Contents - Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
Marian F. Zeitlin, Suganya Sockalingam, Mirette Seireg, and Josefina Bonilla
A new, simplified anthropometric field tool for detecting nutritional stunting, the "tallstick," has been developed in community-based pilot programmes in Nicaragua and Nigeria. The folding stick, marked to indicate height-for-age cut-off points for ages from birth to seven years below which the child is considered nutritionally stunted, is compared with arm circumference measurements as a simplified field tool. Sensitivity-specificity analyses contrasting the ability of height-for-age and arm-circumference cut-off points to detect malnutrition showed height for age to be as good as arm circumference for detecting low weight for age, weaker for low weight for height, and stronger for a composite score of any degree of malnutrition in 1,070 children 0-72 months old in urban Managua with a prevalence of wasting of 0% -4%, of stunting of 17%-38%, and of underweight of 4%-22%. The importance of detecting stunting, the likelihood of reducing stunting rates, and the feasibility of using the tallstick and its variants are discussed.
Key words: anthropometric field tool, pre-school malnutrition, nutritional stunting, height for age.
A pilot test is being conducted in Nicaragua and Nigeria of a simple new anthropometric tool for community-based detection of stunting-the "tallstick," shown in use in figures 1-3. It resembles a folding yardstick and is marked off in graduations indicating heights for children by age, from birth to seven years old, below which the child is considered nutritionally stunted. One side is blue for boys and the other pink for girls. Infants under 12 months old are measured lying down.
This paper presents and discusses the malnutrition prevalence rates and the sensitivity-specificity analyses supporting the decision to develop and use the tallstick in Nicaragua, and looks at some of the advantages of its use and possible variations on the concept.
Two lines of reasoning support the introduction of a new simplified measure to assess stunting. The first rests on the ability of stunting to rival arm circumference as a proxy measure for low weight for age and as a predictor of mortality risk. The second rests on the intrinsic benefits of detecting stunting, particularly in child populations in which stunting rates are high and wasting rates too low to be a community health problem.
A sample survey of 1,070 households with children below the age of six years was conducted in low income neighbourhoods of urban Managua between June and December 1988, using procedures approved by the Tufts University Committee for Protection of Human Subjects. Every fourth household was sampled systematically from starting points representative of urban settlements targeted by the Ministry of Health as high-priority areas for primary health care services. In each household with more than one preschool child, one index child was selected randomly according to probabilities that increased the sample sizes of children in the first and second years of life relative to the numbers between two and six years old. Ages were determined from birth certificates. Supine length was measured for children under two years old and standing height for children two years old and older, using techniques recommended by WHO , with a portable measuring board (Zeitlin, MIT, Cambridge, Mass., USA). Weight was measured using spring balances (ITAC Model 800, ITAC Corp., Silver Spring, Md., USA). Circumferences were taken at the midpoint of the upper left arm between the acromion and the olecranon, using a commercial fibreglass tape measure. Heights and weights were transformed into Z scores of the NCHS reference standards as specified by WHO .
For children over 18 months old, a cut-off point of -2 standard deviations (SD) in height for age (HAZ) was selected for the tallstick because this value delimits stunting according to standards commonly used for international surveillance , and also because it selected slightly more than a third of the children, a proportion judged suitable to motivate community-based action and thus potentially able to mobilize community resources. This choice was in keeping with recommendations that the best level for screening under conditions of scarcity is the level that will deliver the number for which resources will suffice [3; 4]. The stick was prepared for trial in community-based assessment and action programmes, from which the data would be forwarded to regional levels for comparative inspection using surveillance standards.
For children below 18 months, age-specific points above-250 were selected. The need for length cutoff points for detecting malnutrition to vary with age was noted by Waterlow , who observed that it is important to detect an inadequate rate of growth before malnutrition becomes severe. The levels chosen were Z-score values corresponding to the following percentiles: 0-8 months, 10th percentile; 9-14 months, 5th percentile; 15-17 months, 3rd percentile. These cut-off points yielded a constant detection rate of about 37% from 0 to 6 months and from 18 to 60 months, and a marginally higher rate of 42% between 6 and 18 months, when the stunting process was most active.
