Data from EHRs Can Predict and Pinpoint Cases of Clostridium difficile

The generalized model can work across hospitals.

Erica Shenoy, MD, PhD

A new hospital-specific model for identifying patients with high risk for Clostridium difficile (C. difficile) can target infection prevention strategies, according to recent findings.

Researchers from the University of Michigan, Massachusetts General and MIT, wrote that a one size fits all method limits the risk factors that are measured when analyzing risk assessment models. Past literature, however, demonstrates that electronic health record can perform better, which resulted in the researchers building a C. difficile risk stratification model that can be tailored to individual healthcare facilities.

The investigators gathered data from more than a quarter million adult admissions across 2 hospitals in order to show that their facility-specific model is effective. The team collected patient demographics from the electronic health records, plus admission details, patient history and daily hospitalization details.

They split the variables into 2 categories: time-invariant (those that do not change over the course of hospital admission) and time-varying. From these variables, the researchers created a model of each patient admission day, where they could also see the date of C. difficile test and date of patient discharge.

The researchers proved that this model can work at various hospital sites, even when the data they extracted from each electronic health record is represented differently.

“We used a general approach leveraging all the electronic health record contents, and applied it to 2 different hospitals with different electronic health records, different patient populations,” study author Erica Shenoy, MD, PhD, told MD Magazine. “We believe we have demonstrated that institution-specific models and not a one-size-fits-all model approach has major advantages.”

The model allowed the researchers to identify patients at the highest risk for C. difficile infection; more importantly, in those they predicted correctly, they did so 5 days prior to when the patients were actually diagnosed by their doctors, on average. Earlier intervention, of course, can lead to improvements in patient outcomes and less severe disease, Shenoy continued. She added that a model like this can be useful in antimicrobial stewardship efforts and stewardship interventions.

“I see tools and approaches such as these as potentially very effective at reducing the incidence of C difficile,” Shenoy said. “If we are able to diagnose patients earlier, we can implement effective infection prevention tools to prevent transmission to other patients.”

Isolated C. difficile patients can be housed with other C. difficile patients, and providers in these rooms should wear gowns and gloves to prevent spread to other hospital patients, Shenoy added. Hand washing with soap and water, plus the use of alcohol-based sanitizers, may also achieve this. Finally, bleach-based products instead of standard ammonium can kill C. difficile spores that can contaminate rooms. Shenoy said that they can only implement these tools when they know about infections, or when they suspect a case.

“The possibility of identifying patients earlier is very exciting from an infection prevention point of view,” she concluded.

The paper, titled “A Generalizable, Data-Driven Approach to Predict Daily Risk of Clostridium difficile Infection at Two Large Academic Health Centers,” was published in the journal Infection Control & Hospital Epidemiology.