SINCE Stark and colleagues 1. An Evaluation of a System-change Training Model to Improve Emergency Department Response to Battered Women

ACADEMIC EMERGENCY MEDICINE February 2001, Volume 8, Number 2 131 An Evaluation of a System-change Training Model to Improve Emergency Department Response to Battered Women JACQUELYN C. CAMPBELL, PHD, RN, JEFFREY H. COBEN, MD, ELIZABETH MCLOUGHLIN, SCD, STEPHEN DEARWATER, MS, GREGORY NAH, MA, NANCY GLASS, MPH, MSN, RN, DEBBIE LEE, NANCY DURBOROW, MS Abstract. Objectives: To evaluate a system-change model of training from the Family Violence Prevention Fund and the Pennsylvania Coalition Against Domestic Violence for improving the effectiveness of emergency department (ED) response to intimate partner violence (IPV). Methods: An experimental design with outcomes measured at baseline, 9 12, and 18 24 months post-intervention. Twelve hospitals in Pennsylvania and California with 20,000 40,000 annual ED visits were randomly selected and randomly assigned to experimental and control conditions. Emergency department teams (physician, nurse, social worker) from each experimental hospital and a local domestic violence advocate participated in a two-day didactic information and team planning intervention. Results: The experimental hospitals were significantly higher than the control hospitals on a staff knowledge and attitude measure (F = 5.57, p = 0.019), on all components of the culture of the ED system-change indicator (F = 5.72, p = 0.04), and in patient satisfaction (F = 15.43, p < 0.001) after the intervention. There was no significant difference in the identification rates of battered women (F = 0.411, p = 0.52) (although the linear comparison was in the expected direction) in the medical records of the experimental and control hospitals. Conclusions: A system-change model of IPV ED training was effective in improving staff attitudes and knowledge about battered women and in protocols and staff training, as well as patient information and satisfaction. However, change in actual clinical practice was more difficult to achieve and may be influenced by institutional policy. Key words: domestic violence; model; system; training; intimate partner violence; battered women; public health. ACADEMIC EMERGENCY MEDICINE 2001; 8:131 138 SINCE Stark and colleagues 1 class study of battered women in the emergency department (ED), several studies have documented significant proportions of female patients in the ED as abused. Intimate partner violence (IPV) is a major cause of both injury and non-injury visits to the ED by women. 2 5 As reported in a prior article, the prevalence of IPV in this sample of women aged 18 years and older was 2.2% for acute trauma from abuse (15% for past-year physical or sexual IPV From Johns Hopkins University, School of Nursing (JCC, NG), Baltimore, MD; Department of Emergency Medicine, Center for Violence and Injury Control, Allegheny General Hospital (JHC), Pittsburgh, PA; San Francisco General Hospital (EM, GN), San Francisco, CA; Children s National Medical Center, Washington, DC (SD); Family Violence Prevention Fund (DL), San Francisco, CA; and Pennsylvania Coalition Against Domestic Violence (ND), Pittsburgh, PA. Received September 17, 1999; revisions received April 8, 2000, and August 23, 2000; accepted August 25, 2000. Supported by Centers for Disease Control and Prevention R49 CCR310553. Address for correspondence and reprints: Jacquelyn C. Campbell, PhD, RN, FAAN, Johns Hopkins University, School of Nursing, 525 North Wolfe Street, Baltimore, MD 21205-2110. Fax: 410-614-8285; e-mail: jcampbel@jhmi.edu and 36% for lifetime emotional or physical abuse). 3 Health care providers have recognized the need to improve the effectiveness of their response to battered women who present to the ED. 6,7 Early studies conducted in large urban hospitals demonstrated a significant increase in staff identification of battered women after training. 5,8 In more recent evaluation studies, primary care clinicians who participated in a Family Violence Prevention Fund (FVPF) one-day training program were more likely to report carrying out the recommended treatment actions, feeling more informed about IPV, and feeling confident in their ability to screen than participants in a comparison group. 9 Olson and colleagues 10 compared baseline abuse identification rates with those after a one-hour staff educational lecture and the addition of an IPV chart prompt. The proportion of cases identified during the intervention months was 1.8 times higher than that during the baseline month, but the chart prompt rather than the lecture was the differing factor. 10 The abuse identification rate in a primary care setting rose from 0% using discretionary inquiry alone to 11.6% when the health history form

