- Artificial Intelligence Tool Helps Combat Antibiotic Resistance, Misuse
- AHRQ Program Helped Long-Term Care Facilities Reduce Antibiotic Use
Identifying when antibiotic allergic-type events occur and subsequently linking these reactions to particular drug exposures is crucial, particularly for patients who are hospitalized or undergoing surgery. However, standardized processes to flag these events are limited.
To address this, researchers designed an algorithm to improve the detection of antibiotic allergy adverse events.
The research team gathered data from 36,344 patients at Veteran Affairs (VA) hospitals who underwent cardiovascular implantable electronic device (CIED) procedures and received periprocedural antibiotic prophylaxis from October 1, 2015, to September 30, 2019.
Patients were split into training and testing cohorts, and the researchers manually reviewed each case to determine the presence of allergic-type reactions and their severity.
The research team selected variables deemed potentially indicative of allergic-type reactions—including reported and observed allergies contained in the VA’s Allergy Reaction Tracking (ART) system, allergy diagnosis codes, medications given to treat an allergic reaction, and text searches of clinical notes for phrases that may indicate an antibiotic allergy reaction— to develop the tool.
The model was developed on these data from the training cohort and then applied to the testing cohort.
Overall, the tool identified antibiotic allergic-type reactions with an estimated probability of 30 percent or more, a positive predictive value of 61 percent, and a sensitivity of 87 percent.
The final algorithm included seven variables to detect antibiotic allergic-type reactions: entries in the VA’s ART, both reported and observed; antihistamine administration; keyword detection in clinical notes; and PheCodes for “urticaria,” “symptoms affecting skin,” and “allergy or adverse event to an antibiotic.”
These findings indicate that an informatics tool could potentially improve the detection of periprocedural antibiotic allergy adverse events, which could help provide clinicians with feedback concerning the harms that can result from unnecessarily prolonged antibiotic use.
This research is the latest to leverage data analytics and health informatics to advance antibiotic stewardship.
Last year, University of Florida (UF) Health researchers developed a clinical decision-making tool that can determine whether a case of pediatric diarrhea is caused solely by a virus or by bacterial infection, which influences whether a patient should be treated with antibiotics.
Most pediatric diarrhea cases in developing countries are treated with antibiotics regardless of their cause, leading to antibiotic resistance. Using the tool, the researchers found that when the predicted probability of viral-only diarrhea increased, antibiotic prescriptions fell.
By 