3RESULTS

Key Question 1. Does Screening Asymptomatic Women in Health Care Settings for Current, Past, or Increased Risk for Intimate Partner Violence Reduce Exposure to Violence? Reduce Physical or Mental Harms or Mortality?

Summary. A large cluster RCT of the effectiveness of screening for IPV in clinical settings included 6,743 Canadian women randomized to screening or nonscreening groups.⁸⁷ The primary outcomes were exposure to abuse and quality of life in the 18 months after screening. Secondary outcomes included depression, PTSD, alcohol and drug abuse, global mental and physical health, and use of health and social services. Women in both the screened and nonscreened groups had reductions over time in IPV recurrence, PTSD symptoms, and alcohol problems, as well as improvements in scores for quality of life, depression, and mental health. These outcomes were not significantly different between women in the screened versus nonscreened groups.

Limitations of the trial, rated fair quality, include high loss to followup, and the women lost to followup had higher IPV scores and more risk factors for IPV than women retained in the study. Differences between screened and nonscreened groups were also compromised by the absence of a specific intervention, low number of screen-positive women actually having discussions about IPV with their clinicians during their clinic visits, and lack of differences between groups in accessing additional services during followup. Women randomized to the nonscreening group were provided with information cards of locally available resources for women with IPV, and underwent extensive questioning about IPV that could have increased their awareness and influenced outcomes of the trial.

Evidence. A large cluster RCT of the effectiveness of screening for IPV in clinical settings met inclusion criteria for Key Question 1 (Appendixes C1 and C2).⁸⁷ The trial included 6,743 English-speaking women between the ages of 18 and 64 years randomized to screening or nonscreening groups. The primary outcomes were exposure to abuse and quality of life in the 18 months after screening. Secondary outcomes included depression, PTSD, alcohol and drug abuse, global mental and physical health, and use of health and social services. Harms of screening were actively monitored.

Participants were recruited when they presented for a health care visit at one of 12 primary care, 11 acute care, and three obstetrics/gynecology clinic sites in Ontario, Canada. The units of randomization were days or shifts for sites using shifts, and the screening and nonscreening units were balanced across different times of the day and days of the week. Clinicians at all study sites received standardized training in responding to IPV prior to the beginning of the trial. All women had universal access to health care in accordance with local practice. Women were eligible for the trial if they were ages 18 to 64 years, had a male partner at some time during the previous 12 months, presented for their own health care visit, were able to separate themselves from anyone else accompanying them, lived within 120 km of the clinic site, could speak and read English, were not too ill to participate, and were able to provide informed consent.

On screening days, before seeing their clinicians for the intended health care visit, participants provided baseline information and self-completed the Women Abuse Screening Tool (WAST). The WAST is an eight-item instrument measuring physical, sexual, and emotional abuse in the last 12 months. A score of ≥4 was considered positive in the trial. Results of the WAST were provided to the clinicians prior to the health care visit for women with positive scores. Discussion of positive findings, referrals, or treatment was left to the discretion of the treating clinician according to usual practice. After their visits and regardless of their WAST scores, all women completed the Composite Abuse Scale (CAS). The CAS is a 30-item validated research instrument to measure IPV. A score of ≥7 indicates exposure to IPV. The same procedures were followed for nonscreening days, except that participants completed both the WAST and CAS at the end of the visit; however, clinicians could inquire about abuse during the clinic visit if there were indications to do so.

Women with positive scores on both the WAST and CAS in the screened and nonscreened groups were followed for 18 months. Interviewers blinded to group assignment met with participants within 14 days of the initial clinic visit for a baseline interview, and again at 6, 12, and 18 months. At followup, participants self-completed several instruments, including the CAS; World Health Organization (WHO) Quality of Life–BREF; Center for Epidemiologic Studies Depression Scale; SPAN (Startle, Physiological Arousal, Anger, and Numbness) instrument to measure PTSD; alcohol abuse and dependency tool (TWEAK); Drug Abuse Severity Test; Short-Form 12-Item Health Survey, Version 2; Consequences of Screening Tool (COST); and modified version of the Health and Social Service Utilization Questionnaire. Additional services included visits to physicians, nurses, psychologists, or social workers or use of crisis phone lines, sexual assault crisis centers, advocacy or counseling services, women’s shelters, or other type of services.

The 12-month prevalence of IPV at the initial clinic visit was 13 percent in the screened group and 12 percent in the nonscreened group. During the initial clinic visit, 44 percent of screened women and 8 percent of nonscreened women discussed IPV with their clinician. Women in both the screened and nonscreened groups had reductions over time in IPV recurrence, PTSD symptoms, and alcohol problems, and improvements in scores for quality of life, depression, and mental health.

Outcomes for women in the screened versus nonscreened groups at 18 months included statistically nonsignificant reduction in risk for IPV recurrence (OR, 0.82 [95% CI, 0.32–2.12]), more rapid improvement in quality of life (3.74 points higher [95% CI, 0.47–7.00]), and reduced depressive symptoms (−2.32 [95% CI, −4.61 to −0.03]). Results for quality of life and depression were not statistically significant using multiple imputation analysis, and other secondary outcomes were not different between groups. No measures of harm were associated with screening in women with or without IPV exposure.

