Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

Methodology

Much of the following material is for the benefit of those planning anthropometric surveys. Emphasis is on large-scale, cross-sectional surveys, and issues regarding the collection of serial data on individual children are not discussed.

Personnel Selection and Training

The personnel selected to be trained in anthropometry should read and write with ease and have good notions of arithmetic. If at all possible, one should aim for persons with some secondary school education. While some less-educated individuals may perform adequately, measurement errors will be less of a problem with, for example, high school graduates who can also be involved with quality control procedures. On the other hand, more highly educated personnel (e.g., physicians) may find the task of maintaining constant measurement procedures boring and may for this reason not be as reliable.

Women are preferred, for it may not be proper in some cultures for men to measure women, and children may be more relaxed if examined by women.

To train anthropometrists, one obviously needs a person with experience in measuring to demonstrate the proper techniques. In addition, a manual or handbook describing all procedures, ideally including simple illustrations, should be available during training.

The period of training should last as long as it takes to achieve acceptable levels of reliability and accuracy. "Reliability" can be understood as the degree to which the measurer is able to reproduce measurements, while "accuracy" refers to the degree to which the measurement approximates "true" values. Habicht (13) has devised simple procedures for training personnel. (These procedures are recommended by WHO in a manual for use in developing countries 191.) Typically, subjects are measured twice by each student and by the supervisor. From these data, the error variance of each student is estimated and compared to that of the supervisor to monitor reliability; and systematic differences among students and between each student and the supervisor are used to monitor accuracy. At specified levels of performance as suggested by Habicht, students are judged to be ready for data collection. The widespread availability of inexpensive electronic calculators should make these simple computations even easier.

Concern with measurement error should continue throughout the duration of the study and should involve three types of exercises. At periodic intervals, the supervisor and the anthropometrists should meet to discuss problems encountered in the field, to review measurement procedures, to calibrate instruments and to repeat the initial training exercises in order to monitor reliability and accuracy. Secondly, replicate measures by each anthropometrist should also be carried out under field conditions. The actual measurement process, particularly if the examination only includes five measures as recommended in this chapter, takes very little time; it is getting to the subject or getting the subject to come to the measurement centre which is generally time-consuming. Thus it is highly recommended that each subject be measured twice by the anthropometrists. As suggested in figure 3.2 (see FIG. 3.2. Anthropometry Form), the first set of measurements should go on the front of the form and the second on the reverse side in order to maximize the independence of each set of observations. If possible, other data (e.g. clinical signs) could be collected in between each set of values to further improve measurement independence. Double measurements provide ongoing reliability estimates under field conditions and, as explained later in the quality control section, result in better data. Finally, the supervisor will need to visit periodically the various anthromometrists at their working sites. On such occasions, the supervisor should also measure a small number of subjects (i.e.. five subjects). At the end of the survey, a sufficient number of cases (i.e., 50 subjects) would be available to determine accuracy under field conditions for each anthropometrists.

Table 3.3. (see TABLE 3.3 Measurement Error Standard Deviation in an INCAP Longitudinal Study (Preschool Children)) shows measurement error data from an INCAP longitudinal study for the five variables recommended in this chapter. Error estimates derived from standardization exercises are, as expected, less than those encountered under field conditions. Comparisons of the field measurement error variance to the population variance for each variable shows that the five measures chosen have high reliability. The error variance for other measurements such as chest circumference, for example, represents over 50 per cent of the population variance.

FIG. 3.2. Anthropometry Form

TABLE 3.3 Measurement Error Standard Deviation in an INCAP Longitudinal Study (Preschool Children)

The measurement error standard deviation was estimated by [S (a - b)²]1/2 / 2n where a and b are first and second observations, respectively, and n is the number of subjects measured.

Source: Martorell et. al. (14)

Measurement Techniques

The most widely-accepted techniques of measurement are those proposed by the International Biological Programme (10, pp. 8-12), as follows:

1. Stature (measured with a stadiometer or anthropometer; used for children two years old or older). The subject should stand on a platform with his heels together, stretching upward to the fullest extent, aided by gentle traction by the measurer on the mastoid processes. The subject's back should be as straight as possible, which may be achieved by rounding or relaxing the shoulders and manipulating the posture. The marked Frankfort plane must be horizontal. Either the horizontal arm of an anthropometer, or a counter-weighted board, is brought down on the subject's head. If an anthropometer is used, one measurer should hold the instrument vertical with the horizontal arm in contact with the subject's head, while another applies the gentle traction. The subject's heels must be watched to make sure they do not leave the ground.

