Call for tender for project: Grants for health research and development projects 2021: Promotion of health research

Funding body: Department of Health, Basque Government

Record no.: 2021111083

PI: Urko Aguirre

Funding awarded: €71,874

Description: Antibiotic resistance is a growing challenge for global Public Health and threatens advances in the treatment of bacterial infections. Infections due to antibiotic-resistant organisms are associated with increased morbidity and mortality, partly due to inadequate initial empirical antibiotic treatment. Antibiotic Use Optimisation Programmes (AUPs) often contain recommendations for the empirical use of empirical antibiotics. In most cases, these recommendations are derived from expert opinion and informal syntheses of available evidence. Decision support-based adverse event alarm systems seek to identify patients at high risk of infection with multidrug-resistant pathogens. The use of logistic regression-based predictive modelling as part of clinical decision support systems for antimicrobial stewardship remains essential.

Objective:

1.- To identify, independently, in hospitalised patients and in Primary Care: a) the occurrence of infections by microorganisms with resistance; b) inappropriate use according to local guidelines or according to local anti-biotherapy flora.

2.- Based on the data collected with the previous objective 1, develop predictive models and clinical prediction rules for poor evolution (readmission, prolonged hospital stay), complications derived from antibiotic resistance, and inappropriate prescription of antibiotics in hospitalised patients.

3.- To establish a continuous monitoring system based on the electronic medical records of patients attending the emergency department that allows alerts to be established at different levels for antibiotic resistance and prolonged stay.

4.- To identify factors associated with inter-hospital and Primary Care variability as well as to detect problems of equity (according to age, gender, and area where one lives) in terms of treatments, use of anti-biotherapy and antibiotic resistance. 

Hypothesis: 

The hypothesis is that the application of modern machine learning approaches to easily collected patient data can outperform those based on logistic regression or simple decision trees and obtain patient-specific antibiotic susceptibility predictions.

Improved predictions that direct empirical antibiotic therapy can contribute to better patient outcomes and to improve patient outcomes and avoid overuse of inappropriate antibiotics that select for resistance. Optimisation of antibiotic use programmes (PROAs) often contain recommendations for the empirical use of empirical antibiotics. In most cases, these recommendations are derived from expert opinion and informal syntheses of available evidence. Decision-aid-based adverse event alarm systems seek to identify patients at high risk of infection with multidrug-resistant pathogens.

Results:  

The use of logistic regression-based predictive modelling as part of clinical decision support systems for antimicrobial stewardship remains essential. We hypothesise that the application of modern machine learning approaches to collected patient data can easily outperform those based on logistic regression or simple decision trees to obtain patient-specific predictions of antibiotic susceptibility. Improved predictions that direct empirical antibiotic therapy can contribute to improved patient outcomes and avoid overuse of inappropriate antibiotics that select for resistance.

Roles: Kronikgune – Project Coordinator