Accurate height and length estimation in hospitalized children not fulfilling WHO criteria for standard measurement: a multicenter prospective study

In hospitalized children, height should be measured. When world health organization (WHO) height measurement gold standards is impossible, the ideal height estimation technique is still unclear. We conducted an international prospective study in eight different pediatric intensive care units to assess the accuracy, precision, practicability, safety, and inter-rater reliability of 12 different height estimation techniques, based on body segment measurement extrapolation, or other calculations using previous or projected heights. All extrapolation techniques were performed on each child, and later compared to their WHO gold standard heights. A total of 476 patients were enrolled. In the < 2-year subgroup, board length use and growth chart extrapolation performed best. In the ≥ 2-year subgroup, growth chart extrapolation and parents’ report were the most accurate, followed by height measurement alongside the body with a tape measure. In both groups, body segment extrapolations were poorly predictive and showed mean bias and limits of agreement that varied a lot with age. Most body segment-based techniques presented with frequent measurement difficulties, but children’s safety was rarely compromised. The inter-rater reliability of body segment measurement was low in the < 2-year subgroup. Conclusions: To accurately estimate height in hospitalized children, health care professionals should integrate the accuracy, precision, practicability, and reliability of each measurement technique to select the most appropriate one. Body segment-based techniques were the least accurate and should probably not be used. Simple techniques like growth chart extrapolation, or measurement alongside the body (and length board measurement in the youngest) should be implemented in daily practice. Trial Registration: The study protocol was registered (12th April 2019) on the clinical-trial.gov website (NCT03913247). What is Known: • Height should be measured in hospitalized children to assess nutritional status and calculate various clinical parameters. • Many hospitalized children cannot be measured using WHO conventional height measurement methods. The ideal height estimation method has not been identified yet. What is New: • Most estimation methods based on body segment measurement extrapolation fail to accurately predict height. • Board length use and growth chart extrapolation performed best in young children (≥2 years). Growth chart extrapolation and parents’ report were the most accurate, followed by height measurement alongside the body with a tape measure in older children. Supplementary Information The online version contains supplementary material available at 10.1007/s00431-024-05692-3.


Introduction
In hospitalized children, weight and height should be measured at admission, especially in pediatric intensive care units (PICU) [1,2].Children's height evolves with age and requires monitoring as standard follow-up.Height allows for nutritional status assessment to calculate body mass index (BMI) and height-for-age z scores.It is also used to determine resting energy expenditure (Schofield equations), which is essential to prescribe adequate nutritional support and avoid over-and underfeeding which are associated with impaired outcomes [1,2].Height is necessary for the calculations of clinical parameters such as corporal surface area (CSA), which is important to prescribe some medications like chemotherapy, for burn scores, or for the interpretation of CSA indexed parameters like cardiac output.Height is also used to calculate ideal and adjusted weight in obese children, which is important to prescribe some medications with body composition distribution variation or to set mechanical ventilation parameters.Errors in height measurement will induce errors in all these parameters: for example, a 3.5% error would lead to a 2%, 7.5%, and 5% error respectively in CSA, BMI, and Schofield energy requirement calculations, which accuracies are already impacted by weight measurement errors (overhydration, indwelling devices, etc.) and by the 4% mean bias of Schofield equations [3].Thus, reducing height measurement errors is important.
The world Health organization (WHO) has published gold standard procedures for children's height and length measurement.In children below 2 years of age, length should be measured on a straight lying child using a length board.In children above 2 years of age, height is measured standing upright, using a height board [4].In hospitalized children, however, these WHO conventional anthropometric approaches ("WHO gold standard") for length and height measurement cannot always be undertaken.Some children cannot stand upright because of their clinical condition; others are equipped with tubes, drains, plasters, plaster casts, etc. that may interfere with height measurements or compromise their safety.
In children who cannot be measured following the WHO gold standards (e.g., cerebral palsy), other height estimation or extrapolation techniques have been proposed, including body segment length extrapolation or other estimation techniques [5][6][7][8][9].In this study, we aim to assess, in hospitalized children, the accuracy, the reliability, and the safety of these techniques, as they have not been validated in this setting yet.

