Our protocol and the development of the intervention are more fully described elsewhere . The Improving Maternal and Child Health Care for Vulnerable Mothers (MOVE) project was a cluster randomised trial with MCH teams the unit of randomisation, as the intervention aimed to alter DV screening and care practice across the whole MCH team, not only at the centre or individual nurse level. We also randomised at team level to minimise contamination across centres. We involved eight teams previously randomised in 2005 for a trial (Mothers’ Advocates In the Community, MOSAIC) of mentor mother support . The MOSAIC DV trial tested a non-professional, mother-to-mother support intervention for pregnant or abused recent mothers. MCH nurses participated by recruiting women they had identified into the peer support program. Despite six hours of DV training, nurse identification of abused women was low. In process evaluation, nurses requested improved ways of working to identify and support women and children experiencing violence. MOVE developed from this previous trial .
Randomisation and masking
MOSAIC study randomisation had involved the eight MCH teams stratified by size (numbers of births per annum) using opaque envelopes. Selection was made at a public forum by someone outside the study team. To build on MCH nurse feedback from the MOSAIC trial, all eight teams were recontacted in 2009 to participate in a follow-up study with reverse randomisation (previous intervention teams would become comparison teams and vice versa). Managers of MCH teams gave consent through new signed Memoranda of Understanding on behalf of the eight MCH teams for this MOVE study and to participate in reverse randomisation. Blinding of MCH teams to intervention status was not possible given the participatory nature of the intervention with MCH nurses in the intervention arm engaged in design and delivery of the model, but mothers attending were blinded.
Participants included eight MCH teams in the disadvantaged north-west suburbs of Melbourne . Four intervention and four comparison teams and the postpartum women with babies ≤ 12 months who attended them participated. The intervention was delivered to postpartum women attending intervention centres at scheduled visits over one year.
MOVE model development commenced on 30 June 2009 with four intervention team nurse consultants. Normalisation Process Theory (NPT)  was the theoretical framework for the design, implementation and evaluation of the MOVE model, as it aims to strengthen sustainability in health care behaviours, work practices and systems. NPT is based on sociological theory that extends individual explanation of behaviours to predict facilitators and barriers to ‘normalisation’ of new clinical practices. The theory conceptualises types of ‘work’ required for implementation, embedding and integration of complex interventions.
The participatory research process involved all-day meetings between intervention team nurse consultants and research staff held each month for six months during the development phase to discuss and conceptualise perceived nurse DV screening and care problems and solutions using the NPT framework. Following the meeting, each nurse consultant discussed strategies within their own team to improve DV practice by individual nurses, their team and their local government employer and brought them back for discussion. Along with the action research, a systematic search was undertaken of controlled interventions and evidence-based guidelines that aimed to improve clinician responses to abused women and their children. Findings were shared with advisory group members and nurse consultants, to facilitate development of the MOVE clinical resources. Evidence suggested that women prefer self-completion screening methods rather than face-to-face/direct asking . This informed the design and use of the self-completion maternal health checklist used at three to four months. Utilising results from the unpublished evidence review of the community nursing DV practice and iterative development process, nurse consultants and research staff jointly designed the consensus model described below.
The MOVE intervention
The enhanced screening and care model included nurse mentors, designated DV regional liaison workers based in DV services, a self-completion maternal health and wellbeing checklist and a clinical pathway and guidelines . The checklist asked questions about physical symptoms, for example, sore nipples and backache, as well as DV questions asked face to face by nurses in the comparison arm and outlined below. Nurse mentors’ roles included assisting with nurse safety (for example, accompanying nurses home visiting where violence was suspected), supporting colleagues with difficult consultations and enhancing liaison with DV services. Additional MOVE screening was implemented later than mandated, at a non-routine specific maternal health visit at three months in three teams. Additional local government funding had to be found for these visits. In one team, 15 minutes was added instead to the funded routine four month visit. Research staff visited intervention teams once to bring resources, outline yearlong intervention processes and introduce the DV service liaison worker. Research staff administered an online survey for nurses at six to eight months into, and three months after the intervention period (process and impact evaluation) and collected checklists quarterly. The intervention period was 1 April 2010 to 31 March 2011.
Supplementary intervention checklists
The checklist included DV questions asked face to face by nurses in the comparison arm (outlined below) with the following additions:
Do you have any problems in your relationship or intimacy with your partner?
Has anyone in your household ever humiliated you or tried to control what you can or cannot do? (This was recommended in training but not in the MCH manual.)
