Healthcare-associated infections (HAIs) continue to be an ongoing threat to patients. “Despite the best care, patients get infections,” Brian Gross, business lead, genomics for infectious disease at Philips, told MD+DI. Because they are “technically avoidable, there’s pressure on hospitals to drive their numbers down.”
Healthcare system professionals have struggled with two main problems when trying to identify and act on infection risks, Gross explained. “The problem is two-fold. First, because of current tools, all infections look the same,” he said. For instance, if a hospital diagnoses three babies with MRSA, the NICU may be shut down out of fear that the infection is spreading from infant to infant, but it may in fact have come from a parent or visitor. Such action is often taken before the source of infection is identified, Gross said. “Infections may look the same under the microscope—they are only differentiable if you look at their genomics,” he said.
The second problem hospital systems often experience involves “false negatives,” Gross explained. “Infections are being transmitted, but no one knows it until there are multiple patients infected.”
Better data analysis might be able to help hospitals tackle these issues and fight the spread of infection. A recent study published in Infection Control and Hospital Epidemiology looked at the use of an informatics tool to analyze patient and lab data as well as pathogen genomics to identify potential instances of transmission.
Philips IntelliSpace Epidemiology AI-driven tool was used in the study. Part of Philips’s cloud-based IntelliSpace platform, a collection of informatics applications developed to gather and present healthcare data, the Epidemiology tool is meant to be used across all departments, as HAIs are a “broader problem to solve,” said Gross. “Using AI algorithms, prospective data mining, and genomics could enable professionals to piece together” the source of infection, he added. It could potentially “improve the efficiency of the infection control team,” and after a second patient, prevent further infection.
Gross said that users don’t need the entire IntelliSpace suite to be able to use the Epidemiology tool. “This product is used through a subscription,” he said. “At a minimum, the tool needs access to lab results as well as discharge and transfer information to show where patients are or have been in the hospital system. It also has the ability to interact with OR/surgery, radiology, and nurse documentation.” The system maps all transactions of care, including caregiver contact and devices used, he said.
Genomic sequencing of pathogens is then performed to identify whether there has been a transmission. “It is a mathematical problem that we can do in microseconds in our cloud infrastructure,” he said. “The more information we give the system, the more ‘crisp’ our recommendations become. And the more data you give AI, the smarter it gets.”
Gross said the study demonstrated the use of pathogen genomics for the first time, focusing on four specific bacteria – S. aureus, E. faecium, P. aeruginosa, and K. pneumonia. “We looked at infections and clinical data for one year, for 850 patients,” Gross said. “We saw 34 transmission events, where previously they had only known about one. With the benefit of AI algorithms, the system showed 33 others.”
The authors concluded that the new Philips IntelliSpace Epidemiology solution could enable clinicians to intervene and interrupt the chain of infection transmissions, potentially resulting in fewer infections. Using genomic sequencing of bacteria was key to uncovering the transmissions, as it allowed identification of distinct bacterial strains. “In developing IntelliSpace Epidemiology, scientists at Philips studied slight changes in a bacteria’s genetic code,” Philips explained in a press release on the study. “Bacteria of a single species from a single patient collected over a number of days differed only slightly in their genetic code due to normal bacterial mutation. But bacteria of the same species found on different patients showed much larger variations. Using this analysis, the researchers were able to define thresholds to determine related vs. unrelated bacterial strains. From these genetic data, integrated with other hospital-derived epidemiologic data, IntelliSpace Epidemiology aims to determine if a new bacteria isolated from a patient should be included or excluded in a grouping (or cluster) of related infections. This new method of combining diverse and rich clinical and genomic datasets in order to track and help control HAIs is termed Precision Infection Prevention in recognition of the hospital infection prevention team’s role in preventing HAIs.”
Some potential users have been hesitant about genomic sequencing over concerns that it could be “a challenge, but it is quite easy,” Gross said. “It can be daunting for a hospital system to sequence, but if it is already sequencing for oncology or personalized medicine, this system capability can be added.” Philips also offers sequencing as a service.
Gross described the “science behind genomics” as “rapidly becoming a clinical reality. This paper presented a mechanism to determine whether infections are related or not. It can reduce the complexity of what clinicians are seeing, and the science is in the cloud.”
IntelliSpace Epidemiology is essentially the fusion of clinical informatics with genomic informatics,” he concluded. “We are bringing them together for a precision infection prevention program.”