The consequences of teenage pregnancy on teen mothers and their children are significant.1,2 Teenage mothers are more likely to have pregnancy-related disorders such as preeclampsia, preterm deliveries, growth-restricted neonates, and stillbirths.3,4 Children and adolescents of teenage mothers are more likely to experience poor nutrition, abuse, academic difficulties, developmental delay, depression, early sexual activity, and higher rates of criminal activity.5 Disproportionally high rates of poverty exist for teenage mothers. In 2010, 48% of teenage mothers lived below the poverty line.6 Educational achievement rates were low as well, and only 51% achieved high school diplomas by age 22 years (including high school equivalency certificates) as compared with 89% of their peers.6 Teen pregnancy has a major effect on the national economy and cost taxpayers in the United States approximately $9.4 billion in 2010.6
Although teenage birth rates have declined among all racial and ethnic populations, there continue to be notable racial disparities, with Hispanic women having the highest teenage birth rates.7 There are several social factors that are thought to play a role in the racial and ethnic disparities, including limited education and economic disadvantages.8 Data from the American Community Survey examining teenage birth rates at a county level noted lower educational attainment, lower income, and higher unemployment in counties with higher teenage birth rates.
Teenage birth rates are known to vary by geographic region. The purpose of this study was to determine whether certain areas in the contiguous United States could be identified as “hot spots” for teen birth. Our hypothesis was that these “hot spots” would remain, even after adjusting for poverty and educational attainment. Poverty was defined by U.S. census track data, and educational attainment was defined by percentage achieving a high school diploma by age 25 years.
MATERIALS AND METHODS
A retrospective cohort study was performed to identify geographic clusters of high (or low) teenage birth rates. This project was institutional review board–exempt because publicly available epidemiologic databases were used to access the information. These were studied with disease surveillance software (SaTScan) that uses a geographic information system to analyze geocoded data on teenage births. A cluster is defined as a greater than expected number of defined events that occurs within a group of people in a geographic area over a period of time. Geographic cluster mapping is a method that identifies certain areas as “hot spots.” These areas, or clusters, are identified as geographic areas in which several adjacent geographic units share high or low rates of teenage births. This method has been used in various fields to identify areas of the United States or areas in communities that can be targeted for interventions. For the purposes of this project, we used disease surveillance software to focus on the associated demographic variables in clusters of high teenage birth rates in the contiguous United States. Alaska and Hawaii were intentionally not included as a result of their distance from the mainland that would make a geographic study meaningless.
The original count data for teenage births by county are taken from the National Center for Health Statistics and is a total count from 2006 to 2012. The data were accessed through the County Health Rankings website.9 One hundred four missing county values were estimated by using an average of counties with original data that share borders with the missing county. The teenage population was defined as those females aged 15–19 years for the purposes of this project. Data for teenage girls younger than 15 years of age do exist; however, they were not used in this analysis because they are unreliable owing to there being too many suppressed county values for this small population. The count data were made into a rate from the 2010 Census (accessed through the Area Health Resources File) initially before using them to estimate counties.10 At least three counties were used to estimate the missing county's value (if there were not at least three counties with original data that share a border, the next closest county was chosen). The federal information processing standard codes that identified each county in the United States for all estimated counties, and how each one was estimated, can be found in Table 1.
The majority of the data for the variables chosen for investigation were available in compiled form from the Area Health Resources File. This file is compiled by the Department of Health and Human Services and contains more than 6,000 health indicators at the county level.11 Poverty data for the years from 2006 to 2012 were averaged to match the teenage birth data. Disease surveillance software was used to perform a spatial analysis on data to detect clusters. It tested whether teenage births were randomly distributed over space or whether there were statistically significant nonrandom clusters of teenage births. A P value was calculated for each individual cluster through a Monte Carlos study testing. The P values were then assigned to each cluster. The purpose of the P value in this analysis is to determine whether a cluster is the result of random causes. SaTScan does this by gradually scanning a predetermined window of geographic area (a circle), taking note of the number of observed and expected events and highlighting areas of elevated risk. In the case of this study, the maximum cluster size was set at a value of 10% of the total female population aged 15–19 years. The clusters were also limited to clusters with statistical significance (P value) <.01.
SaTScan uses a variety of models to perform the analysis, chosen based on the underlying distribution of the data. The discrete Poisson model was used because the teenage birth data are counting the number of events (teenage births) in a geographic area. The location given for each county is the geographic centroid of the county, labeled with latitude and longitude.
The actual cluster map identifies every significant cluster in this study and each map has a matching table to assist with interpretation. We focused on clusters in large population centers, defined as having a teenage population of at least 100,000 people. Poisson regression was used to adjust for poverty and high school diploma rates. Spearman rank correlation was determined to measure associations between variables.