132 BATTERED WOMEN Campbell et al. ED RESPONSE TO BATTERED WOMEN included a question on lifetime abuse. 11 The Abuse Assessment Screen (AAS) detected significantly more violence in the prior year (15% vs 3%) and during the current pregnancy (10% vs 1%) than a routine social service evaluation in the same prenatal care agency. 12 Changing forms to include the questions from the AAS and thus systematizing routine assessment significantly increased the prevalence of IPV detected and provider satisfaction in another prenatal care setting. 13 These studies suggest that a change in documentation forms to include explicit questions about abuse is important in changing practice. The perceptions of battered women about health care system interventions have been elicited in at least three small studies. All three research teams 11 14 reported that battered women experienced the majority of providers as appearing uninterested, uncaring or uncomfortable. 15 In a survey of 90 primarily African American women, racism and discrimination because of poverty were perceived as significant barriers to seeking help. 14 The structure of the health care system was perceived to be the barrier to seeking and receiving health care in another small qualitative analysis of data from battered women. 15 One woman stated that it is not part of their job, they are there to fix injuries. The few published evaluations of the implementation of training for ED health care professionals 8 and/or the implementation of new domestic violence protocols for EDs 5,8 10 have been single-hospital studies and, except for one, 8 lack a true experimental design. The majority of research on abused women in EDs has been conducted in large urban teaching and/or trauma medical centers. The purpose of this study was to evaluate the effectiveness over time of an intervention designed for mid-sized hospital ED staff to improve both staff and institutional responses to battered women, as measured by medical record review, staff and patient surveys, and institutional observation along with a post-training staff interview process evaluation. The following four hypotheses were tested by comparing outcomes measured preand post-intervention in experimental and control hospitals. Hypothesis 1: There is a difference in the ED culture about IPV in experimental EDs compared with control EDs. Hypothesis 2: There is a difference in ED personnel knowledge of IPV and attitudes toward battered women in experimental EDs compared with control EDS. Hypothesis 3: There is a difference in the proportion of women identified as battered in experimental EDs compared with control EDs. Hypothesis 4: There is a difference in female patients satisfaction with ED care in experimental EDs compared with control EDs. METHODS Study Design. This evaluation used an experimental design with baseline (pretest), immediate (9 12 months), and long-term (18 24 months) post-assessments in 12 mid-sized hospital EDs randomly assigned to experimental and control groups. Individual hospitals and coordinating institutions in Pennsylvania and California received approval for performing this study from their respective human subjects review boards. Study Setting and Population. The 12 hospitals were randomly selected from a list of 39 hospitals within a 100-mile radius of Pittsburgh, PA, and San Francisco, CA, that had previously indicated they would be interested in training on IPV. None of the 12 hospitals EDs reported prior IPV training. All female patients aged 18 years and older who entered the EDs during monitored shifts were entered on the study intake sheets and were invited to participate in an anonymous survey on patient satisfaction and safety. The ED staff (physician and nurse) at the 12 participating EDs completed a self-report measure of knowledge of IPV and attitudes towards battered women pre- and post-ipv training. Study Protocol. The intervention was a two-day training and planning program 17 designed and implemented by the FVPF and the Pennsylvania Coalition Against Domestic Violence (PCADV). The model used a team training approach that addresses systems change and coalition building as well as provider attitudes and skill building. 17 Each of the EDs was asked to identify a team consisting of a physician, nurse, social worker, and administrator (hospital or ED) who were interested in the issue and willing to attend the two-day training. A representative from a local domestic violence service organization joined each team of hospital personnel. The teams were asked to have one meeting before the training and to meet soon afterwards to further develop and implement the action plan made at the session. All ED staff were unaware of the hospital s treatment status. The first day of the training session consisted of a half-day didactic instruction on the background dynamics of IPV and the appropriate ED response, including assessment, documentation, and referral. The second half of day 1 was devoted to role-playing of assessment and intervention, legal mandates, and the beginning of team planning. The second day was dedicated to the system change portion of the training, with the team working to developing a written action plan that included planning appropriate protocols for the

