By Jonathan Snowden, Ph.D. and Ellen Tilden, Ph.D., CNM
Last October, on my flight back to Seattle from the Lamaze International Advocacy Summit in Washington DC, I was seated next to a lovely woman. As oft is the case with strangers seated together, we got to talking and it turns out that she was Ellen Tilden, a Ph.D. researcher, and Certified Nurse Midwife in Portland, OR, and we had several colleagues in common. Ellen shared some exciting research that she had been collaborating on with Jonathan Snowden, another Portland researcher, and it sounded simply fascinating. (Jonathan Snowden had been interviewed about home birth on Science & Sensibility before.) Ellen and Jonathan's research has to do with the ability to conduct effective research on pregnant and laboring people with the current best practice of randomized control trials. They hypothesized that a more effective and accurate method of antepartum and intrapartum research might utilize a method called causal inference. Casual inference was a new term for me and I was fascinated to learn more. Ellen shared the three papers that they wrote on this topic (see below). I am grateful that both she and Jonathan accepted my request to help childbirth educators and others understand how the causal inference methodology might be a new and appropriate approach to conducting research on physiological childbirth. Have you ever wondered if the current methods of research are the most appropriate for use on pregnant and laboring people? Could there be a better way? Please read this introduction to causal inference and let me know what you think in the comment section below. - Sharon Muza, Community Manager, Science & Sensibility.
Are current research methods best suited to study birth?
Randomized controlled trials boast many strengths for generating strong science. This approach is often considered the “gold standard” of health research. However, it may be unfeasible or unethical to use randomized controlled trials for certain kinds of research regarding people during pregnancy and labor.
Problems can arise for several reasons, including:
Many people have strong preferences about what happens to them during pregnancy or labor and so are not willing to be assigned to a specific kind of care for a study (aka- randomized). Evidence of this concern can be found in many experimental studies of healthy childbearing processes that include smaller samples sizes. For example, a study randomizing women to home vs. hospital delivery included a final sample size of 11 (Dowswell, Thornton, Hewison, Lilford, Raisler, MacFarlane, ... & Settatree, 1996).
1. As a result, the people who are studied in randomized controlled trials tend to be very different from most people. This is a recognized problem in all fields of health research that use randomized controlled trials. It means that sometimes the results of the “best” studies do not apply to all people. This is a particular concern given that the people who participate in research tend to be healthier than the average person, and these people are less likely to be people of lower socioeconomic status and/or people of color.
2. Even if people are willing to be randomized to a specific kind of care, for example receiving doula care during labor vs. not receiving doula care during labor, they may find that once they are in labor, they are not able or willing to remain in the study. This can occur especially frequently during active labor among those who do not wish to use an epidural as the intensity of labor can require an individual’s full attention on what their body needs to birth. Midwives, doulas, nurses, and childbirth educators often describe this as ‘labor brain’. And let’s not forget about hormones. When we ask a laboring person to engage with complex information during labor - oxytocin can decrease, catecholamines rise, and endorphins diminish. For these reasons, asking a laboring person to maintain certain kinds of research protocols might deeply impact labor and thus potentially change the phenomenon we wish to study.
These two factors mean that those who are able to complete a study may be different in important ways from most childbearing people- this makes it difficult to confidently assume those study findings are relevant to all childbearing people.
Another scientific challenge is that because most healthy childbearing people will labor and birth with excellent outcomes under all circumstances, poor outcomes are very rare. As a result, scientific teams need to include very large numbers of people in order to meaningfully study what maternity care patterns lead to poor outcomes. If it is very difficult to convince childbearing people to both participate in a study and also very difficult for these individuals to complete the study, having a large enough sample size to analyze associations between maternity care and bad outcomes is a difficult goal to meet.
Could causal inference be a better research choice?
These challenges have led us to deeply consider approaches beyond randomized controlled trials for studying childbearing people, resulting in a three-part series recently published in the Journal of Midwifery and Women’s Health. This paper series reviews a scientific framework, causal inference, that encompasses several alternative ways of planning and conducting science. Causal inference is the process of determining that a cause led to an effect. Causal inference is multi-disciplinary and is shaped by philosophy, statistics, epidemiology, economics, and computer science. We think that for answering certain questions, especially questions related to normal pregnancy, labor, and birth processes, the causal inference approach may be more optimal than randomized controlled trials. These papers also review four methods within the causal inference framework and describe concepts regarding how to conduct this kind of science.
