. 2016 Aug 16;11(8):e0159260. doi: 10.1371/journal.pone.0159260

Table 2. Comparison of weight estimation methods (n = 453,990).

Variable			BT 2007 [B]	BT 2011 [A]	MUAC [Hong Kong]	MUAC1	HEIGHT1
Variable			Percentage Difference between True Weight and Estimated Weight Mean (SD)^*
Weight (kg)	(0,25]	All children	-4.58 (10.38)	-8.95 (11.24)	-7.68 (27.70)	+0.62 (29.92)	+0.49 (10.33)
	(0,10]	Children ≤ 10 kg	-9.64 (11.44)	-13.15 (11.12)	-11.48 (34.95)	+2.53 (39.08)	+1.55 (11.05)
	(10,25]	Children > 10 kg	-1.82 (9.12)	-6.54 (10.40)	-5.71 (24.25)	-0.16 (25.90)	-0.10 (9.88)
MUAC (mm)	< 115	Severely wasted	-22.76 (12.34)	-27.14 (12.36)	+64.27 (21.10)	+88.84 (20.79)	-9.39 (12.02)
	115 ≤ MUAC < 125	Moderately wasted	-15.46 (9.74)	-19.48 (10.25)	+24.18 (17.60)	+42.30 (16.16)	-4.87 (10.98)
	MUAC ≥ 125	Normal	-3.12 (9.76)	-7.46 (10.23)	-12.41 (25.17)	-5.02 (26.36)	-1.30 (9.99)
WHZ	WHZ < -3	Severely wasted	-31.08 (8.06)	-36.81 (7.73)	+33.96 (39.70)	+53.85 (45.89)	-21.63 (10.58)
	-3 < WHZ < -2	Moderately wasted	-20.05 (5.50)	-25.45 (5.29)	+15.13 (26.20)	+29.17 (28.39)	-13.27 (7.56)
	WHZ ≥ -2	Normal	-2.48 (9.09)	-6.69 (9.79)	-10.91 (26.07)	-3.31 (27.70)	2.34 (9.26)
HAZ	HAZ < -3	Severely stunted	-5.11 (11.47)	-8.54 (12.02)	-5.80 (30.28)	+5.52 (34.61)	3.01 (9.98)
	-3 < HAZ < -2	Moderately stunted	-3.92 (10.85)	-7.58 (10.59)	-10.15 (26.59)	-1.26 (29.56)	1.82 (9.61)
	HAZ ≥ -2	Normal	-4.67 (10.25)	-9.55 (11.01)	-7.30 (27.13)	0.00 (28.61)	-0.78 (10.46)
WAZ	WAZ < -3	Severely underweight	-\8.07 (10.90)	-22.00 (10.99)	+15.39 (34.42)	+32.74 (39.25)	-7.44 (10.39)
	-3 < WAZ < -2	Moderately underweight	-10.67 (9.00)	-14.55 (9.78)	-1.56 (26.50)	+10.16 (28.71)	-3.56 (9.74)
	WAZ ≥ -2	Normal	-1.56 (9.06)	-6.01 (9.93)	-11.64 (26.41)	-4.90 (27.76)	+2.48 (9.82)
Estimates accurate to within ± 25% of true weight^**			95.58%	91.36%	61.78%	59.35%	97.36%
			(95.52%; 95.64%)	(91.28%; 91.44%)	(61.64%; 61.93%)	(59.21%; 59.49%)	(97.40%; 97.46%)
Estimates accurate to within ± 10% of true weight^**			62.93%	51.67%	27.89%	26.25%	66.19%
			(62.79%; 63.08%)	(51.53%; 51.81%)	(27.77%; 28.03%)	(26.14%; 26.37%)	(66.05%; 66.32%)
Estimates in same BT weight class^**^†			64.83%	61.53%	30.37%	28.43%	63.23%
			(64.69%; 64.97%)	(61.39%; 61.67%)	(30.24%; 30.50%)	(28.30%; 28.56%)	(63.09%; 63.37%)
Estimates accurate to within ± 1 BT weight class^**^†			99.23%	98.94%	76.43%	73.00%	98.92%
			(99.20%; 99.26%)	(98.91%; 98.97%)	(76.31%; 76.55%)	(72.87%; 73.13%)	(98.89%; 98.94%)
Weighted Kappa^**			0.8835	0.8756	0.6048	0.5922	0.8884
			(0.8807; 0.8863)	(0.8727; 0.8784)	(0.6030; 0.6066)	(0.5905; 0.5940)	(0.8856; 0.8913)
Bland-Altman bias and 95% limits of agreement^‡			-0.44 (-2.47; 1.59)	-0.93 (-3.25;1.40)	-0.80 (-6.90; 5.30)	0.068 (-6.45; 6.59)	0.05 (-2.15; 2.24)

^Calculated as: $p e r c e n t a g e d i f f e r e n c e = \frac{t r u e w e i g h t - e s t i m a t e d w e i g h t}{t r u e w e i g h t} \times 100$

* Mean and SD were estimated using Huber M estimators of location and scale.[52] The mean percentage difference is a measure of systematic bias or accuracy. The SD percentage difference is a measure of precision. The difference in accuracy between any pair of methods can be assessed using the ratio of the absolute values of their mean percentage difference. The difference in precision between any pair of methods can be assessed using the ratio of their SD percentage differences. For example, comparing BT 2011 (A) and HEIGHT 1 in all children: $Δ a c c u r a c y = \frac{| - 8.95 |}{| 0.49 |} = 18.27 \times i m p r o v e m e n t; Δ precision = \frac{11.24}{10.33} = 1.09 \times i m p r o v e m e n t$ Values above one indicate better performance. Values of one indicate no difference in performance. Values below one indicate worse performance.

** Point estimate and 95% confidence interval

^† Appropriate BT classes for BT 2007 [B] and BT 2011 [A]. BT 2011 [A] classes are used for MUAC1 and HEIGHT1 results.

^‡ Bias (mean of true—estimated weight, or mean error) and 95% limits of agreement were calculated following the method of Bland & Altman[55]