Introduction
The question of whether there is a relationship between having an abortion and subsequent mental health has been the focus of much policy and research over the past 20 years. While some researchers have found that having an abortion is not associated with subsequent depression (Major et al. Reference Major, Cozzarelli, Cooper, Zubek, Richards, Wilhite and Gramzow2000; Pope et al. Reference Pope, Adler and Tschann2001; Schmiege & Russo, Reference Schmiege and Russo2005; Warren et al. Reference Warren, Harvey and Henderson2010; Steinberg et al. Reference Steinberg, Becker and Henderson2011; Biggs et al. Reference Biggs, Upadhyay, Mcculloch and Foster2017), self-esteem (Russo & Zierk, Reference Russo and Zierk1992; Pope et al. Reference Pope, Adler and Tschann2001; Warren et al. Reference Warren, Harvey and Henderson2010; Steinberg et al. Reference Steinberg, Becker and Henderson2011), anxiety (Pope et al. Reference Pope, Adler and Tschann2001; Steinberg & Russo, Reference Steinberg and Russo2008; Steinberg & Finer, Reference Steinberg and Finer2011; Steinberg et al. Reference Steinberg, Mcculloch and Adler2014; Biggs et al. Reference Biggs, Upadhyay, Mcculloch and Foster2017; van Ditzhuijzen et al. Reference Van Ditzhuijzen, Ten Have, De Graaf, Van Nijnatten and Vollebergh2017), mood disorder (Steinberg & Finer, Reference Steinberg and Finer2011; Steinberg et al. Reference Steinberg, Mcculloch and Adler2014; van Ditzhuijzen et al. Reference Van Ditzhuijzen, Ten Have, De Graaf, Van Nijnatten and Vollebergh2017), suicidal ideation (Steinberg et al. Reference Steinberg, Mcculloch and Adler2014), stress (Harris et al. Reference Harris, Roberts, Biggs, Rocca and Foster2014), or post-traumatic stress (Biggs et al. Reference Biggs, Rowland, Mcculloch and Foster2016), limited research has indicated that women who have abortions may be more likely to report depression (Cougle et al. Reference Cougle, Reardon and Coleman2003; Rue et al. Reference Rue, Coleman, Rue and Reardon2004; Coleman et al. Reference Coleman, Coyle, Shuping and Rue2009), anxiety (Cougle et al. Reference Cougle, Reardon and Coleman2003), substance abuse (Coleman et al. Reference Coleman, Reardon, Rue and Cougle2002; Reardon et al. Reference Reardon, Coleman and Cougle2004; Coleman et al. Reference Coleman, Coyle, Shuping and Rue2009) and receiving psychological counseling (Coleman, Reference Coleman2006). Three critical reviews (Charles et al. Reference Charles, Polis, Sridhara and Blum2008; Major et al. Reference Major, Appelbaum, Beckman, Dutton, Russo and West2009; Robinson et al. Reference Robinson, Stotland, Russo, Lang and Occhiogrosso2009) have attributed these disparate findings to significant methodological flaws in studies that find an association between having an abortion and mental health. These flaws included the use of inappropriate comparison groups; failure to adjust for mental health status prior to pregnancy; lack of or inadequate adjustment for co-occurring and pre-existing risk factors, particularly demographic, social and structural variables that may be associated with both abortion and mental health outcomes; sampling bias; measurement and temporality of abortion and mental health variables; and use of inappropriate statistical analysis techniques. Despite the lack of empirical support for such policies, 20 states currently mandate that women seeking abortions be counseled on potential psychological consequences of abortion, with six states emphasizing information on negative emotional responses (Guttmacher Institute, 2017a ).
