Texas Won’t Study How Its Abortion Ban Impacts Women, So We Did

by Andrea Suozzo, Sophie Chou and Lizzie Presser

ProPublica is a nonprofit newsroom that investigates abuses of power. Sign up to receive our biggest stories as soon as they’re published.

A first-of-its-kind analysis by ProPublica found that the sepsis rate in second-trimester pregnancy loss hospitalizations increased by more than 50% after Texas’ near-total abortion ban went into effect in September 2021. The analysis also identified at least 120 in-hospital deaths of pregnant or postpartum women in 2022 and 2023 — an increase of dozens of deaths from a comparable period before the COVID-19 pandemic.

Neither the CDC nor states are investigating deaths or severe maternal complications related to abortion bans. And although the federal government and many states track severe complications in birth events using a federally established methodology, far less is known about complications that arise during a pregnancy loss. There is no federal methodology for doing this, so we consulted with experts to craft one.

We acquired Texas hospitalization data from 2017 through 2023, giving us more than two years of data after the implementation of the state’s six-week abortion ban in September 2021, and more than a year of data following its full abortion ban, which went into effect in August 2022.

We spoke with dozens of researchers and clinicians to adapt the federal algorithm for birth complications to focus on severe complications in early pregnancy, before fetal viability.

This methodology lays out the steps we took to complete this analysis, to help experts and interested readers understand our approach and its limitations.

Identifying Second-Trimester Hospitalizations

We purchased seven years of inpatient discharge records for all hospitals from the Texas Department of State Health Services. These records contain de-identified data for all hospital stays longer than a day, with information about the stay, including diagnoses recorded and procedures performed during the stay, as well as some patient demographic information and billing data.

Within this dataset, we opted to focus on second-trimester pregnancy loss, because first-trimester miscarriage management often occurs in an outpatient setting. In the future, we plan to look at outpatient data as well.

To examine outcomes in the second trimester, we first identified hospitalizations where a pregnancy ended. We used a methodology to identify severe complications in birth events developed by the Health Resources and Services Administration, the Centers for Disease Control and Prevention, the Agency for Healthcare Research and Quality, and the Alliance for Innovation on Maternal Health, an initiative of the American College of Obstetricians and Gynecologists. The method is outlined in statistical code published by HRSA, and it first identifies every hospitalization with a live birth, stillbirth or an “abortive outcome” (which refers to an intended or unintended pregnancy loss before 20 weeks). Rather than excluding those abortive outcomes to focus on birth, as the HRSA code directs, we included them to look at all hospitalizations where a pregnancy ended. This narrowed our list of hospitalizations to an average of 370,000 per year.

The HRSA methodology further filters hospitalizations to only patients who are female and between the ages of 12 and 54. Our dataset had five-year age ranges, so we filtered to ages between 10 and 54. This brought our hospitalization list to 364,000 each year, on average.

For each hospitalization where a pregnancy ended, we looked for a diagnosis code recording the gestational age of the fetus. In cases where a long hospitalization had multiple gestational week codes recorded over the course of the stay, we took the latest one.

We excluded pregnancy-end hospitalizations without a gestational week code from our analysis — removing about 49,500 hospitalizations, or 1.9% of our dataset. More than two-thirds had coding that indicated a birth, likely to have occurred after 20 weeks.

Based on conversations with doctors and researchers, we narrowed our focus to hospitalizations where a pregnancy ended in the second trimester before fetal viability, from the start of the 13th week through 21 weeks and six days. While pregnancies that end at 20 and 21 weeks are often coded as births, rather than abortive outcomes, we included those weeks in our definition of pregnancy loss because experts told us it’s extremely unlikely that a baby born at 21 weeks could survive. This brought our list of hospitalizations to 15,188.

The number of second trimester hospitalizations, and characteristics of the women hospitalized, was largely stable from 2017 through 2023, the years of our analysis. In 2023, however, as the number of births in the state increased, the number of hospitalizations in our window declined to 2,036, below the yearly average of 2,169.

The race and ethnicity of patients each year, as well as the proportion of these hospitalizations in which the patients were covered by Medicaid or uninsured, did not change significantly after the state’s 2021 abortion ban, known as SB 8, went into effect.

Determining Sepsis Rates

Within these hospitalizations, we looked for diagnoses of sepsis, a life-threatening complication that can follow delays in emptying the uterus. The CDC defines a list of sepsis codes associated with severe maternal complications, which formed the basis of our definition. However, that list of codes is developed to look at sepsis in birth events, the vast majority of which occur much later in a pregnancy than our hospitalizations. We identified five sepsis codes associated with early pregnancy events like ectopic pregnancy and miscarriage, adding them to the existing list of sepsis codes to develop a definition that more fully captured early pregnancy complications.

