Sepsis: it s a little discussed condition that packs a deadly punch. In fact more Americans die annually from sepsis than from AIDS, breast cancer, and prostate cancer combined! What s worse is that unlike cancers, sepsis can kill within a matter of hours after its onset.
Sepsis is a condition in which there is a systemic overreaction by the immune system to an infectious agent, i.e. a bacteria, virus, etc, in the blood. The response is often so severe that it can progress into septic shock, which is a life threatening condition characterized by severely low blood pressure.
The speed at which sepsis and septic shock progress leaves doctors with little room or time for error when a patient starts exhibiting symptoms. Currently, the protocol for determining sepsis and septic shock are to monitor symptoms like risk factors (such as age), extreme temperatures, irregular heart rates, and blood tests such as those that detect the presence of an organism in the blood. However, not all of these are diagnostic. In particular, the blood tests for organism presence are often inconclusive because bacteria do not grow well in cultures from patients that are already on antibiotics.
However, a new study in Science Translational Medicine describes a computer algorithm that may turn the tides against these conditions. Researchers working at Johns Hopkins have developed the algorithm based on 27 predictors of septic shock and are now reporting a high level of success. In reviewing patient medical records from over 16,000 files from patients who were admitted to intensive care units at Beth Israel Deaconess Medical Center in Boston, MA, the algorithm correctly predicted which patients would progress to septic shock 85 percent of the time. This represents a 60 percent increase over previous methods of prediction.
The next step for the system (which is named Targeted Real-time Early Warning Score or TREWScore) is to determine how best to incorporate it into patient care practices. One method that the researchers are looking into is programming the TREWScore into patient electronic medical records. The algorithm would monitor those 27 factors in all patient files, and if a patient s TREWScore crossed above the threshold to a point where septic shock is likely, it would alert the doctor.
The study s leader, Suchi Saria, an assistant professor of computer science in Johns Hopkins' Whiting School of Engineering and of health policy in the Bloomberg School of Public Health, said of the future of TREWScore the tricky issue is thinking about how the clinical team is provided with the information ¦ but we have to do this in a way that it is well-integrated into the existing clinical workflow and does not cause alarm fatigue."