A family in Brazil is blaming an AI-powered medical system for the untimely death of 32-year-old Rebeca Cardoso Tenente Molina.
Brazilian news outlet MG1 reported on Molina’s death late last week. Her family alleges that a state-run AI system used to manage hospital bed allocation incorrectly assessed her condition and waited too long to transfer Molina to an intensive care unit. Molina died just hours after reaching the ICU.
“What we saw was that doctors lost the autonomy to decide if a patient is very seriously ill,” Sâmela Cardoso Tenente Furtado, a lawyer and Molina’s twin sister, told MG1.
Too low a score
According to MG1, Molina was first hospitalized on June 2 with what was believed to be gallstones. She ended up at a hospital in São João Nepomuceno, a municipality in the Brazilian state of Minas Gerais. Her condition quickly deteriorated, and Molina reportedly requested a transfer to an ICU.
Last month, Minas Gerais changed over to a new management system—called Core-MG—in its state hospitals, which incorporates AI. And the family claims this system wrongly downgraded the severity of Molina’s health problems, delaying the care she needed. At one point, they even went to court to try compelling a speedier transfer. Due to this downgrade, the family argues, she had to wait five days until she was transferred to a hospital ICU in another municipality 186 miles (300 kilometers) away, where she soon died.
Molina’s cause of death is currently listed as septic shock, but doctors are still investigating whether other conditions, such as botulism, may have played a role, according to the family.
The state’s response
The State Health Department of Minas Gerais told MG1 that Core-MG has not fundamentally changed the criteria for managing someone’s care or searching for vacant hospital beds. The department further claimed that Molina was immediately registered into the system and that the choice of allocated hospital beds isn’t only affected by geographic proximity but also by the availability of beds according to a patient’s clinical needs.
The family, however, argues that Core-MG failed to accurately assess Molina’s health, even as worsening test results came in, and that it did a poorer job than trained medical professionals would have in the same situation.
“She would have been a 10, and the system only accepted her as a 6.8,” Furtado told MG1. “She wasn’t just a number or a protocol within the system. She had a family, she had dreams, and a whole life ahead of her.”
This isn’t the only instance of people blaming AI for worse medical care. While some studies have suggested that AI systems might be better than doctors at diagnosing some health issues, others have suggested AI can reinforce existing biases within health care. Other recent research has also indicated that popular consumer-facing systems like ChatGPT Health are prone to underestimating medical emergencies.