For sensitivity and specificity analyses, the 18-72 month levels of HAZ were varied between -4 SD and -0.5 SD. At each level, the under-18-month values remained in the proportions noted above in relation to each other and to the 18-72-month values. Weight-for-age (WAZ) cut-off points were created using the same age-specific cut-off points used for HAZ. The cut-off points for weight for height (WHZ) remained the same across all ages.
In accordance with current practice in Nicaragua , levels of - 1 SD and -2 SD were used on all three indicators, HAZ, WAZ, and WHZ, to define mild and moderate malnutrition respectively, incorporating the below-18 month adjustments noted above. The three indicators were also combined into a composite measure of "any malnutrition," defined as being below-1 SD (mild) or -2 SD (moderate) on any indicator.
The sensitivity and specificity analyses used simple arm-circumference cut-off points that do not change with age or height and varied these circumferences from 13.5 to 16 cm. The decision to use simple arm circumference was based, first, on observations by Seireg  that health volunteers in Nicaragua encountered difficulties in using the UNICEF-Brasilia arm circumference strip, which has multiple age-specific cut-off points, and, second, on findings that simple arm circumference, as popularized in the Shakir strip , has as satisfactory sensitivity and specificity for detecting low WAZ and WHZ as arm circumference for age or arm circumference for height [8; 9]. Simple arm circumference was judged to provide an adequate screening tool between 6 and 60 months. based on findings that it detects infants between 6 and 12 months who are less malnourished than those above 12 months but nevertheless at high risk .
Feasibility testing for the earliest version of the tallstick was conducted in Managua by the same trained field workers who conducted the household survey mentioned above. These workers took duplicate measurements on 71 recumbent infants between the ages of 8 and 18 months. They first used the tallstick, as illustrated in figure 1 (they stood their notebooks behind the child's head and held the foot closest to positive predictive value the stick perpendicular to the stick). They then remeasured the lengths using the length board.
TABLE 1. Definition of sensitivity, specificity, and positive predictive value
|Screening indicator results||True malnutritiona||Total|
|Positive||(A) true positives||(B) false positives||A + B|
|Negative||(C) false negatives||(D) true negatives||C + D|
|Total||A + C||B + D||N|
|Sensitivity = A/(A + C)|
|Specificity = D/(B + D)|
|Positive predictive value = A/(A + B)|
TABLE 2. Prevalence of malnutrition (percentages) by indicator and age group
< -2 SD of Z score
Arm circumference (cm)
|UNICEF Brasilia stripb|
|0-5||105||17 (37)||1||4 (20)||69||77||85||90||4|
|6-11||135||27 (42)||2||21 (33)||36||44||53||67||11|
|12-17||113||35 (43)||2||22 (30)||10||19||41||59||2|
a. Values in parentheses represent percentages below the
following age-specific cut-off points: 0-8 months. 10th
percentile: 914 months. 5th percentile; 15-17 months. 3rd
b. Age-specific cut-off points: 0-1 month, 9 cm; 1-5 months, 10.5 cm; 6-30 months, 12.5 cm; 31-60 months, 13.5 cm (over 60 months undefined).
The SPSS-X programme (Statistical Packages for the Social Sciences)  was used to compute variables defining malnutrition, generate simple age specific prevalence rates, and create cross tabulations required for sensitivity and specificity analyses. Sensitivity, specificity, and positive predictive value were defined as shown in table 1.
Table 2 presents malnutrition prevalence rates within six-month age groups as defined by -2 SD of the NCHS/WHO standard  for HAZ, WAZ, and WHZ; by the age-adjusted cut-offs defined below 18 months for HAZ and WAZ (in parentheses after HAZ and WAZ on the table); by simple arm circumferences ranging from 13.5 to 15 cm; and by the UNICEF-Brasilia arm-circumference strip currently used in Nicaragua , which has the following age specific cut-off points for defining malnutrition: 0-1 month, 9 cm; 2-5 months, 10.5 cm; 6-30 months, 12.5 cm; and 31-60 months, 13.5 cm.
