Artificial Intelligence Benchmarking of Stillbirth Risk Prediction from Fundal-Height Z-Scores Under Unmeasured Gestational Dating Error: A Healthcare Decision-Curve Evaluation

Fundal-height Z-scores are widely used to screen for fetal growth restriction in low-resource antenatal settings, yet their predictive value for stillbirth is confounded by gestational dating error that is rarely measured and almost never reported. The framework separates three sources of error—dating error, Z-score construction error, and decision-threshold error—and shows that the net benefit of a risk-stratified stillbirth pathway is bounded above by the reliability of the dating variable rather than by the sophistication of the classifier. A replication-based power analysis of the before-after-control-impact estimator was conducted across 2000 design cells with between-cluster SD 6.0, residual SD 4.5, and intraclass correlation 0.64; 52 clusters were required to detect a 5-percentage-point effect at 80% power. The analysis indicates that decision-curve evaluation of stillbirth risk models must report dating-error sensitivity alongside discrimination and calibration, and it proposes a reporting standard that makes the dating assumption auditable. The contribution is a methodological benchmark, not an empirical estimate of stillbirth risk.