Investigating Risk Factors Affecting Teenage Pregnancy Rates in the United States

Teenage pregnancy is an issue in the United States and signifies a growing public health concern. Health efforts and research on potential risk factors for teenage pregnancy is essential to the public health sector. In this study, risk factors affecting the teenage pregnancy rate are being investigated using artificial intelligence methods, Artificial Neural Network (ANN) and Support Vector Machine (SVM) to assist with teenage pregnancy education and prevention. Data was collected on a national level from the U.S. Census Bureau, the Center for Disease Control and Prevention (CDC), and the U.S. Departments of Labor and Commerce from 1972 to 2008. The important risk factors found in the ANN and SVM models were consistent with the literature findings, and are major contributors to the teenage pregnancy rate. It is evident that ANN and SVM models have successfully demonstrated their model validity and applicability. Therefore, preventative measures should focus on per capita income, unemployment rate, alcohol consumption, poverty rate, and cigarette consumption to reduce the risk and incidence of teenage pregnancy in the United States. INTRODUCTION Teenage pregnancy has not only become a public health issue, but also a media focal point. After a thorough review of the literature, it was determined that current research regarding teenage pregnancy is limited. The rate of teenage births in the United States has fallen since the late 1950s from 96 births per 1,000 women aged 15-19 to 49 births per 1,000 in 2006. Teenage pregnancy and births among girls in their adolescent years has declined and may be explained to a change in trends of marriage. Marriage in teen years is now not as common as it was in the 1950s. In the 1950s, 13% of teens had children without being married, but in 2000 that rate was 79%. In the 1990s, the typical age of first marriage in the United States was 25 for women and 27 for men. “Shotgun marriages” were also common, but few pregnant teens marry before the baby is born. Teenage pregnancy is still a major issue whether a teenager is married or not, but the change in trends of pregnancy may be explained by change in trends of other factors such as marriage. In the 1970s and early 1980s, the U.S. teen pregnancy rates increased. They remained steady through the 1980s, even as sexual activity among teens increased, due to the use of contraceptives among teenagers. In 1999, the Alan Guttmacher Institute did an analysis of teenage pregnancy and found that approximately onequarter of the decline in teenage pregnancy in the United States between 1988 and 1995 was due to increased abstinence. Also, overall contraceptive use increased—but only slightly, from 78% in 1988 to 80% in 1995. Teenagers in 1995 were choosing more-effective methods. A significant proportion of girls have switched to some long-acting hormonal methods that were introduced to the U.S. market in the early 1990s. By 1995, more than one in eight teen contraceptive users (13%) was using a long-acting method, European International Journal of Science and Technology ISSN: 2304-9693 www.cekinfo.org.uk/EIJST 42 and primarily because of this shift, sexually active teens became increasingly successful at avoiding pregnancy (The Guttmacher Report, 2002). The rates declined 19% from 117 pregnancies per 1,000 women aged 15-19 in 1990 to 93 per 1,000 in 1997—the lowest rate in 20 years. The recent decline is particularly encouraging, because—as with the teen birthrate decline—all population groups followed a similar pattern, regardless of young women's age, marital status, race or ethnicity (National Center for Health Statistics, 2001). However, in 2008, the National Institutes of Health reported that teen pregnancies in the United States increased for the first time since 1991. In 2005, teen pregnancy rates edged upward from 21 births per 1,000 teenage girls to 22 per 1,000 in 2006. Researchers reported uncertainty of the cause of the rate increase (National Institutes of Health, 2008). Teenagers in the United States continue to experience substantially higher pregnancy rates and birthrates than do teens in other Western industrialized countries (Gold, Kawachi, Kennedy, Lynch, & Connell, 2001). The adolescent pregnancy rate in the United States, for example, is nearly twice that of Canada and Great Britain and approximately four times that of France and Sweden. Moreover, teen birthrates have declined less steeply in the United States than in other developed countries over the last three decades. While we are far from having a complete understanding of why teens have children, available research has identified four conditions associated with teenage childbearing. Teens most likely to have children are those: (1) from economically disadvantaged families and communities; (2) not doing well in school and having low aspirations for their own educational achievement; (3) from dysfunctional families; and (4) with substance abuse and behavioral problems (Keplinger, Lundbreg & Plotnick, 1998). In the United States, some of a teenagers’ risk for pregnancy includes factors such as poverty, parents with low levels of education, low performance in school, and single-parent families. The Tennessee Department of Health listed the following as factors contributing to teenage pregnancy: lack of family and community support, lack of involvement in school and recreational activities or after school programs, lack of feeling connected to school, family, and community, use of alcohol and other drugs, lack of health education, and limited knowledge about sex and sexuality (Department of Health, 2011). Teenage mothers are more likely to drop out of high school, be and remain single parents, and