Improving dermatology access by direct-to-patient teledermatology and computer-assisted diagnosis

背景：在退伍军人事务部，皮肤科的获取仍然是一个重要的问题 (VA)尤其是在COVD-19大流行期间。为了满足这一需求，VA将部署一个异步 远程皮肤病学移动的应用程序-My VA Images-允许新皮肤病患者安全地提交病史 和他们的皮肤照片进行评估该应用程序最终也可能为患者提供一个提交 通过人工智能（AI）驱动的计算机视觉， 皮肤科医生。 意义：该项目解决了以下差距：1）直接面向患者的远程皮肤病学对 获得皮肤科治疗的机会以及患者和卫生保健提供者对这种护理的满意度， 2）目前还没有开发出人工智能驱动的计算机视觉工具， 针对患者生成的图像进行了验证; 3）大型医疗保健组织（如VA）的准备情况，以及 他们的利益相关者参与直接面向患者的远程皮肤病学和人工智能是未知的。 创新和影响：将测试两项相关创新：1）直接面向患者的远程皮肤病学， 患者和2）通过AI驱动的计算机视觉评估患者提交的皮肤图像。这些单独 有可能将远程访问转变为VA中的专家皮肤护理，并且一起具有潜在的协同作用。 在项目结束时，我们预计有一个系统的了解如何直接对病人 技术的性能，以及在此之前VA需要解决的操作差距 技术可以在企业范围内实施。目标是建立一个关键的学术和业务 基金会安全地走向一个变革的愿景，退伍军人将不再被束缚在一个固定的时间 和地方的照顾，而是将有选择的自我导向，方便和快速的访问专家级 皮肤科护理，无论何时何地，他们需要它。 具体目标：1。评估直接面向患者的远程皮肤病学对访问和卫生系统的影响 利用率2.对患者提交的图像进行评估、改进和增强计算机辅助评价。 3.评估退伍军人事务部和退伍军人接受实施直接对病人护理的准备情况。 方法：目标1将使用I型混合语用学研究设计来比较直接对 患者远程皮肤病学干预相对于通常的面对面和通常的咨询远程皮肤病学 转介，主要通过VA中央数据仓库的数据来衡量访问。目标1和3将衡量 患者满意度和准备改变使用调查工具和访谈。目标2将包括 测试，培训和改进人工智能驱动的计算机视觉和测量与 皮肤科医生人口：退伍军人提到皮肤科在三个VA医疗设施。干预措施： 符合条件和医学上适当的患者将可以选择将病史和图像提交给 皮肤科使用My VA Images应用程序。比较：干预将与常规护理进行比较（在- 个人和咨询性远程皮肤病学）团体。我们还将比较两个人工智能驱动的计算机视觉 有皮肤科医生诊断的模特5年期间的结果：1）时间和 皮肤科护理的地理可及性; 2）患者满意度; 3）AI与皮肤科医生的一致性 4）组织和患者准备远程和计算机辅助皮肤病护理;以及 5)直接治疗新患者远程皮肤科流程的实施和可持续性。 后续步骤/实施：该项目的成功完成将为VA的互联护理办公室提供 以及其他负责利用关键数据增强专科护理可及性的办公室， 扩大直接面向患者的异步远程皮肤病学项目。该项目还将提供VA 利用关键数据来评估人工智能驱动的计算机视觉在未来远程护理战略中的作用。

Background: Access to dermatology remains a significant problem in the Department of Veterans Affairs (VA), particularly during the COVD-19 pandemic. To address this need, VA will deploy an asynchronous teledermatology mobile app-My VA Images-which allows new dermatology patients to securely submit history and photos of their skin for evaluation. The app may also eventually provide a conduit for patients to submit skin images at will for analysis and triage by artificial intelligence (AI)-powered computer vision to a dermatologist. Significance: This project addresses the following gaps: 1) The impact of direct-to-patient teledermatology on access to dermatology and on the satisfaction with such care by both patients and health care providers has not been systematically studied; 2) Currently no AI-powered computer vision tool has been developed and validated for patient-generated images; 3) The readiness of large healthcare organizations, such as VA, and their stakeholders to engage in direct-to-patient teledermatology and AI is unknown. Innovation and Impact: Two related innovations will be tested: 1) Direct-to-patient teledermatology for new patients and 2) Evaluation of patient-submitted skin images by AI-powered computer vision. These separately have the potential to transform remote access to expert skin care in VA and together are potentially synergistic. At the conclusion of the project, we anticipate having a systematic understanding of how direct-to-patient technologies perform and of the operational gaps that will need to be addressed by VA before these technologies can be implemented enterprise-wide. The goal is to establish a critical scholarly and operational foundation to safely move toward a transformative vision where Veterans will no longer be tied to a fixed time and place for care, but instead will have the choice of self-directed, convenient and rapid access to expert-level dermatology care wherever and whenever they need it. Specific Aims: 1. Assess the impact of direct-to-patient teledermatology on access and health system utilization. 2. Assess, refine and augment computer-assisted evaluation of patient-submitted images. 3. Assess readiness of VA and Veterans' acceptance to implement direct-to-patient care. Methodology: Aim 1 will use a Type I hybrid pragmatic study design to compare the impact of the direct-to- patient teledermatology intervention relative to usual in-person and usual consultative teledermatology referrals, measuring access chiefly by data from VA's Central Data Warehouse. Aims 1 and 3 will measure patient satisfaction and readiness for change using survey instruments and interviews. Aim 2 will include both testing, training and refinement of the AI-powered computer vision and measure concordance with dermatologists. Population: Veterans referred to Dermatology at three VA medical facilities. Intervention: Eligible and medically appropriate patients will be offered the option to submit history and images to Dermatology using the My VA Images app. Comparison: The intervention will be compared to usual care (in- person and consultative teledermatology) groups. We will also compare two AI-powered computer vision models with dermatologist diagnoses. Outcomes over a 5-year period: 1) Multiple measures of temporal and geographic access to dermatologic care; 2) Patient satisfaction; 3) Concordance of AI with dermatologist diagnoses; 4) Organizational and patient-readiness for remote and computer-assisted dermatologic care; and 5) Implementation and sustainability of the direct-to-new patient teledermatology process. Next Steps/Implementation: Successful completion of the project will provide VA’s Office of Connected Care and other offices tasked with enhancing access to specialty care with critical data that will justify further expansion of the direct-to-patient asynchronous teledermatology program. The project will also provide VA with critical data to evaluate the role of AI-powered computer vision in future remote care strategies.