Highlights
In Brazil, 3,480 maternal deaths from 2010 to 2021 had pregnancy or puerperal status documented on the death certificate but were not coded as maternal deaths in the official system.
Reclassifying these uncoded deaths increased the median maternal mortality ratio from 58.2 to 67.5 per 100,000 live births, indicating meaningful national underreporting.
Uncoded maternal deaths were disproportionately associated with lower educational attainment, younger and older maternal age groups, indirect causes of death, and fewer formal death investigations.
The findings suggest that improvements in death certification, coding quality, and maternal death surveillance could materially alter national performance monitoring and policy prioritization.
Background
Maternal mortality remains one of the most sensitive indicators of health system performance, social inequity, and access to timely obstetric care. The World Health Organization defines maternal death 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, but not from accidental or incidental causes. Accurate classification is essential because maternal mortality ratios are used to evaluate progress toward national and global targets, guide resource allocation, and identify failures in emergency obstetric care, referral systems, and chronic disease management during pregnancy.
Brazil has long maintained a national mortality information system and has invested in maternal death committees and surveillance processes. Even so, underreporting has remained a persistent concern, especially when deaths are attributed to nonspecific clinical syndromes, indirect medical causes, or incomplete death certification. In middle-income settings with complex referral networks and regional disparities, the distinction between a documented pregnancy-related death and a formally coded maternal death can be lost at multiple points: by the certifying physician, during registry data entry, or at the stage of International Classification of Diseases coding.
The study by Pinheiro and colleagues addresses a clinically and policy-relevant question: how many maternal deaths are present in the Brazilian mortality database but omitted from the official maternal mortality count because of coding errors? This is not simply a technical issue. If systematic undercounting is concentrated among socially vulnerable women or among deaths from indirect causes such as cardiovascular disease, infection, or other medical disorders aggravated by pregnancy, then national statistics may underestimate both the scale and the changing profile of maternal risk.
Proposed Article Structure
This topic is best interpreted through five linked domains: the public health burden of maternal mortality, the mechanics of mortality coding and surveillance, the observational study design, the quantitative impact of reclassification on national maternal mortality estimates, and the clinical-policy implications for improving ascertainment. The sections below follow that logic.
Study Design and Methods
Design and data source
This was a retrospective observational study using Brazilian mortality data from 2010 through 2021. The investigators analyzed two groups of deaths: officially reported maternal deaths and deaths among women aged 10 to 49 years in whom pregnancy-puerperal status was indicated on the death certificate but who were not coded as maternal deaths.
Population
The study focused on women of reproductive age, 10 to 49 years, a standard surveillance population for maternal mortality assessment. The inclusion of this broad age range is important because miscoding may occur at both ends of the reproductive spectrum, where pregnancy can be less expected or less consistently recognized in administrative datasets.
Exposure and comparison
The central comparison was between maternal deaths that were officially coded as maternal and maternal deaths that were identifiable from pregnancy-puerperal information on the death certificate but remained uncoded in official maternal mortality statistics. In practice, the study tested how reclassification of these missed cases would alter the maternal mortality ratio.
Main endpoint
The primary endpoint was the difference between the official maternal mortality ratio and a recalculated ratio that included reclassified maternal deaths. The maternal mortality ratio is conventionally expressed as maternal deaths per 100,000 live births.
Secondary analytic questions
The investigators also examined characteristics associated with uncoded compared with coded maternal deaths, including education, age, type of cause of death, and whether the death underwent investigation. These variables are highly relevant because they can reveal structural weaknesses in surveillance rather than random administrative noise.
Key Findings
Magnitude of underreporting
Brazil recorded 21,670 official maternal deaths over the 12-year study period. In addition, the authors identified 3,480 uncoded maternal deaths among women whose death certificates indicated pregnancy or puerperal status. This means that a sizable number of maternal deaths were already present in the data environment but did not contribute to the official count because of coding failure.