The trial met criteria for fair quality because loss to followup was high (43 percent of screened and 41 percent of nonscreened participants). Importantly, women lost to followup had lower levels of education, higher scores on the WAST and CAS, and were more likely to be married compared with those retained in the trial. Women lost to followup in the screened group had the highest WAST and CAS scores among trial participants.

Significant differences in major outcomes between screened and nonscreened groups may have been attenuated for several reasons, including the absence of a specific intervention, low number of screen-positive women actually having discussions about IPV with their clinicians during their clinic visits, and lack of differences between groups in accessing additional services during followup. Most women in the trial had contact with at least one additional service during the course of the trial, despite their screening group, indicating that referral and access to services may be a generally accepted practice among women in a universal health care system. Also, in this trial, women randomized to the nonscreening group were provided with information cards of locally available resources for women with IPV—itself an IPV intervention in other studies. In addition, women in the nonscreening group underwent extensive questioning about IPV by completing the WAST, CAS, and other instruments over the 18 months of the trial. These experiences could increase their awareness of IPV in their own lives, affect their utilization of services, and influence outcomes of the trial creating a substantial Hawthorne effect (i.e., the phenomenon that study participants change their behavior as a result of being involved in the study).

Key Question 2. How Effective Are Screening Techniques in Identifying Asymptomatic Women With Current, Past, or Increased Risk for Intimate Partner Violence?

Summary. Fifteen studies that evaluated the diagnostic accuracy of 13 screening instruments for identifying IPV in asymptomatic adult women in health care settings met inclusion criteria. Most instruments assessed current and/or past abuse rather than risk factors for future abuse, including the Abuse Assessment Screen (AAS); Partner Violence Screen (PVS); Hurt, Insult, Threaten, Scream (HITS); WAST; Humiliation, Afraid, Rape, Kick (HARK); Ongoing Abuse Screen/Ongoing Violence Assessment Tool (OAS/OVAT); Slapped, Threatened, and Throw (STaT); Childhood Trauma Questionnaire–Short Form (CTQ-SF); Secure, Accepted, Family, Even, Talk Survey (SAFE-T); Parent Screening Questionnaire (PSQ); one personal safety question; and five items with nongraphic language.

Screening instruments demonstrating the highest sensitivity and specificity in the included diagnostic accuracy studies include the HITS (English version, 86 percent sensitivity and 99 percent specificity; Spanish version, 100 percent sensitivity and 86 percent specificity), OVAT (93 percent sensitivity, 86 percent specificity), STaT (89 percent sensitivity, 100 percent specificity), HARK (81 percent sensitivity, 95 percent specificity), modified CTQ-SF (85 percent sensitivity, 88 percent specificity), and WAST (88 percent sensitivity, 89 percent specificity). Positive responses on the PVS significantly predicted verbal aggression and violence during the 4 months after screening. Two instruments specifically evaluated in pediatric settings had relatively low sensitivity but high specificity (PSQ, 19 to 29 percent sensitivity and 91 to 93 percent specificity; Zink five questions, 40 percent sensitivity and 91 percent specificity). Sensitivity, specificity, and other measures varied between studies and across instruments.

Evidence. Fifteen studies that evaluated the diagnostic accuracy of 13 screening instruments for identifying IPV in asymptomatic adult women in health care settings met inclusion criteria (Tables 4 and 5, Appendixes C3 and C4). Most instruments assessed current and/or past abuse rather than risk factors for future abuse. Screening instruments included the AAS,⁸⁸ PVS,^89,90 HITS,⁹¹ WAST,^90–92 HARK,⁹³ OAS/OVAT,^94,95 StaT,^96,97 CTQ-SF,⁹⁸ SAFE-T,⁹⁹ PSQ,¹⁰⁰ one personal safety question,¹⁰¹ and five items with nongraphic language.¹⁰² Five instruments and their modifications were used as reference standards, including the Index of Spouse Abuse (ISA),^91,94,95,97 WAST,⁹¹ Conflict Tactics Scale (CTS),^{88,89,100–102} PVS,⁹⁹ and CAS.^90,92,93 Structured⁹⁸ and semi-structured interviews⁹⁶ were used as reference standards in two studies. Additional descriptions of these instruments are provided in Table 4.

Three studies met criteria for good quality,^90,92,98 and 12 for fair quality.^{88,89,93–96,97,99,100,101–103} The most common methodologic limitations of these studies include patient spectrum was too narrow or had limited applicability (e.g., single site, not community based, not United States or Canada); sampling method was not random or consecutive, or not described; referencence standard was not credible; or dissimilar groups at baseline, high rates of attrition, and lack of information regarding whether the reference standard was independently interpreted. Details of the quality ratings are described in Appendix C4.

