Computer system aids antibiotic therapy decisions

[unseennude]

TREAT, a computerized decision-support system, can help predict the presence of bacteremia in hospitalized patients, Israeli and other researchers report in the May 1st issue of Clinical Infectious Diseases.

"The mortality associated with severe infections in hospitals is about 40%," lead investigator Dr. Mical Paul told Reuters Health. Appropriate antibiotic treatment improves survival nearly two-fold, but is administered in time only among 60% to 70% of patients. At the same time, unnecessary antibiotic use drives up antibiotic resistance."

The TREAT system, he added, "was built with the intention of improving current practice." It covers 155 pathogens and includes 214 other clinical variables.

Dr. Paul, of E. Rabin Medical Center, Petah-Tikva, and colleagues in Denmark, Germany and Italy developed the system. It was tested in Israel, Germany and Italy in 2 cohorts totaling more than 2500 patients.

The team used the first cohort to define 3 bacteremia risk groups. The prevalence ranged from 2.4% in the low-risk group to 29.9% in the high-risk group.

The researchers applied this classification to the second, independent cohort and achieved good performance. In total, 1.3% of the low-risk patients had bacteremia compared to 28.1% in the high-risk group.

"TREAT can identify patients with a high likelihood for bacteremia needing urgent evaluation and treatment and low-risk patients that can be managed without blood tests," Dr. Paul said.

"In a clinical trial, TREAT prescribed appropriate antibiotics more frequently than clinicians in three countries, while using less ecologically expensive antibiotics," he said. "TREAT can help improve patients' treatment while limiting antibiotic resistance development in hospitals."