Put differently, the uncoded deaths represented a substantial increment over the official tally. Even without a more detailed yearly breakdown in the abstract, the scale is large enough to change the interpretation of national trends and the estimated distance from maternal mortality reduction targets.
Effect on the maternal mortality ratio
The official median maternal mortality ratio over the study period was 58.2 per 100,000 live births. After inclusion of reclassified maternal deaths, the median maternal mortality ratio rose to 67.5 per 100,000 live births. This absolute increase of 9.3 deaths per 100,000 live births is epidemiologically important. In national benchmarking, changes of this magnitude can alter international comparisons, state-level priority setting, and assessment of whether prior interventions have achieved meaningful reductions.
For clinicians and health system leaders, the implication is straightforward: official statistics may paint an overly optimistic picture of maternal mortality performance if they rely on incomplete coding capture rather than comprehensive ascertainment of pregnancy-associated deaths.
Who was more likely to be missed?
Uncoded maternal deaths were more likely than coded maternal deaths to occur in women with lower education. This finding is consistent with broader evidence that socially marginalized patients are more likely to experience both worse outcomes and poorer documentation quality. Administrative invisibility can therefore compound clinical vulnerability.
The study also found greater undercoding at younger and older maternal ages. These age extremes are already associated with elevated obstetric risk, although often through different mechanisms. Adolescents may face barriers to prenatal care, stigma, and fragmented documentation, whereas older pregnant patients more commonly present with complex comorbidity and indirect causes of death that may be coded under the underlying medical condition rather than the maternal context.
Indirect causes are especially vulnerable to miscoding
Uncoded maternal deaths were more likely to involve indirect causes. This is one of the most clinically important observations in the study. In many countries, the epidemiology of maternal death has shifted from predominantly direct obstetric causes such as hemorrhage, hypertensive disorders, and sepsis toward a growing contribution from indirect causes including cardiovascular disease, thromboembolism, respiratory disease, malignancy, and other preexisting or concurrent disorders aggravated by pregnancy.
Indirect maternal deaths are inherently easier to miss in coding systems because the certifier may emphasize the terminal event or chronic disease diagnosis while failing to link the death to pregnancy. If surveillance systems undercount indirect deaths, then national strategies may underinvest in preconception care, medical-obstetric co-management, and referral pathways for high-risk pregnant patients with chronic disease.
Role of death investigation
Uncoded maternal deaths were also associated with fewer death investigations. This finding reinforces the value of structured review processes, including maternal mortality committees and hospital or municipal investigation workflows. Investigation often clarifies whether pregnancy contributed to the death, resolves ambiguous certification, and supports recoding when appropriate. Where investigations are less frequent, missed maternal deaths are likely to persist in official statistics.
Clinical and Public Health Interpretation
This study does more than quantify an administrative gap. It shows that underreporting is patterned, not random. The women most likely to be omitted from official maternal mortality counts appear to be those already at elevated risk of poor outcomes: women with lower educational attainment, those at age extremes, and those dying from indirect causes. That pattern matters because it means coding error can distort both the level and the composition of maternal mortality.
From a health systems perspective, three operational lessons emerge. First, pregnancy status fields on death certificates need to be treated as critical surveillance variables, not optional administrative details. Second, coder training should emphasize maternal death rules for indirect causes and the postpartum period. Third, linkage and review systems should be strengthened so that deaths among reproductive-age women trigger routine evaluation of recent pregnancy when clinical circumstances are compatible.
These findings are also highly relevant to interpretation of temporal trends. Between 2010 and 2021, Brazil experienced evolving patterns in obstetric care, chronic disease burden, and, at the end of the period, the COVID-19 pandemic. Any attempt to interpret changes in maternal mortality over time must account for whether case ascertainment remained stable. Apparent improvements or deteriorations can be misleading if the proportion of uncoded deaths changes because of registry quality, investigation intensity, or shifts toward more complex indirect causes.
Strengths and Limitations
Strengths
The major strength of this study is its national scope and its practical relevance. Rather than estimating underreporting indirectly through modeling assumptions, the investigators identified deaths that had pregnancy-puerperal information documented but were omitted from official maternal coding. This approach addresses a real-world surveillance failure that is amenable to intervention.
