BACKGROUND: Evidence-based obstetric care is widely promoted in developing countries, but the success of implementation is not known. Using selected childbirth care procedures in four hospitals in Shanghai, we compared practice against evidence-based information, and explored user and provider views about each procedure. METHODS: Observational study. Using the Cochrane Library, we identified six procedures that should be avoided as routine and two that should be encouraged. Procedure rate determined by exit interviews with women, verified using hospital notes. Views of women and providers explored with in depth interviews.The study sites were three hospitals in Shanghai and one in neighbouring province of Jiangsu. 150 women at each centre for procedure rate, and 48 in-depth interviews with women and providers. RESULTS: Vaginal births were 50% (303/599) of the total. Of the six practices where evidence suggests they should be avoided as routine, three were performed with rates above 70%: pubic shaving (3 hospitals), rectal examination (3 hospitals), and episiotomy (3 hospitals). Most women delivered lying down, pain relief was rarely given, and only in the urban district hospital did women routinely have a companion. Most women wanted support or companionship during labour and to be given pain relief; but current practice is insufficient to meet women's needs. CONCLUSION: Obstetric practice is not following best available evidence in the hospitals studied. There is a need to adjust hospital policy to support the use of interventions proven to be of benefit to women during childbirth, and develop approaches that ensure clinical practice changes.