Behind those COVID-19 numbers

“My feet hurt,” says Denise Begaye, an X-ray technician with the Monument Valley Health Center, as she sits on a curb and takes a break from COVID-19 testing outside of the center in Oljato-Monument Valley, San Juan County on Thursday, April 16, 2020. The mobile testing team tested 581 people on Thursday. (Photo by Kristin Murphy, Deseret News via AP)

Jourdan Bennett-Begaye

‘Physicians lie awake at night worrying about numerators. Epidemiologists lie awake at night worrying about the denominators’

Jourdan Bennett-Begaye

Indian Country Today

The numbers are stark.

Navajo Nation health officials reported six new deaths Friday and 180 new cases of COVID-19. That adds up to 1,590 known coronavirus infections across the Navajo Nation resulting in 58 deaths.

That topline number places the Navajo Nation at 45 out of 55 states and territories, ahead of ten other jurisdictions.

But what do those rankings mean? What’s the story behind the data? The state and territory data from the Centers of Disease Control and Prevention includes three columns: Column A for the state or territory (which should include tribal nations, by the way); the number of confirmed cases; and the number of deaths. Again, the numbers are stark.

But if you insert a new tab, population, then the story grows even more significant. The Navajo Nation has an infection rate 10 times that of the Indian health system and ranks among the top three when compared to states. (Spreadsheet)

Welcome to the world of the denominator.

“Denominators are key in epidemiology and often difficult to measure,” said Jessica Atwell, infectious disease epidemiologist at the Center for American Indian Health at Johns Hopkins Bloomberg School of Public Health. “The underlying population structure is really important.”

Why does that matter? Remember infection rate and death rates? Health officials need population data to do that math. Those ratios, percentages or proportions are also influenced by a country’s capability to test citizens for COVID-19.

“We know right now, given the limited testing capabilities, that our denominators are wrong,” Atwell said. “We just don't know how wrong. We don't know that order of magnitude that they're incorrect.”

Let's just focus on the population data. An example is Italy and Germany. Italy had a larger population of older people with comorbidities, or chronic diseases.

“So they had more deaths proportionately right, than Germany did and part of that was because older people with comorbidities are more likely to die,” Atwell said. The larger the population of older people with chronic diseases, more of the population is going to die.

“But if you don't compare those things and you don't know, right?” Atwell said. “You just think it somehow is worse in one place versus another, but it has to do with the people.”''

When people compare 50 cases out of a population of 100 to 500 cases out of 10,000 people, some may think the 500 case is worse, Atwell said. “It’s not.”

We’ll use the positive confirmed cases we know, which is the numerator. Our tracker shows 2,045 cases confirmed and 76 deaths in the Indian health system as of April 24.

For the denominator, there are at least five options: the 2010 Census, American Community Survey, total tribal enrollment of all tribal nations, and the Indian Health Service user population or service population.

The Indian Health Service figures its “service population” at 2,562,290. That means the reported infection rate is 0.08 percent. But if you calculate based on the IHS user population of 1,662,834, then the rate is 0.12 percent.

To calculate the case fatality rate, the proportion of people who tested positive and died from the disease, the math works out to 3.72 percent in the Indian health system using Indian Country Today’s data. Epidemiologist Carl Heneghan at the University of Oxford told BBC the case fatality rate is “how many people doctors can be sure are killed by the infection.”

The World Health Organization estimates 2,311 cases per 1 million people and 110 per 1 million people.

Using the WHO model and the IHS service population means it is expected the Indian health system will eventually reach 5,921.5 cases and 282 deaths.

The numbers are different if the 2010 Census with the American Indian and Alaska Native population at 5.2 million is used with the WHO model. The Indian health system could expect 12,017.2 cases and 572 deaths.

But all of that changes if a different denominator is used. There is the 2010 Census, tribal enrollment figures, or even state and county level data.

Top 10 - COVID19 - April 24, 2020

New COVID-19 study in Indian Country

Four Indigenous researchers found in a new study that the rate of COVID-19 cases per 1,000 people on a reservation is more than four times higher than the United States as a whole.

Findings from the “American Indian Reservations and COVID-19: Correlates of Early Infection Rates in the Pandemic” study confirms that these positive COVID-19 cases were more likely to occur in tribal communities with a higher proportion of homes lacking indoor plumbing, prevalent household crowding, and households where only non-English languages are spoken. The study will be published in the Journal of Public Health Management and Practice.

