(Image by Anemone123 from Pixabay)
One of the big points of contention between pro-choice and pro-life people is whether abortion has a positive or negative impact on women’s mental health, overall. While no (reasonable) person would deny that there must surely be some women whose mental health was harmed by abortion, and there must be other women for whom abortion had no impact on their mental health, and there must be other women again for whom abortion improved their mental health, the question is about the total overall impact of abortion on women’s mental health. Does abortion generally improve, disimprove, or have no impact on women’s mental health?
It’s important to point out that this question is a separate question from whether abortion is right or wrong, or even good or bad. Even if we found that abortion always leads to poor mental health outcomes, that wouldn’t be a good enough argument to make abortion illegal. Similarly, even if we found that abortion always improves mental health, that wouldn’t be a good enough reason to make abortion legal. Nonetheless, the question of the impacts of abortion on mental health are an important consideration for pro-life and pro-choice people alike, as it informs our advocacy as well as the supports and services that women require, regardless of the abortion regime in place.
A question such as the impact of abortion on mental health is what’s known as an empirical question – it’s not a question that we can answer simply using logic or first principles. Instead, we have to use observations or experiences about the real world. Such observations can mean anything from a large-scale study to my personal experience. However, we have to be very careful when using personal experience, or even the experience of a large set of people, to make any kind of claim about people in general. This is because we have to make sure that the experiences we are considering can be generalised to the entire population. This is very often not the case.
Let’s say I study literacy levels amongst children and find that children who live in houses with lots of books have higher literacy levels. I could be tempted to conclude that giving families books will boost children’s literacy levels. However, there are good reasons to think that there are other differences between families with lots of books and families without lots of books, and these other reasons might drive the difference in literacy levels. Maybe children who have higher literacy levels in the first place are more likely to ask for and be given books as presents, or maybe parents with higher literacy levels are more likely to buy books for their children and are also more likely to give birth to children with an aptitude for reading.
In order to get around these issues, scientists try to design studies in clever ways to compare two groups of people who are alike in every respect except one: whatever it is we wish to study. If the two groups have different outcomes, we can reasonably conclude that the difference is driven by the one thing they did not have in common: namely the thing we wished to study. These studies are called randomised controlled trials (RCTs). So if we wanted to study whether having children’s books in a home improved children’s literacy, we would take a large group of families and randomly assign them into two groups (in other words, assign them based on no particular criterion, rather than assigning them based on the age of the children or the income of the parents, for example). Then we would give a set of children’s books to all the families in one group and no books to the families in the other group, and finally we would check all the children’s literacy levels, in the book group and the non-book group alike, over time. If the literacy levels across the two groups stayed the same, we would conclude the books had no effect. If the literacy levels diverged, we would consider that as evidence that the books did have an impact on literacy.
One of the problems that plagues public health research is that RCTs are often unethical. In the case of abortion, it is easy to see why. It would be completely unethical to randomly assign pregnant women to two groups, and perform abortions on one group and not the other. So if we want to measure the impacts of abortion, we can’t use an RCT. However, if instead we simply compare women who had abortions to women who didn’t, we are introducing all sorts of problems: there are lots of differences between women who get abortions and those who don’t. They differ by age, marital status, employment status, family, status, race, ethnicity and many other factors. How can we be sure that any differences between women who had abortions and women who didn’t are driven specifically by the abortion, rather than all those other differences? Many public health studies, including many studies on abortion, feature this problem, which means we simply cannot have any confidence in the results.
Scientists have various ways of exploiting what are called natural experiments to get around these issues. A natural experiment is where you study two groups based on some naturally-occurring criterion that is unlikely to drive differences between two groups. Say for example you want to see whether attending a particular elite school, which has an entrance exam, has an impact on educational outcomes. If you simply compare students who attended the school to those who did not, you run into all the problems we mentioned before. However, let’s say you had to score 95% on their entrance exam to attend the school. If we compare students who score 95% on the exam (and subsequently attended the school) to those who scored 94% (and therefore didn’t attend the school), then that’s a nice natural way of comparing two groups of people who are probably pretty similar in all the ways we care about, but who differ on whether they attended the school. In other words, we are not randomly assigning students into an “attended the school” and “didn’t attend the school” group, but we’re exploiting a natural outcome (test scores) that is almost like randomly assigning them to the school or not. There’s a good chance then that any differences in outcomes between these two groups is, indeed, driven by the fact that they attended the school.
There is one famous study that attempts to use a natural experiment to examine the mental health impacts of abortion. This study is called the Turnaway Study. Next week, we’ll look at what the main findings of the study were, and why it is often cited by people on both sides of the abortion debate.
This is a very clear account as far as it goes, but it underestimates how much can be achieved by fitting statistical models to the data which control for other important variables such as previous medical history, ethnicity, socioeconomic status etc. When randomised clinical trials are not possible, for reasons explained above, it is standard statistical practice to fit these models instead, and numerous studies have done so for the relationship between mental health and abortion. A very interesting study, which also gives a very useful summary of other studies in the area, is by Sullins in Medicina: https://pubmed.ncbi.nlm.nih.gov/31731786/. This study also has the very interesting finding that mental health outcomes are worst for women who abort wanted babies.