BACKGROUND: Mothers are at risk of domestic violence (DV) and its harmful consequences postpartum. There is no evidence to date for sustainability of DV screening in primary care settings. We aimed to test whether a theory-informed, maternal and child health (MCH) nurse-designed model increased and sustained DV screening, disclosure, safety planning and referrals compared with usual care. METHODS: Cluster randomised controlled trial of 12 month MCH DV screening and care intervention with 24 month follow-up. The study was set in community-based MCH nurse teams (91 centres, 163 nurses) in north-west Melbourne, Australia. Eight eligible teams were recruited. Team randomisation occurred at a public meeting using opaque envelopes. Teams were unable to be blinded. The intervention was informed by Normalisation Process Theory, the nurse-designed good practice model incorporated nurse mentors, strengthened relationships with DV services, nurse safety, a self-completion maternal health screening checklist at three or four month consultations and DV clinical guidelines. Usual care involved government mandated face-to-face DV screening at four weeks postpartum and follow-up as required. Primary outcomes were MCH team screening, disclosure, safety planning and referral rates from routine government data and a postal survey sent to 10,472 women with babies </= 12 months in study areas. Secondary outcomes included DV prevalence (Composite Abuse Scale, CAS) and harm measures (postal survey). RESULTS: No significant differences were found in routine screening at four months (IG 2,330/6,381 consultations (36.5 %) versus CG 1,792/7,638 consultations (23.5 %), RR = 1.56 CI 0.96-2.52) but data from maternal health checklists (n = 2,771) at three month IG consultations showed average screening rates of 63.1 %. Two years post-intervention, IG safety planning rates had increased from three (RR 2.95, CI 1.11-7.82) to four times those of CG (RR 4.22 CI 1.64-10.9). Referrals remained low in both intervention groups (IGs) and comparison groups (CGs) (<1 %). 2,621/10,472 mothers (25 %) returned surveys. No difference was found between arms in preference or comfort with being asked about DV or feelings about self. CONCLUSION: A nurse-designed screening and care model did not increase routine screening or referrals, but achieved significantly increased safety planning over 36 months among postpartum women. Self-completion DV screening was welcomed by nurses and women and contributed to sustainability. TRIAL REGISTRATION: Australian New Zealand Clinical Trials Registry, ACTRN12609000424202 , 10/03/2009.