To compare rates before and after the implementation of SB 8, we grouped the nine quarters of data we had after the implementation of the ban (October 2021 through December 2023) and compared it with the nine quarters immediately before (July 2019 through September 2021). Our dataset gives us the quarter in which a patient was discharged from the hospital but not the exact date, so the “before” group contains one month of data from after SB 8 went into effect on Sept. 1, 2021.

Identifying Fetal Demise

The standard of care for second-trimester miscarriage or rupture of membranes prior to fetal viability is to offer patients a dilation and evacuation or an induction to end the pregnancy — even if there is still a fetal heartbeat. In our reporting, we’d heard that because of the language of Texas’ abortion law, some hospitals and doctors were waiting for the fetal heartbeat to stop or the mother to develop a life-threatening illness, whichever occurred first. To look into this, we wanted to separate hospitalizations in which doctors would have theoretically been able to offer a termination immediately under the law — ones where the patient had a diagnosis indicating that there was no fetal heartbeat at the time of admission to the hospital — from ones where doctors may have waited to provide care.

We determined that about half of our second-trimester hospitalizations did not have a fetal heartbeat on admission. We identified these cases by focusing on two sets of diagnosis codes: Prior to 20 weeks gestation, a diagnosis of “missed abortion” refers to a miscarriage where the fetus has stopped developing, but the body has not yet expelled the tissue. After 20 weeks, a diagnosis of “intrauterine death” indicates that the fetus has died. For both diagnoses, we included only those that were marked as “present on admission.”

Sepsis Rate Findings

Our analysis found that the sepsis rate in second-trimester pregnancy loss hospitalizations increased after the state’s ban went into effect, and the surge was most pronounced in cases in which the fetus may still have had a heartbeat when the patient arrived at the hospital.

In the nine quarters before SB 8 went into effect, the sepsis rate in second-trimester pregnancy loss hospitalizations was 2.9%. In the nine quarters after SB 8 went into effect, the sepsis rate was 4.5%, an increase of 55%.

Since our total number of sepsis cases was relatively small, we measured whether the two groups of data were significantly different using a t-test. We calculated sepsis rates for second-trimester hospitalizations in the nine quarters after SB 8 went into effect and compared that with sepsis rates during the nine quarters immediately prior. We found that increase to be statistically significant (p-value < 0.05).

Sepsis Rate Increased Over 50% for Second-Trimester Pregnancy Loss Hospitalizations After SB 8

We compared the nine quarters after SB 8 went into effect — from October 2021 through December 2023 — to the nine quarters before the ban went into effect — July 2019 to September 2021.

Note: For hospitalizations involving a pregnancy loss between 13 weeks’ gestation and the end of the 21st week.

Sepsis is a reaction to an infection, and the most common additional infection diagnosis in sepsis hospitalizations was chorioamnionitis, an infection of the amniotic fluid that can also cause early rupture of membranes. Rates of chorioamnionitis in sepsis cases remained stable before and after SB 8.

Our analysis also showed that patients admitted while their fetus was still believed to have a heartbeat were far more likely to contract sepsis.

Sepsis Rates Spiked for Patients Whose Initial Diagnosis Didn’t Include Fetal Death

For patients in Texas hospitals who lost a pregnancy, about half were not diagnosed with fetal demise when they were admitted, meaning that their fetus may still have had a heartbeat at that time. Those patients saw a dramatic increase in sepsis after the state banned abortion.

Note: For hospitalizations involving a pregnancy loss between 13 weeks’ gestation and the end of the 21st week. We identified patients whose fetus had no heartbeat when they were admitted by looking for a diagnosis of “intrauterine death” or “missed abortion.” Rates are annual. (Lucas Waldron/ProPublica)

In the nine quarters prior to the implementation of SB 8, the rate of sepsis was nearly twice as high for those with no fetal demise diagnosis on admission compared with those with a fetal demise diagnosis on admission. After SB 8, the rate increased in both groups, and the gap between them widened.

Again, since the number of total sepsis cases was relatively small, we used a t-test to see if there was a statistically significant difference before and after SB 8 in both groups. We found the increase in rates to be significant on both counts (p < 0.05).

Sepsis Rates for Hospitalizations With Fetal Demise on Admission Sepsis Rates for Hospitalizations Without Fetal Demise on Admission Notes: For hospitalizations involving a pregnancy loss between 13 weeks’ gestation and the end of the 21st week. We compared the nine quarters after SB 8 went into effect to the nine quarters before the ban went into effect. Sepsis Rate Analysis Limitations

Maternal health experts noted that discharge data offers a limited window into the details of patient care. Changes in the frequency of a diagnosis code can signal a change in patient health but also a change in coding practices. Our analysis can’t isolate changes in outcomes from changes in sepsis coding practices over time or doctors taking additional documentation steps to show they’ve complied with the law. And billing records offer no detail into a patient’s history and medical wishes or the decisions that medical staff make in the course of care.

