Objectives: To investigate appropriate measurements to predict height in children with physical impairments to facilitate the accurate assessment of nutritional status in field studies.
Design: Case–control cross-sectional study.
Setting: Dharavi, a large slum in Mumbai, India.
Subjects: In total, 141 children with mixed disabilities and 162 nondisabled control children, aged 2–6 y.
Methods: Height/length, armspan, arm length and tibia length were measured to the nearest 0.1 cm using standard procedures. The relations between armspan, arm length and tibia length with height in controls were investigated using linear regression.
Results: Armspan (R2=0.93, P<0.001, n=158), arm length (R2=0.81, P<0.001, n=162) and tibia length (R2=0.72, P<0.001, n=161) were found to be strong predictors of height based on data from nondisabled control children. These measurements could be used to determine a more accurate height for children with physical impairments where the nature of the impairment may interfere with height measurements.
Conclusions: Armspan, arm length and tibia length can be used to determine accurate height for children with physical impairments, for example, children with a nonambulatory status or those with kyphosis or scoliosis of the spine.
Sponsorship: Department for International Development (DFID), UK.
Anthropometric measurements provide simple, noninvasive methods to assess the nutritional status of children (Gibson, 1990). However, standard anthropometric techniques are not always appropriate for children with disabilities (Stallings & Zemel, 1990). Many research groups have acknowledged difficulties in obtaining accurate assessments of height in children of nonambulatory status, or in those with kyphosis or scoliosis in the spinal column (Pai et al, 2001; Dahl et al, 1996; Dahl & Gebre-Medhin, 1993; Thommessen et al, 1991). Some research groups investigating nutritional status chose not to record height, while others have results that may underestimate height. Thommessen et al (1991) advised caution when interpreting height data from nutrition studies on children with disabilities. Dahl et al (1996) found that seven out of 35 (20%) heights in children with disabilities had to be discarded because of lack of reliability.
Problems can be exacerbated in field studies compared with clinical studies because specialised equipment to facilitate height measurement available in clinics may not be practical in a field study. Knee height measured with callipers was used to measure recumbent length for adults with cerebral palsy and mobility impairments in a clinical setting (R2=0.78) (Hogan, 1999). Body length mats are available for infants that measure length to the nearest 0.5 cm, but are not suitable for older children of nonambulatory status.
The current UN (1986) guidelines on how to weigh and measure children for assessing nutritional status in household surveys are inadequate in offering guidance for assessing children with disabilities. The guidelines advise not to weigh or measure physically ‘deformed’ children because the measurements will interfere with the accuracy of the survey.
Do not weigh or measure a child if the child is physically deformed which will interfere with or give an incorrect measurement. To be kind you may want to measure a child and make a note of the deformity on the questionnaire. (UN, 1986)
These guidelines require revisions to ensure that children with disabilities are not excluded from nutrition surveys. Suitable methods of assessing nutritional status are necessary for both research purposes and household nutrition surveys so that cases of malnutrition among children with physical disabilities are not missed.
Measurements that require only a tape measure are feasible for field studies. Demi-span (sternal notch to finger roots with arms outstretched) was found to have a significant correlation with height in a study using healthy European adults (R2=0.55 for males, R2=0.56 for females) (Bassey, 1986). De Lucia et al (2002) found that armspan (distance between the longest tip of one arm to the other) among Ethiopian adults was a suitable proxy for height to estimate body mass index, although sex and ethnicity cutoffs needed to be applied. Detailed arm length measurements (humerus+forearm+hand length) were found to be suitable measurements as an alternative to height in an elderly population (R2=0.76 for males, R2=0.94 for females) (Mitchell & Lipschitz, 1982). In a large field study assessing the nutritional vulnerability of older people in the developing world, measurements of armspan and half-span (distance between the tip of the middle finger of one arm to the midsternal notch) were used (HelpAge International and LSHTM, 1997). The study concluded that height estimated from armspan could be used for elderly populations. However, it was also observed that there is a need for height and armspan data on young adults in different populations in order to be able to estimate true height from armspan for the elderly (HelpAge International and LSHTM, 1997). Changing skeletal mass and increased physical impairments can complicate nutritional assessments of the elderly population.
