Using old cultures to choose new antibiotics

Results from previous hospitalizations were used to guide empiric therapy.


When physicians prescribe empiric antibiotics, they have two conflicting goals. “On the one hand, we're trying to maximize the likelihood of coverage of the patient's infection. On the other hand, we're trying to minimize unnecessary broad-spectrum antibiotic treatment,” said Nick Daneman, MD, MSc, a clinician scientist at Sunnybrook. “Our research program is trying to help reconcile these opposing goals.”

In an effort to help prescribers balance these priorities, the hospital developed a system to give clinicians prescribing suggestions based on prior microbiology culture results from the patients they were treating, effectively creating a precision medicine approach to empiric antibiotic therapy, said Dr. Daneman. “In this way, we can both increase the chance that a patient receives adequate coverage for their infection and, at the same time, decrease overall broad-spectrum antibiotic use in the hospital.”

How it works

The audit-and-feedback intervention used simple rules to identify patients whose antibiotic regimens were likely to be unnecessarily broad or ineffective. All of the patients had blood cultures in process for their current infections, but the results were not back yet. “These rules were based on the patient's historic culture results, which have been shown to predict for resistance in future infection,” said Marion Elligsen, BScPhm, an antimicrobial stewardship pharmacist at Sunnybrook who served as the study pharmacist.

She reviewed the patients' electronic medical records (EMRs) to determine whether their initial empiric antibiotic therapy was discordant with their microbiology history for three cultures: methicillin-resistant Staphylococcus aureus (MRSA) colonization, previous extended-spectrum beta-lactamases, and resistant gram-negatives. If she found an opportunity to optimize empiric antibiotics, she provided suggestions to prescribers on stopping or altering therapy.


The researchers compared the nine-month intervention period (October 2018 through June 2019) to a nine-month retrospective control period during the prior year. They looked at how often the empiric therapy showed clinically significant discord with prior cultures and found 99 cases in the preintervention period and 86 in the intervention period. The proportion of patients who received concordant therapy significantly increased from 73% (72 of 99) to 88% (76 of 86) under the intervention. The intervention significantly decreased the primary outcome, time to concordant therapy, by more than 30 hours (55.2 hours preintervention vs. 24.5 hours during the intervention; adjusted hazard ratio, 1.95 [95% CI, 1.37 to 2.77]), a finding that was consistent across all three subgroups of cultures, according to results published last September by Clinical Infectious Diseases.

Overall, prescribers accepted 72% of the 86 suggestions for alternative therapy. They are accustomed to receiving feedback from the antimicrobial stewardship pharmacists, said Ms. Elligsen, lead author of the study, so she was not surprised that they were willing to accept her recommendations. One finding that did surprise her, however, was within the MRSA subgroup: The median duration of vancomycin in patients whose previous cultures didn't support its use significantly decreased from 36.8 to 10.7 hours after the intervention. “I didn't know how much ‘unnecessary’ vancomycin was being used,” she said, “and being able to reduce patient exposure to this drug with the intervention was a pleasant surprise.”


Ultimately, the intervention proved too time-intensive to sustain, Ms. Elligsen noted. “We've paused the intervention at our site,” she said. “Automating the screening process to identify patients on discordant therapy could easily overcome this challenge and is something we are exploring.”

Lessons learned

That said, having a pharmacist provide the suggestions likely helped with uptake, Ms. Elligsen said. “I think this was necessary up front to help prescribers gain comfort with the intervention, but having a mechanism to automate some of the process—either though a [computerized physician order entry] system or EMR—would certainly help with sustainability.”

Next steps

Although the intervention has stopped temporarily at Sunnybrook, researchers at the Ottawa Hospital in Ontario, along with Sunnybrook and Ontario-based Trillium Health Partners, plan to adapt it to improve sepsis care, said Derek MacFadden, MD, a clinician scientist at the Ottawa Hospital and co-lead author of the study. “[This] is when we have the greatest potential to improve both adequacy of antibiotic therapy and reduce unnecessary antibiotic use,” he said. “We aim to develop, implement, and study this intervention across multiple different institutions in a cluster randomized fashion, to show that this approach is generalizable and effective.”