Maternal and Child Health Journal

, Volume 16, Issue 5, pp 1063–1071

Teen Motherhood and Long-Term Health Consequences


    • Department of Healthcare Organization & PolicyUniversity of Alabama at Birmingham
  • Bisakha Sen
    • Department of Healthcare Organization & PolicyUniversity of Alabama at Birmingham

DOI: 10.1007/s10995-011-0829-2

Cite this article as:
Patel, P.H. & Sen, B. Matern Child Health J (2012) 16: 1063. doi:10.1007/s10995-011-0829-2


The objective of this article is to examine the association of teen motherhood and long-term physical and mental health outcomes. The physical and mental health components (PCS and MCS) of the SF-12 Healthy Survey in the NLSY79 health module were used to assess long-term health outcomes of women who experienced teenage motherhood. Various familial, demographic, and environmental characteristics were indentified and controlled for that may have predicted teen motherhood and long-term health outcomes. The two comparison groups for teen mothers were women who experienced teen-pregnancy only and women who were engaged in unprotected sexual activity as a teenage but did not experience pregnancy. Multivariate ordinary least squares regression was used for analysis. The average PCS and MCS for teen mothers was 49.91 and 50.89, respectively. Teen mothers exhibited poorer physical health later in life compared to all women as well as the comparison groups. When controlling for age, teen mothers had significantly lower PCS and MCS scores compared to all other women. Furthermore, when controlling for familial, demographic, and environmental characteristics, teen mothers exhibited significantly lower PCS and MCS scores. When comparing teen mothers to the two comparison groups, PCS was not statistically different although MCS was significantly lower in the teen-pregnancy group. Teen motherhood does lead to poorer physical health outcomes later in life. On the other hand, poorer mental health outcomes in later life may be attributed to the unmeasured factors leading to a teen pregnancy and not teen motherhood itself. Additional research needs to be conducted on the long-term consequences of teen motherhood.


Teen motherhoodPhysical healthMental healthHealth outcomes


The US has continued to experience higher rates of teenage pregnancy and motherhood than most developed nations [1]. Each year, approximately 750,000 women aged 15–19 experience a pregnancy [2] and the majority of these pregnancies result in live births. After falling to a low of 40.5 births per 1,000 women in 2005, the average birth rates for 15–19 year olds increased in 2006 to 41.9 births per 1,000 women. Approximately 10% of all US births in 2006 were to teenagers [3].

The total annual government expenditures on public aid to teen-mothers in 2006 were $11.3 billion [3]. But apart from the costs imposed upon taxpayers, teen-mothers themselves exhibit adverse outcomes subsequently in life, including poor educational and economic outcomes. In this study, we extend that literature by exploring the relationship between teen-motherhood and mid-life health outcomes for women. Little research has been done regarding this particular outcome. However, rising health costs are a grave concern in the US, and if teen-motherhood is associated with poorer health outcomes in later life, then increases in teen birthrates may have important consequences to society in terms of the greater health needs and costs in future decades.


It is fairly well established in the social science literature that experiencing teen-motherhood is associated with subsequent adverse life outcomes for women. Several studies find strong associations between teen-motherhood and poor education, underemployment, and lower socioeconomic status for women [411]. British teenage mothers had a significantly higher level of depression in medium term postpartum compared to older mothers [10]. Additionally, teen-mothers often face stigmatization and criticism for contributing to adult poverty [8].

While experiencing teen-motherhood is correlated with adverse outcomes, it is also fairly well-established that experiencing teen-motherhood in itself can be a manifestation of poor socioeconomic status (SES) during childhood, as well as a combination of household factors, environmental stressors, and psychological factors [6, 7, 1214]. Teen-mothers also exhibit lower cognitive scores than their counterparts [15] and score higher on the Dean Romanticism Scale [16] suggesting that they may have more naïve beliefs about romantic love and the permanency of relationships.