Table 2 shows that stunting is the only form of malnutrition with a high prevalence (17%) of infants below -2 SD during the first six months of life. As expected, all of the arm-circumference cut-off points select a high proportion of normal babies in the under-six-month age group. The prevalence of stunting increases as the prevalence of malnutrition detected by arm circumference decreases. Stunting is more than twice as prevalent as low weight for age. There is virtually no wasting as defined by weight for height below -2 SD, and very little malnutrition detected by the UNICEF-Brasilia strip.
Table 3 reports sensitivity-specificity analyses comparing the ability of low HAZ and arm circumference to detect low WAZ, WHZ, and "any malnutrition." Similar analyses of HAZ over the age range from birth through six years were virtually identical to those shown, which were limited to the 6-60-month age window across which simple arm circumference can be used. Figures 4-6 present the sensitivity and specificity curves for the comparisons in table 3.
Section A of table 3 and figure 4 show that HAZ and arm circumference have very similar sensitivities, specificities, and positive predictive values as measures of underweight in this population, and that height for age has slightly higher sensitivity than arm circumference at specificities between 65 and 90. Section B and figure 5 indicate that arm circumference has higher specificity and sensitivity than HAZ in identifying low WHZ. However, section C and figure 6 show that HAZ has far higher sensitivity and specificity than arm circumference for detecting "any malnutrition." In fact, HAZ at the tallstick cut-off points of-2 SD identifies "any malnutrition" with a specificity and a positive predictive value of 100.
In the pretest of the stick by the Managua survey team, the sensitivity of the tallstick to detect stunting as defined by accurately measured HAZ scores was 81, the specificity was 98, and the positive predictive value was 98.
We first evaluate how well HAZ (for which the tallstick is a simplified measure) compares with arm circumference in detecting "true malnutrition" as defined by WAZ, WHZ, and the composite measure of "any malnutrition" according to the NCHS reference standards. We then consider the merits of defining HAZ as a measure of true malnutrition. Finally, we address the feasibility of using the tallstick and variants of the tallstick concept.
TABLE 3. Sensitivity, specificity, and positive predictive value of height for age compared with arm circumference for detecting low weight for age, low weight for height, and "any malnutrition" in children 660 months old
|A. Weight for age||B. Weight for height||C. Any malnutritiona|
|Detection of levels below - 1 SD|
|Ht/age c -3 SD||27||99||96||24||86||20||20||100||100|
|Arm circ. < 14 cm||27||98||93||48||90||40||19||98||96|
|Ht/age < -2 SD||61||87||85||47||63||16||51||100||100|
|Arm circ. < 15 cm||57||88||85||81||71||29||44||89||92|
|Ht/age< - 1 SD||92||53||70||76||30||14||93||100||100|
|Arm circ. < 16 cm||85||61||71||97||41||20||72||61||85|
|Detection of levels below2 SD|
|Ht/age < -3 SD||56||94||70||45||85||7||36||100||100|
|Arm circ. < 14 cm||52||93||64||82||86||7||27||93||92|
|Ht/age < -2 SD||87||72||42||55||62||2||94||100||100|
|Arm circ. < 15 cm||80||74||42||100||65||4||52||76||60|
|Ht/age < - 1 SD||99||36||26||82||29||2||99||49||50|
|Arm circ. < 16 cm||96||45||28||100||37||2||77||45||50|
Sns = sensitivity. Spc = specificity. Ppv = positive
a. Below - 1 or -2 SD on any of the three indicators weight for age, height for age, or weight for height.
