|Home | About | Journals | Submit | Contact Us | Français|
Research has suggested the importance of pregnancy desire in explaining pregnancy risk behavior among adolescent females. Much of the literature, however, uses cross-sectional study designs to examine this relationship. Because bias may strongly influence these results, more prospective studies are needed to confirm the relationship between pregnancy desire and pregnancy incidence over time.
Non-pregnant adolescents aged 14-19 years (N=208) completed baseline interviews and interviews every 6 months thereafter for 18 months. Logistic regression was used to examine demographic and psychosocial correlates of pregnancy desire. Cox regression analysis was used to determine if pregnancy desire predicted pregnancy incidence over time after controlling for potential confounders.
Twenty-four percent of participants either desired pregnancy or were ambivalent towards pregnancy in the next year. Pregnancy desire was associated with older age, relationship duration of less than 6 months, and higher perceived stress. After accounting for potential confounders, pregnancy desire doubled the risk of becoming pregnant over the 18 month follow-up period (RR=2.00, 95% CI=0.99-4.02). Additionally, a heightened risk for pregnancy was found among those who expressed some desire for pregnancy and who were not in school compared to those who expressed no desire for pregnancy and who were in school (RR=4.84, 95% CI=1.21-19.31).
Our analysis reinforces the importance of evaluating pregnancy desire among sexually active adolescent females. Interventions should target young women in new romantic relationships and who are not in school to improve pregnancy prevention efforts. Additionally, improving coping abilities may help reduce feelings of pregnancy desire among adolescent females.
Since the 1990's, the United States has experienced significant declines in teenage pregnancy and birth rates. Between 1991 and 2005 birth rates among women 15 to 19 years old dropped 34% from 61.8 to 40.5 births per 1,000 women. More recently, however, national data have suggested increases in the annual rate and number of births to teenage women. Over 445,000 babies were born to young women 15 to 19 years old in 2007 alone (Hamilton, Martin, & Ventura, 2009).
It has been commonly assumed that teenage pregnancies are unintended and are simply the result of inconsistent or no contraception use. Although adolescent pregnancies are largely unintended or unplanned, several studies have found that a significant proportion (24%-65%) of adolescents feel some desire or ambivalence about becoming pregnant, suggesting the importance of these feelings in understanding sexual risk behavior (Cowley & Farley, 2001; Davies et al., 2003; Finer, 2010; Heavey, Moysich, Hyland, Druschel, & Sill, 2008a; Hellerstedt et al., 2001; Kelly, Sheeder, & Stevens-Simon, 2004).
Much of the research examining the relationship between pregnancy desire or ambivalence and sexual risk or pregnancy incidence has relied on cross-sectional designs (Davies et al., 2004;Heavey et al., 2008a; Kelly et al., 2004; Kinsella, Crane, Ogden, & Stevens-Simon, 2007; Raine, Harper, Paukku, & Darney, 2002). These study designs, however, can be subject to several types of biases. For instance, cross-sectional designs cannot indicate temporality. Although it may be assumed that desire precedes pregnancy, young women who receive positive pregnancy test results may increase their desire from natural tendencies to feel positive about a new pregnancy (Bankole & Westoff, 1998; Heavey et al., 2008a). Likewise, adolescents who use contraception inconsistently may modify their desire according to their perceived risk of pregnancy (Kelly et al., 2004). This measurement of desire may also be subject to social desirability bias, thereby encouraging young women to report more positive feelings towards pregnancy than they actually experience if they truly suspect they may be pregnant.
Few studies have examined the ability of pregnancy desire or ambivalence to predict sexual risk or pregnancy incidence over time. Among a sample of African American adolescents, Davies et al. found pregnancy desire at baseline to be a significant predictor of inconsistent contraception use at 6 months (Davies et al., 2006). Using a similar sample, Zabin and colleagues demonstrated that pregnancy desire predicted contraception use but not pregnancy over a two year period (Zabin, Astone, & Emerson, 1993). Crosby et al. also did not find a significant relationship between participant pregnancy desire and pregnancy outcome among African-American adolescents (Crosby, DiClemente, Wingood, Davies, & Harrington, 2002). This body of literature, therefore, provides somewhat mixed results and offers limited generalizability.
