Domestic violence is a major cause of injury to women1 and its presentation is highly prevalent in primary health care settings.2,3 However, the sensitivity surrounding this topic prevents the wide-spread implementation of effective management strategies for cases of domestic violence in primary care. Studies have demonstrated low levels of knowledge, awareness, and training and high levels of misperceptions and prejudicial attitudes about domestic violence among primary care providers and office staff.4,5 Other provider-related barriers to the management of domestic violence in primary care include lack of confidence, fear of offense, and time limitations.6 Recent studies suggest that patient-related barriers, including patients' fear of retaliation, lack of disclosure, and cultural differences, may be more prevalent than are provider-related barriers.7,8 Although recommended and included in practice guidelines, routine screening for domestic violence is infrequent in primary care; providers tend to rely on complaint-based screening (e.g., during visits with injury as the presenting complaint).7 However, interventions initiated by health care providers can be important turning points for victims. A qualitative study suggests that acknowledgment of abuse and confirmation of the victim's worth by health care providers is a powerful intervention that enables the victim to move toward safety.9
Our objective was to implement and evaluate a multifaceted domestic violence management program in a rural health network involving three hospitals and 19 regional sites located in Central New York State. The challenge was to address the attributable risk of domestic violence in a large population and create an efficient large-scale intervention not limited to the medical system. We prepared the medical home first and then addressed the general public through health education. Components of diffusion of innovation theory, community organization principles, and social marketing methods, as well as institutional changes in office practice and documentation enablers (modifications that facilitate documentation) were combined to implement this program. We hoped to strengthen the effectiveness of health professionals' training by changing social norms in the community. We worked for change in the community by implementing a public health campaign that employed multichannel and multifactor education. We report here the impact of this program on the knowledge, attitudes, beliefs, and behaviors of health care providers regarding the identification, management, and referral of domestic violence victims in the context of primary and emergency care.
The program's elements are described below. The sequencing of interventions and assessments is portrayed in Figure 1.
Health Professionals' Training
Beginning in May 1997, a “train the trainer” approach was used to train targeted groups of health care providers in a rural health network (three hospitals and 19 clinics). A total of 20 trainers were trained one day a week over five weeks; five were attending physicians, three were nurse practitioners, one was a physician assistant, nine were registered nurses, one was a medical office assistant, and one was a social worker. Domestic violence experts from the New York State Office of the Prevention of Domestic Violence conducted the training, which covered the medical, psychological, legal, and social aspects of managing domestic violence using an empowerment model10 and emphasizing a coordinated community response.
The 20 trainers, in turn, targeted staff in their work areas for shorter training sessions. Training occurred in six departments (emergency, primary care, pediatrics, psychiatry, obstetrics and gynecology, and outpatient surgery), medical and surgical residencies, 19 regional sites, and two county public health departments. The training usually involved two sessions, the first devoted to the nature and dynamics of domestic violence and the second to the identification, management, and appropriate referral of victims. The second session usually ended with a presentation by staff from the local domestic violence advocacy organization who could field questions and describe local services, as well as distribute brochures and materials for patients. Thus, the second training session allowed health care providers to meet face-to-face with advocates for victims of domestic violence in the hope that this contact would facilitate future contacts. A clinical protocol for the identification, management, and referral of domestic violence victims was distributed to all primary care sites to reinforce training and aid recall of procedures when needed.
Following this initial wave of training, additional specialty groups in the network also requested training, including inpatient nursing staff. Although the project was focused on outpatient and primary care, it expanded to inpatient and specialty care work areas in order to respond to unmet needs in these areas.
Nurse triage screening was implemented in two of the network hospitals' emergency departments (EDs). The ED nurses used a four-part screening tool11 as part of the triage process:
* Within the last year, have you been hit, slapped, kicked, or otherwise physically hurt by someone?
* Have you ever been emotionally abused by your partner or someone important to you?
* Within the last year, has anyone forced you to have sexual activities?
* Are you afraid of your partner or anyone you listed above?
Patient-education materials that were specifically designed for each location were produced, including a resource and educational booklet for domestic violence victims and palm cards containing phone numbers for the region being served. Bathroom posters accompanied by palm cards listing local and state domestic violence hotline numbers were displayed. Brief public health education announcements about domestic violence were published weekly in the network's Staff Bulletin, in both written and electronic forms. A bimonthly newsletter containing more detailed domestic violence updates was distributed throughout the rural health network. This newsletter described research updates, local developments, and legislative changes, of which there were several during the project period. Last, two hospital policies, one on violence in the workplace and another on domestic violence assessment and reporting, were written, adopted, and disseminated throughout the organization.