FIG. 4. Sensitivity and specificity of height for age and arm circumference in detecting weight for age below -1 or -2 SD (height for age from -4 to -0.5 SD; arm circumference from 13 to 16.5 cm)
FIG. 5. Sensitivity and specificity of height for age and arm circumference in detecting weight for height below ---1 or-2 SD (height for age from -4 to -0.5 SD; arm circumference from 13 to 16.5 cm)
FIG. 6. Sensitivity and specificity of height for age and arm circumference in detecting any mild or moderate malnutrition, any Z score below - 1 or -2 SD (height for age from -4 to -0.5 SD; arm circumference from 13 to 16.5 cm)
Comparison of the tallstick against arm circumference as a measure of true malnutrition
At present, arm circumference is the only simple proxy measure for WAZ and WHZ. Simple arm circumference can be used by community workers with little or no formal education . It identifies many, but not all, of the same children who are malnourished according to weight-for-age and weight-for-height criteria [13; 14; 15], and it appears to identify children who are at risk of mortality as accurately as weight for age and more accurately than weight for height [16; 17].
In both clinic-based and community-based programmes WAZ is widely recognized to be the best single indicator for detecting and monitoring malnutrition. This recognition is manifested in the global emphasis on growth monitoring in child-survival strategies  and the use of weighing for community assessment of nutritional status I 19]. Improvement in thinness and in linear growth are both reflected in weight gain. However, weighing programmes in developing countries often have had poor success rates because weighing requires an accurate portable scale, knowledge of the child's age, and ability on the part of the health worker to graph weight against age. Therefore, it is necessary to seek simpler measurements that reflect WAZ or that perform as well as WAZ.
Weight for height is an indicator of current acute malnutrition that is invaluable during famine or other situations of overall food scarcity [20; 21; 22]. Measuring weight for height tends to be at least as problematic for field workers with low levels of formal education as measuring weight for age. Although knowledge of the age of the child is not required, height or length measurements must be made and graphed against weight measurements. An error in either height or weight distorts the outcome.
HAZ cut-off points, which can be used starting at birth, pass the test of competing with arm circumference cut-off in detecting low weight for age. Yet neither HAZ nor arm circumference has good sensitivity in detecting low WAZ when specificity and positive predictive value are high, which suggests that these indicators detect different dimensions of inadequate weight status.
HAZ cannot compete with arm circumference in detecting low WHZ. This finding is not surprising, given that WHZ and arm circumference both primarily reflect the presence of soft tissue, whereas height reflects skeletal length. Detection of low WHZ lacks relevance for the Nicaraguan population, however, because of the very low prevalence of malnutrition as measured by WHZ. This low prevalence is reflected in the poor positive predictive value of arm circumferences as an indicator of low WHZ, as explained by Habicht .
Our final, inclusive, definition of true malnutrition considers whether the children are malnourished according to any of the NCHS Z scores. Here we note from section C of table 3 that arm circumference has very poor sensitivity at the 14 and 15 cm cut-off points at which specificity and positive predictive value are acceptably high. In other words, arm circumferences that are small enough to eliminate truly normal children fail to detect the majority of those who should be classified as malnourished. Low HAZ, on the other hand, has good sensitivity and perfect specificity and positive predictive value because it is measuring itself.
The intrinsic merits of detecting stunting
Stunting, which is far more prevalent than wasting [23; 24], is a public health problem in its own right. The myth of the "small but healthy" child  is no longer tenable. It has been shown repeatedly that extreme short stature of preschool children in developing countries is not genetic but is a consequence of poor socio-economic conditions [26; 27; 28] and is achieved at an unacceptable cost [29; 30]. Stunted infants and young children in the process of stunting have been found to be at increased risk of mortality [16; 31]. Stunting reflects inadequate recovery from illness . Stunted children have more prolonged infections than children of normal stature . Stunted children also are at increased risk of cognitive disabilities [34; 35; 36; 37]. Moreover, stunted children can recover to statures and growth rates within the normal range . Therefore, stunting should be detected and its progress prevented.
The appropriateness of using height for age as a measure of malnutrition in community-based programmes
As detailed by Habicht and Habicht et al. [3; 4], the merits of an indicator are dictated by its uses. A major purpose for developing simplified indicators for use by less-educated community members is to empower action at the grassroots level. Indicators used for this purpose should be able to measure improvement in response to intervention, in order to provide positive feedback to the community. We believe that HAZ will demonstrate improvement in response to interventions that (1) improve diet quality directly, (2) increase the levels of parental and community investment in young children in a manner that is nonspecific to nutrition, or (3) improve overall socioeconomic indicators.
