Induction of labor is increasing in the United States; the most recent national estimate is that approximately 1 in 5 births follow a labor that was medically induced. Although the reasons for this increase are not entirely clear, labor induction when there is no medical indication to effect delivery may play a role in this increase. The issue of elective induction is particularly important for deliveries at late preterm and early term gestations. Late preterm births (34-36 weeks of gestation) make up 71% of all preterm births in the United States. Although infants who are born at 37-38 weeks of gestation traditionally are considered to be “term,” neonatal morbidity continues to decrease with each week of gestation until 39 weeks. Hence, the American College of Obstetricians and Gynecologists clinical management guidelines require that there be documentation of gestational age of at least 39 weeks at the time of induction in the absence of maternal and/or fetal indications for delivery. In this context, it is an important public health issue to be able to track late preterm and early term births accurately after induced labor where there is no apparent medical reason for delivery and to use that information to inform clinical policy.

In this issue of the American Journal of Obstetrics and Gynecology , Dr Bailit presents findings from data that were gathered by the Ohio Perinatal Quality Collaborative that highlight the limitations of the use of readily available vital statistics information to track induced labor where there are no medical indications. After identifying all scheduled inductions between 36 and 38 completed weeks of gestation in the 20-hospital collaborative, she compared birth certificates with corresponding medical records to determine potential indications for induction. There were no potential indications on 11% of the birth certificates compared with 1% of the medical records. In other words, had birth certificates been used to estimate the rate of nonmedically indicated inductions, the rate would be inflated dramatically when compared with the gold standard of the medical record. Moreover, there was notable variation among hospitals regarding the recording of birth certificate indications for induction compared with hospital records.

There are several possible explanations for the reported differences. Birth certificates do not contain all of the elements that might be construed as indications for induction. Moreover, indications for induction can be found in a variety of places within the medical record, but they might not be found by persons who record information on certificates. The findings in this report are consistent with other literature that demonstrates a general lack of sensitivity for the identification of medical conditions and pregnancy risks on birth certificates when compared with hospital information. Kahn et al reported that the overwhelming majority of births by cesarean delivery that had no indication of a risk factor on the birth certificate did have an indication of risk in the administrative codes on the hospital discharge abstract.

Public health surveillance is the ongoing, systematic collection, analysis, interpretation, and dissemination of data regarding a health-related event for use in public health action to reduce morbidity and mortality rates and to improve health. Improving the quality of care requires timely and accurate information. Knowing the baseline levels of indicators is critical for setting rational goals and monitoring the effects of quality improvement initiatives. At the population level, this is a raison d’être for public health surveillance. Each live birth receives a birth certificate; therefore, the systematic collection of information on these certificates about births functions as a surveillance system. However, as demonstrated by the report of Bailit and other reports, at best, some of the information requires cautious interpretation; at worst, inferences that are generated by the analysis of birth certificate data can be misleading, which calls into question the validity of the use of vital statistics data to inform and monitor quality improvement initiatives.

Will the efforts of state-based quality initiatives be hindered because of a lack of accurate population-based data? As can be seen from the efforts in Ohio, nuanced information about indications for induction can be gathered accurately within institutions, and that information can be used to design interventions and effect change. Obviously, a major limitation to such efforts is that they are resource intensive. The vital statistics system has existing infrastructure that is both ongoing and systematic. Although not explicitly designed to investigate all aspects of perinatal care, which includes indications for inductions where definitions on certificates may differ from definitions of indications found in medical records, it would be unfortunate if we were not able to use the best features of our existing system to monitor and improve the quality of care for women and infants. As Bailit suggests, efforts should be made to understand how information gets from bedside to certificate to inform the establishment of best practices for recording accurate information on certificates. To this end, numerous birth certificate data quality initiatives are now underway to evaluate the quality of items on the birth certificate, to identify best practices, and to develop tools to assist hospitals to improve reporting. The National Center for Health Statistics is also collaborating with its state partners and with information technology experts on projects to develop standardized protocols for the collection of vital records data with electronic health records. These efforts should enhance the quality of these data in the years to come. Moving into the future, the establishment of links between electronic hospital records and electronic birth records has the potential to increase the consistency between these 2 sources. Two hospitals in the collaborative had discrepancies between hospital records and birth certificates of less than 5%. This finding raises several questions. Are there unique features about these facilities, are their recording processes different, and do they represent best practices that can be disseminated and incorporated into other hospitals? It would be interesting to investigate whether state-based perinatal collaboratives, such as the initiative in Ohio, might be able to design simple interventions to improve the accuracy of information that is recorded on birth certificates. However, as we all work to improve the information on birth certificates and to the extent that the Ohio experience can be generalized to other states, the use of birth certificate data to infer that there was or was not an indication for obstetric intervention should be done with caution and, as with all data sources, with a thorough understanding of strengths and limitations.