Long-term sequelae of intimate partner violence identified in hospital settings in New York State: a network analysis approach - PROJECT ABSTRACT Many survivors of intimate partner violence (IPV) continue to suffer consequences even after the abuse has ended, with a constellation of long-term health problems that are poorly understood and may be mis- diagnosed, delaying needed treatment. Advanced methodological approaches are needed that can help to identify individuals exposed to IPV who may be at greatest risk for long-term adverse outcomes. We propose to apply network analysis methods to characterize the structure of diagnoses at index IPV-related hospital visits and examine how features of these network structures predict longer-term outcomes, including mortality. Our central hypothesis is that understanding the strength and patterns of connections between comorbidities at these index visits will shed light on longer-term trajectories, including repeat IPV visits and death. We will test this hypothesis by pursuing three specific aims: (1) Identify networks of comorbid diagnoses that characterize index IPV-related emergency department (ED) visits and hospitalizations, compared to visits among matched controls; (2) Characterize long-term sequelae across features of index IPV diagnosis networks, compared to diagnosis networks among matched controls; and (3) Evaluate community-, hospital-, and individual-level characteristics as potential moderators of relations between features of index diagnosis networks and long- term sequelae. To achieve these aims, we will conduct a retrospective cohort study using data from the Statewide Planning and Research Cooperative System (SPARCS), which includes information on all inpatient, ED, and hospital outpatient clinic visits in New York State, linked to mortality records. Social network analysis and machine learning methods will be used to characterize IPV-related diagnosis networks and identify the key features of index diagnosis networks that predict more severe long-term outcomes. We will also use multi-level analyses to examine the potential buffering effects of hospital- and community-level characteristics. Our approach is innovative because it applies network analysis approaches to not only characterize baseline diagnosis networks but to connect them to subsequent outcomes. Our project is significant because it addresses the under-explored but potentially devastating consequences of IPV and may uncover key red flags at index visits that could be used to identify individuals at greatest risk for adverse outcomes, prompting additional screening and safety planning. Our findings will serve as preliminary data for an R01 proposal to scale-up these methods for a nationwide examination of the long-term sequelae of IPV, including additional applications of systems science approaches to inform the development of interventions for the prevention of long-term health effects in this understudied, high-risk population.
