Objective To investigate whether vaginal Group B Streptococcus (GBS) colonisation or other baseline characteristics of women with preterm premature rupture of membranes (PPROM) can help in identifying subgroups of women who would benefit from immediate delivery. Design Secondary analysis of the PPROMEXIL trials. Setting Sixty hospitals in the Netherlands. Population Women with PPROM between 34 and 37 weeks of gestation. Methods Random assignment of 723 women to immediate delivery or expectant management. Main outcome measures Early onset neonatal sepsis. Results Vaginal GBS colonisation status was the only marker which was significantly associated with the benefit of immediate delivery (P for interaction: 0.04). GBS colonisation was observed in 14% of women. The risk of early onset neonatal sepsis in GBS‐positive women was high (15.2%) when they were managed expectantly but this risk was reduced to 1.8% with immediate delivery. The early onset neonatal sepsis risk was much lower in neonates of GBS‐negative women: 2.6% after expectant management and 2.9% with immediate delivery. We estimated that by inducing labour only in GBS‐positive women, there would be a 10.4% increase in term delivery rate, while keeping neonatal sepsis and caesarean delivery rates comparable to a strategy of labour induction for all. Conclusions Our post hoc findings suggest that women with PROM between 34 and 37 weeks might benefit from immediate delivery if they have GBS vaginal colonisation, while in GBS‐negative women labour induction could be delayed until 37 weeks.