Straub Heather, Adams Marci, Silver Richard K
Division of Maternal Fetal Medicine, NorthShore University HealthSystem, 2650 Ridge Ave, Walgreens Building, Suite 1507, Evanston, IL, 60201, USA,
Matern Child Health J. 2014 Nov;18(9):2134-40. doi: 10.1007/s10995-014-1461-8.
To explore the potential of an integrated outpatient electronic health record (EHR) for preconception health optimization. An automated case-finding EHR-derived algorithm was designed to identify women of child-bearing age having outpatient encounters in an 85-site, integrated health system. The algorithm simultaneously cross-referenced multiple discrete data fields to identify selected preconception factors (obesity, hypertension, diabetes, teratogen use including ACE inhibitors, multivitamin supplementation, anemia, renal insufficiency, untreated sexually transmitted infection, HIV positivity, and tobacco, alcohol or illegal drug use). Surveys were mailed to a random sample of patients to obtain their self-reported health profiles for these same factors. Concordance was assessed between the algorithm output, survey results, and manual data abstraction. Between 8/2010-2/2012, 107,339 female outpatient visits were identified, from which 29,691 unique women were presumed to have child-bearing potential. 19,624 (66 %) and 8,652 (29 %) had 1 or ≥2 health factors, respectively while only 1,415 (5 %) had none. Using the patient survey results as a reference point, health-factor agreement was similar comparing the algorithm (85.8 %) and the chart abstraction (87.2 %) results. Incorrect or missing data entries in the EHR encounters were largely responsible for discordances observed. Preconception screening using an automated algorithm in a system-wide EHR identified a large group of women with potentially modifiable preconception health conditions. The issue most responsible for limiting algorithm performance was incomplete point of care documentation. Accurate data capture during patient encounters should be a focus for quality improvement, so that novel applications of system-wide data mining can be reliably implemented.
为探索整合式门诊电子健康记录（EHR）在孕前健康优化方面的潜力。设计了一种基于电子健康记录的自动病例查找算法，以识别在一个拥有85个站点的整合医疗系统中进行门诊就诊的育龄妇女。该算法同时交叉引用多个离散数据字段，以识别选定的孕前因素（肥胖、高血压、糖尿病、致畸剂使用，包括血管紧张素转换酶抑制剂、多种维生素补充、贫血、肾功能不全、未治疗的性传播感染、HIV阳性以及烟草、酒精或非法药物使用）。向随机抽取的患者样本邮寄调查问卷，以获取他们关于这些相同因素的自我报告健康状况。评估算法输出、调查结果和手工数据提取之间的一致性。在2010年8月至2012年2月期间，共识别出107339次女性门诊就诊，其中推测有29691名独特女性具有生育潜力。分别有19624名（66%）和8652名（29%）女性有1种或≥2种健康因素，而只有1415名（5%）女性没有健康因素。以患者调查结果为参考点，算法结果（85.8%）和图表提取结果（87.2%）的健康因素一致性相似。电子健康记录就诊中不正确或缺失的数据录入是观察到不一致的主要原因。在全系统电子健康记录中使用自动算法进行孕前筛查，识别出了一大批患有潜在可改变的孕前健康状况的女性。限制算法性能的最主要问题是护理点文档不完整。在患者就诊期间准确的数据采集应成为质量改进的重点，以便能够可靠地实施全系统数据挖掘的新应用。
