Proportion of under-five children whose height-for-age is below minus two (2) standard deviations (moderate and severe) and minus three (3) standard deviations (severe) from the median height-for-age of the reference population
Proportion of under-five children whose weight-for-age is below minus two (2) standard deviations (moderate and severe) and minus three (3) standard deviations (severe) from the median weight-for-age of the reference population
Proportion of under-five children whose weight-for-height is below minus two (2) standard deviations (moderate and severe) and minus three (3) standard deviations (severe) from the median weight-for-height of the reference population
Proportion of under-five children whose weight-for-heights is above two (2) standard deviations (overweight and obese) from the median weight-for-height of the reference population
Prevalence of underweight, stunting and wasting among under-five children is estimated by comparing actual measurements to an international standard reference population. In April 2006, the World Health Organization (WHO) released the WHO Child Growth Standards to replace the widely used National Center for Health Statistics (NCHS)/WHO reference population, which was based on a limited sample of children from the United States of America. The new standards are the result of an intensive study project involving more than 8,000 children from Brazil, Ghana, India, Norway, Oman and the United States. Overcoming the technical and biological drawbacks of the old reference population, the new standards confirm that children born anywhere in the world and given the optimum start in life have the potential to reach the same range of height and weight. It follows that differences in children's growth to age five are more influenced by nutrition, feeding practices, environment and health care than by genetics or ethnicity.
The new standards should be used in future assessments of child nutritional status. It should be noted that because of the differences between the old reference population and the new standards, prevalence estimates of child anthropometry indicators based on these two references are not readily comparable. It is essential to have all estimates based on the same reference population (preferably the new standards) when conducting trend analysis.
Adjusting country-level estimates
Before conducting trend analysis of child nutritional status, it is very important to ensure that estimates from various data sources are comparable over time. For example, household surveys in some countries in early 1990s only collected child anthropometry information among children up to 47 months of age or even only up to 35 months of age. Prevalence estimates based on such data only referred to children under four or under three and are not comparable to prevalence estimates based on data collected from children up to 59 months of age. Some age-adjustment needs to be applied to make these estimates based on non-standard age groups comparable to those based on the standard age range. For more information about age adjustment, please click here to read a technical note. In addition, prevalence estimates need to be calculated according to the same reference population. Those calculated according to the World Health Organization (WHO) Child Growth Standards are not comparable to those calculated according to the National Center for Health Statistics (NCHS) reference population. For more information about the difference between the two references and its implications, please click here to read a technical note.
Estimating country-level progress
Country-level progress in reducing undernutrition prevalence is evaluated by calculating the Average Annual Rate of Reduction (AARR) and comparing this to the AARR needed in order to be able to reduce prevalence by half over 25 years. For more information about how to calculate country-level AARR, please click here to read a technical note.
Estimating regional trends by multilevel modelling
Estimation of regional and global trend was based on a multilevel modelling method [de Onis et al. in JAMA 2004]. For the most recent trend analysis presented below, a total of 606 data points from 141 countries over the period from 1985 to 2010 were included in the model. This set of trend data points was jointly reviewed by UNICEF and WHO in June and July 2011 to ensure that it is nationally representative of the under-five children, processed using standard algorithm and comparable vertically and horizontally. Regional trend modelling and graphing were carried out using STATA 11 on two do files (the data set and do files are available on request).
Model-based trend analysis of underweight prevalence, by UNICEF regions
How to read this chart:
Each circle (bubble) represents a prevalence estimate from a country in a data year. The size of the circle is proportional to the under-five population in that country in the data year. For each region, a solid line indicates the regional trend as modeled by the multilevel logit regression on all the available data points in the region. The dashed lines in each region indicate the lower and upper bounds of the regional trend line corresponding to the confidence interval of the estimated trend.
Click here for a high-resolution version of the above chart.
UNICEF, 'Technical Note: Age-adjustment of child anthropometry estimates', UNICEF, New York, 2010.
de Onis, Mercedes et al., 'Estimates of Global Prevalence of Childhood Underweight in 1990 and 2015', JAMA 2004, 291: 2600 2606.
de Onis, Mercedes et al., 'Comparison of the World Health Organization (WHO) Child Growth Standards and the National Center for Health Statistics/WHO international growth reference: implications for child health programmes', Public Health Nutrition,2006: 9(7), 942947.
UNICEF, 'Technical Note: How to calculate Average Annual Rate of Reduction (AARR) of Underweight Prevalence', UNICEF, New York, 2007.