Our analysis also does not account for changes in health care outside of hospitals. Though births typically take place in a hospital, other early pregnancy care often occurs in an outpatient setting and does not require a hospitalization, so we can only see a small subset of this type of care — specifically, the most severe cases. We also can’t account for how closures of reproductive health care clinics in the wake of Texas’ abortion ban changed the role hospitals play in miscarriage care.

We cannot see when hospitals turn patients away rather than admitting them. And if a patient who is miscarrying has an inpatient stay at one hospital and is then transferred to another hospital for another inpatient stay, that patient would be double-counted in our analysis, since we can’t connect patients across visits. This could potentially inflate the number of hospitalizations in our dataset, artificially pushing the sepsis rate down.

Our dataset is missing a handful of records from the fourth quarter of 2023; in a small number of cases — about 300 per quarter, or 0.04% of records — providers submit data on a hospitalization late, and that record is released in the dataset for the following quarter.

Billing data is widely used by researchers to study maternal health. While it will never tell the whole story, in aggregate, particularly in a state with a large population, it can paint a picture of changing health outcomes. Our analysis gives us a broad view of care at Texas hospitals before and after a major policy change.

More than a dozen maternal health experts reviewed ProPublica’s findings and said our analysis adds to mounting evidence that the state’s abortion ban is likely leading to dangerous delays in care. Many said the ban is the only explanation they could see for the sudden jump in sepsis cases.

Pregnancy-Associated Hospital Deaths

We found 120 women who died while hospitalized during pregnancy or up to six weeks postpartum in 2022 and 2023 in the inpatient billing data. The Texas Maternal Mortality and Morbidity Review Committee will not review deaths from these years, stating that they will skip to 2024 in an effort to get a more “contemporary” view of deaths, a choice that faced widespread criticism. (The committee chair said there was “absolutely no nefarious intent” behind the decision.)

To identify inpatient deaths in the Texas hospital discharge data, we included all records with a “patient status” of “expired” and with a diagnosis or procedure code indicating that the patient was pregnant or up to six weeks postpartum, with a specific postpartum complication based on the “Identifying Pregnant and Postpartum Medicaid and CHIP Beneficiaries” code list by the Centers for Medicare & Medicaid Services. The CDC looks at deaths up to within one year of a pregnancy’s end, but our dataset doesn’t explicitly identify pregnant or recently pregnant patients, so we were limited in the hospitalizations we could identify through codes.

Our tally does not include those who died in a hospitalization that took place separately from the end of a pregnancy, unless the patient was diagnosed with a specific postpartum complication. We did not filter for age and gender for our death records, as that data was less reliably filled out than the diagnosis and procedure codes.

Our count of inpatient deaths does not attempt to determine what role a person’s pregnancy or the state’s abortion ban played in their death. That type of analysis would require access to medical records. Our tally would include, for example, a person who was hospitalized after a car crash but who was also pregnant. Experts advised us to leave these cases in, because without investigation by the maternal mortality committee, it’s impossible to know, for example, if there was any relationship between the patient’s pregnancy and the cause of the accident, or if there were any delays in maternal care after the accident.

We found that deaths increased sharply during the height of the COVID-19 pandemic and peaked in 2021, and that many cases in 2020 and afterward included COVID-19 diagnostic codes. More than 60% of the deaths that we analyzed had a diagnosis of COVID-19 in 2021, and 27% had a COVID-19 diagnosis in 2022. The COVID-19 diagnostic code was not introduced until October 2020, several months after the pandemic began, and was updated in January 2021. The coding changes, combined with changes in hospital protocols around identifying COVID-19 cases, make it impossible to filter out all COVID-19 related deaths during this time period.

Texas and National Rates of Maternal Mortality

The hospital billing data only includes information about Texas, so to compare with national rates, we used data from the CDC’s WONDER portal, which is based on birth and death certificates. For this analysis, we used a definition of maternal death recommended by CDC research guidelines for this data source. Our denominator includes all live births. For statewide rates, we use the state of residence of the mother in both the numerator and denominator. Rates are reported per 100,000 births.

Between 2019 and 2023, we found a 33% increase in maternal mortality rates in Texas, compared with a decrease of 7.5% nationally during the same time.

While both nationally and in Texas rates of maternal mortality peaked in 2021 during the COVID-19 pandemic and have dropped since, rates in Texas remain higher than before the pandemic.

Missing Documents

The federal methodology we used as a basis for our analysis of severe complications in pregnancy hospitalizations was outlined in a document available for download from HRSA’s Maternal and Child Health Bureau. The instructions included statistical code that we adapted to do our own analysis, and they were accompanied by a spreadsheet of maternal and child outcome measures over time for all 50 states and nationally.

As of early February, both the instructions and the spreadsheet had been replaced by documents noting that the files were “currently under construction and not available.”

Lucas Waldron contributed graphics.

This content originally appeared on ProPublica and was authored by by Andrea Suozzo, Sophie Chou and Lizzie Presser.