Four systematic reviews of screening instruments for IPV were identified by the searches.^54,104–106 Reviews provided descriptive summaries of screening instruments; however, many studies were not included in this update because their publication dates predated 2003.^54,104–106 Also, inclusion criteria for individual studies differed from this update by including studies of men,^104,106 instruments designed to identify perpetrators,¹⁰⁶ studies that did not report screening test performance outcomes,^54,105,106 studies comparing instruments with unclear or inadequate reference standards,^54,105 or instruments intended for research settings.⁵⁴

Instruments Evaluated in Maternity, Emergency, or Primary Care Clinical Settings. The AAS, a four-item instrument that considers sexual coercion, lifetime abuse, current abuse, and abuse during pregnancy, was evaluated in two studies.^88,94 When compared with the Conflict Tactics Scale 2 (CTS2) among mothers of premature infants in maternity wards, sensitivity and specificity were 32 and 99 percent for minor violence, 61 and 98 percent for severe violence, and 32 and 99 percent for both.⁸⁸ In a prospective study comparing the AAS and the OAS/OVAT with the ISA as the reference standard in a population of men and women presenting to an emergency department, the AAS had a sensitivity and specificity of 93 and 55 percent.⁹⁴

The OAS, a five-item instrument measuring current and past abuse, had sensitivity and specificity of 60 and 90 percent when compared with the ISA in emergency department patients.⁹⁴ After factor analysis, this instrument was edited to a four-item version (OVAT) that demonstrated sensitivity and specificity of 93 and 86 percent in the same population.⁹⁴ A subsequent study comparing OVAT with the ISA among patients presenting to an emergency department also reported high sensitivity and specificity (86 and 83 percent).⁹⁵

Two studies evaluated the three-item PVS, which measures past physical violence and perceived personal safety.^89,90 A prospective study screened women presenting to an emergency department with the PVS and contacted them by telephone 4 months later to administer the CTS.⁸⁹

Participants who screened positive on the PVS had increased risks for verbal aggression (relative risk [RR], 7.25 [95% CI, 3.2–16.2]) and physical violence (RR, 11.3 [95% CI, 5.0–26.3]) compared with those with negative screens. These findings were similar when a single item from the PVS was used (“Have you been hit, kicked, punched, or otherwise hurt by someone?”). This study reported significant findings despite high loss to followup (53.2 percent) and differences in baseline abuse prevalence rates between participants retained and lost to followup (9.4 vs. 21.8 percent).

Results of the PVS and the WAST were compared with the CAS for 2,461 women presenting to family practice, emergency department, and women’s health clinics in Canada.⁹⁰ Sensitivity and specificity were 49 and 94 percent for the PVS and 47 and 96 percent for the WAST. In a subsequent study of a larger sample of women from the same Canadian clinics, the WAST demonstrated higher sensitivity (81 to 88 percent) and specificity (89 percent) using the CAS again as the reference standard.⁹² The WAST identified a 12-month IPV prevalence rate of 22 percent, while the CAS found 14 percent.

English and Spanish versions of the four-item HITS instrument were compared with the ISA and the Spanish-version of the WAST in a cross-sectional study.⁹¹ In a sample of women attending a family practice clinic, the sensitivity and specificity of the English-version HITS was maximized at a cut-point score of 10.5 (86 and 99 percent) and 5.5 for the Spanish-version HITS (100 and 86 percent).

The three-item STaT instrument was evaluated among women presenting to primary care clinics in two studies. Compared with a semi-structured interview, STaT demonstrated sensitivity and specificity of 96 and 75 percent for ≥1 positive responses, 89 and 100 percent for ≥2, and 64 and 100 percent for ≥3.⁹⁶ In a larger sample using the ISA as the reference standard, STaT had sensitivity and specificity of 95 and 37 percent for ≥1 positive responses, 85 and 54 percent for ≥2, and 62 and 66 percent for ≥3.⁹⁷

The four-item HARK instrument was compared with the CAS among women presenting to primary care in London.⁹³ For a positive response on any question, the sensitivity and specificity were 81 and 95 percent.

The SAFE-T instrument includes five questions about a woman’s relationship with her partner (secure at home, accepted by partner, family likes partner, disposition of partner, and talks with partner to resolve differences). A comparison of responses from SAFE-T with one item from the PVS among women in emergency departments indicated sensitivity and specificity of 54 and 81 percent.⁹⁹ Responses on one personal safety question were compared to those on the modified CTS among women in an urban family medicine clinic.¹⁰¹ Sensitivity and specificity were 9 and 91 percent.

Two items from the CTQ-SF, an instrument designed to detect a history of physical or sexual abuse in childhood, were compared with the Evaluation of Lifetime Stressors structured interview among women in a health maintenance organization.⁹⁸ Using a single item from the CTQ-SF, sensitivity and specificity were 70 and 94 percent for physical abuse and 82 and 89 percent for sexual abuse. Using two items to screen for physical or sexual abuse resulted in sensitivity and specificity of 85 and 88 percent.

Instruments Evaluated in Pediatric Clinical Settings. The PSQ, which contains three questions about current and lifetime physical abuse and safety, was compared with the CTS2 in a study of mothers of children <6 years old in a pediatric community clinic.¹⁰⁰ Sensitivity and specificity were 19 and 93 percent for physical assault, 29 and 91 percent for injury, and 27 and 92 percent for psychological aggression.

A five-item screening instrument was designed for administration in pediatric settings by presenting questions with nongraphic language (e.g., “How do you and your partner work out arguments?”).¹⁰² Sensitivity and specificity were 40 and 91 percent when compared with the CTS2 in a cohort of women with children presenting to pediatric clinics or family practices in Ohio.

Key Question 4. For Screen-Detected Women With Current, Past, or Increased Risk for Intimate Partner Violence, How Well Do Interventions Reduce Exposure to Intimate Partner Violence? Reduce Physical or Mental Harms or Mortality?

