Despite the utmost importance of body height in evaluating nutritional status, it is not always possible to obtain its measurement and height may have to be estimated. The objective of the study was to formulate and cross-validate a regression equation to predict height using hand length measurement and also to determine if predicted height (PH) will lead to significant errors when used in body mass index (BMI) calculation.
A cross-sectional study was conducted using a consecutive sample of 465 inpatients (19–91 years), from a university hospital. Participants were randomly divided into a development sample of 311 individuals and a cross-validation one. A linear regression model was used to formulate the equation. Intraclass correlation coefficients (ICCs) for single measures and differences between measured height (MH) and PH and between BMI calculated with MH (BMIMH) and with PH (BMIPH) were determined.
The regression equation for PH is: PH (cm)=80.400+5.122 × hand length (cm)—0.195 × age (years)+6.383 × gender (gender: women 0, men 1) (R=0.87, s.e. of the estimate=4.98 cm). MH and PH were strongly correlated, ICCs: 0.67-0.74 (P<0.001). Differences were small, mean difference±s.d., ⩽−0.6±4.4 cm (P⩾0.24). BMIMH and BMIPH were strongly correlated, ICCs: 0.94-0.96 (P<0.001). Differences were small, ⩽0.3±1.7 kg/m2 (P⩾0.10).
The formulated regression equation using hand length, age and gender provides a valid estimation of height and is useful in the clinical context. PH from this regression equation can be used in BMI calculations as misclassification is small.
Despite the utmost importance of body height in evaluating nutritional status, it is not always possible to obtain its measurement. Although an individual’s height is reported in the identity card, the accuracy of this measurement is questionable and the magnitude of the error is not known. Therefore, for bedridden patients or individuals with any other conditions limiting their ability to stand, it may be necessary to resort to alternative anthropometrical indicators to estimate their height.1 Some of the indicators described include length,1 arm span,2, 3, 4, 5, 6, 7 half-span,2, 8, 9, 10, 11 knee height,1, 2, 11, 12, 13, 14 ulna length11, 15 or hand length.16, 17, 18 Length measurement provided for more accurate estimates of height.1 However, beds are not always completely horizontal and not all patients can be positioned in the supine position. Recumbent length is also more time consuming and requires more skilled technicians.
Cockram and Baumgartner12 equations for predicting height from knee height in older adults present a strong correlation with standing height, R2⩾0.83 and a s.e. of the estimate (SEE) equal to 3.56 cm for men and 2.74 cm for women. Despite the small mean difference of −0.6 cm between body height and height determined from knee height reported by Hickson and Frost2 in a sample of older adults, it was also reported that an application for various groups of Chumlea et al.19 regression equations to estimate body height from the knee height produce errors up to 8.8 cm. Moreover, measuring knee height is time consuming.2 While ulna length and body height are also significantly correlated (R=0.963), equations for predicting height from ulna length results presented small mean differences of 0.3%±s.d. 2.7% of height but higher SEE than Cockram and Baumgartner equations, varying from 4.3 cm to 4.8 cm.15
The relation between height and hand length was described for the first time by Marcus Vitruvius Polliono (BC) and popularized by Leonardo da Vinci’s ‘Vitruvian Man’ from 1487.20, 21, 22 Hand length is an attractive alternative for estimating body height, as it is easier to measure than other anthropometrical indicators used to estimate body height such as ulna length or knee height. The hand is more accessible, its measurement requires a minimum of motion and cooperation by the individual and can be obtained without the necessity of mobilizing the patient, causing practically no discomfort. As hand length is a measurement of smaller magnitude compared with other alternative anthropometrical indicators used to estimate body height, it implies the use of either a segmometer, a pocket ruler or an anthropometrical tape and therefore would be more practical and less time consuming for daily practice. Consequently, surrogate estimates of body height would be quicker and economical. Nutritional status of hospitalized patients is often not identified by the medical staff.23, 24 Easier and less time-consuming practices, as the development of accessible recumbent estimates of height, would possibly lead to better practices.
