Measuring Malnutrition The Role of Z Scores and the Composite Index of Anthropometric Failure (CIAF)
Author(s): N Seetharaman, TV Chacko, SLR Shankar, AC Mathew
Vol. 32, No. 1 (200701  200703)
ABSTRACT
Background: The current WHO recommendation is to use the ZScore or SD system to grade undernutrition which allows
us to measure all the three indices and express the results in terms of Z scores or standard deviation units from the median
of the international reference population.
Objectives: To estimate the prevalence of undernutrition among underfive children in Coimbatore slums, using the ZScore
system of classification and the recently constructed Composite Index of Anthropometric Failure (CIAF). 2. To compare the
ZScore system with the Indian Academy of Pediatrics (IAP) classification of undernutrition.
Methods: Nutritional assessment was done using anthropometry and clinical examination. Children were weighed and
measured as per the WHO guidelines on Anthropometry. EpiInfo 2002 software package was used to calculate the Z scores
and for statistical analysis.
Results: Only 31.4% of the children studied were normal; 68.6% were in a state of “Anthropometric Failure”. As per the Z
score system, 49.6% were underweight (21.7% severely); 48.4% were stunted (20.3% severely) and 20.2% were wasted
(6.9% severely). Whereas, as per IAP criteria, 51.4% were undernourished and 3.2% were severely undernourished. Using
Underweight (low weightforage) as the only criterion for identifying undernourished children (as done in the Integrated Child
Development Services currently) may underestimate the true prevalence of undernutrition, by as much as 21.9%.
Conclusions: More widespread use of the ZScore system is recommended for identifying all the facets of undernutrition.
Estimates of the true prevalence of undernutrition must incorporate a composite index of anthropometric failure.
Keywords: ZScore Classification, Composite Index of Anthropometric Failure (CIAF), Anthropometry, Malnutrition, ICDS.
India has the highest percentages of undernourished children in the world^{1}. In any community, underfive children are one of the most vulnerable groups for nutritional deficiencies, owing to many factors ranging from Low Birth Weight to maternal ill health to socioeconomic and environmental factors^{2}. Most of these problems are accentuated and are highly unfavorable for a child growing in today’s slums. It must be stated here that, although the words undernutrition and malnutrition are being used interchangeably, malnutrition is a broader term that includes undernutrition and overnutrition. In this study, the term undernutrition is used.
There have been quite a few attempts at grading the degrees of undernutrition. Weightforage classifications (Gomez, I.A.P) are the most commonly used. Heightforage and Weightforheight classifications (McLaren, Waterlow’s) have been used less frequently^{3}. Each of these classifications use different sets of reference data and each system employ different cutoff points to decide who is normal and who falls under mild, moderate, or severe undernutrition. The cutoff points are usually a certain percent of the mean/median or a percentile, of the reference population. Most of the cutoff points are admittedly arbitrary and do not carry a prognostic significance for any given individual child^{4}. Employing different reference values and cutoff points has been a major hindrance in comparing data across various studies and countries. In India, the nationwide Integrated Child Development Services (ICDS) program uses the IAP criteria to grade undernutrition.
None of these classifications address all the three indices of undernutrition – Stunting, Wasting and Underweight. Stunting (Low heightforAge) is an indicator of chronic undernutrition due to prolonged food deprivation and/or illness; Wasting (Low weight for height) is an indicator of acute undernutrition, the result of more recent food deprivation and/or illness; Underweight (Low weightforage) is used as a composite measure to refl ect both acute and chronic undernutrition, although it cannot distinguish between them^{4}.
The current WHO recommendation is to use the ZScore or SD system to grade undernutrition. This system allows us to measure all the three indices and express the results in terms of Z scores or standard deviation units from the median of the international reference population, developed from anthropometric data collected in the United States by the National Center for Health Statistics (NCHS). Children who are more than 2 SD below the reference median (i.e. a ZScore of less than 2) are considered to be undernourished i.e. to be stunted, wasted or to be underweight. Children with measurements below 3 SD (a ZScore of less than3) are considered to be severely undernourished^{4}.
Although widely recommended, the Z Scores have not been widely in use in India, especially in communitybased studies.
