Medicaid expansion changed who paid for 12 million births, not severe maternal morbidity

The Affordable Care Act's Medicaid expansion is one of the largest coverage changes in American history, and maternal health is one of the areas it was most expected to improve. A study published in Obstetrics & Gynecology on August 27 tested that expectation against nearly 12 million delivery hospitalisations and did not find it [s1].

The design

This was a retrospective, repeated cross-sectional study using Healthcare Cost and Utilization Project data from the Agency for Healthcare Research and Quality [s1]. All delivery hospitalisations from 2010 to 2018 in 26 US states were analysed; 16 of those states expanded Medicaid in January 2014 and 10 did not [s1].

The outcome was the severe maternal morbidity rate as defined by Centers for Disease Control and Prevention criteria, excluding blood transfusion [s1]. Excluding transfusions is standard practice in this literature, because transfusion thresholds vary between hospitals enough to swamp the signal.

The analysis compared changes from the pre-expansion period (2010–2013) to the post-expansion period (2015–2018) in expansion versus non-expansion states, using generalised synthetic control models to estimate the adjusted average treatment effect on the treated [s1]. The synthetic control approach builds a weighted composite of non-expansion states to stand in for what the expansion states would have looked like — a stronger comparison than a simple before-and-after.

The sample included 11.98 million delivery hospitalisations, among which 82,903 women — 69.2 per 10,000 — experienced severe maternal morbidity [s1].

What happened

The coverage change landed. Medicaid expansion was associated with a 5.2% increase (95% CI 2.1–8.2) in the proportion of delivery hospitalisations paid for by Medicaid [s1]. That is the policy working as designed: fewer uninsured deliveries, more publicly financed ones.

The health outcome did not move. There was no statistically significant change in the severe maternal morbidity rate: 0.02% (95% CI −0.01 to 0.04) [s1]. Results were consistent in subgroup analyses by maternal race and ethnicity and in sensitivity analyses [s1].

The authors' conclusion is that expanded access to public health insurance was not associated with a reduction in severe maternal morbidity during delivery hospitalisations, and that additional clinical and policy interventions may be required to improve maternal health outcomes [s1].

What a null like this can and cannot mean

The confidence interval here is narrow, which makes this a more informative null than most. Between −0.01% and 0.04% is not "we could not tell"; it is closer to "if there was an effect at the delivery hospitalisation, it was very small".

But the outcome window is doing a lot of work. Severe morbidity during the delivery hospitalisation is a narrow slice of maternal health, and it is the slice least likely to respond to insurance status, because childbirth is the one point in the maternity pathway where American hospitals treat people regardless of coverage. Effects of coverage on preconception health, antenatal care, chronic disease control and postpartum care could all be real and would be largely invisible here.

The window also ends in 2018, before the postpartum coverage extensions that several states adopted later.

The measurement problem underneath

There is a second reason to hold this null loosely, and it applies to every study in the field: severe maternal morbidity is not one thing, and which definition is used changes the answer.

A national cohort study published on August 4 applied three definitions to 5,939,919 English maternities delivered at 20 weeks or later between January 2013 and March 2023 [s2]. Under the US CDC definition without blood transfusion, severe maternal morbidity occurred in 1.4% of maternities; under the Canadian Perinatal Surveillance System definition, 2.7%; under the English Maternal Morbidity Outcome Indicator, 1.1% [s2].

Agreement between them was moderate at best. Chance-adjusted agreement was 0.67 for CDC and EMMOI, 0.50 for CPSS and EMMOI, 0.52 for CDC and CPSS, and 0.55 across all three [s2].

Measured inequalities moved with the definition too. For Black women compared with White women, relative risks were 1.96 (95% CI 1.90–2.01), 1.71 (1.67–1.74) and 1.98 (1.92–2.04) under the CDC, CPSS and EMMOI definitions respectively; for women in the most versus least deprived areas, 1.37 (1.33–1.40), 1.18 (1.16–1.21) and 1.22 (1.19–1.26) [s2].

Much of the divergence traces to one code. Among maternities flagged as having severe morbidity, 55%, 41% and 35% respectively had a diagnostic code for sepsis, and sepsis was the only qualifying component for 51%, 44% and 33% of them [s2]. Only 0.3% to 0.6% of those sepsis codes carried an accompanying code for shock [s2].

A composite in which up to half of cases are a single administrative code, largely without evidence of severity, is a fragile instrument for detecting a policy effect.

What to watch

Whether anyone re-runs the expansion analysis with postpartum outcomes and a longer window, and whether the definitional work feeds into a standard the surveillance systems actually adopt. Until they measure the same thing, studies of severe maternal morbidity will keep disagreeing for reasons that have nothing to do with maternal health.