To date, few studies examining abortion and mental health in the USA have focused specifically on the experiences of young women, particularly using nationally representative data. The definition of young adulthood varies, but it commonly encompasses ages 18–29, with some scholars extending the maximum age to 40, and others identifying ages 18–25 as a distinct transitional period between adolescence and young adulthood (Levinson, Reference Levinson1986; Erikson & Erikson, Reference Erikson and Erikson1998; Arnett, Reference Arnett2000; Arnett et al. Reference Arnett, Žukauskienė and Sugimura2014). Understanding the relationship between abortion and mental health during the various developmental stages of the life course is critical, given the normative timing of key events related to relationships, sexual behavior and reproduction, and ubiquity of instability and uncertainty during the transition to adulthood (Arnett, Reference Arnett2000; Arnett et al. Reference Arnett, Žukauskienė and Sugimura2014). For example, 76% of pregnancies among 18–19-year-old women were classified as unintended in 2012, compared with 45% among all women of reproductive age (Finer & Zolna, Reference Finer and Zolna2016); and women younger than age 30 constitute 72% of all abortion patients (Jerman et al. Reference Jerman, Jones and Onda2016). In the USA, young adults experience mental disorders at higher rates than other age groups (Kessler et al. Reference Kessler, Birnbaum, Demler, Falloon, Gagnon, Guyer, Howes, Kendler, Shi, Walters and Wu2005). While some research has examined abortion during adolescence and subsequent mental health, this relationship has not been investigated during young adulthood, when many women experience their first pregnancies (Zabin et al. Reference Zabin, Hirsch and Emerson1989; Pope et al. Reference Pope, Adler and Tschann2001; Warren et al. Reference Warren, Harvey and Henderson2010; Leppälahti et al. Reference Leppälahti, Heikinheimo, Kalliala, Santalahti and Gissler2016). In particular, it is unclear whether the impact of having an abortion after an unwanted first pregnancy on subsequent mental health varies by age of first pregnancy. Populations experiencing the greatest social and health inequities – such as women identifying with racial or ethnic minority groups or of low-income status—have their first pregnancies, on average, earlier than their more advantaged counterparts in the USA (Finer & Zolna, Reference Finer and Zolna2016; Romero et al. Reference Romero, Pazol, Warner, Cox, Kroelinger, Besera, Brittain, Fuller, Koumans and Barfield2016). These same groups are more likely to experience trauma, such as adverse childhood experiences and intimate partner violence, which has important implications for subsequent mental health (Lipsky et al. Reference Lipsky, Caetano and Roy-Byrne2009; Slopen et al. Reference Slopen, Shonkoff, Albert, Yoshikawa, Jacobs, Stoltz and Williams2016).
Two studies have utilized data from the National Longitudinal Study of Adolescent Health (Add Health) to examine the impact of abortion during adolescence in the USA. Using the first two waves of Add Health data, Coleman (Reference Coleman2006) found that adolescents who had abortions had increased odds of ever receiving counseling for psychological counseling and sleep difficulties compared with adolescents who gave birth after an unintended pregnancy. However, in this study, both the pregnancy and mental health measures reflected the last 12 months, thus making it impossible to determine if the abortion occurred before or after the outcome. In the second analysis, Warren et al. (Reference Warren, Harvey and Henderson2010) examined a subsample of 289 female adolescents who experienced an unintended pregnancy between the first and second waves of data collection. Compared with adolescents whose unintended pregnancies resulted in live births, adolescents who had abortions had no increased risk of subsequent depressive symptoms or self-esteem.
The present analysis builds on the prior research with Add Health and utilizes longitudinal data capturing pregnancies experienced after adolescence, when more than half of female respondents experienced their first pregnancy. In particular, this analysis aimed to understand whether there was a relationship between having an abortion and subsequent depressive symptoms among women who experienced unwanted first pregnancies in young adulthood. Analyses focused on first pregnancies in order to most accurately determine the sequence of pregnancy and depressive symptoms (Steinberg et al. Reference Steinberg, Becker and Henderson2011). Moreover, this study was informed by the common risk factors framework, which posits that factors beyond the immediate pregnancy context, such as socioeconomic status, prior mental health, or history of violence victimization, are critical in post-abortion mental health. Neglecting to account for such factors may generate a spurious association between having an abortion and subsequent mental health (Steinberg & Finer, Reference Steinberg and Finer2011).
Methods
Data from the 15-year National Longitudinal Study of Adolescent Health, commonly known as Add Health, were utilized. Add Health is a school-based study, nationally representative of students in grades 7–12 during the 1994–1995 school year (Harris, Reference Harris2013). This rich, longitudinal dataset has been utilized by thousands of researchers to study wide-ranging, policy-relevant topics, including physical activity, youth violence, and sexual behavior. Schools were the primary sampling unit, with the sample frame derived from the Quality Education Database. Using this frame, 80 schools were selected for a sample stratified by urbanicity, region, school type (public or private), racial/ethnic composition, and school size. At Wave 1 (1994–1995), a core sample of 20 745 students was selected to complete in-home questionnaires. Parents of these adolescents were recruited to complete a separate parental survey. One year later (1996), Wave 2 in-home data were collected for all participants who had not yet graduated high school. Approximately 5 years later (2001–2002), Wave 3 data were collected, when participants were between the ages of 18 and 25. Finally, Wave 4 data were collected when the Add Health participants were between ages 24 and 32 (2007–2008). This analysis utilized the full, restricted-use dataset, available to certified researchers only by contractual agreement with the University of North Carolina at Chapel Hill. The Committee for Protection of Human Subjects at the University of California, Berkeley approved the study protocol.
