Facebook Twitter YouTube

15641

An Improved Pediatric Weight Estimation Strategy

Saturday, October 20, 2012
Room 346-347 (Morial Convention Center)
Susan Abdel-Rahman, Pharm.D., Clinical Pharmacology and Medical Toxicology, Children's Mercy Hospital, Kansas City, MO

Purpose :  Pediatric medical interventions and nutritional assessment rely heavily on the weight of the patient yet, scales suitable for obtaining accurate infant and child weights are unavailable in many remote, resource constrained health-care settings. Though there exist numerous weight estimation strategies to address this problem, most perform poorly in children at the extremes of weight. We developed and validated a pediatric weight estimation strategy (the Mercy method) that addresses the limitations of currently available weight estimation methods.  

Methods Demographic and anthropometric data on children 2 months to 16 years of age were extracted from the National Health and Nutrition Examination Survey. Datasets were randomly partitioned into a method development set (n=17,328) and a method validation set (n=1,938). The final weight estimation method was constructed using humeral length and mid-upper arm circumference. The predictive performance of this method was compared with the performance of 13 previously published weight estimation methods (Table).

Results :  The Mercy method outperformed the 13 other published methods when evaluated by mean error (ME, p<0.002), mean percentage error (MPE, p<0.04), root mean square error (RMSE, p<0.001) and goodness-of-fit (r2=0.98 vs. 0.69 to 0.87; slope=0.97 vs. 0.43 to 0.96; intercept 0.9 vs. 3.1 to 11.8). Moreover, the Mercy method returned a weight for 100% of the children in the validation set and predicted weight within 10% or 20% of actual for a larger proportion of children (Table). These results did not differ within BMI percentile subgroups (e.g. infant, underweight, normal, overweight, obese). 

Conclusion :  The Mercy method performs robustly independently of age and length over a broad range of weights. This method represents a significant improvement over existing age-based, length-based and habitus-based weight estimation strategies. A device developed and based on the Mercy method is currently under investigation.

 

Method

MPE

(%)

ME

(kg)

RMSE (kg)

% predicted

% in agreement within:

 

 

 

 

 

10%

20%

Mercy

-0.46

-0.40

3.65

100

78.6

98.0

APLS

-14.88

-4.87

8.97

52.3

17.8

33.2

ARC

-16.84

-9.49

16.96

90.4

27.5

51.0

Argall   

-4.55

-1.78

6.84

52.3

19.8

34.4

Best Guess

0.98

-1.65

9.96

80.6

31.2

54.0

Broselow          

-6.05

-1.85

4.88

62.6

37.4

53.9

Cattermole

3.51

0.73

3.42

27.3

15.5

23.9

Leffler

-5.52

-2.61

7.44

61.9

24.8

42.9

Luscombe & Owens

0.82

-0.78

6.65

52.3

20.2

35.9

Nelson

-7.69

-3.19

9.33

72.2

25.9

49.4

Shann

-17.38

-10.88

18.78

90.4

26.3

48.6

Theron

9.64

2.23

7.720

52.3

19.6

33.4

Traub-Johnson

-7.13

-4.52

11.93

90.4

45.2

69.3

Traub-Kichen

-9.48

-6.18

13.23

90.1

45.3

67.8