ACADEMIC EMERGENCY MEDICINE February 2001, Volume 8, Number 2 133 setting, how to organize the adoption of the protocols, training of the ED staff, and securing administrative support. Cultural sensitivity and culturally competent care were stressed throughout the training. The FVPF and PCADV employees were available for telephone technical assistance during the implementation of the action plans but did not initiate contact with the hospitals.* Measures. The culture of the ED (hypothesis 1) was measured by the multiple observable outcomes of: 1) appropriate protocols [met Joint Commission on Accreditation of Healthcare Organizations (JCAHO) standards for definition, assessment, documentation, referral, and clinical staff responsibility], 2) materials related to IPV present in the ED [IPV brochures and posters displayed, assessment (intake) forms that included questions on IPV, medical record intervention checklists and referral information on IPV available to staff], and 3) staff training (regular, mandated, and ongoing). A total score of 16 points was possible on the culture measure (7 points for appropriate protocols, 5 points for appropriate materials, and 4 points for appropriate training). The nurse manager from each ED submitted the written protocols both preand post-intervention and was interviewed by the research team regarding staff training, IPV educational materials, and intervention checklists in medical records. Observational measures of the culture were also obtained annually in each ED by the site coordinators in Pennsylvania and California. The culture measure was administered at the six experimental and six control hospitals in each state at baseline, 9 12 months, and 18 24 months post-intervention. The staff knowledge and attitude survey (SAS, or Staff Attitudinal Survey) is a self-report survey developed to address the ED staff attitude and knowledge outcome (hypothesis 2). This anonymous survey was administered twice to a total 649 ED staff (physicians and nurses), at baseline (n = 336), and at the conclusion (n = 313) of the study. The surveys were distributed and collected by the ED nurse manager. The SAS consisted of five core items from Saunders and associates 18 that has published support for reliability and validity to measure staff attitudes toward battered women. The SAS also contained 12 items covering critical knowledge domains and ten items addressing common myths about IPV from the training manual and agreed upon by both research and partner organization staff. Four questions were deleted after *The basic outline of the training and materials used are available in a manual from the National Health Resource Center on Domestic Violence (1-888-792-2873). administration because of their ambiguity, leaving a 23-item Likert scale instrument with an internal consistency of 0.73. Higher scores (92 total possible score) represented less blaming attitudes toward battered women and more knowledge about IPV and health care professionals roles in addressing IPV. The rate of identification of battered women (hypothesis 3) was determined by the ratio of the number of women who were identified as abused during the medical record review to the number of women who self-reported IPV on an anonymous patient survey. These abuse rates were then compared by hospital treatment status. The prevalence of IPV by patient self-report was determined by the anonymous Patient Safety Satisfaction Survey (PSSS), which was administered to female ED patients aged 18 years and older who agreed to take part in the survey and met the study criteria. The PSSS is an adaptation of the AAS, 19 a screen used in several other ED studies with established validity and test retest reliability. The three waves of survey administration covered 309 shifts between Friday and Tuesday. Eight- to 12-hour shifts were covered depending on the pattern of staffing. These shifts were chosen because they traditionally represent the periods during the week with highest patient volume. A more detailed description of the methods of PSSS administration has been previously reported. 