Furthermore, limited work has examined the psychosocial and demographic characteristics of pregnancy desire or ambivalence in adolescents (Cowley & Farley, 2001; Davies et al., 2003; Heavey et al., 2008a; Hellerstedt et al., 2001). Given the growing salience of pregnancy desire in predicting risk, understanding its correlates is particularly important to effectively target intervention efforts. Therefore, data from a larger cohort study designed to explore sexual behavior over time were used to accomplish two distinct aims: (1) to document how psychosocial and demographic characteristics relate to pregnancy desire and (2) to determine whether pregnancy desire predicts incident pregnancy among adolescent females.
Participants were adolescent females recruited through 10 hospital clinics, community health care centers, and high school-based clinics in New Haven, Bridgeport, and Hartford, Connecticut. These three cities have the highest rates of teenage pregnancy, HIV/AIDS, and STDs in the state (Connecticut Department of Public Health, 2006). Between June 1998 and March 2000, adolescent women were approached in the clinic waiting rooms by study staff, referred by an enrolled participant or health care provider, or responded to study promotional material in the clinics. Both pregnant and non-pregnant adolescents were recruited and screened for eligibility in the same manner. Eligibility criteria included being female between the ages of 14 and 19, ever having had sexual intercourse, being HIV negative, and being nulliparous. Out of the 534 eligible adolescents, 411 agreed to participate and were interviewed (response rate 77%). Adolescents who agreed to participate were more likely to be African-American (p<0.05) and younger (p<0.05) than those who refused participation. Participation did not vary by any other measured characteristic. Based on the aims of our study, we limited our sample to the non-pregnant cohort (n=208).
Non-pregnant participants completed a baseline interview shortly after enrollment. Each participant was then interviewed every 6 months thereafter for 18 months. At each assessment, a member of the study team administered structured, face-to-face interviews lasting approximately 90 minutes with each participant. The survey instrument was pilot-tested for age- and cultural-appropriateness with a similar population of adolescents. It ascertained demographic information, sexual histories, current sexual risk behaviors, psychological measures, and partnership characteristics. This study was approved by the Yale University Human Investigation Committee and by Institutional Review Boards at study clinics. Participation was voluntary and confidential and did not affect the provision of health care in any way. All participants provided written informed consent and were paid $25 for each interview.
We derived measures for this study from previous research examining pregnancy desire and sexual risk behavior among adolescents (Brown et al., 2006; Crosby et al., 2002; Davies et al., 2006; DiClemente et al., 2001).
We used independent t-tests and chi-square tests to examine demographic and psychosocial correlates of pregnancy desire for continuous and categorical variables, respectively. We then entered variables with p-values less than 0.2 into a multivariate logistic regression model to examine the association of these factors entered simultaneously with pregnancy desire. The p-value of 0.2 has been suggested for use in observational studies with small sample sizes (Jewell, 2004). A final multivariate regression model was constructed using backwards elimination guided by the likelihood ratio test (inclusion p≤0.05; exclusion p≤0.10).
We used Cox regression analysis to examine the ability of pregnancy desire to predict pregnancy incidence. This approach allowed us to include participants who did not complete the entire 18 month follow-up. First we used independent t-tests and chi-square tests to compare demographic and psychosocial characteristics between those with and without an incident pregnancy during follow-up. Risk factors with significance levels of less than 0.2 were entered simultaneously in a multivariate Cox regression model and backwards elimination using the likelihood ratio test was used to derive the final covariates. After the predictors of incident pregnancy had been determined, we entered pregnancy desire into the model to examine its impact on pregnancy incidence after accounting for these other associated factors. We also explored two-way statistical interactions between pregnancy desire and the other predictor variables.
A very small proportion of participants had missing data for some predictor variables (all <4%). Missing values were imputed based on means for continuous variables and modes for categorical variables. All analyses were conducted using SPSS 16.0 (SPSS Inc., Chicago, 2007).
Participant demographic and psychosocial characteristics at baseline are shown in Table 1. The mean age of participants was 17.2 years (SD=1.46). Approximately 40% were non-Hispanic black, 33% were Hispanic, and 16% non-Hispanic white. Most were still in school and not working at baseline. Approximately one-third of participants were involved in relationships lasting longer than one year. Seventy-six percent of participants disagreed with the statement, “You would like to get pregnant next year;” almost 12% neither agreed nor disagreed; and 12% agreed to the statement.
