Inconsistent diagnosis of acute malnutrition by weight-for-height and mid-upper arm circumference: contributors in 16 cross-sectional surveys from South Sudan, the Philippines, Chad, and Bangladesh

Table 3 Sensitivity and specificity of MUAC₁₂₅ and WHZ₋₂ to detect slim children

		Total N	Slim n	WHZ₋₂ (n = 2 022)						MUAC₁₂₅ (n = 983)
Covariates				Sensitivity	95 % CI		Specificity	95 % CI		Sensitivity	95 % CI		Specificity	95 % CI
Long legs	Yes	4,309	847	84.2	75.9	92.5	82.9	73.7	92.1	32.2	26.4	38.0	96.8	90.2	100
Long legs	No	6,079	536	68.3	63.9	72.8	95.2	90.2	100	32.0	28.0	36.1	97.5	92.9	100
Sex	Male	5,020	720	73.4	67.4	79.3	93.6	86.9	100	25.9	20.9	30.9	99.2	93.7	100
Sex	Female	5,368	663	68.8	62.7	75.0	93.4	86.6	100	33.6	28.7	38.6	100	94.8	100
Age < 24 months	Yes	3,754	664	75.0	69.9	80.2	95.0	89.2	100	57.4	51.1	63.7	87.7	80.6	94.8
Age < 24 months	No	6,634	719	65.8	59.1	72.5	94.1	86.9	100	18.1	14.0	22.1	97.6	93.1	100
Total		10,388	1,383	70.6	65.4	75.9	93.9	88.3	99.6	31.0	26.8	35.2	98.4	93.9	100

Slimness was defined as a WAZ<−2 in children with no linear growth retardation, i.e. with HAZ≥−2. Adjusted sensitivity and specificity are derived from multivariate multi-level linear regression models, with “survey” as random effect. The regression coefficients represent the absolute % in the dependent variable (WHZ-2 or MUAC125) between the categories of the independent variable (slimness). Sensitivity is computed by summing the regression coefficient with the constant of the regression model. Specificity is computed by using the reciprocal of the dependent and independent variables in the model. All models were adjusted for all covariates. Interactions between covariates and slimness were tested all together through a chunk test. The chunk test was highly significant for both indicators, and all covariates had a modifying effect on the sensitivity and specificity

ISSN: 1475-2891