Attempted Femicide: A Mixed Methods Approach To Deriving An Operational Definition for the Fedisa Modikologo Study, South Africa

Abstract Purpose Women often do not recognize when they experience attempted femicide, so another way of measuring it is needed for research. We examined narrative accounts of femicide attempts, and analysed data to determine whether we could systematically expand the definition of attempted femicide, for the purposes of research. Methods We recruited a non-probability sample of 3742 women into the Fedisa ModikologoStudy. They were 18–45 years and had experienced severe intimate partner violence in the previous year. They completed an interview with two closed and one open question about attempted femicide experiences. Results Strangulation, suffocation, stabbing, bludgeoning, gun shots, fire, mocked suicides and accidents were all used in attempted femicides, which 32.4% (1211/3742) of women said they had experienced. These were significantly more common among this sub-group of women. Gunshot injuries and loss of consciousness due to strangulation were the most strongly associated with a recognized femicide attempt, however other injuries were also associated. Assuming unrecognized femicide attempts were broadly similar, we conducted an analysis with recognized femicide attempts as a putative gold standard, and examined change in the sensitivity, specificity and ROC AUC of including different elements in the attempted femicide definition. After incrementally adding elements, we optimized the model with a sensitivity to 85.0%, specificity of 52.7% and ROC AUC of 67.1. Conclusions Our findings point to a working definition of attempted femicide as having a recognized femicide attempt, or injuries leading to loss of consciousness, acts of strangulation, suffocation, gunshots, burns, or stabbing to the neck, torso or genitals.