Making sense of US maternal mortality data

Accurately identifying US maternal deaths has significant challenges. Data is collected and measures reported independently by two different national organisations and over 50 different state and local jurisdictions. There are at least two different definitions widely used: WHO Maternal Mortality Rate, used by the National Center for Health Statistics, limits deaths during or up to 42 days post-end of pregnancy and the CDC Pregnancy-Related Mortality Rate also used by state maternal mortality committees that include deaths up to a year after the end of the pregnancy. The very low rate of maternal deaths in both measures makes them especially sensitive to changes in definition and user error in the completion of vital records. The article by Joseph and colleagues, ‘Maternal mortality in the United States: are the high and rising rates due to changes in obstetrical factors, maternal medical conditions, or maternal mortality surveillance?’, discusses the consequences of these issues in detail.1 Here we provide some additional insights into the difficulties and successes of measuring maternal deaths. Measuring maternal mortality is altogether different from measuring infant mortality. Very limited data is required for generating infant mortality rate including the date of death, the date of birth and the gestational age (to allow for cut-offs at either 20 or 28 weeks). Minimal or no judgements are needed. The definition has remained unchanged for generations. In contrast, accurate assessment of maternal deaths requires multiple data sources and a series of judgements. First, we need to identify deaths among all pregnant persons and persons who delivered in the prior 42 days, and even more difficult, persons who delivered in the last year following the end of a pregnancy. Even the linkage of all death records to birth records (which can lead to significant delays) will miss maternal deaths that occurred before 20 weeks of gestation. But the hardest judgement is determining, without review of the medical records, whether the death from a medical cause was ‘related to or aggravated by the pregnancy.’ Even in a maternal mortality review committee with access to autopsy results and full case records, it can still be a very subjective and hotly debated decision. To make matters even worse, the maternal mortality definitions and even the fields on the death certificate have changed multiple times in the last 40 years.2 Our current challenges are directly related to these issues. An oft taught rule in public health is that an abrupt large change in population outcomes is usually associated with a change in either data collection or measure definition. The change from ICD-9 to ICD-10 for Death Certificates in 1999 is a case in point. Under ICD-9 rules, a death could only be counted as pregnancy-related if pregnancy was identified in the list of immediate and underlying causes (US Death Certificate Part 1). Under ICD-10, the definition of pregnancy-related death was expanded to additionally include deaths where pregnancy was noted as ‘conditions that contributed to death but was not part of the chain of events that led directly to death’ (US Death Certificate Part 2).2 The inclusion of these deaths where the pregnancy was not on ‘the casual pathway’ led to an immediate and significant increase in the rate of maternal deaths from ‘primarily medical causes’ (indirect deaths) from 1.8 to 8.1 per 100,000. However, not often acknowledged, the actual rate of ‘primarily obstetric causes’ has remained fixed at 7.9 per 100,000 (Table 1).3 Multiple authors have also noted the steady increase in the rates of maternal comorbidities which has also contributed to the rising rates of severe maternal morbidity and maternal deaths due to ‘primarily medical causes’.1, 4, 5 In this same time period, public health leaders were concerned about underreporting of maternal deaths because of lack of recognition of pregnancy on the death certificate. As many as 30%–50% of maternal deaths were identified as being missed.6 Early reports from 14 states indicated that additional reminders on the death certificate could capture these cases without resulting in over-reporting.6 A revised US standard death certificate was released in 2003 that included a recommended standard ‘pregnancy checkbox’ but it has taken over 10 years for all states to adopt it. Unfortunately, as Joseph et al. and others have noted, since the pregnancy checkbox has been in widespread use, it appears to have led to significant over-calling of maternal deaths due to ‘primarily medical causes’. We are currently seeking the proper balance of under-reporting versus over-reporting maternal deaths. There are three different reasons to collect maternal mortality data: To trend a key public health indicator internally; to compare to other countries and to recognise improvement opportunities. An international comparison requires the use of the WHO definition. However, it is unknown whether the maternal mortality surveillance systems in other countries follow the same coding rules and have undergone the same scrutiny as we have in the United States. We have already seen how the changing rules and the checkbox have made trending US maternal mortality imprecise. I had the honour of leading the largest maternal mortality review committee in the United States for over 15 years. From the outset in 2006, we chose to review every maternal death with the goal to identify clinical and policy improvement opportunities. Even more importantly, we were able to turn ‘data into action’ by taking lessons uncovered in the case reviews and turning them first into topic-specific Quality Improvement toolkits and then implementing them in a series of large-scale quality collaboratives (each over 100 hospitals) that met together monthly in small groups with structure, process and outcome metrics.7 California maternal death rates using either the Maternal Mortality or the Pregnancy-Related Mortality measures at 42 days is now half that for the United States.8 The toolkits also inspired and serve as resources for the US National Safety Bundles in the Alliance for Innovation on Maternal Health (AIM) initiative based at the American College of Obstetricians and Gynecologists (ACOG) (https://saferbirth.org/patient-safety-bundles/). In the United States, public health and vital records are largely decentralised to individual state departments of health. The CDC has now funded in 46 states standardised maternal mortality review committees with the explicit goal of ‘review to action’.9 The California toolkits and AIM Safety Bundles have focused on the Primarily Obstetric (Direct) causes of mortality. Less success has been made with the Primarily Medical (Indirect) causes of maternal deaths particularly in the late postpartum period (>42 days to a year post end of pregnancy) where primarily medical causes dominate. Postpartum is now a focus of CDC efforts and there are real opportunities to improve the care of women following birth. Policy changes include extension of Medicaid coverage to a year and the introduction of care coordinators/care navigators. There are also calls to expand the definition of pregnancy-related deaths to include overdose deaths, suicides and homicides. However, there are reasons for pause. As the months pass from delivery it gets harder to establish pregnancy relatedness for primarily medical causes. Indeed, the five most common causes of death in non-pregnant reproductive age women are (in order), accidental overdose, malignancies, cardiac disease, suicide and homicide. All are at considerably higher rates per year than in the postpartum year.10 Even suicide rates are no more common in the postpartum year than other years for reproductive age women.8 While late postpartum deaths potentially generate more attention but they are not less or more tragic than other deaths among young women. All preventable deaths should be reduced but it remains unclear whether these deaths should all be combined into maternal death measures. While this discussion has illustrated many of the challenges with measures of maternal mortality, public attention to the measures has generated new funding and intense clinical and public health focus on maternity care. Furthermore, it cannot be overlooked that every version of maternal mortality measures shows disparities of at least threefold for Black and Native American mothers. Inequities remain one of the greatest challenges for US perinatal healthcare. The author is supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Develop- ment of the National Institutes of Health under Award Number UH3HD108053. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. No disclosures or conflicts. Dr. Elliott K. Main was the founding Director of the California Maternal Quality Care Collaborative (CMQCC) and chaired the largest maternal mortality review committee in the United States (California Pregnancy-Associated Mortality Review Committee) for over 15 years. California has 1 of every 8 births in the United States (with a peak of over 500,000 annual births). He has led multiple large-scale maternal quality improvement projects, including collaboratives with over 100 hospitals addressing haemorrhage, hypertension, sepsis, and reducing primary caesarean deliveries. He is a Professor of Obstetrics and Gynecology at Stanford University School of Medicine, CA. There is no unpublished data presented in this commentary.