2. Supine length (measured with an infant measuring table; used for children younger than two years of age). With the infant lying supine, one measurer holds the infant's head in the Frankfort plane and applies gentle traction to bring the top of his head into contact with the fixed headboard. A second measurer holds the infant's feet, toes pointing directly upward, and also applying gentle traction. brings the movable footboard to rest firmly against the infant's heels.

3. Upper arm circumference (measured with a tape). The subject's arm hangs relaxed, just away from his side, and the circumference is taken horizontally at the marked level. (The horizontal mark on the left arm is measured half way between the inferior border of the acromion process and the tip of the olecranon process.)

4. Skinfold thick-nesses (measured with skinfold caliper). The skinfold is picked up between thumb and forefinger and the caliper jaws applied at exactly the level marked. The measurement is read two seconds after the full pressure of the caliper jaws is applied to the skinfold; if a longer interval is allowed the jaws may "creep" and the reading will be inaccurate. Two measurements are used:

- Over triceps. The skinfold is picked up at the back of the arm about 1 cm above the level marked on the skin for the arm circumference and directly in line with the point of the elbow, or olecranon process.
- Subscapular. The skinfold is picked up under the angle of the left scapula. The fold should be vertical, or pointing slightly downwards and outwards.

5. Weight (measured with a scale). Weighing should preferably be done with the subject in the nude or clothed only in lightweight shorts (which may be provided by the investigator). In the latter circumstance the measurement can be corrected accordingly by adjusting the machine to read zero when a sample garment is placed on it. In all other circumstances, including when trousers are worn, the weight of a representative garment should be entered on the form, for subtraction later. The presence of visible oedema should be recorded.

Knowing the correct age is essential for interpreting anthropometric data. In its absence, for example, indicators of stunting cannot be generated. Age data are easy to obtain when records are available, or when mothers recall birth dates reliably. in such cases, the date of birth and the date of examination should be recorded; hasty calculations of age in the field should be avoided.

It would be a serious mistake to exclude systematically all children of unconfirmed ages, as they most likely will be in poorer nutritional status. However, the problem of estimating age in such cases is a difficult one. Some options are available if the child is yew, young, as Jelliffe notes:

Sometimes, the mother may not know the child's age, but may be able to recite the month of birth, and occasionally the day as well. If this is so, the mother will often recall details of the youngest child only, not those of older siblings. (16)

The examiner in the above situations can estimate the year of birth confidently if the child is an infant or toddler.

The use of dental eruption data as an aid in estimating age in the first two years of life has also been proposed. However, there is considerable variability in eruption times and normative data are not available for many groups.

Estimating the age of children older than two years is the difficult aspect of the problem. Jelliffe offers some advice:

Often the only practicable method may be to construct a locally relevant calendar... based on events in the preceding years, including agricultural, climatic and political occurrences, as well as natural or man-made disasters... However, such a calendar takes weeks to prepare and pre-test in the field, while its use in survey circumstances is laborious, time-consuming, and least satisfactory with the unsophisticated communities for which it is intended. Calendars will plainly have to be specific for different communities. (16)

All other data should be collected even when age proves impossible to estimate. Assessments of wasting, as noted earlier, do not require that age be known.

Equipment

The Ninth Handbook of the International Biological. Programme (15) as well as Zerfas (17) discuss the selection of equipment and list the addresses of some instrument distributors.

The lightest and least expensive piece of equipment will be the tape needed for measuring arm circumference. Cloth and paper tapes are not recommended because they wear out and stretch easily. Fiberglass or steel tapes are preferable.

A number of calipers are available for measuring fat folds. The two most widely used are the Lange (US made) and the Harpenden (British-made) models, either of which is suitable. These two calipers possess rectangular jaws and exert a constant pressure of 10 gm/mm2 Prices range between US $150 and US $200 for most calipers.