This is in spite of the fact that the nationally representative National Family Health Survey 2 (NFHS 2) uses Z scores to grade undernutrition^{5}. The Z score system is not yet being widely used among researchers, probably because of the perceived difficulty in calculating the ZScores. The EpiInfo 2002 software developed and distributed by the Center for Disease Control, Atlanta^{6} has eliminated this problem. Besides being sexspecific and able to measure all the three indices, the main advantage of Z scores is that it allows comparison across indicators and countries. Another advantage is that, ZScores can be subjected to summary statistics (such as Mean and SD), which help us to assess the quality of the data collected.
The three indices stunting, wasting and underweight refl ect distinct biological processes and their use is necessary for determining appropriate interventions^{4}. It must be remembered here that these indices overlap – i.e. a child who is underweight may also have wasting and/or be stunted and other similar combinations. On the one hand, none of the three indices is able to provide a comprehensive estimate of the total number of undernourished children in a community and on the other hand – since they overlap – we cannot add these three indices to get the overall prevalence. For a comprehensive measure of the overall prevalence of undernutrition, what is needed is a single aggregate indicator that incorporates all undernourished children, be they wasted and/or stunted and/or underweight.
Such an aggregate indicator – the Composite Index of Anthropometric Failure (CIAF) – has been proposed by Svedberg^{7}. His original model suggests six subgroups of anthropometric failure (labeled AF); to which one more subgroup (labeled Y) has been added by Nandy et al^{8}. The anthropometric subgroups of the children are as follows: A – No Failure, B – Wasting only, C – Wasting + Underweight, D – Wasting + stunting + Underweight, E  Stunting + Underweight, F – Stunting only and finally, Y – Underweight only. The sum of the children in groups B to F provides the CIAF. As a single indicator, CIAF provides a single number to the overall estimate of undernourished children in a population, which none of the current indicators do. Furthermore, using the CIAF, anthropometric data can be disaggregated for further analysis, e.g. to analyze the specific risk factors & correlates or the mortality/morbidity patterns for the different types of anthropometric failure. Such disaggregation enables the identification of groups of children missed by conventional indices.
The current study uses the ZScore system and the CIAF to estimate the magnitude of undernutrition among underfi ve children in the slums coming under the field practice area of the PSG Urban Health Center, Coimbatore. The appropriateness of using low weightforage as the only indicator for detecting childhood undernutrition has been analyzed. The calculations are in reference to the WHO/ NCHS International population.
Material and Methods
Ten slums coming under the fieldpractice area of the Urban Health Centre, PSG Institute of Medical Sciences & Seetharaman N, et al: Malnutrition by Z Scores and CIAF Research, Coimbatore formed the study area. The Study population comprised of Children less than five years of age residing in the abovementioned slums. The total number of underfive children in these 10 slums was 625. Sample size for the crosssectional prevalence study was calculated using the formula Sample size (n) =4PQ/d2. With an expected prevalence of undernutrition (P) of 50% and a relative precision (d) of 10% of P, the required sample size was calculated as 400.
To arrive at the required sample size of 400, six out of the ten slums were randomly selected and all the underfive children in the six selected slums were included in the study. The actual number of children in these six slums was 405 and this was taken as the study population (n = 405). All the children up to 59 months of age living in the selected slums were included for the study. Children who were not resident of the slum, but visiting and children of families who had moved into the slum within the past 1 month were excluded from the study.
The exact age of the child was computed from the child’s date of birth. When data on the exact date of birth was not available, the age as told by the mother was used, corrected to the nearest month. A regional localevents calendar was used to assist the mothers for better recall. Nutritional assessment was done using anthropometry and clinical examination. Children were weighed and measured as per the WHO guidelines on Anthropometry^{4}. For children less than two years, the recumbent length was measured with the children lying down. Data collection was done over a period of two months.
Statistical analysis was done using EPIINFO 2002 software package, from CDC. The Zscores for the different nutritional indices – weightforage, heightforage and weightforheight were calculated in reference to NCHS International reference population by using the EPINUT component of the software. The prevalence of underweight (low weightforage), stunting (low heightforage) and wasting (low weight for height) were calculated at the cutoff level of < 2 SD (ZScore <2) and the prevalence of severe underweight, stunting and wasting at cutoff level of < 3 SD (ZScore <3) of the NCHS reference median values. Svedberg and later, Nandy et al have used these indicators to construct the CIAF. This index provides us with a single number to the overall prevalence of undernutrition in the community.
ChiSquare test was used to verify the statistical significance of the associations. P value of less than 0.05 was considered statistically significant.
