Table 4.
Referral eligibility prevalence, % (n/N) | Sensitivity for identifying deaths (95% CI) | Specificity for identifying deaths (95% CI) | Positive predictive value for mortality (95% CI) | Negative predictive value for mortality (95% CI) | Likelihood ratio positive (95% CI) | Likelihood ratio negative (95% CI) | AUC | Diagnostic odds ratio for identifying deaths (95% CI) | Likelihood ratio test* | |
---|---|---|---|---|---|---|---|---|---|---|
WHO IMCI without pulse oximetry | 8·1% (313/3848) | 25·0% (9·7–46·7) | 92·0% (91·1–92·8) | 1·9% (0·7–4·1) | 99·5% (99·2–99·7) | 3·1 (1·5–6·2) | 0·8 (0·6–1·0) | 0·58 (0·49–0·67) | 3·8 (1·5–9·6) | Ref |
WHO IMCI with SpO2 <90% | 10·5% (403/3848) | 41·7% (22·1–63·4) | 89·7% (88·7–90·7) | 2·4% (1·2–4·5) | 99·6% (99·3–99·8) | 4·0 (2·5–6·5) | 0·6 (0·4–0·9) | 0·65 (0·55–0·75) | 6·2 (2·7–14·1) | 7·9 |
WHO IMCI with SpO2 <90% or failed measurement | 12·0% (461/3848) | 54·2% (32·8–74·4) | 88·3% (87·2–89·3) | 2·8% (1·5–4·7) | 99·7% (99·4–99·8) | 4·6 (3·1–6·7) | 0·5 (0·3–0·8) | 0·71 (0·61–0·81) | 8·9 (3·9–20·0) | 15·2 |
WHO IMCI with SpO2 <94% | 17·5% (674/3848) | 33·3% (15·6–55·3) | 84·9% (83·8–86·1) | 1·3% (0·5–2·6) | 99·5% (99·2–99·7) | 2·2 (1·2–3·9) | 0·7 (0·5–1·0) | 0·59 (0·49–0·68) | 2·8 (1·2–6·6) | 11·0 |
WHO IMCI with SpO2 <94% or failed measurement | 19·0% (732/3848) | 62·5% (40·6–81·2) | 81·3% (80·0–82·5) | 2·0% (1·1–3·3) | 99·7% (99·5–99·9) | 3·3 (2·4–4·5) | 0·4 (0·2–0·7) | 0·71 (0·62–0·81) | 7·2 (3·1–16·5) | 18·6 |
WHO IMCI guidelines refer to children categorised into having WHO IMCI severe pneumonia. AUC=area under the receiver operating characteristic curve. IMCI=Integrated Management of Childhood Illnesses. SpO2=peripheral oxyhaemoglobin saturation.
χ2 test statistic for the likelihood ratio test comparing the reduced model (WHO IMCI without pulse oximetry) and the full model (WHO IMCI with pulse oximetry [SpO2 threshold specified by table row]). p values for all likelihood ratio test comparisons <0·01.