A Bot for Sepsis and Cancer Care in the Cloud

Two global generators of massive amounts of healthcare data are collaborating on digital tools for some of the world’s biggest health challenges

The average hospital creates 50 petabytes of data per year – roughly the size of 10 million iCloud storage accounts. This includes clinical notes, lab tests, medical images, sensor readings, genomics, and operational and financial data. Yet less than three percent of it is used. [1]

For example, the heart rate makes it to the EMR, but the high-fidelity ECG waveform it comes from never does. Lab values might show that white blood cell count is normal, but the microscopes in the lab can see changes in the white cells shape and internal components.

“It’s like visiting the Mona Lisa and sending home a post card that reads, ‘I saw a painting of a woman,’” said Paul Mullen, General Manager of Acute Care at GE Healthcare, which is working closely with clinical partners globally to solve these operational and technology challenges. “It’s true, but not enough.”

Lack of effective data integration can prevent clinicians from understanding the root cause of an illness and may hinder informed decision-making when it is most crucial, for example in acute situations, such as the onset of sepsis. Without quick treatment, sepsis, a common but serious complication arising from an infection, can cause multiple organ failure. It claims the lives of 250,000 patients in the United States each year.

“The early detection of sepsis has eluded us even with the most state-of-the-art biomarkers and diagnostics,” said Dr. Matthew J. Bankowski, Ph.D., MS, D (ABMM), HCLD/CC (ABB), Clinical and Molecular Microbiology Laboratory Director at Baptist Medical Center in Jacksonville, Florida. “We desperately need a solution for early detection.”

To address this challenge, GE Healthcare and Roche Diagnostics are working to create an AI-enabled “Virtual Collaborator”[2] to integrate data from the EMR and other patient-information systems and provide insights into the status and trends of patients who are at-risk for sepsis-related deterioration.

While most current applications of AI are reactive in nature, with the technology responding to questions or inputs, GE Healthcare and Roche say their AI-enabled bot takes things one step further – beginning to infer the “why” behind clinicians’ questions and serving relevant information beyond what was asked.

For example, a clinician treating a patient for heart failure will have different reasons for asking about blood pressure than a clinician treating a trauma patient, even though the question -- “What is the blood pressure?” -- is the same for both.

“Imagine a clinical scenario in which a patient’s condition is deteriorating, but it’s unclear why or how quickly,” Mullen said. “Clinicians are left to forage through mounds of data and to connect the dots – when there’s little time to waste. A doctor could ask the virtual collaborator a question, and it will ‘understand’ and provide additional context, otherwise buried in the EMR, to help the clinician better treat the patient.”

Part of a partnership between GE Healthcare and Roche Diagnostics that was announced in early 2018, the sepsis solution is one example of the way industry is coming together in less traditional ways.

“I believe that a collaborative endeavor like the one between diagnostic tools from Roche and analytic tools for patient monitoring from GE Healthcare is the type of joint thinking we need for early detection,” Dr. Bankowski said.

The collaboration is also working on cloud-based software to fundamentally change the process of tumor boards – meetings that bring together providers from multiple disciplines to discuss the diagnosis and treatment plan for a patient with cancer.

Reports show that the process of preparing for, conducting and documenting tumor board meetings is frequently suboptimal and non-standardized,[3] with each specialist aggregating data in a silo. As a result, clinicians can develop perspectives based on an incomplete view of the patient, and meetings are spent switching back and forth between systems and portable technologies as each specialist gives their presentation.[4] These inconsistencies and inefficiencies can lead to wasted time, decreased engagement, and could even negatively impact patient outcomes.[5]

“Time is the most important factor in quality of care, and it is a scarce commodity in medicine,” said Dr. Sergi Serrano, a pathologist at Hospital del Mar in Barcelona, Spain. “Time to test, time to review results, time to listen to the patients, time to consider the best therapeutic option.”

That’s why the two companies are partnering to combine in-vitro diagnostic data (biomarkers, tissue pathology, genomics, sequencing) and in-vivo diagnostic data (medical imaging) with clinical context for a more streamlined and holistic view of the patient.

“NAVIFY Tumor Board streamlines and standardizes clinical workflow and aggregates relevant data into a holistic patient dashboard, helping oncology care teams collaborate to deliver personalized care more efficiently,” said
Dr. Okan Ekinci, Chief Medical Officer of Roche Diagnostics Information Solutions.


[2] Technology in development that represents ongoing research and development efforts. These technologies are not products and may never become products. Not for sale. Not cleared or approved by the U.S. FDA or any other global regulator for commercial availability

[3] Patkar, Vivek, et al. “Cancer Multidisciplinary Team Meetings: Evidence, Challenges, and the Role of Clinical Decision Support Technology.” International Journal of Breast Cancer 2011 (2011).