"Maternal Mortality Estimation: Methodology, Politics, and the Numbers Behind the Numbers"

EXECUTIVE SUMMARY

Maternal mortality estimation is the most methodologically complex and politically sensitive data exercise in UNFPA's mandate. The headline number — an estimated 287,000 maternal deaths in 2020, corresponding to a global MMR of approximately 223/100,000 — is produced by the Maternal Mortality Estimation Inter-Agency Group (MMEIG), comprising WHO, UNFPA, UNICEF, the World Bank, and UNDESA. This number is not a direct count but a modelled estimate, and understanding the methodology behind it is essential for anyone using maternal mortality data for advocacy, planning, or accountability.

The estimation process involves: collecting available data from CRVS, surveys, and census-based methods; classifying data quality; applying a Bayesian regression model that adjusts for known biases and produces estimates with uncertainty intervals; and reviewing results with countries through a formal consultation process. The resulting estimates are the authoritative global reference, cited in SDG reporting, UNFPA strategic plans, and donor decision-making.

The political dimension is significant: countries may dispute MMEIG estimates that are higher than their own national figures (implying worse performance than the government claims) or lower (implying that the government's investment in maternal health is not reflected). The 2023 report finding that global MMR essentially stagnated between 2016 and 2020 was politically uncomfortable — it meant that the first five years of the SDG period saw no progress on the most fundamental maternal health indicator.

KEY FACTS

MMEIG members: WHO, UNFPA, UNICEF, World Bank, UNDESA
Latest report: "Trends in Maternal Mortality 2000–2020" (published February 2023)
Global MMR (2020): 223 per 100,000 live births (95% UI: 202–256)
Total maternal deaths (2020): 287,000 (95% UI: 261,000–328,000)
Stagnation finding: Global MMR was 227/100,000 in 2015 and 223/100,000 in 2020 — essentially no decline during the SDG period
Regional variation: SSA: 545/100,000; South Asia: 140/100,000; Europe: 10/100,000
Data sources: Only approximately 60 countries have complete CRVS with medical certification of cause of death; the remainder depend on surveys, census, or modelling
Uncertainty: Country-level estimates have wide confidence intervals — for many SSA countries, the 95% uncertainty interval spans a factor of 2 or more
Methodology: Bayesian multilevel regression model; inputs include GDP, skilled birth attendance, general fertility rate, and available MMR data points
Country consultation: MMEIG shares preliminary estimates with countries before publication; countries can provide additional data but cannot change the model
SDG target: MMR below 70/100,000 by 2030 — unreachable at current trajectory

DETAIL

Why Direct Measurement Is Difficult

Maternal death is defined as "the death of a woman while pregnant or within 42 days of termination of pregnancy, from any cause related to or aggravated by the pregnancy or its management." This definition requires: (a) knowing that the woman was pregnant or recently pregnant, (b) determining the cause of death, and (c) assessing whether the cause was related to pregnancy.

In settings with complete CRVS and medical certification: this information is available from death certificates with ICD-coded cause of death. Approximately 60 countries have this capacity.

In settings without complete CRVS: maternal deaths must be identified through household surveys (sisterhood method, direct sisterhood method), census (indirect estimation), or verbal autopsy (non-physician assessment of cause of death). All of these methods have significant limitations — they require large sample sizes (maternal death is relatively rare), produce estimates with wide confidence intervals, and may misclassify deaths.

The MMEIG Model

The Bayesian regression model used by MMEIG:

Collects all available data points (CRVS, surveys, census) for each country
Classifies data quality (some sources are more reliable than others)
Adjusts for known biases (misclassification of deaths in CRVS, underreporting in surveys)
Uses a multilevel regression model with covariates (GDP, SBA, GFR) to produce estimates for all countries and years from 2000 to 2020
Generates posterior distributions with uncertainty intervals

The model's strength is that it produces comparable estimates for all countries even where data is sparse. Its limitation is that for data-scarce countries (many in SSA), the estimate is driven more by the covariates and the regional pattern than by country-specific data.

Political Sensitivity

Maternal mortality is highly politically salient — it is used as an indicator of health system performance, government commitment to women's health, and development progress. Disputes between MMEIG estimates and national figures are common and can be contentious. Some countries have declined to endorse MMEIG estimates, producing their own alternative figures.

UNFPA navigates this by: participating in the MMEIG process, supporting countries in strengthening their own data systems (so that future estimates are based on better country data), and using MMEIG estimates transparently in reporting while acknowledging uncertainty.

SOURCES

WHO, UNFPA, UNICEF, World Bank, UNDESA: "Trends in Maternal Mortality 2000 to 2020" (2023) — full report and statistical annex
Alkema, L. et al.: "Global, regional, and national levels and trends in maternal mortality between 1990 and 2015" (Lancet, 2016)
Say, L. et al.: "Global causes of maternal death: a WHO systematic analysis" (Lancet Global Health, 2014)