Emergency Medicine (2002) 14, 160–165

Blackwell Science, LtdPaediatric Emergency MedicineOriginal Research

Are methods used to estimate weight in children accurate?Karen Black,1 Peter Barnett,1 Rory Wolfe2 and Simon Young1

1Departments of Emergency Medicine, Royal Children’s Hospital, Parkville, and 2Epidemiology and Preventive Medicine, Central and Eastern Clinical School, Monash University and the Alfred Hospital, Prahran, Victoria, Australia

Abstract

Objective: The exact weight of a child undergoing resuscitation is usually not known. Severalmethods to estimate a child’s weight have been proposed. We evaluated six of thesemethods to determine their accuracy and clinical usefulness across a range of weightsand ages.

Method: Children attending a tertiary paediatric emergency department on 30 non-consecutivedays were weighed, a length or height was obtained and an estimation of bodyhabitus (slim, average or heavy) made by a single investigator, Karen Black (KB). Allchildren less than 145 cm were also measured using the Broselow tape. Six methods ofweight estimation were employed and the calculated weight compared to the true weight.The weight estimation methods evaluated were the advanced paediatric life support(APLS) method, Broselow tape, devised weight estimation method (DWEM), Oakleytable, Traub–Johnson and the Traub–Kichen methods.

Results: Four hundred and ninety-five children were included in the study. Children were evenlydistributed among the weight groups of less than 10 kg, 10–25 kg, 25–40 kg and over40 kg. The methods with the best overall performance were the devised weightestimation method and Broselow tape although the latter has a length limitation of145 cm. The other methods performed well in the middle two weight groups but poorlyoutside these groups.

Conclusions: The most accurate methods of weight estimation in children are the Broselow tape andthe devised weight estimation method . We recommend the use of either of these methodsin emergency situations where direct weighing is not possible.

Key words: Broselow, children, estimation, resuscitation, weight.

Correspondence: Dr Simon Young, Department of Emergency Medicine, Royal Children’s Hospital, Flemington Road, Parkville, Vic. 3052, Australia; Email: youngs@cryptic.rch.unimelb.edu.au

Karen Black, MD FRCP, Clinical fellow; Peter Barnett MBBS FRACP FACEM, Deputy Director of Emergency Services; Rory Wolfe, PhD, BSc Senior Lecturer in Biostatistics; Simon Young MBBS FACEM Director of Emergency Services.

Accuracy of estimating children’s weight

161

Introduction

Knowledge of a child’s weight is often necessaryto enable accurate treatment. Many drug doses,intravenous fluid therapy and direct current countershocks are all calculated on a weight basis. As itmay not be possible to accurately weigh a seriouslyill or injured child during resuscitation, a numberof methods to estimate a child’s weight have beendevised. In this paper, we test the accuracy and clinicalperformance of six of these methods in childrenbetween the ages of 1 month and 18 years presenting toa tertiary paediatric emergency department.

Method

This is a prospective observational study comparingthe accuracy and clinical performance of six weightestimation methods.

Setting and population sample

The Royal Children’s Hospital, Melbourne is a 300 bedpublic children’s hospital providing a wide rangeof paediatric services for the state of Victoria. Itsemergency department treats approximately 58 000children per year from a wide variety of socioeconomicand ethnic groups.

Children who attended the emergency depart-ment on 30 non-consecutive days during the months ofMay to August 1999 between the hours of 08 00 and18 00 were eligible for the study. Children needingimmediate resuscitation, those clinically dehydrated,and those with severe joint contractures wereexcluded. The study was approved by the ethicscommittee of Women’s and Children’s Health.

Prior to the study commencing it was estimatedthat a sample size of approximately 500 children withequal numbers in four different weight groups wasrequired. This would enable the study to ensureaccurate assessment of agreement between the weightcalculated by each estimation method and themeasured weight to within 2% error on 95% limits ofagreement.1

Measurements

All children were registered, weighed, measured forlength and assigned a body habitus by the sameinvestigator, Karen Black (KB).

The true weight of the child was determined withthe child undressed on a seated scale (ABLE scale typeASD, capacity 150 kg) or a baby scale (Kubota digitalmodel DS 30A). The scales were tested for accuracyusing 1 kg, 10 kg and 22.4 kg weights. The seatedscale was accurate throughout the range, and the babyscale was accurate to within 5 g at 10 kg. Accuracywas tested at the beginning and end of the study.