are less likely to attend college than those who become pregnant at an older age (Guttmacher Institute, 2004). In turn, there are repercussions from the teenage mother that heavily impact the life of the child, including: lower scores in math and reading into adolescence, repeat of a school grade, high school dropout, giving birth as a teenager, and being unemployed or underemployed as a young adult. Overall, being a teen mother perpetuates the cycle of teen motherhood (Guttmacher Institute, 2006). Teen pregnancy affects education in such a way that only a third of teenage mothers earn their high school diploma. And only 1.5% acquires a college degree by age 30. Adolescent childbearing is usually inconsistent with mainstream societal demands for making the transition into adulthood, gaining education and work experience, and financial stability (Guttmacher Institute, 2004). This paper addresses the risk factors, population, per capita income, poverty rate, education and unemployment, and cigarette and alcohol consumption as key components of teen pregnancy rate. Population Based on the most recent data, teenage birthrates were highest in the states of Mississippi, Texas, Arizona, Arkansas and New Mexico; and states with the lowest birthrates were in New Hampshire, Vermont, Massachusetts, North Dakota and Maine. States with the largest number of teens also had the greatest number of teen pregnancies (O’Hare, 1999). Current trends in teenage pregnancies varied in different parts European International Journal of Science and Technology Vol. 2 No. 2 March 2013 43 of the U.S. in relation to prevailing socioeconomic conditions (Smith et al, 2007). Teens who live in urban or rural areas are more likely to give birth between 15-19 years old (Smith et al, 2007). Per capita income Further literature exploration indicated that young people growing up in disadvantaged economic, familial and social circumstances are more likely than those who did not to engage in risky behavior and have a child during their adolescent years. However, at all socioeconomic levels, American teenagers are less likely than their peers in the other study countries to use contraceptives and more likely to have a child. Among teens aged 15–17, income inequality and per capita income were independently associated with birth rate; the mean birth rate was 54 per 1,000 in counties with low income and high income inequality, and 19 per 1,000 in counties with high income and low inequality. Among older teens (aged 18–19) only per capita income was significantly associated with birth rate. Although teen childbearing is the result of individual behaviors, these findings suggest that community-level factors such as income and income inequality may contribute significantly to differences in teen birth rates (Gold, Kawachi, Kennedy, Lynch & Connell, 2001). In comparison to previous research which concluded that early childbearing leads to lower levels of investment in education and labor market experience, which in turn leads to depressed socioeconomic status, other research concluded that teen mothers are now more likely than in the past to complete high school or obtain a GED, but they are still less likely than women who delay childbearing to go on to college (Keplinger, Lundbreg & Plotnick, 1998). Poverty Rate Teen mothers are more likely to be homeless at a younger age and homeless more often when compared to homeless adult mothers. In 2002, teen mothers were 75% more likely to have reported being homeless more than once and they were over three times more likely to have first become homeless before the age of 18. Many homeless teen mothers grew up in unstable environments were removed from parental care, or witnessed domestic violence as a child; and one in four were actually abused as children themselves. The relationship between teen pregnancy and poverty in the US is strong and well documented (Guttmacher Institute, 2002). Nearly 60% of teens who become mothers are living in poverty following the birth of their infant (Hofferth, 2001). This research supports the conclusion that family income has selective but, in some instances, quite substantial effects on child and adolescent well-being. Family income appears to be more strongly related to the child’s ability and achievement than to their emotional outcomes. Children who live in extreme poverty or those have lived below the poverty line for multiple years appear, all other things being equal, to suffer the worst outcomes. The timing of poverty also seems to be important for certain child outcomes. Education and Unemployment The instability of the economy affects adults as well as teenagers and children. Teenage pregnancy cast as a social problem was believed to be the result of lower levels of education, welfare dependency, and lowpaying jobs, as well as greater health troubles for these teens and their babies. An environment that creates doubt in one’s economic sphere may also create doubt in other areas (Winters & Winters, 2012). European International Journal of Science and Technology ISSN: 2304-9693 www.cekinfo.org.uk/EIJST 44 Conduct problems in school can deflect obtaining an education. Educational level has been determined to be a risk factor in teenage pregnancy. The relationship between conduct problems at younger ages and teenage pregnancy by age 18 was analyzed and a statistically significant association found between early conduct problems and later risk of teenage pregnancy. The results suggest that the higher rate of teenage pregnancy among girls with early conduct problems reflected their tendencies have a risky taking behavior in their adolescence years (Woodard & Fergusson, 2000). Manlove (1998) found that factors relevant to teens’ school experiences—including characteristics of their