Simplified models to assess newborn gestational age in low-middle income countries: findings from a multicountry, prospective cohort study

Introduction Preterm birth is the leading cause of child mortality. This study aimed to develop and validate programmatically feasible accurate approaches estimate newborn gestational age (GA) in low resource settings. Methods The WHO Alliance for Maternal Newborn Health Improvement (AMANHI) recruited pregnant women from population-based cohorts five countries (Bangladesh, Ghana, Pakistan, Tanzania Zambia). Women <20 weeks gestation by ultrasound-based dating were enrolled. Research staff assessed newborns for: (1) anthropometry, (2) neuromuscular/physical signs (3) feeding maturity. Machine-learning techniques used construct ensemble models. Diagnostic accuracy was areas under receiver operating curve (AUC) Bland-Altman analysis. Results 7428 liveborn infants included (n=536 preterm, <37 weeks). Ballard examination biased compared with ultrasound (mean difference: +9 days) 95% limits agreement (LOA) −15.3 33.6 days (precision ±24.5 days). A model including 10 characteristics (birth weight, head circumference, chest foot length, breast bud diameter, development, plantar creases, skin texture, ankle dorsiflexion infant sex) estimated GA no bias, LOA ±17.3 an AUC=0.88 classifying preterm infant. that last menstrual period (LMP) had ±15.7 high diagnostic (AUC 0.91). An alternative simpler weight LMP ±16.7 AUC 0.88. Conclusion best machine-learning (10 neonatal LMP) within early dating. Simpler models performed reasonably well marginal increases prediction error. These hold promise estimation when unavailable.