3 Medical records were abstracted for every female patient in the ED during the monitored shifts. The medical record abstraction provided information on patient demographics, discharge diagnoses, description of abuse, and patient disposition in order to count all female patients who were identified as being abused by an intimate partner or ex-intimate partner by any of the ED personnel during the course of the visit. The medical record review was conducted by research team members using a standardized review form across all participating hospitals. The PSSS had three additional questions that addressed the patients satisfaction with ED care they received that day. The satisfaction questions were 1) Did the ED staff seem concerned about you? 2) Were you satisfied with your care at the ED? 3) Would you come back to this ED again if you had a medical problem? Hypothesis 4 assumed that female patients at the experimental EDs would express more satisfaction with ED care than women in the control EDs. A qualitative interview was conducted at the conclusion of the study with the personnel from the experimental EDs who attended the training and were responsible for implementation at their ED. It was hoped that this interview would elucidate the facilitators and barriers each hospital encountered toward instituting comprehensive training and provide feedback for potential modifications to

134 BATTERED WOMEN Campbell et al. ED RESPONSE TO BATTERED WOMEN Figure 1. Observational measure of emergency department (ED) culture by treatment and wave of administration, Pennsylvania and California community hospital EDs, 1995 97. the training. The interview consisted of 15 oral open-ended questions designed to address the trainees perceptions of the program and assess their success in implementing an IPV training program. Data Analysis. A power analysis ( = 0.05, = 0.90) was conducted, and a sample of 300 women per treatment group per wave of administration would be needed to detect a significant effect of the intervention on rate of identification of battered women. We had estimated that 10 15% of our sample would present to the ED because of acute trauma from IPV; therefore, we projected that approximately 300 acutely abused women would present to the ED during the study period. SPSS statistical software (SPSS Inc., Chicago, IL) was used for the hypotheses testing. Analysis of covariance (ANCOVA) was used to test hypothesis 1 following an analysis for homogeneity of regression slope, while analysis of variance (ANOVA) was used to test hypotheses 2, 3, and 4. In addition, a process evaluation consisting of anonymous telephone interviews by research staff to 16 (84%) of the 19 ED staff members at the experimental hospitals who underwent training was conducted. RESULTS Six of seven Pennsylvania hospitals and six of nine California hospitals that were initially contacted agreed to participate in the study. The three California hospitals that refused to participate had recently merged with other hospitals or were no longer interested when contacted. The lone Pennsylvania hospital that refused to participate expressed concern about discussing such a sensitive topic with patients who might be their neighbors. Only one of the experimental hospitals actually sent a complete team as called for in the system change model, consisting of a physician, nurse, social worker, and hospital administrator (nurse managers attended from three hospitals), to the two-day training. Two hospitals did not send a physician, and social workers were sent from five of the six hospitals. All but one (94%) of the health care professionals who attended the original training and were interviewed for the process evaluation thought that the training program was a positive and worthwhile experience that would enable them to implement a screening program to increase identification of battered women in their EDs. Two of the three Pennsylvania experimental hospitals did implement some form of routine screening, but this took between seven months and one year (average ten months) after the original training to accomplish. Yet our first outcome measurement (time 2) was nine to 12 months following the training. The results from the hypotheses testing were as follows: The assumption of homogeneity of regression slope was not violated, with no significant interaction effect between the covariate (baseline culture score) and treatment (F = 0.174, p = 0.688); therefore, the analysis of hypothesis 1 proceeded using ANCOVA. The experimental EDs had significantly higher scores on the summary score of the culture criteria at time 2 and time 3 than the control EDs when controlling for the culture score at time 1 (Fig. 1). There was no significant effect of the intervention by state in the analysis. For hypothesis 2, a total sample