The association between height and hand length has already been studied in forensic and legal medicine,16, 17, 18 showing moderate-to-strong correlations (R>0.51), whereas SEE from regression equations varied between 4.1 and 5 cm. Although these equations were developed in individuals aged between 18 and 35 years from Mauritius and India, it is documented that ethnic differences have an influence on height. Therefore, one should be cautious when applying predicting equations developed in a specific ethnic group in other ethnic groups.25, 26 Moreover, with increasing age, a narrowing of the spinal discs and a decrease of the spine length occurs,27, 28 justifying the need to test whether hand length can be used to predict height of Caucasians adults and older adults.
The present study aimed to formulate and cross-validate a regression equation to predict height using hand length measurement. It also aimed to evaluate the effectiveness of height estimation based in this regression equation and to predict how it will lead to significant errors when used in body mass index (BMI) calculation.
Subjects and methods
Subjects and design
A cross-sectional study was conducted using a consecutive sample from a university hospital. Patients were admitted if they were ⩾18 years old, Caucasian, with an expected length of stay >24 h, conscious, cooperative and able to provide written informed consent. Patients with critical illness, defined as failure of at least one vital organ29 admitted to intensive care units were not included. We also excluded pregnant women, individuals in isolation, those who were admitted for procedures that could put them in a critical situation, those with hemodynamic instability at the time and if body mass and half-span were impossible to measure. We therefore included inpatients from angiology and vascular surgery, cardiology, digestive, non-digestive and hepato-biliary surgeries, endocrinology, gastroenterology, internal medicine, nephrology, orthopedics, otorhinolaryngology and urology wards. From the daily list of patients admitted to each service, those who met inclusion criteria were invited to participate in the study, until the number of patients had reached the total number of beds of the ward. To have a representative sample of the hospital, data collection took place until the number of patients corresponded to twice the total number of beds of all the selected wards (n=470).
As it was not possible to measure either dominant or nondominant hand length in five individuals, they were not included. The study sample is composed of 465 participants aged between 19 and 91 years, consisting of 98.9% of the eligible sample.
Participants were randomly divided into a development sample consisting of 311 individuals and a cross-validation sample of 154 individuals. The development sample was used to formulate the generalized regression equation to predict height from hand length and was cross-validated with the second sample.
The study was conducted according to the guidelines laid down in the Declaration of Helsinki and approved by the Institutional Review Board and the Ethics Committee of Centro Hospitalar do Porto. Written informed consent was obtained from all study participants.
Age and clinical history of each participant was obtained by consulting the patient clinical file.
Standing height, half-span, body mass, dominant and nondominant hand length were measured according to standardized procedures.7, 30 The measurement of hand length was taken as the shortest distance from the marked midstylion line to the most distal point of the third digit (dactylion),30 with a segmometer, instrument used to measure segment lengths and selected heights, with 0.1 cm resolution and 20.0 cm of range (Kennon Instruments, Vignola, Modena, Italy). One branch of the segmometer was placed on the marked midstylion line while the other branch was positioned on the dactylion.30 Dominant and nondominant hand length were highly correlated (R=0.93, P<0.001) and nondominant was chosen to formulate and cross-validate the generalized regression equation to predict height from hand length. When it was impossible measuring nondominant hand length, dominant hand was used (n=26). Standing height (cm) and half-span (cm) were measured with a metal tape Rosscraft Innovations Incorporated (Surrey, BC, Canada) with a 0.1 cm resolution and a headboard was also used for measuring standing height. Mass (kg) was measured with a calibrated portable beam scale with a 0.5 kg resolution.
When it was impossible to measure standing height, half-span was converted to height multiplied by two.7 Standing height was measured in 325 participants and half-span was measured in 140 participants. The intra-observer technical error of measurement was obtained in 16 patients and was 0.2% for weight, 0.2% for height, 0.3% for half-span and 0.6% for hand length and therefore acceptable for a trained anthropometrist.31
Means±s.d. and 95% confidence intervals (CIs) were calculated. Kolmogorov-Smirnov test was used to evaluate the normality of variables distribution.