Public Health Campaign
Following the professional training period of the project, a public health education campaign was conducted beginning in October 1998 and ending in April 1999. Its purpose was to increase public awareness and to influence provider practice in domestic violence management. The target audience was potential victims and bystanders. We used focus-group testing and a social marketing approach in our production of radio and print advertisements. The main messages focused on recognizing the full spectrum of domestic abuse, from verbal to physical, the effects of domestic violence, the promotion of active public disapproval, and actions to take. All ads included the slogan “No one deserves to be abused.” One radio ad featured a scenario of a victim disclosing her abusive situation to her doctor. Paid radio advertising was linked to other activities as described below.
Over the six months, the public health campaign involved included approximately 4,000 30-second radio ads, 12 weeks of community public service announcements on television, ten newspaper articles, 36 print ads, 15 speaking engagements, 105 bulletin board posters, 55 clinic bathroom posters, and various mailings to community groups, schools, libraries, and clergy. Tee-shirt art (tee shirts with painted personal messages) designed by victims to depict their experiences with domestic violence was displayed in “clothesline” form for two weeks in the lobby of the Mary Imogene Bassett Hospital in Cooperstown as well as in other parts of the county. Portions of the public health education campaign were independently replicated by the Rural Health Education Network in two neighboring counties where the network has affiliated hospitals and clinics.
Surveys. The Centers for Disease Control's (CDC's) questionnaire on health care professionals' knowledge, attitude, beliefs, and behaviors (KABB) was used for a pre- and post-intervention survey to measure the impact of our program. This Likert-type questionnaire was developed to evaluate health professionals' training efforts,12 and it uses 70 statements with seven-point-scaled responses from “strongly disagree” to “strongly agree.” This tool has 13 scales: perceived ability to identify and document domestic violence, specifically intimate partner violence; self-reported screening behavior; workplace resources; making referrals; provider self-efficacy; health care role; perceptions concerning the need for victim autonomy; understanding the victim's circumstances; understanding and compliance with legal requirements; staff preparation; limitations to assistance when the patient does not disclose; the belief that there is no need for training; and feeling either too busy or otherwise unable to help.13
The pre-intervention survey of KABB was conducted in the summer through the winter of 1997. It was administered before the training sessions or at administrative meetings. Following psychometric testing, the first survey was shortened to 47 items, and the post-intervention survey was conducted two years after the first as a mailed survey or distributed at administrative meetings. Both of the surveys were self-administered, anonymous, and could be matched by using the respondents' self-administered code numbers.
Based on an unpublished study,13 maximum-likelihood-factor analysis with oblique rotation was used to test the psychometric properties of the CDC survey. Alpha coefficients were used to determine the internal consistency reliability within each scale. The total scale and the subscales had acceptable reliability (alphas ranged from 0.52 for limitations to 0.80 for self-efficacy). The shortened instrument also had high validity as measured by the Rand coefficient (.92).14
Record audits. At the end of the intervention, project staff used a chart-abstraction form to review the medical records of all women who had positive domestic violence screens administered by an ED nurse to assess whether the cases of domestic violence were managed appropriately. The results of the public health campaign are reported elsewhere.15
Analysis. Statistical analyses were performed using standard statistical software packages. Bivariate analysis included calculation of odds ratios with 95% confidence intervals to test the strength of the associations between certain KABB variables (e.g., receipt of domestic violence training and identification of a victim in the preceding year). Scale scores were obtained by summing responses to individual items within the scale. Not all respondents had matching identification codes at pre- and post-intervention surveys. Therefore, to be conservative, rather than pooling all participants, multiple analysis of variance (MANOVA) tests were run using factors for both time and group, where group categories involved the matched set of participants versus participants that could not be matched from pre- to post-intervention survey results.
Additionally, multiple regression analyses were run to consider the effects of demographic and experiential variables (age, sex, number of years working in the health care field, percentage of full-time work status, and prior domestic violence training experiences, including having read domestic violence information or seen a video, having received brief training, having received several brief trainings, or having received the equivalent of a semester of training or more) on post-intervention KABB scores.
A total of 709 health care professionals received training for a total of two to three hours. Their professional categories are summarized in Table 1. County-level or rural-health-network—level denominators were not available for all the professional categories that were trained.
The pre-intervention survey received 380 responses (67%). The respondents were mostly women, full-time employees (see Table 2). A total of 38% of the respondents had received some past training relative to domestic violence. Overall, 36% had identified a victim in the preceding year. Physicians and physician assistants reported the same rates of victim identification in the preceding year as did nursing staff, nurse practitioners, and nurse midwives (42%). Those without prior training were less likely to have identified a victim within the year (OR = 0.19; 95% CI 0.11–0.33).