The length of the child was determined by using arecumbent board for children under 100 cm, a standingstadiometer (Holtain Ltd, Pembs, UK) for children over100 cm and a tape measure for bigger children unableto weight bear. All children under 145 cm were alsomeasured with the Broselow tape.2

A body habitus (slim, average or heavy) wasassigned by KB for each child. This was necessary forthe devised weight estimation method (DWEM).3

Weight estimation methods

The child’s estimated weight was calculated by each ofsix methods. These methods were:1. The advanced paediatric life support course

method (APLS).42. The Broslow tape.23. The devised weight estimation method (DWEM).34. The Oakley tables.55. The Traub–Johnson method.66. The Traub–Kichen method.7The first four methods were selected as they werespecifically designed to be used during paediatricresuscitation. The Traub–Johnson and Traub–Kichenmethods were developed for non-emergency use butwere included in the study for comparison and to testfor potential application in the emergency setting.

The calculations necessary for each weight estima-tion method together with any limitations are shownin Table 1.

Statistics

The assessment of accuracy for the purpose of weightestimation requires analysis on two points. The firstis whether a weight estimation method on averagecalculates the correct answer. The second is howmuch variability is there in the estimated weights ofa method when applied to many different children ofalmost identical weight. These points are, respectively,the measurement bias and precision of that method.

The bias inherent in each method was estimatedby calculating the mean percentage error (PE) of that

K Black et al.

162

method by the formula: PE = 100 × (estimated weight− true weight) / true weight.

The standard deviation (SD) of the PE was usedto estimate the precision in the method for approx-imating true weight.

Mean and standard deviation of percentage errorswere calculated for each method in the weightgroups: < 10 kg, 10–25 kg, 25–40 kg, > 40 kg. Theseweight groups were selected to allow analysis ofinfants (generally children < 12 months old are< 10 kg) in one group, the older child whose weight isoften more difficult to estimate in another. They alsoallowed exclusion of the methods that were notintended for less than 1 year of age or over 10 years.

The assessment of body habitus is necessary forthe DWEM method. Inter-rater agreement for theassignment of body habitus was assessed by showing70 slides of children undressed (from a collectionof clinical slides for various conditions) and askingeach of 26 medical and nursing staff (the raters) to tickthe appropriate box (slim, average, heavy). To assessintra-rater agreement, six of the medical staff wereasked a week later to repeat the rating exercise withthe slides presented in a different random order. Inter-and intraobserver agreement was summarized with κstatistics. A value of the κ statistic between 0.61 and0.80 is regarded as showing substantial agreementbetween observers whereas a value between 0.41 and0.60 is regarded as showing moderate agreement.8

All analyses were performed using Stata9 exceptfor the calculation of confidence interval (CI) for the κstatistic when SPSS for Windows was used.10

Results

Five hundred and two patients were eligible for thestudy, four were too uncooperative to get an accuratelength measurement and three parents declined toallow their child to participate. Four hundred andninety-five patients completed the study. The sampleconsisted of 283 boys and 212 girls and wasrepresentative of a wide range of ages and weights.The details of each group are shown in Table 2.

The weight for a given height of girls wascompared with boys in the weight groups < 10 kg,10–25 kg, 25–40 kg and > 40 kg. Despite boys in the< 10 kg group being slightly heavier at a given length,analysis of the weight estimation methods for eachsex separately yielded identical conclusions. The resultspresented are therefore for boys and girls combined.

The mean PE and 95% CI for each method areshown in Table 3. The mean PE and SD of PE for eachmethod are shown in Fig. 1.

The results indicate that the most accurate methodsof estimating weights in this paediatric populationwere length-based methods such as the DWEM andthe Broselow tape. The Broselow tape showed the

Table 1. Methods of weight estimation of children requiring resucitation

Name of method Applicability of method Method of calculation

Advanced paediatric life support (UK)4 Age range: 1–10 year Weight in kg = 2 × (Z + 4) Z = age in years (to nearest half year)

Broselow tape2 Length range: 46–143 cm Read weight from measuring tapeDevised weight estimation method3 Length range: 50–175 cm (i) measure length

(ii) assign habitus (iii) read weight from table

Oakley5 Length range: 50–160 cm (i) measure length (ii) read weight from graph

Traub–Johnson6 Age range: 1–18 year Weight in kg = 2.05 × e0.02×X Or, for males > 60 inches Weight in kg = 39 + 2.27 × (Y − 60) Or, for females > 60 inches Weight in kg = 42.2 + 2.27 × (Y − 60) X = height in cm Y = height in inches

Traub–Kichen7 Length over 74 cm and age range: 1–17 year

Weight in kg = 2.396 × 1.0188X X = height in cm

Accuracy of estimating children’s weight

163

least bias and the best precision within its limits (foruse between 46 and 143 cm) and was the best methodin the under 10 kg group. The DWEM proved to be themost useful over the whole spectrum of ages andweights, having the greatest range and showing theleast bias and greatest precision in the heavier groups.

