Screening for marasmus: a discriminant analysis as a guide to choose the anthropometric variables

In a cross-sectional study, children 0-6 yr of age from eight different population groups in Africa and Asia were examined. Clinical assessment defined 8750 children as being well nourished and 194 as having marasmus. Height, weight, arm circumference (AC), and triceps skinfold thickness were measured; the latter two measurements and the clinical assessment were done by the same observer. Based on data from normal children, local growth curves were computed for each group. Each child's growth was expressed in standard deviation scores (SDS) of his own group. On the basis of the results of a discriminant analysis, all variables were ranked by their decreasing power to discriminate between normal and marasmic children. For 83% of the children one measurement (AC/age) is sufficient to classify them definitely; for the others several variables are needed. This strategy yields an overall sensitivity of 80%, a specificity of 97%, and a positive predictive value of 38%.