Two scales were correlated in the baseline survey: self-efficacy (feelings of comfort and capability related to responding to domestic violence) and familiarity with referral resources. These two dimensions are relevant for programmatic and intervention efforts. Health care providers' feelings of self-efficacy in responding to domestic violence victims may be closely related to their abilities to offer the victim access to additional resources that they know respond to her needs.
There were 273 respondents (56%) to the post-intervention survey. These respondents were demographically similar to those responding to the pre-intervention survey (see Table 2). Of the respondents, 32% had attended a brief training session, and 43% had attended several brief training sessions, grand rounds, or an inservice. Thirty-four percent had read the recommended protocol, 25% had watched a video on domestic violence for training purposes, and 25% had read and researched the topic on their own. Only 4% of the respondents reported never receiving any training.
Among all staff, the reporting of victim identification within the preceding year increased from 36% to 39%. Among physicians and physician assistants, 51% reported identifying a victim within the year, compared with a pre-intervention rate of 42%. Among the nursing staff, the victim identification rate was 40%, compared with 42% before the intervention. Those with prior training were more likely to have identified a victim within the year (OR = 2.86; 95% CI 1.73–4.74). Of those who had identified a victim, 62% reported that they had counseled the patient about community service options, 39% had helped the patient develop a personal safety plan, and 46% had referred the patient to a domestic violence advocate or social worker.
Combined KABB Results
Multiple analysis-of-variance tests were performed using all participants with complete data (n = 232). Participants were grouped according to those with matched pre-to-post surveys and those that could not be matched. The results demonstrated significant overall improvement across all scales for both groups (Wilks' lambda = 0.865, F = 5.380, df = 13, p < .001). Considered separately, nine of the 13 KABB scales revealed significant increases from pre- to post-intervention surveys for screening behavior, workplace resources, making referrals, provider self-efficacy, perceptions concerning the need for victim autonomy, understanding the victim's circumstances, understanding and compliance with legal requirements, staff preparation, and feeling too busy or otherwise unable to help (see Table 3). Among these significant findings, results indicated a significant group effect on one scale—understanding the victim's circumstances, and significant interactions between time and group on two scales—making referrals and staff preparation. For understanding the victim's circumstances, the group with matched surveys had higher scores on both pre- and post-intervention responses than did the group whose surveys could not be matched. For making referrals and staff preparation, the matched group began with lower scores and ended with higher scores than the unmatched group.
Stepwise multiple regression was conducted on the responses of the 232 participants with complete data to identify demographic and experiential predictors of the total outcome score (sum of scale scores). Even at the 85% level of significance, only 10.24% of the variance in the outcome was accounted for. Four of the eight variables listed above were in the final model: number of years working in the health care field, age, having received several brief domestic violence training sessions, and (to a lesser extent) having read domestic violence information or seen a video (see Table 4).
A consecutive sample of 271 medical records with positive domestic violence screens were reviewed at two EDs. Of these, the presenting problems were directly related to domestic violence (18%) or suspected to be (8.5%). Among these 72 cases, 86% had documented physical assessments, 62% had domestic violence documented in the narratives, and 37% had quotes included in the narratives. Nineteen percent had had photos taken, 42% had documented safety plans, and 35% had documented lethality assessments. Thirty percent had documented that a victim's rights statement had been given to the victim. Forty-two percent had documented referrals made, most often to police or domestic violence advocates.
This study showed significant changes in the knowledge, attitude, beliefs, and behaviors of health care providers following a multifaceted intervention. However, it is difficult to attribute the changes observed to any one component of the intervention or to any of the demographic characteristics or experiences for which data were collected. In addition, how these self-reported changes relate to the outcomes experienced by victims interacting with the health system is unknown. While our medical record audit documented some improvement in management in the ED setting, the outcomes of these cases could be determined only by longitudinal follow-up studies of the victims. More subtle changes in health care providers' behaviors or public attitudes may have occurred but not been captured by self-reported measures or the medical record audit.
Given the demands on their time that health care providers face, as well as competing training needs, it is important to achieve the most parsimonious mix of content issues when designing training interventions. The content and length of training we used are similar to those used in other studies.16,17 In Thompson's study,16 the dynamics of domestic violence, the identification and management of victims, and the availability of internal and community resources, as well as a discussion of legal issues, were the training components primary health care providers rated most highly nine months after training. Our project incorporated these elements as well as attempted to facilitate community referral by bringing providers “face to face” with domestic violence advocates. Strong links to domestic violence advocates and other referral sources both inside and outside an organization should strengthen a program.
