Ten hours into the future: Predicting ICU deterioration for earlier intervention
First for SA: Electronically enabled algorithm supports better clinical decisions
A South African first of its kind tool is transforming patient care locally by leveraging Netcare’s advanced electronic medical records system and abundant clinical data. This will help clinicians identify risk of deterioration from common causes earlier among intensive care patients so that treatment can start sooner.
Doctors in Netcare ICUs now have the support of a scientifically developed algorithm that predicts a person’s risk of deteriorating from causes including heart failure, respiratory instability or compromise, infection, sepsis or acute cardiac arrhythmia.
“The prediction algorithm uses automatically recorded real-time heart rate, blood pressure, oxygen saturation and respiratory rate data to detect a person’s chance of deteriorating. This provides vital information that doctors can use to commence therapy much earlier, when such interventions tend to be most effective,” explains Professor Reitze Rodseth, head of clinical data innovation and research at Netcare.
Underpinned by complex mathematics, the artificial neural network applies a machine learning algorithm to analyse electronic information on patients’ vital signs and anticipate the risk of a patient deteriorating eight to ten hours before it is otherwise clinically identifiable.
International collaboration
The sophisticated algorithm was developed over years by researchers from institutions including Netcare, Charité – Universitätsmedizin Berlin, one of the largest academic hospitals in Europe, Telehealth Competence Centre (TCC) in Germany, the University of KwaZulu-Natal, University of Minnesota School of Medicine and Emory University School of Medicine in Atlanta, United States, and DigitalOn Tech (Pty) Ltd from South Africa.
The algorithm, which goes live in all Netcare ICUs from 28 May 2025, provides an early indication to the treating doctors who are able to assess and adjust medication to prevent their patients’ condition from deteriorating. In the case of sepsis, which is the body’s reaction to an overwhelming infection, this is critical as treatment is especially time sensitive for securing the best outcomes for the patient.
“Crucially, this means that we are supporting doctors to identify this risk hours in advance, and commence potentially lifesaving treatment much earlier – providing an opportunity to address this leading cause of clinical deterioration before it progresses,” says co-author of the article published in The Journal of Clinical Medicine and chief executive officer of the Netcare Group, Dr Richard Friedland.
Turning back the clock
“This insight is of considerable human and clinical value, as once sepsis has reached the stage where the person is showing symptoms, the risk of mortality is tragically as high as 20%. Turning back the clock for the initiation of sepsis treatment with this advanced machine learning algorithm provides clinicians the opportunity to address this leading risk for intensive care patients earlier,” Dr Friedland says.
Digital integration of medical equipment and devices in ICUs and theatres is already well established in all Netcare hospitals through the international award-winning CareOn electronic medical records system, another South African healthcare first that Netcare initiated. With approval from the South African Health Products Regulatory Authority (SAPHRA), this prediction algorithm has been embedded into the CareOn system.
“Strategically, our Group-wide digitisation focus laid a foundation that now enables us to use this technology meaningfully in the clinical setting. The data derived from digitising our operating platforms informs the development and implementation of innovative analytics and algorithms that informs doctors’ decisions at the bedside, improving quality and safety of care, as well as cost-effectiveness,” adds Dr Anchen Laubscher, Group medical director of Netcare.
“The World Health Organization notes that treatment of sepsis is most effective when initiated early, and this evidence based tool seeks to provide an early warning system for doctors, adding an extra layer of protection that is especially significant for persons at higher risk of infection and sepsis, including the elderly, long-term ICU patients and people with certain co-morbidities,” she explains.
During the pilot phase, limited to certain Netcare ICUs, it was observed that even among patients whose vital signs were not yet abnormal enough to typically warrant concern, the machine learning algorithm was able to identify an increased risk in the quick sequential organ failure assessment (qSOFA) score, prompting doctors to evaluate this new information as part of their management of intensive care patients for timeous treatment.
Now, the system has been implemented in all Netcare ICUs as part of an ethics approved clinical study to measure the impact of this technology. The system always protects the privacy of patients, and no personal data ever leaves Netcare’s strictly safeguarded information technology environment.
“We believe early warning systems such as this risk prediction tool represents the future of proactive healthcare. Through strategically harnessing evidence-based medicine that is data driven and digitally enabled, we see this groundbreaking application of the algorithm in our ICUs as another important step in our mission to provide the best and safest care, centred on the needs of the individual,” Dr Friedland concludes.
What is sepsis?
Sepsis is defined as a life-threatening extreme reaction of the body’s own defences against any form of infection, bacterial, viral, fungal or even parasitic.
“When the immune system overreacts to infection in this way, it can cause tissue damage and organ failure. Without prompt treatment, sepsis may progress to septic shock and can be fatal,” explains Dr Anchen Laubscher, Netcare’s Group medical director.
Sepsis accounted for 19.7% of all deaths worldwide, according to research by Rudd et al. published in The Lancet in January 2020, cited by the World Health Organization.
While anyone can develop sepsis, some groups are more at risk:
- Pregnant women
- Babies
- Older persons
- People with compromised immune systems, such as those with cancer and HIV+ individuals
- Patients with chronic health conditions, such as kidney disease or diabetes
- People in ICUs or long-term hospital care
‘Research supports earlier intervention as crucial in treating sepsis, which is why Netcare is pioneering the use of this machine learning algorithm technology in its electronic medical records system to help doctors identify this leading risk early among patients in intensive care units,” Dr Laubscher says.
Technology supporting clinical decision-making explained
“When we set out on the research, our aim was to provide a clinical decision support algorithm able to flag patients at risk of deterioration from any cause, most common being respiratory compromise, infection and suspected sepsis, or heart failure, who may benefit from early intervention,” explains Professor Reitze Rodseth.
Through international collaboration with academic institutions and Telehealth Competence Centre (TCC) in Germany, and DigitalOn Tech in South Africa, an artificial intelligence (AI)-driven model was developed to predict patient deterioration.
This sophisticated machine learning algorithm known as an artificial neural network learns from patterns of data, including an ICU patient’s heart rate, blood pressure, oxygen saturation and breathing rate, to indicate a patient’s risk of deteriorating in the next eight to ten hours.
Doctors in Netcare ICUs now have the benefit of this real-time indicator as another tool to help support better informed clinical decisions for patients who are at high risk of deterioration. In this way, technology is enhancing medical professionals’ ability to make time sensitive decisions earlier, delivering better and safer care to our patients.
