Algorithms have long been used in medicine to evaluate whether targeted workups are needed for different conditions. They are computations that help providers determine the likelihood that a patient’s symptoms are due to one or another condition. Algorithms save lives by helping decision making as calculators for pulmonary embolism, or pneumonia severity. Alternatively, they may help provide treatment plans by calculating a patient’s BMI. They require providers to input the data, so the condition needs to at least be on the provider’s differential diagnosis. A system that automatically calculates risk scores for certain conditions may shortcut the decision making and expedite the treatment process, for patients where time is essential.

A sepsis algorithm recently developed to automatically calculate scores from from electronic health records has demonstrated a remarkable overall 20% drop in sepsis related deaths. Researchers and doctors at Johns Hopkins developed and fine-tuned a system that automatically identifies the symptoms most closely related to sepsis development.

Sepsis is the body’s reaction to a major infection; the response itself may cause organ failure, limb loss, and even death. Approximately 1.7 million American adults develop sepsis every year; some 270,000 die. Even patients who do not die may have long lasting or permanent injury. The condition is closely associated with hospital patients – often develops in response to a hospital acquired infection. Sepsis is also the leading cause of in-hospital death in the U.S..

Sepsis has proliferated in part because it can develop so rapidly. The difference of a couple of hours often has a significant affect on the patient’s outcome. Sometimes the condition is not recognized until the patient is in septic shock, which often results in permanent injury or death. An automatic warning will note vital sign, lab results, or orders placed by providers, and trigger an alert to providers that the patient must be urgently evaluated for sepsis. Implementation of this system has led to antibiotic therapy commencing just two hours faster, and already led to a nearly 20% decrease in deaths due to sepsis.

Systems such as this TREWS system, derived from machine learning, may lead to other alerts for other conditions that can save future lives.