Validation of the Mercy TAPE

In 'real-world' health care settings there exist a number of circumstances where the weight of a child is desirable or even necessary but unavailable. The most conspicuous of these settings can be found in developing countries where many medical clinics lack suitable scales to obtain accurate infant and child weights. Though resource restrictions are less of an issue in developed countries, scenarios still exist where weight assessment is problematic. For example, accurate estimates of a child's weight are rarely available during emergency or trauma situations, and in some in-patient settings (e.g. critical care units, orthopedic clinics) obtaining an accurate patient weight can be impaired by the presence of external hoses, tubing, casts, and/or other medical equipment. Irrespective of the environment, the challenge that each of these settings present is the same; namely, the provision of age-appropriate, weight-based interventions which remain the most accurate approach to delivering therapy in children. Thus, techniques which permit accurate weight estimation address a critical medical need in both developing and developed countries.

Numerous weight estimation strategies have been described with each used to varying degrees in clinical practice. Many of the published techniques have distinct advantages. For example; simple age-based equations can be used without the need for reference materials, strategies that utilize preprinted tables or tools limit the risk of calculation errors. Other techniques present unnecessary complexities for the end-user including; the need for subjective assessments of habitus, the requirement to solve exponential equations, the call for multiple formulae delineated by age bracket, or the reliance on one or more reference charts. Irrespective of their simplicity or complexity, almost all of the reported techniques have significant limitations. Relatively few methods have been evaluated in pediatric populations of varying races, ethnicities and nationalities and essentially no single previously described method provides accurate estimates of weight across broad age- and weight-bands.

Apart from parental recall which can vary in accuracy, the most commonly used strategies for estimating weight rely on the child's age, length, or a combination of the two parameters. While simple and easy to integrate into a weight estimation technique, age based strategies fail to account for the extremes of body composition and stature that are observed in children of the same age. Similarly, length based strategies do not take into consideration that two children of the same height may demonstrate markedly discrepant weights based on underlying nutritional status (e.g. malnourished, underweight, overweight, obese). Consequently, many of the currently available weight estimation strategies perform well in only a small subset of children. As such, there remains a critical need for weight estimation methods that are accurate across a wide range of pediatric ages, weights, lengths, nationalities and body compositions despite the relative abundance of strategies that already exist.

Investigators at Children's Mercy Hospitals and Clinics recently developed and validated a weight estimation method (the Mercy MethodTM) that addresses the principal limitations of previously published methods, requires no subjective assessment and performs robustly independently of age and length over a broad range of weights. As with other strategies, the Mercy Method incorporates growth velocity but uses humeral length as a surrogate for total body length. Total body length will be discrepant depending on whether the measurement is obtained with the child standing or lying down and can be difficult to obtain in a child who is uncooperative or obtunded. The Mercy Method also incorporates body habitus as a quantitative variable which improves the accuracy of the overall length-based weight estimate and removes the subjective nature of categorizing the child's body type into one of a few alternatives (e.g. "slim," "average," or "heavy"). By developing a model with these considerations in mind we were able to expand the age range to which our weight estimation method can be applied and remove length restrictions which are typically imposed because of the disproportionate increase in weight-for-height observed as children get older.

In brief, demographic and anthropometric data on children 2 months to 16 years of age were extracted from the NHANES database and individual datasets were randomly assigned into a method development (n=17,328) or a method validation (n=1,938) set. Humeral length (HL) and mid-upper arm circumference (MUAC) were used to develop a weight estimation method by 1) collapsing length and habitus measurements into discrete bins, 2) examining the median population weight for each bin-pair, 3) statistically weighting the bin-pairs for age and sample size, and 4) calculating a fractional weight for each HL and MUAC. An individual weight estimate is generated by the simple addition of the MUAC and HL fractional bin value that corresponds to that individual child's measurements. The predictive performance this method was evaluated using the internal validation set and compared with the performance of 13 previously published weight estimation methods applied to the same data.

The Mercy Method outperformed the 13 other published methods when evaluated for goodness-of-fit, mean error, mean percentage error, root mean square error and percentage of children in agreement within 10% of actual weight. Most of the age-and length-based strategies examined overestimated weight in children classified, by BMI, as underweight and significantly underestimated weight in children classified as overweight or obese. The degree to which this occurred depended largely on the constants driving their mathematical equations, with some methods biased toward more accurate prediction in children of lower weight (e.g. Broselow) and others performing better among children in the higher weight brackets (e.g. Theron). Irrespective of directionality, the bias observed with some methods at the extremes of weight represented as much as a 3-fold error between predicted and actual weights. Discrepancies of this magnitude can be dangerous, and potentially life-threatening, depending on how 'forgiving' the intervention or treatment that is being administered.

The singular habitus-based method (i.e. Cattermole) ranked among the best (after the Mercy Method) with respect to absolute bias; however, it performed only moderately well when precision and MPE were factored into the assessment. This method, which was developed in Chinese children consistently overestimated weight at lower absolute weights and underestimated weight at higher absolute weight irrespective of BMI percentile. This suggests that while the relationship between weight and MUAC tends to be linear within any given population, the mathematical constants that define the relationship differ between populations having different height-for-weight averages. Given the nature of the data used to develop and validate the Mercy Method, comparative performance of the Devised Weight Estimation Method (DWEM, the only other method to incorporate both body length and body habitus) could not be assessed. Notably, the DWEM involves a subjective rating of "slim," "average," or "heavy". While DWEM has been shown to outperform other age-based methods, the categorical assignment of habitus coupled with inconsistencies in subjective assessment between and within observers [inter-rater agreement- 78% (range: 58-93%); intra-rater agreement- 86% (range: 81-94%)] contributed to bias and precision estimates that were larger than observed with strategies based solely on length.

While the Mercy Method can be used as a reference table, a more practical application was the development of a simple and inexpensive device that can perform the two required measurements simultaneously and report the predicted weight directly from the device as opposed to consulting a separate table or chart. Consequently, the 3D Mercy TAPE was developed to perform both measurements simultaneously requiring no external references to arrive at the weight estimate for a given child. An alternative 2D Mercy TAPE was also designed . It requires two serial measurements with the same simple addition used with the 3D TAPE but does not require any folding or manipulation when removed from its packaging. Both devices are intended to be printed on any flexible, non-stretchable medium (e.g. paper, plastic coated paper, fiberglass) so as to be disposable or semi-permanent, inexpensive to mass produce and easy to store.

In its numeric form, the Mercy TAPE would be expected have limited utility in settings where care providers are illiterate or do not use a written language. However, the tool can be easily revised with colors and/or symbols whose combination would correspond to a given dose, intervention strategy or weight target. While the Mercy Method is expected to perform well in U.S. children given its creation using data from a U.S. database, external validation of the in non-U.S. settings is currently ongoing with support of the World Health Organization to gauge its utility in children of varying ethnicity and geographic origin. The related 2D and 3D Mercy TAPE still awaits prospective evaluation. The requisite study to satisfy the validation requirements are described herein under the hypothesis: The Mercy TAPE will demonstrate the same predictive performance as the Mercy method in an independent pediatric assessment.