Standard deviation of anthropometric Z-scores as a data quality assessment tool using the 2006 WHO growth standards: a cross country analysis
Zuguo Meia, Laurence M Grummer-Strawna
Nutritional status, especially in children, has been widely and successfully assessed by anthropometric measures in both developing and developed countries.1 Height and weight are the most commonly used measures, not only because they are rapid and inexpensive to obtain, but also because they are easy to use. Once a child’s height and weight have been correctly measured and their age has been recorded, a clinician or researcher can assess the child’s growth and general nutritional status by using a standardized age- and sex-specific growth reference to calculate height-for-age Z-scores (HAZ), weight-for-age Z-scores (WAZ), weight-for-height Z-scores (WHZ) and body-mass-index-for-age Z-scores (BMIZ).
An international reference is useful since the growth in height and weight of well fed, healthy children under 5 years of age from different ethnic backgrounds and different continents is reasonably similar.2–4 In April 2006, the World Health Organization released new global child growth standards for infants and children up to the age of 5 years. These new standards were developed in accordance with the idea that children, born in any region of the world and given an optimum start in life, all have the potential to grow and develop to within the same range of height and weight for age. The new WHO child growth standards, which will be used worldwide, provide a common basis for the analysis of growth data.5,6
The Z-score system expresses anthropometric values as several standard deviations (SDs) below or above the reference mean or median value. Because the Z-score scale is linear, summary statistics such as means, SDs and standard errors can be computed from Z-score values. Z-score summary statistics are also helpful for grouping growth data by age and sex. The summary statistics can be compared with the reference, which has an expected mean Z-score of 0 and a SD of 1.0 for all normalized growth indices.1 On the basis of the 1978 WHO/National Center for Health Statistics (NCHS) growth reference, WHO has previously indicated that the SD of Z-scores of these indicators is reasonably constant across populations, irrespective of nutritional status, and thus can be used to assess the quality of anthropometric data.1 However, there has been no evaluation of whether these Z-score ranges still apply to data attained from application of the 2006 WHO growth standards.
For this analysis, we used all available Demographic and Health Survey (DHS) data that have height and weight measurements for children from birth to 59 months of age to assess whether the SD of height- and weight-based Z-score indicators derived from the 2006 WHO growth standards can still be used as data quality indicators.
The DHS are nationally representative surveys with large sample sizes. Supported by the United States Agency for International Development (USAID), the surveys provide data on a wide range of monitoring and impact evaluation indicators in the areas of population, health and nutrition of women and children in developing countries.
The surveys are based on household interviews of women of childbearing age, and they record standardized anthropometric measurements and some biochemical measurements. Standard core questionnaires are used for all DHS surveys to obtain information about respondent’s backgrounds, the children they have given birth to, their knowledge and use of family planning methods, the health of their children, their reproductive health and other issues helpful to policy-makers and administrators in health and family planning. Results are comparable between countries because the surveys use standard questionnaire modules and standardized sampling procedures, fieldwork methods (i.e. similar schedules of field work, standardized training for each survey worker, the same number of survey workers and type of equipment for each survey team) and data processing techniques. Four rounds of surveys have been completed to date: DHS I (1984–1989), DHS II (1988–1993), DHS III (1992–1997) and DHS+ (1997–2003). Over the past 20 years, the DHS project has coordinated surveys in more than 70 countries throughout sub-Saharan Africa, north Africa, Asia, Latin America and the Caribbean, and parts of Europe. Details of the DHS survey design and methodology have been published elsewhere.7–12
According to the protocols used by the DHS, for children younger than 60 months of age, height or length is measured to the nearest 0.1 cm. A measuring board is used to measure the child’s recumbent length if the child’s age is younger than 24 months; standing height is measured for children 24 months or older. Weight is measured to the nearest 0.1 kg on a paediatric scale or other beam balance scale. All survey team members receive standardized training on how to weigh and measure children. The protocol requires that two measurements of a child’s length or height agree within 0.1 cm, and two readings of weight agree within 0.1 kg.10,11 These measurement instructions are based, with some alteration, on the United Nations manual on how to weigh and measure children.13 All records are entered onto a standardized paper form or into an automated computer system in the field. After the records are computerized at the country level, they are transferred to Macro International Incorporated, an Opinion Research Corporation company (ORC Macro), for inclusion in the DHS database. For our analysis, all Z-scores – HAZ, WAZ, WHZ and BMIZ – were recalculated using the 2006 WHO growth standards with a SAS Macro program.14
Altogether, 94 surveys from 46 countries were available from the DHS website (http://www.measuredhs.com). Of these 94, we excluded 11 surveys that had missing anthropometric data and 32 surveys that did not collect data for children aged 0–59 months. The final sample for this analysis was from 51 DHS surveys representing 34 developing countries (Table 1).
