Artificial intelligence helps prevent deaths, according to hospitals

Gradually, artificial intelligence advances in Brazilian hospitals. And, according to the institutions that adopt it, the technology helps prevent deaths.

In Brazil, hospitals that use it to monitor patients include Israelita Albert Einstein and Sírio-Libanês, as well as Rede D’Or.

Sírio-Libanês uses the tool integrated with the electronic medical records to estimate the length of stay. The algorithm collects information about previous admissions, the reason for the current one, and test results. The technology makes it possible to manage ICU resources, says Felipe Veiga, medical director of information technology in medical images at the institution.

Rede D’Or, owner of Hospital São Luiz and Nova Star in São Paulo, also uses artificial intelligence to estimate length of stay. When technology provides for a long stay of an intubated patient, for example, he receives early mobilization treatment, which consists of moving him or, if possible, helping him to get up from the bed.

“This reduces length of stay, mortality and, after hospital discharge, this patient has better cognition”, says Jorge Salluh, intensive care physician and researcher at the D’Or Institute for Research and Teaching.

In 2018, Einstein implemented the Care Monitoring Center, which tracks patient data and uses artificial intelligence to predict the worsening of cases and issue alerts to teams.

This nucleus monitors patients at the Einstein and also at the Hospital Municipal Vila Santa Catarina, in the south of São Paulo, which belongs to the public network and is managed by the group. According to doctor Leonardo Carvalho, from the Precision Medicine sector at Einstein, the network algorithm becomes more accurate when it is fed with data from people with different contexts.

Developing inclusive artificial intelligence models is a challenge. To ensure accuracy, the algorithms need to be “trained” with data representative of the population, says lawyer Christian Perrone, law coordinator at the Institute of Technology and Society (ITS).

Another issue is the risks of artificial intelligence tools in relation to information security, since these systems process a large volume of sensitive data.

The General Data Protection Law (LGPD) guarantees the right to privacy of personal information in physical or digital media and covers electronic medical records and artificial intelligence systems.

According to Perrone, the law offers protection mechanisms, but there is a need for more specifications. “Some processes are not yet fully regulated, such as using large datasets to train tools.”

Albert Einstein and Rede D’Or state that they employ, as a security measure, the anonymization of health data. All hospitals mentioned claim to be in compliance with the LGPD.

The Einstein system cost R$ 1.2 million. Claudia Laselva, director of the Morumbi hospital unit and the hospital’s care practices, says that technology reduces costs by improving patient safety.

“Since we implemented it in the operating room, more than 1,500 days ago, we have eliminated catastrophic and serious events related to anesthesia and reduced 30% of other adverse events”, says Laselva. A catastrophic event is one that leads to permanent damage or death, while a serious event generates important but temporary harm to the patient.

The use of artificial intelligence in hospitals is still limited in the public network. The initiatives exist, but, according to Marco Bego, executive director of the innovation and technology center at the USP Hospital das Clínicas (HC-USP), they are mostly in the research phase. The Center for Research and Solutions in Artificial Intelligence in Health, at USP, develops artificial intelligence tools and validates those developed externally for use at HC-USP.

Bego says that it is necessary to contemplate three conditions to implement the tool in the SUS: increasing access to health, service quality and sustainability of the public network. According to him, technology can guide decisions based on real data, making the investment worthwhile.

sepsis mortality

In July, researchers at Johns Hopkins University, in the US, published three studies showing how artificial intelligence reduced mortality from sepsis (generalized infection) by 18% among patients in five American hospitals.

The artificial intelligence developed at the university was called Trews (“targeted real-time early warning score”, whose free translation in Portuguese is early warning score in real time).

The system analyzed vital signs, medication use and test results. When parameters such as blood pressure and platelet count exceeded predetermined thresholds, the algorithm triggered a septic shock risk alert at a range of 11 to 94 hours in advance.