Begum Tahmina, Veeranki Bhagyavalli, Chike Ogenna Joy, Tamang Suzanne, Simard Julia F, Chen Jonathan, Chaichian Yashaar, Mackey Sean, Darnall Beth D, Falasinnu Titilola
University of North Carolina, Charlotte, NC, USA.
Division of Immunology and Rheumatology, Department of Medicine, Stanford University School of Medicine, Stanford, CA, USA.
J Pain. 2025 Mar;28:104775. doi: 10.1016/j.jpain.2025.104775. Epub 2025 Jan 3.
The use of electronic health records (EHR) for chronic pain phenotyping has gained significant attention in recent years, with various algorithms being developed to enhance accuracy. Structured data fields (e.g., pain intensity, treatment modalities, diagnosis codes, and interventions) offer standardized templates for capturing specific chronic pain phenotypes. This study aims to determine which chronic pain case definitions derived from structured data elements achieve the best accuracy, and how these validation metrics vary by sociodemographic and disease-related factors. We used EHR data from 802 randomly selected adults with autoimmune rheumatic diseases seen at a large academic center in 2019. We extracted structured data elements to derive multiple phenotyping algorithms. We confirmed chronic pain case definitions via manual chart review of clinical notes, and assessed the performance of derived algorithms, e.g., sensitivity/recall, specificity, positive predictive value (PPV). The highest sensitivity (67%) was observed when using ICD codes alone, while specificity peaked at 96% with a quadrimodal algorithm combining pain scores, ICD codes, prescriptions, and interventions. Specificity was generally higher in males and younger patients, particularly those aged 18-40 years, and highest among Asian/Pacific Islander and privately insured patients. PPV was highest among patients who were female, younger, or privately insured. PPV and sensitivity were lowest among males, Asian/Pacific Islander, and older patients. Variability of phenotyping results underscores the importance of refining chronic pain phenotyping algorithms within EHRs to enhance their accuracy and applicability. While our current algorithms provide valuable insights, enhancement is needed to ensure more reliable chronic pain identification across diverse patient populations. PERSPECTIVE: This study evaluates chronic pain phenotyping algorithms using electronic health records, highlighting variability in performance across sociodemographic and disease-related factors.
近年来，使用电子健康记录（EHR）进行慢性疼痛表型分析受到了广泛关注，人们开发了各种算法以提高准确性。结构化数据字段（如疼痛强度、治疗方式、诊断代码和干预措施）为捕获特定的慢性疼痛表型提供了标准化模板。本研究旨在确定从结构化数据元素得出的哪些慢性疼痛病例定义具有最佳准确性，以及这些验证指标如何因社会人口统计学和疾病相关因素而有所不同。我们使用了2019年在一个大型学术中心就诊的802名随机选择的患有自身免疫性风湿病的成年人的EHR数据。我们提取结构化数据元素以得出多种表型分析算法。我们通过对临床记录进行人工图表审查来确认慢性疼痛病例定义，并评估所推导算法的性能，例如敏感性/召回率、特异性、阳性预测值（PPV）。仅使用ICD代码时观察到最高敏感性（67%），而结合疼痛评分、ICD代码、处方和干预措施的四模式算法的特异性最高达到96%。男性和年轻患者，特别是18 - 40岁的患者，特异性通常更高，在亚洲/太平洋岛民和私人保险患者中最高。PPV在女性、年轻患者或私人保险患者中最高。PPV和敏感性在男性、亚洲/太平洋岛民和老年患者中最低。表型分析结果的变异性强调了在EHR中完善慢性疼痛表型分析算法以提高其准确性和适用性的重要性。虽然我们目前的算法提供了有价值的见解，但仍需要改进以确保在不同患者群体中更可靠地识别慢性疼痛。观点：本研究使用电子健康记录评估慢性疼痛表型分析算法，突出了社会人口统计学和疾病相关因素在性能方面的变异性。
