[Skip to Navigation]
May 7, 1997

Accuracy of 3 Brief Screening Questions for Detecting Partner Violence in the Emergency Department

Author Affiliations

From the Departments of Emergency Medicine (Dr Feldhaus) and Psychiatry (Dr Norton), Denver Health Medical Center; Colorado Emergency Medicine Research Center, Division of Emergency Medicine, Department of Surgery (Ms Koziol-McLain and Drs Lowenstein and Abbott), University of Colorado School of Medicine (Ms Amsbury), University of Colorado Health Sciences Center, Denver.

JAMA. 1997;277(17):1357-1361. doi:10.1001/jama.1997.03540410035027

Objective.  —To devise a brief screening instrument to detect partner violence and to partially validate this screen against established instruments.

Design.  —Prospective survey.

Setting.  —Two urban, hospital-based emergency departments.

Participants.  —Of 491 women presenting during 48 randomly selected 4-hour time blocks, 322 (76% of eligible patients) participated. Respondents had a median age of 36 years; 19% were black, 45% white, and 30% Hispanic, while 6% were of other racial or ethnic groups; 54% were insured.

Interventions.  —We developed a partner violence screen (PVS), consisting of 3 questions about past physical violence and perceived personal safety. We administered the PVS and 2 standardized measures of partner violence, the Index of Spouse Abuse (ISA) and the Conflict Tactics Scale (CTS).

Main Outcome Measures.  —Sensitivity, specificity, and predictive values of the PVS were compared with the ISA and the CTS as criterion standards.

Results.  —The prevalence rate of partner violence using the PVS was 29.5% (95% confidence interval [CI], 24.6%-34.8%). For the ISA and CTS, the prevalence rates were 24.3% (95% CI, 19.2%-30.1%) and 27.4% (95% CI, 21.7%-33.6%), respectively. Compared with the ISA, the sensitivity of the PVS in detecting partner abuse was 64.5%; the specificity was 80.3%. When compared with the CTS, sensitivity of the PVS was 71.4%; the specificity was 84.4%. Positive predictive values ranged from 51.3% to 63.4%, and negative predictive values ranged from 87.6% to 88.7%. Overall, 13.7% of visits were the result of acute episodes of partner violence.

Conclusion.  —Three brief directed questions can detect a large number of women who have a history of partner violence.