PTSD and Autoimmune Disease: Towards Causal Effects, Risk Factors, and Mitigators

创伤后应激障碍（PTSD）是一种常见的，慢性的，使退伍军人衰弱的精神疾病。 除了精神病学特征外，PTSD还与多种身体健康状况有关， 行为和生物过程的失调，如免疫失调和慢性炎症。 先前的证据表明PTSD和自身免疫性（AI）疾病风险之间存在关联， 超过80种涉及自身免疫反应的复杂疾病。然而，将PTSD和AI联系起来的研究 疾病风险主要集中在少数流行的AI条件，尚未估计潜在的因果关系， 关系，一直在欧洲主要是白色样本，并没有检查风险或缓解因素。 因果方法，如边际结构模型，可以解释观测中的时变因素， 数据，以更好地估计因素之间的因果关系，提供更精确的推论比以前的关联 问题研究此外，还需要研究来确定PTSD和所有AI疾病之间的联系， 在很大程度上是异质的，但共享潜在的病因。事实上，确定PTSD和某些 AI失调的形式可能指向构成疾病风险的免疫过程模式。给予更高 PTSD和一些AI疾病在种族或少数民族群体中的发病率，有必要探索潜在的 PTSD和AI疾病之间的健康差异。此外，其他重要风险或保护 影响创伤后应激障碍中AI疾病风险的因素可以通过利用大的临床样本进行经验性检查， 在机器学习上下文中测试多个预测器。与此相关，没有研究确定是否 PTSD的治疗，如抗抑郁药或循证心理治疗，可能会减轻AI疾病的风险 创伤后应激障碍患者中。本研究旨在通过估计文献中的这些差距， PTSD和AI疾病之间的因果关系在一个大的，不同的美国退伍军人样本。第一个目标是 为了估计PTSD对AI疾病风险的因果影响（例如，任何AI疾病，个体AI病症）和 检查精神并发症的影响（例如，多项精神病诊断）。第二 目的是确定种族和民族是否改变PTSD和AI疾病之间的关联， 使用数据驱动的方法来探索增加或减轻AI疾病风险的临床因素， 创伤后应激障碍第三个目的是调查接受治疗的人（例如，抗抑郁药物， 与未接受治疗的PTSD患者相比，PTSD患者接受心理治疗的风险降低。 为了实现所有目标，将从大约900万退伍军人的国家VA电子健康记录（EHR）中收集数据。 访问和分析，以确定PTSD，AI疾病的诊断和相关协变量随时间的变化。我们 将应用边缘结构模型，机器学习算法进行特征选择和逻辑回归 用倾向分数匹配来解决目标。与研究目标相一致，培训目标将 支持我作为一个独立的研究人员的发展，包括发展：1）临床PTSD的知识 病理学和治疗; 2）PTSD的心理神经免疫学过程的专业知识; 3）理解 AI障碍及其病因学;以及4）精通大数据方法，包括实施因果推理 以及大规模EHR数据中的机器学习。我的研究和培训目标将得到一个优秀的 导师团队的跨学科研究人员，并将在旧金山弗朗西斯科退伍军人事务部进行 卫生保健系统。这个职业发展奖是关键的下一步，我的整体科学和 职业目标，将数据科学和流行病学应用于VA数据，以了解关系 创伤、创伤后应激障碍和身体疾病之间的关系，以改善患有创伤后应激障碍的退伍军人的健康状况。

Posttraumatic stress disorder (PTSD) is a common, chronic, and debilitating psychiatric condition in Veterans. Beyond psychiatric features, PTSD has been linked multiple physical health conditions due to poorer health behaviors and dysregulation of biological processes such as immune dysregulation and chronic inflammation. Prior evidence has indicated an association between PTSD and risk for autoimmune (AI) conditions, a group of over 80 complex diseases involving self-reactive immune responses. However, research linking PTSD and AI disease risk has largely focused on only a few prevalent AI conditions, has not estimated potential causal relationships, has been in European mostly White samples, and has not examined risk or mitigating factors. Causal methods, such as marginal structural modeling, can account for time-varying factors in observational data to better estimate causal links between factors, providing more precise inferences than prior associational studies. Additionally, research is needed to determine associations between PTSD and all AI diseases, which are largely heterogeneous but share underlying etiology. Indeed, determining links between PTSD and certain forms of AI dysregulation may point to patterns of immune processes that underlie disease risk. Given higher rates of PTSD and some AI diseases in racial or ethnic minority groups, it is necessary to explore potential health disparities in associations between PTSD and AI disease. Moreover, other important risk or protective factors influencing AI disease risk in PTSD can be examined empirically by utilizing a large clinical sample and testing multiple predictors in a machine learning context. Relatedly, no studies have determined whether treatment for PTSD, such as antidepressants or evidence-based psychotherapy, may mitigate AI disease risk among individuals with PTSD. This study is designed to respond to these gaps in the literature by estimating causal associations between PTSD and AI disease in a large, diverse sample of US Veterans. The first aim is to estimate the causal impact of PTSD on AI disease risk (e.g., any AI disease, individual AI conditions) and examining the effect of psychiatric comorbidity (e.g., multiple psychiatric diagnoses) on AI disease. The second aim is to determine whether race and ethnicity modify the association between PTSD and AI disease and to use data-driven methods to explore clinical factors that increase or mitigate risk for AI disease in those with PTSD. The third aim is to investigate whether receiving treatment (e.g., antidepressant medications, psychotherapy) for PTSD attenuates risk for AI disease compared to those with PTSD not receiving treatment. For all aims, data from national VA electronic health records (EHR) of approximately 9 million Veterans will be accessed and analyzed to identify diagnoses of PTSD, AI disease, and relevant covariates across time. We will apply marginal structural models, machine learning algorithms for feature selection, and logistic regression with propensity score matching to address the aims. Aligned with the research aims, the training aims will support my development as an independent researcher, including to develop: 1) knowledge of clinical PTSD pathology and treatment; 2) expertise in psychoneuroimmunological processes in PTSD; 3) understanding of AI disorders and their etiology; and 4) proficiency in big data methods including implementing causal inference and machine learning in large-scale EHR data. My research and training aims will be supported by an excellent mentorship team of interdisciplinary researchers and will be conducted at the San Francisco Veterans Affairs Health Care System. This Career Development Award is the critical next step towards my overall scientific and career goals, which are to apply data science and epidemiology to VA data to understand relationships between trauma, PTSD, and physical disease in order to improve the health of Veterans with PTSD.