of 649 ED personnel (336 pretest, 313 post) was used for the analysis (75% overall response rate), with 330 nurses and physicians from the experimental EDs completing the instrument and 319 from the control EDs. This instrument was administered at pre- and post (18 24 months)-intervention only. The ED personnel in the experimental hospitals [pre mean = 64.4 (95% CI = 63.1 to 66.6); post mean = 69.0 (95% CI = 68.6 to 70.0)] had significantly higher (F = 5.57, p = 0.019) scores on the SAS than did those in the control hospitals [pre mean = 68.0 (95% CI = 67.0 to 69.0); post mean = 68.1 (95% CI = 67.0 to 69.4)]. Both gender and state had significant main effects on the SAS, with both pre- and post-training differences and no interaction effect. Females scored significantly higher than males on the survey [female mean = 68.1 (95% CI = 67.5 to 68.7); male mean = 66.5 (95% CI = 65.4 to 67.6); F = 6.5, p = 0.011], and California personnel scored significantly higher than those in Pennsylvania hospitals [CA mean = 68.3 (95% CI = 67.6 to 68.9); PA mean = 66.3 (95% CI = 65.3 to 67.3); F = 10.5, p = 0.001]. Hypothesis 3 addressed prevalence of self-identified battered women documented in the medical

ACADEMIC EMERGENCY MEDICINE February 2001, Volume 8, Number 2 135 records. The ratio of women documented as abused in the medical record (n = 40) to those who selfreported themselves as presenting with acute trauma (n = 83) from abuse on the PSSS in experimental vs control hospitals was nonsignificant (F = 0.411, p = 0.52). Although lacking the statistical power to show a significant difference, the trajectory of change for the analysis addressing hypothesis 4 was in the expected direction. There was an increase in the experimental hospitals from 27% at baseline to 55% at time 2, leveling off to 53% at time 3, while in the control hospitals, the percentages decreased from 73% at baseline to 47% and 25% at times 2 and 3, respectively (Fig. 2). Self-reporting patient satisfaction was not affected by baseline differences. The women treated in the experimental hospitals reported significantly more satisfaction [experimental mean = 95% (95% CI = 93% to 97%); control mean = 90% (95% CI = 88% to 92%); F = 15.43, p < 0.001] with their care than the women treated at the control hospitals. The process evaluation was completed with 19 ED staff members who attended the training. The trainees generally agreed that a combination of factors was crucial: 1) strong staff support and ED director with vested interest in IPV intervention; 2) administrative backing to incorporate the protocol into the ED policies and procedures; and 3) development of charting prompts on ED face sheet to facilitate screening. Barriers to implementation included: 1) not enough mentoring to ensure timely adherence to action plan; 2) the six-month implementation goal was too rapid; and 3) time and money are significant factors at smaller community hospitals with respect to providing ongoing inservice training and modifying existing medical records to include charting prompts. DISCUSSION The results of this study indicate that a model of staff training that emphasizes system change can be effective in improving the IPV-related culture of the ED and provider attitudes toward survivors of domestic violence. The experimental hospitals in this evaluation significantly improved the culture of their EDs. Culture was defined as having protocols regarding IPV and having IPV information available both for battered women in terms of brochures and relevant posters, and for staff in terms of intervention check lists, and regular, routine and mandated IPV training. The experimental hospitals also improved in their identification of women injured from abuse, although the difference was not significant. Importantly, all of these changes were maintained over time. This model of Figure 2. Ratio of the prevalence of self-identified battered women on an anonymous survey (n = 83) to the prevalence of women documented as abused in the medical record (n = 40) by treatment and wave of survey administration, California and Pennsylvania hospitals, 1995 97. training is innovative in several important ways: 1) the interdisciplinary team of both shelter and key ED personnel attending the training, 2) the emphasis on system change throughout the training, and 3) the inclusion of significant time in the team training sessions for facilitated system change planning with technical assistance available after the training. Universal Screening. The evaluation also demonstrated that the symptom-based assessment for abuse that was taught in this training and implemented by the majority of U.S. hospital EDs results in low rates of identification of women who are abused but