Gestational age estimation on United States livebirth certificates: a historical overview
Conflicts of interest: the authors have declared no conflicts of interest.
Megan L. Wier, MPH, Program on Health Equity and Sustainability, Environmental Health Section, San Francisco Department of Public Health, 1390 Market St., Ste. 910, San Francisco, CA 94102, USA. E-mail: firstname.lastname@example.org
Gestational age on the birth certificate is the most common source of population-based gestational age data that informs public health policy and practice in the US. Last menstrual period is one of the oldest methods of gestational age estimation and has been on the US Standard Certificate of Live Birth since 1968. The ‘clinical estimate of gestation’, added to the standard certificate in 1989 to address missing or erroneous last menstrual period data, was replaced by the ‘obstetric estimate of gestation’ on the 2003 revision, which specifically precludes neonatal assessments. We discuss the strengths and weaknesses of these measures, potential research implications and challenges accompanying the transition to the obstetric estimate.
Gestational age at birth and birthweight are associated with morbidity and mortality through childhood and beyond.1–3 Over 65% of infants who die in the first year of life are born low birthweight (<2500 g) and/or preterm (<37 weeks' gestation).4 Epidemiologists compare rates of preterm birth and low birthweight across both populations and time to assess the prevalence of factors that foster or impede infant health.5,6 Researchers also use gestational age and birthweight, which are measures of time (maturity) and size, respectively, to create other perinatal health indicators including small- and large-for-gestational-age measures and adequacy of prenatal care indices.5 In the US, the birth certificate collected from the States as part of the US Vital Statistics System is the only source of national data for gestational age and birthweight.
Because there are no measures of infant maturity at birth based on neurological criteria and physical characteristics that have been validated for the US population, birth certificates do not currently contain data regarding maturity at birth. Of the two, gestational age is a better proxy for maturity than birthweight, asbirthweight can vary greatly for a given level of maturity and gestational age.7 While birthweight is measured relatively reliably at birth, it is more difficult to estimate gestational age because it is based on one or a combination of imperfect criteria, including last menstrual period (LMP) and size- and maturity-based clinical indicators, such as prenatal ultrasound biometry and neonatal (i.e. postnatal) neurophysical assessments.5,8
This paper reviews gestational age estimation on birth certificates, beginning with a historical perspective of birth registration and gestational age estimation on the US Standard Certificate of Live Birth, the model for State birth certificates. We then review the underlying assumptions, strengths and limitations, and potential research implications for the two existing birth certificate-based gestational age estimates: ‘date last normal menses began’ (LMP) and ‘clinical estimate of gestation’ (CE), and describe the ‘obstetric estimate of gestation’ (OE), which is replacing the CE over the next several years. We conclude with a discussion of research challenges accompanying the transition to the OE.
Gestational age on the Birth Certificate
A historical overview
The US Standard Certificate of Live Birth was introduced in 1900. In an ongoing effort to improve collection of comparable and relevant nationwide birth data, the Centers for Disease Control and Prevention's National Center for Health Statistics (NCHS) collaborates with States' offices of vital statistics to revise the birth certificate every 10–15 years.9,10 Foremost a legal document used to establish date of birth, citizenship and identity, the birth certificate has evolved to serve secondary purposes including providing population-based statistical information on perinatal health.11 The US federal government has cooperative agreements with State-level agencies to obtain birth certificate data, and relies on the States and independent registration areas (District of Columbia, New York City, and territories) to use their legislative and regulatory bodies to ensure vital registration and data completeness, accuracy and comparability.9
Table 1 describes the gestational age items on the US Standard Certificate of Live Birth, years of inclusion, instructions and critiques. The 1930 birth certificate revision included the first gestational age-related variable, ‘premature or full term’. This variable was changed to ‘months of pregnancy’ in 1939, further refined to ‘length of pregnancy (completed weeks)’ in 1949, and retained on the 1956 revision.9 Instructions did not state how the physician should estimate gestational age, although LMP, birthweight (a birthweight ≤2500 g was a sufficient criterion to determine ‘prematurity’),12 or signs of fetal development and growth13 potentially informed gestational age assessment. Time, size and maturity were used interchangeably, with resulting gestational age estimates biased towards assessment of mature or ‘full-term’ infants.14,15