Summary. Six RCTs reported in eight publications met inclusion criteria, including three trials of interventions targeted to pregnant and postpartum women^{29,30,107–109} and three trials of women enrolled without regard to pregnancy status.^110–112

The National Institutes of Health–District of Columbia (NIH-DC) Initiative to Reduce Infant Mortality in Minority Populations is a randomized trial of counseling interventions to reduce multiple risks factors during pregnancy and postpartum compared with usual care.^29,30,109 Screening for cigarette smoking, environmental tobacco smoke exposure, depression, and IPV with the AAS was done using an anonymous computer interview. Results indicated that compared with usual care, women in the counseling intervention group had significantly fewer recurrent episodes of IPV during pregnancy and postpartum, and better birth outcomes, including fewer very preterm neonates (≤33 weeks’ gestation), lower rates of very low birth weight neonates (<1,500 g), and increased mean gestational age.

A randomized trial of home visitation enrolled women in Hawaiian hospitals who gave birth to an infant evaluated as at risk for maltreatment.¹⁰⁷ Primary outcomes were mothers’ IPV victimization and perpetration towards their domestic partners over the subsequent 9 years of followup. Women randomized to the intervention group received home visitation by paraprofessionals for 3 years. During the program, the intervention group had lower rates of IPV victimization that were of borderline statistical significance, and lower rates of IPV perpetration. Results were similar for physical assault victimization and perpetration. Long-term followup rates of overall IPV victimization and perpetration decreased with nonsignificant between-group differences, although rates of verbal abuse suggest an increase for victimization and perpetration among the intervention group.

A cluster randomized trial of pregnant women and mothers of children ≤5 years old evaluated the effectiveness of mentor support versus usual care in reducing IPV and depression.¹¹² The trial enrolled women in primary care clinics in Australia who disclosed IPV or had behavioral symptoms suggestive of abuse. Results indicated that abuse scores were significantly reduced in the intervention versus usual care group, the odds of experiencing violence at followup when adjusted for baseline abuse was reduced, and other differences were not significant (e.g., depression, physical wellbeing, mental wellbeing, parenting stress).

A cluster randomized trial evaluated a counseling intervention compared with usual care in reducing abuse related to pregnancy coercion. Coercion is defined as a lack of control over a woman’s reproductive health, including compromised decisionmaking or limited ability to enact contraceptive use and family planning, and fear of condom negotiation.¹¹¹ Results indicated that women randomized to the intervention who reported recent IPV at baseline had decreased pregnancy coercion at followup compared with usual care. Women in the intervention group were also more likely to discontinue an unhealthy or unsafe relationship compared with women in the usual care group, regardless of recent IPV status.

A randomized, two-arm trial compared use of a wallet-sized referral card with a nurse management protocol in reducing IPV.¹¹⁰ Two years after the interventions, both groups reported fewer threats of abuse, assaults, danger risks for homicide, and events of work harassment, and there were no significant differences between groups.

Evidence. Five systematic reviews of studies evaluating interventions for IPV have been published since 2003 and were reviewed for this update.^{54,105,113–115} Many of the studies cited in the reviews did not meet inclusion criteria because they were either originally published before 2003 or were outside the scope of Key Question 4.

Six RCTs reported in eight publications met inclusion criteria for Key Question 4 (Table 6; Appendixes C5 and C6).^{29,30,107–112} Three trials evaluated interventions targeted to pregnant and postpartum women,^{29,30,107–109} and three trials enrolled women without regard to pregnancy status and were conducted in primary care,^110,112 Women, Infants, and Children (WIC) clinics,¹¹⁰ and family planning clinics.¹¹¹ One trial met criteria for good quality,³⁰ while two other publications of this trial were rated fair quality.^29,109 Four additional trials met criteria for fair quality^{107,110–112} and one for poor quality.¹⁰⁸ Trials were limited primarily by enrollment of dissimilar groups at baseline, high and/or differential loss to followup, lack of power, analysis was not intention-to-treat or not described, use of self-reported measures, recall bias, missing data, lack of blinding, and lack of true control groups, since asking about abuse is itself an intervention. Trials included narrowly-defined patient populations that may not be applicable to broader populations. Although these limitations are important in the interpretation and application of the study results, many are unavoidable in this field of research. The good- and fair-quality rated studies are described below.

Interventions for Pregnant and Postpartum Women. The NIH-DC Initiative to Reduce Infant Mortality in Minority Populations is a randomized trial of counseling interventions during pregnancy and postpartum compared with usual care.^29,30,109 This trial enrolled 1,044 women at six prenatal care sites in the District of Columbia. Women were eligible for the trial if they were African American, at least 18 years old, ≤28 weeks pregnant, a District of Columbia resident, and English speaking. Screening for cigarette smoking, environmental tobacco smoke exposure, depression, and IPV with the AAS was done using an anonymous computer interview. Additional and followup information was collected by a telephone interviewer blinded to randomization group designations at baseline, 22–26 weeks’ gestation, 34–38 weeks’ gestation, and at an average of 10.3 weeks postpartum. Exposure to IPV was determined using scores from the CTS, which was also used to categorize women as having minor or severe and physical or sexual IPV. These designations were further defined by the study. Birth outcomes were determined by reviewing participants’ medical charts.