school and classroom, their family background, and individual engagement—were associated with the risk of school-age pregnancy leading to a live birth. A set of proportional hazards models indicated that for all racial and ethnic groups, high levels of school engagement were associated with postponing pregnancy. Cigarette and Alcohol Consumption A study done by Cavazos-Rehg et al. (2011), determined that alcohol users, no clear pattern of age or intensity of use, were at increased risk of sexual intercourse with and without a history of teenage pregnancy. Experimental cigarette use increased the risk of having sexual intercourse but did not reach significance for pregnancy. However, daily cigarette smokers had an increased risk for sexual intercourse with and without a history of pregnancy. The study also found, earlier age at substance use initiation and higher intensity of use may be associated with increased risk for teenage pregnancy. Zimmer-Gembeck & Helfand (2008) stated, “Adolescents who are potentially at greater risk for teenage pregnancy are those who use substances; yet, the bulk of current pregnancy prevention strategies do little to target or educate youth on the sexual health risks associated with substance use despite growing evidence demonstrating the efficacy of these strategies.” SIGNIFICANCE OF THE STUDY The teenage pregnancy rate is an undeniably powerful indicator of the overall well being in a population. This is a concern to public health practitioners because majority of adolescents are among unmarried and unemployed young women. The investigation of teenage pregnancy provides an opportunity for developing preventive strategies to improve population health. To our best knowledge, this study may be the first to use the Artificial Neural Network (ANN) and Support Vector Machine (SVM) to rank the important risk factors contributing teenage pregnancy in the U.S. If the most important predictors of teenage pregnancy in the U.S. can be identified, targeted interventions can be developed and the efficacy of pre-existing programs may be improved. Policy makers could further use this information to determine which social programs (i.e. smoking cessation campaigns, unemployment programs, and teen pregnancy prevention programs) should receive more government support. Also, ranking the most important risk factors at the national level could help identify national trend and regions affected by the same risk factors. OVERVIEW OF ANN AND SVM MODELS ANN is an information processing paradigm inspired by the function of the human brain (Freeman & Skapura, 1991; Hinton, 1992; Zutada, 1992). The network consists of many processing units (neurons or nodes) that are linked together and embedded in the input, hidden, and output layers. These links multiply the neurons of the input and hidden layers by an individual weighing factor, which is a value analogous to the connection strength at a synapse (Dreiseitl & Ohno-Machado, 2002; Freeman & Skapura, 1991). The synaptic process of the ANN stores the knowledge needed to solve specific problems. European International Journal of Science and Technology Vol. 2 No. 2 March 2013 45 The ANN model is constructed with activation functions that establish nonlinear relationships between input layer neurons and hidden layer neurons, and nonlinear relationship between hidden layer neurons and output layer neurons. These activation functions allow for the identification, ranking, and interpretation of associations between teenage pregnancy and related risk factors. The SVM classifier is operated on the principle of structural risk minimization (Vapnik, 1995). It was designed to minimize true risk of misclassifying examples during the model training. The SVM classifier has its advantage in the practical application for small sample and generalization because of structural risk minimization (Vapnik, 1995; Zeng, Xu, Gu, Liu, & Xu, 2008). It can be applied to the prediction of continuous outcome variables (Cristianini & Shawe-Taylor, 2000). SVM allows researchers to construct nonlinear classifier as the solution to a quadratic problem in order to yield a minimum of error function. It seeks an optimal hyperplane to separate data from different categories through the computational short-cut of kernel functions (Cortes & Vapnik, 1995; Cristianini & ShaweTaylor, 2000). The basic role of the kernel function is to calculate inner product values through a transformation in high-dimensional feature space, and ultimately maximize the margin of separation to yield high accuracy of data prediction. In this study, ANN and SVM models used logistic and linear functions to determine the most important risk factors for teenage pregnancy. STUDY DESIGN AND DATE SOURCES This retrospective study was designed to analyze teenage pregnancy rate in the 1972-2008 national natality dataset of the United States (See Figure 1), compiled by Centers for Disease Control and Prevention Wonder. The data, at the county level, was pre-coded according to uniform specifications, and passed through vigorous statistical quality checks (Centers for Disease Control and Prevention, 2011). As shown in Table 1, additional data in this study was taken from the U.S. Census Bureau the Office of Management and Budget, National Institute of Alcohol Abuse and Alcoholism, and the CDC on the national level for the following factors: U.S. per capita income, U.S. poverty rate, U.S. unemployment rate, U.S. cigarette consumption, U.S. alcohol consumption, and the percentage of the U.S. population who had less than 9 years of education. Figure 1. U.S. Teen Pregnancy Rates, Years 1972-2008 European International Journal of Science and Technology ISSN: 2304-9693 www.cekinfo.org.uk/EIJST 46 Table 1. Independent Variables for Teenage Pregnancy Rate