Examining Maternal Mortality Rates in the US
It is widely known that the United States has abysmal rates of maternal mortality compared to other Western countries, especially amongst Black women. I decided to use data on various determinants of health in order to examine which factors are most predictive of high state maternal mortality rates. This article will explore variables, models, and graphs, and examine how those determinants contribute to a specific health outcome, that being maternal mortality rates.
Before I begin, a brief definition of determinants of health from the Office of Disease Prevention and and Health Promotion:
The range of personal, social, economic, and environmental factors that influence health status. It is the interrelationships among these factors that determine individual and population health.
Pictured below, an overview of state-by-state maternal mortality rates from 2021 (Vermont, New Hampshire, and Delaware did not have data available). Using geopandas, I plotted the following:
Rates in the past year were highest in the South, Midwest, and New Jersey. The average maternal mortality rate in the data I used was 22.05 per 100,000 births.
Determinants of Health
I chose 19 predictors as determinants of maternal mortality rates. There are an incredible amount of social and political determinants which effect maternal mortality. Health disparities, costs, and outcomes are effects of these social and political realities, which are shaped by the policies and decisions of those in power. Based on data available to me from 2019, provided by the Kaiser Family Foundation’s State Reports, the following are the 19 determinants of maternal mortality I analyzed, with their descriptions and relationships to social and political determinants of heath.
Health Insurance Coverage
These values provide each state’s proportion of the population covered by Medicare, Medicaid, and those who are uninsured. Health insurance coverage is an essential component of health outcomes, as those without insurance are less likely to seek out and receive care, increasing their chance of health risks. Those on Medicare and Medicaid versus private insurance can be limited in their choices of doctors, facilities, and procedures. Affordable health insurance provided by the government to low income individuals is an aspect of healthcare that fluctuates based on political power, and is determined by economic status.
Medicaid Eligibility Limits for Pregnant Women
This value provides the income eligibility limit as a percent of the Federal Poverty Level (FPL). Pregnant women have a higher income cutoff to receive Medicaid than the general population. The higher this rate is per state, the more healthcare coverage pregnant women will be eligible to receive. The mean eligibility limit is 203% of the FPL. This percent is again directly tied to income status and varies from state to state.
Abortion Rate
The number of reported legal abortions in women ages 15–44. The graph to the left plots a best-fit line between abortion rate and maternal mortality rate, and we can see a slightly negative correlation. The impact of legal abortions on maternal mortality has been researched, with some studies finding that access to safe abortions decreases maternal mortality rates.
Total Health Care Expenditures
This gives the spending, in millions, of all publicly and privately funded healthcare services per state. Of course, this will vary from state to state based on size and demand. Expenditures and budgets for public institutions are tied to policy and power.
Hospital Expenses per Day
This value provides an estimate of average expenses incurred in inpatient and outpatient care. The high costs of staying in a hospital can cause barriers to those unable to afford the care, who therefore can’t receive medical treatment. Those who have just given birth may leave earlier than necessary in order to avoid paying more, which could increase risk of maternal mortality. Indeed, we can see a negative relationship between hospital costs and maternal mortality rates, and the two variables have a moderate correlation at around -0.36.
Lack of Doctor
This value provides the proportion of adults who reported not having a personal doctor. This is a function of many social determinants, including income status, access to care, and availability of medical professionals.
Racial Demographics
The demographic proportion for racial minorities. Those included in the data are percentages by state of people who identify as Black, Hispanic, Asian, American Indian/Alaska Native, or Multiple Races. Race is an essential social determinant of health due to systemic racism in healthcare which causes many tangible downstream disparities. For example, in terms of maternal mortality, doctors are less likely to be responsive to reported pain of Black women. Racial proportions by state are shown in the heat maps below:
Poverty Rate
The percent of adults in the state falling below the FPL. Poverty and class in general are extremely important social determinants of health, effecting other determinants such as housing conditions, educational attainment, smoking rates, and many more. Seen on the left, there is a positive relationship between poverty rate and maternal mortality rate.
Median Annual Household Income
Another indicator of class distribution, median household income ties together with financial status, as well as costs of living which vary greatly between states.
State Political Parties
At the root of political determinants: who has power and what policy and resources those in power choose to use or ignore. In the US, Democrats and Republicans have opposing views on the role of government in healthcare, policy to reduce inequities, and even simple acknowledgments of the existence of inequities. The data provides state-level information about which party is in charge in the governor’s office, and who has majorities in the state house and senate. The mean maternal mortality rate per 100,000 births in states with Republican governors is 25.17 versus 18.65 in states with Democratic governors.
Unemployment Rate
Unemployment leads to economic instability, which is a clear determinant of health. Additionally, lack of employment may leave people without employer-based insurance, and thus without a means of paying for healthcare.
Plots with best-fit lines between each predictor variable and maternal mortality rate:
Graph showing correlations between each pair of variables:
The determinants which had the highest absolute correlations (≥ .4 or ≤ -.4) with maternal mortality are the Black population, Median Annual Household Income, and State Senate and Assembly Majority Political Affiliations.
Building a Model
I created a multiple linear regression model with the above predictor variables. The model has an R² score of 0.53, meaning that 53% of the outcomes of maternal mortality rate can be explained by the predictors.
This graph shows a heat map of the predicted maternal mortality rates from my model, as compared to the actual shown at the top of the article.
Additionally, I used the umap package in order to reduce the dimensionality of the model. In this way, I could reduce the predictor determinants from 19 variables to 2, in order to better assess their grouping.
In this graph, the size and color of the dots represent differing maternal mortality rates. There does seem to be a general grouping of states with higher rates on the left side of the graph, however it is not very distinct. This is reflective of the only moderate accuracy of the model.
Takeaways
An issue such as maternal mortality rate doesn’t exist in a vacuum; it is the result of many different determinants working simultaneously. It is easy to think of maternal mortality as merely a factor of racism, inadequate maternal care, or lack of resources. In fact, it is all of those factors, and more. Through the data analysis I have performed, the relation of many different variables to maternal mortality rates have been demonstrated. Though the model was not extremely accurate, I think that can be attributed to the fact that many of my variables were tied to the same broad social determinant: income. Given data on other determinants, such as distance to medical facilities and education level, I believe I would have been able to create a more accurate model.
Of course, people’s lives should not be condensed to mere statistics, graphs, and models. These tools do, however, allow us to clearly see the effects of inequities through determinants of health. Just as no one variable has a 100% correlation with maternal mortality rates, there is no singular solution to reducing those deaths. It will take comprehensive legislation, policy and training of medical professionals in order to reduce maternal mortality in the US.
Thanks for reading! This project was created for GOVT 3251: Health Equity, Politics, and Policy at Cornell University, taught by Professors Jamila Michener and Isabel Perera.
Code and data can be found at https://github.com/esalamon17/Maternal-Mortality-Data-Analysis.