Table 1. Gestational age estimation on the US Standard Certificate of Live Birth, 1930–present
|Premature or full term||1930–1938||No item-specific instructions. Certificate was to be completed by the physician, preferably at the bedside.50||Item definition uses age and maturity interchangeably; estimation method not specified;a dichotomous variable.|
|Months of pregnancy||1939–1948||No item-specific instructions. Certificate was to be completed by the physician, preferably at the bedside.50||Broad age categorisation; estimation method not specified.a|
|Length of pregnancy, completed weeks||1949–1955, 1956–1967||No item-specific instructions. Certificate was to be completed by the physician, preferably at the bedside.50||Estimation method not specified.a|
|Date last normal menses began, month/day/year (LMP)||1968–1977||Enter the exact date (month, day and year) of the beginning of the mother's last normal menstrual period, as obtained from the hospital record or the mother herself. Spell out or abbreviate the name of the month. Do not use a number to designate the month. If the exact day is unknown, but the month and year are known, obtain an estimate of the day from the mother or her physician or enter only the month and the year. Do not use a date given by any informant other than the mother or physician. Enter ‘unknown’ if the date cannot be determined.51||Assumes LMP represents a date 14 days prior to conception; reliance on maternal recall and correct identification of menstrual bleeding.|
|1978–1988||Enter the exact date (month, day and year) of the beginning of the mother's last normal menstrual period, as obtained from the physician or the medical record. If the information is unavailable from these sources, obtain it from the mother. Enter the full or abbreviated name of the month (Jan., Feb., March, etc.). Do not use a number to designate the month. If the exact day is unknown, but the month and year are known, obtain an estimate of the day from the mother or her physician. If an estimate of the date cannot be obtained, enter only the month and the year. Enter ‘unknown’ if the date cannot be determined. Do not leave this item blank.b,52||See above|
|1989–2002||Enter the exact date (month, day and year) of the first day of the mother's last normal menstrual period, as obtained from the physician or hospital record. If the information is unavailable from these sources, obtain it from the mother. Enter the full name of the month – January, February, March, etc. Do not use a number or abbreviation to designate the month. If the exact date is unknown, but the month and year are known, obtain an estimate of the day from the mother, her physician or the medical record. If an estimate of the date cannot be obtained, enter the month and year only. Enter ‘unknown’ if the date cannot be determined. Do not leave this item blank.b,20||See above|
|2003–present||Enter all known parts of the date the mother's last normal menstrual period began. If no parts of the date are known, write ‘unknown’. The instructions note that the first (i.e. preferred) source is the Prenatal Care Record, under Menstrual History or Nursing Admission Triage Form; the 2nd source is Admission – History & Physical, under Medical History.b,19||See above|
|Clinical estimate of gestation, weeks||1989–2002||Enter the length of gestation as estimated by the attendant. Do not compute this information from the date last normal menses began and the date of birth. If the attendant has not done a clinical estimate of gestation, enter ‘None’. Do not leave this item blank.20||Informed by estimates that assume a direct correlation between time and size and/or maturity; estimation method not specified;a potentially informed by postnatal assessments with questionable validity.|
|Obstetric estimate of gestation at delivery, completed weeks (OE)||March 2003−April 2006||The obstetric estimate of the infant's gestation in completed weeks based on the birth attendant's final estimate of gestation. This estimate of gestation should be determined by all perinatal factors and assessments such as ultrasound, but not the neonatal exam. Enter the obstetric estimate of the infant's gestation in completed weeks. If the obstetric estimate of gestation is not known, enter ‘unknown’ in the space. Do not complete this item based on the infant's date of birth and the mother's date of the last menstrual period. The instructions note that the first (i.e. preferred) source is the Obstetric Admission History & Physical, under Weeks or Gestational Age.19||Informed by estimates that assume a direct correlation between time and size and/or maturity; estimation method not specified.c|
|May 2006−March 2007||The best obstetric estimate of the infant's gestation in completed weeks based on the birth attendant's final estimate of gestation. This estimate of gestation should be determined by all perinatal factors and assessments such as ultrasound, but not the neonatal exam. Ultrasound taken early in pregnancy is preferred. Enter the best obstetric estimate of the infant's gestation in completed weeks. If a fraction of a week is given, (e.g. 32.2 weeks) round up to the next whole week (e.g. 33 weeks). If the obstetric estimate of gestation is not known, enter “unknown” in the space. Do not complete this item based solely on the infant's date of birth and the mother's date of the last menstrual period. The instructions note that the first (i.e. preferred) source is the Obstetric Admission History & Physical, under Weeks or Gestational Age.19||See above. Additionally, rounding up when fractions of weeks are reported is contrary to clinical convention.|