In our final sample for each DHS survey, we excluded children without data on height, weight, age or sex. We used the new cut-offs recommended by the WHO for data exclusion.5 Thus, data were excluded if a child’s HAZ was below –6 or above +6, WAZ was below –6 or above +5, WHZ was below –5 or above +5, or BMIZ was below –5 or above +5, because these extreme values were most likely a result of errors in measurement or data entry.5
In our analysis, we calculated the Z-score means and SDs of the four nutritional indicators (HAZ, WAZ, WHZ and BMIZ) for each survey and summarized them by region (Africa, Asia and Latin America). Then we repeated the analysis by age group (0–35 months and 36–59 months). All data were analysed with SAS software version 8.2,15 and the original sample weights for each survey were applied to all the results.
Table 1. Mean and standard deviation of anthropometric Z-scores derived from Demographic and Health Survey data
Of the 51 DHS surveys, 32 were from 23 African countries, four were from three Asian countries and 15 were from eight Latin American countries (Table 1). Z-score means and SDs of height-for- age, weight-for-age, weight-for-height, and body-mass-index-for-age are summarized in Table 1 and also shown in Fig. 1, Fig. 2, Fig. 3 and Fig. 4, respectively. For all four indicators, Z-score means in Latin American countries were generally higher than in African and Asian countries (Fig. 1, Fig. 2, Fig. 3, Fig. 4 and Table 2). However, the SDs for all three indicators were relatively stable and did not vary with the Z-score means.
Table 2 shows Z-score means and SDs stratified by age group and region. In general, the Z-score means in older children (36–59 months) were lower than those in younger children (0–35 months), but the SDs in older children were smaller than those in younger children. These findings were true for all four indicators and in all three regions.
For children aged 0–59 months, the 5th and 95th percentiles of the SDs were 1.35 and 1.95 for HAZ, 1.17 and 1.46 for WAZ, 1.08 and 1.50 for WHZ, and 1.08 and 1.55 for BMIZ (Table 3). The ranges of SDs in children in the younger age group (0–35 months) were much wider than those of children in the older age group (36–59 months). For children aged 0–35 months, the 5th and 95th percentiles of the SDs were 1.46 and 2.03 for HAZ, 1.23 and 1.56 for WAZ, 1.13 and 1.62 for WHZ, and 1.15 and 1.66 for BMIZ. For children aged 36–59 months, they were 1.16 and 1.67 for HAZ, 1.02 and 1.29 for WAZ, 0.94 and 1.33 for WHZ, and 0.95 and 1.36 for BMIZ.
Table 2. Mean and standard deviation of anthropometric Z-scores derived from Demographic and Health Survey data, stratified by age group
Table 3. Standard deviation ranges for data quality assessment and the 5th and 95th percentiles of standard deviations from the Demographic and Health Surveys (DHS) for four anthropometric Z-scores
Height- and weight-based anthropometric measurement is an excellent tool to gauge general nutritional status in a population. However, informative assessments of nutritional status rely on the data being of high quality. There are several ways to evaluate the quality of data for anthropometric indicators, including percentage of missing data, digit preference in recorded measurements, the percentage of measurements that are biologically implausible and the SDs of the Z-scores.1
On the basis of the 1978 WHO/NCHS growth reference, WHO recommended an acceptance range on the SDs for each indicator for data quality assessment.1 Those acceptance ranges were valid only if the same data exclusion criteria recommended by WHO also applied (i.e. data were excluded if a child’s HAZ was below –5 or above +3, WAZ was below –5 or above +5, or WHZ was below –4 or above +5).1 When WHO released the new growth standards in April 2006, they recommended new cut-offs for data exclusion based on the 2006 WHO standard (i.e. data were excluded if a child’s HAZ was below –6 or above +6, WAZ was below –6 or above +5, WHZ was below –5 or above +5, or BMIZ was below –5 or above +5).5 The new cut-offs for data exclusion for the 2006 WHO standards were wider than those for the old cut-offs for the 1978 WHO/NCHS reference, particularly in height-for-age measurements.
