Discriminatory ability
The ability of both charts to predict which fetuses would vswould not go on to develop developmentally vulnerability was poor (AUC
of 53.0% for INTERGROWTH-21st; 52.8% for WHO;
sensitivity of all cut-points < 67% [Tables 4 and
5] ). For example, the 10th centile of the
INTERGROWTH-21st had 3% sensitivity and 38% positive
predictive value, while the 10th centile of the WHO
chart had 6% sensitivity and 36% positive predictive value in
identifying developmental vulnerability. It is noteworthy that
the 10th percentile of both charts identified
considerably less than 10 percent of our cohort as small-for-gestational
age: 2.2% of our population was below the 10thpercentile of the INTERGROWTH-21st chart, and 4.9%
was below the 10th percentile of the WHO chart.
However, discriminatory ability at the percentile of the charts that
classified 10 percent of our cohort as small-for-gestational age (the
33rd percentile of the INTERGROWTH chart, the
24th percentile of the WHO chart) was equally poor.
The statistically-optimal threshold (i.e., that which maximizes both
sensitivity and specificity, based on Youden’s Index) was the
80th percentile of the
INTERGROWTH-21st chart, and the 91stpercentile of the WHO chart.