Background
Evidence about optimal mode of delivery for preterm birth is lacking, and there is thought to be considerable variation in practice.

Objective
To assess whether variation in hospital preterm caesarean section rates (Robson Classification Group 10) and outcomes are explained by casemix, labour or hospital characteristics.

Materials and Methods
Population‐based cohort study in NSW, 2007–2011. Births were categorised according to degree of prematurity and hospital service capability: 26–31, 32–33 and 34–36 weeks’ gestation. Hospital preterm caesarean rates were investigated using multilevel logistic regression models, progressively adjusting for casemix, labour and hospital factors. The association between hospital caesarean rates, and severe maternal and neonatal morbidity rates was assessed.

Results
At 26–31 weeks’ gestation, the caesarean rate was 55.2% (seven hospitals, range 43.4–58.4%); 50.9% at 32–33 weeks (12 hospitals, 43.4–58.1%); and 36.4% at 34–36 weeks (51 hospitals, 17.4–48.3%). At 26–31 weeks and 32–33 weeks’ gestation, 81% and 59% of the variation between hospitals was explained with no hospital significantly different from the state average after adjustment. At 34–36 weeks’ gestation, although 59% of the variation was explained, substantial unexplained variation persisted. Hospital caesarean rates were not associated with severe maternal morbidity rates at any gestational age. At 26–31 weeks’ gestation, medium and high caesarean rates were associated with higher severe neonatal morbidity rates, but there was no evidence of this association ≥32 weeks.

Conclusion
Both casemix and practice differences contributed to the variation in hospital caesarean rates. Low preterm caesarean rates were not associated with worse outcomes.