Investigating Complex Relationship Between Circumstances of Suicide in Perinatal Women using the NVDRS: A Network Analysis

Body of Abstract: Title: Investigating Complex Relationship Between Circumstances of Suicide in Perinatal Women using the NVDRS: A Network Analysis

Track: Research & Innovation

IRB Approval: N/A

In the Works - New for this year

Abstract:
- Research Aims
The U.S. has the highest maternal mortality rate among developed countries, with 80% of maternal deaths deemed preventable. Suicide during the perinatal period is a leading cause of U.S. maternal mortality. According to the Life-course Framework, suicide risk is shaped by the timing and accumulation of known contextual risk factors, such as interpersonal relationship issues, health, employment, and housing. Our previous work found that perinatal suicide decedents, compared to nonperinal decedents, had higher odds of interpersonal problems, depressed mood, substance abuse, and physical health problems. However, how these circumstances interact with each other remains unclear. Therefore, exploring the complex interactions of circumstances among perinatal decedents is important for understanding suicide and informing prevention policy and interventions.
Network analysis is a novel tool to visualize complex systems and explore complex relationships that has been successfully applied in clinical psychopathology. Using network analysis methods, this study aimed to 1) understand how circumstances from different life-course domains related to each other in a network among all perinatal descendants and 2) investigate whether the network structure of circumstances differ between decedents who were pregnant (i.e., pregnant at time of death) versus postpartum (i.e., pregnant within a year of death) at the time of death.
- Methods
This cross-sectional observational study used data from the National Violent Death Reporting system (NVDRS), a state-based surveillance system of violent deaths. Data included suicides and deaths of undetermined intent identified in the NVDRS from January 1, 2003 to Decemeber 31, 2021. Variables of interest included binary circumstance variables (i.e., contextual risk factors) cited in coroner, medical examiner, or law enforcement case narratives as a contributing factor. The analytic sample included decedents aged 10 to 50 years who were either pregnant or postpartum at time of death. A total of 13 circumstance variables across 6 Life-course domains (physical health, mental health, bereavement, interpersonal relationship, financial and job issue, and substance abuse) were included as nodes of the networks. The relationship between the circumstances variables were visualized using network analysis. L1-regularized regression was used to estimate the width of edges (i.e., edge weight) between nodes. Key metrics estimated included betweenness centrality (assessing how often a node lies on the shortest path connecting any two other nodes), strength centrality (assessing the total weight connected to a node) to understand the role of each node in the network. Two networks were generated for pregnant decedents and postpartum decedents respectively. A Network Comparison Test was used to compare the network structure, and a Pearson Correlation test was used to estimate the correlation of edges between two networks.
- Results
This study included 1,150 perinatal descendants identified in the NVDRS. The mean age was 29.1 years (SD: 7.4). 456 (39.6%) were pregnant at time of death and 694 (60.4%) were postpartum. In the network of all perinatal decedents, history of mental illness treatment had the highest strength centrality. Depressed mood was in the most central position in the network of circumstances and had the highest value of betweenness centrality. The stability test showed that the network was reasonably stable. The visualizations of the individual networks for pregnant and postpartum decedents showed differences, though the network comparison test showed that the two networks of circumstances were similar in terms of correlation of edges (Pearson r= 0.95) and network structure (Global Strength Invariance, p=0.09)
- Conclusions
The network of perinatal decedents is consistent with the Life Course Framework that suggests different domains of risk factors are not isolated from each other. However, the individual network structure of circumstances may not substantially differ by pregnancy status.