Drey, Eleanor A. MD, EdM1,2; Foster, Diana G. PhD1,2; Jackson, Rebecca A. MD1,2,3; Lee, Susan J. JD4; Cardenas, Lilia H.5; Darney, Philip D. MD, MSc1,2
Second-trimester abortion has received considerable political attention recently with the enactment of state1 and federal legislation banning so-called “partial-birth” abortion.2 Second-trimester procedures account for approximately 12% of abortions performed in the United States.3 Procedures performed after 12 weeks of gestation may be costlier for women in many respects—financially,4 emotionally,5 and medically—posing greater risks of medical complications and mortality than abortions performed earlier.6 Second-trimester procedures are also more difficult to obtain because fewer providers offer them, limiting access.4 Understanding reasons for abortion delay may encourage the improvement of referral networks and facilitate the development of health education programs that reduce the need for second-trimester abortions. Such education may help women recognize unwanted pregnancy earlier, thus increasing a woman's options for pregnancy termination by rapid referrals to clinics and by enabling a woman to choose abortion by medication.
The literature on the causes of abortion delay in the United States is outdated; many articles are more than two decades old. Most of these studies primarily analyzed demographic factors correlated with overall delay and found that younger, unmarried women with less education and no previous pregnancies tended to seek abortions later.7–11 A few studies examined reasons for delay at key points in the process of obtaining an abortion, such as when pregnancy was suspected, when it was confirmed, and when the woman first attempted to schedule an abortion.11–13 These studies came to differing conclusions about which step in the process contributed most to delay, an important determination for prioritizing policy-based solutions. One study found that the longest delay occurred between the last menstrual period and the first suspicion of pregnancy,12 whereas two other studies found that the most substantial delay occurred between the first suspicion of pregnancy and seeing a doctor.11,13 The most recent comprehensive study of delay in the United States was based on data collected in 1987, but this study did not analyze delay by steps leading to abortion.14 In the last two decades, US studies of abortion delay have focused more narrowly on race15 and delayed abortions among teens.16,17
Using audio computer-assisted self-interviewing, we assessed a cohort of women obtaining abortions from 5 to 23 weeks of gestation. We sought to identify factors associated with abortion delay overall and during six time intervals, beginning with suspicion of pregnancy and ending with the abortion appointment. We evaluated a comprehensive list of demographic, reproductive, logistic, relationship, and emotional factors. We asked participants to prioritize which factors caused the most delay. We hypothesized that unknown date of last menstrual period and difficulty in getting an appropriate referral would be associated with abortion delay.
MATERIALS AND METHODS
We conducted a cross-sectional analysis to determine which demographic, medical, reproductive, and other factors were associated with abortion delay. We recruited consecutive English- and Spanish-speaking patients presenting for abortion from 5 to 23 weeks of gestation at the Women's Options Center, an urban, hospital-based abortion clinic from September, 2001, through March, 2002. The Women's Options Center serves a local community of primarily Latina and African-American women and accepts referrals from throughout Northern California. Women who are referred are typically in the second trimester, are low income, and/or have medical complications. Women were excluded from the study if they were obtaining an abortion because of fetal anomalies or demise or if they were unable to learn to use audio computer-assisted self-interviewing. The study was approved by the institutional review board of the University of California, San Francisco. To keep the numbers of first- and second-trimester patients roughly equivalent throughout the enrollment period, if either group outnumbered the other by more than ten, recruitment for the larger group was suspended until numbers equalized. Because gestational duration was determined after enrollment, four women were found to be beyond the clinic's 23-week limit, could not terminate their pregnancies, and were excluded from analysis. Four were excluded because the gestational duration was not available.
