Continuous vital measurements for early sepsis at the emergency department to predict patient deterioration
Continuous vital measurements for early sepsis at the emergency department to predict patient deterioration
Sepsis is a life-threatening dysregulated host response to an infection. It can result in organ failure and may require admission to the intensive care unit (ICU). Early recognition is essential, as prompt treatment with antibiotics improves survival. However, diagnosing sepsis is challenging because its symptoms vary widely between patients.
This thesis of Raymond van Wijk investigates how patients presenting to the emergency department (ED) with an infection can be identified earlier as being at risk of clinical deterioration, ICU admission, or death. The research was based on data from Acutelines, a large clinical database and biobank of ED patients at the University Medical Center Groningen (UMCG).
Several commonly used early warning score, were compared. NEWS and NEWS2 showed the best performance in predicting adverse outcomes. In addition, the results demonstrated that changes in vital signs over time provide important information beyond a single measurement, improving the prediction of patient deterioration.
The thesis also explored novel approaches using routinely collected physiological signals, such as the electrocardiogram (ECG) and photoplethysmography (PPG), a non-invasive measurement of blood flow. These signals were found to contain valuable information about a patient’s condition. A machine learning model that analyzed the complete ECG waveform predicted clinical deterioration more accurately than existing risk scores.
Overall, the findings show that continuous monitoring and advanced analysis of routinely available patient data can improve the early identification of high-risk patients and may contribute to more timely treatment and better outcomes for patients with severe infections.