There have been several news reports about various factors that contributed to deaths due to COVID-19. I discussed some of these factors in my last post. Based on your comments, I have investigated a few others. Deaths due to the COVID-19 vary significantly from state to state. New York and surrounding states have been hit hard, but many southern states have not had nearly as many reported deaths so far. Using statistical modeling of state-level data for deaths due to the novel coronavirus, I have found several factors that explain the state-level variation, and several others that do not.
Things that do NOT explain variation:
- Smoking
- Obesity
- Diabetes
- Age
- Income/wealth
- Race
Some of these are surprising to me, as several news reports have discussed these as contributing factors to COVID-19 deaths. The percent of obese people in state is actually negatively correlated with deaths (albeit weakly).
Things that do contribute:
- Population density (+)
- Percent of the population that are citizens (-)
- Sunshine (-)
- Number of nursing homes in a state that has reported an outbreak (+)
- High rise buildings (+)
- Miles of interstate (+)
The plus and minus signs indicate whether the variable is positively or negatively correlated with deaths due to COVID-19. For example, the more tightly packed people are in a state, the more likely the virus will be spread to many others.
The percent of the population that are citizens could be because many non-citizens work in the meat processing industry, which has had several widespread outbreaks at a number of plants.
Sunshine is the vitamin D connection. Higher levels of vitamin D are reported to help with immune system function.
The nursing home statistic is a sad one. It is not very good at predicting the future, but it explains many of the deaths. The New York Times recently reported that 35% of deaths due to COVID-19 have been nursing home residents or staff.
High rise buildings is an interesting one. The more people who live and/or work in a high-rise building, the greater the opportunity for the disease to spread, including through the ventilation system of these buildings.
More miles of interstate means a significant number of people traveling within the state or through the state, leading to more spread.
There is a significant amount of uncertainty, but these six factors explain 92% of the variation in state-level data on COVID deaths. See the graph below for a comparison of the estimated deaths per state vs. the actuals.
This is probably the most interesting explanation I have seen throughout this pandemic. Thanks for your investigation and sharing. Amazing!
Thanks!
Population density, interstate, and high rises definitely make sense. What is it about the nursing homes? Patient care and sanitation protocol lapses?
Large amounts of elderly people not in good health confined in close quarters. The ventilation system of these large homes is also likely spreading the virus rapidly once there is a case.
Could be sanitation protocol.
Formal Dutch figures show 5500 Covid-19 deaths, however, compared to the same months in other years more than 9000 extra deaths have been registered. In Russia the number of official Covid-19 deaths is very low (compared to the people infected), but many people died of peneumonia..
In other words, don’t use the official Covid-19 deaths but the excess number of deaths (compared to the regular mortality) and check what happens with state-level variation.
Thanks for the comment. That is a good point about the undercount. Do you have a source for excess death statistics?
Wonderful work. This goes a long way towards debunking/undercuting the political blame game. Much of this is what you would call ‘facts on the ground.’ There are a few issues that can become political discussion points, such as: 1) the percent that are citizens – this is not likely to be a decision variable; making folks citizens will not directly affect their housing, density and income; 2) the number of nursing homes, which is already a hot topic because of NY having forced admissions and the denials issued and retracted; 3) miles of Interstate, which will reinforce the desire for Reservation barricades. As to the demographics, part of their impact was perhaps explained by population density (race and income.) Other parts may not vary much by state (obesity, smoking, diabetes and age) or may likewise be explained by population density. Well done. Too bad the volume and intensity of rhetoric on the news media will drown this out. Not to mention (apophasis) that folks don’t want to believe it’s not fault of “the *other* party.” Especially the politicians, who have the microphone and the TV lens at their disposal. Shame on us for applauding and egging them on, or booing and holding up signs. Our reactions feed the blame game. Ignore them, they may realize that the appeal is gone and they may stop. We can try, anyway.
Thanks for the excellent analysis!
Comments are closed.