Background Accurate, reliable and efficient measures of pain related presentations are essential to evaluate and improve pain care in the ED. Estimates of pain prevalence on arrival to the emergency department (ED) vary depending on the methods used. Artificial intelligence (AI) approaches are likely to be the future for identifying patients in pain from electronic health records (EHR). However, we need a robust method to identify these patients before this can occur. This study aims to identify patients presenting in pain to the ED using binary classification and to describe the population, treatment and outcomes. Methods This study employs a cross sectional design using retrospective data routinely collected in the EHR at a single ED. A random sample of 10 000 patients was selected for inclusion over three years. Triage nursing assessment underwent binary classification by three expert clinicians. The prevalence of pain on arrival is the primary outcome. Patients with pain were compared to those without pain on arrival regarding demographics, treatment and outcomes. Results The prevalence of pain on arrival was 55.2% (95%CI 54.2%, 56.2%). Patients who presented in pain differed from those without pain in terms of age, country of birth, socioeconomic status, mode of arrival, urgency and discharge destination. The median time to first analgesic medication was 65min (IQR 38, 114 min), and 45.6% (95% CI 44.3%, 46.9%) of patients arriving in pain received analgesic medication. Conclusions The prevalence of pain on arrival compares well with previously reported figures using similar methods. Differences in the cohort presenting in pain compared to the population may represent differences in the prevalence or be an extension of previous bias seen in the documentation of pain. This work has set a rigorous methodology for identifying patients presenting with pain from the EHR. It will form the basis for future applications of AI to identify patients presenting in pain to the ED.

中文翻译：

---

识别出现在成人急诊科疼痛的患者：二元分类任务和患病率描述

背景 疼痛相关表现的准确、可靠和有效的测量对于评估和改善 ED 中的疼痛护理至关重要。到达急诊科 (ED) 时对疼痛患病率的估计因所使用的方法而异。人工智能 (AI) 方法很可能成为从电子健康记录 (EHR) 中识别疼痛患者的未来。但是，我们需要一种可靠的方法来识别这些患者，然后才能发生这种情况。本研究旨在使用二元分类来识别出现在急诊室疼痛的患者，并描述人群、治疗和结果。方法 本研究采用横断面设计，使用在单个 ED 的 EHR 中常规收集的回顾性数据。在三年内选择了 10 000 名患者的随机样本进行纳入。分诊护理评估由三位临床专家进行二元分类。到达时疼痛的发生率是主要结果。在人口统计学、治疗和结果方面，将有疼痛的患者与抵达时没有疼痛的患者进行比较。结果到达时疼痛的发生率为55.2%（95%CI 54.2%, 56.2%）。在年龄、出生国家、社会经济地位、到达方式、紧迫性和出院目的地方面，出现疼痛的患者与没有疼痛的患者不同。首次使用镇痛药物的中位时间为 65 分钟（IQR 38、114 分钟），45.6%（95% CI 44.3%、46.9%）的患者接受镇痛药物治疗。结论 到达时疼痛的发生率与以前使用类似方法报告的数字相比，效果很好。与人群相比，出现疼痛的队列差异可能代表患病率的差异，或者是先前在疼痛记录中看到的偏见的延伸。这项工作为从 EHR 中识别出现疼痛的患者制定了严格的方法。它将成为人工智能未来应用识别急诊科疼痛患者的基础。