The accuracy of LUS to predict respiratory support in late
preterm and term infants
The ROC analysis for any respiratory support needs using high-risk
patterns yielded an AUC of 0.95 (95% CI, 0.92-0.98, p<0.001).
Correspondingly, the ROC analysis for any respiratory support needs
using low-risk patterns yielded an AUC of 0.89 (95% CI, 0.85-0.93,
p<0.001). In contrast, the ROC for RR (respiratory rate),
which we conventionally use to predict respiratory support needs,
yielded an AUC of 0.70 (95% CI, 0.64-0.76, p<0.001). The ROC
curves for LUS (high-risk patterns) and RR were significantly different
(p<0.01) (Figure 3). Table 2 shows the accuracy data for LUS
and RR prediction of any need for respiratory support. A ROC analysis
within respiratory support subgroups was also conducted, and the result
for hood oxygen support is shown (Table 2). However, due to an
insufficient number of patients who only obtained CPAP, MV, and PS
treatment (23/74, 18/74, and 29/74, individually), the ROC curves for
these analyses may not reliable and results are not shown.