Subjects were enrolled before obtaining an abortion but after receiving counseling from trained pregnancy advisors and signing a consent. We used audio computer-assisted self-interviewing (Sensus Q&A 2.0; Sawtooth Technologies, Evanston, IL, 1998) to administer the questionnaire to improve the accuracy of responses for sensitive topics.18 The questionnaire was developed in consultation with psychologists expert in instrument design and included characteristics identified in previous studies to be associated with abortion delay, including demographic, reproductive, socioeconomic, and insurance factors. In addition, we added questions about the timing of menses, pregnancy symptoms, relationship factors, social support, attitudes about abortion, and number of prior providers consulted regarding this pregnancy. We also included closed- and open-ended questions about reasons for delay. Finally, we asked participants to identify seven dates leading to the abortion appointment: 1) first day of last menstrual period, 2) suspicion of pregnancy, 3) positive pregnancy test, 4) decision to abort, 5) first telephone call to an abortion clinic, 6) first call to our clinic, and 7) abortion date. These timing questions were completed with the help of a research assistant using a calendar. All other questions were completed in a private room with the subjects encouraged to request help as needed. We assumed the first missed menstrual period would have occurred 28 days after the last menstrual period.
The instrument was pretested for clarity with 10 English-speaking and 10 Spanish-speaking patients. Subjects were trained in audio computer-assisted self-interviewing and computer use as necessary with sample questions. The median time to complete the survey was 18 minutes. Research assistants abstracted additional demographic and medical data from each subject's medical record. Subjects received $15 for their participation.
The primary outcome variable was gestational duration at the time of abortion as determined by ultrasonography, which was dichotomized to second (≥ 13 weeks) versus first trimester for the logistic regression. Secondary outcomes included elapsed days for each of the six intervals between the consecutive steps defined above and proportion of women who were in the second trimester at the end of that step. We used multivariable logistic regression to examine factors associated with delay in the overall time to abortion. The model was constructed using a set of fixed covariates describing demographic characteristics: race/ethnicity, foreign-born status, marital status, age, education, income, previous abortions, and previous births. In addition, we included variables expected to be associated with abortion delay based on our clinical practice, available literature, and those found to be significant at a .05 level in bivariate analysis. All analyses were conducted using STATA 8.2 (College Station, TX). A Cox multivariable hazard analysis was also performed with the primary outcome of time to abortion (results not shown). Similar results were obtained; therefore, for simplicity, we present the results of the multivariable logistic regression.
Sample size calculations originally were based on guidelines for the sample size needed for multiple linear regression, which suggest enrolling twenty subjects per independent variable.19 Based on previous studies, an estimated 15 variables were expected to require analysis. Thus at least 300 subjects were deemed necessary. In addition, because of anticipated colinearity between many of the variables, we estimated we would need an additional 30% for a total of 390.
According to study design, subjects were divided evenly between the first and second trimesters. Subjects in the second trimester were more likely to have been referred from other clinics and to have had difficulty finding an abortion provider (Table 1). They were also more likely to be less educated, to live farther from the clinic, and to have had difficulty arranging transportation. Although both first- and second-trimester patients predominantly relied on state funding (Medi-Cal) for their abortions, second-trimester patients had more difficulty obtaining Medi-Cal. Second-trimester subjects were also more likely to have had a previous second-trimester abortion, to be unsure of their last menstrual period, to experience fewer pregnancy symptoms, and to have used drugs and/or alcohol. More than 80% of first-trimester subjects were local residents and consistent with the demographics of the clinic's neighborhood, and they were more likely to have a household income of less than $20,000 and to be foreign-born and Latina. More than two thirds of all women having abortions were using contraception at the time of conception (Table 1).
Women having second-trimester abortions presented an average of 70 days (10 weeks) later than women having first-trimester abortions (Table 2). Compared with women having first-trimester abortions, among women having a second-trimester abortion, each of the six steps leading to abortion was significantly longer (P < .001 at all steps). The largest delay occurred in the first step–delay in suspecting pregnancy after missing a period was responsible for nearly a third (22 days) of the total difference between the two groups. Another 19% of the difference was due to difficulty locating an abortion provider. More than half (58%) the women were already in the second trimester by the time they obtained a pregnancy test.
All subjects were asked to identify from a list of 21 factors which ones, if any, had caused delay in obtaining an abortion and which of these had caused the most delay (Table 3). On average, first-trimester subjects reported that two factors had delayed their abortions whereas those in the second trimester reported more than 3 delaying factors (P < .001). One third (36%) of first-trimester subjects and 14% of second-trimester subjects reported that nothing had slowed them down (P < .001). Comparing broad categories of reasons for abortion delay, women with second-trimester abortions reported more logistical delays (63%), such as difficulty locating a provider, initially being referred elsewhere, or difficulty arranging transportation, compared with 30% in the first-trimester group (P < .001). An initial referral elsewhere was the single most frequently reported delay-causing factor by second-trimester patients (47%). Second-trimester patients were also more likely to be delayed because they did not suspect they were pregnant (34% versus 20%, P < .001). Emotional factors such as fear, depression, uncertainty, and a sense that abortion is “wrong” were cited by 51% in the second trimester and 42% in first trimester (P = .06). Similar portions of both groups attributed delay to interpersonal and financial factors overall, although more second-trimester patients reported difficulty obtaining Medi-Cal (7.3% versus 1.6%, P < .01).
