Singapore, 10 September 2021 – Sepsis is a life-threatening medical emergency when the body has an extreme response to an infection. A person with a weakened immune system, elderly or infants and those with medical conditions such as cancer, diabetes, liver or kidney diseases, are at higher risk of developing sepsis. If not detected and treated early, the death rate may be up to 50 per cent when the infection progresses to severe sepsis and septic shock.
To determine the risk of sepsis-related death quickly and accurately, researchers at the Singapore General Hospital (SGH) and Duke-NUS Medical School leveraged artificial intelligence (AI) and developed a predictive model. The model uses AI analysis of simple clinical information and electrocardiogram (ECG) analysis of Heart Rate Variability (HRV), which is the time variations between patient’s heartbeats. From the readings produced, the team was able to give an accurate risk stratification in five minutes.
“Our Emergency Department sees more than 1,000 patients yearly who present with sepsis. In a busy environment, it helps to be able to quickly stratify them according to severity and mortality risk so that the department can efficiently redirect necessary resources and care to prevent patients with a higher risk from going into septic shock,” said Professor Marcus Ong, Senior Consultant, Department of Emergency Medicine, SGH, and senior author of the study. Prof Ong is also Director of the Health Services and Systems Research (HSSR) Programme at Duke-NUS Medical School.
Every year, sepsis kills more than five million people worldwide. In Singapore alone, close to 5,000 deaths were attributed to sepsis from pneumonia and urinary tract infection in 2019.
The most accurate method to assess mortality risk of a patient with sepsis presently is a blood test. However, the results may take two to four hours, resulting in delay to give the appropriate treatment. The team felt there was much potential in using HRV analysis due to its quick turnaround time of five minutes without the need for invasive tests, and decided to expand its use.
With data from a study involving 340 sepsis patients who came to SGH’s Emergency Department (ED) between September 2014 and April 2017, they created this predictive model comprised of clinical variables such as respiratory rate, blood oxygen level as well as Heart Rate n-Variability (HRnV), a novel approach to HRV analysis. Like HRV, HRnV can be derived from ECG reading but it allowed researchers a more detailed look at the heartbeat variations to determine a value that will alert the ED team to take action. The study also established that the new predictive model outperformed other established sepsis risk scoring models currently available.
“HRV analysis is a useful tool for assessing sepsis patients’ mortality risk. However, the amount of information that could be extracted is limited. HRnV, on the other hand, provides us with additional information to improve prediction. With the availability of more biomarkers derived from HRnV analysis, advanced artificial intelligence approaches can be used for accurate and reliable patient risk assessment,” said Associate Professor Liu Nan, from Duke-NUS’ HSSR Programme and Centre for Quantitative Medicine. Prof Liu is also the study’s corresponding author.
The study was published in peer-reviewed scientific journal PLOS One, in August 2021.
To further validate the effectiveness of the HRnV model, the team is now conducting a larger study involving 1,100 patients at SGH and the National University Hospital. It will also determine if the technology has potential to be developed into bedside devices for other uses in the ED, specialist outpatient clinics, general practitioner (GP) clinics and even in homes for monitoring purposes.
For media enquiries, please contact:
M VikneswaranCommunications Singapore General HospitalEmail: email@example.com
Federico GracianoCommunicationsDuke-NUS Medical SchoolTel: 6601 3272Email: firstname.lastname@example.org
Subscribe to our mailing list to get the updates to your inbox