Results
Majority (93.6%) of the study population were Hindus; 70.8% were living in nuclear families; 65.2% were practicing openair defecation; 74.8% were living in overcrowded dwellings and 86.4% had percapita incomes less than Rs.750 per month. Table 1 shows the agesex distribution of the study population. Female children constituted 54.3% of the study group. The maximum numbers of children (21.5% each) were seen in the 2435 and the 4859 months age group.
Table 1: AgeSex distribution of the study population (n=405)
Age Group (months)  Male (%)  Female (%)  Total (%) 

0 – 11  36 (19.5)  32 (14.6)  68 (16.8) 
12 – 23  43 (23.2)  37 (16.8)  80 (19.8) 
24 – 35  38 (20.5)  49 (22.3)  87 (21.5) 
36 – 47  39 (21.1)  44 (20.0)  83 (20.5) 
48 – 59  29 (15.7)  58 (26.4)  87 (21.5) 
Total  185 (100)  220 (100)  405 (100) 
Table 2 presents the distribution of undernutrition among the children studied. Female infants (011 months) had a significantly lower prevalence of undernutrition compared to male infants (p<0.001). The Overall gender difference was not statistically significant.
Table 2: Prevalence of undernutrition by sex and age group (n = 405)
Age (Months) 
Male (N) 
Under Nourished n (%) 
Female (N) 
Under Nourished n (%) 
Total (N) 
Under Nourished n (%) 

0 – 11  36  31 (86.1)  32  15 (46.9)  68  46 (67.6) 
12 – 23  43  38 (88.4)  37  21 (56.8)  80  59 (73.8) 
24 – 35  38  16 (42.1)  49  38 (77.6)  87  54 (62.1) 
36 – 47  39  19 (48.7)  44  39 (88.6)  83  58 (69.9) 
48 – 59  29  19 (65.5)  58  42 (72.4)  87  61 (70.1) 
Total  185  123 (66.5)  220  155 (70.5)  405  278 (68.6) 
Table 3 shows the prevalence of underweight, wasting and stunting among the children studied. Conventional growth monitoring activities detect only those children with underweight (46.7%). By Using Z scores, we can further identify children with wasting (20.2 ) and stunting (49.6).
Table 3: Agewise Prevalence of Indices of Undernutrition (n = 405)
Age (Months) 
No. of Children  % of Normal Children  Underweight % below  Wasting % Below  Stunting % Below  

< 2SD  < 3SD  <2SD  < 3SD  <2SD  < 3SD  
0 – 11  68  32.4  14.7  7.4  14.7  7.4  60.3  39.7 
12 – 23  80  26.3  42.5  25.0  17.5  6.3  56.3  12.5 
24 – 35  87  32.9  50.6  21.8  21.5  10.3  40.2  11.5 
36 – 47  83  30.1  56.6  28.9  30.1  6.0  39.8  22.9 
48 – 59  87  29.9  62.1  10.3  16.1  4.6  54.0  25.3 
Total  405  31.4  46.7  19.0  20.2  6.9  49.6  21.7 
“Below2 SD” denotes undernutrition and “below3 SD” denotes severe undernutrition.
CIAF permits us disaggregation of the undernourished children in to different subgroups, as shown in Table 4. Overall, only 127 (31.4%, group A) of the children studied were anthropometrically normal; 278 (68.6%) of the children were suffering from one or other form of “Anthropometric Failure”. That is, summing up the children in groups B to F provides the CIAF (68.6%). By using low weightfor age (underweight) as the sole criterion for undernutrition, we can identify children from groups C, D, E and Y (189 children in our study) but will be missing those in groups B and Fchildren who are stunted or wasted but not underweight.
In the current study, 89 such children (accounting to 21.9% of the total study population) would be missed out as ‘not undernourished’, if we use the popularly used low weightfor age as the only indicator of undernutrition.
Table 4: Classification of children with anthropometric failure (n=405)
Group  Anthropometric status^{a,b}  Number of Children (%) 

A  No Failure  127 (31.4) 
B  Wasting only  11 (2.7) 
C  Wasting + Underweight  48 (11.9) 
D  Wasting + Stunting + Underweight  23 (5.7) 
E  Stunting + Underweight  100 (24.7) 
F  Stunting only  78 (19.3) 
Y #  Underweight only  18 (4.4) 
Total  405 (100) 
  Subgroups A to F as proposed by Peter Svedberg5 and subgroup Y
as suggested by Nandy et al^{4}.
