Multidimensional Nature of Undernutrition: A Statistical Approach

The statistical assessment of undernutrition is usually restricted to a pairwise analysis of anthropometric indicators. The main objective of this study was to model the associations between underweight, stunting and wasting and to check whether multidimensionality of undernutrition can be justified from a purely statistical point of view. 3742 children aged 0 to 59 months were enrolled in a cross-sectional household survey (2004 Cameroon Demographic and Health Surveys (DHS)). The saturated loglinear model and the multiple correspondence analysis (MCA) showed no interaction and a highly significant association between underweight and stunting (P=0), underweight and wasting (P=0); but not between stunting and wasting (P=0.430). Cronbach's alpha coefficient between weight-for-age, height-for-age and weight-for-height was 0.62 (95% CI 0.59, 0.64). Thus, the study of these associations is not straightforward as it would appear in a first instance. The lack of three-factor interaction and the value of the Cronbach's alpha coefficient indicate that undernutrition is indeed (statistically) multidimensional. The three indicators are not statistically redundant; thus for the case of Cameroon the choice of a particular anthropometric indicator should depend on the goal of the policy maker, as it comes out of this study that no single indicator is to be used for all situations.