Project 2

项目摘要--项目2 服务不足的乳腺癌筛查人群，包括主要由种族/族裔少数群体、收入较低、受教育程度较低和农村人口组成的人群，乳腺癌发病率和死亡率负担仍高于相应人群。这些人群在异常筛查后的随访率往往较低，更多的癌症遗漏，以及诊断时更晚期的疾病。造成不平等的原因可能是多方面的，不仅包括妇女一级的促成因素，还包括社区一级的健康社会决定因素和设施一级的因素，这些因素影响获得和使用高质量的筛查、及时诊断评估和治疗。国家少数民族健康和健康差异研究所代表美国国立卫生研究院报告说，实现健康公平的一个主要障碍是，以前的差异研究工作侧重于个人促成因素，而不是社区或医疗保健提供因素。随着包括人工智能(AI)在内的新筛查技术在临床实践中的迅速采用，了解乳房成像设施层面的不平等驱动因素的影响尤为重要。如果较新的技术不能在社区间公平传播，持续存在的乳腺癌差异可能会进一步加剧。我们的总体项目目标是确定导致乳腺癌筛查差异的可修改的乳腺成像设备级别的因素。使用观察性队列研究设计和模拟建模，我们将探索旨在增加常规筛查的获取和使用以及有针对性地使用人工智能以提高成像解释准确性的有针对性的设施水平变化如何促进筛查结果的更大公平性。我们将利用美国八个地区乳腺成像登记机构的强大、纵向、多水平的乳腺癌监测联盟数据，以实现以下具体目标：目标1)执行多水平分析，以确定导致筛查性能和结果差异的设施水平因素(例如，现场技术)。目的2)使用回顾性配对病例对照设计和五种商业可用的人工智能技术，评估商业上可用的用于自动乳房X光检查解释的人工智能工具是否可以帮助表现不佳的机构达到或超过国家乳房X光检查性能基准。目的3)使用已建立的三个微观模拟模型和AIMS 1和2的结果，估计为美国总体筛查人群和未得到充分服务的亚人群实施设施级护理质量干预措施(例如，提高绩效的人工智能)的长期、人群水平的好处、危害和成本。提高低绩效设施的护理质量有可能使人口一级的筛查效益更大、危害更小，同时也促进健康公平。

PROJECT SUMMARY – Project 2 Underserved breast cancer screening populations, including those that are predominantly composed of racial/ethnic minorities, lower income, lower educated, and rural populations, continue to have a higher breast cancer morbidity and mortality burden than their counterparts. These populations tend to have lower follow-up rates after abnormal screening, more missed cancers, and more advanced stage disease at the time of diagnosis. Drivers of inequities are likely multi-factorial and include not only woman-level enabling factors but also neighborhood-level social determinants of health and facility-level factors that influence access to and use of high quality screening, timely diagnostic evaluation, and treatment. The National Institute of Minority Health and Health Disparities, on behalf of the NIH, reports that a major barrier in achieving health equity is that prior disparities research efforts have focused on individual enabling factors rather than neighborhood or healthcare delivery factors. Understanding the impact of breast imaging facility-level drivers of inequities is particularly important as new screening technologies, including artificial intelligence (AI), are rapidly adopted in clinical practice. If newer technologies do not diffuse equitably across communities, persistent breast cancer disparities may be further exacerbated. Our overall project objective is to identify modifiable breast imaging facility-level factors that drive breast cancer screening disparities. Using an observational cohort study design and simulation modeling, we will explore how targeted facility-level changes that aim to increase access to and use of routine screening and targeted use of AI for improved imaging interpretation accuracy can promote greater equity in screening outcomes. We will leverage the robust, longitudinal, multi-level Breast Cancer Surveillance Consortium data across eight regional U.S. breast imaging registries to pursue the following specific aims: Aim 1) Perform multi-level analyses to identify facility-level factors (e.g., on-site technologies) that drive disparities in screening performance and outcomes. Aim 2) Using a retrospective matched case control design and five commercially available AI technologies, evaluate whether commercially available AI tools for automated mammography interpretation can aid low-performing facilities to meet or exceed national mammography performance benchmarks. Aim 3) Using three established microsimulation models and results from Aims 1 and 2, estimate the long-term, population-level benefits, harms, and costs of enacting facility-level quality-of-care interventions (e.g., AI for higher performance) for the overall U.S. screening population and for underserved subpopulations. Elevating the quality-of-care at low-performing facilities has the potential to tip the balance towards greater screening benefits and less harms at the population-level, while also promoting health equity.