Q&A: Predictive AI can help to prevent sepsis Special

Posted Sep 28, 2019 by Tim Sandle
Healthcare facilities, like Virginia’s Sentara Healthcare and Alabama's Southeast Healthcare, are using AI to predict which patients are likely to become septic and prevent the life-threatening condition. Expert Dr. John Showalter explains how.
File photo: Zion Harvey had his hands and feet amputated at the age of two  following a sepsis infec...
File photo: Zion Harvey had his hands and feet amputated at the age of two, following a sepsis infection
, Children's Hospital of Philadelphia/AFP
Sepsis is a major medical issue. In the next week, an estimated 5,000 people will die from sepsis in the U.S. alone, and one third of all hospital deaths are related to sepsis (according to U.S. Centers for Disease Control and Prevention figures). These deaths are preventable, but by the time sepsis is detected, it’s often already too late.
One way to reduce incidences of sepsis is with the application of artificial intelligence. The staff at Sentara Healthcare are using an AI-enabled prescriptive analytic tool developed by Jvion, which identifies who is at risk of sepsis, alerts clinicians and suggests interventions tailored to each patient’s needs.
Sentara had previously created its own sepsis alert system that used nine data points to detect the first signs of sepsis in patients. The new system analyzes about 4,500 pieces of medical, environmental, and socioeconomic data per patient to predict sepsis before it takes hold. In addition, a second health facility - Alabama's Southeast Healthcare - has achieved a 25.5 percent average monthly reduction in sepsis after implementing the Jvion Machine into their workflow.
To understand how AI is working in practice, Digital Journal spoke with Dr. John Showalter, Chief Product Officer, Jvion.
Digital Journal: How widespread is sepsis?
John Showalter: Sepsis is a pervasive problem effecting people from all walks of life. It is one of the leading causes of death in the United States. Each year over 1.7 million adults in America will be diagnosed with sepsis and nearly 270,000 will die from it. According to the CDC, Sepsis accounts for a third of all hospital deaths. However, as many as 80% of these deaths could be avoided with rapid diagnosis and treatment, but often by the time sepsis is detected, it’s too late.
DJ: Are cases of sepsis increasing?
Showalter: The mortality rates for sepsis have been decreasing, but overall are still quite high, especially for a treatable condition.
DJ: Which types of bacteria are the main causes of sepsis?
Showalter:Bacterial sepsis is the most common form of sepsis, but you can get it from viruses (i.e. the flu), and fungal infections. The most common bacteria are Staphylococcus aureus, Enterococcus, and Staphylococcus epidermidis.
DJ: To what extent are the public knowledgeable of sepsis?
Showalter:In 2009, according to an article published, 81% of people in the US had never heard of sepsis. Of those that said they had heard the term sepsis only 44% could identify infection as the cause.
Now, in 2019, in a recent survey by the Sepsis Alliance, we see that while awareness of the term itself is at an all-time high (65%), 91 million adults say they do not know the symptoms of sepsis demonstrating a lack of awareness around the prevalence and severity of the condition.
So, generally, I think we still have alot of education to do with the public around this often-preventable condition.
DJ: To what degree is sepsis preventable?
Showalter:It is mostly about preventing the infections that lead to sepsis. Only 66%-70% of adults get recommended vaccines for pneumonia and the flu, so in there is a large portion of sepsis that could be prevented with vaccines.
DJ: How is sepsis currently detected?
Showalter:Sepsis is detected with a physical exam leading to a diagnosis of infection, an elevated/low white blood cell count, abnormal heart rate, rate of breathing, and temperature with low blood pressure or lab tests demonstrating organ damage
DJ: How can new technologies improve detection?
Showalter:At hospitals around the country, including Sentara Healthcare in Virginia and Southeast Health in Alabama, doctors and nurses are using AI-enabled prescriptive analytics provided by Jvion to predict which patients will get sepsis and the specific interventions that can be taken to stop it in its tracks. Since integrating the Jvion into their workflow, Southeast Health has seen a 25.5% average monthly reduction in sepsis cases. The AI-enabled prescriptive analytics tool looks at 4,500 data points per patient to see who is at risk of developing the life-threatening condition, alerts clinicians when a patient needs help, and suggests interventions tailored to the patient’s needs.
Jvion’s technology improves upon existing solutions such as EHR systems by being purpose-built to address this challenge, leveraging new architectural frameworks that enable artificial intelligence (AI) and machine learning (ML) and supporting deep data analytics. Jvion can ingest data from a myriad of publicly and commercially available sources, such as de-identified patient data, Medicare, clinical research, commercial payor, social, socio-economic, medical literature and census data, to create the underlying mapping needed to generate patient-specific clinical interventions. This is far beyond what traditional solutions, including EHRs, are able to incorporate and analyze.
DJ: How did you develop and test your technology?
Showalter:Jvion’s inception was driven by a passion to build a prescriptive analytics solution that stops preventable harm. A solution that precisely and comprehensively foresees risk and—more importantly—provides the best recommended actions that will improve outcomes.
Jvion helps healthcare systems prevent patient harm and associated costs by enabling clinical staff to focus attention, resources, and individualized interventions on patients whose outcomes can be improved. Unlike traditional AI and predictive analytic solutions that merely identify high-risk patients and cause alarm fatigue, Jvion pinpoints the impactable patients who are on a risk trajectory that can be changed and provides the patient-specific recommendations that will drive to a better outcome.
The Jvion MachineTM is a combination of Eigen-based mathematics, dataset of more than 16 million patients, and software that can be quickly applied to any of 50 preventable harm vectors, including sepsis, without the need to create new models or to have perfect data. Jvion has proven effective in clinical settings for nearly a decade, with hospitals reporting average reductions of 30% in preventable harm incidents and avoidable cost savings of $6.3 million a year.
DJ: What are the next phases for testing or roll-out of the solution?
Showalter:Jvion’s solution is generally available now and is in use by over 70 hospitals across the North America including Mercy Medical Center, Southeast Health, Grady Memorial Hospital, Baptist Health, Novant, Northwell and Geisinger among others.