Material and methods
We undertook a multicenter prospective cross-sectional study.It was conducted in PICU as we considered critically ill children would be the most likely to present with contraindications to or difficulties to perform height measurement as per WHO gold standards (in relation to their severe critical condition and multiple equipment); we also hypothesized that the results could be extrapolated to less severe patients in pediatric wards.
Eight PICUs participated from Belgium, France, and Switzerland, from 2019 to 2022.Children (28 days to 18 years) were included if they did not fulfill the WHO gold standard criteria for height measurement at the time of enrolment (because of their clinical condition or any indwelling equipment).Each participant was later measured according to WHO criteria and acted as his own control to compare height estimation techniques to WHO gold standards measurements.The period between enrollment time point and WHO measurement timepoint had to be shorter than 5% expected natural height growth, based on WHO height and length velocity growth charts (supplemental digital content 1 provides these delays for various age ranges, based on WHO height velocity curves).Children with abnormal skeletal presentations (e.g., nanism, scoliosis, limb abnormalities) or retractions were excluded.Two groups of children were distinguished, based on their age (< 2 years, ≥ 2 years), as WHO criteria differ.Parental consent was obtained.Ethical clearance was obtained from the "OUEST III protection of persons" committee (7 th October 2019).The study protocol was registered (12 th April 2019) on the clinical-trial.govwebsite (NCT03913247).
During the child's PICU stay, height estimation techniques were performed for each child: All techniques are described in detail in Supplemental Material 2. Local investigators followed a 1-day in-person training to perform measurements following a written protocol, provided by the principal investigator to optimize measurement reproducibility.Ulna, tibia, arm span, and knee-heel measurements were performed with both a tape measure and a standard caliper (Cescorf Large Bone Anthropom-eter©, Nutriactiva, Mineapolis, USA).Measurements were performed by a trained local investigator, with the help of the nurse in charge of the patient.
When the child recovered and fulfilled WHO criteria, height was measured as per the WHO gold standard and each child served as his own control.Our main objective was to compare estimation techniques to the WHO gold standard in terms of "overall accuracy" or bias (defined as a mean relative error < 3.5%).We also considered the "individual accuracy" or precision of the techniques based on the percentage of patients presenting with a relative error < 3.5%, its variation with children's height (considered a surrogate of children's age), the dispersion of measurements, if these were 0-centered, and if numerous outliers were encountered.
We also aimed to assess each measurement technique safety (based on the number of patients presenting with clinical condition worsening during measurements or indwelling device accidental removal) and applicability (based on the number of patients presenting with impossible or difficult measurements such as reading result difficulty, child's position holding difficulty, indwelling devices being obstacle to measurements, body segments difficulty to locate).
We further aimed to assess the inter-rater reliability.In two of the eight centers, all measurements were performed twice on the same child, by two different trained operators, blind to each other, and results were compared.
Patient characteristics and indwelling devices were collected to describe the population.
This study was reported using STARD, the most appropriate EQUATOR reporting checklist [10].

Statistical analysis
A 3.5% relative error in height measurement was considered to significantly impact on CSA, BMI, and Schofield equation accuracy.The standard deviation of the relative error was expected to be equal to 4%, assuming 10% of children for whom either a technique or the WHO gold standard would not be available; thus, the inclusion of 231 children per group (< and ≥ 2 years) would provide a precision in the mean relative error estimate of ± 0.5% (mid-width of the 95% confidence interval).The absolute mean relative error was tested for each technique against 3.5% with a one-sided normal test.The percentage of children for whom the relative error was < 3.5% was described.The mean relative bias was estimated with its associated 95% confidence interval, and the Bland and Altman concordance plot was provided for each method.The intraclass correlation coefficient was calculated for each method to assess the inter-rater reliability.The percentage of children with a least one difficulty, and at least one safety issue was calculated per technique.
The accuracy analysis was performed on children with WHO gold standard measured within the acceptable time period defined previously; the safety and practicability analysis were performed on all the children for whom the measurements were attempted and the reliability analysis on those who were measured twice regardless their WHO gold standard measurement.Missing data were not imputed, and analysis was performed on available data.R software version 4.1.1 was used.A p-value less than 0.05 was considered significant.

Patient characteristics
In total, 476 patients were enrolled (244 and 232 in the < 2-year and ≥ 2-year age subgroups respectively).Of them, 239 and 223 in each respective group could have their height measured as per WHO criteria and were further analyzed.In each age subgroup, 244 and 232, and 47 and 72, were analyzed regarding the safety-practicability and interrater reliability of the measurement respectively (see patient flow chart in Supplemental Material 3).
Table 1 (and Supplemental Material 3) presents patients' characteristics and their indwelling devices or condition that compromised their height measurement.All children were equipped with at least one device and 73% were mechanically ventilated.Tables 2 and 3 present the percentages of patients who could have their height estimated for each technique: those were high (> 95%) except for parent's report, and missing data were rare.  2 and 3 present the mean bias and the relative error compared to the WHO gold standard of each technique assessed to estimate height (detailed results: in Supplemental Material 3 and 4).In the < 2-year subgroup, body segment length extrapolations were poorly predictive of patients' length, and board length (− 0.3 cm [95% CI − 0.5; 0.0]) and growth chart extrapolation (− 0.1 cm [95% CI − 0.4; 0.1]) performed best.In the ≥ 2-year subgroup, body segment length extrapolations were better than in younger children but remained poorly accurate, and growth chart extrapolation (− 0.7 cm [95% CI − 1.7; 0.3]) and parents' report (− 0.42 cm [95% CI − 0.9; 0.0]) were the most accurate.However, the mean bias and the limits of agreement of each technique varied with age (height being a surrogate of age) as shown in their Bland and Altman graphs presented in Figs. 3 and 4.