Other studies have addressed different types of training programs for dealing with domestic violence. Saunders17 suggests that more intensive training that uses patient simulations may be more effective in long-term improvement of providers' communication skills and management of domestic abuse cases. The Massachusetts Medical Society has produced a CD-ROM that includes virtual interviews that electronically give the trainee feedback about correct or incorrect questions and responses during the interview.18
Training programs that address batterers are another approach. In our study, during training and public speaking events, questions about batterers would arise that speakers had to be prepared to address. The paucity of batterer programs in rural areas as well as the lack of efficacy of these programs preclude a comprehensive approach to domestic violence training that includes batterer interventions.
We used a four-stage model (awareness, adoption, implementation, and institutionalization) based on organizational change theory to create and sustain changes in domestic violence management within this rural network. Such a comprehensive intervention is probably cost-effective, however, given that physical and emotional abuse is associated with increased health care use19 and that health care costs for domestic violence victims are 92% higher than those for controls matched for age and gender.20 For example, in a general health plan in Minnesota, the mean cost of health care for domestic violence victims was found to be $3,635 in 1994.20 Given an average prevalence of 25% for domestic violence among women,21 a health care plan serving 25,000 women could expect to pay a cumulative health care cost of $22 million (in 1994 dollars). The training costs for this project were approximately $75,000 per year for two years and the public health campaign cost approximately $30,000. While many assumptions in this cost comparison can be challenged, and much more detailed studies are needed to substantiate the cost savings potential of a domestic violence training program, it is helpful to consider such preventive program costs as potentially averted health care costs.
The relatively small change in the rates of victim identification between our pre- and post-intervention surveys may reflect the considerable time (two years) between training and measurement, the degree of training and experience needed, the time required to change providers' behaviors, and avoidance of this sensitive topic. Our study found that the clinicians' rate of victim identification in the last year following the intervention, 50%, was lower than that reported for urban areas, (e.g., 70% among clinicians in Seattle5).
The low response rates to the surveys may be a function of the length of the questionnaire, the sensitivity of the topic, or other health care provider constraints, such as having to provide patient care at the time the surveys were administered. The baseline response rate was similar to that of Rodriquez's survey of physicians in California (69%) that used three mailings.7 Our survey differed in that it included other disciplines and the questionnaire was distributed only once. In addition, the network staff turnover was 52% during the two-year interval between the two surveys (however, this includes all staff and is not limited to primary care). The high rates of turnover of health care staff in rural areas combined with downsizing has limited our ability to track personnel and their changes in KABB over time. In addition, the self-assigned coding of the KABB forms was problematic in terms of reliability, so another method needs to be devised. Because staff felt that the 47-item post-intervention survey was still too long, shortening it to a five-minute assessment may increase its use as a practical evaluation tool.
Universal screening in primary care settings was found to be unacceptable to most health care providers despite the fact that domestic violence is more common than hypertension, thyroid disease, and colon cancer.22 The learning curve among health professionals for accepting and adopting new practices tend to be gradual, and the sensitive nature of this issue slows its acceptance even more. Given its overlap with criminal justice, domestic violence screening may need to be legislated by state-level departments of health in order to speed its integration.
Physician—administrative coding of adult maltreatment using the ICD-9CM code for adult maltreatment (995.80) rarely occurs. Insurance verification practices, i.e., mailing notification of medical visits and the reason for the visit to the home, may compromise victim confidentiality. For this reason, we did not encourage the use of ICD-9 CM coding for domestic violence. This led to mixed messages to providers to document domestic violence in the medical note, but not code domestic violence as a reason for the visit. This problem obviously further complicates the ability to track changes in provider identification following any intervention, because ICD-9-CM codes cannot be used.
However, our findings do suggest that domestic violence training programs for health care providers in primary care can be effective in increasing awareness, improving case management, and increasing rates of referral. Integrating domestic violence management into a primary care system requires training, reinforcement, office enablers (small office practice modifications that facilitate patient—doctor interaction), public education, and specific links to community resources. Through this program, we attempted to bridge prevention and practice as well as to integrate a variety of disciplines and practice levels to address this public health problem.23 Thus, even with the documented impediments, domestic violence management and prevention programs can be integrated into health care network operations using a professional and public health education approach, and can be an important component of population-based management of a prevalent and serious public health problem.
On a more pessimistic note, the problem with domestic violence identification and management in health care settings may be symptomatic of deeper problems within health care delivery, including the health care provider's inability to communicate about sensitive issues, lack of mental health care capacity in primary care, lack of expertise in chronic disease management, lack of continuity of care and established primary care relationships, lack of accountability, and a medical system that responds poorly to social issues.