A linear regression model was used to formulate the equation to predict height and the method Enter was used for results presentation. Hand length, age and gender were entered the model as independent variables and standing height or height determined from half-span, hereinafter referred as measured height (MH), was entered as the dependent variable.
BMI was calculated using MH (BMIMH), and using predicted height (PH, BMIPH). BMI classes according to the World Health Organization standards were created.32
Differences in the proportion of women and men between samples were compared using χ2 test. Differences in the other samples characteristics were compared using Student’s t-test for independent samples or Mann–Whitney test. Differences between MH and PH (MH−PH) and differences between BMIMH and BMIPH (BMIMH−BMIPH) were compared using Student’s t-test for paired samples. Error percentages between MH and PH and between BMIMH and BMIPH were also calculated.
Intraclass correlation coefficients (ICCs) for single measures between MH and PH and between BMIMH and BMIPH were determined. In addition, the visual agreement between MH and PH and between BMIMH and BMIPH was evaluated by Bland–Altman plots.33 Limits of agreement were calculated as mean of the difference minus two s.d. and mean of the difference plus two s.d. The level of agreement between BMI classes when BMI was calculated using MH and using PH was assessed by agreement percentage and kappa with quadratic weighting.
As MH was obtained either by standing height or by height determined from half-span, differences between each one of these approaches and PH were compared using Student’s t-test for independent samples. Results were considered significant when P<0.05. All statistical analyses were carried out using the Software Package for Social Sciences (SPSS) for Windows (version 21.0, 2012, SPPS Inc., Chicago, IL, USA) and Microsoft Excel (version 2010, Microsoft, Redmond, WA, USA).
Study sample characteristics are presented in Table 1. No significant differences were found between development and cross-validation samples in the proportion of women and men (P=0.86) and also regarding age, hand length, body mass, MH, PH, BMIMH or BMIPH.
The proposed regression equation formulated for predicting height (cm) is:
Height (cm)=80.400+5.122 × hand length (cm)–0.195 × age (years)+6.383 × gender
gender: women, 0; men, 1
R=0.87, SEE=4.98 cm
Hand length, age and gender were all significant predictors (P<0.001).
Standing height ranged from 143 cm to 191 cm and height determined from half-span ranged from 140.4 cm to 190.2 cm. PH ranged from 139.6 cm to 184.2 cm.
For women, measured and predicted heights were strongly correlated with ICCs ranging from 0.71 to 0.74 (P<0.001), respectively, for the development and the cross-validation samples. Mean differences and 95% CIs between MH and PH were −0.01±4.9 (−0.8; 0.8) cm (P=0.97) and −0.6±4.4 (−1.7; 0.4) cm (P=0.24), respectively, for the development and the cross-validation samples. Mean error and 95% CI between MH and PH was −0.1±3.1 (−0.6; 0.4) % for the development sample and −0.5±2.8 (−1.2; 0.2) % for the cross-validation sample.
For men, measured and PHs were also strongly correlated with ICCs ranging from 0.73 to 0.67 (P<0.001), respectively, for the development and the cross-validation samples. Mean differences and 95% CIs between MH and PH were −0.02±5.0 (−0.8; 0.7) cm (P=0.97) and −0.2±4.9 (−1.2; 0.8) cm (P=0.70), respectively, for the development and the cross-validation samples. Mean error and 95% CI between MH and PH was −0.1±3.0 (−0.6; 0.3) % for the development sample and −0.2±2.9 (−0.8; 0.4) % for the cross-validation sample. The Bland–Altman plot from MH against PH for the entire sample is presented in Figure 1, revealing a high agreement and similarly a quite small mean difference of −0.1±4.9 cm between them.
