EHRs Can Be Used to Predict Sepsis Risk and Mortality

UC Davis researchers have found that routine information—blood pressure, respiratory rate, temperature, and white blood cell count— from the EHRs of hospitalized patients can be used to predict the early stages of sepsis, a leading cause of death and hospitalization in the United States. They also determined that just three measures—lactate level, blood pressure, and respiratory rate—can pinpoint the likelihood that a patient will die from the disease.

"EHRs have become essential resources for providing relevant information on patients' medical histories and improving the quality of care," says study coauthor Tim Albertson, chair of UC Davis Department of Internal Medicine. "We have shown that they can also be powerful resources for identifying best practices in medicine and reducing patient mortality."

Sepsis is an immune system response to infection that can damage organs and cause permanent physical and mental disabilities. It is associated with increased blood levels of lactate, an acid produced when organs receive too little oxygen.

Patients are rarely screened for blood lactate levels, because sepsis is very hard to distinguish in its early stages. The blood test also lacks specificity, as many patients with elevated lactate do not have sepsis.

While early treatment with broad-spectrum antibiotics and intravenous fluids is associated with better outcomes for those with sepsis, the potential harm from those treatments for low-risk patients far outweighs the benefits.

"Finding a precise and quick way to determine which patients are at high risk of developing the disease is critically important," says study coauthor Hien Nguyen, an associate professor of internal medicine and medical director of EHRs at UC Davis. "We wanted to see if EHRs could provide the foundation for knowing when aggressive diagnosis and treatment are needed and when they can be avoided."

In conducting their investigation, the researchers analyzed data from the EHRs of 741 patients with sepsis at UC Davis Medical Center during 2010. They found that vital signs combined with serum white blood cell count—measures routinely taken for hospitalized patients—could accurately predict high lactate levels and sepsis. They also found that lactate level, blood pressure, and respiratory rate could determine a patient's risk of death from sepsis.

The research team is now working on a specific sepsis-risk algorithm that can be automatically calculated in the EHR.

"The electronic health record has been a transformative development for the delivery of health care with enormous potential," says Ilias Tagkopoulos, an assistant professor of computer science at UC Davis and senior author of the study. "Rather than using a 'gut-level' approach in an uncertain situation, physicians can instead use a decision-making tool that 'learns' from patient histories to identify health status and probable outcomes. Another benefit of the sepsis predictor is that it is based on routine measures, so it can be used anywhere—on the battlefield or in a rural hospital in a third-world country."

The study, titled "From Vital Signs to Clinical Outcomes for Patients with Sepsis: A Machine Learning Basis for a Clinical Decision Support System," was published in the Journal of the American Medical Informatics Association.

Source: UC Davis Health System