MCHB/EPI Miami Conference — December 7 - 9, 2005

Measurement of Gestational Age: Challenges to Research and Surveillance

JOYCE MARTIN: Good morning. Let me get myself situated here. I'm very pleased to be with you today, to have the opportunity to speak on this very important topic. I will describing the measures of gestational age available from the U. S. birth certificate and how these data have changed over the last decade and the potential impact of the revised data and (inaudible) systems on gestational age data.

First of all I'd like to acknowledge my new colleague Sharon Kirmeyer. Why spend a full session on a subject of measuring gestational age? Well, gestational age data, especially those based on birth certificates has become much more widely used in recent years to track trends in parinatal health and to study the causes of poor pregnancy outcome. Further fueling the interest in gestational as a measure of maternal and child health is the fact that the U. S. preterm rate has been on the rise essentially since national data on the subject have been available. National vital statistics data indicate that the preterm birthrate has climbed nearly one-third over the last two decades. Just released preliminary data for 2004 show that for the first time in U.S. history we had more than half a million infants born preterm for that year. Let's try this. But how reliable are these numbers? Numerous studies have demonstrated substantial quality issues of vital statistic gestational age data. Although NCHS edits these data for public release these edits traditionally are intended to resolve only the most egregious problems. A number of more rigorous editing procedures have been developed over the years but none have gained wide acceptance or use. Two measures of gestational age, the date of the last normal menses and the clinical estimate of gestation are available from the 1989 revised certificate of live birth, the standard currently in use by most states. All states report the date of the LMP, all states except California report the clinical estimate of gestation. The gestational age measure used by NCHS for all data releases is primarily based on the difference between the date of the LMP and the date of birth. Unfortunately, the date of the LMP is too often missing or invalid.

To reduce the bias that high levels of missing data are introduced we use two basic approaches of imputing or substituting gestational age. First, where the day of the LMP is missing but the month and the year are stated and valid we impute a number of weeks of gestation based on a previous record with similar characteristics. That is the same computed months of gestation, race and birth weight within 500 grams. This imputation was performed for just over five percent of 2002 records. Secondly, where the month and the year of the LMP date or unknown or invalid or inconsistent with birth weight, the clinical estimate of gestation is substituted for the LMP. This occurred for 5.4 percent of all 2002 births. For only 0.05 percent of all births was the clinical estimate substituted for the LMP because the LMP base estimate was inconsistent with birth weight.

This slide demonstrates the decline in levels of missing LMP dates by birth weight. For 2002, 5.3 percent of all birth records had a valid month and year of LMP but were missing the day of the LMP, a substantial decline from the 1990 level of 12 percent. As you can see the largest declines in missing data were among births weighing less than 3,000 grams. Still, data were most likely to be missing at the extremes of the birth weight ranges, that is for the higher risk births. A dramatic drop in missing data is seen for each group, a slightly larger decline is seen for non-Hispanic black mothers down 58 percent. Black mothers, however, still had the largest proportion of missing data of any group.

However, where we do have valid days of LMP reported we have other potential data issues. That is, we see substantial data preference. This tendency is particularly apparent at the 1 st and 15 th day of the month. Unfortunately, there appears to be very little change in digit preference over the last decade. We see similar patterns and trends for each of the racial groups that's not shown here.

Given the clear quality issues associated with LMP dating it's often suggested that we use the clinical estimate more to measure gestational age. The clinical estimate, however, also has some serious flaws. This slide shows our 1989 handbook instruction for the completion of the clinical estimate. It offers little or no guidance to the attendant as to what factors this measure is to be based except that it is not to be based on the interval between the LMP and the date of birth. This item was never intended as a primary measure of gestational age but was intended essentially as a back up item to be used only where the LMP was missing or invalid. Not unexpectedly anecdotally we understand that the clinical estimate is based on any number of factors, often ultra sound but also many other factors including the date of the LMP. Whatever factors are currently used to determine the clinical estimate they certainly have appeared to have changed in recent years. This slide shows the gestational age distribution of births based only on the clinical estimate. The distribution of this measure is far less peaked in 2002 compared with only 12 years earlier. In 1990, 40 percent of births were reported to have occurred at exactly 40 weeks of gestation compared with 28 percent in 2002. Although this shift likely also reflects a true shift towards shorter gestational ages the less peak distribution at precisely 40 weeks suggests that reporting of this item has changed over time.

There maybe a relationship between what appears to be improved reporting of the clinical estimate of gestation and the increase in the use of ultra sound to date pregnancy. This slide shows vital statistics data on ultrasound use for 1990 and 2002. These data likely do under report ultra sound usage and do not indicate when in the pregnancy the ultrasound was taken. However, it does indicate that the use of ultrasound has increased over the last decade and for all groups. Also that the gap among groups has narrowed somewhat.

This slide compares the distribution of the clinical and the LMP based gestational estimates for 2002. Although concordance between the two measures improve slightly over the study period the clinical estimate continues to show a substantially lower proportion of preterm and post term deliveries compared with the LMP based measure.

This slide shows preterm birth rates for the clinical estimate and the LMP for 1990 and 2002. For all groups the preterm rate based for all births I should say. The preterm rate based on the clinical estimate is substantially lower than that based on the LMP. Both the LMP and the clinical estimate, however, show preterm rates on the rise for all births and for white and for Hispanic births. However, among black births the two measures indicate preterm rates heading in slightly different directions. The clinical estimate shows a small but significant rise in the preterm rate where as the LMP shows a slight decline.

Let's talk briefly about the potential impact of the revised birth certificate and reengineered birth systems on the future of vital statistics gestational age data. The revised birth certificate also includes two measures of gestational age, the date of the LMP and new item or modified item called the obstetric estimate of gestation. This obstetric estimate replaces the clinical estimate. The name was changed to encourage clinicians to base their reporting on the obstetric estimate, not the pediatric estimate of age. The definition was also refined and now states explicitly that the estimate should be based on the attendants' best estimate as determined by all parinatal factors and assessments such as ultra sound but not the neonatal exam.

To further improve data quality separate work sheets have been developed. One work sheet from the mother to capture basic demographic and legal information and one for the hospital staff to complete to capture health and medical information from medical records. The intent is to encourage the use of the best sources and to better standardize these sources. We have also developed a detailed guidebook for the medical items, which includes recommended sources for the items, key words to look for and common abbreviations. Detailed specifications for the electronic birth systems have also been developed. The new systems include automatic edits and queries of data entry that is at the hospital where the data are entered. The systems do not allow entry of invalid dates inquiry out of expected ranges, entries such as very short gestation and high birth weight infants.

I'd like to conclude by summarizing the important findings. The good news is that is does appear that reporting of the LMP date has improved over the last decade. Unfortunately, there is little question that high levels of misclassification persist, particularly among high-risk groups. The lack of improvement and (inaudible) preference for the day and month of the LMP is also troubling. Reporting of the clinical estimate also appears to have improved likely related to greater reliance on ultrasound. The derivation of the clinical estimate for the most part, however, remains unclear. The two measures also agree on the direction of the trend in preterm birth rates both over all and for non-Hispanic white and Hispanic births but for black births the clinical estimate and the LMP base estimates deviate. Finally, although the implementation of the new birth certificates has been somewhat delayed we are confident that the new systems offer an opportunity to substantially enhance data quality. At NCHS we are very interested in continuing to improve upon the quality of these data.

Any thoughts you have on the subject from improving data at the sources to improving our editing procedures would be very welcome. Thank you.