A group of researchers in 12 cities across the country partnered with police departments, district attorney offices, domestic violence shelters, and medical examiners to conduct the study. A case control design compared information from interviews with proxy informants for females killed by an intimate partner (cases) with information from abused women (abused controls). Sampling quotas for cases and controls for each of the 12 cities in the study were determined by annual rates of intimate partner femicides. The study encompassed 220 femicide cases and 356 abused controls. In addition to administering the DA, the interview solicited information on demographic and relationship characteristics, including type, frequency, and severity of any violence; psychological abuse and harassment; alcohol and drug use; and weapon availability. Scales that measured partners' controlling behaviors and stalking were constructed based on factor analysis of the risk factor items. Bivariate logistic regression was used to estimate the independent association between each of the hypothesized risk factors from the DA and the risk of intimate partner femicide. Psychometric analysis of the DA included internal consistency and discriminant group validity using mean scores. The study found that 15 of the 17 items of the DA distinguished intimate partner femicide victims from abused women. The factor with the strongest risk (highest odds ratio) was the use (or threatened use) of a weapon. Women who had been threatened with being killed were almost 15 times more likely to be among the femicide victims rather than among the abused controls. Perpetrator drug abuse and serious alcohol abuse also differentiated batterers who killed from those who did not, as did prior gun ownership. The two DA items that did not significantly differentiate intimate partner femicide victims from abused women pertained to suicidality. Internal consistency of the DA was acceptable among the femicide cases and among the controls. Researchers are continuing to analyze the data to determine a DA cutoff score. Implications of the findings are drawn for practitioners and for researchers. (Abstract) Record #4262