Refining and Validating Borderline Personality Disorder Phenotypes Through Factor Mixture Modeling

点击翻译按钮获取中文摘要

The proposed research seeks to clarify the symptomatic heterogeneity of borderline personality disorder (BPD) by examining BPD phenotypes through advanced latent variable modeling. A second, innovative aim is to validate these findings through intensive longitudinal assessment in daily life. BPD is associated with high rates of emergency room visits and costly healthcare service utilization, affecting 10-20% of psychiatric outpatients and 20-40% of psychiatric inpatients. BPD also contributes to impaired social and occupational functioning and significant suicide risk, with 1 in 10 individuals with BPD completing suicide. Recent research has aimed to enhance treatment effectiveness for BPD by identifying prototypical patterns of symptom manifestation that may suggest ideographic treatment targets. However, no research has simultaneously included: a) a sufficiently large patient sample; b) ecologically sound validation of results; and c) use of appropriate statistical techniques. The proposed project builds on this research through two aims. Aim 1: Utilize a model comparison approach to identify BPD phenotypes in a large psychiatric outpatient sample assessed via semi-structured diagnostic interviews (Study 1). Aim 2: Validate the results of Study 1 by applying phenotype classification algorithms produced in Study 1 to a smaller sample of patients who have completed 21 days of momentary surveys on symptoms and clinical outcomes (Study 2). To address Aim 1, factor mixture modeling (FMM)—a novel, flexible, and integrative latent variable modeling approach—will be compared to standard factor analysis and latent class analysis in order to evaluate the dimensional and categorical structure of BPD. We expect a single-factor, multi-class FMM will best explain heterogeneity in BPD, over and above other sources of heterogeneity (e.g., gender, comorbidity). To address Aim 2, we will use a prototype-matching approach to algorithmically assign patients in the validation sample to phenotypes identified in Aim 1 and determine their predictive validity in terms of daily clinical outcomes. Results of this project will provide empirically grounded personalized prediction tools for BPD intervention and treatment development, in line with the NIMH’s goal of “developing, testing, and refining tools and methodologies… for personalized risk and trajectory prediction and intervention.” This fellowship will allow the applicant to receive tailored consultation from experts in methodology, data analysis, and BPD theory and assessment, as well as advanced statistical training and grantsmanship courses and workshops. This training will be enhanced by the resource-rich environment and explicit support of student research and funding provided by the Pennsylvania State University, as well as the support of Dr. Kenneth Levy and his lab. This promising young researcher will gain training in computational modeling, proficiency in working with “big data,” increased understanding of conceptual and nosological models of BPD, and further skills in disseminating research findings through publication and presentation, as vital steps towards an independent research career in translational clinical science.