Social demographer Desi Rodriguez-Lonebear, health services researcher Nicolas E. Barcelo, economist Randall Akee, and public health researcher Stephanie R. Carroll were concerned about the “massive spread” on the Navajo Nation and throughout the pueblos.

Rodriguez-Lonebear, Northern Cheyenne, said her reservation in Montana hasn’t had a case that she knows of but so many other reservations do.

“So we're kind of just like waiting ‘cause we know it's gonna come. It's like sitting ducks just kind of waiting for it to happen,” she said in a phone interview. She is frustrated by the lack of public health money for tribal communities to better respond to COVID-19.

The researchers from the University of California at Los Angeles and the University of Arizona were curious to see about any correlation between community and household characteristics and the rate of spread of COVID-19 on tribal lands.

“And so we were just like there has to be either somewhere they're talking about high rates of mortality among Blacks but we know there's going to be high rates of mortality amongst Indians as well,” Rodriguez-Lonebear said. “We have similar experiences with just institutional racism that has created a sicker population.”

Akee, Native Hawaiian, came across the Indian Country Today data that gave the researchers enough information to mine.

The team took the Indian Country Today data, that is publicly available, and merged it with the 2018 American Community Survey 5-year data. Akee said the survey had the population demographics they needed rather than a 10-year-old census. On April 10, Indian Country Today gathered 861 positive confirmed COVID-19 cases in 280 tribal communities within the Indian health system.

Researchers analyzed the correlation of the data in two different ways.

The first looked at 214 cases on 253 American Indian Reservations, excluding the Navajo Nation and Oklahoma tribes. The second time, they included the cases from the Navajo Nation and the 25 Oklahoma tribes. The Navajo Nation and 25 Oklahoma tribes had a total of 647 cases as of April 10. Researchers said the Navajo and Oklahoma tribes were “large, but important data outliers.”

“We wanted to exclude the two outliers to make sure they weren't responsible for all of the results,” Akee said. The results held up and when the researchers added back in the Navajo Nation and Oklahoma tribes “the results seem to intensify.”

As a public health researcher, Carroll, Ahtna from the Native Village of Kluti-Kaah, now looks to the future.

“What does this mean on the ground? And so there's some there's some more kind of short-term impacts, right? So what can we do right now to change things? So how do we change messaging so that we're messaging in language and in culturally-relevant ways and through the media that people actually receive the message on the ground in different tribal communities?” she said. Facebook and radio are better for different communities to spread the physical distancing message, she said. “The community knows best what those mechanisms are.”

There are complexities to water access, too. “In some communities, you might choose not to have plumbed water right to your house. And so how do we assure access to water in a crisis situation?”

Access to water has been heavily covered in mainstream media and the percentage of lack of indoor plumbing has fluctuated for these news reports. Similar reports and presumptions have been made about overcrowded housing and language barriers.

“So, as always, we can know from our experiences in the past, but whether it plays out in the current situation is another story and so likely also there are other things that are associated that haven't been looked at yet,” Carroll said. Stay at home orders, as part of the tribe’s public health codes, need to be considered as well as infection rates.

“I'm interested in the [tribal public health codes] and together kind of build this narrative, which is another piece of the puzzle and so being able to build up from there,” Carroll said. Tribes have public health authority and can implement public health-related guidelines such as quarantine or stay-at-home orders. 

Carroll said having these tribal public health codes could give tribes "the opportunity to act quicker and more strongly around some of these stay at home orders or around the relationships to work with states around data and infections."

Use what we know

“So you start with something you know, right?” Carroll said. “Or you think you know, you test it, and then you can, you can keep adding into the model.”

As in any study, there are limitations.

One drawback is the small data set collected by Indian Country Today. Data is gathered from tribes by phone calls and submissions by emails and Google form, as well as press releases.

Another limitation is the lack of data or methods to count urban Indians, Carroll said. “Without cases and a real ability to link those cases, especially at the time that we pulled the data to urban areas, you can't do that same type of analysis.”

There are multiple factors to consider when counting urban Indians.

The 2010 Census reported that approximately 70 percent of American Indians and Alaska Natives lived in urban areas. Native people are already undercounted in the census, Rodriguez-Lonebear said.

