California Department of Public Health logo: three likenesses of people colored blue, green, and orange  
Sign-In  
Infant in incubator

Related Links



Join our list

Get updates on our project activities and new features of our website. Sign up for our newsletter here.


Contact Us

California Environmental Health Tracking Program

850 Marina Bay Pkwy, P-3
Richmond, CA 94804

(510) 620-3038
E-Mail Us
Last Edited: 9/30/2010

Prematurity and Growth Retardation: Measurement and Limitations

The following information explains where we get the data and how we calculate rates.


How is prematurity measured?

Estimates of the length of a pregnancy (“gestational age”) are included in most birth certificate records. The due date of a pregnancy is considered to be 40 complete weeks following the first day of the last menstrual period prior to the pregnancy. If delivery occurs prior to 37 complete weeks, it is considered preterm.

It should be noted that for many pregnancies the last menstrual period date is not known or not accurately recorded,1 and that such inaccuracies can have substantial impacts on calculated rates and disparities between rates.2

 

Back to top

 


How are rates of prematurity calculated?

Most commonly, preterm birth rates are expressed as percents. This is the number of preterm births in a population, divided by the total number of births, multiplied by 100. For our calculations, we exclude birth records for multiple births (twins, triplets, etc) and those for which the recorded gestational age is implausible considering the birthweight.3

In the data on this website, we refer to preterm birth rates calculated in this way as “conventional rates.” Due to the statistical properties of these numbers, rates based on small populations are subject to uncertainty. The degree of uncertainty for any given rate is represented by its confidence intervals. Although there are many ways to calculate these intervales, we used a common modification of Wilson’s approach.

 

Back to top

 


How is growth retardation measured?

Newborn weights are recorded in most birth certificate records. Any newborn weighing less than 2,500 grams (about 5.5 pounds) is considered low birthweight.

Most newborns having low birthweights are small because they are born preterm, although some are small because conditions of illness or nutrition interfered with their ability to grow during pregnancy. We suggest that viewers interested primarily in low birthweight due to prematurity become familiar with measurement of prematurity (above). One measure of growth restriction during pregnancy is the rate of low birthweight among full term pregnancies, or term low birthweight, which we discuss here.

Estimates of the length of a pregnancy (“gestational age”) are included in most birth certificate records. The due date of a pregnancy is considered to be 40 complete weeks following the first day of the last menstrual period prior to the pregnancy. If delivery occurs after to 37 complete weeks, it is considered full term.

It should be noted that for many pregnancies the last menstrual period date is not known or not accurately recorded,1 and that such inaccuracies can have substantial impacts on calculated rates and disparities between rates.2

 

Back to top

 


How are rates of growth retardation calculated?

Most commonly, term low birthweight rates are expressed as percents. This is the number of full term low birthweight births in a population, divided by the number of full term births, multiplied by 100. For our calculations, we exclude birth records for multiple births (twins, triplets, etc) and those for which the recorded gestational age is implausible considering the birthweight.3

In the data on this website, we refer to term low birthweight birth rates calculated in this way as “conventional rates.” Due to the statistical properties of these numbers, rates based on small populations are subject to uncertainty. The degree of uncertainty for any given rate is represented by its confidence intervals. Although there are many ways to calculate these intervales, we used a common modification of Wilson’s approach.

 

Back to top

 


How does spatial modeling produce estimates for areas with sparse data?

The most common form of health surveillance involves calculation of rates for specific geographic units (counties, census tracts, etc.). For health outcome mapping, each rate is then assigned to a category represented by a color as dictated by the map legend. Many alternative approaches to health outcome surveillance exist. These alternative methods are most useful when creating health outcome maps, as they allow the user to visualize the information. These methods have been used to improve visualization of prematurity and growth retardation on our MIH data query.

  • Why might I be interested in looking at modeled rates like this?

Modeled rates of this kind can be helpful if:

    • The location that interest you have been suppressed because the numbers are too small for the calculation of rates
    • Many of the high and low values that you see in a table, chart, or map are potentially related to random variations rather than genuine geographic disparities
    • The geographic scale that interests you is larger than the units you have; that is, you believe that groups of counties may be put together to demonstrate a regional trend, or a group of census tracts may be put together to show trends within counties
  • What is the name of the method we are using?

This approach is a type of hierarchical Bayesian modeling originally proposed by Besag, York, and Molié (the “BYM approach”). These researchers were motivated by the general problem of image reconstruction, in which random noise needs to be filtered to reveal an underlying image (potentially a disease map but also a satellite or microscopic image).

  • What is the reasoning behind this method?

A computer algorithm calculates the degree to which outcomes in spatial units (counties or tracts) appear to be influenced by the outcomes in their neighboring units. Once this is known, the rate for each unit can be updated using information from surrounding units. Units with lots of information (for example, those with large populations) are generally left alone, while those with less information are allowed to be informed by the rates of their neighbors. What results can be considered a “best guess” of what the rates would look like if there were no random variation cluttering up the map image.

  • What else should I know about this method?

This method has difficulty representing single counties or tracts with substantially high or low rates (“outliers”) unless the populations in those units are large. Unless these high or low rates are part of a regional trend, they tend to be adjusted towards the population average.

 

Back to top

 


Where do these data come from?

In the U.S., states are responsible for issuing birth certificates and recording and maintaining the data included in them. Birth certificates are considered one type of vital record (others include deaths, fetal deaths, and marriage). In California, the Office of Health Information and Research (OHIR) is responsible for stewardship and distribution of vital statistics data and provides written reports and data tables analyzing these data. Since several of the important functions of the CEHTP include the analysis and processing of these records, we maintain our own databases consisting of records produced by OHIR and then subject to further processing, most notably regarding address and other geographic information fields.

 

Back to top

 


Limitations

For many pregnancies the last menstrual period date is not known or not accurately recorded,1and such inaccuracies can have substantial impacts on calculated rates and disparities between rates.2 Information such as maternal race and ethnicity or place of residence are generally provided by hospitals and other providers of obstetric services throughout the state. As such, the methods of collecting this information and the categories chosen may vary. Even when geographic residential information is accurate, it may not serve as a reflection of where the mother spends the majority of her time during or after her pregnancy or be useful when inferring exposures to environmental hazards.

 

Back to top

 


1. Vahratian A, Buekens P, Alexander G. State-specific trends in preterm delivery: Are rates really declining among non-Hispanic African Americans across the United States? Maternal and Child Health Journal. 2006;10(1):27-32.
2. Wingate M, Alexander G, Buekens P, Vahratian A. Comparison of gestational age classifications: Date of last menstrual period vs. clinical estimate. Annals of Epidemiology. 2007;17:425-430.
3. Alexander G, Himes J, Kaufman R, Mor J, Kogan M. A United States national reference for fetal growth. Obstetrics and Gynecology. 1996;87(2):163-168.