Dr Rachel Scott @rachelhvscott
Underreporting
Abortions are known to be underreported in surveys. This has important consequences for demographic research. If reporting of abortions in surveys is incomplete, there are knock on effects on other pregnancy data collected in surveys. Conceptions comprise births, miscarriages and abortions; Incomplete data on abortions means incomplete data on conceptions. This has implications for the measurement of unplanned pregnancies and pregnancy outcomes. If reporting patterns differ between different groups, then this will lead to biases in analyses of characteristics associated with abortion and other pregnancy outcomes. It is valuable, therefore, to consider whether reporting of abortions can be improved.
What We Did
We analysed data from three nationally representative surveys, all three of which collect data on sexual and reproductive health and behaviours, including abortion. The National Surveys of Sexual Attitudes and Lifestyles (Natsal) have been conducted in Britain roughly every decade since 1990. We analysed data from Natsal-2, conducted in 2000, and Natsal-3, conducted in 2010. FECOND is a French survey on fertility, contraception and sexual dysfunction that was conducted in 2010 and follows many surveys that have taken place in France since the nineteen seventies. We measured the extent of underreporting of abortions in these three surveys by calculating the yearly abortion rate from the abortions reported by women in the survey, and comparing this to the rate obtained from national statistics, which are considered to be complete in both Britain and France.
One of the aims of this research was to examine whether differences in data collection methods affect the extent to which abortions are underreported. Data on abortion was collected in different ways in the two surveys. Natsal-2 asked a direct question on abortion; women were asked if they had ever had an abortion, and if they had they were asked how old they were at the first one and at the last one. In Natsal-3 and FECOND, information on abortion was collected through a pregnancy history module; women were asked if they had ever been pregnant, and if they had, they were asked about the date of the pregnancy and how it ended, for each pregnancy in turn.
Asking the right question
Our findings suggest that a direct question might elicit greater reporting of abortion than a pregnancy history module. There are several reasons why this might be the case.
- A pregnancy history module can be quite burdensome to complete, with many follow up questions for each pregnancy that a person reports. People may therefore omit some of their pregnancies to avoid being asked so many questions.
- Women for whom abortion is a sensitive event may be less likely to disclose a pregnancy ending in abortion in a pregnancy history module than in response to a direct question.
- For others, it may be that abortion is a less salient event that they would not ‘count’ in their reproductive history, and so not something that they would think to bring up if not asked directly.
It is important to note the limitations to this research. Firstly, the uncertainty around the rates we estimated in the surveys is high. Secondly, even if we are seeing a ‘real’ change in reporting between the surveys, there were other differences between the surveys besides question wording that likely also played a role in that change. Natsal-2 and Natsal-3 surveyed the same general population, but ten years apart. The later survey had a lower response rate, which might affect reporting of abortions. The stigma around abortion might also have increased during that time, leading to lower reporting. Both Natsal surveys asked questions about abortion in a self-administered section of the interview, whereas FECOND was conducted over the phone, which respondents might consider a less anonymous setting.
A pregnancy history is extremely useful for collecting data on respondents’ reproductive biographies, but it may not be the best way to collect accurate data on abortions. Supplementing the pregnancy history with some abortion-specific direct questions might lead to higher overall reporting of abortions. Whilst these questions can obtain less detailed information than a pregnancy history module, data on whether the respondent has ever had an abortion, how many, and at what age(s), enables estimation of lifetime prevalence and incidence in the last year, age patterns, and associations with sociodemographic characteristics and behaviours. Survey designers might consider the advantages of collecting a smaller amount of data, which is more complete, over more detailed information that is subject to substantial underreporting.
In the meantime, it is vital that all researchers using survey data on pregnancy outcomes consider the implications of underreporting of abortions for their findings.
Comparing reporting of abortions in three nationally representative surveys: methodological and contextual influences was published in BMJ Sexual and Reproductive Health.