Factors that determine COVID-19 mortality

By Dr. Liji Thomas, MDJun 14 2020

A potentially important new study by MIT researchers and published on the preprint server medRxiv* in June 2020 shows that COVID-19 death rates in U.S. states correlate with several factors like commuting to work, temperatures, and geographical location, but not with ICU beds, obesity or poverty. This work, if validated, could help identify the actual contributing factors that shape mortality rates across the USA.

The current paper focuses on correlations between the COVID-19 death rates per 1,000 population at the county level with four important variables: socioeconomic status (SES), commuting modes, health variables, and climate and pollution patterns.

What does and does not correlate with COVID-19 death rates. Image Credit: Photo Spirit

This news article was a review of a preliminary scientific report that had not undergone peer-review at the time of publication. Since its initial publication, the scientific report has now been peer reviewed and accepted for publication in a Scientific Journal. Links to the preliminary and peer-reviewed reports are available in the Sources section at the bottom of this article. View Sources

The Factors Examined

SES includes the proportion of African-Americans and Hispanics, as well as whites and other ethnic groups; poverty; age over 65; living in self-owned houses; and the home value in financial terms.

Commuting mode includes the proportion of workers who drive, use public transport, use bikes, walk, work from home, or do not commute at all. Health variables include the percentage of diabetics, smokers, and obese individuals, as well as those who lack health insurance, and the number of ICU beds per capita.

Climate and weather variables include average air pollution levels, average summer and winter temperatures.

Correlation is Not Causation

At the outset, the researchers caution, “It is important to understand that this research, and other observational analyses like it, only identify correlations: these relationships are not necessarily causal. However, these correlations may help policymakers identify variables that may potentially be causally related to COVID-19 death rates and adopt appropriate policies after understanding the causal relationship.”

Potential sources of error include differences in test availability and criteria for testing, which would affect the ability to record all deaths. However, the researchers have used an analytical model that will allow relationships to be judged as such, and not as the result of other confounding factors for which adjustments have already been made.

The study period was from April 4, 2020, to May 27, 2020, but the researchers omitted the five counties of New York City from the principal analysis.

Using State Fixed Effects for Inter-State Comparison

The researchers also used “state fixed effects”, a type of dummy variables that allow the computation of baseline death rates for each state, by adjusting for variables common to all counties in the same state. This means that the death rates can be compared across states irrespective of the other variables, giving an idea of how any state’s mortality rates compare with those of others.

The use of state fixed effects also allows for correlations to be measured across counties in the same state, but without their use, correlations are measured both within a state and across states. The researchers explain that these differences are important in that if a correlation is found across states but not within a state, this could mean that the first type of correlation is because of specific state-related factors which have not made their way into the model.

Do African Americans Have a Higher Death Rate?

First of all, the researchers found that without using state fixed effects, the death rates were higher when the proportion of African-Americans was higher; this effect became insignificant but remained when the state fixed effects were in place. For instance, a state like Louisiana, which has a higher proportion of African Americans, has a higher death rate than one like Tennesse, which does not. However, within Louisiana itself, counties with more African Americans and a lower proportion of them do not show this disparity.

By way of instance, they quote a death rate 1.262 higher in a county with all African-Americans, compared to one which has zero. In comparison, the average death rate at the county level is 0.119.

They show that death rates between two counties where one has zero African-Americans, and one at 87% (the highest observed) varies by almost tenfold, to 1.10.

If the state fixed effects are now considered, the death rates are tripled, but this increase is not significant anymore. Nonetheless, the researchers say they “encourage policymakers and researchers to investigate the causal relationship between death rates and the share of African American residents.”

Other Factors Correlated with Higher Death Rates

By controlling for obvious confounding factors, the correlations allow a narrower set of factors to be focused on to identify the real reasons for such links.

Diabetes is also associated with a higher death rate but again becomes insignificant when state fixed effects are in place. An older population is significant, if marginally associated with higher death rates, as are higher home values, hotter climates, and lower winter temperatures.

Public Transit and Death Rates

The relationship between the death rate and the use of public transit is significant and strong. About 0-0.26% of people across the sample use public transit, and if the use of public transit increases by an over 20-percentage point increase, the death rate goes up by almost 1.0, which is almost ten times the mean death rate of 0.119 across all counties.

The researchers point out two facts. One is that public transit is linked to the highest death rate, but on the other hand, all commute modes other than biking are related to higher death rates than telecommuting. Given this, the researchers speculate that this increase is due to the type of transit itself and to the type of job such workers do. This latter reason also applies to the people who walk or drive and may also explain the increased death rate in their case – both these classes may have jobs that involve more interaction with others.

The recommendation would, therefore, be to increase the safety of public transit and ensure better social distancing at work.

Why are the death rates higher for those who don’t commute at all? The researchers attribute this to the large proportion of children and unemployed in this segment, which points to possible higher exposure as well as the toll of unemployment.

What’s New, and How it Matters

The new findings are a lack of correlation between pollution, obesity, ICU bed numbers, and poverty, with death rates. Moreover, the comparison of multivariate-controlled death rates show they are highest in the Northeastern states, and in Michigan, Louisiana, Iowa, Indiana, and Colorado, while being lowest in California.

The study offers a preliminary overview of some important correlations. It may add to the research-based identification of variables that possibly play a causal role in COVID-19 death rates, helping to shape policies appropriately and thus reduce the impact of the outbreak.

This news article was a review of a preliminary scientific report that had not undergone peer-review at the time of publication. Since its initial publication, the scientific report has now been peer reviewed and accepted for publication in a Scientific Journal. Links to the preliminary and peer-reviewed reports are available in the Sources section at the bottom of this article. View Sources