|April 2007−present||The best obstetric estimate of the infant's gestation in completed weeks based on the birth attendant's final estimate of gestation. This estimate of gestation should be determined by all perinatal factors and assessments such as ultrasound, but not the neonatal exam. Ultrasound taken early in pregnancy is preferred. Enter the best obstetric estimate of the infant's gestation in completed weeks. If the obstetric estimate of gestation is not known, enter “unknown” in the space. Do not complete this item based solely on the infant's date of birth and the mother's date of the last menstrual period. The instructions note that the first (i.e., preferred) source is the Obstetric Admission History & Physical, under Weeks or Gestational Age.19||See above. Critique regarding rounding up no longer applicable, as that instruction has been omitted.|
The 1968 birth certificate revision was the first to include ‘date last normal menses began’, replacing ‘length of pregnancy’.9 Researchers viewed calculating the number of days from LMP to birth as opposed to clinically estimating completed weeks of pregnancy as an improvement in precision and consistency.15 LMP-based gestational age estimation became a priority as scientific evidence indicated that infants classified as premature based on a birthweight (<2500 g) criterion were not all gestation <37 weeks (and vice versa), with notable between-group differences in infant health outcomes for specific gestational age-birthweight categories.16,17 Health researchers and paediatricians focused on developing criteria that best classified at-risk infants using separate measures of birthweight and gestational age.18‘Date last normal menses began’ has been retained on the birth certificate through the 1978, 19899 and 2003 revisions.19
‘Clinical estimate of gestation (weeks)’ was added on the 1989 revision9 as a source of ‘. . . information on gestational age when the item on date last normal menses began contains invalid or missing information. For a record with a plausible date last normal menses began, it provides a cross-check with length of gestation based on ultrasound or other techniques’.20 On the 2003 revision, the ‘obstetric estimate of gestation at delivery (completed weeks)’ replaced the CE. Like the CE, OE instructions dictate that it should not be computed from LMP but also further detail that it ‘. . . should be determined by all perinatal factors and assessments such as ultrasound, but not the neonatal exam’.19 The CE and OE reintroduce gestational age dating concepts from birth certificate revisions prior to 1968.
Historically, States simultaneously implemented revisions to the birth certificate. However, the 2003 revisions will be gradually implemented over several years.10 Despite the large scale of data collection, complexities of implementation, and corresponding limitations and opportunities for measurement error, the birth certificate remains a critical source of national data on infant health indicators11 including LMP-, CE-, and OE-based gestational age estimates.
Last menstrual period (1968 birth certificate revision – present)
LMP is the primary gestational age source on the birth certificate, with documentation of its clinical use to estimate gestational age dating prior to the 1850s.21 LMP is the only time-based gestational age measure. This method of estimating pregnancy duration is widely used, with data collection based on maternal self-report. However, variation in menstrual cycle duration, non-menstrual vaginal bleeding, imperfect maternal recall, and clerical errors affect LMP accuracy as a population-level gestational age indicator.
The LMP-based gestational age estimate implicitly assumes a 28-day menstrual cycle with ovulation (and conception) occurring mid-cycle, and follicular and luteal phases each consisting of 14 days.22 However, the presence of menses and menstrual cycle regularity, length and timing of ovulation varies between and within women, due to genetic, environmental, sociodemographic, and behavioural factors.23–27 A study of women's menstrual cycles over the course of a year found that 18.6% and 29.5% of women had at least one short (<21 days) or long (>35 days) menstrual cycle, respectively, although only <1% of the women reported a cycle length that was usually short (0.5%) or long (0.9%).28 Based on data excluding long cycles, follicular cycle lengths reported in three studies ranged from 10 to 23 days [90–95% confidence intervals (CI)] with averages of 13–15 days, while luteal phase lengths ranged from 8 to 17 days [90–95% CI], with averages of 13–14 days.26Table 2 details the assumed cycle for LMP-based gestational age calculations, and an example of a cycle with a longer follicular phase that would, given conception, result in an overestimate of gestational age.