Measures
Depressive symptomatology was both the outcome variable and a primary independent variable of interest. Several systematic reviews have indicated that failure to adjust for prior mental health in statistical models is a critical methodological flaw in studies examining abortion and mental health (Charles et al. Reference Charles, Polis, Sridhara and Blum2008; Major et al. Reference Major, Appelbaum, Beckman, Dutton, Russo and West2009; Steinberg & Russo, Reference Steinberg and Russo2009). Thus, data from all four waves of Add Health were utilized in this analysis. Depressive symptoms in the last 7 days were measured at each wave using a condensed version of the Center for Epidemiological Studies Depression Scale (CES-D). The CES-D is a well-validated, self-report scale for assessing depressive symptomatology (Radloff, Reference Radloff1977). The full CES-D includes 20 items assessing negative affect, positive affect, somatic complaint and interpersonal relations. The CES-D has been frequently used to examine mental health in the Add Health study population (Goodman, Reference Goodman1999; Shrier et al. Reference Shrier, Harris, Sternberg and Beardslee2001; Rushton et al. Reference Rushton, Forcier and Schectman2002; Swallen et al. Reference Swallen, Reither, Haas and Meier2005; Spriggs & Halpern, Reference Spriggs and Halpern2008; Primack et al. Reference Primack, Swanier, Georgiopoulos, Land and Fine2009; Warren et al. Reference Warren, Harvey and Henderson2010; Frisco et al. Reference Frisco, Houle and Lippert2013). Previous research has indicated that use of as few as four CES-D items provides predictive validity similar to the full scale (Grzywacz et al. Reference Grzywacz, Hovey, Seligman, Arcury and Quandt2006). The number of CES-D items included in Add Health varied by survey wave, with 19 items in Waves 1 and 2; 9 items in Wave 3; and 10 items in Wave 4. Nine items appeared in all four waves of Add Health. Respondents were asked how often in the prior 7 days the following had occurred: they felt sad, enjoyed life, felt as good as other people, had trouble keeping their mind on what they were doing, were depressed, felt that people disliked them, were bothered by things that do not usually bother them, and could not shake off the blues. Respondents selected how frequently each item occurred (never, rarely, most of the time, or all of the time). After reverse coding positively worded items, CES-D scores were computed using these nine items. If respondents were missing five or fewer items, the mean of the answered items was substituted for the missing items (Hall et al. Reference Hall, Richards and Harris2017). For each wave, the internal consistency (Cronbach's alpha) was calculated for all women surveyed and was greater than 0.80. The nine-item CES-D scores for Waves 1, 2 and 4 each had correlations greater than 0.95 with the scores using the full sets of available items, indicating high criterion validity of the briefer version. The CES-D score was utilized in two ways: (1) as a continuous variable; and (2) to create a binary variable indicative of moderate to severe depressive symptoms. Following previous research using Add Health data, respondents who scored 11 or higher on the nine-item CES-D were classified as having moderate or severe depressive symptoms (Spriggs & Halpern, Reference Spriggs and Halpern2008; Primack et al. Reference Primack, Swanier, Georgiopoulos, Land and Fine2009; Frisco et al. Reference Frisco, Houle and Lippert2013; Nkansah-Amankra & Tettey, Reference Nkansah-Amankra and Tettey2015). Sensitivity analyses were conducted to examine using a cutoff score of 10 with the 9-item measure, or one standard deviation above the mean using all CES-D items available at each wave (Gotlib et al. Reference Gotlib, Lewinsohn and Seeley1995). The relationship between abortion and the binary depressive symptoms variable was consistent across the various measurement approaches.
The key independent variable was the outcome of unwanted first pregnancy (abortion or live birth). At Wave 4, each respondent provided a complete, retrospective pregnancy history. For each pregnancy, participants were asked, ‘Thinking back to the time just before this pregnancy with [partner's initials], did you want to have a child then?’ Pregnancies were considered unwanted when participants responded ‘no’ to this question. The pregnancy outcome (including abortion and live birth) was also solicited.
