DYNAMIC TOPIC LANGUAGE MODEL ON HETEROGENEOUSCHILDREN'S MENTAL HEALTH CLINICAL NOTES

• 成果类型: Article
• 署名作者: Ye, Hanwen; Moreno, Tatiana; Alpern, Adrianne; Ehwerhemuepha, Louis; Qu, Annie
• 署名单位: University of California System; University of California Irvine; Childrens Hospital of Orange County
• 刊物名称: ANNALS OF APPLIED STATISTICS
• ISSN/ISSBN: 1932-6157
• DOI: 10.1214/24-AOAS1930
• 发表日期: 2024
• 页码: 3165-3184

摘要

Mental health diseases which affect children's lives and well-beings havereceived increased attention since the COVID-19 pandemic. Analyzing psy-chiatric clinical notes with topic models is critical to evaluating children'smental status over time. However, few topic models are built for longitudinalsettings, and most existing approaches fail to capture temporal trajectoriesfor each document. To address these challenges, we develop a dynamic topicmodel with consistent topics and individualized temporal dependencies onthe evolving document metadata. Our model preserves the semantic mean-ing of discovered topics over time and incorporates heterogeneity amongdocuments. In particular, when documents can be categorized, we proposea classifier-free approach to maximize topic heterogeneity across differentdocument groups. We also present an efficient variational optimization pro-cedure adapted for the multistage longitudinal setting. In this case study, weapply our method to the psychiatric clinical notes from a large tertiary pedi-atric hospital in Southern California and achieve a 38% increase in the overallcoherence of extracted topics. Our real data analysis reveals that children tendto express more negative emotions during state shutdowns and more positivewhen schools reopen. Furthermore, it suggests that sexual and gender minor-ity (SGM) children display more pronounced reactions to major COVID-19events and a greater sensitivity to vaccine-related news than non-SGM chil-dren. This study examines children's mental health progression during thepandemic and offers clinicians valuable insights to recognize disparities inchildren's mental health related to their sexual and gender identities