Intra-household correlations in maternal-child nutritional status in rural Guinea: implications for programme-screening strategies.

Introduction It is widely recognized that the collective state of ill-health among women and children in developing countries results from the joint influences of nutritional, biological, and social deprivation (1--3). Since rates of physical growth and achieved body size closely reflect the underlying depletive processes, anthropometric indicators provide useful markets for the degree of current and cumulative deprivation experienced by individuals, communities, and populations. In addition, in view of the strong associations between maternal and child nutritional status at a given time and subsequent health outcomes (incidence of infectious disease, adverse pregnancy outcomes, and mortality), anthropometry also provides a useful screening tool for more general health risk among individuals and populations (4--7). Recently, however, concerns have been raised about appropriate uses and interpretation of anthropometric data. For example, the United Nations Administrative Committee on Coordination--Subcommittee on Nutrition (ACC/SCN) has enumerated the various uses of anthropometric data and the limitations of anthropometry, in general, and of various indices in different applications (2). Clear distinctions are made in the use and interpretation of various indicators for different purposes (i.e., one-time versus trend assessments, individual-level versus population-level applications, and emergency versus non-emergency situations). The present article extends the discussion of appropriate uses of anthropometric indices by considering to what extent anthropometric indicators for a particular household member (e.g., a child) can be used to make inferences about the nutritional status of other household members (e.g., the child's mother or siblings). In other words, to what extent can households be identified as being at risk of maternal--childhood malnutrition on the basis of a nutritional assessment of a single household member? The answers to these issues have far-reaching implications for the use of anthropometry for screening nutrition programmes. To the extent that there are high within-household correlations in nutritional status among women and children, routine screening programmes or periodic assessments could provide accurate and cost-efficient means of identifying households for nutrition and/or more broadly-based maternal and child health programme interventions. In public health practice it is generally assumed that households are relatively homogeneous with respect to nutritional status; indeed, this assumption underlies the notion of a "high-risk" household. However, the magnitude of within-household homogeneity with respect to nutritional status and the implications of this for programme screening purposes have not been systematically investigated. This study provides evidence on these issues through the analysis of within-household associations in nutritional status among women and children in rural Guinea. Cross-sectional data from a moderately sized household survey undertaken in 1990 were used to examine associations between the nutritional status of mothers and their children and between siblings residing in the same household. In addition to the examination of within-household correlations, the efficiency of using anthropometric indices for a single household member as a screening tool for household maternal--child nutritional risk was formally evaluated by analysing the sensitivity, specificity, and positive--negative predictive values of various indicators for women and children of different ages. Materials and methods Survey data The data were obtained from a cross-sectional sample survey of women with surviving children under the age of 5 years in five provinces in central Guinea in 1990. The survey was designed to provide baseline data on nutrition and health problems in the region and related information useful for the design of health and nutrition intervention programmes. …