Analyzing Online Reviews to Evaluate Quality of Care at Substance Use Disorder Treatment Facilities

在美国，估计有2000万成年人被诊断为吸毒 精神障碍(SUD)。患者面临的一个主要挑战是找出最合适和最好的治疗方法 结果。患者在门诊和住院设施接受护理，但质量指标处于这些水平 个别设施不向公众开放。 在数字数据时代，互联网和P2P资源往往是第一个 个人希望找到有关医疗资源的信息。谷歌和Yelp等网站上的在线评论 提供有关医疗设施的说明，并给予易于理解的星级评级，范围从1到 五星级的。这些对医疗设施的审查提供了有关患者体验、结构(例如 物理 教育) 设施、组织特征、支付方式)、流程(例如诊断、治疗、患者 和结果(例如，知识、与健康相关的生活质量、发病率、死亡率)。 先前的工作已经证明，在线对医院的评级与国家评级相关 医院消费者对医疗服务提供者和系统的评估(HCAHPS)调查。然而，几乎没有什么工作 评估了未经验证、自发生成但可广泛访问的SUD在线评论的效用 治疗设施。这些审查可以填补一个利基市场，为患者及其家人提供 关于对他们来说最重要的问题和经验的有用信息。或者这些评论 可能是有限的，并提供错误的信息或只提供有偏见的观点。到目前为止，人们对 这一新兴数据源的潜在价值或有用性。 对于这项建议，我们的目标是研究美国SUD治疗设施的在线审查，以确定 被报告为对患者和家属最重要的护理领域。对于目标1，我们将首先 提取大约50,000条关于SUD治疗设施的在线评论，并使用以下工具对它们进行编码 定性方法，包括使用机器学习进行手动编码和大数据分析 自然语言处理。我们假设这些在线叙述将包括关于以下的定性数据 患者体验以及南加州大学提供的护理类型和质量(例如循证治疗) 治疗设施。对于目标2，我们将评估星级评级如何区分设施与结构 以及《全国药物滥用治疗服务调查》中报告的处理措施。 总体而言，新兴的在线数据资源有可能提供有关关键问题的新信息 受阿片类药物影响的弱势人群，更广泛地说，受到SUD危机的影响。我们的项目力求严格地 研究这些数以百万计的个人查看的以患者为中心的新数据源，以便更好地 了解患者和家属的需求。这些努力可以为今后的工作奠定基础。 在为至关重要的医疗保健资源--SUD治疗设施制定质量措施时。

In the United States (US), an estimated 20 million adults have been diagnosed with substance use disorder (SUD). A major challenge for patients is to identify the most appropriate treatment with the best outcomes. Patients receive care in outpatient and inpatient facilities yet quality metrics at the level of these individual facilities are not publicly available. In the era of digital data, the Internet and peer-to-peer resources are often the first place where individuals look to find information about healthcare resources. Online reviews on sites like Google and Yelp provide narratives about healthcare facilities and assign easily interpretable star ratings ranging from one to five stars. These reviews of healthcare facilities provide information about patient experience, structure (e.g. physical education) facility, organizational characteristics, payment methods), process (e.g. diagnosis, treatment, patient and outcomes (e.g. knowledge, health-related quality of life morbidity, mortality). Prior work has demonstrated that online ratings of hospitals correlate with ratings from the national Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey. Little work however has evaluated the utility of unvalidated, spontaneously generated, but widely accessible online reviews of SUD treatment facilities. These reviews could fill a niche with providing patients and their family members with useful information about the questions and experiences that are most important to them. Or these reviews could be limited and provide misinformation or only biased perspectives. To date, less is known about the potential value or usefulness of this emerging data source. For this proposal we aim to study online reviews of SUD treatment facilities in the US to identify the areas of care that are reported as most important to patients and family members. For Aim 1, we will first extract approximately 50,000 online reviews of SUD treatment facilities and code them for themes using qualitative methodologies that involve manual coding and big data analytics using machine learning and natural language processing. We hypothesize that these online narratives will include qualitative data about patient experience and the type and quality of care (e.g. evidence-based treatments) provided at SUD treatment facilities. For Aim 2, we will then assess how star ratings differentiate facilities relative to structure and process measures reported in the National Survey of Substance Abuse Treatment Services. Overall, emerging online data resources have the potential to provide new information about a critically vulnerable population affected by the opioid and more broadly the SUD crisis. Our project seeks to rigorously study these new patient-centric data sources viewed by millions of individuals for the purposes of better understanding the needs of patients and family members. These efforts can lay the groundwork for future work in developing measures of quality for a critically important healthcare resource, SUD treatment facilities.