Other methods estimated weight reasonably well inthe 10–25 kg and 25–40 kg groups, but were limited intheir range and lost accuracy in the< 10 kg and > 40 kggroups. The Oakley method tended to overestimateweights in the lighter children especially those under10 kg and underestimate the weights of heavierchildren. Its precision in all weight groups was com-paratively poor. The APLS method underestimatedthe weight with the highest mean percent error, andhad the least precision of the methods studied.

The interrater agreement for assigning bodyhabitus as slim, average or heavy, was assessed bycomparing each rater’s answers with each of the other25 sets of answers (a total of 325 distinct pairings).The proportion of exact agreement and a κ statisticwere calculated for each of these pairs. Across thesepairs, the mean proportion of exact agreement was78% (SD 6%, range 58–93%) with a mean κ statisticof 0.65 (SD 0.09). The lowest κ statistics indicating theleast agreement were observed between junior nursingand medical staff whereas the highest κ statisticswere found with more experienced paediatricians and

senior nurses. The overall κ statistic for agreementwithin the slim category was 0.73 (95% CI: 0.72, 0.75),in the average category 0.57 (95% CI: 0.56, 0.59), andin the heavy category 0.66 (95% CI: 0.65, 0.67). Themean (across six raters) of the observed intra-raterexact agreement was 86% (range 81–94%). Kappastatistics for these raters ranged from 0.69 (95% CI:0.54, 0.84) to 0.87 (95% CI: 0.76, 0.99) with a mean κstatistic of 0.75. This indicates substantial agreement.

Discussion

All commonly used algorithms for paediatric cardiacarrest rely on estimation of a child’s weight. Vital com-ponents of resuscitation attempts such as adrenalinedose, joules of energy and intravenous fluid bolusesare calculated using the child’s weight and are allnecessary during circumstances where weighing thechild may be totally impractical. Although there isscant evidence regarding optimum drug doses, delayin obtaining a weight or calculating an inaccurate onehas the potential to adversely affect a resuscitationoutcome. Access to an easy, unbiased and accuratemethod of weight estimation is likely to assist inreducing the potential mistakes in resuscitation,especially in centres where there are few paediatrictrained staff, or few paediatric emergencies.

Table 2. Age range and number of patients in each weight group

Weight group n (kg) Number in group Median age in months Minimum age in months Maximum age in months

≤ 10.0 121 7 1 2110.1–25.0 132 41 7 12025.1–40.0 121 109 36 183≥ 40.1 121 160 89 225

Table 3. Bias of six methods of paediatric weight estimation

Method APLS Broselow DWEM Oakley Traub–Johnson Traub–Kichen

< 10 kg –0.6 (–2.8, 1.5) 7.0 (5.0, 9.1) 16 (13, 18)n = 12110.1–25kg –4 .7 (–6 .7, –2.7) –0.4 (–2.0, 1.3) 0.7 (–0.8, 2.1) 10 (8.8, 12) –2.6 (–4.2, –1.0) –0.7 (–2.2, 0.7)n = 13225.1–40kg –20 (–22, –17) –6.4 (–9.1, –3.7) –2.0 (–4.2, 0.2) –1.0 (–3.6, 1.7) –4.0 (–6.4, –1.5) –6.5 (–8.7, –4.2)n = 121 n = 102 n = 86> 40kg –42 (–46, –38) –3.9 (–5.9, –1.9) –11 (–15, –7.1) –15 (–18, –13) –16 (–18, –12)n = 121 n = 21 n = 110 n = 68

Mean percentage error (estimate of bias) is given with 95% CI in parentheses. The number in each group varies due to the limitationsof each method with respect to height, weight and age. Where numbers are less than the full group, n is indicated.

K Black et al.

164

All emergency room personnel must therefore haveaccess to a method capable of estimating a child’sweight that is reasonably accurate for the populationthey treat. Oakley reported that house officers andregistrars, especially those working in emergencydepartments are inaccurate when asked to providethe endotracheal tube size, adrenaline and atropinedoses, and joules for DC conversion in different agedchildren.5 Harris examined estimates of paediatric weightsmade by doctors, nurses and parents, and found thatthey were all unreliable, with only 42% of parents, 30%of doctors and 25% of nurses estimating to within5% of the child’s weight. Ranges varied from under-estimating by 55.6% to overestimating by 292%.11

Our study has determined that the DWEM andthe Broselow tape were the most accurate methods ofweight estimation in our study population. Both

methods are based upon measurement of a child’slength with the DWEM requiring an additionalestimate of body habitus. Both methods are quick,simple to apply and do not require any equipmentbeyond the tapes and charts.