While teen childbearing is found to have a strong association with a variety of subsequent adverse outcomes in women, teen childbearing itself seems to be predicted by a variety of negative childhood experiences, family background characteristics, and personal psychological factors. Thus, it remains unclear whether it is experiencing teen-motherhood per se that leads to the subsequent poor outcomes, or whether these outcomes are more a result of the same underlying factors that predict teen-motherhood in the first place. It is important from a policy perspective that this be better analyzed. If the poor educational, economic and health outcomes can be attributed to experiencing teen-motherhood per se, then resources directly targeted towards preventing teen-motherhood will benefit the women. However, if the poor outcomes are more a result of the underlying factors predicting teen-motherhood per se, then merely preventing the latter may not yield the expected improvements in outcomes for these women. Thus, the challenge for researchers is to decipher whether outcomes for teen-mothers differ from their counterparts who had virtually identical familial, environmental and personal characteristics, but who happened not to become teen-mothers.

Researchers have attempted to address this through a variety of methods. One approach is to include as comprehensive a set of control variables as data permits for childhood and family background. Studies have examined the outcomes of teenage mothers while controlling for various background characteristics and have shown that the association between teenage motherhood and subsequent outcomes were reduced [17, 18].

Another approach is to find an appropriate ‘control’ group likely to have circumstances very similar to teen-mothers, but simply not experiencing the actual event of teen-motherhood themselves. This allows researchers to infer whether differences in outcomes between the control group and teen-mothers are attributable to teen childbearing per se [19]. Sisters of teen-mothers who were raised in the same household (thus, presumably, experiencing the same household and environmental stressors) but did not experience a teenage birth are one favored control group [2022]. Another control group is teenagers who had become pregnant but did not actually experience a birth [3].

Very little research currently exists on the association of teen-motherhood with long-term health outcomes. With this article, we help fill that gap. We use data from the NLSY79, and derivations of some of the methods established in the literature to help control for underlying confounders that may both lead to teenage births and subsequent poor health outcomes. We describe further details of the data used in the study below.

Data and Methods

National Longitudinal Survey of Youth 1979

The NLSY79 is a nationally representative sample comprised of 12,686 young people aged 14–22 years old on the date they were first surveyed in 1979, and who were residing in the US. The respondents were interviewed annually from 1979 through 1994. After 1994, the survey was conducted biennially. The NLSY79 initially comprised of three subsamples:
  1. (1)

    A cross-sectional sample (n = 6,111) designed to represent non-institutionalized civilian youths aged 14–22 years.

  2. (2)

    A supplemental sample (n = 5,295) that oversampled blacks, Hispanics, and poor Non-black Non-Hispanics aged 14–22 years.

  3. (3)

    A military sub-sample (n = 1,280) representative of the population who enlisted in an active military branch by September 30, 1978 aged 17–21 years by 1979.


Due to funding constraints, the military subsample and the poor Non-black Non-Hispanic respondents in the supplemental sample were subsequently dropped from the survey.

Outcome Variables

Beginning in survey year 1998, an extended Health module was created to accurately assess the occurrence of chronic health problems in the aging NLSY79 cohort. It was administered to NLSY79 respondents in the first survey they took after reaching age 40. This occurred anywhere from 1998 to 2006, depending on the respondents’ year of birth. We use the physical and mental health components of the SF-12 Health Survey in the NLSY79 Health module The SF-12, which stands for “short-form 12-question”, is a brief inventory of self-reported mental and physical health using twelve standardized questions. The health concepts include physical functioning, role functioning physical, general health, bodily pain, social functioning, vitality, role functioning emotional, and mental health (ALSFRS, 2009). The exact questions and the distributions of responses in our sample are in Table 1. Results of the SF-12 are summarized in terms of two meta-scores in the NLSY79 dataset, the Physical Component Summary (PCS) and the Mental Component Summary (MCS). Scores are created in NLSY79 according to the manual by Ware et al. [23], though the precise scoring formula is kept confidential. These meta-scores are our main outcomes of interest. Higher scores represent better health.
Table 1

SF-12 questions and distributions in the sample


Frequency (N)

Percentage (%)

Assessment of general health




 Very Good












Does health limit moderate activities?

 Yes, limited a lot



 Yes, limited a little



 No, not limited at all



Does health limit climbing stairs?

 Yes, limited a lot



 Yes, limited a little



 No, not limited at all



Have accomplished less than would like in past 4 weeks?







Does health limit kind of work or other activities?







Emotional problems caused to accomplish less in past 4 weeks?