Accuracy
As a majority (64.6%) of the patients were enrolled from two centers, a sensitivity analysis was conducted to describe the difference with the other centers.Compared to the WHO gold standard, the accuracy of the estimation techniques was higher in these two centers (Supplemental Digital Content 5).

Safety and feasibility
Safety issues and difficulties encountered during measurements are presented in Tables 2 and 3 (and in Supplemental Material 3 and 4).Safety issues (almost all in relation with caliper use) were rarely encountered and were not severe.Difficulties were more frequently encountered in the youngest (p = 0.03): in the < 2-year subgroup, tibia measurement was the most difficult to perform, followed by body segments and ulna measurement (tibial lateral condyle and ulna styloid process locating difficulty was the most frequently reported).In the ≥ 2-year subgroup, length measurements alongside the body and body segment measurements were the most difficult to perform followed by tibia length.Caliper use also generated more difficulties than tape measure use in both groups (p ≤ 0.01).Measurements caused pain in a few children and the caliper frightened some of them.Difficulties to position the patient and for them to maintain the position during measurements were more frequent for the half of the arm span but rare for other measurements.Finally, locating body segments and tibia and ulna extremities were the main cause of the reported difficulties.

Inter-rater reliability
In the inter-rater reliability analysis, 119 patients were enrolled.The inter-rater reliability is presented in Tables 2  and 3 (and in Supplemental Material 3 and 4).Body segment measurements had the lowest inter-rater reliability.In the < 2-year subgroup, tibia and ulna measurements were the less reliable techniques, and in the ≥ 2-year subgroup, all measures had high intraclass correlation coefficients.

Discussion
This is the first study that has tested several techniques to estimate height or length when the WHO conventional anthropometric approaches could not be performed.They presented with a large variation in their accuracy, precision, practicability, and inter-rater reliability, but were all safe.Body segment-based techniques are the least accurate and should not be used, and simple techniques like growth chart extrapolation or measurement alongside the body and length board measurement in the youngest were the most useful.
Systematic measurement of height is good practice in hospitalized children [11].However, limited reports are available in the literature to assess how this is done in children not fulfilling WHO criteria, such as PICU patients, and from any pediatric ward.In a study conducted in PICU, height measurement increased from 32 to 99% of the patients after a training program and the implementation of ulna and tibia length extrapolation [12].However, the accuracy of such techniques remained questionable.
Body segment measurements to extrapolate height have been validated in specific children populations (e.g., neuromuscular weakness, cerebral palsy).Their accuracy and inter-rater reliability were tested in school age healthy but at an individual level, they may present with a lack of precision which limits their use in clinical practice [7,8].Other studies found heterogenous accuracies of these techniques [13,14] and a recent review concluded that most studies showed no correspondence between extrapolated and real height [9].Not surprisingly, in our youngest subgroup of children, these techniques were of overall poor quality, as most of the extrapolation calculation formulas  were developed in children above 5 years of age.However, in the older subgroup, their precision remained low.We found inter-rater reliability was high in most techniques, after a short training period of the local investigators.The safety of all measurements was very high even in critically ill children.A lot of difficulties were encountered, which was expected in PICU.However, most measurement were possible with few missing data.The use of a dedicated caliper did not increase measurement accuracy (compared to a simple inexpensive tape measure as shown by Spender et al. in 1989 [15]), and generated more difficulties, and we do not recommend its use.
The precision, safety, reliability, and practicability of the technique should be taken into consideration when choosing which one to implement in daily clinical practice.Accuracy was sometimes satisfactory in the overall subgroup, but with large variations within the children's ages, which compromises its precision at the individual level.We pragmatically presented in Fig. 5a qualitative summary of the features of all techniques assessed, to allow selecting the optimal one to implement in clinical practice.In the youngest subgroup, length board measurement was considered a high-quality technique and was possible in most cases (expect in those with scalp catheter or head dressings).Measurement alongside the body and growth chart extrapolation also performed well.In the older subgroup, measurement alongside the body, growth chart extrapolation, and parental report achieved the highest overall quality.However, if no previous measurement is available or in countries where regular height measurement is not standard of care, growth chart extrapolation and parental report may be impossible.
Ulijaszek et al. [16] reviewed anthropometric measurement errors in children, finding that weight and height measurements were accurate and reproductible, but other anthropometric measurements were not (e.g., waist and hip circumferences, skinfolds).Some limb measurements used to extrapolate height or length in our study similarly presented poor inter-rater reliability as per Ulijaszek definition (R > 0.95) and thus failed to accurately predict height.