a – Another theoretical combination would be “wasted and stunted”, but this is not physically possible since a child cannot simultaneously experience stunting and wasting and not be underweight.
b – Adding up children in the subgroups B to F gives us the CIAF (68.6 %)
Table 5 attempts a comparison between the IAP system and the Z score system of grading undernutrition. As per IAP criteria, 208 children (51.4%) were undernourished and 13 (3.2%) were severely undernourished (Grade III & IV). Compared to this, as per the Z score system 189 children (46.6%) were undernourished and 77 (19%) were severely undernourished. Although the overall prevalence of undernutrition is higher (by 4.8%) as per the IAP criteria, the Z score system identifies a much higher prevalence of severe undernutrition compared to IAP system. In the present study, 64 out of the 77 children graded as “Severely undernourished” (by Z score system) seem to fall under Grade II (the “Moderately undernourished” category) of the IAP system.
Table 5: Comparison of the IAP and the ZScore Systems (n = 405)
IAP  Z Scores  

Normal  “Undernutrition” (< 2SD) 
“Severe Undernutrition” (< 3SD) 
Total (%) 

Normal  197  0  0  197 (48.6) 
Mild (Grade I)  19  93  0  112 (27.7) 
Moderate (Grade II)  0  19  64  83 (20.5) 
Severe (Grades III, IV)  0  0  13  13 (3.2) 
Total (%)  216 (53.3)  112 (27.7)  77 (19.0)  405 (100.0) 
Discussion
In India, there is still a paucity of communitybased studies on childhood undernutrition using the Z score system. Using the Z score classification, Studies from Haryana^{9} and Punjab^{10} report comparable prevalence levels. The relatively high prevalence of wasting observed among the children in the current study is indicative of a state of acute undernutrition, indicative of recent food deprivation and/or illness. NFHS 2 uses the Z score system of classification to grade undernutrition among Indian children. At the national level, the prevalence of underweight, stunting and wasting were 47%, 45.5 % and 15.5% respectively and the corresponding values for Tamilnadu were 36.7%, 29.4% and 19.9%.
Currently the IAP classification based on weightforage, is followed in the ‘anganwadi’ centers throughout the country to grade undernutrition at the grassroot level for the Government of India’s project on the Integrated Child Development Services (ICDS). As revealed by Table 5 the IAP system identifies 4.8% more children as undernourished, whereas the Z score system identifies significantly more children as severely undernourished. These “Severely undernourished” children are the ones who get additional nutritional supplementation under the ICDS. In our study, 64 out of the 77 children graded as “Severely undernourished” by Z score system fall under the Grade II “Moderately undernourished” category as per the IAP system. This has high practical significance, in light of the fact that priorities of nutritional supplementation through ICDS are inclined towards the “severely undernourished” children – Grades III & IV of the IAP system. A similar study^{11} in West Bengal comparing the IAP and the Z score systems, found comparable results  61% of the children were undernourished (3.9% severely) as per IAP criteria, whereas 46.6% were undernourished (6.9% severely) as per Z score system.
As seen from Table 4, underweight children form only one subgroup of the total number of undernourished children i.e. children who show evidence of “anthropometric failure”. Nandy et al have improved on the CIAF (originally proposed by Svedberg) which they have applied to the entire NFHS 2 dataset. In their study, “Children with no failure” (Group A) account for 40.2% while children with “Wasting and Stunting and Underweight” (Group D) account for 7.2% and children with “Stunting only” (Group F) account for 10.1%. In the current study, considerably fewer children – only 31.4% were normal or had ‘no anthropometric failure’ and 5.7% of children had “Wasting and Stunting and Underweight”. The prevalence of “Stunting only” is relatively high – 19.3% among the slum children studied.
As evidenced by the current study, the use of underweight (low weightforage) as the sole criterion for identifying undernourished children may be underestimating the true load of undernutrition. Use of the CIAF helps us to visualize the extent of underestimation. Nearly 22% of the present study population – 89 undernourished children – would be missed if low weightforage is considered as the only indicator of undernutrition. CIAF provides an overall estimate on the number of undernourished children in a population, which none of the conventional indices provide. Attempts at estimating the overall prevalence of undernutrition in the population must integrate such an aggregate index of undernutrition. This could be a tool of considerable interest to health planners and policy makers – especially considering the fact that to compute the CIAF, the only additional data that needs to be collected is the height of the child. Measurement of the child’s height as part of the routine ICDS growth monitoring is worth considering.