Linear growth failure is a non-specific response to the deficiency of a variety of nutrients, and these nutrients tend to be found together in naturally occurring diets that include modest amounts of animal protein foods . Growth retardation also is mediated by, and may be a non-specific response to, illness and environmental stress. The nonspecific contribution of many factors to inadequate growth invalidates the attempt to use sensitivity-specificity analysis on a variable-by-variable basis to estimate the adequacy of a given indicator for detecting a growth response to an integrated community-based intervention, although such analysis would theoretically be desirable . For practical purposes, however, both univariate and multivariate analyses examining the correlates of malnutrition against the indicators as dependent variables serve a purpose similar to the creation of sensitivity distributions. The dependent indicator that shows the greatest change in response to variability in the dietary and environmental independents should be the most sensitive indicator for registering improvement.
Height for age was the most sensitive dependent indicator in such analyses in our moderately malnourished sample in Nicaragua. HAZ varied more sensitively than WAZ or WHZ with measures of feeding (excluding breast-feeding) and of mother-child interaction and with most of the common socio-demographic determinants of malnutrition. It was therefore reasonable to assume that interventions that address child feeding and care in the context of the household environment and family coping mechanisms would be more measurable in improved HAZ than in WAZ or WHZ. The design of the tallstick represents the application of these conclusions for programmatic purposes, according to the positive deviance research approach  used in the Nicaraguan study.
Given the relatively low malnutrition rates in our samples, these conclusions are consistent with Habicht's findings  that increments in height are more sensitive than increments in weight in identifying the benefits of an improved diet in moderately malnourished children, whereas weight is more useful in severe malnutrition. They also are consistent with Keller's observation  that a common, although not universal, course of nutritional development at the population level starts with the elimination of wasting, moves to the elimination of stunting, and ultimately confronts the risk of obesity.
The ability of height for age to measure improvement should depend on the kinds of dietary changes that occur with intervention. Studies of pre-school children in industrialized countries [40; 41] demonstrate that growth in stature of children on nutritionally adequate diets, according to US and Dutch recommended daily allowances, depends on the amount of animal food in these diets. Children whose diets are fully adequate in protein and energy but who receive minimal amounts of animal food still grow at rates significantly below the reference standards [40; 41]. Extrapolation from heights reported for children on macrobiotic diets in a study in Holland , where average child growth exceeds the NCHS reference standards, suggests a 24% rate of stunting (below -2 SD of the Dutch reference standard, for 1-3-year-old children whose median stature is at the 10th percentile, assuming normality and the same standard deviation as in the reference population). These childrens' diets were fully adequate in energy and protein but were low in calcium and vitamin D. Moreover, they were born of mothers who did not grow up on macrobiotic diets.
Golden , reviewing the growth effects of supplementary feeding programmes, notes that most studies that have reported a height gain have used a milk- or soya-based supplement. He surmises that poor growth effects of other programmes have resulted partially from failure to take into account the variety of "fellow traveler" nutrients, apart from protein, that are contained in milk and other animal foods and that are necessary for growth. He makes particular reference to Malcolm , who documented different growth responses of children in boarding schools in Papua New Guinea to different diet supplements and found that the growth response was proportional to the amount of milk in the supplement.
Stunting does not appear to be a hopeless consequence of poverty that cannot be prevented or remedied by community action. The expensive diets consumed in North America and western Europe should not be necessary to reduce stunting and its negative consequences. Recent research indicates, for example, that even among low-income populations, those children who receive modest amounts of foods of animal origin show better linear growth [36; 37; 43] as well as better cognitive performance [37; 44] (also M. Sigman, personal communication) than children who receive little or no animal foods. Dietary data from the Nicaraguan survey  indicate that children of all ages above six months who were normal in their height drank significantly more cow's milk than the children who were stunted. Controlling for age, children with normal height received more foods of animal origin other than milk and more frequent feeding than did stunted children.
