A. Krishna Rao, MD, MS, an infectious diseases specialist who treats patients at the Michigan Medicine Taubman Health Center and University Hospital, and the co-founder and medical director of the Fecal Microbiota Transplantation Program, is making significant strides in doing just that.
Dr. Rao’s research work is focused on the diagnosis and treatment of Clostridium difficile infection (also known as C. diff or CDI), infections from multidrug resistant organisms (MDROs), and studying how the gut microbiome mediates the onset and course of infections. His goal is to improve and create new biomarker-based predictive models for health care-associated infections that will tell whether a patient is predisposed to developing an infection, the severity of the infection, and the chance of recurrence.
“If we can develop accurate predictive models that can tell whether a patient is likely to develop a disease, clinicians can provide earlier and better treatment which will improve patient outcomes,“ says Dr. Rao.
Health care-associated infections (HAIs) and their challenges
HAIs are infections people get while they are receiving health care for another condition and can be caused by bacteria and other pathogens. These infections can lead to major illnesses and death, and can be attributed to a significant cost burden on the health care system. Many of these infections are preventable.
Risk factors for contracting a HAI include catheters, surgery, injections, health care settings that aren’t properly cleaned and disinfected, communicable diseases passing between patients and health care workers, and the overuse or improper use of antibiotics.
Common HAIs include central line-associated bloodstream infections (CLABSI), Clostridium difficile infection, pneumonia, surgical site infections, and urinary tract infections.
One challenge is identifying the best treatment strategy for an infection that will result in good outcomes. While there are many treatment choices available for HAIs such as CDI, they cannot be used in every patient due to cost, invasiveness, or experimental nature, where safety and efficacy are uncertain. Better tools are needed, Dr. Rao points out, to identify the patients who will benefit most from specific therapies.
In addition, as more and more infections are becoming resistant to antibiotics making them more difficult to treat, and diseases occurring as a result of antibiotic use, focus on and adherence to antimicrobial stewardship is critical.
“Health care providers first need to really look at whether a patient’s illness requires an antibiotic, and then, if one is needed, the goal should be to balance the need to appropriately treat the infection with the need to avoid the development of antimicrobial resistance,” Rao says. “This is best done by choosing antibiotics that treat the infection well, but are narrowly focused on the pathogen or syndrome in question, rather than ones that broadly cover many different pathogens.”
Using biomarkers in the diagnosis and management of HAIs
Another challenge in the diagnosis and management of HAIs is identifying, at the earliest stages of a disease, which patients will develop an infection, like Clostridium difficile, and will benefit the most from targeted treatments.
In a study where Dr. Rao was the principal investigator, it was demonstrated for the first time that procalcitonin, a cytokine elevated in bacterial infections, is a predictor of severe CDI and adverse outcomes. He also led a study that was first to characterize the systemic inflammatory response to CDI, which identified hepatocyte growth factor as a new and potentially significant predictor of CDI.
Applying machine learning to improve predictive modeling
Machine learning is a method of data analysis that automates analytical model building based on the idea that systems can learn from data, identify patterns, and make decisions with minimal human intervention. Predictive models are built using machine learning.
Dr. Rao utilizes machine learning for the majority of his research. He believes that leveraging machine learning in the study of infectious diseases enables more accurate and specific predictive models that will aid providers in predicting which individuals are more likely to develop an infection and to what degree and severity.
“Machine learning approaches can be superior to traditional statistical models in wrangling complex, multidimensional data”, says Rao. “It plays an important role in creating better tools for health care providers to risk stratify patients, allocate limited resources, and guide management decisions.”
Finding answers in the gut microbiome
Dr. Rao’s latest research looks to the gut microbiome to understand its role in infections at the various stages, from colonization to the ultimate outcome of the infection. There he hopes to find features that could indicate a susceptibility to an infection and recurrence.
He recently conducted a study which consisted of healthy volunteers who were undergoing elective surgery and were to receive antibiotics to prevent infection. His team collected stool samples one month prior to receiving the antibiotic and stool samples up to three months after receiving the antibiotic.
The results from the comparison were significant. He found that even healthy adults can have long-term disruptions in their microbiome from antibiotics – the stool samples after taking the antibiotic looked different three months later, even in those who received just one dose.
In another of Dr. Rao’s studies, collaborating with Dr. Michael Bachman, MD, PhD from the Department of Pathology, he looked at the gut microbiome to determine if there are factors that could predispose someone to get an Enterobacteriaceae infection. There he found a family of bacteria, Enterobacteriaceae, that when existing in high levels in the stool, led to an increased risk of infection.
In a third study, again in collaboration with Dr. Bachman, Dr. Rao compared the microbiomes in patients who had developed sepsis to those patients who didn’t. Differences in the gut microbiome were discovered between the two groups. The patients who had developed sepsis had a lower level of butyrate-producing bacteria. Butyrate is essential to colonic health, with low levels of it associated with a myriad of conditions. Correspondingly, he found that the lower the level of butyrate-producing bacteria, the higher the risk of the patient dying from sepsis.
Next steps
Dr. Rao plans to continue developing predictive models, using machine learning, biomarkers, and next generation sequencing, for earlier intervention of HAIs and ultimately the prevention of them.
“The combination of machine learning tools and predictive modeling will lead to new advances in precision medicine, which I strongly believe is the way forward. As someone who enjoys mathematics and computers, this is an exciting time in infectious disease research for me,” says Dr. Rao.
“I first became interested in infectious diseases when I was in high school where I was able to take a course in microbiology taught by a professor from a local university. My interest grew and I eventually decided to specialize in this field because I saw it as a major untapped area where I could take my passion and make a difference in improving people’s lives, both locally and globally.”
Learn more about Dr. Rao and his research on ORCID.
Find out more about the Michigan Medicine Division of Infectious Diseases.