Emotional problems made actions less careful?







Pain interfered with normal work in past 4 weeks?

 Not at all



 A little bit






 Quite a bit






How often felt calm and peaceful in past 4 weeks?

 All the time



 Most of the time



 A good bit of the time



 Some of the time



 A little of the time



 None of the time



How often had a lot of energy in past 4 weeks?

 All the time



 Most of the time



 A good bit of the time



 Some of the time



 A little of the time



 None of the time



How often felt down-hearted and blue in past 4 weeks?

 All the time



 Most of the time



 A good bit of the time



 Some of the time



 A little of the time



 None of the time



Physical/Emotional problems interfere with social activities in past 4 weeks?

 All the time



 Most of the time



 A good bit of the time



 Some of the time



 A little of the time



 None of the time



Summarized results from the SF-12 scores are given in the NLSY79 dataset as the Physical Component Summary (PCS) and Mental Component Summary (MCS) scores. Scores are computed based on algorithms in the manual by Ware et al. [23], but the precise formula is kept confidential

The PCS and MCS scores were designed such that the representative US population mean scores are 50, and the standard deviations 10. Thus each one-point difference above and below 50 corresponds to one-tenth of a standard-deviation (ALSFRS, 2009). Because the US population mean score includes the elderly for whom scores decline rapidly, a better comparison for the NLSY79 sample might be the scores for the sub-group of US populations aged 35–44 years, for whom the mean PCS and MCS scores, respectively, are 52.18 (std. dev. 7.30) and 50.1 (std. dev. 8.62).

Analytical Approach

The participants in this study include 4,271 NLSY79 female respondents for whom PCS and MCS scores were reported by 2006. Our primary group of interest is women who had a live birth as a teenager (“teen-mother”). When comparing PCS and MCS scores for teen mothers with other women in a multivariate regression framework, we attempt to control for underlying confounders that may predict both teen-motherhood and future health outcomes using the following tactics. First, we control for an extensive array of measured familial, demographic and environmental characteristics available in the NLSY79 dataset (described later under “other variables”). Next, we attempt to identify women who are likely to share unmeasured characteristics with teen-mothers, but who do not actually become teen-mothers. We consider two comparison groups: (1) women who became pregnant as teenagers but then experienced a miscarriage, abortion, or stillbirth (“teen-pregnancy-only”); and (2) women who reported unprotected sexual activity as a teenager, but did not experience a teen-pregnancy (“teen-unprotected-sex”). We identify respondents in the teen-mothers, teen-pregnancy-only and teen-unprotected-sex groups using information from the NLSY79 Fertility module, which informs on contraceptive use and pregnancy outcomes. Specific fertility variables, including age of first (and subsequent) pregnancies and outcomes of these pregnancies were added to the NLSY79 beginning in 1984. We use information from the 1984, 1985 and 1986 surveys to identify the groups as described below.


We use “age 18 or less” to define a teenager. Therefore, respondents who meet this criterion for “age of pregnancy” are identified. In 1984, a majority of respondents were already past the age of 18, in which case our identification is based on information provided retrospectively. If in 1984 the respondent reports a past pregnancy where age of pregnancy is 18 or younger, and the outcome of the pregnancy is a live birth, then they are identified as a ‘teen-mother’. For respondents 18 years or younger and currently pregnant in 1984 (1985), data from survey year 1985 (1986) are used to identify the outcome of the pregnancy. If it is a live birth, then they are also identified as a ‘teen-mother’. Otherwise the variable is set to 0. Finally, all responses are compared across the 3 years for consistency. This procedure results in 1,310 respondents qualifying as ‘teen-mothers’ for our analysis. These numbers are consistent with Hotz et al. [3].


