The Center for Health Research, Kaiser Permanente Northwest, 3800 N. Interstate Avenue, Portland, OR 97227, United States.
Vaccine. 2013 Jun 12;31(27):2898-903. doi: 10.1016/j.vaccine.2013.03.069. Epub 2013 Apr 30.
The need for research on the safety of vaccination during pregnancy is widely recognized. Large, population-based data systems like the Vaccine Safety Datalink (VSD) may be useful for this research, but identifying pregnancies using electronic medical record (EMR) and claims data can be challenging.
We modified an existing data processing algorithm to identify pregnancies within seven of the ten VSD sites. We validated the algorithm by calculating the agreement in pregnancy outcome type, end date, and gestational age between the algorithm and manual medical record review. At each participating site, we randomly sampled 15 episodes within four outcome type strata (live births, spontaneous abortions, elective abortions, and other pregnancy outcomes) for a total of 60 episodes per site. We also developed and validated methods to link mothers to their infants in the electronic data.
We identified 595,929 pregnancy episodes ending in 2002 through 2006 among women 12 through 55 years of age. Of these pregnancies, 75% ended in live births, 12% in spontaneous abortions, and 9% in elective abortions. We were able to confirm a pregnancy within 28 days of the algorithm-estimated pregnancy start date for 99% of live births, 93% of spontaneous abortions, 92% of elective abortions, and 90% of other outcomes sampled. The agreement between the algorithm-identified and the abstractor-identified outcome date ranged from 70% (elective abortion) to 96% (live birth) depending on outcome type. When gestational age was available in the EMR, agreement ranged from 82% (other) to 98% (live birth) depending on outcome type. We confirmed 100% of the 350 sampled mother-infant linkages with manual medical record review.
The VSD algorithm accurately identifies pregnancy episodes and mother-infant pairs across participating sites. Additional manual record review may be needed to improve the precision of the pregnancy date estimates depending on specific study needs. These algorithms will allow us to conduct large, population-based studies of the safety of vaccination during pregnancy.
人们广泛认识到有必要对孕期接种疫苗的安全性进行研究。疫苗安全数据链（VSD）等大型基于人群的数据系统可能对此类研究很有用，但使用电子病历（EMR）和理赔数据来确定妊娠可能具有挑战性。
我们修改了现有的数据处理算法，以在十个 VSD 地点中的七个地点内识别妊娠。我们通过计算算法和手动病历审查之间妊娠结局类型、结束日期和胎龄的一致性来验证算法。在每个参与地点，我们在四个结局类型（活产、自然流产、人工流产和其他妊娠结局）内随机抽取 15 个病例，每个地点共抽取 60 个病例。我们还开发并验证了将母亲与其婴儿在电子数据中进行关联的方法。
我们在 2002 年至 2006 年期间，从 12 岁至 55 岁的女性中确定了 595929 例妊娠结局。这些妊娠中，75%以活产结束，12%以自然流产结束，9%以人工流产结束。我们能够在算法估计的妊娠开始日期后 28 天内确认 99%的活产、93%的自然流产、92%的人工流产和 90%的其他抽样结局的妊娠。算法确定的和摘要确定的结局日期之间的一致性因结局类型而异，范围从 70%（人工流产）到 96%（活产）。当 EMR 中有胎龄信息时，一致性因结局类型而异，范围从 82%（其他）到 98%（活产）。我们通过手动病历审查确认了 350 个抽样母婴关联中的 100%。
VSD 算法可准确识别参与地点的妊娠病例和母婴对。根据特定的研究需求，可能需要额外的病历记录审查来提高妊娠日期估计的精度。这些算法将使我们能够开展大型基于人群的孕期接种疫苗安全性研究。
