EHR-Based Prediction Models to Improve PrEP Use in Community Health Centers

项目概要 艾滋病毒新发感染率异常高，而暴露前预防 (PrEP) 的采用率却很低 美国黑人、拉丁裔和保险不足的个人 安全网社区健康的医疗保健提供者 中心 (CHC) 为保险不足率较高的不同种族人群提供护理。然而， 提供者列举了 PrEP 处方的障碍，包括缺乏识别 PrEP 候选人的工具。没有 帮助提供者识别有 HIV 感染风险的患者并在适当时开出 PrEP 处方的实用工具， PrEP 在人群层面的益处不太可能实现。使用数据的电子临床决策支持 嵌入患者的电子健康记录（EHR）有可能满足这一需求。 EHR 包含丰富的 有助于识别感染艾滋病毒高风险患者的数据，包括人口统计、诊断、检测 健康的模式、处方和社会决定因素。在我们之前的工作中，我们开发并验证了 使用来自马萨诸塞州和加利福尼亚州两个大型医疗保健系统的 EHR 数据进行预测模型，其中 1.1 和 430 万患者群体，以确定感染 HIV 的高风险患者。这些机器 学习模型具有很强的预测性能，C 统计量高达 0.91。本提案的目的 是为了检验这样一个假设：包含 HIV 风险预测模型的临床决策支持工具可以 帮助提供者识别 HIV 感染高危患者并改进 PrEP 处方。我们的学习环境是 拥有 280 万患者的全国 CHC 网络 (OCHIN)。我们将首先定制我们的艾滋病毒预测模型 这个诊所网络，然后与提供者进行形成性工作，以告知我们警报的开发和 额外的 PrEP 决策支持工具将是有效且受欢迎的。研究团队包括以下领域的专家 CHC 中的 HIV、PrEP 实施、预测分析和医疗保健服务。我们的具体目标是 1) 优化使用 EHR 数据识别潜在 PrEP 候选者的预测模型， 社会经济和地理上不同的患者群体； 2) 探索供应商的观点 PrEP 处方的障碍以及他们对 PrEP 决策支持的偏好，以便为制定 基于 EHR 的 CHC 决策支持工具； 3) 进行试点以评估可行性、可接受性、 基于 EHR 的临床决策支持干预对 CHC 中 PrEP 相关护理的初步影响。 我们将评估对整个 PrEP 护理连续体指标的影响，包括处方、持续性、临床 艾滋病毒和其他性传播感染的监测、检测和诊断。这个提议是 其创新之处在于使用预测分析和临床决策支持来优化 PrEP。该项目是 意义重大，因为我们的干预措施将可扩展到全国的 CHC 和其他医疗保健系统 还因为它涉及通过在高危地区扩大 PrEP 来结束艾滋病毒流行的联邦倡议 发生率设置。预期结果是整群随机试验的基础，以测试 EHR 是否 基于 PrEP 的决策支持可以在国家 CHC 网络中预防新的 HIV 感染。

PROJECT SUMMARY Rates of new HIV infections are disproportionately high, and uptake of preexposure prophylaxis (PrEP) low, in Black, Latino, and underinsured individuals in the U.S. Healthcare providers at safety net community health centers (CHCs) provide care to racially diverse populations with high rates of underinsurance. However, providers cite barriers to PrEP prescribing, including lack of tools to identify candidates for PrEP. Without practical tools to help providers identify patients at risk for HIV infection and prescribe PrEP when appropriate, the population-level benefits of PrEP are unlikely to be realized. Electronic clinical decision support using data embedded in patients’ electronic health records (EHRs) has the potential to fulfill this need. EHRs contain rich data that can help identify patients at high risk of HIV acquisition, including demographics, diagnoses, testing patterns, prescriptions, and social determinants of health. In our prior work, we developed and validated prediction models using EHR data from two large healthcare systems in Massachusetts and California, with patient populations of 1.1 and 4.3 million, to identify patients at high risk for incident HIV. These machine learning models had strong predictive performance, with C-statistics up to 0.91. The objective of this proposal is to test the hypothesis that a clinical decision support tool that incorporates an HIV risk prediction model can help providers identify patients at high risk for HIV infection and improve PrEP prescribing. Our study setting is a national network of CHCs with 2.8 million patients (OCHIN). We will first tailor our HIV prediction models to this clinic network, and then conduct formative work with providers to inform our development of alerts and additional PrEP decision support tools that will be effective and welcomed. The study team includes experts in HIV, PrEP implementation, predictive analytics, and healthcare delivery in CHCs. Our specific aims are to 1) optimize prediction models that use EHR data to identify potential PrEP candidates in racially, socioeconomically, and geographically diverse patient populations; 2) explore providers’ perspectives on barriers to PrEP prescribing, and their preferences for PrEP decision support, to inform development of an EHR-based decision support tool for CHCs; and 3) conduct a pilot trial to assess the feasibility, acceptability, and preliminary impact of an EHR-based clinical decision support intervention on PrEP-related care in CHCs. We will assess impact on metrics across the PrEP care continuum, including prescriptions, persistence, clinical monitoring, and tests and diagnoses of HIV and other sexually transmitted infections. This proposal is innovative in its use of predictive analytics and clinical decision support to optimize PrEP. The project is significant because our intervention will be scalable across CHCs nationally and to other healthcare systems with EHRs, and because it addresses the federal initiative to end the HIV epidemic by scaling up PrEP in high- incidence settings. The expected outcome is the foundation for a cluster randomized trial to test whether EHR- based decision support for PrEP can prevent new HIV infections in a national network of CHCs.