Objective This study aimed to evaluate the association between clinical and examination features at admission and late preterm birth. Study Design The present study is a secondary analysis of a randomized trial of singleton pregnancies at 34 0/7 to 36 5/7 weeks' gestation. We included women in spontaneous preterm labor with intact membranes and compared them by gestational age at delivery (preterm vs. term). We calculated a statistical cut-point optimizing the sensitivity and specificity of initial cervical dilation and effacement at predicting preterm birth and used multivariable regression to identify factors associated with late preterm delivery. Results A total of 431 out of 732 (59%) women delivered preterm. Cervical dilation ≥ 4 cm was 60% sensitive and 68% specific for late preterm birth. Cervical effacement ≥ 75% was 59% sensitive and 65% specific for late preterm birth. Earlier gestational age at randomization, nulliparity, and fetal malpresentation were associated with late preterm birth. The final regression model including clinical and examination features significantly improved late preterm birth prediction (81% sensitivity, 48% specificity, area under the curve = 0.72, 95% confidence interval [CI]: 0.68-0.75, and p -value < 0.01). Conclusion Four in 10 women in late-preterm labor subsequently delivered at term. Combination of examination and clinical features (including parity and gestational age) improved late-preterm birth prediction.