Only two field studies were found that specifically investigated alternative height measurements for children with disabilities. Tompsett et al (1999) initially investigated demi-span, but fieldworkers found this a difficult measurement to take accurately compared with half-span. Mean weight/height and height/age data were reassessed with height derived from half-span. No significant differences were identified between the original data and derived data; for example, weight/height data for children with poliomyelitis changed from –0.4 to 0.6Z scores. In the Philippines, Socrates et al (2000) used armspan (half-span × 2) to derive height for children with cerebral palsy. However, the authors observed armspan had limited use where spasticity was present in both arms. The results of these studies are important for nutritional assessments of children with disabilities everywhere. The usefulness of these measurements requires further confirmation. Lower leg length as a measurement to derive height must also be field-tested.
The purpose of the study was to investigate appropriate measurements to predict height in children with physical impairments to facilitate the accurate assessment of nutritional status in field studies.
A case–control cross-sectional survey to investigate the nutritional status of children with disabilities in Dharavi, a large slum in Mumbai, India was conducted in 1999–2001 (Yousafzai, 2001). In total, 141 children with disabilities, aged 2–6 y, were identified using the ‘10 Questions Screen’ (Zamen et al, 1990) through convenience sampling. The 10 Questions Screen is an epidemiological tool developed and validated in South Asia, which is designed for the identification of the five major areas of disability (movement disorders, seizures, cognitive impairments, visual impairments and hearing impairments). A total of 162 neighbourhood children without any disabilities, aged 2–6 y, were recruited as controls. The age of the child could be given within 6 months by most carers. For carers who were not able to remember the child's exact age, estimates were made with the help of local events calendars and discussions with family and neighbours.
All anthropometric measurements were taken by the first author (AKY) to reduce intermeasurer variability. AKY had received anthropometry training at the Growth Clinic of Great Ormond Street Hospital, UK, which includes children with physical and cognitive impairments. The anthropometric measurements were carried out using standard procedures (UN Guidelines, 1986). A carer for the child was present throughout the procedure. Wherever possible a flat well-lit area was used as the measuring site. However, not all homes had sufficient space and the lanes outside were not all flat and, thus, any discrepancies that may have affected measurements were noted.
Length was measured to the nearest 0.5 cm using a measurer mat (Starters, Norwich, UK) for children unable to stand who were less than 92 cm tall. Height was measured to the nearest 0.1 cm using a stadiometer (Leicester Portable Measure, UK). Further measurements were taken to the nearest 0.1 cm in order to predict height using a flexible tape measure (TALC, UK). Armspan measurements were taken from the tip of the middle finger of one arm to the tip of the middle finger of the other arm with the arms outstretched at right angles to the body. Arm length measurements were taken from the tip of the humerus bone to the tip of the middle finger of the left arm. If it was not possible to measure the left arm because of a physical impairment (eg left hemiplegia), the right arm was measured and a note made in the data entry book. Tibia length was measured from the knee joint to the ankle joint of the left leg. Again if it was not possible to measure the left leg because of a physical impairment (eg the leg was affected by poliomyelitis), the right leg was measured and a note made in the data entry book. Measurements were taken in the subject standing whenever possible.
Data management and analysis
The anthropometric data were recorded in a data entry book by AKY in the field and subsequently double entered into Epi-info (version 6.04c). The analysis was carried out using Epi-info (version 6.04c) and the Statistical Package for Social Sciences (SPSS, version 8.0). The relations between armspan, arm length and tibia length with height were investigated in the nondisabled neighbour controls using linear regression. The different regression equations were compared using the standard error of the estimate (SEE), which is the error attached to a prediction made from the equation—the smaller the SEE, the better the prediction. Heights derived from armspan, arm length and tibia length were compared to actual measured heights in the cases using paired t-tests.
The study was approved by the Ethics Committee of Great Ormond Street Hospital in the UK and permission was granted from the local urban health centre in Dharavi, Mumbai for fieldwork to proceed. The study was explained to families participating in the study and informed verbal consent was obtained. Care was taken not to distress children during anthropometry. Confidentiality was adhered to in all stages of data management.
Table 1 provides a summary of the subjects and measurements taken. Difficulties were encountered in measuring height in 21% of children with disabilities (eg including children with amputations, cerebral palsy, postpolio syndrome and clubfoot). These children were unable to stand correctly for height measurements to be taken accurately; for example, as a result of scoliosis, kyphosis or having a nonambulatory status.