The bivariate analysis (Table 1) indicated that those young women who desired pregnancy were significantly older and less likely to be in school than those who did not desire pregnancy within the next year. These adolescents also had a greater number of lifetime sexual partners and experienced significantly more perceived stressed than adolescents who did not desire pregnancy in the next year (all p<0.05).
In the multivariate logistic regression model, three participant characteristics were significantly related to pregnancy desire: participant age, age at first sex and relationship duration (Table 2). Each year increase in age was associated with almost two-fold greater odds of desiring pregnancy. Additionally, older ages at first sex were associated with reduced odds of desiring pregnancy. And thirdly, adolescents who were in a relationship of less than six months at baseline had odds of desiring pregnancy 5.35 times greater those who were not in a relationship. Longer relationships (6 months or more) were not statistically associated with pregnancy desire in the multivariate model. Perceived stress was marginally significant (p=0.054) with pregnancy desire; every unit increase in perceived stress almost doubled the odds of desiring pregnancy.
Table 3 describes participant follow-up and number of reported pregnancies over the follow-up period. Seventy-eight percent of participants completed the first follow-up visit. Of these participants, almost 90% either reported a pregnancy or completed both remaining follow-up assessments. Of the 162 adolescents who completed at least one follow-up visit, 38 (23.5%) became pregnant at least once during the study.
Table 4 displays the bivariate analysis used to examine the associations between participant characteristics and pregnancy incidence. Variables that were significant [p<0.2; school status, employment status, relationship duration (follow-up), ever been pregnant, marijuana use, depression, anxiety, and negative coping ability] were used to develop the final model. School status, employment status, and ever been pregnant were retained in the mulitvariate Cox model for their significant effects on pregnancy incidence. Pregnancy desire was then added to this model. Desire for pregnancy doubled the risk of becoming pregnant over the 18 month follow-up (RR=2.00, p=0.053) after accounting for potential confounders (Table 5). Figure 1 presents the multivariate Cox regression survival curves stratified by pregnancy desire. Investigating further, we found a significant statistical interaction between pregnancy desire and school status (RR=4.84, p=0.026). The risk of pregnancy was almost 5 times higher among those who were not in school and expressing some desire for pregnancy compared to those in school and expressing no desire for pregnancy.
Approximately 24% of participants reported feelings of desire or ambivalence towards pregnancy within the next year. Our analysis demonstrates that some desire for pregnancy in this sample of adolescents is significantly associated with pregnancy incidence over the 18 month follow-up period. This study, therefore, is one of the first to suggest that expressions of desire for pregnancy indicate higher risk for subsequent pregnancy among a racially and ethnically diverse sample of young women.
Adolescents who expressed pregnancy desire or ambivalence were likely to be older, have been younger at first sex, in shorter-term relationships and have greater perceived stress than those who expressed no desire for pregnancy. The relationship between pregnancy desire and relationship length is particularly interesting. Pregnancy desire may be creating the drive for new partners or a new relationship may be contributing to pregnancy desire. Ideas surrounding pregnancy in these shorter term relationships may be more romanticized and influenced by the strong intensity of feelings associated with new love (Aron, Norman, Aron, & McKenna, 2000). These shorter adolescent relationships, however, tend to be less stable by nature, thereby introducing a higher chance of family instability if a pregnancy were to occur and be carried to term. This instability has been linked with less positive behavioral and cognitive child development. Clinicians and family planning practitioners may consider strategies for intervening among this particular adolescent subpopulation, specifically upon presentation for contraceptive regimens.
Additionally, higher levels of stress among those desiring pregnancy also warrant additional attention. These results suggest that adolescents may be using thoughts about pregnancy as a coping mechanism to escape stressful environments. Other literature corroborates this notion, proposing that pregnancy and childbirth may be a strategy used among certain subpopulations to find value and purpose in otherwise depressing or tumultuous environments (Geronimus, 1991; Stevens-Simon & Lowry, 1995). Again, improving coping strategies among adolescent females may be a beneficial approach for reducing sexual risk behavior.