The limitations of this study include – the approximation of children’s weight to the nearest 500 grams, which might have had an infl uence on the prevalence estimates. The date of birth as told by the mother has been used; crosschecking with records could not be done for all of the children. There have been concerns about the appropriateness of using the NCHS data as the reference population for Indian children^{12}. To address this concern, WHO is in the process of developing a more appropriate reference population, which would be available soon^{13}.
Conclusions
Overall, only 31.4% of the underfive children studied were anthropometrically normal. In other words, more than two thirds of the children were undernourished. This is a very serious problem, by any scale. Under such conditions, our intervention efforts need to be broader than providing supplementary nutrition alone.
More widespread use of the ZScore system of classification, especially in communitybased studies, is recommended. This system enables us to estimate/express the prevalence of undernutrition using all the three indices – underweight, stunting and wasting. This also allows meaningful comparisons with the nationally representative NFHS 2 database. The process of calculating the Z scores has been made very simple by the use of Epiinfo software package developed and distributed freely by the CDC.
Current measures of undernutrition are, on their own, unable to give a reliable estimate of the overall number of undernourished children in a population. This issue has been addressed with the construct of the new indicator, CIAF. Findings from the current study suggest that conventional measures of undernutrition may be missing out a considerable proportion of undernourished children present in the population. The proportion of children identified as “severely undernourished” receive additional nutritional supplementation under the ICDS. Hence, underestimating this proportion might prevent undernourished children from receiving the benefit of the extra supplementation they deserve. The dissagregation of undernourished children in to different subgroups (as done in CIAF) allows the researcher to further examine the relationship between particular combinations of undernutrition and poverty or morbidity/ mortality data (when available). Studies have shown that children with double or triple failures are more likely to be from poorer families and also have a higher morbidity risk than children with single failures(^{8}. Identification of these children with multiple failures has obvious implications in antipoverty policies. A comprehensive measure of the total load of undernutrition – such as the Use of the Composite Index of Anthropometric Failure discussed in this paper – must be incorporated in our attempts at quantifying undernutrition.
Acknowledgement
The authors would like to thank Mr.Nanjappan for his help in data collection, Miss.Narmada for her help in data analysis and Dr.YSS.Sivan for his continual efforts in improving the paper.
References
 UNICEF. The state of the world’s children Report 2003. New York: UNICEF; 2003.
 Maurice H. King. Nutrition for Developing Countries. Reprint 1976.Nairobi: Oxford University Press; 1976.
 Sachdev HPS. Assessing Child Malnutrition – Some Basic Issues. Nutritional Foundation of India archives 156. New Delhi: NFI; 1996.
 World Health Organization. The use and interpretation of Anthropometry – Report of WHO Expert committee. WHO Tech Rep Series 854. WHO, Geneva. 1995.
 International Institute of Population Sciences and ORCMacro. National Family Health Survey India. (NFHS 2) 199899: India. Mumbai: I.I.P.S; 2000.
 Center for Disease Control, Atlanta. http://www.cdc.gov/ epiinfo/
 Svedberg P. Poverty and undernutrition: theory, measurement and policy. New Delhi: Oxford India Paperbacks; 2000.
 Shailen Nandy, Michelle Irving, David Gordon, Subramanian SV, George Davey Smith. Poverty, Child undernutrition and morbidity: new evidence from India. Bulletin of the World Health Organization 2005; 83(3): 210216.
 Kumar R, Agarwal A K, Iyengar S D. Nutritional Status of Children: Validity of MidUpper Arm Circumference for Screening Undernutrition. Indian Pediatrics 1996; 33: 189 196.
 Laxmaiah A, Rao M K, Brahmam GNV, Kumar S, Ravindrenath M, Kasinah K et al. Diet and Nutritional Status of Rural Preschool Children in Punjab. Indian Pediatrics 2002; 39: 331338.
 Ray S K, Halder A, Biswas B, Mishra R, Kumar S. Epidemiology of undernutrition. Indian Journal of Pediatrics 2001; 68: 1025 30.
 Evelth P, Tanner J. Worldwide variations in human growth. Cambridge: Cambridge University Press; 1990.
 de Onis M, Garza C, Habicht JP. Time for a new growth reference. Pediatrics 1997; 100:5.
Department of Community Medicine, PSG Institute of Medical
Sciences and Research Peelamedu, Coimbatore, Tamilnadu
Email: [email protected]
Received: 19.08.05