Assumed cycle length and ovulation timing in last menstrual period (LMP)-based gestational age calculations and associated errors
When mistaken for LMP, non-menstrual vaginal bleeding may also contribute to gestational age misclassification. Vaginal bleeding during early pregnancy would contribute to gestational age underestimation. In one study of planned pregnancies, 9% of women (n = 151) reported bleeding during the first 8 weeks of pregnancy; however, the study authors questioned whether the bleeding would be mistaken for a menstrual period given its generally short duration and light intensity.29Table 2 includes an example of a cycle with vaginal bleeding in early pregnancy that would result in a gestational age underestimate, given that bleeding was mistaken for LMP. Researchers have also postulated that early spontaneous abortion immediately prior to pregnancy and intermenstrual vaginal bleeding after withdrawal from hormonal contraceptives may contribute to gestational age error.30–32
Researchers must also assume that women accurately self-report LMP and that it is correctly recorded on the birth certificate. Maternal uncertainty regarding LMP contributes to digit preference for LMP day (preferred numbers: 1, 5, 10, 15, 20, 25, 28), with day 15 the most commonly reported.32–34 Social desirability may also motivate inaccurate LMP reporting (e.g. premarital conception).35 In addition, errors are probably introduced into LMP dates during data recording, transcription and data entry. Missing or inaccurate LMP, based on assessments of digit preference error, clerical error, or biologically implausible birthweight-for-gestational-age, is also more frequent among younger, poor, less educated, publicly or uninsured, non-white, unmarried women with late or no prenatal care.32,34,36–38
While LMP-based gestational age is widely used to date pregnancies, it has systematic and non-differential biases that potentially contribute to biased population estimates of pre- and post-term birth rates, as well as outcome study results.
Clinical estimate (1989 birth certificate revision)
For national reports, NCHS currently uses the CE to measure gestational age when LMP month and yearare missing (when only day is missing, it is imputed), or when the CE is compatible with birthweight and the LMP is not.39 The method used to estimate the CE is not reported, making assessments of its potential biases necessarily speculative. Potential sources include prenatal and neonatal clinical assessments, including ultrasound.
Prenatal and neonatal clinical assessments that potentially inform the CE assume a direct correlation between pregnancy duration and fetal or infant maturity or size. In the prenatal period, size-based gestational age estimates include fundal height and ultrasound measures; maturity-focused assessments may use clinical signs, such as quickening. Postnatally, gestational age may be crudely estimated from size-based measures, i.e. birthweight, while maturity-oriented assessments are based on infant physical characteristics (e.g. nipple formation, pupillary membrane clearance) and/or neurological criteria (e.g. scarf sign, arm recoil), including combination neurophysical assessments, such as the Dubowitz and Ballard scores.5,7,8
Clinical assessments of gestational age may therefore be inaccurate among infants whose growth or timing of maturation occurs at a rate markedly different from the reference used.5,40 For example, variation in fetal growth may potentially bias ultrasound and other size-based gestational age assessments. Characteristics associated with small fetal size and therefore potential underestimation of gestational age include younger maternal age, female gender, smoking, multiple gestation and high altitude,41–43 while obesity and diabetes are associated with larger fetal size and overestimation of gestational age.44
Clinical assessments also assume that size and maturity norms are represented by the reference used.5 Neonatal neurophysical assessments, developed primarily based on term infants, have been associated with increased gestational age measurement error in preterm and post-term infants.7,8 Furthermore, gestational age references historically used the LMP, and therefore may have themselves been biased by LMP-based gestational age errors.5,45
Because variation in growth increases with gestational age, clinical assessments conducted earlier in gestation generally have less estimation error.7,8 However, women with missing or questionable LMP-based gestational age, associated with delayed/no prenatal care and high risk for adverse pregnancy outcome, are less likely to have a CE from early pregnancy. Demographic differences in those who are lacking a CE and differences in CE measurement and documentation methods at the clinical, medical institutional and State level may contribute to systematic bias.40
How does the CE compare with LMP-based gestational age among women with both estimates? Two studies have compared LMP-based and CE-based gestational age using 1989–91 birth certificate data, one utilising State data from South Carolina (SC, n = 150 898)40 and the other from Illinois (IL, n = 476 034).46 The two studies had strikingly similar findings, despite differences in missing data (missing CE: SC = 6.5%, IL = 0.4%; missing LMP: SC = 2.9%, IL = 13.9%). Compared with the LMP, the CE distribution had a 1-week higher median (40 vs. 39 weeks), a smaller standard deviation and a narrower range between the 5th and 95th percentiles. The narrower CE distribution, with subjects concentrated around the 40-week median, parallels physicians' gestational age estimates on the early birth certificate with a bias towards term births.14,15
The LMP distribution had higher percentages of preterm (SC: 11.4% vs. 8.4%; IL: 9.8% vs. 6.7%) and post-term (SC: 9.6% vs. 4.2%; IL: 3.5% vs. 0.4%) cases compared with the CE distribution. The CE was more highly correlated with birthweight than the LMP,23 and there were fewer birthweight-inconsistent gestational age values40,46 which is not surprising given that the CE is potentially informed by size-based measures. The two studies found very similar overall exact concordance (SC: 47.3%; IL: 46%), with differences varying systematically by gestational age (greater differences in early and late gestational ages),40,46 sociodemographic, prenatal care, and hospital characteristics.40 As both studies were conducted shortly after the introduction of the CE, the findings may not be generalisable to more recent CE data given that ultrasound use for dating has increased over time.
