Unintended Pregnancy and Abortion Worldwide: Country-Level Estimates Explained
Unintended pregnancy and abortion are experiences shared by people worldwide, irrespective of region, country income level or legal status of abortion. Unintended Pregnancy and Abortion Worldwide is an established study by the Guttmacher Institute and the World Health Organization (WHO) that aims to provide stakeholders with evidence about and deeper insights into sexual and reproductive health disparities in low-, middle- and high-income countries.
In 2020, the researchers used a statistical methodology to generate the first-ever internationally comparable country-specific estimates of unintended pregnancy and abortion incidence. These estimates are essential for informing national decisions on investment in access to contraception and comprehensive abortion care, because country-level estimates may differ substantially from available data, such as regional averages. This explainer answers some anticipated questions from policymakers and advocates as to how to interpret and use these country-specific model-based estimates.
What are model-based estimates?
The country-specific numbers and rates of unintended pregnancies and of abortions, and the proportion of those pregnancies ending in abortion, that were generated by the study are referred to as model-based estimates because they are produced using a statistical model. We use the term model-based estimates to contrast the information generated by the model with the data and estimates—from country-specific official statistics, studies and surveys—that informed the model.
This model integrates and is informed by all available data on pregnancy intentions and abortion among women of reproductive age collected across 166 countries and territories. While modeling offers an additional approach to approximating data that are difficult to collect, it does not replace the need for robust in-country data collection systems and studies, which can provide more comprehensive information on sexual and reproductive health outcomes at the national level and for population groups and areas within a country.
What data did you use to generate these estimates?
The model-based estimates are generated using all available data from country-specific studies, surveys and official statistics, along with other information on contraceptive needs and use by union status. The model estimates all pregnancy outcomes jointly. Key input data used in the model are numbers of live births for all 195 countries, and data on the number of abortions and/or the proportions of births that were unintended (available for 166 countries). The model generates country-specific estimates for each of these outcomes, accounting for their relationships with one another. Country-specific estimates are published for 150 countries that had some data on abortion and/or the proportion of births that were unintended. Estimates were not released for 45 countries: Twenty-nine lacked data on both of these key input measures; the other 16 countries belong to a subregion (North Africa and the Middle East) that lacks reliable abortion data. Note that while country-specific estimates are published for 150 countries, the global, regional and subregional averages are based on and represent the populations of all 195 countries and territories.
How can estimates be produced for countries with unreliable or limited data?
Of the 166 countries from which we obtained data on abortion incidence or on the proportions of births from unintended pregnancies, somewhat fewer than half (75) had reliable abortion data for one or more years. Three-quarters of reproductive-age women lived in these 75 countries, reflecting the fact that larger countries are more likely to have reliable abortion data. We also obtained data from 139 countries on the proportions of births that were from unintended pregnancies. Overall, then, 150 countries—in which 95% of the population of reproductive-age women live—had reliable data on either or both outcomes.
Births from unintended pregnancies and abortions are related, as they both follow from the incidence of unintended pregnancy and the proportion of such pregnancies that end in abortion. For this reason, we developed a model that utilized key determinants of unintended pregnancy as predictors and that concurrently estimated all outcomes. Differences between countries—and, within countries, over time—in the availability and quality of data are reflected in 80% and 95% uncertainty intervals published with the estimates. With the 80% intervals, there is a 20% chance that the true value is outside the interval, with equal probability (10%) that the truth is below or above the bounds of the interval. With the 95% intervals, there is a 5% chance that the true value is outside the interval—a 2.5% chance that the value lies below the interval and a 2.5% chance that it lies above it.
To validate the approach, we conducted a number of tests. We produced estimates excluding random subsamples of the underlying data, as well as systematically excluding all of the data for each country, one at a time. These exercises showed that the model is well calibrated. The model clustered countries that were expected to have similar relationships between outcomes and predictors. Since the availability and quality of data may systematically differ between these groups, we also evaluated model performance for each of these clusters. Because we had no reliable abortion data for countries in the cluster of North Africa and the Middle East with which to confirm model calibration, we did not release country-specific estimates for any countries in that cluster. However, those countries are included in the published average estimates for the regions and subregions that they are part of.
Were countries consulted when these estimates were produced?
A WHO country consultation process was conducted, led by members of the research team at the UNDP/UNFPA/UNICEF/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction (HRP). During this process, the team shared a translated technical note with WHO’s regional offices and country-level focal points from health ministries and national statistical offices, along with country-specific estimates from the model, as well as a list of all data and sources used in the model.
Through this process, the study team invited country-level focal points to comment on the data we included for their country and on the methodology, and, if applicable, to share any additional data that they had for their country; and we responded to their questions. The criteria described in the study protocol determined whether and how to include any new data submitted for potential inclusion in the model during this process. As a result, this consultation process does not necessarily imply that countries approved the model-based estimates.
How are estimates presented?
All rates, percentage values and numbers are annual averages for a five-year period, and the country-specific estimates are available for six time periods from 1990 to 2019. The unintended pregnancy rate and the abortion rate are the annual numbers of abortions and unintended pregnancies, respectively, per 1,000 women aged 15–49. Unintended pregnancies are those that are mistimed or not wanted at all. The total number of pregnancies includes live births, abortions and fetal losses.
How do we interpret the estimates?