By applying the 2006 WHO growth standards and the new cut-offs for data exclusion, our analysis confirms the WHO assertion that the SD remains in a relatively small range for each indicator, no matter the Z-score mean. This finding suggests the SD of Z-scores could still be used as a data quality indicator for evaluation of anthropometric data. However, the observed ranges of SD for all four indicators from our analysis were consistently wider (especially in the high value of the range for each indicator and in height-for-age measurements) than those recommended by WHO.1
One of the strengths of our analysis is that we have used data from 51 national representative DHS from 34 developing countries with a large number of standardized measurements. However, a potential limitation is that we included only four surveys from three Asian countries and thus may not have fully captured anthropometric status in this region.
To assess whether the wider SDs that we observed in the DHS data were due to the switching of the growth reference from 1978 WHO/NCHS to 2006 WHO or because of the difference in the cut-offs for data exclusion recommended by WHO for each reference, we repeated our analysis, applying both references and cut-offs to the same DHS data we examined. As shown in Table 3, even with the old 1978 WHO/NCHS reference and the old cut-offs, the observed SDs were wider than recommended by WHO in 1995. Using the old cut-offs based on the 1978 WHO/NCHS reference to clean the DHS data first, then assigning new Z-scores using 2006 WHO standards without any data exclusion, we observed SDs for each indicator that were similar to those obtained with use of the old 1978 WHO/NCHS reference with old cut-offs (Table 3). This finding suggests that the wider range of SD values that we observed, using the 2006 WHO standards with the new cut-offs that WHO recommended for data exclusion, were mostly due to the new wider cut-offs for data exclusion proposed by WHO. ■
- Physical status: the use and interpretation of anthropometry. Geneva: WHO; 1995. WHO technical report series.
- PL Graitcer, EM Gentry. Measuring children: one reference for all. Lancet 1981; 2: 297-9.
- JP Habicht, R Martorell, C Yarbrough, RM Malina, RE Klein. Height and weight standards for preschool children. How relevant are ethnic differences in growth potential? Lancet 1974; 1: 611-4.
- Martorell R. Child growth retardation: a discussion of its causes and of its relationship to health. In: Blaxter KL, Waterlow JC, eds. Nutritional adaptation in man. London: John Libbey;1985:13-30.
- WHO child growth standards: length/height-for-age, weight-for-age, weight-for-length, weight-for-height and body mass index-for-age: methods and development. Geneva: WHO; 2006.
- M de Onis, C Garza, CG Victora, MK Bhan, KR Norum. The WHO Multicentre Growth Reference Study (MGRS): Rationale, planning, and implementation. Food Nutr Bull 2004; 25: S1-89.
- RJ Lapham, CF Westoff. Demographic and Health Surveys: population and health information for the late 1980s. Popul Index 1986; 52: 28-34.
- Model “A” questionnaire with commentary for high contraceptive prevalence countries: MEASURE DHS+ basic documentation number 1. Calverton: ORC Macro; 2002.
- Model “B” questionnaire with commentary for low contraceptive prevalence countries: MEASURE DHS+ basic documentation number 2. Calverton: ORC Macro; 2002.
- Interviewer’s manual for use with model “A” questionnaire for high contraceptive prevalence countries: MEASURE DHS+ basic documentation number 3. Calverton: ORC Macro; 2002.
- Interviewer’s manual for use with model “B” questionnaire for low contraceptive prevalence countries: MEASURE DHS+ basic documentation number 4. Calverton: ORC Macro; 2002.
- Supervisor’s and editor’s manual for use with model “A” and “B” questionnaires: MEASURE DHS+ basic documentation number 5. Calverton: ORC Macro; 2002.
- How to weigh and measure children: assessing the nutritional status of young children in household surveys. New York: UN Department of Technical Co-operation for Development and Statistical Office; 1986.
- WHO child growth standards SAS macro. Geneva: WHO; 2006. Available from: http://www.who.int/childgrowth/software/readme_sas.pdf
- SAS/STAT user’s guide (version 8, 1st ed). Cary: SAS Institute Inc; 1999.
- Maternal and Child Nutrition Branch, Division of Nutrition and Physical Activity, Centers for Disease Control and Prevention, Mailstop K-25, 4770 Buford Highway, Atlanta, GA 30341-3724, USA.