When asked which single factor caused the greatest delay in getting an abortion, the 3 most common factors cited by both groups were the same: 1) initial referral elsewhere (17% in the second trimester versus 8% in the first trimester, P = .004); 2) difficulty deciding (10% versus 7%, P = .4), and 3) fear (8% versus 6%, P = .6). Overall, logistical factors caused the greatest delay for more second- than first-trimester patients (30% versus 19%, P = .02), as did factors associated with not suspecting pregnancy (16% versus 7%, P = .005).
Using multivariable logistic regression, we examined the covariates associated with second-trimester abortion after adjusting for demographic factors (Fig. 1). Factors independently associated with second-trimester abortion were prior second-trimester abortion (odds ratio [OR] 5.9), delay in obtaining Medi-Cal (OR 4.4), difficulty locating a provider (OR 4.1), initial referral elsewhere (OR 2.3), and unsure last menstrual period (OR 2.3). Factors associated with decreased likelihood of second-trimester abortion were presence of nausea/vomiting (OR 0.5), prior abortion (OR 0.4), and use of contraception (OR 0.4). Emotional and interpersonal factors were not associated with second-trimester abortion in the multivariable model.
Similar to other studies, women who have second-trimester abortions typically discover relatively late that they are pregnant.12,14 In our study, more than half (58%) the patients having second-trimester abortions had already delayed beyond the first trimester by the time they obtained a pregnancy test. Half of the 70-day difference between the average gestational durations in first- trimester and second-trimester abortions was due to later suspicion of pregnancy and administration of a pregnancy test. Earlier studies also found that the most significant delays occurred early in the process, with later suspicion of and testing for pregnancy.11,12,14 Second-trimester patients were less certain about their last menstrual periods and had fewer pregnancy symptoms, which if present, may have prompted these women to test sooner. In contrast to previous studies that found oral contraception to be associated with abortion delay,9,11 hormonal contraception was actually associated with less delay in our sample.
This initial delay preceded further delays once a woman decided to obtain an abortion. In fact, women obtaining second-trimester abortions took significantly longer to complete each step of the process. By the time an abortion provider was contacted for the first time, 71% of the second-trimester group was already in the second trimester. Thereafter, an additional 15 days elapsed before contacting our clinic. Delay in this last step was associated with being referred to other clinics before ours. Referrals were associated with a 4-fold increased risk of second-trimester abortion and were the most important delaying factor cited by second-trimester subjects. Delays due to referrals and other trouble locating a provider suggest a link between the scarcity of second-trimester providers and increased delay. Trouble with Medi-Cal was more often cited by second-trimester patients as a delaying factor and was associated with a 4-fold increased risk of second-trimester abortion. These financial barriers may have been more onerous for second-trimester patients, given the increased cost of second- versus first-trimester procedures.4 Fewer providers are available for women seeking second-trimester abortions, especially those with public funding. For example, in Northern California, ours is virtually the only clinic to accept patients with Medi-Cal for late second-trimester abortions.
One strength of our study is the extensive list of potential delaying factors that we examined. The audio computer-assisted self-interviewing design allowed us to collect and assess numerous factors that might have caused delay and then to re-present them to subjects to assess whether they experienced that factor as having caused delay. For example, subjects were initially asked if they experienced a factor, such as fear. If they said yes, they were later asked if fear was a delaying factor. In this way, we were better able to prompt women to obtain a more complete list of delaying factors. In addition, we asked open-ended questions about delay to ensure there were no major causes of delay that we had omitted from the list.