do not present with acute trauma. Assessment for IPV in the women presenting with trauma related to abuse reached 50% in the experimental hospitals, not ideal but better than prior to training and better than the control hospitals. Yet the majority (76%) of the women who self-reported physical assault by a partner within the preceding year did not present with trauma at this ED visit, and only 9% of the women overall were asked about abuse by ED staff, even after training. This is the rationale behind recommendations for routine universal screening of all women in EDs rather than only assessing women presenting with trauma or other obvious symptoms of IPV. 20 At the same time, there is continued uncertainty about the value of universal routine screening, because it has yet to be directly demonstrated that identification and subsequent referral in the ED result in better health and/or safety outcomes for battered women. 8,21,22 This kind of research has been conducted in prenatal care settings, but is urgently needed in other settings. 21 Data from other studies 11 13 as well as this eval-

136 BATTERED WOMEN Campbell et al. ED RESPONSE TO BATTERED WOMEN uation suggest that routine screening is dependent on forms being changed, and that form revisions take time for clearance with various hospital committees and executives. Financial constraints among some of the hospitals studied here necessitated completely using current forms before implementing new ones. Some of the experimental hospitals still had not been able to get new forms printed that reflected the inclusion of domestic violence screening questions by the second follow-up (18 months post-training). Other hospitals have addressed this problem by using a stamp to insert the questions on old forms before new ones are printed. 13 Another issue that may have impeded the extent of improvement in identification achieved is that only one of our hospitals sent a hospital administrator to the training. As demonstrated by the staff attitudinal survey, study ED personnel report positive attitudes toward battered women; however, positive attitudes have not translated into a change in clinical practice and improvement in identification of battered women. Staff turnover is one potential limitation to implementation of a system change model to improve clinical practice. However, the process evaluation included the observations that increased resources and strong and continued administration support were necessary to effect lasting system change. This has been supported in other research 22 and is optimally begun by including administration in the team training, as was originally designed. LIMITATIONS AND FUTURE QUESTIONS There were several historical threats to internal validity in the study that probably contributed to a higher rate of identification of battered women at baseline than at time 2 and time 3 in the control hospitals, and may have differentially affected the California hospitals in their pattern of documentation. The Simpson trial was taking place during baseline measurement and initial training, which significantly increased attention to and changed attitudes about domestic violence all over the country, 23 perhaps even more so in California. Also in California, two laws were implemented in 1995, the same year of the training, which were directly aimed at the management of domestic violence in EDs. California Law AB 1652 mandated reporting of domestic violence to criminal justice authorities, with a $1,000 fine for noncompliance; and AB 980 mandated training and the development of protocols for screening and documentation in hospital EDs (no sanctions attached). Although there was insufficient sample size for significance, California personnel in both experimental and control hospitals were less likely than those in Pennsylvania to document abuse in the medical record. The mandatory reporting provision is resisted by health care providers and policy makers, 21,24 26 and medical record documentation may have been assumed to result in (or be accompanied by) women being reported to the criminal justice system even if they didn t want this action. California providers may have also been concerned that recording without reporting would result in legal sanctions. Thus their practice may have been to appropriately assess but not to record these actions, yet our major measure of improvement was through the medical record. The medical record reviewers in each state were not blinded to the hospital s treatment status, but had no knowledge of an individual woman s responses to acute abuse on the PSSS. The FVPF was also conducting other training in various health care settings in California at approximately the same time as this intervention