The solutions to domestic violence from the health care perspective may involve revamping a broad range of health system issues that adversely affect not only the management of domestic violence but other health problems as well. Addressing social determinants, improving communication abilities and training, and promoting the referral of victims to qualified professionals may enhance health care providers' capacity to interact more effectively with all patients, and may facilitate their role as multidisciplinary team members in the response to victims of domestic violence.
1. Sisley A, Jacobs LM, Poole G, Campbell S, Esposito T. Violence in America: a public health crisis—domestic violence. J Trauma Inj Infect Crit Care. 1999;46:1105–13.
2. Naumann P, Langford D, Torres S, Campbell J, Glass N. Women battering in primary care practice. Fam Pract. 1999;16:343–52.
3. Eisenstat SA, Brancroft L. Domestic violence. N Engl J Med. 1999;341:886–92.
4. Tilden V, Schmidt T, Limandri B, Chiodo G, Garland M, Loveless P. Factors that influence clinicians' assessment and management of family violence. Am J Public Health. 1994;84:628–33.
5. Sugg NK, Thompson RS, Thompson DC, Maiuro R, Rivara FP. Domestic violence and primary care: attitudes, practices and beliefs. Arch Fam Med. 1999;8:301–6.
6. Sugg N, Inui T. Opening Pandora's box: primary care physician's response to domestic violence. JAMA. 1992;267:3157–60.
7. Rodriquez MA, Bauer HM, McLoughlin E, Grumbach K. Screening and intervention for intimate partner abuse: practices and attitudes of primary care physicians. JAMA. 1999;282:468–74.
8. Derk S, Reese D. Rural healthcare providers' attitudes, practices and training experience regarding intimate partner violence—West Virginia, March 1997. MMWR. 1998;47(32):670–3.
9. Gerbert B, Abercombie P, Caspers N, Love C, Bronstone A. How health care providers help battered women: the survivor's perspective. Women and Health. 1999;29(3):115–35.
10. Warshaw C. Intimate partner abuse: developing a framework for change in medical education. Acad Med. 1997;72(1 suppl):S26–S37.
11. Soeken K, Parker B, McFarlane J, Lominak MC. The Abuse Assessment Screen; a clinical instrument to measure frequency, severity and perpetrator of abuse against women. In: Campbell J (ed). Beyond Diagnosis: Changing the Health Care Response to Battered Women and Their Children. Newbury Park, CA: Sage Publications, 1995.
12. Short LM, Cotton D, Hodgson CS. Evaluation of the module on domestic violence at the UCLA School of Medicine. Acad Med. 1997;72(Suppl 1):S72–S92.
13. Short LM, Fanslow JL, Gadomski A. Psychometric properties of a survey to assess health care provider knowledge, attitudes, beliefs and intended behaviors concerning intimate partner violence. Unpublished manuscript.
14. Rand W. Objective criteria for the evaluation of clustering methods. J Am Stat Assoc. 1971;66:846–50.
15. Gadomski A, Tripp M, Wolff DA, Lewis C, Jenkins P. Impact of a rural domestic violence prevention campaign. J Rural Health [in press].
16. Thompson TS, Meyer, BA, Smith-DiJulio K. et al. A training program to improve domestic violence identification and management in primary care: preliminary results. Violence and Victims. 1998;13:395–410.
17. Saunders DG, Kindy P. Predictors of physicians' responses to woman abuse: the role of gender, background and brief training. J Gen Intern Med. 1993;8:606–9.
18. Massachusetts Medical Society. Seminar Series on Domestic Violence. An Interactive Educational Curriculum for Physicians. Waltham, MA: Massachusetts Medical Society, 1997.
19. Sansone RA, Wiederman MW, Sansone LA. Health care utilization and history of trauma among women in a primary care setting. Violence and Victims. 1997;12:165–72.
20. Wisner CL, Gilmer TP, Saltzman LE, Zink TM. Intimate partner violence against women: do victims cost health plans more? J Fam Pract. 1999;48:439–43.
21. Tjaden P, Thoennes N. Prevalance, incidence and consequences of violence against women: findings from the National Violence against Women Survey. Washington, DC: U.S. Department of Justice: National Institute of Justice and Centers for Disease Control, 1998.
22. El-Bayoumi G, Borum ML, Haywood Y. Domestic violence in women. Med Clin North Am. 1998;82:391–401.
23. Rosenberg ML, Fenley MA, Johnson D, Short L. Bridging prevention and practice: public health and family violence. Acad Med. 1997;72(1 suppl):S13–S18.