Women randomly assigned to the intervention group received prenatal behavioral counseling for two to eight sessions, with up to two postpartum sessions provided by professional counselors. The intervention was delivered during routine prenatal care visits at the clinics by social workers or psychologists trained to respond specifically to each identified risk. The intervention sessions averaged 35 minutes in length. Counseling for IPV emphasized danger assessment, safety behaviors, and information on community resources. Smoking and depression were also specifically addressed for participants with these problems.

At baseline, 32 percent of women reported IPV in the previous year, and rates were similar for intervention and usual care groups. The mean age of participants was 24.5 years, participants initiated prenatal care at an average of 13 weeks’ gestation, most were single, 68 percent had at least a high school education, and 79 percent were enrolled in Medicaid. Regarding other risk factors, 22 percent of participants smoked, 78 percent had environmental smoke exposure, 62 percent were depressed, 32 percent used alcohol, and 17 percent used illicit drugs.

Results indicated that women in the intervention group had significantly fewer recurrent episodes of IPV during pregnancy and postpartum (adjusted OR, 0.48 [95% CI, 0.29–0.80]).^29,30 Further analysis showed that reduction in IPV was confined to minor physical violence, but not severe or sexual violence. Alcohol use and depression at baseline were significantly associated with recurrent episodes of IPV (alcohol use: OR, 1.85 [95% CI, 1.09–3.12]; depression: OR, 1.90 [95% CI, 1.11–3.25]).³⁰

Women in the intervention group also had better birth outcomes, including fewer very preterm neonates (≤33 weeks) (1.5 vs. 6.6 percent; p=0.03), lower rates of very low birth weight neonates (<1500 g) (0.8 vs. 4.6 percent; p=0.52), and increased mean gestational age (38.2 vs. 36.9 weeks; p=0.016). A later publication of this trial also reported fewer very preterm neonates (2.2 vs. 5.0 percent in intervention vs. usual care groups; OR, 0.43 [95% CI, 0.20–0.95]; number needed to treat, 36 mothers).²⁹ The trial also evaluated other risk factors for adverse pregnancy outcomes and found that smoking and environmental smoke exposure were the only other risk factors to decrease in the intervention group.¹⁰⁹ It is unclear how modification of these risk factors influenced birth outcomes reported in the subsequent publications.^29,30

A randomized trial of home visitation enrolled 685 English-speaking mothers in Oahu, Hawaii hospitals who gave birth to an infant evaluated as at risk for maltreatment.¹⁰⁷ Primary outcomes were mothers’ IPV victimization and perpetration towards their domestic partners. Newborn risk was determined by chart review and score on the Kempe’s Family Stress Checklist for screening. Eligible families were not involved with Child Protective Services.

Women randomized to the intervention group received home visitation by paraprofessionals for 3 years. The content of home visits was designed to promote child health and decrease child maltreatment by linking families to appropriate community services, teaching about child development, role modeling positive parenting and problem-solving strategies, and offering emotional support. The intervention was offered by three community agencies and participants averaged 13.6 visits in the first year.

Outcome measures were obtained by interviewers blinded to the group assignment. These occurred within 1 week postpartum, annually during the intervention period when the child was 1 to 3 years old, and annually during the followup period when the child was 7 to 9 years old. Measures included the CTS at baseline and the CTS2 at subsequent data points, with four sexual coercion questions omitted. Additional outcomes were measured by the Mental Health Index for anxiety and depressive symptoms, and questions about alcohol and drug use.

During the program, the intervention group had lower rates of IPV victimization that were of borderline statistical significance (incidence rate ratio [IRR],¹¹⁶ 0.86 [95% CI, 0.73–1.01]), and lower rates of perpetration towards their domestic partners (IRR, 0.83 [95% CI, 0.72–0.96]). Results were similar for physical assault (victimization IRR, 0.85 [95% CI, 0.71–1.00]; perpetration IRR, 0.82 [95% CI, 0.70–0.96]). Long-term followup rates of overall IPV victimization and perpetration decreased with nonsignificant between-group differences. Rates of verbal abuse suggest an increase for victimization (IRR, 1.14 [95% CI, 0.97–1.34]) and perpetration (IRR, 1.08 [95% CI, 0.92–1.26]) among the intervention group. Sexual violence and injury were not significantly associated with group assignment, although low prevalence rates of self-reported sexual abuse and injury may have limited the comparisons.

Results were likely influenced by several factors. Characteristics of participants at baseline differed between groups. In the intervention group, fewer participants used alcohol and had poor mental health, and more were employed. The specific elements relating to IPV in the intervention were minimal and varied by the needs of the family. Participation in the intervention decreased over time (70 percent at 6 months, 49 percent at 12 months, 25 percent at 36 months). Although >85 percent of participants completed the final interview when the child was 9 years old, women lost to followup differed from women retained in the trial (women lost to followup were more likely to be Asian, less likely to be Native Hawaiian). Also, IPV improved for both intervention and control groups as the children aged, consistent with epidemiologic data indicating IPV prevalence is highest for young women.^117,118

Interventions for Women Regardless of Pregnancy Status. A cluster randomized trial of pregnant women and mothers of children ≤5 years old evaluated the effectiveness of mentor support versus usual care in reducing IPV and depression.¹¹² The trial enrolled 215 English or Vietnamese speaking women in primary care clinics in Melbourne, Australia who disclosed IPV or had behavioral symptoms suggestive of abuse. Symptoms included depression, anxiety, frequent attendance without obvious causes, and other signs indicative of abuse. Methods of IPV disclosure were not described, except that participating clinicians underwent 6 hours of training to improve their capacity to identify, respond to, and refer women with IPV or at risk for IPV to community-based services.

