Design: single-centre before-and-after quality-improvement study with adjusted analysis
Population: 2,674 births with quantitative blood loss recorded (1,309 before, 1,365 after implementation), excluding placenta accreta spectrum and under 24 weeks
Primary outcome: postpartum haemorrhage, quantitative blood loss 1,000 mL or more
Effect: 14.3% to 11.2% overall, adjusted relative risk 0.77 (95% CI 0.64 to 0.93); vaginal births 8.5% to 5.2%, adjusted relative risk 0.60 (95% CI 0.42 to 0.86)

A single centre changed its third-stage protocol from 30 units of oxytocin over an hour followed by 3.6 units over a second hour, to 60 units over one hour, and compared quantitative blood loss across 2,674 births either side of the change: 1,309 before, 1,365 after. Implementation was the ordinary kind — order sets, infusion pumps, staff training — which is what makes the result portable.

Postpartum haemorrhage, defined as quantitative blood loss of 1,000 mL or more, fell from 14.3% to 11.2%, adjusted relative risk 0.77 (95% CI 0.64 to 0.93). Median blood loss fell from 444 mL to 396 mL, adjusted median difference −43 mL (95% CI −72 to −14). The effect sat entirely in vaginal births: 8.5% to 5.2%, adjusted relative risk 0.60 (95% CI 0.42 to 0.86). After caesarean there was no reduction, 23.4% versus 21.8%, adjusted relative risk 0.89 (95% CI 0.71 to 1.11).

The split by mode of delivery is the actionable part. A protocol change worth making on the labour ward is not automatically worth making in theatre, where the causes of blood loss and the other uterotonics in play are different. A unit reviewing its third-stage regimen can reasonably take this to vaginal births and leave caesarean protocols where they are pending better evidence.

In Indian practice, where postpartum haemorrhage remains a leading cause of maternal death and quantitative blood loss measurement is uneven, the measurement discipline in this study is as transferable as the dose. The signal was only detectable because loss was quantified rather than estimated.

Review the third-stage oxytocin regimen for vaginal births specifically, rather than changing a single hospital-wide protocol
Introduce or tighten quantitative blood loss measurement first — estimated loss will not show a 43 mL median shift
Update infusion pump presets and order sets at the same time as the protocol, as the failure mode is administration, not intent
Do not extrapolate the reduction to caesarean birth; the data do not support it
Keep the rest of the haemorrhage bundle intact — this changes a dose, not the escalation pathway

Don't overread it

This is a before-and-after quality-improvement study, not a randomised trial — other changes over the same six months could have contributed to the fall.

The statistics, in plain English

An adjusted relative risk of 0.60 for vaginal births, with an interval from 0.42 to 0.86, is a clear effect on a common outcome — roughly three fewer haemorrhages per hundred vaginal births. The caesarean interval, 0.71 to 1.11, includes 1.0, so no benefit was shown there; that is not proof of no effect, but it is not grounds for changing practice either. The 43 mL median difference is statistically solid and clinically trivial on its own — the outcome that matters is the proportion crossing 1,000 mL, not the shift in the middle of the distribution.