A proposed new method of estimating weight deficits in children

Summary of paper presented at the Intenational Child Health Meeting on 'Managing Childhood Malnutrition' held in Birmingham

At the recent Birmingham meeting reported on in Field Exchange 3, André Briend from INSERM made a presentation on a new method of estimating weight deficits in children. In his presentation he argued that it is unsatisfactory to classify children as malnourished from nutritional surveys if their weight for height expressed in Z scores is below -2. His reasoning was that some children who were initially above the median Z score may lose considerable amounts of weight, without crossing the threshold, while others may be constitutionally thin and may have lost little weight and yet fall below this limit. He suggested that it made more sense to calculate the 'average weight deficit' of children with a given Z score.
He presented nutritional survey data from an ACF survey in Kabul, which showed that most children have weight deficits in populations that are under nutritional stress. This finding is apparently quite common in most nutritional surveys. However, he stressed that this type of finding does not challenge the frequently described association between a low Z score and a high risk of death, or the strategy of targeting nutritional interventions at children with Z scores below -2. He did however argue that Z scores may not be that effective in detecting children with an actual weight deficit, and that this may partly explain the failure of most nutritional programmes to avoid leakage of food distributions targeted at children with weight for height Z scores below - 2
In his presentation he put forward a new methodology which would allow calculation of the average weight deficit of children with a given Z score. The method is based on two simple hypotheses:

i) all children if well fed would form a population with a normal distribution of weight for height Z score with a mean of 0 and a standard deviation of 1 and

ii) the rank of Z score of children (i.e. the position of their Z score in relation to other children - 5th, 26th etc), is the same as the rank of the Z score they would have if well fed.

In the calculation of weight deficit one has to calculate the Z score a child would have if well fed by estimating the Z score at the same rank in the normal distribution. First, one calculates the 'cumulative frequency' (rank/sample size) corresponding to the observed Z score found during the survey (Zs), then the Z score with the same cumulative frequency in the normal distribution is estimated (Zn). The postulated weight deficit is Zs-Zn.

More details of the methodology can be obtained from Dr. André Briend INDERM U 290, Hospital Saint Lazare, 107 rue du Fauborg Saint Denis, 75475 Paris Cedex 10, France. E-mail: briend@ext.jussieu.fr