While the application of these concepts and methods are just emerging for the study of healthy childbearing processes, causal inference approaches have been used to importantly advance other areas of science facing similar challenges. For example, causal inference has been used to study how features of residential neighborhoods may affect health. Just as for certain scientific questions about childbearing it would be unethical or unfeasible to randomize women to different arms of an experiment, it would generally be unfeasible and unethical to randomize families to reside in a specific neighborhood. This makes it difficult to understand how the features of a neighborhood influence the health outcomes of its residents. A study by Sampson and colleagues used causal inference to study the effect of disadvantaged neighborhoods on children’s verbal cognitive ability (Sampson, Sharkey, & Raudenbush, 2008). Because they couldn’t randomize families to live in different neighborhoods, they applied a casual inference approach called inverse probability of treatment weighting. This study showed that living in a severely disadvantaged neighborhood had a negative causal impact on the verbal cognitive ability of children. They demonstrated that growing up in a disadvantaged neighborhood had a verbal cognitive effect equivalent to missing one year or more of school.
An invite to researchers to consider applying the principles of causal inference to research needs
Our focus is particularly on science that uses pre-existing databases of routinely collected healthcare information to understand labor and birth processes and outcomes. These “big data” approaches have become more common but are still relatively rare in the science of childbirth. We aim to address the challenges to understanding and adopting these approaches: the daunting terminology and the highly specialized methods and statistics involved. It is our hope that these papers will help make causal inference clear for anyone who is interested, including perinatal and midwifery scientists and those who are helping the individuals and families they are working with interpret science regarding childbearing processes. This link connects to all three papers.
Dowswell, T., Thornton, J. G., Hewison, J., Lilford, R. J., Raisler, J., MacFarlane, A., ... & Settatree, R. S. (1996). Should there be a trial of home versus hospital delivery in the United Kingdom?. BMJ: British Medical Journal, 312(7033), 753.
Sampson, R. J., Sharkey, P., & Raudenbush, S. W. (2008). Durable effects of concentrated disadvantage on verbal ability among African-American children. Proceedings of the National Academy of Sciences, 105(3), 845-852.
Snowden, J. M., & Tilden, E. L. (2018). Further applications of advanced methods to infer causes in the study of physiologic childbirth. Journal of midwifery & women's health.
Snowden, J. M., Tilden, E. L., & Odden, M. C. (2018). Formulating and Answering High‐Impact Causal Questions in Physiologic Childbirth Science: Concepts and Assumptions. Journal of midwifery & women's health.
Tilden, E. L., & Snowden, J. M. (2018). The Causal Inference Framework: A Primer on Concepts and Methods for Improving the Study of Well‐Woman Childbearing Processes. Journal of midwifery & women's health.
About Jonathan Snowden
Jonathan Snowden, Ph.D., is an Assistant Professor in the new joint School of Public Health at Oregon Health & Science University/Portland State University, and in the Department of Obstetrics & Gynecology at OHSU.
Dr. Snowden is an epidemiologist whose research focuses on the content areas of perinatal health, healthy pregnancy and birth, and sexual/reproductive health disparities. Dr. Snowden is a methodologist whose cross-cutting focus is improving causal inference from observational data. He implements analytical methods from a variety of disciplines to devise innovative approaches for answering high-impact questions in public health, medicine, and health policy. He conducts research using large databases with secondary data (i.e., not designed for the research question at hand), and focuses on the optimal methods to capture, clean, manage, and analyze such datasets for extracting valid scientific information. His overarching research goal is to generate actionable evidence to improve population health and to close disparities, while also respecting the limitations of the empirical data.
In his trans-disciplinary research program, Dr. Snowden collaborates with clinicians from fields including obstetrics-gynecology, nurse-midwifery, and pediatrics, as well as population researchers from a variety of fields. Drawing on expertise from a variety of fields informs Dr. Snowden’s substantive research focus on maternal-child health, and provides multiple professional perspectives for preventing harm and promoting optimal health in women and infants. The broad goals of his research are: (1) to enable a safe delivery for every mother and baby (2) to close disparities and improve the reproductive health of racial minorities and sexual minorities.
About Ellen Tilden
Ellen Tilden, Ph.D., CNM, is an Assistant Professor in the School of Nursing Department of Nurse-Midwifery and School of Medicine Department of Obstetrics and Gynecology at Oregon Health and Science University (OHSU).
Her midwifery training began with an apprenticeship within a home and birth center maternity care team in Berlin, Germany and she subsequently received her nursing and nurse-midwifery training at the University of California, San Francisco, graduating in 2000. Since this time she has been practicing full-scope nurse-midwifery in a variety of settings.
Ellen completed her PhD in 2015. Her postdoctoral research and training is currently supported by a women’s health BIRCWH career development award from the National Institutes of Health Office of Research on Women’s Health and National Institutes of Child Health and Development.
Ellen is a health services researcher focused on healthcare systems factors that impact obstetric procedure use, particularly modifiable drivers of cesarean delivery. She has published her research in the Journal of Midwifery and Women’s Health, Birth, the American Journal of Obstetrics and Gynecology, and the New England Journal of Medicine. Her research approach is inter-disciplinary and she works closely with perinatologists and epidemiologists, employing tools from economics, causal inference, and other disciplines.
Her overarching research goal is to define risk-appropriate care for healthy women and their children in the U.S.