Characterization of Marital Violence Using Data from Six Countries in the World Health Organization World Mental Health Survey Initiative

Intimate partner violence (IPV) is a pervasive global public health problem that occurs in all settings and cultural groups. Traditionally, IPV research has largely focused on identifying risk factors from those who have already been victimized. In contrast to descriptive statistics, this project utilized predictive modeling methods to develop a robust model to predict risk for IPV, defined as moderate physical violence occurring within current marriages. Data for this project come from six countries participating in the World Health Organization (WHO) World Mental Health (WMH) Survey Initiative. Analyses capitalized on the availability of data containing detailed pre-marital factors from both members of currently married couples and considered both independent and joint effects. All potential predictors were broken into four defined predictor groups; demographics and relationship characteristics, adverse experiences in childhood, violence in dating relationships, and pre-marital psychiatric disorders.;Among the 1,515 couples within our sample, 14.4% (se, 0.98) experienced female victimization of IPV as reported by either the husband or the wife. Separate analyses for each predictor group resulted in ten significant variables; three demographics and relationship characteristic predictors, two childhood adversity predictors, two dating experience predictors, and three mental disorder predictors. All ten predictors were used to construct a final predictive model. Predicted probabilities of marital violence for each couple were then calculated from the final model\u27s coefficients. Given the possibility of overfitting our model, we then used the method of replicated 10-fold cross-validation with 20 replicates and generated predicted probabilities of marital violence for each couple in this simulated data set (20 times our original sample size, n=30,300). A Receiver Operating Characteristic and Area Under the Curve were calculated to quantify overall prediction accuracy of the model in the observed and simulated data sets. Model fit indices were strong as the estimated Area Under the Curve for the observed data was 0.75 and 0.70 for the simulated data. The top 5% of respondents with the highest predicted risk included 18.6% of all cases of marital violence. This is just under four times the proportion expected by chance.;The World Mental Health survey findings advance our understanding of the extent to which marital violence varies within the context of the couple. Traditionally, research on IPV utilize report from one person, typically the female victim. Our results suggest that this practice does not adequately describe IPV as it is inherently a dyadic experience. These results are valuable in providing a foundation for more targeted primary prevention efforts