As patient satisfaction is heavily linked to their choice of provider and medical outcomes, hospital administrations routinely consider a bevy of factors to improve patient satisfaction. These considerations are complex, so targeting the most important areas for improvement is challenging. However, consumers’ online reviews of their hospital experience provide a vital lens into the factors associated with their satisfaction. In this study, we use a large dataset of Facebook reviews to construct a taxonomy of potential service attributes that consumers discuss online. We find partial overlap between this taxonomy and prior works and more traditional survey measures; the specific mix of service attributes found in these reviews is unique. Next, we utilize regression modeling to determine which service attributes are most closely associated with star ratings, which we use to measure overall satisfaction. This study demonstrates that mentions of waiting times, treatment effectiveness, communication, diagnostic quality, environmental sanitation, and cost considerations tend to be most associated with patients’ overall ratings. Finally, we construct text analyses to rapidly detect consumers’ mentions of these service attributes in an automated manner. We derive a set of “smoke terms,” or terms especially prevalent in posts that mention specific service attributes. We find that these are generally non-emotive terms, indicating limited utility of traditional sentiment analysis. Managerially, this information helps to prioritize the areas in greatest need of improvement. Additionally, generating smoke terms for each service attribute aids health care policy makers and providers in rapidly monitoring concerns and adjusting policies or resources to improve service.

中文翻译：

---

Facebook 医院评论：自动服务质量检测和与患者满意度的关系

由于患者满意度与他们选择的提供者和医疗结果密切相关，因此医院管理部门通常会考虑一系列因素来提高患者满意度。这些考虑因素很复杂，因此针对最重要的改进领域具有挑战性。然而，消费者对其医院体验的在线评论提供了一个重要的视角来了解与他们的满意度相关的因素。在这项研究中，我们使用 Facebook 评论的大型数据集来构建消费者在线讨论的潜在服务属性的分类。我们发现这种分类法与以前的工作和更传统的调查措施有部分重叠；在这些评论中发现的特定服务属性组合是独一无二的。下一个，我们利用回归建模来确定哪些服务属性与星级评分最密切相关，我们用它来衡量整体满意度。这项研究表明，对等待时间、治疗效果、沟通、诊断质量、环境卫生和成本考虑的提及往往与患者的总体评分最相关。最后，我们构建文本分析，以自动方式快速检测消费者对这些服务属性的提及。我们推导出一组“烟雾术语”，或在提及特定服务属性的帖子中特别流行的术语。我们发现这些通常是非情感术语，表明传统情感分析的效用有限。在管理上，这些信息有助于确定最需要改进的领域的优先级。此外，