children than fresco, which resembled a Kool-Aid-type drink. Fresco may have been perceived as a "cool" drink (as the Spanish name implies) and considered inappropriate for consumption by younger children. These perceptions are particularly important within the hot-cold classification system that guides mothers' feeding and health habits throughout Latin America .
The two supplements were perceived and consumed differently. At the older ages, even though fresco children attended about the same percentage of days as the atole children, they were consuming far greater amounts of supplement. Besides the different characteristics of the supplements described, these results may be due to climatic variations as well; Espíritu Santo is hotter and dryer than the other villages.
The temporal changes in terms of rates of attendance were somewhat expected. It takes some time for a target audience to become accustomed to a programme such as this. The modest yet steady increase in attendance might suggest increased acceptance of the supplements.
A number of factors predicted participation in and attendance at the centre. Proximity resulting in higher attendance was expected, a finding also observed in Tanzania . Fortunately, distance affected participation for both the atole and the fresco groups, although the magnitude of coefficients was somewhat larger for atole.
Large family size was also associated with increasing attendance for both groups, a finding identified in an Indian study as well . Explanations for this may be that, because increasing family size is often associated with a decline in food consumption per family member [12, 13], larger families were relying on the supplementation to a greater extent to complement home diets. Alternatively, children from larger families were more likely to have older siblings accompany them to the centre.
The strong relationship between lower SES and higher attendance for the atole children, a phenomenon absent for the fresco children, is perhaps the most disturbing result of this analysis. Because lower SES has been associated with decreased in (non-participants included), by supplement type and socio-economic status takes of energy and poorer nutrition status [13, 14] as well as more monotonous diets , it is likely that the more frequently attending atole children were at increased nutritional and health risk.
Because this result is so important, additional analyses were done to confirm the relationship. Figure 7 (see FIG. 7. Mean attendance of children 1-3 years old) presents an example of these analyses, comparing attendance at the centre for children 1-3 years old according to socio-economic tertile. Clearly, for atole, lower SES children were attending at higher rates than high SES children, a pattern not seen for fresco.
It is unclear why lower SES atole mothers and children would choose to attend more than lower SES fresco participants. Perhaps atole was perceived as more nutritionally beneficial. If mothers did not perceive fresco as nutritionally beneficial, they may have been less willing to spend the time attending the centre.
Having identified the asymmetric influence of SES on attendance, the next question is how this could affect results. For some analyses, such as those that relate atole intake to outcomes such as growth and mental development, that the worse-off children received atole should bias the results against the hypothesis of finding a positive impact with atole. To remedy this, SES should always be controlled for when carrying out these types of analyses.
Sex bias in terms of family food allocation and nutrition status is often observed in Asia ; no bias against females was identified in this study. It could be that older females attended more often to receive atole because they were not receiving adequate food at home, but additional analyses would be required to confirm this.
Nutrition status as a predictor for attendance and consumption was not addressed in these analyses. As anthropometric measures were one of the primary outcomes of the study and changed over time, properly modelling their predictive influence is outside the scope of this article. In a similar analysis of the factors predicting attendance for atole children during the first year of the study, anthropometric indicators were not independently associated with participation .
Maternal education and literacy are associated with a variety of child health variables . Unfortunately, variables that may be used as indicators for maternal education in this study were missing for a disproportionate number of the non-participants and mothers who lived farthest from the centre. Entering this variable in the multivariate models reduced the sample size significantly and caused distance to become non-significant. Other analyses using this data base have found maternal literacy to have no impact on participation  or to affect only certain subgroups (e.g., those from larger families) .
Methodologically, missing values are of concern in the present analysis. In attempting to identify the determinants of participation, the fact that non-participants are most likely to have missing data may bias results. In multivariate models of the type presented in tables 3 and 4, non-participants are often missing data on one or more variables and thus are eliminated from the analyses. The influence of missing values on participation and the consumption of supplement is discussed in depth elsewhere . As may be expected, whether bias is introduced by missing values depends on the variables included in the model.
In conclusion, participation and attendance in this study were very high and actually increased over the life of the study. Patterns of attendance, however, did vary by supplement type during the first three years of life. Lower socio-economic status was independently associated with increased rates of attendance for atole but not fresco children. Analyses using these data should consider the possible biases introduced by differential self-selection between supplement types, particularly those potentially confounded by attendance or socioeconomic status. Future analyses should always control for socioeconomic status.
Data analyses were supported by NIH grant HD22440, a collaborative effort involving investigators at several institutions: R. Martorell (principal investigator, originally at Stanford University, now at Cornell University), J. Rivera (INCAP, Guatemala), E. Pollitt (University of California at Davis), and J. Haas (Cornell University). The authors would like to thank Jean-Pierre Habicht (Cornell University), Edward Frongillo (Cornell University), and Marie Ruel (INCAP, Guatemala), whose comments were particularly valuable in the preparation of this article.
Contents - Previous - Next