Using electronic medical record (EMR) data to examine the impact of prenatal drug exposure: Evaluating availability, accuracy and utilty of exposure information

具体目标 拟议研究的总体目标是检查产前检查的可用性、准确性和实用性 现有电子病历(EMR)中的物质暴露数据。 关于怀孕期间使用药物对孩子结局的影响的研究依赖于大量的前瞻性研究， 提供关于暴露和结果的准确数据的纵向研究。这导致了知识的产生 严重滞后，特别是在阿片类药物迅速流行和最近合法化的情况下 商业级大麻。基于人口的数据集产生了流行率和更普遍的 信息，但不包含足够精细的暴露信息，以可靠地回答特定问题 研究问题。单个患者的EMR包含丰富的临床信息，这些信息可能会被使用 然而，出于研究目的，这种与产前暴露相关的潜力在很大程度上还没有研究过。我们的工作，以及 其他人的工作，展示了使用电子病历数据来回答有关 产前接触物质对儿童结局的影响，同时也强调了限制和 未知数。急诊室提供了将儿科门诊记录与多达三种其他类型的 EMRs(新生儿住院记录、产妇住院分娩记录和产妇门诊产前护理 记录)，为许多患者扩大了有关暴露的可用信息。 我们的研究将涉及以下具体目标： 1.要准确确定在四种电子病历类型中的每一种中都有哪些暴露数据，以及 频率和曝光数据源之间的一致性。 2.从不同的数据源构建复合暴露变量，并检验其程度 每一项指标：a)与脐带测试结果相关，b)与儿童结局相关。 3.确定预测暴露数据的可用性、准确性和实用性的因素。 4.开发耐久的研究方法材料，包括暴露变量编码 算法。 这项研究将利用来自大型但相当农村社区的儿科门诊诊所的急诊医生， 和一个大型的城市学术儿科门诊服务，这将产生大量的种族和 社会经济多样性，将涉及美国最常见的两个EMR系统(Epic和Cerner)。一个 将在两个地点确定至少2500名6岁及以下的儿科门诊患者进行研究 包括，至少40%的人预计曾接触过某种类型的产前物质(烟草、酒精、 滥用处方药和/或非法药物)。为每个参与者检查的四种电子病历类型将是 儿科门诊病历、新生儿住院病历、产妇住院分娩病历和产妇住院分娩病历 门诊产前护理记录。除了背景、合并症和结果信息外，任何 有关产前物质暴露的信息将摘要为：ICD-10诊断和程序 密码、尿药筛查、脐带血/组织/胎粪分析、附加生化测试、自我 关于药物使用/暴露的报告/筛查问题、处方药清单、新生儿停药 评估/治疗，以及备注字段中的其他信息。具体目标将通过以下方式进行评估 描述性和双变量统计、敏感性/特异性分析和预测性建模 产品开发耐久的研究方法材料，包括暴露变量编码 算法。 如果我们能够找出方法，EMR数据可以可靠地用于不仅识别儿科患者 产前物质暴露，而且开发用于研究目的的有效暴露变量， 可以招募研究人员参与检查产前暴露和以后结局之间的联系 在结果评估点上。这将大大缩短在这方面进行有效研究所需的时间 通过经过验证的方法利用现有数据。研究结果还将提供有价值的方法 为检查产前影响的最终黄金标准前瞻性纵向研究提供信息的见解 暴露在物质中。

SPECIFIC AIMS The overall goal of the proposed study is to examine the availability, accuracy, and utility of prenatal substance exposure data in the existing electronic medical records (EMR). Research on the impact of substance use during pregnancy on child outcomes has relied on large prospective, longitudinal studies to provide accurate data on exposures and outcomes. This has resulted in knowledge lagging significantly, especially problematic with the rapidly evolving opioid epidemic and recent legalization of commercial-grade marijuana. Population-based datasets have produced prevalence and more general information, but do not contain exposure information that is granular enough to reliably answer specific research questions. Individual patient EMRs contain a wealth of clinical information that can potentially be used for research purposes, however this potential related to prenatal exposures is largely unstudied. Our work, and the work of others, has demonstrated the possibilities for using EMR data to answer questions about the impact of prenatal substance exposures on child outcomes, while also highlighting the limitations and unknowns. EMRs provide the opportunity of linking pediatric outpatient records with up to three other types of EMRs (newborn inpatient record, maternal inpatient delivery record, and maternal outpatient prenatal care record), expanding available information about exposures for many patients. Our study will involve the following Specific Aims: 1. To determine exactly what exposure data is available in each of the four EMR types, with what frequency, and the congruence between exposure data sources. 2. To construct composite exposure variables from different data sources and examine the degree to which each: a) correlates with cord test results, and b) correlates with child outcomes. 3. To determine factors that predict availability, accuracy, and utility of exposure data. 4. To develop enduring study methodology materials including exposure variable coding algorithms. The study will utilize the EMRs from both a large but fairly rural community based pediatric outpatient practice, and from a large urban academic pediatric ambulatory service, which will produce substantial racial and socioeconomic diversity and will involve the two most common EMR systems in the U.S (Epic and Cerner). A minimum of 2,500 pediatric outpatients age 6 and under will be identified across the two sites for study inclusion, with at least 40% expected to have had some type of prenatal substance exposure (tobacco, alcohol, prescription drugs of abuse, and/or illicit drugs). The four EMR types examined for each participant will be the linked pediatric outpatient record, newborn inpatient record, maternal inpatient delivery record, and maternal outpatient prenatal care record. In addition to background, comorbidity, and outcomes information, any information about prenatal substance exposure will be abstracted to include: ICD-10 diagnosis and procedure codes, urine drug screens, cord blood/tissue/meconium analysis, additional biochemical testing, self- report/screening questions for drug use/exposure, lists of prescribed medications, newborn withdrawal assessment/treatment, and additional information from notes fields. Specific Aims will be evaluated with descriptive and bivariate statistics, sensitivity/specificity analyses, and predictive modeling, with the end product the development of enduring study methodology materials including exposure variable coding algorithms. If we are able to identify ways EMR data could be reliably used to not only identify pediatric patients with prenatal substance exposure, but also to develop valid exposure variables to use for research purposes, participants in studies examining the links between prenatal exposures and later outcomes could be recruited at the point of outcome assessment. This would significantly shorten the time needed to do valid studies in this field by leveraging existing data via validated methods. Study findings will also provide valuable methodological insights to inform the ultimate gold-standard prospective longitudinal studies examining the impacts of prenatal substance exposure.