Significant correlations (P<0.001) were observed between height and measurements of armspan, arm length and tibia length. Linear regression was used to predict height from armspan, arm length and tibia length (all measured in cm) using the nondisabled neighbour control group. The regression equation using all three variables is shown below:
The SEE of 2.8 cm indicates that height predicted from the equation will be within 2.8 cm of the true value 68% of the time, and within two SEEs (5.6 cm) 95% of the time.
Regression equations using each variable separately are shown below:
2. Arm length:
3. Tibia length:
The regression coefficient for age with height as the dependent factor was R2=0.53 (P<0.001). Age, as an additional covariate, significantly improved the strength of the association with height and the SEE only with tibia length. When sex was included as an additional covariate, it was not significant in any model.
Armspan had the strongest association (R2=0.93) and the smallest SEE (3.2 cm). The authors acknowledge difficulties in obtaining accurate armspan measurements for children with spasticity or contractures in both arms. Missing data were common for armspan (n=30), which was the most difficult of the three measures investigated to obtain accurately. This factor must also be considered when interpreting the results. Tibia length measurements had the weakest association (R2=0.72). The SEE (6.6 cm) was also greater than the other models investigated, indicating a larger error when using this prediction equation.
The actual heights for the case group were then compared with predicted heights by paired t-tests (Table 2). The actual heights were less than those predicted from the equations. Therefore, it is useful to have methods to predict height, which take into account the height loss observed when children are unable to stand straight. The difference was significantly lower (P<0.001) when either arm length or tibia length variables were used, but not significantly lower when either armspan or all three variables were used.
Armspan, arm length and tibia length were all highly correlated with height. The regression equation using all three variables had the strongest degree of association (R2=0.95) and the best fit (SEE=2.8 cm). Therefore, this model can be recommended as a means of obtaining height in situations where direct height measurement is not possible and the other three variables are available.
However, it is not always practical to measure all three variables. Therefore, it is useful to have separate models available for each variable. Armspan had the strongest association and the smallest SEE. The authors acknowledge difficulties in obtaining accurate armspan measurements for children with spasticity or contractures in both arms, which was the most difficult of the three measures investigated to obtain accurately. This factor must also be considered when interpreting the results.
Stevenson et al (1995) found that tibia length could be reliably measured with a standard tape measure and was a useful proxy for height. However, Socrates et al (2000) commented that further work was required to assess whether tibia length rather than armspan is the preferential proxy for height in community surveys. In this study, tibia length measurements had the weakest association. The SEE was also greater than the other models investigated, indicating a larger error when using this prediction equation.
The mean actual height for the case group was found to be lower than mean heights predicted from each regression equation. However, it may be useful to have methods to predict height, which take into account the height loss observed when children are unable to stand straight.
Table 3 summarises results from the present investigation and previous investigations that use nondisabled control children to produce regression models in order to predict height for children with physical impairments. All three studies (Tompsett et al, 1999; Socrates et al, 2000; Yousafzai et al, present study) looked at a wide age range, which may have an impact on the results of the regression analysis. Tompsett et al (1999) found half-span to be the best measurement. Socrates et al (2000) found armspan (actually half-span multiplied by 2) had a strong linear association with height (R2=0.98). The present study also found armspan to be the best variable to use, if measured accurately. The mean height predicted from armspan for children with disabilities was higher than the actual mean height, but not significantly different when compared using paired t-tests. A means of overcoming difficulties encountered when obtaining armspan is to measure half-span and multiply the result by 2.
The limitations in measuring height can be overcome using suitable alternative long-bone measures when investigating the nutritional status of children with physical impairments, and the regression equations will be specific for the population being investigated with regard to age group and ethnicity. The final selection of measurements may be based on both the study population (ie nature of physical impairment encountered), social acceptability of a particular measure and the confidence of fieldworkers (eg demi-span was found to be difficult for fieldworkers in a study conducted by Tompsett et al (1999)). Armspan measurements have the strongest association with height and the smallest SEE. Any difficulties in taking armspan measurements may be overcome by measuring half-span instead.
An alternative to the prediction of actual height could be the development of accurate length mats similar to the Starters Mat (UK) suitable for field studies, which could measure older children with a nonambulatory status. The measurement of height is only one unresolved issue in assessing the nutritional status of children with disabilities (Stallings & Zemel, 1990). Work is also required in assessing and interpreting data on body fat and muscle (particularly in children with cerebral palsy) (van den Berg-Emons et al, 1998) and nutritional status from mid-arm circumferences (particularly in children with orthopaedic or motor difficulties who may have increased upper body strength to facilitate movement) (Tompsett et al, 1999). It is hoped that guidelines on anthropometry for children will include revisions to include this vulnerable group.
