Sepsis is the body’s overwhelming response to infection. The onset of sepsis can occur very quickly, and every hour a patient goes untreated, their chance of mortality rises by 8 percent. About one in five deaths worldwide result from sepsis. The treatment of sepsis puts substantial strain on hospitals, requiring facilities and skilled healthcare professionals that in the U.S. cost an estimated $57.47 billion to maintain in 2019 alone.
If healthcare professionals could detect sepsis events earlier, before a patient’s condition deteriorates, they could reduce the adverse effects, improving patient mortality and the strain on the healthcare system. One way providers seek to meet sepsis head-on is through artificial intelligence (AI)-based methods to interpret patients’ vitals. For an AI algorithm to be successful in this new and blossoming field, it must assess an optimized group of patient parameters, particularly early signs of sepsis, such as changes in body temperature, which could be detected by continuous temperature monitoring (CTM). With the correct combination, this technology may soon make it possible to identify sepsis earlier to streamline the delivery of life-saving interventions to more patients.
Severe Sepsis Can Be Preventable
Sepsis begins with an infection of the bloodstream resulting in a drop in blood pressure and an increase in heart rate and temperature. Sepsis can be dangerous or life-threatening if untreated. Early detection can be the difference between life and death.
Numerous lab tests may indicate a patient’s risk of developing sepsis and can be measured to follow the rate of decline. Automated systems have the potential to streamline this process to gain insight as early as possible using AI algorithms designed to synthesize lab results to predict when an individual is at an early risk of developing sepsis. By using an automated system, nursing workflow can be streamlined. Thus, this strategy has the potential to improve patient safety significantly. Moreover, an ideal AI-powered system capable of assessing myriad variables at once could help providers push the limit of early sepsis detection, minimizing the number of patients who develop severe sepsis or septic shock.
However, these AI technologies are still in the early phase. Several hospitals recently reported that one widely-used algorithm for predicting the likelihood of sepsis missed about two-thirds of septic patients. It employed circular logic that precluded the identification of septic patients until doctors had identified the problem and prescribed antibiotics while simultaneously causing a disarming number of false alerts.
Ideal sepsis-predicting algorithms will flag at-risk patients with a low false positive rate. No one symptom alone can indicate whether or not a person has sepsis, but optimizing criteria is vital. For instance, current algorithms underutilize body temperature data, but adding this parameter could potentially bolster AI-based strategies in the future to benefit patients and providers alike.
Tuning into Body Temperature
As an individual becomes septic, their body temperature fluctuates, manifesting as a fever or hypothermia. However, body temperature is the only vital sign not continuously monitored outside the Intensive Care Unit (ICU). Nurses measure inpatients’ temperatures manually about once every four to eight hours. Since body temperature can vary over time and manual measurements are subject to human error, a patient with abnormal body temperature may go unnoticed for some time. As each hour goes by and sepsis remains undetected, the risk of developing severe sepsis or septic shock rises, as does the risk of mortality.
An efficient way to monitor body temperature and tune into a patient’s health status is through the use of wearables capable of CTM. Devices on the market today take measurements multiple times per minute and upload the data wirelessly to the central nursing station. Alone, CTM can help identify a fever hours earlier than manual measurements, but when assessed by an AI algorithm for predicting sepsis, CTM data can do so much more.
Trend Detection for Early Sepsis Prediction
AI algorithms can parse out diagnoses based on biological trends. Healthy bodies follow specific patterns; if variations are detected, the AI concludes that disease may be present. For example, one AI algorithm assesses breathing patterns to detect early Parkinson’s Disease. Another can predict the development of a fever at least 3.5 hours in advance based on circadian rhythm. Body temperature also varies in a tightly regulated pattern in healthy individuals, making it possible for AI software to evaluate irregularities like increased fluctuation and loss of rhythm and identify the onset of sepsis.
When integrated into an algorithm encompassing layers of patient data amassed at the nurses’ central monitoring station, continuous body temperature data could help hone the system, enabling quicker detection of actual sepsis cases and reducing false positives the algorithm discovers.
Sepsis Prediction that Meets Regulatory Standards
There are many variations of sepsis prediction software available on the market designed to assess multiple vital signs at once, but all are missing crucial CTM data that would allow for faster, more reliable sepsis detection. Furthermore, much of this software has been developed with little or no oversight by regulatory bodies like the U.S. Food and Drug Administration (FDA). FDA-approved technology is more likely to give providers results they can trust. Therefore, any software for sepsis prediction and wearable technology for measuring vitals should undergo the same stringent requirements as other medical technologies to reduce inconveniences and impracticalities generated by the algorithms in a hospital setting.
When selecting sepsis prediction software for use in a hospital, administrators should favor algorithms that consider a variety of relevant parameters, including CTM. They should consider software that meets FDA requirements, likely more streamlined, accurate, and ready for patient use. By selecting the best parameters to power sepsis detection algorithms and making it simple for healthcare providers to use that information daily, medical technology companies can save lives and reduce the burden on the healthcare facilities they assist.
John Gannon is President and CEO of Blue Spark Technologies.