We estimate that the global average abortion rate declined by 12% between 1990–1994 and 2000–2004, from 40 to 35 per 1,000 women aged 15–49, before increasing by 11%, to 39 per 1,000, in 2015–2019. The estimates we present are medians from model-estimated distributions of possible values; for example, there is an equal probability (50%) that the true global average abortion rate in 2015–2019 was smaller than or greater than 39 per 1,000.
The model-based estimates for each country have varying degrees of uncertainty over time and across measures, reflecting differences between time periods in the quantity and quality of the country’s data. In turn, averages across groups of countries—whether the global average or an average across all countries in a region, subregion or income group—reflect the uncertainty in their country-specific estimates. For this reason, we calculated uncertainty intervals to be used when interpreting the estimates. When we write that the global average abortion rate in 2015–2019 was 39 per 1,000 and indicate that the 95% uncertainty interval extends from 34 to 46, we mean that there is a 95% probability that the global average rate in this time period was between 34 and 46, a 2.5% chance that it was less than 34 and a 2.5% chance that it was greater than 46.
We produced uncertainty intervals for comparisons between countries and across time periods. But with such comparisons, we also calculated probabilities of change in the median estimated direction. (Where the estimate is positive, the probability of increase is shown; where the estimate is negative, the probability of decrease is shown.) For example, the decline in the global abortion rate from 40 to 35 per 1,000 between 1990–1994 and 2000–2004 was associated with a 96% probability of decrease—i.e., there was a 96% chance that the abortion rate decreased, or, equivalently, a 4% chance that it increased. If we compare the global average abortion rates for the time periods 1990–1994 and 2015–2019 (40 and 39 per 1,000), the probability of decrease was 61%, meaning there was a 61% chance that the abortion rate declined and a 39% chance that it increased over those 30 years. Hence, we write that there is little evidence that the abortion rate differed between 1990–1994 and 2015–2019. However, this does not mean that there were no significant trends in the global average abortion rate over the entire analysis period; in fact, the probability that the abortion rate declined from 1990–1994 to 2000–2004 was 96%, and the probability that the abortion rate increased from 2000–2004 to 2015–2019 was 92%.
Why use model-based estimates when in-country data are available?
These model-based estimates provide valuable information that is additional to in-country data. For example, they provide trend information for a 30-year period (1990–2019) for key outcomes, including unintended pregnancy and abortion. Many countries have reliable data on some pregnancy indicators and not on others, and some countries do not have any reliable data. These new estimates also make possible international comparison—for example, comparing one country with another country, or with a regional average—because all estimates are harmonized. Thus, the model provides a more comprehensive and integrated set of indicators of pregnancy and its outcomes for each country.
Countries and territories for which we published country-level estimates
Angola, Benin, Botswana, Burkina Faso, Burundi, Cabo Verde, Cameroon, Central African Republic, Chad, Comoros, Congo, Côte d'Ivoire, Democratic Republic of the Congo, Djibouti, Eritrea, Eswatini, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mayotte, Mozambique, Namibia, Niger, Nigeria, Réunion, Rwanda, Sao Tome and Principe, Senegal, Sierra Leone, Somalia, South Africa, South Sudan, Togo, Uganda, United Republic of Tanzania, Zambia, Zimbabwe
Canada, United States
Armenia, Azerbaijan, Bangladesh, Bhutan, Cambodia, China, Georgia, Hong Kong (China), India, Indonesia, Japan, Kazakhstan, Kyrgyzstan, Lao People's Democratic Republic, Maldives, Mongolia, Myanmar, Nepal, Pakistan, Philippines, Republic of Korea (South Korea), Singapore, Sri Lanka, Tajikistan, Thailand, Timor-Leste, Turkmenistan, Uzbekistan, Viet Nam
Albania, Belarus, Belgium, Bosnia and Herzegovina, Bulgaria, Croatia, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Italy, Latvia, Lithuania, Montenegro, Netherlands, North Macedonia, Norway, Poland, Portugal, Republic of Moldova, Romania, Russian Federation, Serbia, Slovakia, Slovenia, Spain, Sweden, Switzerland, Ukraine, United Kingdom
Latin America and the Caribbean
Argentina, Barbados, Belize, Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, Dominican Republic, Ecuador, El Salvador, French Guiana, Guadeloupe, Guatemala, Guyana, Haiti, Honduras, Jamaica, Martinique, Mexico, Nicaragua, Panama, Paraguay, Peru, Puerto Rico, Saint Lucia, Suriname, Trinidad and Tobago, Uruguay
Australia, New Zealand, Papua New Guinea, Samoa, Solomon Islands, Tonga, Vanuatu
Countries and territories for which we did not publish country-level estimates
Afghanistan, Algeria, Antigua and Barbuda, Aruba, Austria, Bahamas, Bahrain, Brunei Darussalam, Curaçao, Cyprus, Democratic People's Republic of Korea (North Korea), Egypt, Equatorial Guinea, Fiji, Grenada, Guam, Iran, Iraq, Ireland, Israel, Jordan, Kiribati, Kuwait, Lebanon, Luxembourg, Libya, Malaysia, Malta, Mauritius, Morocco, Oman, Qatar, Saint Vincent and the Grenadines, Saudi Arabia, Seychelles, State of Palestine, Sudan, Syrian Arab Republic, Tunisia, Turkey, United Arab Emirates, United States Virgin Islands, Venezuela, Western Sahara, Yemen
This fact sheet was made possible by UK Aid from the UK Government and grants from the Dutch Ministry of Foreign Affairs and the Norwegian Agency for Development Cooperation. The findings and conclusions contained within do not necessarily reflect the positions or policies of the donors.