Height difference in cm; mean length in cm
Bland & Altman graphs for other techniques are available in supplemental material 3 Fig. 3 Bland and Altman graphs, by method and length (surrogate of age) in children < 2 years Outside the cerebral palsy setting, we found few studies that have assessed the reliability of height estimation techniques confirming our results.Scalercio et al. compared newborn length measurements using a length board and a tape measure and found that weighted kappa coefficient and intraclass correlation coefficient indicated good to excellent agreement, but no other techniques were tested [17].In children with learning difficulties, Hardy et al. [18] found disappointing wide bland-Altman limits of agreement between measured supine length and that estimated from segmental measurements.Young et al. performed a systematic review [19] of weight estimation techniques in children (in both hospitalized and healthy children) and showed that parent's estimation and length-based techniques predicted the most accurately children's body weight (which may be difficult to measure), which was confirmed in a recent study conducted by O'Leary et al. [20].Consequently, this is crucial to measure length accurately to correctly estimate weight in hospitalized children who cannot be weighed easily or accurately.
Our study was conducted in PICU.However, a lot of hospitalized children outside the PICU also have clinical conditions or indwelling devices impairing height measurement.The translation of these PICU results to the ward setting might thus be possible.The choice of a simple technique such as the measurement alongside the body with a tape measure appears the most useful and is easier than all other sophisticated techniques.
Limitations The interrater reliability of growth charts, WfA z-score, and genetic target extrapolation was not tested, but these techniques do not depend on anthropometric measurements and strictly following a written protocol should have limited the bias.However, this is the first study that has tested numerous techniques to estimate height in hospitalized children, using a standardized and homogeneous procedure for each of them to limit potential bias, and focusing on various key features, relevant to clinical practice (precision, practicability, safety, reliability).Children with limb abnormalities or abnormal skeletal presentations or Height difference in cm; mean length in cm Bland & Altman graphs for other techniques are available in supplemental material 3 Fig. 4 Bland and Altman graphs, by method and height (surrogate of age) in children ≥ 2 years significant retractions were excluded and results should not be extrapolated to this population.Two thirds of the measurements were performed by the same operator; however, his higher accuracy rate mainly reflects the impact of the operator' experience, as homogenous training to measurements techniques was part of the study design.

Conclusions
Numerous techniques have been proposed to estimate hospitalized children's height when the WHO conventional anthropometric approaches are not applicable.Health care professionals should consider a number of factors (accuracy, precision, practicability, and reliability) in their selection of the ideal technique to use for height measurement in their specific practice settings.Body segment-based techniques are the least accurate and should probably not be used.Simple techniques like growth chart extrapolation or measurement alongside the body and length board measurement in the youngest are the most useful.New extrapolation equations have recently been proposed [21] but still require adequate validation prior to clinical implementation.
the corresponding author upon reasonable request.Data are located in controlled access data storage at Hospices Civils de Lyon.

Figures 1
Figures 1 and 2 present the relative error (%) of each estimation technique compared to the WHO gold standard.Figures 3 and 4 present the mean bias and limits of agreement variations with height of 7 key techniques (results of all techniques are shown in Supplemental Material 3).Table2and 3 present the mean bias and the relative error compared to the WHO gold standard of each technique assessed to estimate height (detailed results: in Supplemental Material 3 and 4).In the < 2-year subgroup, body segment length

Fig. 1 Fig. 2
Fig. 1 Boxplot of height relative errors compared to WHO gold standard, in the < 2-year subgroup

Table 2
Children < 2 years.Concordance, percentage of absolute value of the relative error, number of difficulties or safety issues, and intraclass correlation coefficients of each extrapolation or estimation of height or length methods a Number of patients (and its percentage of the 239 included patients) who could be measured according to the WHO gold standard and allow for comparison to each technique (these figures were used for mean bias, p-value, and % of children with relative error < 3.5%) b % based on all the children who were assessed by the estimation technique, regardless of the WHO comparability

Table 3
Children ≥ 2 years.Concordance, percentage of absolute value of the relative error, number of difficulties or safety issues, and intraclass correlation coefficients of each extrapolation or estimation of height or length methods a Number of patients (and its percentage of the 223 included patients) who could be measured according to the WHO gold standard and allow for comparison to each technique (these figures were used for mean bias, p-value, and % of children with relative error < 3.5%) b % based on all the children who were assessed by the estimation technique, regardless of the WHO comparability