This analysis confirms the suspected link between pregnancy desire and pregnancy over time and reinforces the importance of evaluating this construct among sexually active adolescents. Our multivariate model also suggests that being in school and being employed are important factors that may protect young women from pregnancy. Being in school or employed may represent more responsible behavior or may simply result in less time and opportunity for risk behavior to occur. Being employed may also suggest a greater financial need. The meaning of these variables, however, may differ according to the participant's age. For instance, attending school before graduating high school is reasonably different from attending school after high school. Furthermore, our results highlight the synergistic risk for pregnancy among young females who express some desire for pregnancy and who are not in school. Our study therefore helps interventionists target higher risk subpopulations and provides constructive strategies for improving pregnancy prevention efforts.
After establishing this relationship between pregnancy desire and pregnancy incidence, we tested whether contraception use mediated this relationship using three criteria established by Baron and Kenny (Baron & Kenny, 1986). Contraception use was measured on a scale from 0 to 4 to describe the number of occasions participants used contraception inconsistently. First, contraception use was significantly associated with pregnancy desire (p<0.001); second, contraception use was significantly associated with pregnancy incidence (p<0.001); and third, when contraception use was added to our final multivariate model predicting incident pregnancy, pregnancy desire was no longer significant (p=0.485) and contraception use was highly significant (p<0.001). These results suggest that the relationship between pregnancy desire and pregnancy is mediated by contraception use, and thus highlights risk not only for pregnancy, but also for sexually transmitted infections, including HIV.
This study, however, is not without limitations. First, our sample includes sexually active adolescents and therefore may not generalize to a non-sexually active adolescent population. This population, however, was most appropriate for our study because they represent those at highest risk for pregnancy. Second, our pregnancy desire variable is limited as it uses a single item to assess this construct and therefore may have issues of validity and reliability. Our relationship duration variable for follow-up was also limited as it is pertains only to the relationship reported at baseline. We do contend, however, that this variable contains valuable information for the half of participants who did not initiate new partnerships during the follow-up period. Other attempts at measuring relationship dynamics (e.g., number of partners during follow-up, new partnerships during follow-up) did not meaningfully alter results. Additionally, all measures are self-reported and are subject to bias and misclassification. We were also unable to use biological markers or a chart review to confirm pregnancy status. Furthermore, the measures assessed at baseline – including pregnancy desire – assume stability over time, which may not have been the case. Future research designs should account for changes in these potentially dynamic variables. We also lack information on partner characteristics, which may play an important role in shaping feelings towards pregnancy and pregnancy risk among adolescent females (Cowley & Farley, 2001; Crosby et al., 2002; Heavey, Moysich, Hyland, Druschel, & Sill, 2008b; Hellerstedt et al., 2001). And lastly, the rate of any follow-up among participants (77.9%) may have been a limitation, and we cannot rule out the possibility of selection bias.
Our study, however, has a number of strengths including its prospective design with a long follow-up period. Additionally, several potential confounders were considered, such as behavioral risk factors and psychological characteristics, which have previously been associated with sexual risk behavior (Brown et al., 2006; Crosby et al., 2002; DiClemente et al., 2001). Furthermore, this racially and ethnically diverse sample improves the generalizability of this research beyond the African American samples used in previous research. Lastly, these results provide strong evidence for the importance of assessing pregnancy desire among adolescents.
It is crucial for future work to continue exploring the characteristics and complexity of pregnancy desire to determine potential strategies for reducing or delaying such sentiment. For instance, research should investigate the importance of the partner's desire for pregnancy and the potential interaction of desire in the dyad. Research should also explore the link between pregnancy desire and expectations for the future, which has been associated with pregnancy desire and sexual risk behavior (Hellerstedt et al., 2001; Kotchick, Shaffer, Forehand, & Miller, 2001; Singer et al., 2006). And lastly, intervention efforts should target young women who desire or are ambivalent towards pregnancy, such as those involved in new romantic relationships, and promote positive coping skills to help reduce rates of adolescent pregnancy in the US.
The project described was supported by Award Number T32MH020031 from the National Institute of Mental Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIMH or the NIH. Heather Sipsma had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.