Extending the common risk factors approach to investigating the relationship between abortion and mental health, another key independent variable is an index of traumatic experiences before age 18 (Steinberg & Finer, Reference Steinberg and Finer2011). Eight items drawn from the Waves 3 and 4 surveys were included in the trauma index. These included five types of traumatic experiences occurring before age 18 (psychological, physical, and sexual abuse by a parent or adult caregiver; death of a parent or parental figure; and incarceration of a parent or parental figure); two items capturing physical and non-physical experiences of forced sex by someone other than a parent or caregiver prior to the measurement of the outcome variable; and experience of intimate partner violence in the past year. The trauma index was constructed as an ordinal variable (0, 1, 2, or 3+ traumatic experiences reported).
Additional covariates included current age (Wave 4); age at first pregnancy (Wave 4); length of time between the pregnancy and measurement of the outcome variable, in months (Waves 3 and 4); receipt of public assistance by a household member before age 18 (reported at any wave or in the Wave 1 parental survey); race/ethnicity (Hispanic, non-Hispanic Black, non-Hispanic White, non-Hispanic other; Wave 1); current educational attainment (Wave 3 or 4); and relationship to partner at the time of pregnancy (Wave 4).
Sample
First, data from all women who completed the Wave 4 survey and were not missing data on pregnancy end dates were used to understand the distribution of age at first pregnancy in the Add Health sample (unweighted n = 7870). Second, to investigate the relationship between having an abortion after an unwanted first pregnancy and subsequent depressive symptoms, pregnancy histories of women who completed all four Add Health surveys and were not missing the end date of their first pregnancy were used to identify two analytic subsamples. The Wave 3 subsample included 856 women who experienced an unwanted first pregnancy between Waves 2 and 3, while the Wave 4 subsample included 438 women whose unwanted first pregnancy occurred between Waves 3 and 4 (unweighted n’s). Because the models were organized around respondents’ first pregnancy, there was no overlap between the two subsamples, nor with a previous analysis examining pregnancies prior to Wave 2 (Warren et al. Reference Warren, Harvey and Henderson2010). To ensure the creation of an appropriate comparison group, women whose first pregnancies were considered wanted or resolved in outcomes other than abortion or live birth were excluded (Charles et al. Reference Charles, Polis, Sridhara and Blum2008; Steinberg & Russo, Reference Steinberg and Russo2008; Major et al. Reference Major, Appelbaum, Beckman, Dutton, Russo and West2009). Additionally, any woman who gave birth three months prior to an interview date was excluded, as the depressive symptoms measure may be indicative of postpartum depression (Charles et al. Reference Charles, Polis, Sridhara and Blum2008). There were six cases excluded for data quality because multiple, singleton first pregnancies with different pregnancy resolution outcomes and/or wantedness were reported in the same month. Finally, women missing more than five CES-D items at any wave, the relevant grand sample weight, pregnancy end dates, the wantedness or outcome of first pregnancy, or covariate data were excluded from the subsamples.
Statistical analyses
All analyses were conducted using Stata statistical analysis software, version 13.1 (College Station, TX: StataCorp LP). F tests were employed to assess statistically significant differences between the timing of first pregnancy and demographic characteristics reflecting social disadvantage (race/ethnicity, educational attainment, and maternal educational attainment), as well as to compare characteristics of women included in each subsample. Population-averaged lagged logistic and linear regression models were used to examine the association of unwanted first pregnancy outcome with subsequent depressive symptoms. Multivariable models adjusted for lagged depressive symptoms, the trauma index, sociodemographic characteristics, relationship, status and age when the pregnancy ended. For each subsample, four logistic and four linear regression models were estimated: (1) an unadjusted model; (2) an adjusted model including lagged depressive symptoms and covariates; (3) an adjusted model including the trauma index and covariates; and (4) a fully adjusted model including lagged depressive symptoms, the trauma index, and covariates. To adjust for the complex design features of Add Health, Stata's SVY commands were employed for all analyses, including application of sampling weights, adjustments for clustering and usage of Taylor series linearized standard errors (Chantala & Tabor, Reference Chantala and Tabor2010).