Women randomized to the intervention received 12 months of weekly home visitation from trained nonprofessional mentors offering advocacy, parenting support, and referrals. Outcome measures obtained at baseline and at 12 months included abuse measured by the CAS, depression (Edinburgh Postnatal Depression Scale), wellbeing (Short-Form 36-Item Health Survey, Parenting Stress Index–Short Form), and social support (Medical Outcomes Study–Short Form) at baseline and followup.

Results indicated that abuse scores from the CAS were significantly reduced in the intervention compared with usual care groups (adjusted difference, −8.67 [95% CI, −16.2 to −1.15]); the odds of experiencing violence at followup when adjusted for baseline abuse was 0.47 (95% CI, 0.21–1.05]). Other differences were not significant (depression, physical wellbeing, mental wellbeing, parenting stress). Results of the trial were influenced by low recruitment numbers, imbalance in recruitment to the intervention and usual care arms, dissimilar characteristics of participants at baseline (more depression and parenting stress in the intervention group), and drop off in participation with mentoring. Also, the CAS was developed as a categorical outcome, not as a measure of change along a continuum. The implications of changes in scores are unclear.

A cluster randomized trial evaluated a counseling intervention compared with usual care in reducing abuse related to pregnancy coercion.¹¹¹ Coercion is defined as a lack of control over a woman’s reproductive health, including compromised decisionmaking or limited ability to enact contraceptive use and family planning, and fear of condom negotiation.¹¹¹ The trial enrolled 906 English or Spanish speaking women ages 16 to 29 years in urban family planning clinics in California. All women in the participating clinics were screened for IPV using two questions on an intake form, with a standard clinic protocol as part of usual care. For the trial, the intervention clinics also provided a counseling intervention that educated patients about reproduction coercion and provided information about local IPV and sexual assault resources. Measures included items from the CTS2 and Sexual Experiences Survey, questions about awareness and recent use of IPV services, and relationship changes from baseline. Outcome measures were obtained by computer-assisted followup surveys between 12 to 24 weeks after the baseline survey.

Results indicated that women randomized to the intervention who reported recent IPV at baseline had decreased pregnancy coercion at followup compared with usual care (adjusted OR, 0.29 [95% CI, 0.09–0.91]). Women in the intervention group were also more likely to discontinue an unhealthy or unsafe relationship compared with the usual care group (p=0.013), regardless of recent IPV status. The trial was limited by its small sample size (four clinics), underpowered to assess additional outcomes, and restricted in its applicability. Also, followup was short term and demographic characteristics of clinics differed.

A randomized, two-arm trial compared use of a wallet-sized referral card with a nurse management protocol in reducing IPV over the subsequent 2 years.¹¹⁰ A total of 360 English or Spanish speaking women ages 18 to 45 years who reported physical or sexual abuse in the past 12 months were recruited to the study from urban primary care public health clinics and WIC clinics in the United States. A positive response to either of the two questions on the AAS was used to determine IPV exposure. Women were randomized to receive either a wallet-sized referral card with a safety plan and resources for IPV services or a 20-minute nurse case management protocol (March of Dimes). This protocol includes a brochure with a 15-item safety plan, supportive care, anticipatory guidance, and guided referrals.

Measures were obtained by interviews at baseline and at 6, 12, 18, and 24 months post-baseline using the Safety Behavior Checklist, Community Resources Checklist, Severity of Violence Against Women Scale, Danger Assessment Scale, and Employment Harassment Questionnaire. Two years after the interventions, both groups reported fewer threats of abuse (p<0.001), assaults, danger risks for homicide, and events of work harassment, and there were no significant differences between groups. Compared with baseline, both groups adopted more safety behaviors by 24 months, and community resource use declined for both groups (p<0.001), with no significant differences between groups. Interpretation of the effectiveness of the counseling intervention is limited by the lack of a true control group, although ethical issues required addressing IPV in an acceptable usual care arm. The counseling intervention did not demonstrate superiority in reducing IPV compared with the information card in this trial. The trial was also limited by its small size and restricted applicability.

Key Questions 3 and 5. What Are the Adverse Effects of Screening for Intimate Partner Violence and Interventions to Reduce Harm From Intimate Partner Violence?

Summary. Harms related to IPV screening and interventions were reported in three studies included for Key Questions 1 and 4, and in 11 descriptive studies identified by search strategies and systematic reviews.^54,115

Harms were actively monitored in a randomized trial of 6,743 women that evaluated screening versus no screening in Canadian primary care, acute care, and specialty care sites.⁸⁷ Results of the analysis of a measure developed to monitor harms for this study indicated no differences in reported harms for screened women who were either exposed or not exposed to IPV, and no harms were associated with screening for either group. A randomized trial of a 3-year home visitation intervention for at-risk newborns and their mothers suggested increased verbal abuse victimization and perpetration in the intervention group over long-term followup.¹⁰⁷ A randomized, two-arm trial of women receiving either a wallet-sized referral card or a 20-minute nurse management protocol to address IPV found no adverse effects as a result of the intervention.¹¹⁰

Descriptive studies generally indicated low levels of harm related to IPV screening and interventions, but study populations and methods varied widely. Some women indicated discomfort with screening, particularly among those with prior IPV; infringement of privacy; worries about increasing abuse by disclosing IPV; feelings of sadness and depression; and general concerns with IPV screening. These issues were voiced by a minority of respondents in the various surveys and interviews.