This group includes respondents who experienced a first pregnancy at age 18 or less, but it ended in a miscarriage, abortion, or stillbirth, and they did not experience a subsequent pregnancy ending in live birth while still a teenager. Again, 1984, 1985, and 1986 interview data were used to identify the occurrence and outcomes of the teen pregnancy. This procedure resulted in 467 women qualifying for this group. Of them, 129 respondents reported having a miscarriage, 320 respondents reported having an abortion, and 18 respondents reported having a stillbirth as a teenager. It is important to note that data on abortions, miscarriages, and pregnancies in NLSY79 interview data are based on self-reports, and it is also fairly well-established that the numbers of abortions among the young is highly underreported [24]. We speculate that some of the reported ‘miscarriages’ are likely to have been abortions, and that some teen pregnancies would have been terminated via abortion if they did not happen to end in a miscarriage. Thus, we avoid the approach by Hotz et al. [3] of only using information on women who report a miscarriage. Instead, we combine women who report a miscarriage/stillbirth as a teenager with those who report an abortion as a teenager in one group.


This group includes respondents who report engaging in non-contraceptive sexual activity as a teenager, but did not experience a pregnancy. We argue that this is a useful comparison group to teen-mothers, because their unmeasured personal and environmental characteristics lead them to engage in the risky behavior that leads to teen-motherhood, and it is likely a matter of chance that they did not become pregnant. The Fertility module in NLSY79 provides specific contraceptive related questions. Specifically, the respondents identifying themselves as “not using contraception, sexually active” when 18 years and under were included in this group. This procedure resulted in 238 respondents reporting non-contraceptive sexual activity (another 1,480 respondents reported sexual activity with contraception, but we do not consider them an appropriate comparison group). Again, the usual caveats exist about misreporting in self-reported, retrospective data.

We acknowledge that our comparison groups are unlikely to have the identical unmeasured characteristics to teen-mothers. For example, respondents in the teen-unprotected sex group may have timed the sexual intercourse more carefully to avoid pregnancy,—which could suggest differences in cognitive ability or in the desire to get pregnant between them and teen-mothers. Some teen-pregnancy-only respondents may have chosen an abortion or induced a miscarriage because they wanted to avoid the socio-economic consequences of teen-motherhood. Abortions and miscarriages may also be more indicative of experiencing sexual assault and having worse access to prenatal care, respectively. compared to teen-mothers. However, while the unmeasured personal and environmental characteristics may not be identical across the groups, they are likely to have a number of similarities. Taken in conjunction with a list of measureable characteristics we also control for other variables, which are listed below. We believe our approach allows us to minimize the effects of confounders when analyzing the association of teen-motherhood with future health outcomes.

Other Variables

In our multivariate regression framework, we control for several measureable demographic, familial and other background characteristics that are available in the NLSY79 that may be associated with teen-motherhood and the outcome of interest in our study. These include race-ethnicity, region of residence, whether the teenager lived with both parents at age 14, two proxy variables to control for the extent reading and learning were encouraged in the home, the number of siblings, highest grade completed by respondent’s mother, self-reported bad health in 1979, poverty status prior to 1979, if parents were alcoholics, and if the respondents themselves reported substance-use before age 14.


All statistical analyses were conducted using STATA, version 11.0. Table 2 illustrates the mean, and standard deviations for our outcome variables, as well as for the other demographic, familial and other background characteristics available in the NLSY79. The average PCS and MCS among teen-mothers were 49.91 and 50.89, respectively, while for all women they were 50.79 and 51.10, respectively. Notably, while teen-mothers have lower PCS and MCS compared to other women, they also have several other differences in their background characteristics compared to other women. For example, teen-mothers are less likely than counterparts to be living with both parents at age 14 (54% vs. 63%), more likely to belong to a minority group, and much more likely to be in poverty prior to 1979 (49% vs. 19%).
Table 2

Means and standard deviations of key variables



(N = 1,310)

All other women

(N = 2,961)










































Lived with both parents at age 14







Newspaper in household growing upa







Library card in household growing upa







In poverty before 1979







Alcoholic parents







Number of siblings







Highest grade completed







Bad health in 1979







Alcohol at age 14 or younger







Marijuana at age 14 or younger







aThese variables serve as proxy variables for the extent to which reading and learning were encouraged in the household where the respondent grew up

