Maternal Morbidity at Duke and UNC: A Multicenter Examination of Risk Factors and Trends of Pregnancy-Related Morbidity

Summary

Maternal morbidity is disappointingly high in North Carolina with unjust race and ethnic disparities. Severe maternal morbidity (SMM) encompasses events that occur during and after pregnancy that are harmful or dangerous to a patient’s health and may even lead to death. These morbidity events include organ damage, such as heart attacks, strokes, lung failure, blood clots, and kidney failure. Severe maternal morbidity occurs in 1.75% of deliveries in North Carolina, compared to roughly 1.2% of deliveries nationwide. Also in North Carolina, Black patients experience SMM at a rate 1.8 times higher than White patients. While predicting which patients may experience maternal morbidity has the potential to improve health care delivery and save lives, there are currently no effective mechanisms for prediction.

Overarching Goals and Approach

Ultimately, this project aimed to reduce maternal morbidity and mortality at Duke and UNC. Towards that goal, the project specifically intended to create a multicenter dataset of maternal comorbidities (medical diseases before pregnancy) and outcomes. With this dataset, the team 1) used the OB Comorbidity Index (OB-CMI; a weighted risk score that uses 20 diseases present at the time of delivery; Leonard 2020) to determine the risk profile of pregnant patients at Duke and UNC and determine the morbidity events that are most common; 2) determined the SMM rate at Duke and UNC; and 3) explored what racial and ethnic disparities exist in SMM. Duke and UNC both have large health systems and significant referral volume for high levels of care.

The dataset included information on patient outcomes, the OB-CMI, and demographic factors including race. To create the multicenter dataset, electronic medical records of both hospital systems were used to collect data about birth-related hospitalizations from 2016 to 2020 regarding comorbidities (such as chronic hypertension and autoimmune diseases) preceding delivery admission and diagnoses acquired during hospital admission, as well as readmission.

Because the OB-CMI was developed in a different region of the U.S., one of our goals was to examine how the OB-CMI performs in this data- set, and how rates of SMM and non-transfusion SMM compared between patients who delivered in our health systems and to national rates generated from delivery hospitalizations in the 2018 National Inpatient Sample (NIS). The distributions of OB-CMI were similarly compared using weighting designed to predict SMM and non-transfusion SMM across the health systems and births in the 2018 NIS sample. Researchers found that, at both Duke and UNC, there was large variation in burden of comorbid conditions and non-transfusion SMM among birthing patients. For the OB-CMI scores, both health systems had a higher median score when compared to the national sample.

To evaluate the association between race and ethnicity and the incidence of SMM, electronic medical records were used to extract demographic and pregnancy-related data. To ensure an adequate sample size and occurrence of the outcome, the analysis was limited to non-Hispanic Black, Hispanic, and non-Hispanic White pregnant people. The primary outcome of interest was non-transfusion SMM at the time of birth hospitalization, as defined by the Centers for Disease Control and Prevention. The secondary outcome of interest was all SMM. There were significant differences in outcomes among groups by age, payer, gestational age at delivery and route of delivery, and OB-CMI score. Odds ratios (OR) were adjusted for primary payer, delivery rout, and OB-CMI. In addition, non-Hispanic Black (OR 1.32, 95% CI 1.11, 1.56) and Hispanic (OR 1.43, 95% CI 1.14, 1.80) birthing people had significantly higher adjusted odds of non-transfusion SMM than birthing non-Hispanic White people, with the adjusted odds of non-transfusion SMM highest among Hispanic people.

Future Directions

Future directions include validation of SMM events by comparing coding to events included in the electronic health records and developing computable phenotypes (a set of criteria that do not require clinician interpretation, such as laboratory data or vitals for each SMM event) to verify SMM events without opening individual charts. Much retrospective clinical research is performed by individual chart abstraction; however, that is tedious and time consuming. In contrast, coding data is less accurate and does not completely capture the patient’s experience. A computable phenotype would allow for improved accuracy in SMM research and a clearer picture of these rare but serious outcomes. Future directions include ensuring that SMM events are 1) coded correctly by billers, and 2) do not represent a previous diagnosis carried forward from an earlier hospital encounter. Creating universal computable phenotypes (definitions that consist of data elements and logical expressions such as AND, OR, and NOT) that can be interpreted and executed by data algorithms would also be helpful to identify SMM events. This would offer a more valid process than using the ICD 10 codes alone.

Benefits

Demonstrated benefits are those that have been observed and are verifiable.

Potential benefits are those logically expected with moderate to high confidence.

Clinical & medical benefits

Creation of computable phenotypes for SMM will allow for improved data capture to assess maternal morbidity rates and measure improvement.

demonstrated.

Software technologies

Standardized approaches for measuring SMM using both administrative (billing) data and clinical including electronic medical record (EMR) data, and reliable comorbidity indices for obstetric care that assist in risk stratication and risk adjustment at the patient and facility level.

potential.