The Broselow tape or a standard measuring tapecould be part of the resuscitation room equipmentor the resuscitation trolley, with the DWEM chartposted on the trolley or a nearby wall. As measuringlength and reading off values are simple manoeuvre,either method could conceivably be performed prior toarrival at hospital, facilitating preparation for resusci-tation. The use and accuracy of these methods in theprehospital setting would need to be further evaluated.

The APLS method underestimated weight with thehighest mean percentage error and the least precisionof any of the methods studied. It should not be used for

Figure 1. Mean percentage error (estimate of bias) indicated by an open circle; bars are plus and minus two standard deviations of percentage error (estimate of precision). The number in each group varies due to the limitations of each method with respect to height, weight and age. Where numbers are less than the full group, n is indicated.

Accuracy of estimating children’s weight

165

estimating weight when another method is possible.This method is still useful in the preparation stage ofresuscitation where it may not be possible to measurethe child. In this situation the knowledge that theAPLS method significantly underestimates weightespecially in the heavier and older children may be ofimportance to the resuscitation team.

In our study, only one person measured everychild’s length or height. While this assists with theinternal comparison of different weight estimationmethods that rely on a value for length or height, it isa limitation for the generalization of the estimatedmean and SD of PE as there will be some variabilityin the measuring technique between different staff.We also used mostly standing height in ambulatorychildren that could be different from recumbent lengthin a resuscitation situation. However we believe suchdifferences between staff or in measuring techniquesare likely to be very small.

We have shown that there is acceptable agreementbetween observers of different levels of experience inthe assignment of body habitus, reducing the possiblevariability from the DWEM. However, our assessmentof intra-rater reliability using slides instead of realchildren might not completely reflect ‘patient’ habitusreliability, but we felt this to be a minor limitation. Forpopulations that have a different distribution of bodyhabitus than our population, there is an advantage tousing a tiered table such as the DWEM and we wouldexpect that this method may be more accurate acrosspopulations.

Children requiring immediate resuscitation wereexcluded from the population sample as it was notrealistic to obtain the measured weight in thesecircumstances. As all of the estimation methods areeasy to apply, we do not believe that the measure-ment bias or the precision of each method wouldbe significantly altered by the conditions under whichit was used. This assumption, however, may need tobe tested by further research. Although clinicallydehydrated children were excluded, it is likely that asmall number of mildly dehydrated children wereincluded in the study. Their inclusion would beexpected to affect the results because the estimationmethods would overestimate their actual weight at

presentation. We did not exclude this group as theywould realistically be present in a paediatric popu-lation in any emergency department.

Conclusion

The most accurate methods of weight estimation inchildren are the Broselow tape and the DWEM. Werecommend the use of either of these methods inemergency situations where direct weighing is notpossible.

Accepted 5 December 2001

References

1. Bland MJ, Altman DG. Statistical methods for assessingagreement between two methods of clinical measurement.Lancet 1986, February 8, 307–10.

2. Lubitz DS, Seidel JS, Chameides L et al. A rapid method forestimating weight and resuscitation drug dosages fromlength in the pediatric age group. Ann. Emerg. Med. 1988; 17:576–81.

3. Garland JS, Kishaba RG, Nelson DB et al. A rapid and accuratemethod of estimating body weight. Am. J. Emerg. Med. 1986; 4:390–3.

4. Mackway-Jones K, Molyneux E, Phillips B, Wieteska S.Advanced Pediatric Life Support. London: BMJ Books,2001.

5. Oakley PA. Inaccuracy and delay in decision making inpaediatric resuscitation, and a proposed reference chart toreduce error. BMJ 1988; 297: 817–19.

6. Traub SL, Johnson CE. Comparison of methods of estimatingcreatinine clearance in children. Am. J. Hosp. Pharm. 1980; 37:195–201.

7. Traub SL, Kichen L. Estimating ideal body mass in children.Am. J. Hosp. Pharm. 1983; 40: 107–10.

8. Landis J, Koch JR. The measurement of observer agreement forcategorical data. Biometrics 1977; 33: 159–74.

9. Stata Corporation. Stata statistical software: Release 6.0.College Station, Texas: Stata Corporation, 1999.

10. SPSS Inc. SPSS for Windows, Base system user’s guide: Release10.0. Chicago USA, 1999.

11. Harris M, Patterson J, Morse J. Doctors, nurses, and parentsare equally poor at estimating pediatric weights. Pediatr.Emerg. Care 1999; 15: 17–18.