Researchers in Brazil Struggle to Get Solid COVID-19 Death Counts
Researchers in Brazil Struggle to Get Solid COVID-19 Death Counts

Researchers in Brazil Struggle to Get Solid COVID-19 Death Counts

After combing through data on public death notices in Minas Gerais state, scientists say the coronavirus death toll in the country is worse than reported.

Chris Baraniuk
Chris Baraniuk
Jun 4, 2020

ABOVE: Belo Horizonte, Minas Gerais, Brazil

Brazil is a major new epicenter of COVID-19—but the official figures don’t tell the whole story. That’s what researchers have concluded after combing through public death notice data from the second most-populous state in the country and comparing it to officially reported figures. 

A team of scientists examined data gleaned from public notary death notices—legal documents required for arranging a burial—in Minas Gerais state and published their results in a preprint on medRxiv May 23. They found that notaries recorded 201 COVID-19 deaths in March and April, versus 65 COVID-19 deaths reported in official statistics via a government system called The Notifiable Diseases Information System (SINAN). There is generally a lag between these data sources, but the team also found that notaries had recorded an unexpected surge of deaths due to other respiratory illnesses. The authors write that they suspect hundreds of these are in fact additional coronavirus deaths, likely attributed instead to conditions caused by COVID-19, such as pneumonia.

They also argue that Brazil is not testing for COVID-19 widely enough in order to understand the scale of the epidemic and to make decisions on policies such as social distancing. Despite this lack of data, lockdown measures are already being relaxed in Minas Gerais, says coauthor Tafarel Andrade de Souza, a postdoc in health science in the School of Medicine at the Federal University of Uberlândia.

I know for a fact that the excess exists and COVID is causing a ton more deaths than historically expected.

—Flávio Codeço Coelho, Fundação Getúlio Vargas

“We are facing the gradual opening of workplaces and public parks,” he writes in an email to The Scientist. “This worries us.”

He and his colleagues found that public notaries in Minas Gerais had recorded a very large number of deaths caused by severe respiratory symptoms, labelled generically as “SARS” in the notary dataset. Between 2017 and 2019, such deaths averaged around no more than 20 per week during March and April but this year they surged to more than 80 per week at times. There was also an increase of about 5–6 percent in the number of deaths related to pneumonia and “respiratory insufficiency” in the same period this year compared to previous years.

Brazil’s Ministry of Health did not respond to a request for comment.

The robustness of the SINAN system has been scrutinized by epidemiologists in the past. An analysis published in 2008 in Leprosy Review found that, while the SINAN software used by health professionals was functional, there were “serious weaknesses” in terms of, for example, staffing, which meant the quality and timeliness of data reporting were lower than expected.

More recently, in 2016, Flávio Codeço Coelho, a professor of mathematical epidemiology at Fundação Getúlio Vargas and colleagues criticized the functionality of SINAN in The Lancet. They pointed out that it only provides monthly data, which they argued is not good enough for accurate disease modelling. Because of this and other problems, SINAN, they said, risked “complete obsolescence.”

Coelho, who was not involved in the preprint, notes that, pre-pandemic, there was always a lag between the notary data and the figures in SINAN, which are confirmed by health officials and take longer to be updated. In other words, the SINAN data may eventually square more closely with the notary records. 

But notary records have their own problems. Coelho, says that notary documents are not always medically rigorous. “Sometimes what’s written as a cause of death for that certificate, it’s not necessarily a valid assessment from the point of view of medicine,” he says.

See “It’s So Hard to Know Who’s Dying of COVID-19—and When

Coelho adds that the challenges presented by the rising death toll during the pandemic and variation between data sources mean researchers can only obtain a broad impression of what is happening.

Felipe Iani, a bioinformatics graduate student at Ezequiel Dias Foundation, also stresses that SINAN data are gradually updated and might become more accurate over time. He adds that Brazil is undoubtedly facing a serious outbreak of COVID-19. 

To take another example of the discrepancies that exist in Brazil’s COVID-19 death count, the advocacy group Articulation of Indigenous Peoples of Brazil (APIB) recorded more than 125 deaths among indigenous groups by May 24 and a death rate double that of the wider population. Meanwhile, the Health Ministry’s Special Secretariat of Indigenous Health had only recorded 34 deaths by that date.

Coelho says that the preprint by de Souza and colleagues cannot conclusively reveal the true number of COVID-19 deaths in Minas Gerais but he adds that the general findings concur with his own analysis of deaths in Brazil in 2020 versus previous years. 

“I know for a fact that the excess exists and COVID is causing a ton more deaths than historically expected,” he says.

Some states including São Paulo and Rio de Janeiro have been hit especially hard, with more than 10,000 deaths already between them.

Observers outside Brazil are using what data are available to make projections as to how the country will fare in the coming weeks and months. The picture is a disturbing one. Researchers at the University of Washington’s Institute for Health Metrics and Evaluation recently projected that more than 125,000 COVID-19 deaths will have occurred in Brazil by August 4.

As of June 3, the officially reported death toll stands at 31,300.

Correction (June 4): We misidentified de Souza as a graduate student. He is a postdoc. The Scientist regrets the error.