Guidelines

There is a need for examination of racial and ethnic disparities that exist within SMM and the OB-CMI. Determining the underlying comorbidities of pregnant patients at Duke and UNC, and the cumulative risk these comorbidities pose by race and ethnicity, is essential to highlighting modifiable disparities in care. This is the rst step towards prospective application of this index in our inpatient units. By creating a multicenter dataset, researchers were able to determine where race and ethnicity disparities exist within the Duke/UNC cohort and create a Duke/UNC specific risk score for further review.

potential.

Health Equity & Disparities

Community & public health benefits

The creation of a multi-center dataset examining diseases present at the time of delivery and severe maternal morbidity events will help to reduce SMM events and even save lives at our institutions. This model of localized examination of disease states and multi-center collaboration can be applied to other centers to reduce disease burden.

potential.

Disease prevention & reduction

Evaluation of SMM by nationally recognized maternal levels of care will allow for evaluation for pairing patient risk factors with appropriate institution level.

potential.

Health care delivery

SMM is a potential measure of health care quality for pregnancy and delivery.

demonstrated.

Health care quality

Identification of leading regional causes of SMM may inform health education prevention resources for prenatal patients.

potential.

Health education resources

Policy & legislative benefits

SMM is a public health crisis in the United States. Defining the local impact of SMM at two large centers allows for regional public health and policy development.

potential.

Policies

This research has clinical, community, economic, and policy implications. The framework for these implications was derived from the Translational Science Benefits Model created by the Institute of Clinical & Translational Sciences at Washington University in St. Louis.⁶

Lessons Learned

Facilitators of Success

Pilot Award to Junior Researchers

The Pilot award between two institutions, as well as the funder’s willingness to grant funding to two early career investigators, contributed significantly to the success of the project.

Historical Collaboration

The primary investigators of this study were introduced by a mentor collaborating between the two institutions. This introduction was facilitated by a historic institutional collaboration between the Duke and UNC Obstetric and Obstetric Anesthesiology departments. Furthermore, the partnership established a strong team of co-investigators and trainees, allowing a variety of knowledge to influence the project independent of specific institutions.

Interinstitutional Collaboration

This Duke/UNC collaboration brings together a team with diverse expertise and allows for more complete data capture than possible with a single institution. Marie-Louise Meng, MD (Duke principal investigator) is an obstetric and cardiothoracic anesthesiologist with expertise in high-risk maternal care research. Johanna Quist-Nelson, MD (UNC principal investigator) is a maternal-fetal medicine physician with expertise in maternal safety and quality improvement. This collaboration recognized the need for partnership between obstetrics and anesthesiology to provide appropriate care to pregnant patients at risk of severe maternal morbidity. Both PIs had mentors who are national leaders in maternal-fetal medicine with expertise in areas such as risk assessment and data-informed strategies for risk-appropriate care, as well as in racial and ethnic disparities in maternal care. This collaboration improved the data capture as the two sites shared patients. The data query involved steps to identify patients who experience SMM on readmission regardless of site of delivery. Without this collaboration, data would have only captured SMM events at a patient’s delivery institution, resulting in less complete data for individual patients. Looking forward, this interinstitutional collaboration will also facilitate implementing strategies to decrease severe maternal morbidity and advance health equity.

Workforce Development

Funding this project directly supported the careers of the co-principal investigators, since both Dr. Meng and Dr. Quist-Nelson are early career investigators. The project also brought together investigators from varying backgrounds, as Dr. Meng is both an obstetric and a cardiac anesthesiologist and Dr. Quist-Nelson is a maternal-fetal medicine specialist.

Challenges and Learnings

Challenge: Data Use Agreement

One of the biggest challenges for this project was the creation of the data use agreement between the two institutions, which involved time and ad- ministrative hurdles. Both PIs are clinicians and junior investigators new to interinstitutional navigation, leading to administrative burden between balancing clinical obligations with researchers and an extended learning process on navigating multiple institutional systems. While the project received IRB approval, the process of the data use agreement was time-consuming and led to delays in the project. The team was able to request and receive a six-month extension to their one-year funding period to accommodate these delays. During the delay, the research team published an article focused on the need for comprehensive, high-quality datasets to improve SMM and racial disparities in obstetrical care and formed a research collaboration with epidemiologists at Harvard, leading to a funded RO1.

Challenge: Data Captured

Because the purpose of this project was to define maternal morbidity in the UNC and Duke populations, the team planned to capture high-level data (i.e. ICD codes). Retrospectively, team members wished that they captured more data points (i.e. disease lab values) to further define problems caused during the childbirth process. The research team plans to dive deeper into the data for their next project. They learned, for future projects, to expand their data set and the details within it.