Evidence. Harms related to IPV screening and interventions were reported in three studies included for Key Questions 1 and 4^87,107,110 and in 11 descriptive studies^55,119–128 identified by search strategies and systematic reviews^54,115 (Table 7). Most studies could not be rated for quality because USPSTF criteria do not apply to descriptive studies.

Harms Reported in Trials of Screening and Interventions. Harms were addressed in a randomized trial of 6,743 women that evaluated screening versus no screening in Canadian primary care, acute care, and specialty care sites.⁸⁷ Harms were actively monitored using the COST instrument administered to the screening group at baseline. COST is a multidimensional questionnaire developed for the study that measures the effect of being asked IPV screening questions. Results of the analysis of the Effects on Quality of Life subscale indicated no differences in reported harms for screened women who were either exposed or not exposed to IPV, and no harms were associated with screening for either group.

A randomized trial of a 3-year home visitation intervention for at-risk newborns and their mothers indicated nonsignificantly increased rates of verbal abuse victimization (adjusted IRR, 1.14 [95% CI, 0.97–1.34]) and perpetration (IRR, 1.08 [95% CI, 0.92–1.26]) in the intervention group over long-term followup.¹⁰⁷ However, this study also found decreased rates of IPV during the 3 years of the program. A randomized, two-arm trial of women receiving either a wallet-sized referral card or a 20-minute nurse management protocol to address IPV found no adverse effects as a result of the intervention.¹¹⁰

Harms Reported in Descriptive Studies of Screening and Interventions. A prospective, observational study screened 3,083 male and female emergency department patients for IPV, provided resources and information for those who screened positive, and subsequently assessed them for IPV exposure, safety, and use of resources.¹¹⁹ Screening was administered using the Universal Violence Prevention Screening Protocol on a touch screen kiosk. None of the participants reported safety issues in the emergency department after participating in screening, whether or not they disclosed IPV. Two of the 216 participants who screened positive for IPV reported safety concerns or emotional distress related to the screening in followup, and one of the 65 telephone interview participants reported an adverse issue related to screening. No additional safety issues or increases in injuries, violence, or calls to the police were reported as a result of participating in screening or followup.

Five studies were based on surveys of women in health care settings. In a study of 198 women receiving services in an urban emergency department in the United Kingdom, responses on a modified WHO Multi-Country Domestic Violence Study questionnaire indicated that 24 percent felt uncomfortable when asked about IPV.¹²⁴ Women with prior abuse had higher levels of discomfort. Some women commented on the need for privacy and safety and had concerns about direct IPV questions. In a study of 95 women in a trauma center who completed a survey about IPV screening, 18 percent thought screening infringed upon their privacy, but most (90 percent) felt it was appropriate to ask about IPV.¹²⁵ Approximately 25 percent of abused women thought reporting would increase their chances of further harm. A survey of 645 women ages 15 to 24 years in U.S. family planning clinics indicated that most women (90 percent) thought universal IPV screening is a good idea, but 36 percent of younger women (ages 15 to 18 years) had concerns.¹²⁶ Although most women felt positive after IPV screening in a retrospective survey of screened women in Australia, 6 percent (7/119) of participants indicated sadness or depression, and one woman experienced further abuse as a result of her disclosure.¹²⁰ A survey of adults in Canada found that only 10 percent thought it would be inappropriate to ask all women visiting the emergency department about violent or threatening behavior at home.¹²⁷

Five studies were based on data collected from interviews. In a study of 36 women interviewed several weeks after IPV screening in New Zealand, 97 percent perceived it as nonthreatening and safe, and none experienced increased risks.¹²¹ Interviews of 27 abused women in the United States indicated no instances of harmful disclosure in any health care setting (emergency department, obstetrics and gynecology clinics, or primary care clinics), although some disclosures were viewed as not helpful to the women.¹²² A study of 519 women used anonymous computer interviews in maternity units that asked about IPV screening and interventions, past disclosure, preferences about screening, and violence during pregnancy. Most women (97 percent) had no feelings of anger or embarrassment and were not offended when screened for IPV.¹²³ Focus group interviews with women who had experienced IPV described potential negative consequences of screening as feeling judged by the health care provider, increased anxiety about the unknown, feeling that the intervention protocol was cumbersome or intrusive, and disappointment in the health care provider’s response to disclosure.⁵⁵ Although no one described adverse effects from IPV screening in a study based on interviews of women followed up after disclosing abuse, 40 percent thought it had minimal impact.¹²⁸

Key Question 1. Does Screening Asymptomatic Elderly and Vulnerable Adults in Health Care Settings for Current, Past, or Increased Risk for Abuse and Neglect Reduce Exposure to Abuse and Neglect? Reduce Physical or Mental Harms or Mortality?

No RCTs or controlled observational studies of screening for abuse and neglect in elderly or vulnerable adults were identified or met inclusion criteria for this review.