Regarding BMI results, for women, BMI calculated with both MH and PH were strongly correlated with ICCs=0.96 (P<0.001) for both development and cross-validation samples. Mean differences and 95% CIs between BMIMH and BMIPH for development and cross-validation samples were quite small, respectively, 0.03±1.8 (−0.3; 0.3) kg/m2 (P=0.85) and 0.3±1.7 (−0.1; 0.8) kg/m2 (P=0.10). Error percentages between BMIMH and BMIPH were again small for the development sample, −0.1±6.2 (−1.2; 0.9) % and for the cross-validation sample, 0.7±5.6 (−0.6; 2.1) %. For men, BMIMH and BMIPH were strongly correlated with ICCs ranging from 0.94 and 0.96 (P<0.001), respectively, for development and cross-validation samples. Mean differences and 95% CIs between BMIMH and BMIPH for both development and cross-validation sample were again small, respectively, −0.03±1.6 (−0.3; 0.2) kg/m2 (P=0.78) and 0.03±1.5 (−0.3; 0.3) kg/m2 (P=0.85). Error percentages between BMIMH and BMIPH were also small for the development sample, −0.1±5.9 (−0.9; 0.8) % and for the cross-validation sample, 0.1±5.8 (−1.1; 1.4) %. The Bland–Altman plot from BMIMH against BMIPH for the entire sample according to gender is presented in Figure 2, and a small mean difference of 0.1±1.8 kg/m2 between them for women and of 0±1.5 kg/m2 for men is displayed. Agreement between BMI classes when BMI was calculated using both MH and using PH was high for the development sample, 79.7% and kappa value was good (0.89). Results are similar for the cross-validation sample, agreement was 76% with a kappa=0.88.
As MH was obtained either by standing height (in 325 participants) and by height determined from half-span (in 140 participants), it was explored if differences between each one of these approaches and PH were comparable. Mean difference between standing height and PH (mean±s.d.: −0.01±4.5 cm) was not significantly different from the mean difference between height determined from half-span and PH (mean±s.d.: −0.4±5.5 cm), P=0.41.
This study explored if a generalized regression equation to predict height from hand length can be used as an alternative method for estimating height and if this approximation will lead to significant errors when used in the BMI calculation. Several authors have developed regression equations to estimate body height from hand length.16, 17, 18 However, these equations were developed in samples of young individuals from Mauritius and India. Ethnic and age differences could limit its application to Caucasian adults and older adults, therefore justifying the present research.
Our generalized regression equation provides a valid alternative measurement of height as MH and PH were strongly correlated (ICC⩾0.67, P<0.001) and mean differences between MH and PH were quite small, ⩽−0.6±4.4 cm.
Regarding the BMI calculus, BMIMH and BMIPH were strongly correlated (ICC⩾0.94, P<0.001) and mean differences between BMIMH and BMIPH were also quite small ⩽0.3±1.7 kg/m2. Agreement percentages between BMI classes were high (⩾76%) and kappa coefficients were good (⩾0.88). The Bland–Altman plot revealed a considerable 2 s.d. ranging from −3.4 to 3.6 kg/m2, which is relevant given that BMI is used not only for categorization but also for continuous and longitudinal measurements. However, despite being high, the discrepancy between BMI calculated from measured and predicted height for the development sample is lower than what has been described for ulna length measurements.15
In the present analysis, half-span was measured and was multiplied by two because measuring span requires that the subject’s arms are entirely straight and ideally two technicians are needed to hold the elbows and pass the tape parallel to the clavicles.5 This measurement is difficult to perform, due to the human resources requirement and the need for collaboration and mobility by the individual who is being measured. Moreover, half-span has shown to provide valid and reliable estimates of body height in young, middle-aged and older adults.3, 4, 7 Although this differential method for obtaining height could be regarded as a study limitation, the mean difference between standing height and PH was not significantly different from the mean difference between height determined from half-span and PH. Knee height and ulna length were not evaluated in this research, which is recognized as a study drawback as, despite not being a study purpose, it would be interesting to compare different surrogate estimates of height within the same patients.