that may have attracted interested providers from both the experimental and control hospitals; this was not occurring in western Pennsylvania. Differential effects in California were supported by the finding that California staff had significantly more positive scores on the knowledge and attitude survey (SAS) in both experimental and control hospitals pre- and post-training. The gender differences (with women having higher scores) have been found in most other studies as well as in national attitude data. 25 An additional historical threat was that 9% of the 398 hospitals in California merged or closed their EDs during the study. The consequent staff turnover and uncertainty in our targeted hospitals may have negatively influenced the ability of the research team to collect data in California and resulted in a lower than optimal response rate of providers on the SAS especially at post-intervention (18 24 months) (83% in PA; 65 in CA). Finally, the JCAHO standards regarding domestic violence were strengthened in each year from 1992 to 1995, perhaps resulting in hospitals differentially paying attention to domestic violence as their accreditation was up for renewal. Another serious issue in this study was the small number of women who presented to the ED with acute trauma from abuse. We had decided to make the patient surveys totally anonymous to increase response rate to this survey on a sensitive topic. We were especially sensitive to the California mandatory reporting laws and the possible concern this might raise in abused women if asked to respond to a survey without anonymity. In fact, our patient response rate was indeed high (only 11% refusal), which reflected the quality of our nurse data collectors, who were sensitive to when patients could be approached to complete the surveys (e.g., after pain medication administration, af-

ACADEMIC EMERGENCY MEDICINE February 2001, Volume 8, Number 2 137 ter initial procedures, after the sense of crisis had passed), as well as our anonymity decision. 3 However, this decision resulted in an inability to link the patient surveys with medical records directly. Therefore, we could not determine exactly which women who reported abuse on the PSSS were identified as battered in the medical record and had to rely on proportional comparisons. We had based our original estimates of power and, therefore, sample size on the literature available at the time of study design, which suggested as many as 18 30% (at least 10% prior to training and 20% post-training) of the women would be in the ED with trauma from abuse. 4,5 Unfortunately, our design included only women at the ED due to injuries from abuse, not women who were in the ED for other IPV-related health problems, for the final outcome analysis. Our sample size of 83 women who self-reported that they were in the ED because of acute trauma from abuse, and the 40 medical records on which domestic violence was documented, was too small to detect significance. We also originally hypothesized that better documentation and medical record indication of improved interventions would occur in the experimental hospitals, but we did not have enough power to detect significance and therefore were unable to test those hypotheses. According to the process evaluation, these hospitals also found it difficult to truly partner with local shelters without dedicated resources and priorities for the initiative on both sides. They also needed external prods and support (and/or internal administrative goal setting and legitimization) to get implementation started once they returned home from the training. This kind of kick start, as one of the interviewees described, could be provided by shelter personnel and/or internal administration, but it was essential that both these players (along with nurses and physicians) be part of the initial team for training in order to forge longterm relationships. These lessons learned from this evaluation are reflected in the revised model of training being implemented by the FVPF in its new ten-state initiative that will also be rigorously evaluated. Even so, in order to establish the rationale for routine screening and ED interventions for domestic violence, further research is needed to document that this kind of secondary prevention intervention can actually improve the long-term health and safety of battered women. CONCLUSIONS This evaluation demonstrated that a systemchange model of domestic violence training for ED personnel and advocate teams can be effective in improving attitudes and knowledge of nurses and physicians about battered women and can change the culture of the ED so that protocols, training regimens, and information available for battered women are significantly