Patterns of Pregnancy Termination According to Contraceptive Use

David Antonio Sánchez-Páez, University of Salamanca
José Antonio Ortega, University of Salamanca

Background: The outcome of pregnancy is either a live birth or a pregnancy termination. The probability of a pregnancy ending in a live birth is a fundamental element in demographic models of reproduction, such as the proximate determinants framework, implicitly assuming the likelihood of pregnancy termination is independent of contraceptive use. Contraception signals if a pregnancy is unintended at the time; hence, failures in its use lead to unwanted pregnancies and induced abortion.

Methods: We use individual-level calendar data of 1,316,885 women aged 15-49 from 94 DHS for 49 low- and middle-income countries to estimate non-parametric probabilities of pregnancy termination by age, marital status, and contraceptive use before pregnancy at the survey, region, and sample level. Also, we provide estimates for the probability of pregnancy loss and induced abortion.

Results: On average, 11.7% of pregnancies end before birth. The odds of termination are higher for unmarried women by 1.2 percentage points. By contraception, users are 11.3 percentage points more likely to have terminations. The probability increases with age from 9.9% for adolescents to 27.4% at ages 45-49. The likelihood of biological loss follows a U-shaped pattern by age. The odds of induced abortion are 9.8% for users and 2.3% for non-users.

Conclusions: Patterns differ by regions and are consistent with demographic risk factors as age and marital status. There is an association between contraceptive use at the time of pregnancy and terminations. Contraceptive failure leads to unintended pregnancies and increases the probability of pregnancy termination, thus, induced abortion. Unmarried women, adolescents, older women, and contraceptive users are the groups of higher risk. Users and non-users should be treated separately from the perspective of demographic modeling of fertility.


1. Background

There are social, biological and health determinants of the probability of a pregnancy loss, including miscarriages and stillbirths. Nevertheless, the likelihood of a pregnancy termination mostly depends on levels of induced abortion, with an incidence between 12% and 39% of total pregnancies. Also, there are clinical and observational studies focusing on the determinants and consequences of pregnancy loss and induced abortion. However, literature fails to explain the patterns of pregnancy termination.

Country studies on abortion show the role of demographic factors and comparative research. The former identifies marital status and age as main factors to have an abortion. The latter finds unplanned pregnancies is the most commonly reported reason for having an abortion. A 40% of total pregnancies in 2012 were unintended, mostly occurring in developing countries. Contraceptive use reduces the likelihood of an unwanted pregnancy, although contraceptive failure is a possibility to face. The incidence of contraceptive failure depends on using traditional methods, discontinuation, or misuse, and increases the odds of abortion, especially in countries where it is legal.

A pregnancy ending in a live birth is fundamental in demographic models of reproduction, such as the proximate determinants framework, even more, when induced abortion is already modifying fertility levels. Our purpose is to explore patterns of pregnancy termination according to age, marital status, and contraceptive use at the time of pregnancy.

2. Methods

We use data from 94 DHS surveys for 49 low- and middle-income countries in Africa, Asia, Europe, and Latin America. We select those including single women and contraceptive calendar data.

We analyze all pregnancies with an outcome reported in the 36 months before the survey. For each pregnancy, we record the outcome and classify women by five-year age-group, marital status, and contraceptive use at the time of pregnancy. Then, we estimate conditional probabilities of pregnancy termination ((T)) as (T=(1- frac{b}{p})times 100) for every group, where (b) is the weighted number of births, and (p) is the weighted number of pregnancies. For surveys with union history, we use marital status at the time of pregnancy; otherwise, we impute based on the date of the first and last union.

In the complete manuscript, we will provide confidence intervals, probabilities for users and non-users of contraception at the time of pregnancy, and the estimates of biological termination and abortion.

3. Results

On average, a pregnancy termination occurs for one of every ten pregnancies, but there are significant differences on the country context and contraceptive use. Average probability of pregnancy termination for non-users is 10.5% compared to 21.8% for users. Since there is no medical evidence of a causal effect of contraceptive use on miscarriage, it indicates a higher likelihood of abortion when the pregnancy is a result of contraceptive failure.

Figure 1 provides average probabilities of termination for users and non-users. There are regional differences, especially in those known to have high levels of induced abortion, like Central Asia and Europe. In countries with a lower incidence of pregnancy termination, like Africa and Latin America, the probability is higher for users than for non-users of contraception. Some of the DHS regional groups, such as North Africa/West Asia/Europe or South & Southeast Asia are very heterogeneous.

Figure 2 provides general patterns of pregnancy termination by using separate boxplots for each combination of region, age-group, marital status and contraceptive use. Many of the categories, particularly older single women, have a small number of cases, thus subject to high sample variation. Findings for non-users in countries with low incidence of pregnancy termination are consistent with demographic patterns of fetal loss, with a typical U-shape by age. The probabilities of termination tend to be lower for married women, consistent with previous studies. Countries having high levels of induced abortion tend to have increased risk of termination by age, consistent with a parity-specific abortion.

4. Conclusions

We find a substantial share of pregnancies not ending in a live birth. The patterns differ widely according to a region and are consistent with demographic risk factors as age and marital status. In most countries, there is an association between contraceptive use at the time of pregnancy and pregnancy termination confirming the likelihood of a contraceptive failure ending in abortion. More efficient contraceptive use would lead to less contraceptive failure, lower unintended pregnancies, and fewer abortions. Users and non-users should be treated separately from the perspective of demographic modeling of fertility.

Presented in Session 1235: Posters