Results
By the Wave 4 interview, nearly two-thirds (63.0%) of female Add Health participants had experienced their first pregnancy (Table 1). The majority of participants had their first pregnancy by age 25. Bivariate analyses indicated that women who were members of groups typically experiencing greater social advantage in the USA were more likely to have never been pregnant or to first become pregnant at an older age. For example, 25.3% of Black women had never been pregnant, compared with 39.6% of White women, 37.3% of Latina women, and 40.5% of women with other racial/ethnic identities (p < 0.001). The majority of women with a college degree (59.4%) had never been pregnant, compared with a minority of women with lower levels of educational attainment (18.5–27.0%, p < 0.001). A similar trend existed for maternal educational attainment.
Table 1. Age at first pregnancy by selected demographic characteristics
Notes: Unweighted n = 7870. Data collected in the USA between 1994 and 2008. ***p < 0.001. Two women were missing data on educational attainment, and 124 women were missing data on maternal educational attainment.
Descriptive statistics for the subsamples of women with unwanted first pregnancies ending in abortion or live birth are provided in Table 2, as well as results of F tests comparing the two subsamples with regards to race/ethnicity, receipt of public assistance before age 18, relationship to partner at the time of first pregnancy, and the binary measures of moderate/severe depressive symptoms at Waves 1–4. The profile of Wave 3 subsample participants was characterized by markers of greater social disadvantage than their older counterparts in the Wave 4 subsample. For example, greater proportions of Black and Latina women were represented in the Wave 3 subsample (p < 0.01). Among the Wave 3 subsample, 28.9% of women reported living in households receiving public assistance before age 18, compared with 21.9% of the Wave 4 subsample (p < 0.05). Greater proportions of women in the Wave 3 subsample experienced depressive symptoms in adolescence (Waves 1 and 2, p < 0.01) and young adulthood (Wave 3, p < 0.05) compared with the Wave 4 subsample. Notably, there was not a statistically significant difference between the two subsamples with regards to pregnancy outcome (abortion or live birth) or Wave 4 depressive symptoms.
Table 2. Characteristics of subsamples
Notes: Wave 3 unweighted n = 848. Wave 4 unweighted n = 438. Data collected in the USA between 1994 and 2008. Moderate or severe depressive symptoms were indicated by a score of 11 or higher on the nine-item Center for Epidemiologic Studies Depression Scale.
*p < 0.05, **p < 0.01, ***p < 0.001.
In Table 3, unadjusted and adjusted results from population-averaged lagged logistic and linear regression models for the Wave 3 subsample are presented. In the unadjusted (1) and adjusted (2–4) models for both binary and continuous outcome measures of depressive symptoms, there was not an association between having an abortion after an unwanted first pregnancy and subsequent depressive symptoms. In the adjusted logistic regression model (2) including lagged depressive symptoms measures and covariates, the most recent (Wave 2) measure of prior depressive symptoms was the strongest correlate of depressive symptoms at Wave 3 (OR 2.72, 95% CI 1.54–4.80). In the linear regression models, both Wave 1 and 2 CES-D scores were positively associated with the Wave 3 CES-D score. In adjusted model 3, the trauma index was positively associated with the binary depressive symptoms outcome (OR 1.41, 95% CI 1.10–1.81) and the CES-D score (β 0.75, 95% CI 0.39–1.10). In the fully adjusted model (4) including lagged depressive symptoms at Waves 1 and 2, the trauma index, and covariates, Wave 2 depressive symptoms (OR 2.65, 95% CI 1.48–4.72) and the trauma index (OR 1.35, 95% CI 1.05–1.73) were positively associated with the binary Wave 3 depressive symptoms outcome. In the fully adjusted linear regression model (4), the trauma index (β 0.57, 95% CI 0.25–0.89), Wave 1 CES-D score (β 0.19, 95% CI 0.11–0.27), and Wave 2 CES-D score (β 0.22, 95% CI 0.12–0.31) were positively associated with the Wave 3 CES-D score.
Table 3. Unadjusted and adjusted population-average logistic and linear regression analysis results for unwanted first pregnancies occuring between Waves 2 and 3
Notes: Unweighted n = 848. Data collected in the USA between 1994 and 2008. Models 2–4 adjust for current age, age at first pregnancy, months passed since first pregnancy, receipt of public assistance by a household member before age 18, race/ethnicity, current educational attainment, and relationship to partner at the time of pregnancy.
*p < 0.05, **p < 0.01, ***p < 0.001, + p < 0.10.
