Objective To assess the effectiveness of a three-stage intervention to reduce caesarean deliveries in a Chinese tertiary hospital. Methods A retrospective study was conducted to assess whether educating staff, educating patients and auditing surgeon practices (introduced in 2005) had reduced caesarean delivery rates. Multiple logistic regression was used to check for a potential association between caesarean rates and rates of admission to the neonatal intensive care unit (NICU).: Findings The caesarean delivery rate ranged from 53.5% to 56.1% in 2001-2004 and from 43.9% to 36.1% in 2005-2011. When 2001-2004 and 2005-2011 were treated as "before" and "after" periods to evaluate the intervention's impact on the mean caesarean section rate, a significant reduction was noted: from 54.8% to 40.3% (odds ratio, OR: 0.56; 95% confidence interval, Cl: 0.52-0.59;X-2 test: P < 0.001). The overall drop in the caesarean section rate was significant (X-2 test: P < 0.001) and inversely correlated with the years (Spearman's p: -0.096; P < 0.001). Although complicated pregnancies increased after 2004, the primary caesarean section rate decreased annually by 20% on average in 2005-2011, after practice audits were implemented. Multiple logistic regression showed a positive association between the caesarean delivery rate and the rate of admission to the NICU (adjusted OR: 1.26; 95% CI: 1.14-1.40). Conclusion Patient and staff education and practice audits reduced the Caesarean section rate in a tertiary referral hospital without an increase in admissions to the NICU.