PS 4-49
DATA-DRIVEN ANALYSIS OF MINED ELECTRONIC HEALTH RECORDS REVEALS ASSOCIATIONS BETWEEN INTIMATE PARTNER VIOLENCE AND ADVERSE HEALTH CONSEQUENCES
Intimidate partner violence (IPV) often results in acute physical injury, sexual assault, and mental health issues. Therefore, it is essential for healthcare providers to better understand the health trends of IPV victims to provide better treatment and prevention. As such, mining Electronic Health Record (EHR) data of IPV victims provides an innovative way to identify key female health issues potentially associated with IPV. Mapping these issues to a network map, we can reveal otherwise hidden symptom associations.
Method:
We obtained two EHR data sets by querying the Explorys platform (Explorys, Inc. gathers national electronic health and billing information from hospitals and healthcare organizations): the domestic abuse cohort with N=5,870, and the non-domestic abuse cohort with N=14,315,140. Using the Chi-Squared test (p < 0.05), we found 2,429 terms were significantly more prevalent among victims of domestic abuse, compared to the general population. These terms were classified into 28 broad categories (e.g. 'acute conditions', 'gastrointestinal', 'mental health’) by two independent research assistants with high interrater reliability.
A subsequent network map was created, taking the 28 categories to decide the strength of relationships among them in order to identify general trends in negative health effects associated with IPV. This was achieved by counting the frequency of each category among IPV-significant terms, noting that a single term could be coded into multiple categories. Therefore, measuring the frequency of co-occurrence of the categories among the coded terms was equally important. 208 pairs of categories were assigned together to at least one of the 2,429 terms.
Result:
Our network visualization created with GePhi 0.8.2 Beta showed that acute conditions are connected to chronic, cardiovascular, gastrointestinal, gynecological, and nervous system conditions, while acute injuries seem to remain isolated from other symptoms. Findings corroborate existing research. However, chronic and acute conditions tend to relate to each other the most, hinting at overlap between those two categories.
Conclusion:
Future research should focus on the underlying role of stress, which is a known health factor for IPV victims and has been shown in other research studies to affect immune response and several other categories in our network map. Furthermore, network mapping and additional research into IPV comorbidity and symptom associations may help the healthcare community to improve IPV screening, identification and treatment.