The measurement of weight, height, and supine length presents special problems. If subjects are to be measured in their homes, the equipment chosen should be light and sturdy. Where subjects are asked to come to a central station or designated location, the bulk and weight of the equipment is of less concern. The most widely used scales suitable for fixed locations are the Detecto baby (up to 16 kg) and adult (up to 140 kg) scales (beam balance). Suitable for use in house-to-house surveys is the British-made Salter scale (spring scale; up to 25 or 50 kg), which can be hung from a tree branch or a house beam. This scale has been used successfully in surveys such as those conducted by the US Center for Disease Control (CDC) in El Salvador. Currently, the portable weighing model 235 sells for £21. Field-worthy but heavier scales (i.e., 19 pounds) are available for measuring adults (15). Common bathroom scales are not reliable and are not recommended for weighing infants and young children, though they are suitable for cross-sectional surveys in adults.

Precise, sturdy instruments for measuring stature in fixed locales include the Harpenden Stadiometer (US $694 per unit). This instrument is impractical (100 pounds in weight) for the type of survey research that is generally carried out in developing countries. Some weight scales, such as the Detecto scales, have built-in stadiometers; but again, they are not suitable for house-to-house surveys A portable stadiometer, suitable for house-to-house surveys, is available in the Harpenden line of instruments (Harpenden Pocket Stadiometer. model 98-605, $46). Where reasonably straight walls are found in homes, a scale (e.g. fibre-glass tape) can be taped to the wall, and, with the use of a flat board to press against the top of the head after positioning the individual correctly, height can be measured with surprising accuracy.

Length in children can be easily measured with portable boards, usually made out of wood. Inexpensive models have been developed by UCLA and the CDC (17), WHO (9). and by INCAP; they can easily be replicated by local carpenters. Infants' boards have a fixed vertical end against which the child's head is positioned and a movable end which is positioned against the soles of the feet. A fiberglass tape is attached along the flat board. Sturdy infant measuring tables are also available commercially; these are, however, more expensive, very bulky and more appropriate for fixed locales. For example, the Harpenden infant measuring table weights 20 pounds and costs US $658. Perspective Enterprises (7466 Thrasher Lane, Kalamazoo, Michigan) has, however, developed a lighter (8 pounds) measuring board for around US $70.

Quality Control

All data collection activities should be monitored by a supervisor. He or she should meet periodically with the anthropometrists to reread the manuals and to review the measuring techniques; periodic visits to the fled are also recommended.

Data flow charts need to be prepared. Forms must be inspected shortly after collection to make sure all information is recorded properly. A record of names and dates or measurements should be kept at the field site, and the number of forms sent periodically to the central offices for data-processing should be counted and recorded. After the data are punched, the original forms should be filed for safekeeping. The punched data should be analysed as soon as possible to monitor reliability (from duplicate measurements) and to scan for outliers. A recent book edited by Jelliffe and Jelliffe (18) contains articles by Garn, Zerfas, and Nichaman. These three articles outline the most common types of errors that lead to outliers.

In scanning for outliers, values can be checked against the appropriate age-sex group distribution. For example, a value may be labelled as an outlier if it is more than two or three standard deviations above or below the mean. In addition, if two measurements are obtained, as recommended in this chapter, the differences between paired values can be compared to the measuring standard deviation. Values that differ by three times or more from the measuring standard deviation would be suspect. Ideally, one should remeasure the subjects whose values are questionable. Where this is not feasible, suspect values should be divided into those that are clearly impossible and those that are unlikely but plausible. The former should be deleted from the data files. Decisions as to what to do with unusual but plausible information are always difficult and arbitrary. If the quality of data is high, only a small percent of values (i.e., less than 1 per cent) may be involved, and little information will be lost if these values are deleted. Another strategy is to code suspect but plausible information so as to investigate its effect on key data analyses. Whatever is done, criteria and procedures should be specified and made available to those using the information.

A final note on quality control: To some, the supervision and quality control procedures suggested here may appear excessive and time consuming. They are certainly not. One impact of quality control procedures is that they foster a careful attitude on the part of field workers. To see "all this fuss" devoted to numbers (which may be largely meaningless to them) develops a sense of pride and responsibility. Otherwise, workers come to believe one is interested primarily in numbers, quality not being an issue.