“You don't have the data when nobody has the data to then be able to include urban Indians in these types of research, which is so vital,” she said. “You end up relying on county-level data, state-level data. And you rely on these county and state agencies, vital statistics to collect information about American Indians and Alaskan Natives in such a way, that is complete. But we know that doesn't happen.”

This is one more reason why the census matters.

As more data becomes available in Indian Country, the public, tribal leaders, health officials and media will have to consider one thing: the source of the population data.

Part of the data talk during COVID-19 includes the population data used to calculate infection rates.

In Indian Country, or mainstream media who doesn’t understand the complexity of Indian Country, the math can be skewed according to the population data they choose to use.

Each set of population data has flaws.

The 2010 Census can be used in which Native people are undercounted and is only done every 10 years. This counts every person in the country once and grabs their race or ethnicity and gender.

The American Community Survey gathers data every year from a smaller population throughout the county. The survey collects specifics of the population like jobs, education, military status, housing, and more.

For Indian Country, there’s the number of enrolled tribal citizens which there is no central database for tribes to submit this. Each tribe keeps track of their enrollment. There’s also the consideration that many non-Natives live on tribal lands or reservations. Some tribes do not allow double enrollment.

The last population data to consider, in this public health crisis, is what numbers the Indian Health Service uses. It calculates both service population and user population.

“The IHS service population is based on the CDC National Center for Health Statistics Bridged Race File, and consists of American Indians and Alaska Natives who reside in geographic areas in which IHS has responsibilities ("on or near" reservations) and is comprised of approximately 58 percent of all AI/AN residing in the U.S.,” wrote IHS in an email. “User population is the number of IHS patients seen at least once in the last three years at an IHS or tribal facility and that live in a service delivery area.”

In 2019, the service population was 2,562,290 and the user population was 1,662,834, IHS said.

Why else do denominators matter?

“The different denominators help us understand different aspects of the outbreak. But denominators are incredibly difficult to measure accurately,” Atwell said.

“Are you trying to look at how bad it is in a given location? Are you trying to look at how many like hospital beds you need? So if you're talking about hospital beds, and the numbers are important, but if you're thinking about severity, then the proportions are important,” Atwell said.

If the census is used, Atwell asks, “how confident are we in the magnitude of that denominator?”

On April 7, Zia Pueblo reported 31 confirmed positive cases with a population of 900 people living in their community. Its infection rate was 3.4 percent. San Felipe Pueblo had 52 cases with a population of 2,200, giving it an infection rate of 2.36 percent.

New York City has a population of 3.2 million with 74,601 cases at the time, or a 2.33 percent infection rate.

Thus both pueblos had a higher infection rate than New York City on April 7.

“This is reminding me of something I heard Tom Frieden give a lecture and years ago, the former director of the CDC,” Atwell recalled. “‘How can you tell the difference between a physician and an epidemiologist?’ He says, ‘Physicians lie awake at night worrying about numerators. Epidemiologists lie awake at night worrying about the denominators.’” 

Indian Country Today - small phone logo

This story has been updated to show the “American Indian Reservations and COVID-19” study found cases were more likely in non-English-speaking households, not English-only households.

Mark Trahant contributed to this report. 

Jourdan Bennett-Begaye, Diné, is the Washington editor for Indian Country Today based in Washington, D.C. Follow her on Twitter: @jourdanbb. Email: jbennett-begaye@indiancountrytoday.com

Indian Country Today LLC is a nonprofit, public media enterprise. Reader support is critical. We do not charge for subscriptions and tribal media (or any media, for that matter) can use our content for free. Our goal is public service. Please join our cause and support independent journalism today. We have an audacious plan for 2020 and your donation will help us make it so. #MyICT #MyVoiceMyICT #SocksAlot

Comments (1)
No. 1-1
Wmat
Wmat

Reservations are more comparable to counties instead of whole states. It's very easy to use STATA and find counties with similar infection rates and similar population sizes and similar rates per 100,000 when compared to a given rez. For instance no one in the nation has compared Minnehaha County to the Navajo Nation but they have very similar populations, numbers of cases, numbers of cases per hundred thousand, and similar dates of first infection. However they vary immensely in regards to number of deaths and ethnicity. Write about that Indian Country but give a shout out to me for doing the STATA coding and filtering.


News

FEATURED
COMMUNITY