Prediction of intima partner violence using administrative data on pregnancy

Abstract

Background Intimate partner violence (IPV) is one of huge problem worldwide, especially IPV during pregnancy need to be prevented because it leads negative health outcome for both the mother and offspring. However, it is hard to find women who suffer from IPV by health practitioner or public health staff due to stigma or hesitation. Thus, it would be useful if we can detect the possibility of IPV using administrative data on pregnancy. The aim of this study is to develop an algorism to predict IPV during pregnancy using administrative data on pregnancy. Methods We used data of pregnancy record registered at Adachi city, Tokyo, in 2016 fiscal year (n = 6008). The administrative pregnancy record include date of registration, location of health center, type of center where pregnancy record submitted, parity, maternal age, BMI, psychiatric disease, lack of insurance or public aid, week of gestation when pregnancy record submitted, marital status, feeling when recognized pregnancy, multiple pregnancy, twin, teenage pregnancy for first baby, existence of social support, having trouble within family, history of abnormal pregnancy or delivery, having smoker within family, drinking, economic situation. IPV was assessed for the first time interview or other opportunity for further assessment by public health nurse. Multiple logistic regression model was used to predict IPV. Results IPV cases were found for 38 (0.65%) cases. Multiple logistic regression analyses showed that weekend registration, location of health center, parity, young maternal age, psychiatric disease, lack of insurance or public aid, late week of gestation when pregnancy record submitted (20 week or later), single marital status, multiple pregnancy (4 or more), lack of social support, having smoker within family, drinking, severe economic situation showed significant association with IPV. These variables predicted IPV with good predictive power (area under receiver operating characteristic curve = 0.95, 95% confidence interval: 0.92 to 0.98). Conclusion We showed strong predictability of IPV during pregnancy using administrative data on pregnancy record. Current study is useful to prevent deterioration of IPV per se, and further adverse health outcome due to IPV during pregnancy, such as postpartum depression, low birth weight or child maltreatment.