Physical growth norms

The design presented in table 3.1. does not necessarily require the use of a norm or reference against which to compare the data collected. However, there is always the need to quantify the severity of nutritional problems and/or ascertain the pace of progress in nutritional status in a relative sense; for these purposes, norms become necessary.

The question as to whether growth data on children of European origin from developed nations are appropriate as norms for developing countries has been heatedly debated (19; 20; 21). A seemingly obvious solution would be for data on the well-nourished of each country, or even better for each ethnic group within a country, to serve as the norm for their ethnically appropriate group. This suggestion is not always practical because high quality data on sufficient numbers of the well-nourished do not yet exist for all ethnic groups. Moreover, some ethnic groups in some countries are so overwhelmingly poor and malnourished that a "well-nourished" group may not even exist. What is more interesting is that when data from well-nourished, preschool children of diverse ethnic groups are compared to the European norms, differences are usually rather small, particularly when viewed against the yew, large differences between the poor and the well-nourished in each ethnic group (19; 21). It would appear therefore, that growth data derived from European populations are justified as "norms" for developing countries - that is, as rough approximations and not as targets. (Appropriate to this point are the remarks of Waterlow et al. (22, p. 491 ) who in proposing NCHS data as norms for developing countries state: "If it is felt that in a particular population even well-nourished children are shorter in stature than children of the North American reference population, then it might be reasonable to set the target for height as 95 per cent of the reference height rather than 100 per cent. Decisions of this kind have to be taken locally, and it is not possible to make international recommendations about them.") Finally, the use of a similar norm by workers in developing countries facilitates the comparison of published results.

The National Center for Health Statistics (NCHS) has prepared new percentile charts for assessing the growth of children in the United States (23). These data have been recommended as the norms to use worldwide by a team of leading scientists from Europe, the United States, and the World Health Organization (22). Their proposal is likely to gain wide acceptance among researchers.

The adoption of the NCHS norms will not lead to very different estimates of the magnitude of malnutrition. The cut-off point of 75 per cent weight for age has been used to identify children with second-and third-degree malnutrition (i.e., < 75 per cent weight for age). Table 3.4 (see TABLE 3.4. Differences (kg) between the NCHS and the Iowa-Harvard Norms with Respect to the Cut-off Point of 75 % Weight for Age ) shows the criterion values using the new NCHS norms and the values corresponding to the most widely used norm of the past, the Iowa-Harvard data. (24) Mean differences are small particularly for boys. The use of the NCHS norms, instead of the Iowa-Harvard data, would result in a slightly reduced prevalence of malnutrition in boys and a small increase in the prevalence in girls.

Norms for arm and skinfold values are not yet available from NCHS. When they do become available, note of the fact that children in the US are heavier and fatter than those in other European-origin populations (e.g.. British children) should be taken into account. Interestingly, it has recently been pointed out that children from "privileged" groups in some developing countries are even fatter than those from the US (21).

TABLE 3.4. Differences (kg) between the NCHS and the lowa-Harvard Norms with Respect to the Cut-off Point of 75 % Weight for Age

* The NCHS norms use data from the Fels Research Institute for children up to 3 years of age and from a national sample thereafter
** 75 per cent of the age-sex-specific 50th percentile values

Analysis

Length for age and weight for length are the key variables for evaluating the impact of nutrition interventions Data on arm circumference, triceps skinfold, arm and fat areas and subscapular skinfold should also be used as we have stressed. When norms of comparison from NCHS become available, the same reporting procedures discussed below for weight for length and length for age should be followed.

Waterlow et al. (22) propose that data on children be analysed by three months groups during the first year of life (i.e., 0-2.99. 3.0-5.99, 6.0-8.99, 9.0-11.99 months), by one-year groups for older children (1.0-1.99, etc.). The recommended sample sizes are 100 children in each age group; failing this, Waterlow et al. recommend collapsing some of the age groups. WHO (9) proposes seven larger age groupings: infants (0-5.9 and 6-11.9 months), preschool children (12-23.9, 2447.9, and 48-71.9 months), and school children (72-95.9 and 96-119.9 months). (The issue of sample size is an important one which is best approached by power analyses (251. Needed for sample size calculations are specifications of expected effects, level of statistical significance, and power.)