Table 3 presents the results of t-tests with unequal variances to test the equality of mean physical and mental scores for three categories. The first column compares PCS and MCS for teen-mothers to all other women. The second column compares teen-mothers to teen-pregnancy only. The last column compares teen-mothers to teen-unprotected-sex. These simple bivariate analyses reject the null hypothesis of equality of means for PCS in all cases. Thus, teen-mothers appear to exhibit poorer PCS later in life compared to all women, but also compared to women who had a pregnancy but not a live birth as a teen, or engaged in unprotected sex as a teen. However, while we reject the null of equality of means for MCS between teen-mothers and all other women, we do not find any statistical differences between the teen-mothers and the teen-pregnancy only groups, and only a weak statistical difference (significant at 10% but not 5%) between the teen-mothers and teen-unprotected-sex group.
Table 3

T-test statistics with unequal variances


PCS score

Mean (Group 1, Group2)

(t statistic)

MCS Score

Mean (Group 1, Group2)

(t statistic)

Teen-mothers (Group1)

compared to All other women (Group 2)

49.82, 51.85


50.89, 52.20


Teen-mothers (Group 1)

compared to teen-pregnancy-only (Group 2)

49.79, 51.08


50.89, 51.12


Teen-mothers (Group 1)

compared to Teen-unprotected-sex (Group 2)

49.82, 51.73


50.90, 51.87


Teen-mother group includes teenagers who had a live birth at age 18 or younger (N = 1,310)

Teen-pregnancy only includes teenagers who had a pregnancy but resulted in an abortion, miscarriage, or stillbirth at age 18 or younger (N = 467)

Teen-unprotected-sex group includes teenagers who engaged in sexual activity at age 18 or younger but did not experience teen pregnancy (N = 238)

aSignificant at the P < 0.10 level

bSignificant at the P < 0.05 level

cSignificant at the P < 0.01 level

Table 4 presents results from three sets of multivariate Ordinary Least Squares (OLS) regressions (referred to as OLS Regression 1, 2 and 3 respectively) with PCS and MCS as the outcome variable. Regression 1 includes a binary measure for teen-mothers, with all other women being the comparison group or omitted category, and it only controls for the age at which PCS and MCS scores were measured. Regression 2 extends Regression 1 by controlling for the full set of variables listed under ‘other variables’. This allows us to inspect how teen mothers differ from all other women in their PCS and MCS scores after accounting for differences in measureable demographic, familial and other background characteristics. Regression 3 adds in separate binary measures for teen-pregnancy-only and teen-unprotected-sex in addition to the binary measure for teen-mothers, with the omitted category now being women not in any of these groups. This allows us to inspect whether the association between teen-motherhood (compared to the omitted-category) and PCS and MCS are statistically different from the associations between the other two groups (compared to the base-category) and these outcomes.
Table 4

Results of multivariate ordinary least square regression analysis

Variable name

Regression 1

Regression 2

Regression 3



(t stat)



(t stat)



(t stat)



(t stat)



(t stat)



(t stat)














Teen-pregnancy only






Teen unprotected sex






Age when MCS, PCS score computed

































Not in SMSA










SMSA not central city










SMSA central city










Region (North central)










Region (South)










Region (West)










Poverty status when young










Drink before 14 years










Smoke before 14 years










Marijuana before 14 years










Bad health condition










Highest grade completed










Newspaper in household










Library card in household










Lived with both parents at age 14










Number of siblings










Tee-mother group includes teenagers who had a live birth at age 18 or younger

Teen-pregnancy-only group includes teenagers who had a pregnancy but resulted in an abortion, miscarriage, or stillbirth at age 18 or younger

Teen-unprotected-sex group includes teenagers who engaged in sexual activity at age 18 or younger but did not experience a teen pregnancy

Comparison group (or omitted-category) in Regressions 1 and 2 are all women who did not become teen mothers. Comparison group (or omitted-category) in Regression 3 is all women who neither became teen mothers, nor experienced a teen pregnancy, nor had unprotected sex as a teen

Race: reference group = Non-Hispanic white

SMSA: reference group = SMSA central city not known

Region: reference group = East

aSignificant at the P < 0.10 level

bSignificant at the P < 0.05 level

cSignificant at the P < 0.01 level

When interpreting the magnitudes of the estimated results, it is useful to remember that a 1 point reduction in the PCS or MCS score is equivalent to a one-tenth of one (population) standard-deviation reduction.