Women were provided with the checklist at the commencement of the three or four month visit and encouraged to complete it themselves while nurses cared for the child. Nurses responded to women’s self-identified concerns in the checklist but responses to the intimate relationship and violence questions were addressed first.
Comparison and intervention teams were both trained and mandated to undertake universal DV screening at four weeks postpartum using recommended standard questions suggested in the MCH service practice guidelines  which covered abuse broader than that from an intimate partner:
Are you in any way worried about the safety of yourself or your children?
Are you afraid of someone in your family?
Has anyone in your household ever pushed, hit, kicked, punched or otherwise hurt you?
Comparison teams did not use self-completion checklists for screening or any other elements of the MOVE model. Screening occurred only via face-to-face nurse questioning as mandated at four weeks and at other times if considered necessary.
Informed consent was negotiated at the MCH team level and supported by a written Memorandum of Understanding to participate. Women were unaware of their nurse’s participation in the trial and women’s return of the survey questionnaire was regarded as consent to participation in the survey. The study was approved by the Human Ethics Committee, La Trobe University (UHEC 08-142) and also by the University of Melbourne and the Victorian Government Department of Education and Early Childhood Development (ADD/07/6733).
Outcome data collection
Routinely collected MCH data
MCH nurses routinely complete anonymised and computerised reports on each episode of DV screening, safety planning and referral, and data are collated annually by each team in report form to the Department of Education and Early Childhood Development. These data were extracted by all eight MCH teams’ administrative staff for the full 12 months of the intervention period, and at three months afterwards, and forwarded to the research team. These same data were later also requested for the two year post-intervention (1 April 2011 to 31 March 2013).
Specifically, the data included a) the numbers of women screened and all consultation numbers, regarded as opportunities to screen (denominator) at four weeks, four months and twelve months; b) numbers of reported safety plans; and c) numbers of referrals for the year and all active infant records representing all those who had attended local government MCH centres (denominator) for both arms of the study.
Supplementary intervention checklists
In the intervention arm, MOVE maternal health screening checklists were collected quarterly from specified collection boxes. Although checklists were to be completed at the three or four month maternal health visits, some were returned that were completed at other time points. Only those recorded at three or four month visits as intended and those fully completed (2,771/4,152) were retained for analysis. Maternal health checklist use and DV screening at three months could not be assessed for the 24 months follow-up period, as MCH teams do not routinely collect and report any three month data.
To achieve β of 0.20 and α at 0.05, the survey sample size (n = 10,000) assumed an 8 to 10 % DV prevalence to predict a 15 % increase in disclosure, taking into account birth rate per team, a 55 % response rate, clustering of clients and an intra-cluster correlation (ICC) of 0.02 from a previous study . Outcomes from client perspectives were measured by anonymous surveys (n = 10,472) about recent mothers’ emotional health. These were mailed in June 2011 by an external data capture company blinded to the study arm to all women who had given birth between 1 May and 31 December 2010, in order to reach the required sample size (when babies were between 6 and 12 months old). DV was measured using the well-validated Composite Abuse Scale (CAS)  with recommended cut-off scores - low levels (a score of 3 to 6) and high (a cut-off of ≥ 7) considered probable and confirmed DV. The survey also asked women whether the nurse inquired about a range of maternal health issues, including DV, during any of the visits, whether the topic was discussed and whether affected women were referred. Other measures included three items from the Consequences of Screening Tool (COST), a measure of harm from screening .
Surveys were returned to the data entry company, cleaned, coded and double-entered into a secure database by company staff blinded to the trial arm and then forwarded to the study statistician, also blinded.
Using routine data at all time points, relative risks between arms were estimated using binomial regression, adjusting for clustering of observations at team level and stratified by time for the proportion of a) women screened (four weeks and four months), b) reports of nurses discussing safety plans and c) reported referrals.
In the intervention arm only, we also calculated the checklist screening proportions by time point (three or four months). The one team using the checklist at an extended routine four month visit provided screening data at this time point to be compared with the comparison team average post-intervention.
For survey respondents, women’s socio-demographic and birth characteristics were compared by trial arm to assess the effectiveness of randomisation. Respondent representativeness was assessed by comparison with available Victorian Perinatal Data Collection routine data for all women giving birth in the region.
An intention to treat  analysis was undertaken to estimate risk ratios with robust variance estimation to adjust standard errors for clustering of study participants within teams. We used contingency table analyses for bivariable and generalised linear modelling (specifying a binomial distribution and log link function) for multivariable models. For primary outcomes as pre-specified, data were adjusted for women’s abuse status and for confounding variables, including women’s socio-economic status. All data were analysed using STATA 11 .