A common procedure in analysing data is to test whether the number of observations falling between specific percentile values of the norm have changed (i.e., between 50th and 60th percentile values, 60th and 70th percentile, etc.). This method is inappropriate for data on stunting indicators from developing countries because many children, often as many as 90 per cent, will have values that fall below the 5th percentile. Many researchers in the past have therefore expressed the observations as a percentage of the mean or median value for age. For example, per cent weight for age is the actual recorded weight divided by the 50th age-sex specific percentile weight value in the norms, the result multiplied by 100. There are at least two problems, one minor and one major, with the use of per cent of mean or median values. First, per cent of median values are not, as the seemingly equal units of measurement would suggest. equivalent across age. For example, with respect to the NCHS norms, 95 per cent of median height corresponds to the 8th percentile at 12 months of age but to the 12th percentile at 48 months of age.

Differences of this magnitude are not generally important. Of greater concern is the fact that per cent of median values are not equivalent across measures, the differences in this case being large. For example, for girls and once again in reference to the NCHS norms, 90 per cent of median corresponds to the 13th percentile for weight for length at 95 cm but to less than 1 percentile for height for age at 36 months which is approximately the age when girls achieve a median length of 96 cm (22). Failure to take this into account has led to fallacious statements such as: "Measure A appears to be more retarded than measure B because the per cent of median value is less for A than for B."

The use of standard deviation scores or Z-scores is an elegant and simple solution to the problems noted in the case of per cent of median value. Standard deviation scores are equivalent across age and across measures. Waterlow et al. (22) have proposed methods of classification based upon Z-scores in the hope that they "will be widely acceptable and thus make international comparisons possible." They include Z-scores grids and suggest that data be presented in summary tables by Z-score categories (i.e., by half or one standard deviation groupings). Waterlow et al. also propose that cross-tabulations of stunting (length for age) and wasting (weight for length) be presented and that tables of means (medians) and standard deviation in the original units (cm or kg) and in Z-score units should be presented for all sex-age groups.

There are many approaches to the testing of the experimental design shown in table 3.1. This simple design, as will be recalled, has a test and a control sample and a baseline and an intervention phase. For example. differences between groups can be tested by an analysis of variance (ANOVA). Independent variables would be "sample" (test/control), " phase " (baseline/intervention), " sex" (male/female), and " age" (0-1, 1-2, etc.). Dependent variables would be each of the anthropometric variables either in the original units (i.e., cm or kg) or Z-score values.

WHO (9) has published a data analysis manual for height and weight data. Numerous illustrative tables and figures from a simulated analysis of a World Food Programme in the country of "Barico" and the "Kimani" project are included.