In spite of significant decreases in the purchasing power of low-income families, stunting should not be irreversible in Nicaragua. Agricultural land and water resources are plentiful with respect to population density, and there is a national commitment to public health and nutrition. The Ministry of Agriculture and Land Reform supports and promotes home, school, and community gardens as well as husbandry of small animals through the Nicaraguan Food Programme. Although self-sufficiency to meet basic food needs is the goal of the programme, donations of powdered milk and canned meat have been made available to selected communities primarily through the World Food Programme and the European Economic Community. These efforts are organized through an extensive network of community workers that could be linked to growth promotion. UNICEF and the Nicaraguan Food Programme have supported food fairs demonstrating the use and preparation of soybeans.
In Nigeria, where the Positive Deviance Project found a 51% prevalence of stunting and a 2% prevalence of wasting among two-year-olds , relatively small differences in the quantity of meat that mothers claimed to be suitable to give to two-year-old children were correlated with the height for age of the child. Ethnographic data (T. A. Aina, personal communication) confirmed that low-income households had the resources to reallocate a greater proportion of animal foods to young children but were severely limiting these foods as a part of the process of training the children to respect and obey their elders.
The positive-deviance intervention in urban Nigeria appeals to the desire of low-income parents for their children to do well in school. Programme materials motivate parents to attempt to achieve minimal protein adequacy standards for their pre-schoolers as established by the joint FAO/WHO/UNU expert consultation group . The programme is testing a simple pictorial scoring sheet, which gives points for portions of animal foods and vegetable protein foods. The sheet asks, "Can you score 100 for your child today?" Given the nature of this intervention, which also stresses greater overall investment in each child, its success may be measurable in reduced stunting.
Use of the tallstick and its variants
Age determination and accuracy of measurement are primary concerns in testing the tallstick. The initial pretest reported above was conducted by anthropometrically trained field workers, who disliked the stick because of its imprecision and its hand-made appearance. The accuracy of the tallstick assessment was analysed against ages determined by computer from birth records. The tallstick is currently being pilot-tested in projects using community-based workers, in which not only its sensitivity and specificity for detecting accurate measures of stunting will be documented but also its acceptability and ease of use to motivate action to improve the nutritional conditions of children.
An immediate advantage of the tallstick over formal measurements is that the stick can be used with minimal cooperation from the child to eliminate from screening all children who are obviously taller than the markings for their age. Children who resist lying down to be measured can remain standing at any age provided they are taller than their age line. Recumbent measurement is needed only below 12 months for those children who cannot stand or whose standing height is shorter than their age marking.
Just as many useful methods have been devised to measure arm circumference , the tallstick concept invites multiple applications. Potential variants include a similarly age-marked sheet of paper or cloth for recumbent infants. For individual growth monitoring, a strip of adhesive paper with the same age markings as those on the stick could be attached to a post or wall in the child's home, aligning the child's current age with his or her current height. Failure to maintain height at the level of age over time would indicate progressive stunting, while height that progressively exceeded age would denote improving growth. An accurate stadiometer could be constructed with markings similar to those on the stick and might be used by trained workers for formal surveillance purposes in situations where computational facilities are limited. The addition of numbers representing ages at appropriate heights along the top of the white lower section of the Nabarro weight-for-height wall chart  would transform this chart into a two-function weight-for-height and height-for-age chart.
If it would be useful for programme purposes to determine two levels of malnutrition, moderate and severe, as incorporated into the Iringa Project in Tanzania , one side of the tallstick could be marked with levels corresponding to -2 SD (moderate) and the other side with levels corresponding to -3 SD (or some other cut-off level measuring severe stunting). In that case, separate sticks could be made for boys and girls, or the same cut-off points could be used for both sexes. Arm circumference might also be used for detecting severe acute malnutrition requiring immediate intervention.
The authors gratefully acknowledge the assistance of Drs. A. O. Grange, Alfred Zerfas, Peter Greaves, and Lindsay Allen and of the Nicaraguan and Nigerian UNICEF Positive Deviance in Nutrition Project field teams.
Contents - Previous - Next