Measuring ulna length requires the identification of the olecranon and the styloid process; for hand length, identifying the end point of the middle finger is immediately obvious, the only subjective point is identifying the midstylion, therefore measuring hand length possibly will diminish measurement error by 50%. It was possible to measure hand length in 98.9% of the eligible sample. It is a fast and simple measurement that can be performed with a segmometer, an anthropometrical tape or a simple pocket ruler. The hand is accessible, requires a minimum of motion and cooperation from the individual, causing practically no discomfort. The small mean differences between MH and PH and the small mean differences and high agreement between BMIMH and BMIPH described for the cross-validation sample strengths this study, empowering the external application of the formulated equation. The wide age range in present sample (19–91 years old) also empowers these study results. Moreover, participants included in this study come from a variety of hospital wards, ensuring a wide spectrum of patients and relevant pathologies, which strengthens the external validity of the study results.
In the clinical setting, height is a highly relevant anthropometrical indicator to predict body composition from bedside methods. However, within the context of modern body composition analyses, height measurement could be avoided.
Although hand length had already been described as a predictor of height,16, 17, 18 as far as we are concerned, this is the first study to formulate and cross-validate a regression equation to predict height from hand length in a Caucasian population of adults and older adults. The formulated regression equation using hand length, age and gender provides a valid alternative measurement of height as mean differences between MH and PH were quite small. PH from this regression equation can be used in BMI calculations as misclassification is small. This reliable alternative approach for estimating height for adults and older adults will hopefully lead to better practices improving the identification of the patient’s nutritional status.
Lee RD, Nieman DC . Assessment of the hospitalized patient. In: Lee RD, Nieman DC (eds) Nutritional Assessment 4th edn Mosby: Chicago, IL, 2007, pp 226–231.
Hickson M, Frost G . A comparison of three methods for estimating height in the acutely ill elderly population. J Hum Nutr Diet 2003; 16: 13–20.
Brown JK, Whittemore KT, Knapp TR . Is arm span an accurate measure of height in young and middle-age adults? Clin Nurs Res 2000; 9: 84–94.
Brown JK, Feng JY, Knapp TR . Is self-reported height or arm span a more accurate alternative measure of height? Clin Nurs Res 2002; 11: 417–432.
de Lucia E, Lemma F, Tesfaye F, Demisse T, Ismail S . The use of armspan measurement to assess the nutritional status of adults in four Ethiopian ethnic groups. Eur J Clin Nutr 2002; 56: 91–95.
Mohanty SP, Babu SS, Nair NS . The use of arm span as a predictor of height: A study of South Indian women. J Orthop Surg (Hong Kong) 2001; 9: 19–23.
Kwok T, Whitelaw MN . The use of armspan in nutritional assessment of the elderly. J Am Geriatr Soc 1991; 39: 492–496.
Bassey EJ . Demi-span as a measure of skeletal size. Ann Hum Biol 1986; 13: 499–502.
Hirani V, Mindell J . A comparison of measured height and demi-span equivalent height in the assessment of body mass index among people aged 65 years and over in England. Age Ageing 2008; 37: 311–317.
Hirani V, Tabassum F, Aresu M, Mindell J . Development of new demi-span equations from a nationally representative sample of adults to estimate maximal adult height. J Nutr 2010; 140: 1475–1480.
Malnutrition Advisory Group. (eds). The ‘MUST’ Explanatory Booklet. BAPEN: Redditch, Worcestershire UK, 2003.
Cockram DB, Baumgartner RN . Evaluation of accuracy and reliability of calipers for measuring recumbent knee height in elderly people. Am J Clin Nutr 1990; 52: 397–400.
Han TS, Lean ME . Lower leg length as an index of stature in adults. Int J Obes Relat Metab Disord 1996; 20: 21–27.
Muncie HL Jr., Sobal J, Hoopes JM, Tenney JH, Warren JW . A practical method of estimating stature of bedridden female nursing home patients. J Am Geriatr Soc 1987; 35: 285–289.
Barbosa VM, Stratton RJ, Lafuente E, Elia M . Ulna length to predict height in English and Portuguese patient populations. Eur J Clin Nutr 2012; 66: 209–215.
Agnihotri AK, Agnihotri S, Jeebun N, Googoolye K . Prediction of stature using hand dimensions. J Forensic Leg Med 2008; 15: 479–482.