Obstetric estimate (2003 birth certificate revision)
In order to address concerns regarding the CE definition and the potential inclusion of neonatal gestational age assessments, the OE was introduced on the 2003 birth certificate revision. Instructions for the OE explicitly exclude use of neonatal examinations as criteria and specifically suggest that the findings from ultrasound be considered along with other perinatal factors (Table 1).19 As of the end of 2006, 19 States (including New York State but not New York City, which is a separate registration area) had begun using this revised birth certificate variable, and several more are expected to add the OE in 2007 (Ventura S, personal communication, 14 March 2007).
Although prenatal ultrasound is the only gestational age assessment referenced in the OE instructions,19 other perinatal factors that could potentially contribute include obstetric clinical estimates, such as those derived from fundal height and/or time of first pregnancy symptoms, heartbeat detection, and quickening and possibly even biochemical tests for fetal maturation (e.g. amniotic fluid lecithin/sphingomeylin ratio).7 It is anticipated that the routine reporting of the OE on the birth certificate will improve the accuracy of gestational age estimates. Standard certificate instructions distributed to States between May 2006 and March 2007 pose a concern for States utilising those instructions, as they specified that fractions of weeks be rounded up (Table 1), contrary to the standard clinical practice of truncating gestational age estimates to completed weeks.47 Such rounding could artificially increase OE-based gestational age estimates by 1 week. Quantifying ways in which the OE will affect birth certificate-based gestational age estimation among the subgroup of women with LMP-based gestational age incompatible with birthweight or missing and assessing the impact of changing instructions will require additional research as the OE is gradually implemented on State birth certificates over the next few years.
Gestational age measurement on US birth certificates continues to evolve in response to concerns regarding data reliability and validity. In the 2002 NCHS birth files (n = 4 021 726), gestational age was estimated with the CE for 4.6% of births, overwhelmingly because LMP was missing (97%, after LMP day imputation) as opposed to inconsistent with birthweight (3%, based on NCHS cut points).39 Addressing the validity and reliability of gestational age on the birth certificate requires attention to both erroneous and missing LMP-based gestational age data.
Accuracy of birth certificate-based gestational age estimation may be improved by adoption of the OE in place of the CE, although it is important to assess OE comparability with other gestational age estimates in light of instruction revisions during its implementation.
The gradual implementation of the OE across States over the next several years raises additional challenges for epidemiologists using the CE/OE variables, particularly for surveillance as there will be differences between States (with implications for summary indicators) and within States (with implications when assessing temporal trends) during the transition. Specifically, how do the CE and OE differ in terms of data completeness, quality and accuracy, particularly among women missing LMP-based gestational age or with inconsistent infant birthweight? These questions should be further explored among subgroups of women with sociodemographic and obstetric characteristics, such as late prenatal care or multifetal gestation, associated with preterm and/or low-birthweight births.40
Understanding the biases in all measures of gestational age on the birth certificate is needed for accurate interpretation of the data for surveillance and research. Comparisons of measures in population-based samples of women can provide important information. For example, in California, as well as other States with centralised prenatal screening programmes, birth certificate gestational age data validity and reliability can be assessed through linkage to prenatal screening programmes with <20 weeks ultrasound-based gestational age estimates.48 From a surveillance perspective, gestational age should be considered in conjunction with other nationally reported health indicators, including, for example, birthweight, fetal and infant mortality, delivery method, congenital anomalies and prenatal care access, in order to obtain a complete picture of infant health. Otherwise, health disparities may be exaggerated or improvements overestimated.49
This paper was partially supported through contract CQ004942-LOS with the Centers for Disease Control and Prevention, Atlanta, GA. The authors acknowledge the helpful comments and suggestions of William M. Callaghan, Patricia M. Dietz, and Lucinda J. England of the National Center for Chronic Disease Prevention and Health Promotion, Division of Reproductive Health, Centers for Disease Control and Prevention, Atlanta, GA. The authors also acknowledge help with references and editorial comments for the historical overview provided by Joyce A. Martin of the Centers for Disease Control and Prevention, National Center for Health Statistics, Hyattsville, MD.