Similarly, for the Wave 4 subsample (Table 4), there was not a relationship between having an abortion and subsequent depressive symptoms in unadjusted (1) or adjusted (2–4) models employing either analytic approach. In model 2, adjusting for prior depressive symptoms and covariates, the most recent (Wave 3) measure of depressive symptoms had the strongest relationship to Wave 4 depressive symptoms for both the binary and continuous outcomes. In the adjusted model (3) including the trauma index, there was a positive relationship between trauma and Wave 4 depressive symptoms as a binary (OR 1.63, 95% CI 1.13–2.35) and continuous outcome (β 0.97, 95% CI 0.54–1.40). For the fully adjusted logistic regression model (4), depressive symptoms at Wave 3 had a positive and statistically significant relationship with Wave 4 depressive symptoms (OR 8.26, 95% CI 3.30–20.65), while the relationship between trauma and depressive symptoms was attenuated. For the linear regression model, both the Wave 3 CES-D score (β 0.42, 95% CI 0.30–0.54) and the trauma index (β 0.53, 95% CI 0.11–0.94) were significantly associated with the Wave 4 CES-D score.
Table 4. Unadjusted and adjusted population-average logistic and linear regression analysis results for unwanted first pregnancies occuring between Waves 3 and 4
Notes: Unweighted n = 438. Data collected in the USA between 1994 and 2008. Models 2–4 adjust for current age, age at first pregnancy, months passed since first pregnancy, receipt of public assistance by a household member before age 18, race/ethnicity, current educational attainment, and relationship to partner at the time of pregnancy.
*p < 0.05, **p < 0.01, ***p < 0.001, + p < 0.10.
Discussion
Using nationally representative, longitudinal data, this analysis found that young women whose unwanted first pregnancies ended in abortion were at no greater risk of subsequent depressive symptoms than their counterparts who gave birth. Similar to an earlier study with Add Health data, this was the case even before adjusting for prior depressive symptoms and trauma history (Warren et al. Reference Warren, Harvey and Henderson2010). For both the Wave 3 and Wave 4 subsamples, prior measures of depressive symptoms were consistently associated with subsequent depressive symptoms in adjusted logistic and linear regression models. In three of the fully adjusted models, the trauma index also was associated with subsequent depressive symptoms. Notably, in comparing the subsample demographic profiles, there were some differences by social advantage, yet the relationship between having an abortion and subsequent depressive symptoms was consistent across the models and analytic approaches.
This research, along with a prior analysis of Add Health data, indicates that women who experienced their first pregnancies in adolescence or young adulthood were not at increased risk of subsequent depressive symptoms (Warren et al. Reference Warren, Harvey and Henderson2010). Both developmental periods are significant. Adolescents under age 18 may face restricted access to abortion, through the requirement of parental notification laws, while women under age 30 constitute the majority of people accessing abortion and experience the highest rates of unintended pregnancy (Finer & Zolna, Reference Finer and Zolna2016; Jerman et al. Reference Jerman, Jones and Onda2016; Guttmacher Institute, 2017b ). Moreover, young women are in a precarious developmental period characterized by greater risk of adverse mental health outcomes (Kessler et al. Reference Kessler, Birnbaum, Demler, Falloon, Gagnon, Guyer, Howes, Kendler, Shi, Walters and Wu2005; Arnett et al. Reference Arnett, Žukauskienė and Sugimura2014). These results stand in contrast to a highly criticized meta-analysis that found an increased risk of depression after an abortion (Coleman, Reference Coleman2011; Goldacre & Lee, Reference Goldacre and Lee2012; Kendall et al. Reference Kendall, Bird, Cantwell and Taylor2012; Steinberg et al. Reference Steinberg, Trussell, Hall and Guthrie2012). These disparate findings are likely due to statistical, measurement, and design flaws that threaten the validity of many studies finding a link between abortion and subsequent depression (Major et al. Reference Major, Appelbaum, Beckman, Dutton, Russo and West2009). Indeed, the present study's results are in line with findings from previous research with adult women and adolescents that do attend to these important methodological issues (Major et al. Reference Major, Cozzarelli, Cooper, Zubek, Richards, Wilhite and Gramzow2000; Pope et al. Reference Pope, Adler and Tschann2001; Schmiege & Russo, Reference Schmiege and Russo2005; Warren et al. Reference Warren, Harvey and Henderson2010; Steinberg et al. Reference Steinberg, Becker and Henderson2011; Biggs et al. Reference Biggs, Upadhyay, Mcculloch and Foster2017). Taken together, this body of work indicates that concerns about subsequent mental health do not provide support for policies to restrict abortion access.