Regression 1 results show teen mothers have lower PCS (β = −2.095, P < 0.01) and lower MCS (β = −1.336, P < 0.01) scores compared to all other women. Regression 2 results show that, when the other demographic, familial and other background characteristics are controlled for, the estimated reductions in PCS (β = −1.596, P < 0.01) and MCS (β = −0.903, P < 0.01) for teen mothers compared to all other women are smaller, but still highly significant. Regression 3 results show that, compared to the omitted-category, teen mothers have significantly lower PCS (β = −1.590, P < 0.01) and MCS (β = −1.065, P < 0.01). In contrast, neither the teen-pregnant-only nor teen-unprotected-sex respondents have PCS that are statistically different than the omitted category. However, in case of MCS, while the teen-unprotected-sex group is not statistically different from the omitted category, the teen-pregnancy-only group shows statistically lower MCS compared to the omitted category (β = 1.379, P < 0.01), with the magnitude of the negative association being even greater than the teen-mothers group. In an F-test, we failed to reject the null hypothesis that teen-mothers and teen-pregnancy-only groups are equivalent in terms of reductions in MCS.

Discussion and Conclusion

Extant literature has explored whether teen-motherhood is linked to outcomes such as educational attainment, or employment, poverty and welfare dependency in early adulthood. To our knowledge, this is the first study to examine whether teen-motherhood has long term consequences in terms of women’s physical and mental health later in life.

We find significant and negative associations between teen-motherhood and women’s physical and mental health in their 40 s as measured by the Physical Component Summary and Mental Component Summary meta-scores from NLSY79 Health module. Our approach involves using regression models that control extensively for measured background factors, and also comparing teen-mothers with women at a very high risk of experiencing teen-motherhood—specifically, women who experienced a teen pregnancy but not a live birth, and women who had unprotected sex as a teen but chanced not to get pregnant. The motivation is to investigate whether it is teen-motherhood per se that leads to adverse health outcomes, or whether the adverse outcomes are a function of measured and unmeasured factors that increase the likelihood for teen-motherhood.

Our results strongly suggest that teen-motherhood does, indeed, lead to poorer physical health in later life. While further research is needed to verify exactly why this happens, we speculate that perhaps the economic consequences of teen-motherhood as well as the stresses of childrearing at a young age leave these women with fewer resources to invest in their own physical health. These results suggest that resources devoted to reducing teen-motherhood may lead to health cost savings in future decades.

Results regarding mental health are more challenging. Both teen-motherhood and teen pregnancies not ending in a live birth have significant and statistically similar negative associations with future mental health. This may indicate that it is not teen-motherhood per se, but the unmeasured factors leading to a teen pregnancy, that actually lead to worse mental health. It may also indicate that experiencing a teen pregnancy per se has a negative effect on mental health, regardless of whether the pregnancy results in a live birth, and that a pregnancy ending in a non-live birth may be worse for mental health than those ending in live births—which is similar to what another set of researchers found using similar data [25].

We acknowledge several shortcomings of this study. There are the usual concerns regarding accuracy in self-reported data, particularly given the sensitive nature of issues like pregnancy outcomes. We do not have information on some crucial factors that may predict teen pregnancy outcomes as well as subsequent health, such as experiences with sexual abuse and rape. Such factors have been found to be associated with women having abortions and also suffering from anxiety/depression and may help explain why mental health outcomes were more negative for the teen-pregnancy-only group compared to teen-mothers in our sample [26, 27]. We also lack information on health insurance status and access to pre-natal care among the teens in our sample. Finally, as acknowledged previously, there is good reason to believe that the teen-mothers and our two comparison groups are not identical in terms of their unmeasured characteristics. Therefore, there may remain other unmeasured confounders, hence even for physical health outcomes we must use caution before interpreting our results as being ‘causal’.

This paper also suggests that more research needs to be done on the long-term consequences of teen-motherhood. The NLSY79 Health module provides a rich array of health information beyond the SF-12, which permits further research into the association of teen-motherhood with different health-outcomes. Finally, longitudinal information is also available the children of the female respondents of NLSY79, which provide a rich resource for comparing outcomes of the children of teen-mothers to the children of their counterparts, to understand whether the effects of teen-motherhood persist across generations.

Copyright information

© Springer Science+Business Media, LLC 2011