• To determine the effectiveness of an intervention delivered in general practice, designed to improve recording and management of domestic violence in primary care, compared with usual practice in terms of the rate of referral to a specialist domestic violence agency providing advocacy.
• To determine the effectiveness of an intervention delivered at the general practice level, designed to improve recording and management of domestic violence in primary care, compared to usual practice in terms of the rate of domestic violence disclosure in relation to the number of women aged 16 or older registered in the practice.
• To estimate the cost effectiveness of the intervention.
• To determine whether individual physician attitudes, knowledge and behaviour in relation to domestic violence change after the intervention.
General practices in two primary care trusts in the south west of England (Bristol) and east London (Hackney) are eligible for participation if they have EMIS LV, EMIS PCS or iSoft Synergy electronic medical record systems. These systems allow the use of clinician prompts and templates within the medical record and were the dominant systems in both areas (>90%).
We have obtained lists of practices in two primary care trusts in the south west of England (Bristol) and east London (City and Hackney) and stratified them by practice characteristics: number of whole time equivalent doctors, training status and presence of a practice counselor. All practices on the lists with the relevant record systems were considered eligible for the trial except for two in which investigators were practicing. A stratified random sample of the general practices in each site was generated from random number tables and those practices were then contacted in rank order. General practices received a postal invitation to participate, including details of the study and a copy of an editorial from a general practice journal highlighting domestic violence as a central primary care issue [50
]. The invitation was followed up within a fortnight by telephone contact with one of the general practitioners (GF in London and DS in Bristol) to discuss possible participation. Practices where initial interest in the study was expressed were posted or e-mailed a further information sheet, and given time to consider participation in the trial. Practices wishing to have further details or clarification were offered a practice visit to explain the study in more detail. Implicit consent from individual clinicians within recruited practices was assumed from their subsequent attendance at training sessions and their completion of study questionnaires.
Study design and practice allocation
As the intervention was targeted at the practice teams (clinicians and reception staff), the trial is cluster randomised with the practice as unit of analysis. The 48 practices recruited in both Hackney and Bristol were allocated (by JR for Hackney practices and DD for Bristol practices) to intervention and control arms of the trial with a computer minimisation programme, including a random component (Minim Version 1.3). Allocation was determined by the programme and the process was concealed from the person entering practice minimisation variables. The variables were the percentage of female doctors in the practice (whole time equivalents), postgraduate training status, the number of patients registered with the practice, and the percentage of the practice population on low incomes.
The training and support intervention within an educational outreach framework [51
], has a theoretical basis in adult learning theory and peer influence [52
]. It focuses on addressing the barriers to asking women patients about abuse and responding appropriately, which includes offering referral to a specialist domestic violence agency that provided advocacy. The primary intervention consisted of two 2-hour multidisciplinary training sessions, scheduled at lunchtime on practice premises, targeted at the clinical team: general practitioners, directly employed practice staff (practice nurses and counselors), and those employed by primary care trusts (midwives, health visitors, district nurses) who have contact with patients registered in the practice. The training sessions were designed to address the expressed and tacit barriers to improving the response of clinicians to women experiencing abuse through improved identification, support and referral to specialist agencies. These sessions incorporated case studies and role play in relation to asking about violence and responding appropriately. They were delivered by an advocate educator based in one of the two collaborating specialist agencies: Nia project in London http://www.niaproject.info/
and Next Link in Bristol http://www.nextlinkhousing.co.uk/
, a clinical psychologist specializing in domestic violence and an academic general practitioner. The advocate educator was central to the intervention, combining a training and support role to the practices with provision of advocacy to women referred from the practices. The training sessions were followed by periodic contact with the practice in clinical meetings, feeding back anonymised practice data on disclosure and referral to the advocacy service, and reinforcing guidance on good practice with regards to domestic violence, as well as ad hoc
telephone conversations with clinicians about referrals or advice. One hour training sessions with administrative staff focused on issues of confidentiality and safety for patients experiencing abuse and introduced the IRIS information materials signposting domestic violence agencies. Ongoing support to clinicians and reception staff in the practices was provided by the named domestic violence advocate educator, with the aim of consolidating the initial training
Intervention practices also were asked to identify a "champion" for the project; with the agreement of the practice, a member of staff from any of the clinical disciplines was invited to attend an additional two days training about domestic violence and to integrate this into the work of the practice.
Other components of the intervention include: a pop-up template linked to diagnoses (such as depression, anxiety, irritable bowel syndrome, pelvic pain and assault) which acts as a prompt to remind clinicians to ask questions about domestic violence and to record this in the electronic medical record, an explicit referral pathway to the named advocate in the Nia project and Next Link and publicity materials about domestic violence visible in the practices (waiting rooms and women's toilets). The template in the electronic medical record is based on HARK, four questions which have been validated for the identification of intimate partner violence in UK general practice [51
Control practices are offered the training sessions once the follow-up period for the intervenrm of the trial is complete.