Strengths of this analysis included use of a 15-year, nationally representative dataset, allowing for adjustment for prior depressive symptoms and utilization of an appropriate comparison group (i.e., women whose unwanted first pregnancy ended in live birth) (Major et al. Reference Major, Appelbaum, Beckman, Dutton, Russo and West2009). By examining respondents’ first pregnancies and using measures of depressive symptoms that captured the past 7 days, this analysis avoided the temporality issues present in previous research investigating the relationship between abortion and mental health (Charles et al. Reference Charles, Polis, Sridhara and Blum2008; Major et al. Reference Major, Appelbaum, Beckman, Dutton, Russo and West2009; Robinson et al. Reference Robinson, Stotland, Russo, Lang and Occhiogrosso2009; Steinberg & Finer, Reference Steinberg and Finer2012). Additionally, this analysis extended the use of the common risk factors approach to more broadly consider trauma, which has been shown to be an important influence on adult mental health (Dube et al. Reference Dube, Anda, Felitti, Chapman, Williamson and Giles2001; Chapman et al. Reference Chapman, Whitfield, Felitti, Dube, Edwards and Anda2004; Herrenkohl et al. Reference Herrenkohl, Hong, Klika, Herrenkohl and Russo2013).
This analysis is not without limitations. Underreporting of abortion is a well-established measurement issue in national surveys (Jones & Kost, Reference Jones and Kost2007). The impact of abortion underreporting on the relationship between abortion and subsequent mental health is unknown, though a few studies have suggested that women with better mental health may be less likely to disclose abortions in surveys (Jagannathan, Reference Jagannathan2001; Schmiege & Russo, Reference Schmiege and Russo2005). At the same time, abortion is highly stigmatized in the USA, and individuals experiencing greater stigma may be less likely to self-report their abortions (Cockrill et al. Reference Cockrill, Upadhyay, Turan and Foster2013). Additionally, this analysis focused on pregnancies classified as unwanted using a retrospective and binary measure that may be sensitive to recall bias and oversimplify the complex nature of pregnancy intentions (Santelli et al. Reference Santelli, Rochat, Hatfield-Timajchy, Gilbert, Curtis, Cabral, Hirsch and Schieve2003). Because this analysis focused on first pregnancies, we were unable to account for the impact of multiple pregnancies on subsequent depressive symptoms. Future analyses utilizing longitudinal data analytic approaches would extend this work to holistically capture the diversity of women's reproductive experiences across the life course. Moreover, while women who chose to give birth after an unwanted first pregnancy are a more appropriate comparison group than all women who gave birth irrespective of wantedness, these two groups may be different in ways beyond the sociodemographic variables adjusted for in multivariable models (Major et al. Reference Major, Appelbaum, Beckman, Dutton, Russo and West2009). Finally, the use of an abbreviated CES-D scale presents additional limitations. The abbreviated measures demonstrated high construct validity with the full set of items available in Waves 1, 2, and 4; nonetheless, the content of these nine items appears to more accurately capture general malaise or distress as opposed to depressive symptoms. One of the CES-D items refers to ‘feeling blue’, a concept that may not contemporarily convey the meaning intended when the scale was developed in the 1970s (Radloff, Reference Radloff1977).
Conclusion
Despite the lack of evidence demonstrating an association between having an abortion and subsequent depression and other mental health outcomes, policies in six states continue to require counseling of women about negative psychological consequences of abortion (Guttmacher Institute, 2017a ). This analysis and the growing body of work that finds no link between having an abortion and subsequent mental health suggests that policy and clinical efforts that aim to improve women's mental health should focus elsewhere. In particular, intervention and policy efforts to promote positive mental health outcomes and reduce exposure to trauma among adolescent and young women—regardless of pregnancy status—may have the greatest impact throughout the life course.
Acknowledgements
This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Information on how to obtain the Add Health data files is available on the Add Health website (http://www.cpc.unc.edu/addhealth). No direct support was received from grant P01-HD31921 for this analysis.The author thanks the Sexual Health and Reproductive Equity writing group for providing feedback on the manuscript.
This work was supported by the Society of Family Planning Research Fund (grant number SFPRF7-14).
Declaration of Interest
None.
