|
Article Excerpt
Teenage pregnancy and childbearing are significant public health problems in most large metropolitan areas of the United States, and rates in Milwaukee are among the highest in the country. (1) The literature generally shows that the risk of adverse birth outcomes--low or very low birth weight, preterm birth and neonatal death--is greater for teenagers than for older women. (2,3) However, studies disagree about what risk factors are associated with these adverse outcomes and whether young maternal age poses an increased risk when other factors are controlled for. (2,4) Other maternal characteristics that have been examined as potential explanatory variables for increased risk of adverse outcomes in teenage births are low socioeconomic status and minority race or ethnicity. (3,5) A number of behavioral and medical risk factors have also been associated with adverse birth outcomes in teenagers: lack of appropriate prenatal care or weight gain, smoking, alcohol use, illicit drug use and sexual risk-taking. (4,6)
A second teenage birth may be even more deleterious to mother and child than a single teenage birth because of compounded socioeconomic impacts and the influence of short interpregnancy intervals. (7) Women with interpregnancy intervals of less than 18 months appear to be at greater risk of having a premature birth or a low-birth-weight baby than women with longer intervals. (8) However, a number of studies that examined outcomes in higher order births to teenagers found preterm birth to be consistently associated with higher parity, while others indicate no increased risk of preterm birth or other adverse outcomes. (9,10) Teenage childbearing has a negative impact on educational attainment, which can lead to long-term economic disadvantage. (11) Teenage mothers who have a second birth within two years are less likely to complete high school than are those who postpone a second pregnancy. (12)
A systematic review of the literature found that 20-37% of teenage mothers had a second birth within 24 months. (13) In a review of programs that targeted pregnant teenagers or teenage mothers, Klerman found that most programs were not able to reduce the proportion of participants who had an additional birth within 24 months to below 20-25%. (7) The most recent national birth data reported that 81,517 teenage births (19%) in 2005 were second or higher order births. (14) Even though this number represents a decrease from previous years, it is still substantial. (15) Additionally, preliminary data indicate that the number of repeat teenage births rose to almost 85,000 in 2006. (16)
The ability to identify births to the same woman is necessary to understand the poor health outcomes associated with repeat childbearing. Many birth databases contain information on previous live births, but do not link births to the same mother. Thus, it is often possible to look at characteristics and outcomes of first and second births, but not first and second births to the same mother. Examining data for the same mother provides control for biological factors that vary among women.
In the study' described here, we used record linkage to identify repeat births to Milwaukee teenagers during the period 1993-2002, and assessed characteristics associated with adverse second-birth outcomes. The study was approved by the University of Wisconsin Health Sciences Institutional Review Board.
METHODS
Background
Milwaukee is the largest metropolitan area in the state; its 596,974 residents represent 11% of the total state population. (17) According to the 2000 census, 45% of the population was non-Hispanic white, 37% non-Hispanic black and 12% Hispanic. (17) The median household income in 1999 was $32,216, and 21% or residents for whom poverty status was determined had family incomes below the poverty level. (17) In 2004, some 19% of teenage births in Wisconsin were repeat births, compared with 20% nationally; (18) in Milwaukee, 25% of teenage births were repeal births. (19) During the study' period, mandatory birth certificate data reported by hospitals to the Wisconsin Office of Vital Records recorded 111,862 births.
To identify repeat births, it was necessary to find records of births to the same mother and create a family or sibling structure--that is, to identify related births and establish their order. The birth certificate records in the master birth file held by the Milwaukee health department did not contain any unique maternal identifiers Therefore, as is common when unique identifiers are not available, we employed a probabilistic matching procedure to link records. (20) The method works well with the kinds of errors often found in large data sets (e.g., misspellings, data entry errors, inconsistent use of abbreviations and missing data), because an exact match is not required.
Record Linkage Procedure
* Creation of birth record database. Birth certificate data for 1993-2002 existed in 11 files in four formats with a variety of field names and coding schemes. The data were consolidated and were extensively cleaned, recoded and reformatted where necessary to reconcile differences among the original files. This step, using Microsoft Access 2000, resulted in the creation of a master database, containing 111,862 birth records, and a data dictionary. A unique identification number was assigned to each birth record. The master file was then duplicated for the probabilistic match.
* Identifying potential matches. The duplicate birth files were matched against each other in FEBRL (Freely Extensible Biomedical Record Linkage), version 0.2.2. (21) To identify potential matches, FEBRL employs user-selected comparators to examine the same data element in two records. A similarity value is assigned on the basis of this examination. Similarity values are summed for each record pair to produce a final weight, which is used to determine if the record pair is a match.
Prior to performing the record linkage, we ran numerous match routines in FEBRL, using random samples from the birth file, and reviewed them manually' until the most accurate comparison method for the data was determined. On the basis of this information, we used three comparators to compensate for typographical errors and an exact comparator to identify identical data elements. The comparators examined mothers' first, middle, last and maiden names; state, day; month and year of birth: and race. Street address and husband's name, if available, were also included. The results of the random file matches also provided information to set upper and lower threshold weights. Record pairs with weights above the upper threshold were designated matches, and those with weights below the lower threshold were nonmatches.
Additionally, before performing the linkage, we used a standardizer to minimize the effect of differences in capitalizations, spacing and abbreviations in fields. The correction lists used by the FEBRL standardizer were modified to address...
|
