Fig. 2

Receiver operating characteristic (ROC) curves showing the incremental improvement in perinatal mortality prediction when social determinants are added to clinical models. Model 1 (blue line) includes only clinical variables and has an area under the curve (AUC) of 0.878 (95% CI: 0.849–0.909), with a detection rate (DR) of 0.566 (95% CI: 0.428–0.658) at a 10% false-positive rate (FPR). Model 2 (red line) adds basic social determinants, increasing the AUC to 0.887 (95% CI: 0.859–0.915) and the DR to 0.579 (95% CI: 0.461–0.711). Model 3 (green line) incorporates social support, further improving the AUC to 0.898 (95% CI: 0.878–0.924) and the DR to 0.618 (95% CI: 0.500-0.724). Model 4 (orange line) includes overcrowding and poverty, achieving an AUC of 0.903 (95% CI: 0.877–0.929) and a DR of 0.658 (95% CI: 0.539–0.776). Model 5 (gray line) adds vulnerability indices, reaching the highest AUC of 0.921 (95% CI: 0.899–0.942) and a DR of 0.671 (95% CI: 0.579–0.789)