Skip to main content

Table 3 Performance characteristics of four candidate tools for screening 2–14 year-old children for HIV testing in Ethiopiaa

From: Adapting strategies for effective and efficient pediatric HIV case finding in low prevalence countries: risk screening tool for testing children presenting at high-risk entry points in Ethiopia

Screening tool

pAUC

Threshold scoreb

Sensitivity (95% CI)

Specificity (95% CI)

NNT

Z-HRST

0.76

2

95.0 (87.5–100)

53.6  (51.5–55.7)

26.3

Exact logistic

0.77

0.0047

95.0 (87.5–100)

61.4 (59.4–63.6)

22.0

Rescaled logistic

0.78

4

95.0 (87.5–100)

65.6 (63.5–67.6)

19.7

  1. aOut-of-sample performance of the logistic tools is based on tenfold cross-validation, whereas the value of the partial area under the receiver operating characteristic curve (pAUC) for sensitivity ≥ 80%, sensitivity, specificity and the average number of tests required to find a single HIV-positive child (NNT) is overly optimistic for the Z-HRST (Zimbabwe High-Risk Screening Tool) method
  2. bSums of affirmative responses for the Z-HRST tool, estimated probability of testing positive for the exact logistic tool, and the sum of the rounded rescaled logistic parameter estimates for the rescaled logistic tool