INTRODUCTION
Research has shown that a resolution of an unintended pregnancy by induced abortion compared to delivery is associated with higher rates of alcohol consumption (1-9) and illegal drug use (7-16). Women with a history of abortion are also more likely to use drugs and alcohol during later wanted pregnancies (17-21), which places their fetuses at increased risk.
The relationship between abortion and substance abuse has received little notice or discussion among medical practitioners, however, because it has been widely assumed that the association is fully explained by common risk factors for both unintended pregnancy and substance abuse. Further, some researchers have suggested that negative emotional reactions following an abortion are likely to be negligible or at least equivalent to those stemming from carrying an unintended pregnancy to term (22,23). Unfortunately, this hypothesis has not yet been tested.
The purpose of this study was to explore any differences in subsequent substance use between women who carry an unintended pregnancy to term and those who have an induced abortion, while controlling for potential confounding factors including prepregnancy psychological state. To provide further information for the interpretation of the results, a control group of women with no reported pregnancies was also used.
METHOD
This analysis is based on data collected from 1979 through 1988 from the National Longitudinal Survey of Youth (NLSY), a survey conducted by the Center for Human Resource Research at Ohio State University. The survey began in 1979 and involved the follow-up of 12,686 youths aged 14-21 at the time of the first interview. The cohort used in this study was a carefully selected cross-sectional sample of noninstitutionalized civilian citizens of the United States, born between January 1, 1957, and December 31, 1964, with a supplementary oversample of blacks, Hispanics, and poor whites.
Women in the NLSY sample were queried regarding their childbirth history every year beginning in 1979. Starting in 1984, women were asked about their abortion history every two years by way of a confidential abortion card. All variables regarding abortion and childbirth outcomes were used to construct a reproductive history profile for each woman. Variables regarding intentionality of each pregnancy were constructed from the data across all years by Joyce, Kaestner, and Korenman (24) and were provided for our use in these analyses.
The 1988 interview of the NLSY included 11 items that tap into alcohol abuse symptoms (see Table 1). The dichotomously scored items required respondents to indicate whether or not the behaviors described reflected their own personal situation. The alcohol items were primarily derived from the National Health Interview Survey, conducted by the U.S. Census Bureau. Four additional substance use items from the 1988 data set were used as outcome measures. One item assessed the number of days the subject drank over the course of the last month and the second asked how many drinks the respondent consumed on days when she drank alcohol. The other two questions asked if the respondent had used marijuana or cocaine over the course of the last month.
In order to control for the effects that social support may have on substance abuse, year 1988 variables were extracted related to marital history and frequency of religious service attendance. Women for this study were categorized according to whether they were 1) in their first or second marriage, or if they were 2) never-married or had not remarried after their first or second divorce. Women who had been married three times were excluded from analyses, since data was not available to determine if they were still married (n = 7) in 1988. Finally, variables pertaining to age, total family income, total years of formal education, and race were extracted for the year 1988 to enable control for these potentially confounding variables.
The Rosenberg Self-Esteem Scale (25) was included in the NLSY 1980 interview. These scores were incorporated into this study as a means for controlling for psychological state prior to the first pregnancy. Scores range from 10 to 40, with higher scores indicating greater self-esteem. High self-esteem is an accepted component of positive mental health, with low self-esteem often correlating with mental health problems including anxiety and depression (26). Fortunately, self-esteem is not the only measure available to researchers for the assessment of psychological well-being in the NLSY data set. In 1979, respondents were presented with an abbreviated version of the Rotter Internal-External Locus of Control Scale. The suitability of the Rotter as a marker for prior psychological history is confirmed by previous investigations demonstrating a relation between external locus of control and greater depression (27-31). Moreover, a theoretical link has been established between externality and the etiology of depression (32). Scores on the abbreviated 4-item scale range from 4 to 16, with higher scores indicative of a more externally controlled individual, and lower scores suggestive of a more internally controlled individual.
To confine our analyses to women with unwanted first pregnancies, we excluded all subjects whose first pregnancy was intended or occurred prior to 1981. The latter condition was imposed so we could control for self-esteem and locus of control. The remaining women were divided into three groups: 1) those whose first pregnancy was unintended (mistimed or unwanted), delivered a child between 1981 and 1988 (n = 535), and did not report an abortion; 2) those whose first pregnancy occurred between 1981 and 1988 and was terminated by induced abortion (n = 213); and 3) women who had no pregnancies prior to 1988 (n = 1144).
Among women who had abortions, the average year of their first abortion was 1984 (SD = 2.3). Among women who had an unintended birth, the average year of their first birth was 1983 (SD = 2.6). Table 2 shows descriptive statistics collected in 1988 for age, marital status, income, and race. Only three classifications for race are provided in NLSY: black, Hispanic, and not black and not Hispanic. The latter is typically reported as white, but in fact includes Asians and other minorities.
Unfortunately not all of the participants in this subsample of the NLSY were surveyed pertaining to the alcohol variables. However, no significant differences were observed relative to the control variables between those who were administered the alcohol-related items and those who were not.
RESULTS
Table 1 provides the percentages from each group who responded affirmatively to each of the alcohol abuse items. For a majority of the items, the unadjusted scores were highest among the abortion group.
The pregnancy variable was found to be significantly associated with age [r(1892) = .05, p < .05], self-esteem [r(1892) = .12, p < .001], locus of control [r(1892) = - .11,p < .001], income [r(1892) = .15,p < .001], marital status [r(1892) = - .27, p < .001], years of education [r(1892) = .32, p < .001], and race [r(1892 = .13, p < .001]. Therefore, these seven variables were used as covariates in all of the primary analyses. A significant association was not detected between the pregnancy history variable and regularity of religious service attendance [r(1892) = -.02, p = .339], measured as a possible indicator of social support.
Two analyses of covariance (ANCOVAs) using the seven control variables listed above were performed with the three pregnancy history groups as the independent variable in each case. The results of these analyses are presented in Table 3. As indicated by the information provided in the table, the overall analysis pertaining to the number of days the respondent reported drinking over the previous 30 days was significant. The univariate F-tests revealed a higher frequency of drinking for the abortion group compared to the unintended pregnancy group and for the no pregnancy group compared to the unintended delivery group. The overall result of the second ANCOVA pertaining to the number of drinks consumed on days when the respondent drank over the past 30 days was not significant. However, examination of the univariate tests revealed that the difference between the abortion group and the unintended delivery group approached significance, with the abortion group indicating consumption of more alcohol.
