A Network Analysis of Borderline Personality Disorder Symptoms: a real-world data (RWD), electronic health record (EHR) study

Abstract: Background Borderline Personality Disorder (BPD) is a heterogeneous disorder. The current study used network analysis of Electronic Health Record (EHR) data to explore associations between symptom groupings in patients diagnosed with BPD. Methods Symptom data recorded by healthcare professionals as part of clinical assessment were extracted from de-identified EHR’s from 25 U.S. mental healthcare providers. Patients with a diagnosis of BPD (ICD-9: 301.83, ICD-10: F60.3) and symptom data recorded within 1 month of diagnosis were included. Symptoms were mapped into 9 BPD-specific constructs (affect instability, problems with relationships, impulsivity, emptiness, problems with anger control/irritability, identity disturbance, (para)suicide, fear of abandonment, dissociation). The network of symptoms was estimated using Enhanced Least Absolute Shrinkage and Selection Operator (eLASSO). Model selection was based on the Extended Bayesian Information Criterion (EBIC). Centrality measures of betweenness, closeness and strength were calculated. Results Data from 2,287 patients (Mean(SD) age=32.7(11.8) years; 85.1% female) with a diagnosis of BPD were used in the current analysis. 301 unique symptoms were recorded. Affective instability (82.8%), problems with relationships (75.3%) and impulsivity (74.9%) were the most frequently reported symptoms. Centrality indices indicated four main clusters, with relationship problems being the most central element in the network. Conclusions Despite the heterogeneity of BPD, affective instability, problems with relationships and impulsivity are key constructs in the disorder. Further exploration into the association between BPD symptom clusters could identify patient subtypes and provide insight into the heterogeneity of BPD and help to personalise treatment pathways.Short Description: This research uses network analysis techniques applied to de-identified electronic health record data to investigate the presentation of borderline personality disorder symptoms at the point of diagnosis in real-world clinical practice.Name of Sponsoring Organization(s): Holmusk Technologies Inc.