Key Question 2. How Effective Are Screening Techniques in Identifying Asymptomatic Elderly and Vulnerable Adults With Current, Past, or Increased Risk for Abuse and Neglect?

Summary. One study evaluating the diagnostic accuracy of the Elder Abuse Suspicion Index (EASI), a six-item dichotomous screening instrument, met inclusion criteria for the systematic review. The EASI was evaluated in a study of elderly men and women in primary care clinics in Canada. Compared with a comprehensive evaluation by a social worker, the sensitivity and specificity of the EASI varied from 9 to 47 percent and from 75 to 97 percent, respectively, depending on the number of positive responses to specific questions.

Evidence. One study evaluating the diagnostic accuracy of the EASI met inclusion criteria for the systematic review as well as criteria for fair quality (Appendixes C7 and C8).¹²⁹ The EASI is a six-item dichotomous screening instrument that measures dependence on assistance for activities such as bathing or shopping; withholding food, care, or other needs; verbal aggression; financial exploitation; physical harm; and physician assessment of visible signs of abuse or neglect during the preceding 12 months. Its use for screening was evaluated in a cohort of 953 elderly men and women in Montreal, Canada.¹²⁹ Patients were seen in university-affiliated teaching family medicine clinics and a government community-based health and social service center. Patients were eligible for the study if they were age ≥65 years, spoke English or French, scored ≥24 on the Mini Mental Status Exam, and were capable of providing informed consent. The EASI was administered by physicians during the course of the clinic visit. Screening took <2 minutes for most participants.

The reference standard was an evaluation by social workers from the Elder Abuse Center using a 1.5- to 3-hour interview protocol for elder abuse assessment that was considered the community standard. Social workers in the study underwent training on the use of the interview protocol to ensure consistency. Social workers were blinded to the results of the EASI when they conducted their interviews, which occurred within 3 weeks of the EASI in the participants’ homes or other locations chosen by the participants. If social workers identified elder abuse in the course of their interviews, they followed a protocol to refer study participants to appropriate services for help. A total of 66 percent of enrolled participants completed both the EASI and social work interview; characteristics of participants retained in the study and lost to followup were similar.

Compared with the evaluation by a social worker, the EASI had sensitivity and specificity of 47 and 75 percent with ≥1 positive responses on questions 1 to 6, 32 and 89 percent with ≥1 positive responses on questions 2 to 6, 14 and 96 percent with ≥2 positive responses on questions 1 to 6, and 9 and 97 percent with a positive response to question 1 and ≥1 positive responses on questions 2 to 6. Nearly one third of the 663 participants who completed all evaluations gave positive responses to at least one question on the EASI.

Key Question 4. For Screed-Detected Elderly and Vulnerable Adults With Current, Past, or Increased Risk for Abuse and Neglect, How Well Do Interventions Reduce Exposure to Abuse and Neglect? Reduce Physical or Mental Harms or Mortality?

Summary. No RCTs or controlled observational studies of interventions for abuse and neglect in elderly or vulnerable adults were identified or met inclusion criteria for this review. One descriptive retrospective study included predominantly male veterans ≥65 years old who received health care services from the West Los Angeles Veterans Affairs (VA) Medical Center. Veterans were identified as possibly exposed to abuse or neglect, referred by their primary care providers to the Geriatric Research, Education, and Clinical Center’s (GRECC’s) Outpatient Clinic, and received social work services. Of 575 veterans evaluated, 41 incidents or situations of elder abuse and neglect among 31 veterans (5.4 percent) were identified and reported to APS over a 3-year period, and 33 veterans received specific services after case management, such as nursing home placement or conservatorship arrangements.

Evidence. No RCTs or controlled observational studies of interventions for abuse and neglect in elderly or vulnerable adults were identified or met inclusion criteria for this review.

One descriptive retrospective study of predominantly male (96 percent) veterans ages 65 to 103 years evaluated outcomes resulting from social work interventions (Appendix C9).²⁰ The study included veterans who received services from the West Los Angeles VA Medical Center, were identified as possibly exposed to abuse or neglect, and referred by their primary care providers to the GRECC Clinic. It is unclear how patients were identified as abused, and whether individuals identified by primary care physicians had symptoms relating to abuse. Forty-eight percent of the veterans had dementia and 35 percent had depression. At the GRECC Clinic, a social worker (case manager) and APS staff provided individualized services as needed. Outcomes included type of abuse, moving from unsafe living situations to nursing homes or board and care facilities, and implementation of conservatorship arrangements.

Of 575 veterans, 41 incidents or situations of elder abuse and neglect among 31 veterans (5.4 percent) were identified and reported to APS over a 3-year period. Abuse was classified as physical (five incidents), psychological (five incidents), neglect (seven incidents), financial (12 incidents), and self-neglect (12 incidents). After case management, four individuals received conservatorship arrangements, six received conservatorship “plus other,” eight were moved to a nursing home and three to board and care/assisted living, seven remained at home with services, five refused services, and six outcomes were unknown.

Key Questions 3 and 5. What Are the Adverse Effects of Screening for Abuse and Neglect of Elderly and Vulnerable Adults and Interventions to Reduce Harm From Abuse and Neglect?

No studies of the harms of screening or interventions for abuse and neglect in elderly or vulnerable adults were identified or met inclusion criteria for this review. Potential harms include shame, guilt, self-blame, and fear of retaliation by perpetrators.²⁰