Bassey JE (1986): Demi-span as a measure of skeletal size. Ann. Hum. Biol. 13, 499–502.
Dahl M & Gebre-Medhin M (1993): Feeding and nutritional problems in children with cerebral palsy and myelomeningocoele. Acta Paediatr. 82, 816–820.
Dahl M, Thommessen M, Rasmussen M & Selberg T (1996): Feeding and nutritional characteristics in children with moderate or severe cerebral palsy. Acta Paediatr. 85, 697–701.
De Lucia E, Lemma F, Tesfaye F, Demisse T & Ismail S (2002): The use of armspan measurement to assess the nutritional status of adults in four Ethiopian ethnic groups. Eur. J. Clin. Nutr. 56, 91–95.
Gibson RS (1990): Principles of Nutritional Assessment. New York, USA: Oxford University Press.
HelpAge International and London School of Hygiene and Tropical Medicine (LSHTM) (1997): Assessing the Nutritional Vulnerabilities of Older People in Developing Countries, eds K Gregory & K Peachy. London, UK: London School of Hygiene and Tropical Medicine.
Hogan SE (1999): Knee height as a predictor of recumbent length for individuals with mobility-impaired cerebral palsy. J. Am. Coll. Nutr. 18, 201–205.
Mitchell CO & Lipschitz DA (1982): Arm length measurements as an alternative to height in nutritional assessment of the elderly. J. Parent Enter. Nutr. 6, 226–229.
Pai M, Alur M, Wirz S, Filteau S, Pagedar S & Yousafzai AK (2001): A pilot study of the nutritional status of disabled children living in a Mumbai slum. J. Indian. Pediatr. 38, 60–65.
Socrates C, Grantham-McGregor SM, Harknett SG & Seal AJ (2000): Poor nutrition is a serious problem in children with cerebral palsy in Palwan, the Philippines. Int. J. Rehab. Res. 23, 177–184.
Stallings VA & Zemel BS (1990): Nutritional Assessment of the Disabled Child in Feeding the Disabled Child, eds Sullivan & Rosenbloom. UK: Cambridge Press.
Stevenson RD, Robets C & Vogtle L (1995): The effects of non-nutritional factors on growth in cerebral palsy. Dev. Med. Child Neurol. 37, 124–130.
Thommessen M, Heiberg A, Kase BF et al (1991): Feeding problems, height and weight in different groups of disabled children. Acta Paediatr. Scand. 80, 527–533.
Tompsett J, Yousafzai AK & Filteau SM (1999): The nutritional status of disabled children in Nigeria—a cross sectional survey. Eur. J. Clin. Nutr. 53, 915–919.
United Nations (UN) (1986): How to weigh and measure children—assessing the nutritional status of young children in household surveys. UN Department of Technical Co-operation for Development and Statistics Office, New York.
Van den Berg-Emons, van Baak MA & Westerterp K (1998): Are skinfold measurements suitable to compare body fat between children with spastic cerebral palsy and health controls? Dev. Med. Child Neurol. 40, 335–339.
Yousafzai AK (2001): The Nutritional Status of Disabled Children Living in Dharavi, an Indian Urban Slum in Mumbai. PhD Thesis, University of London.
Zamen SS, Khan NA, Islam S, Banu S, Dixit S, Shrout P Durkin M (1990): Validity of the ‘ten questions’ for screening serious childhood disability—results from urban Bangladesh. Int. J. Epidemiol. 19, 613–620.
We are very grateful to all the families who willingly participated in the study. We would like to thank the field assistants who supported the data collection work. We also appreciate the collaborative support for the study given by the Spastics Society of India (SSI), Mumbai. Finally, many thanks to Jane Pringle of Great Ormond Street Hospital in the UK for giving time for the anthropometry training. We are grateful to the Department for International Development (DFID), UK for funding the study.
About this article
Cite this article
Yousafzai, A., Filteau, S., Wirz, S. et al. Comparison of armspan, arm length and tibia length as predictors of actual height of disabled and nondisabled children in Dharavi, Mumbai, India. Eur J Clin Nutr 57, 1230–1234 (2003). https://doi.org/10.1038/sj.ejcn.1601705
- nutrition status
- tibia length
- arm length
Irish Journal of Medical Science (1971 -) (2015)