User login
Use of a clinical decision tree predicted antibiotic resistance in sepsis patients infected with gram-negative bacteria, based on data from 1,618 patients.
Increasing rates of bacterial resistance have “contributed to the unwarranted empiric administration of broad-spectrum antibiotics, further promoting resistance emergence across microbial species,” said M. Cristina Vazquez Guillamet, MD, of the University of New Mexico, Albuquerque, and her colleagues (Clin Infect Dis. cix612. 2017 Jul 10. doi: 10.1093/cid/cix612).
The researchers identified adults with sepsis or septic shock caused by bloodstream infections who were treated at a single center between 2008 and 2015. They developed clinical decision trees using the CHAID algorithm (Chi squared Automatic Interaction Detection) to analyze risk factors for resistance associated with three antibiotics: piperacillin-tazobactam (PT), cefepime (CE), and meropenem (ME).
Overall, resistance rates to PT, CE, and ME were 29%, 22%, and 9%, respectively, and 6.6% of the isolates were resistant to all three antibiotics.
Factors associated with increased resistance risk included residence in a nursing home, transfer from an outside hospital, and prior antibiotics use. Resistance to ME was associated with infection with Pseudomonas or Acinetobacter spp, the researchers noted, and resistance to PT was associated with central nervous system and central venous catheter infections.
Clinical decision trees were able to separate patients at low risk for resistance to PT and CE, as well as those with a risk greater than 30% of resistance to PT, CE, or ME. “We also found good overall agreement between the accuracies of the [multivariable logistic regression] models and the decision tree analyses for predicting antibiotic resistance,” the researchers said.
The findings were limited by several factors, including the use of data from a single center and incomplete reporting of previous antibiotic exposure, the researchers noted. However, the results “provide a framework for how empiric antibiotics can be tailored according to decision tree patient clusters,” they said.
Combining user-friendly clinical decision trees and multivariable logistic regression models may offer the best opportunities for hospitals to derive local models to help with antimicrobial prescription.
The researchers had no financial conflicts to disclose.
Use of a clinical decision tree predicted antibiotic resistance in sepsis patients infected with gram-negative bacteria, based on data from 1,618 patients.
Increasing rates of bacterial resistance have “contributed to the unwarranted empiric administration of broad-spectrum antibiotics, further promoting resistance emergence across microbial species,” said M. Cristina Vazquez Guillamet, MD, of the University of New Mexico, Albuquerque, and her colleagues (Clin Infect Dis. cix612. 2017 Jul 10. doi: 10.1093/cid/cix612).
The researchers identified adults with sepsis or septic shock caused by bloodstream infections who were treated at a single center between 2008 and 2015. They developed clinical decision trees using the CHAID algorithm (Chi squared Automatic Interaction Detection) to analyze risk factors for resistance associated with three antibiotics: piperacillin-tazobactam (PT), cefepime (CE), and meropenem (ME).
Overall, resistance rates to PT, CE, and ME were 29%, 22%, and 9%, respectively, and 6.6% of the isolates were resistant to all three antibiotics.
Factors associated with increased resistance risk included residence in a nursing home, transfer from an outside hospital, and prior antibiotics use. Resistance to ME was associated with infection with Pseudomonas or Acinetobacter spp, the researchers noted, and resistance to PT was associated with central nervous system and central venous catheter infections.
Clinical decision trees were able to separate patients at low risk for resistance to PT and CE, as well as those with a risk greater than 30% of resistance to PT, CE, or ME. “We also found good overall agreement between the accuracies of the [multivariable logistic regression] models and the decision tree analyses for predicting antibiotic resistance,” the researchers said.
The findings were limited by several factors, including the use of data from a single center and incomplete reporting of previous antibiotic exposure, the researchers noted. However, the results “provide a framework for how empiric antibiotics can be tailored according to decision tree patient clusters,” they said.
Combining user-friendly clinical decision trees and multivariable logistic regression models may offer the best opportunities for hospitals to derive local models to help with antimicrobial prescription.
The researchers had no financial conflicts to disclose.
Use of a clinical decision tree predicted antibiotic resistance in sepsis patients infected with gram-negative bacteria, based on data from 1,618 patients.
Increasing rates of bacterial resistance have “contributed to the unwarranted empiric administration of broad-spectrum antibiotics, further promoting resistance emergence across microbial species,” said M. Cristina Vazquez Guillamet, MD, of the University of New Mexico, Albuquerque, and her colleagues (Clin Infect Dis. cix612. 2017 Jul 10. doi: 10.1093/cid/cix612).
The researchers identified adults with sepsis or septic shock caused by bloodstream infections who were treated at a single center between 2008 and 2015. They developed clinical decision trees using the CHAID algorithm (Chi squared Automatic Interaction Detection) to analyze risk factors for resistance associated with three antibiotics: piperacillin-tazobactam (PT), cefepime (CE), and meropenem (ME).
Overall, resistance rates to PT, CE, and ME were 29%, 22%, and 9%, respectively, and 6.6% of the isolates were resistant to all three antibiotics.
Factors associated with increased resistance risk included residence in a nursing home, transfer from an outside hospital, and prior antibiotics use. Resistance to ME was associated with infection with Pseudomonas or Acinetobacter spp, the researchers noted, and resistance to PT was associated with central nervous system and central venous catheter infections.
Clinical decision trees were able to separate patients at low risk for resistance to PT and CE, as well as those with a risk greater than 30% of resistance to PT, CE, or ME. “We also found good overall agreement between the accuracies of the [multivariable logistic regression] models and the decision tree analyses for predicting antibiotic resistance,” the researchers said.
The findings were limited by several factors, including the use of data from a single center and incomplete reporting of previous antibiotic exposure, the researchers noted. However, the results “provide a framework for how empiric antibiotics can be tailored according to decision tree patient clusters,” they said.
Combining user-friendly clinical decision trees and multivariable logistic regression models may offer the best opportunities for hospitals to derive local models to help with antimicrobial prescription.
The researchers had no financial conflicts to disclose.
FROM CLINICAL INFECTIOUS DISEASES
Key clinical point:
Major finding: The model found prevalence rates for resistance to piperacillin-tazobactam, cefepime, and meropenem of 28.6%, 21.8%, and 8.5%, respectively.
Data source: A review of 1,618 adults with sepsis.
Disclosures: The researchers had no financial conflicts to disclose.