Due to our clinic's population, we were unable to draw solid conclusions about how delay may be associated with certain demographic factors. In our clinic, women who obtain second-trimester abortions are often referred from a larger geographic region and are therefore more heterogeneous with respect to ethnicity, education and other demographic features. Conversely, women who obtain first-trimester abortions live nearby and are disproportionately Latina or African-American, foreign born and low income (Table 1). Our study also may have been limited by biases associated with observational studies, such as volunteer bias and recall bias. Despite using audio computer-assisted self-interviewing, subjects may have difficulty disclosing sensitive information. Our study's findings necessarily reflect the circumstances affecting a very particular population (that of a referral clinic located in an ethnically diverse population). To increase external validity, the study ideally should be repeated with a larger, truly random population.
Legal and accessible second-trimester abortion services will remain necessary to provide safe medical care. Our study shows that many women seeking second-trimester abortions simply lacked pregnancy symptoms or were unaware of their last menstrual period and therefore took a long time to recognize and test for pregnancy. Legislative measures that may further reduce the availability of abortion services will likely increase delays by making it even more difficult to find a provider, with delay further increasing medical risks. Several public health measures might decrease the frequency of second-trimester abortion. In addition to improving their access to effective contraceptive methods, patients could be educated about the importance of maintaining menstrual records. Facilitating earlier pregnancy testing by providing women with low-cost home pregnancy tests before they suspect pregnancy may also decrease delays. Health care professionals should be encouraged to provide patients with information about options before they become pregnant, as well as facilitating timely referrals and decision-making after pregnancy has been diagnosed. Despite these measures, because of the individual nature of many of the reasons for delay, it is unlikely that public health measures alone can eliminate or substantially decrease the need for access to elective second-trimester abortion.
1. Neb Rev Stat §§28-326, 28-328 (2004).
2. Partial Birth Abortion Ban Act of 2003, 18 USCS §1531 (2005).
3. Strauss LT, Herndon J, Chang J. et al Abortion surveillance: United States, 2001. MMWR Surveill Summ 2004;53:1–32.
4. Henshaw SK, Finer LB. The accessibility of abortion services in the United States, 2001. Perspect Sex Reprod Health 2003;35:16–24.
5. Adler NE, David HP, Major BN. et al Psychological responses after abortion. Science 1990;248:41–4.
6. Lawson H, Frye A, Atrash H. et al Abortion mortality, United States, 1972 through 1987. Am J Obstet Gynecol 1994;171:1365–72.
7. Kerenyi TD, Glascock EL, Horowitz ML. Reasons for delayed abortion: results of four hundred interviews. Am J Obstet Gynecol 1973;117:299–311.
8. Mallory GB Jr, Rubenstein LZ, Drosness DL. et al Factors responsible for delay in obtaining interruption of pregnancy. Obstet Gynecol 1972;40:556–62.
9. Fielding W, Sachtleben M, Friedman L, Friedman E. Comparison of women seeking early and late abortion. Am J Obstet Gynecol 1978;131:304–10.
10. Bracken MB, Swigar ME. Factors associated with delay in seeking induced abortions. Am J Obstet Gynecol 1972;113:301–9.
11. Bracken MB, Kasl SV. Psychosocial correlates of delayed decisions to abort. Health Educ Monogr 1976;4:6–44.
12. Burr WA, Schulz KF. Delayed abortion in an area of easy accessibility. JAMA 1980;244:44–8.
13. Bracken MB, Kasl SV. First and repeat abortions: a study of decision-making and delay. J Biosoc Sci 1975;7:473–91.
14. Torres A, Forrest JD. Why do women have abortions? Fam Plann Perspect 1988;20:169–76.
15. Lynxwiler J, Wilson M. A case study of race differences among late abortion patients. Women Health 1994;21:43–56.
16. Slonim-Nevo V, Anson J, Sova J. Delayed abortion among teenagers: can a population at risk be identified? Health Care Women Int 1995;16:101–12.
17. Poliak J, Morgenthau JE. Adolescent aborters: factors associated with gestational age. NY State J Med 1982;82:176–9.
18. Turner CF, Ku L, Rogers SM. et al. Adolescent sexual behavior, drug use, and violence: increased reporting with computer survey technology. Science 1998;280:867–73.
19. Katz M. Multivariable Analysis: A Practical Guide for Physicians. Cambridge (MA): Cambridge University Press; 1999.
Figure. No caption available.