The relative risk (RR) is a ratio of the average teenage birth rate of the counties in the clusters to the average teenage birth rate in the rest of the contiguous United States. The RR categories were used to group the RR values for the purposes of mapping.
We identified 109 statistically significant clusters with teenage birth rates significantly higher than the average teenage birth rate in the contiguous United States (P<.01). Figure 1 identifies all significant high teenage birth rate clusters with six RR categories colored differently on the cluster map.
In Table 2 we have listed the 10 clusters having the highest RR values for populations exceeding 100,000 people. The cluster with the highest RR in Figure 1 is labeled as Cluster 1, which includes 40 counties having an average teenage birth rate of 67 per 1,000 and an RR=1.87. Cluster 1 has 32.6% of its population living below the U.S. poverty line. The rate of teenage births in Cluster 1 is 87% higher than in the rest of the female population in the contiguous states. The top 10 clusters listed in Table 2 have a poverty rate between 21.6% and 33%. This is in comparison with the 2010 national poverty rate of 14.8%. Four of the 10 clusters in the table were counties in the state of Texas. Figure 2 identifies all significant low teenage birth rate clusters with P value<.01 and with six RR categories colored differently on the cluster map.
Our second research question was to determine whether the geographic clusters of reported teenage births were still present after adjusting for differences in poverty levels. Spearman rank correlation coefficient between teenage birth rates and poverty rates of 0.67 shows a strong positive association between these variables, suggesting that as the poverty rate in a cluster increases, the corresponding teenage birth rate in that county should also increase.
Figure 3 shows the significant clusters after adjusting rates for the corresponding poverty rates. Geographic clusters of reported teenage births are still present, but in different geographic areas after adjusting for differences in poverty levels. The counties with no coloring are counties that were never in any cluster, unadjusted or poverty-adjusted. The green counties are counties that were no longer in a cluster after adjusting for poverty. This indicates that the high teenage birth rates in the green counties are associated with high levels of existing poverty. The blue counties are counties that appeared in clusters after adjusting for poverty. This indicates that these counties had associated teen birth rates lower than the national average rates on the surface, but after adjusting for the poverty level of the county, the rates should have been lower. The purple counties are counties that were in an unadjusted cluster and are still in a cluster after adjusting for poverty. This indicates the high teenage birth rates in these counties are most likely not associated with poverty. These counties can be seen in Figure 3 as included in poverty-adjusted clusters. It is important to note that Figure 3 does not display a reduction or increase in RR. Clusters could have increased or decreased in RR as a result of poverty adjustment.
Table 3 gives the top 10 teenage birth clusters after adjusting for poverty rates. The “Fort Worth, Texas” cluster includes 149 counties with a teenage population of 1,458.831 and an average teenage birth rate of 57 per 1,000. Its poverty rate is approximately 26% with RR=1.58. This cluster has a 58% higher teenage birth rate than the rest of the contiguous United States even after adjusting for existing poverty differences across the contiguous United States. It is clear that adjusting for poverty greatly reduced the RR for all clusters, because there were no longer any clusters above the 2.06 RR thresholds. This indicates an association between poverty and teenage birth rates over the entire United States.
Our third research question was to determine whether the geographic clusters of teenage births were still present after adjusting for differences in education. We used the percentage high school diploma per county as a measure of education level achieved. The Spearman rank correlation coefficient between teenage birth rates and high school diploma rates is 0.58, which indicates that as the high school diploma rate for a county increases, the corresponding teenage birth rate in that county decreases. The cluster map shown in Figure 4 demonstrates that geographic clusters of teenage birth are still present after adjusting for differences in education levels. The green areas on this map identify counties with high teenage birth rates that are associated with the education level. The pink counties have high teenage birth rates that are not associated with the education level, whereas the orange counties have high teenage birth rates after the adjusting for differences in the education level. Table 4 identifies the top 10 clusters for high teenage birth rates after adjusting for high school diploma rates. The “Denver, Colorado” cluster includes only one county, and it has an RR=2.13. This means that this cluster has a teenage birth rate that is more than twice as great as the teenage birth rate for the rest of the contiguous United States after adjusting for education rates. In comparison, the same cluster has RR=1.78 (Table 1) for the raw (unadjusted) teenage birth rates.
The most important aspect of our study is the demonstration that there is a nonrandom distribution in the occurrence of teenage births; rather, “hot spots” exist. There are growing data in the literature describing the significant differences in teenage birth rates that link variables such as ethnicity, age of childbearing, poverty level, educational attainment, and living with both parents, affecting the likelihood of giving birth in the teenage years.3,12,13 However, our analysis showed that when we scrutinize these “clusters” of teenage births, and adjust for two of the strongest associated variables (poverty and educational attainment),14 clusters of elevated risk still remain, well above statistically what would be predicted. This indicates that there are other factors contributing to the high teenage birth rates in these clusters. Further study is needed to define the additional factors contributing to the disproportionately high teenage birth rates in these clusters. Note that these clusters represent outliers, because overall the teenage birth rate in the United States has fallen in all age demographics from 2008 to 2011, including in the study population.15 Although not a focus of our project, an unadjusted analysis in low population centers (less than 100,000 people) demonstrated striking results. Eight of the 10 clusters of teen birth (Table 5) were housed on or contained a Native American reservation. This is important information for the Indian Health Public Service and warrants a deeper look into what is driving high teen birth rates in these areas and what policy can be developed to implement initiatives to reduce the rates.