References

1. R. Chambers, R. Longhurst, D. Bradley and R. Peachem, "Seasonal Dimensions to Rural Poverty Analysis and Practical Implications." Discussion Paper no 142 (Institute of Development Studies at the University of Sussex, Brighton, 1979).
2. S.M Garn. "Optimal Nutritional Assessment," in D.B. Jelliffe and E.F.P. Jelliffe, eds., Human Nutrition: A Comprehensive Treatise, Vol. 2: Nutrition and Growth. (Plenum Press, New York and London, 1979)
3. J.C. Waterlow, "Classification and Definition of Protein-Calorie Malnutrition," British Medical Journal, 3: 566-569 (1972).
4. C. Yarbrough, J.-P. Habicht, R.E. Klein, R. Martorell, A. Lechtig and G. Guzman, "Response of Indicators of Nutritional Status to Nutritional Interventions in Populations and Individuals," Conference on Evaluation of Child Health Services, the John E. Fogarty International Center, NIH, Bethesda, Maryland, March 7-9, 1977, in S.J. Bosch and J. Arias. eds., Evaluation of Child Health Services. The Interface Between Research and Medical Practice, DHEW Publication No. (NIH) 78-1066 (U.S. Government Printing Office, Washington D.C., 1978).
5. C. Yarbrough, J.-P. Habicht, R. Martorell and R.E. Klein, "Anthropometry as an Index of Nutritional Status," Conference on Physical Anthropology and Nutritional Status, Burg Wartenstein, Austria, August 6-15, 1973. in A.F. Roche and F. Falkner, eds.. Nutrition and Malnutrition (Plenum Press, New York, 1974).
6. D.H. Rao and A.N. Naidu, "Nutritional Supplementation - Whom Does It Benefit Most?" American Journal of Clinical Nutrition, 30: 1612-1616 (1977).
7. R. Martorell and R.E. Klein, "Food Supplementation and Growth Rates in Preschool Children," Nutrition Reports International, 21: 447-454 (1980).
8. R. Martorell, R.E. Klein and H. Delgado, "Improved Nutrition and Its Effects on Anthropometric Indicators of Nutritional Status," Nutrition Reports International, 21: 219-230 (1980).
9. WHO, Measurement of Nutritional Impact, WHO/FAO/79.1. /WHO, Geneva, 1979).
10. D. Morley, "Nutritional Surveillance of Young Children in Developing Countries," International Journal of Epidemiology, 5: 51-55 (1976).
11. R. Martorell, C. Yarbrough, A. Lechtig, H. Delgado and R.E. Klein, "Upper Arm Anthropometric Indicators of Nutritional Status," American Journal of Clinical Nutrition, 29: 46-53 (1976).
12. R. Martorell, J.-P. Habicht and R.E. Klein. "Anthropometric Indicators of Changes in Nutritional Status in Malnourished Populations," in B.A. Underwood, ea., Methodologies for Human Population Studies in Nutrition Related to Health, NIH Publication No. 82-2462 (U.S. Government Printing Office. Washington D.C, 1982), pp. 99 110.
13. J.-P Habicht, "Standardization Procedures for Quantitative Epidemiologic Field Methods," in Manual of Internationally-Comparable Growth Studies in Latin America and the Caribbean (Pan American Health Organization, Washington D C., 1973). Revised form in Spanish in Boletin de la Oficina Sanitaria Panamericana. 76: 375-384 (1974).
14. R. Martorell, J.-P. Habicht, C. Yarbrough, G. Guzman and R.E. Klein, "The Identification and Evaluation of Measurement Variability in the Anthropometry of Preschool Children," American Journal of Physical Anthropology, 43: 347-352 (1975).
15. J.S. Weiner and J.A. Lourie, Human Biology. A Guide to Field Methods. (Blackwell Scientific Publications, Oxford and Edinburgh, 1969).
16. D.B. Jelliffe, "The Assessment of the Nutritional Status of the Community," Monograph Series No. 53 (World Health Organization, Geneva, 1966).
17. A.J. Zerfas, "Anthropometric Field Methods: General," in D.B. Jelliffe and E.F.P. Jelliffe, eds., Human Nutrition. A Comprehensive Treatise, Vol. 2: Nutrition and Growth (Plenum Press, New York and London, 1979).
18. D.B. Jelliffe and E.F.P. Jelliffe, eds., Human Nutrition A Comprehensive Treatise, Vol. 2: Nutrition and Growth.
19. J.-P. Habicht, R. Martorell, C. Yarbrough, R.M. Malina and R.E. Klein, "Height and Weight Standards for Preschool Children: How Relevant are Ethnic Differences in Growth Potential?" Lancet 1: 611-615 (1974).
20. H. Goldstein and J.M. Tanner, "Ecological Considerations in the Creation and Uses of Child Growth Standards, " Lancet 2: 582-585. 1980.
21. P.L. Graitcer and E.M. Gentry. "Measuring Children: One Reference for All." Lancet 2: 297-299 (1981).
22. J. C. Waterlow, R. Buzina, W. Keller, J.M. Lane, M.Z. Nichaman and J.M Tanner, "The Presentation and Use of Height and Weight Data for Comparing the Nutritional Status of Groups of Children Under the Age of 10 Years," Bulletin of the World Health Organization, 55: 489 (1977).
23. P.V.V. Hamill, T.A Drizd, C.L. Johnson. R.B. Reed. A P. Roche and W.M. Moore. "Physical Growth: National Center for Health Statistics Percentiles," American Journal of Clinical Nutrition, 32: 607-629 (1979).
24. H.C. Stuart and H.V. Meredity. "Use of Body Measurements in the School Health Programme," American Journal of Public Health, 36: 1365 (1946).
25. J. Cohen, Statistical Power Analysis for the Behavioral Sciences. Rev. ed. (Academic Press, New York, 1977)

Contents - Previous - Next