Ahemad N, Purkait R . Estimation of stature from hand impression: a nonconventional approach. J Forensic Sci 2011; 56: 706–709.
Rastogi P, Nagesh KR, Yoganarasimha K . Estimation of stature from hand dimensions of north and south Indians. Leg Med (Tokyo) 2008; 10: 185–189.
Chumlea WC, Guo SS, Steinbaugh ML . Prediction of stature from knee height for black and white adults and children with application to mobility-impaired or handicapped persons. J Am Diet Assoc 1994; 94: 1385–1388., 1391, quiz 1389-1390.
Sofat S . My Vitruvian Man. Acad Med 2009; 84: 1075.
Naini FB, Cobourne MT, McDonald F, Donaldson AN . The influence of craniofacial to standing height proportion on perceived attractiveness. Int J Oral Maxillofac Surg 2008; 37: 877–885.
Le Floch-Prigent P . [The Vitruvian Man: an anatomical drawing for proportions by Leonardo Da Vinci]. Morphologie 2008; 92: 204–209.
Singh H, Watt K, Veitch R, Cantor M, Duerksen DR . Malnutrition is prevalent in hospitalized medical patients: are housestaff identifying the malnourished patient? Nutrition 2006; 22: 350–354.
Farrer K, Donaldson E, Blackett B, Lloyd H, Forde C, Melia D et al. Nutritional screening of elderly patients: a health improvement approach to practice. J Hum Nutr Diet 2013, e-pub ahead of print 30 April 2013 doi:10.1111/jhn.12073.
Chumlea WC, Guo S . Equations for predicting stature in white and black elderly individuals. J Gerontol 1992; 47: M197–M203.
Madden AM, Tsikoura T, Stott DJ . The estimation of body height from ulna length in healthy adults from different ethnic groups. J Hum Nutr Diet 2012; 25: 121–128.
Cline MG, Meredith KE, Boyer JT, Burrows B . Decline of height with age in adults in a general population sample: estimating maximum height and distinguishing birth cohort effects from actual loss of stature with aging. Hum Biol 1989; 61: 415–425.
Milne JS, Williamson J . A longitudinal study of kyphosis in older people. Age Ageing 1983; 12: 225–233.
ASPEN Board of Directors and the Clinical Guidelines Task Force. Guidelines for the use of parenteral and enteral nutrition in adult and pediatric patients. J Parenter Enteral Nutr 2002; 26: 1SA–138SA.
Marfell-Jones M, Olds T, Stewart A, Carter L (eds) International Standards for Anthropometric Assessment. International Standards for Anthropometric Assessment: Potchefstroom, South Africa, 2006.
Pederson D, Gore C . Anthropometry Measurement Error. University of New South Wales Press: Sydney, NSW, 1996.
World Health Organization. Obesity: Preventing and Managing the Global Epidemic. Report of a WHO Consultation. World Health Organization: Geneva, 2000.
Bland JM, Altman DG . Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307–310.
We thank the Centro Hospitalar do Porto and all ward directors for facilitating the data collection. Rita S Guerra as a Ph.D. student is receiving a scholarship from FCT – Fundação para a Ciência e a Tecnologia under the project (SFRH/BD/61656/2009).
The authors declare no conflict of interest.
About this article
Cite this article
Guerra, R., Fonseca, I., Pichel, F. et al. Hand length as an alternative measurement of height. Eur J Clin Nutr 68, 229–233 (2014). https://doi.org/10.1038/ejcn.2013.220
- body height
- hand length
- nutritional assessment
- body mass index
Journal of Human Nutrition and Dietetics (2021)
Nutrition Research (2021)
Assessment of Stature from Hand Span - A Cross Sectional Descriptive Investigation among Undergraduate Medical Students
Journal of Evolution of Medical and Dental Sciences (2021)
Association Between Anthropometric Indicators of Nutrition Status and Length of Hospital Stay in Hospitalized Patients
Journal of Parenteral and Enteral Nutrition (2021)