A strength of our study is that although epidemiologic studies often adjust for the age distribution within the population being studied by using some regression approach, we “matched” the teenage birth counts for a given age group14–17 with the female population for ages,14–17 which effectively creates “age-adjusting.” Furthermore, this study gives a unique perspective on areas in which teenage birth is still disproportionate even after adjusting for education and poverty. A weakness of our study is although it is a unique epidemiologic look at teenage birth, it does not provide specific insight into the additional factors contributing to the creation of the clusters.
Despite an overall national decline in the teenage birth rate, clusters of elevated teenage birth rates remain. There is a plethora of literature on teenage birth rates and the sexual activity of teenagers, aimed at finding the key to curbing the relatively high teenage birth rates in portions of the United States.16 It is an important observation that recent observed declines in the teenage birth rate have been largely attributed to an increase in contraception use rather than a change in sexual practices of adolescents.15 One would postulate that access to contraception may be a factor in these hot spots and represents an area of further study.17
This analysis demonstrates that teenage birth rates are not randomly distributed, but rather occur at higher rates within specific clusters, even when adjusted for poverty and education. This project will provide a literal road map for directions to focus targeted interventions to work on the reduction of teenage births in these areas.
1. Basch CE. Teen pregnancy and the achievement gap among urban minority youth. J Sch Health 2011;81:614–8.
2. Kearney MS, Levine PB. Why is the teen birth rate in the United States so high and why does it matter? J Econ Perspect 2012;26:141–66.
3. Hamilton BE, Martin JA, Osterman MJK, Curtin SC, Mathews TJ. Births: final data for 2014. Natl Vital Stat Rep 2015;64:1–64.
4. Malabarey OT, Balayla J, Klam SL, Shrim A, Abenhaim HA. Pregnancies in young adolescent mothers: a population-based study on 37 million births. J Pediatr Adolesc Gynecol 2012;25:98–102.
5. Carter DM, Felice ME, Rosoff J, Zabin LS, Beilenson PL, Dannenberg AL. When children have children: the teen pregnancy predicament. Am J Prev Med 1994;10:108–13.
6. National Campaign to Prevent Teen and Unplanned Pregnancy. Counting it up: the public costs of teen childbearing. Available at: https://thenationalcampaign.org/why-it-matters/public-cost
. Retrieved November 20, 2016.
7. Romero L, Pazol K, Warner L, Cox S, Kroelinger C, Besera G, et al. Reduced disparities in birth rates among teens aged 15–19 years—United States, 2006–2007 and 2013–2014. MMWR Morb Mortal Wkly Rep 2016;65:409–14.
8. Jaffee S, Caspi A, Moffitt TE, Belsky J, Silva P. Why are children born to teen mothers at risk for adverse outcomes in young adulthood? Results from a 20-year longitudinal study. Dev Psychopathol 2001;13:377–97.
9. County health rankings & roadmaps. Available at: http://www.countyhealthrankings.org/
. Retrieved November 20, 2016.
10. Area Health Resources Files (AHRF). Available at: http://ahrf.hrsa.gov/
. Retrieved November 20, 2016.
11. Heath Indicators Warehouse. Available at: http://www.healthindicators.gov/Resources/Initiatives
. Retrieved November 20, 2016.
12. Martinez G, Copen CE, Abma JC. Teenagers in the United States: sexual activity, contraceptive use, and childbearing, 2006–2010 national survey of family growth. Vital Health Stat 2011;23:1–35.
13. U.S. Department of Health & Human Services. Trends in teen pregnancy and childbearing. Available at: https://www.hhs.gov/ash/oah/adolescent-health-topics/reproductive-health/teen-pregnancy/trends.html
. Retrieved February 25, 2016.
14. Lou CC, Thomas A. The relationship between academic achievement and nonmarital teenage childbearing: evidence from the panel study of income dynamics. Perspect Sex Reprod Health 2015;47:91–8.
15. Finer LB, Zolina M. Declines in unintended pregnancy in the United States, 2008–2011. N Engl J Med 2016;374:843–52.
16. Finer LB, Zolna MR. Unintended pregnancy in the United States: incidence and disparities. Contraception 2011;84:478–85.
17. Coles MS, Makino KK, Stanwood NL. Contraceptive experiences among adolescents who experience unintended birth. Contraception 2001:84:578–84.