Our primary outcome is the rate of referral to advocacy and/or specialist domestic violence agencies for one year following the educational sessions in intervention practices (measured for each practice in the intervention and control arms). A woman aged 16 or over is counted as a referral within the practice if her general practitioner records referral information. The denominator is the number of women aged 16 or over within the practice.
Domestic violence disclosure rate for one year following the educational sessions in intervention practices: this is measured for each practice (intervention and control) from searches of the practice database for domestic violence codes in the electronic medical record. The denominator is the number of women aged 16 or above within each practice.
Resource use and associated costs of the intervention are collected for estimation of the IRIS programme's cost-effectiveness.
Physician Readiness to Manage Partner Violence Survey (PREMIS) [53
]: this is completed by health professionals within each intervention and control practice at two time points (baseline and twelve months post intervention). We compare sum scores for perceived preparation, perceived knowledge, actual knowledge, opinions and self-reported practices.
Data are collected by research associates from the electronic medical records in each intervention practice twelve months after the second training session. This comprises identification and referral to domestic violence advocacy of women who have experienced or are experiencing domestic violence. We are searching over two time periods: (i) the twelve month period preceding the intervention, and (ii) the twelve month period after the second training session. Use of the trial template and/or domestic violence codes is identified by a series of computer searches, followed by examination of any eligible patient's electronic record that contains these codes, to determine identification or referral to any agency providing domestic violence advocacy. The same data are collected from the control group practices (the only exception being that identification data based on the use of the trial template codes will not be available). At the time of follow-up data collection for each intervention practice, data are collected from a control practice recruited at least 12 months prior to the data collection time point with similar practice size and training practice status to ensure comparability in the timing of data collection from similar practices.
Whether a record of identification or referral is present in a record or not, is determined using a detailed flowchart (available from authors) to reduce subjectivity. If the researchers extracting data from the medical record are uncertain about whether an identification or a referral was made or not, this will be referred to an independent outcomes panel, blinded to the arm of the trial and identity of the practice the case came from, for a decision. The panel members consist of two general practitioners (one with extensive domestic violence research experience) and a trial methodologist. Direct extracts from the medical record are presented to the table and decisions are made by majority voting if there is no unanimity. Validity of referral data collection will be assessed by an independent researcher searching and extracting from the records of four randomly selected practices in each of the two sites.
Some women may self-refer to various agencies for support or be referred by other professionals. As these referrals may be affected by our intervention (either negatively or positively), other sources of referral data will also be collected and sensitivity analyses conducted: (i) self-referral or referral from a non-general practice source recorded in advocacy service logs, (ii) clinician referral recorded in the advocate educators' records, (iii) referral to other domestic violence agencies. After the primary analysis based on referrals in the general practice record, we will conduct sensitivity analyses, combining these referrals sequentially with referrals in (i), (ii), (iii) and finally with all referral sources combined.
Cost data for the intervention are directly recorded: salaries and management costs of advocate educators, materials (posters and cards), practice payments for training sessions and travel costs. Data on domestic violence costs are based on secondary sources [5
The 'Physician Readiness to Manage Partner Violence Survey' (PREMIS) questionnaire is completed by GPs and practice nurses in intervention and control practices at baseline, and again concurrent with outcome data collection one year later.
Due to the nature of the intervention, patients and practice staff cannot be blinded to allocation status, although patients are given no explicit information about the intervention. Research associates collecting outcome data could not be blinded to allocation status, but we have used a detailed flowchart for extraction to minimize this, plus referral to a panel for arbitration over difficult cases. We are also validating a sample of our data collection using a blinded researcher. All analyses will be conducted blind to treatment allocation.
With 24 intervention and 24 control practices, assuming a disclosure rate of 1% in control practices (a conservative estimate based on our survey of 12 Hackney practices [54
] and an intra-cluster correlation coefficient of 0.03, we will be able to detect an increase of 5.2% in the disclosure rate with a power of 80% at a significance level of 0.05. This calculation assumes that there are on average 1600 women in the relevant age group in each practice, and also takes account of variation in cluster size. With this number of practices we will also be able to reliably detect a three-fold difference between intervention and control practices in the referral of women disclosing abuse to domestic violence advocacy services.
Analyses of referral and identification will be by intention to treat with a Poisson regression model. Our independent variable is the number of referrals/identifications for each cluster. The number of women aged 16 or older will be included as the exposure and practice will be included as a random effect. The analysis will be adjusted for minimization factors. Additionally the analysis will be adjusted for the practice baseline rate if considered to improve the precision of our treatment estimate and not adversely affecting the fit of the model.
Directly collected cost data for the intervention will be combined with service costs and costs of domestic violence in relation to imputed benefits of advocacy referral in a Markov model to calculate cost effectiveness in terms of cost per quality adjusted life years. We have used this approach in a pilot study for this trial [55
Before-and-after analyses of scores for each of the three sections of the PREMIS questionnaire (knowledge, opinion, practice) will be carried out using descriptive statistics and multivariable models (including respondent characteristics) to test for the effect of the intervention.