A second strength is the characterization of which deaths were most likely to be missed. That provides actionable direction for policy, training, and quality assurance.
Limitations
The abstract does not provide detailed methods on the exact reclassification algorithm, the ICD coding logic applied, or year-by-year variation in underreporting. Without those details, readers cannot fully assess the possibility of residual misclassification in either direction.
The analysis also appears dependent on what was recorded on the death certificate. Deaths in which pregnancy status was never documented would still be missed, meaning the study may underestimate the true degree of underascertainment. In that sense, the recalculated maternal mortality ratio may still be conservative.
Finally, while the findings are highly informative at national level, Brazil is regionally heterogeneous. The burden and mechanisms of underreporting may differ across states, urban versus rural settings, and referral tiers. Subnational analyses would be useful for targeted intervention.
Relation to Existing Evidence and Guidelines
Global guidance from the World Health Organization emphasizes robust civil registration and vital statistics systems as foundational to maternal mortality measurement. Maternal death surveillance and response frameworks are intended not only to count deaths, but also to investigate their determinants and drive quality improvement. The present study fits squarely within that agenda by showing how failures in coding can suppress the visible burden of maternal death even when traces of pregnancy status exist in the record.
The findings also align with broader literature showing that maternal mortality is often underestimated when relying solely on routine coding, especially for indirect causes and postpartum deaths. In high-, middle-, and low-income settings alike, enhanced case finding through record linkage, confidential enquiries, and committee review generally identifies more maternal deaths than routine administrative counts alone.
Implications for Practice and Policy
For clinicians, accurate completion of death certificates matters. The causal chain should clearly state when pregnancy or the puerperium contributed to death, including in women who die from medical disorders worsened by pregnancy. This is especially important for intensive care, emergency medicine, internal medicine, cardiology, infectious diseases, and anesthesiology teams who may care for critically ill obstetric patients outside the delivery ward.
For hospitals and health departments, routine audit of deaths among women aged 10 to 49 years should be considered, with mandatory review of pregnancy status and postpartum timing. Electronic alerts and linkage between mortality records, live birth records, and hospital discharge datasets may improve ascertainment.
For policymakers, recalibrated maternal mortality estimates should inform planning, workforce distribution, and the design of maternal death surveillance and response systems. If official ratios are undercounted by this magnitude, program targets and resource allocation formulas may require reappraisal.
Conclusion
The analysis by Pinheiro and colleagues indicates that Brazil’s official maternal mortality statistics from 2010 to 2021 likely underestimated the true burden because of coding errors. The identification of 3,480 uncoded maternal deaths increased the median maternal mortality ratio from 58.2 to 67.5 per 100,000 live births. Importantly, underreporting was not random: it was more common among women with lower education, at younger and older ages, in indirect causes of death, and when formal death investigation was lacking.
The central message is clear. Better maternal mortality reduction policy begins with better maternal mortality measurement. Improving death certification, coding quality, and investigation processes is not an administrative afterthought; it is a core component of equitable maternal health care.
Funding and Trial Registration
No funding information or ClinicalTrials.gov registration number is provided in the abstract. Given the retrospective observational use of national mortality data, formal trial registration would not typically be expected.
References
1. Pinheiro AMPAL, Herzog RS, de Aguiar RALP, Rodrigues AS, Francisco RPV. Maternal Mortality Underreporting in Brazil Due to Coding Errors (2010-2021). Obstetrics and gynecology. 2026-04-02;147(6):e160-e163. PMID: 41926768.
2. World Health Organization. Trends in maternal mortality 2000 to 2020: estimates by WHO, UNICEF, UNFPA, World Bank Group and UNDESA/Population Division. Geneva: World Health Organization; 2023.
3. World Health Organization. Maternal death surveillance and response: technical guidance information for action to prevent maternal death. Geneva: World Health Organization; 2013.