A New Approach to an Old Problem: Redesigning Latent Tuberculosis Screening and Treatment

总结/摘要 2015年，结核病（TB）超过艾滋病毒成为全球传染病死亡的头号原因。在 美国，加州是全国发病率最高、结核病病例最多的州， 占2017年所有新发活动性结核病例的四分之一。在美国，超过80%的活动性结核病是由于 再激活，这可以通过筛查和治疗LTBI来预防。然而，采用潜伏性结核病 筛查和治疗指南非常糟糕。筛查指南效率低下， 临床医生几乎无法获得这些数据，治疗指南也令人困惑。此外，本发明还 LTBI患者的治疗开始率和完成率较低， 我们对此知之甚少。为了解决这些错失的机会，我们将使用广泛的电子健康记录 加州两家最大医疗机构的数据以及LTBI的定性数据 利益相关者要实现三个具体目标。对于目标1，我们将研究目前在筛查实践中的差距， LTBI，以及指南本身的差距，通过模拟筛查的情况， 完美实施。为此，我们将收集LTBI的实验室检测数据， 感兴趣，并将使用建模来估计通过当前筛查预防的结核病例数量， 衡量有效性。我们将使用模拟模型来估计最佳筛选的有效性。为 目标2，我们将根据广泛的变量制定和验证新的LTBI筛查和治疗指南 在电子健康记录中。首先，我们将使用机器学习和传统回归来识别 阳性LTBI测试和再活化TB的风险因素。接下来，我们将为每个风险因素分配风险评分， 将使用联合概率分析来确定最需要筛查和治疗的人群。到 估计新提出的策略的性能，我们将使用仿真建模。目标3： 开发和试点一个文化定制的教育视频干预，以改善LTBI治疗的启动， 建成我们将首先通过对患者的定性访谈来确定治疗依从性的障碍 随后将根据访谈结果制作一个短片。我们将 进行个体随机疗效试验，以评估干预对启动和治疗的影响 完成率。这种方法是创新的，因为我们提出了一个完整的重新构建目前的美国。 LTBI控制策略，大大提高了一线供应商的易用性。成果 这项工作将通过提供低成本和易于扩展的解决方案，为公共卫生做出重大贡献 解决当前LTBI护理连续体中持续存在的重大差距。

SUMMARY/ABSTRACT In 2015, tuberculosis (TB) surpassed HIV as the number one cause of infectious disease deaths worldwide. In the U.S., California has the highest incidence and largest number of TB cases in the nation, comprising nearly one-quarter of all new active TB cases in 2017. More than 80% of active TB disease in the U.S. is due to reactivation, which could be prevented via screening and treatment of LTBI. Yet, adoption of the latent TB screening and treatment guidelines has been extremely poor. The screening guidelines are inefficient and rely on data that are almost never available to clinicians, and the treatment guidelines are confusing. Further, treatment initiation and completion rates are low for patients with LTBI and the barriers to successful treatment are poorly understood. To address these missed opportunities, we will use expansive electronic health record data across 2 of the largest healthcare institutions in California as well as qualitative data from LTBI stakeholders to conduct three specific aims. For Aim 1, we will look at current gaps in screening practices for LTBI, as well as gaps in the guidelines themselves, by simulating what it would look like if screening was being perfectly implemented. To do this, we will collect laboratory testing data for LTBI stratified by characteristics of interest and will use modeling to estimate the number of TB cases prevented by current screening as a measure of effectiveness. We will use simulation models to estimate effectiveness of optimal screening. For Aim 2, we will develop and validate new screening and treatment guidelines for LTBI based on variables widely available in electronic health records. First, we will use machine learning and traditional regression to identify risk factors for positive LTBI tests and reactivation TB. Next, we will assign risk scores to each risk factor, and will use joint probability analyses to identify populations at greatest need for screening and treatment. To estimate performance of the newly proposed strategy, we will use simulation modeling. For Aim 3, we will develop and pilot a culturally-tailored educational video intervention to improve LTBI treatment initiation and completion. We will first identify barriers to treatment adherence through qualitative interviews with patients and providers and will subsequently develop a short video based on findings from the interview. We will perform an individually randomized efficacy trial to assess impact of the intervention on initiation and treatment completion rates. The approach is innovative because we propose a complete re-framing of the current U.S. LTBI control strategy in a way that dramatically enhances ease-of-use for frontline providers. Results of this work will make a significant contribution to public health by providing low-cost and easily expandable solutions to address ongoing and substantial gaps in the current LTBI care continuum.