improved. These changes were achieved even with limited technical assistance, scarce resources, and inconsistent administration support. Yet even though the protocols of the experimental hospitals were significantly better than those of the controls, the study suggested that actual practice change is more difficult to achieve. In order to identify all battered women who present in the ED, universal rather than symptom (or trauma)-based screening is needed. Routine screening needs reminders such as form prompts and administration support to achieve change in forms and maintain change in practice. References 1. Stark E, Flitcraft A, Frazier W. Medicine and patriarchal violence: the social construction of a private event. Int J Health Serv. 1979; 9:461 93. 2. Abbott J, John R, Loziol-McLain J, Lowenstein S. Domestic violence against women: incidence and prevalence in an emergency department population. JAMA. 1995; 273:1763 7. 3. Dearwater S, Coben JH, Nah G, et al. Prevalence of domestic violence in women treated at community hospital emergency departments. JAMA. 1998; 280:433 8. 4. Goldberg WG, Tomlanovich MC. Domestic violence victims in the emergency department. JAMA. 1984; 251:3259 64. 5. McLeer SV, Anwar RAH. Education is not enough: a systems failure in protecting battered women. Ann Emerg Med. 1989; 18:651 3. 6. Department of Health and Human Services. Healthy People Two Thousand: national health promotion and disease prevention objective [abstract]. 1991, pp 225 48. Publication no. (PHS) 91-50212. 7. Stark E, Filtcraft A: Women at Risk: Domestic Violence and Women s Health, Thousand Oaks, CA: Sage Publishers, 1996. 8. Tilden VP, Shepard P. Increasing the rate of identification of battered women in an emergency department: use of a nursing protocol. Res Nurs Health. 1987; 10:209 15. 9. Gielen AC, Forney CK, Abede S. Evaluation of a primary care clinic training program: final report to the family violence prevention fund. Unpublished report, 1998. 10. Olson L, Anctil C, Fullerton L, Brillman J, Arbuckle J, Sklar D. Increasing emergency physicians recognition of domestic violence. Ann Emerg Med. 1996; 27:741 5. 11. Freund KM, Bak SM, Blackhall L. Identifying domestic violence in primary care practice. J Intern Med. 1996; 11:4 46. 12. Norton LB, Peipert JF, Zierler S, Lima B, Hume L. Battering in pregnancy: an assessment of two screening methods. Obstet Gynecol. 1995; 85:321 5. 13. Covington DL, Dalton VK, Diehl SJ, Wright BD, Piner JH. Improving detection of violence among pregnant adolescents. J Adolesc Health. 1997; 21:18 24. 14. Campbell JC, Pliska MJ, Taylor W, Sheridan D. Battered women s experiences in emergency departments: need for appropriate policy & procedures. J Emerg Nurs. 1994; 20:280 8. 15. Gerbert B, Johnston K, Caspers N, Bleecker T, Woods A, Rosenbaum A. Experiences of battered women in health care settings: a qualitative study. Women Health. 1996; 24:1 17. 16. McCauley J, Kern DE, Kolodner K, Derogatis LR, Boss EB. Relation of low-severity violence to women s health. J Intern Med. 1998; 13:687 91. 17. Nudelman J, Durburow N, Grambs M, Letellier P. Best Practices: Innovative Domestic Violence Programs in Health Care Settings. San Francisco, CA: Family Violence Prevention Fund, 1997.

138 BATTERED WOMEN Campbell et al. ED RESPONSE TO BATTERED WOMEN 18. Saunders DG, Lynch AB, Grayson M, Linz D. The inventory of beliefs about wife beating: the construction and initial validation of a measure of beliefs and attitudes. Violence Victims. 1987; 2(1):39 57. 19. 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, 1998, pp 195 203. 20. Family Violence Prevention Fund. Preventing Domestic Violence: Clinical Guidelines on Routine Screening. San Francisco, CA: FVPF, 1999. 21. Chalk R, King P. Violence in Families: Assessing Prevention and Treatment Programs. Washington, DC: National Academy Press, 1998. 22. Dienemann J, Trautman D, Shahan JB, et al. Developing a domestic violence program in an inner-city academic health center emergency department: the first 3 years. J Emerg Nurs. 1999; 25:110 5. 23. Klein E, Campbell JC, Soler E, Ghez M. Ending the Violence. Newbury Park, CA: Sage, 1997. 24. Campbell JC. Making the health care system an empowerment zone for battered women: health consequences, policy recommendations, introduction, and overview. In: Campbell JC (ed). Empowering Survivors of Abuse: Health Care for Battered Women and Their Children. Thousand Oaks, CA: Sage, 1998, pp 3 22. 25. Glass NE, Campbell JC. Mandatory reporting of intimate partner violence by health care professionals: a policy review. Nurs Outlook. 1998; 46:1 6. 26. Rodriguez MA, Quiroga SS, Bauer HM. Breaking the silence: battered women s perspectives on medical care. Arch Fam Med. 1996; 5:153 8.