OBJECTIVES: As a consequence of the increased prevalence of the carbapenem-resistant Acinetobacter baumannii, these are classified by the WHO and CDC as an urgent threat. The objectives of the study were: 1) to create an agent-based model; 2) to calibrate the model; 3) to predict the prevalence of resistance of Acinetobacter baumannii to colistin.

METHODS: Using the demographic information of the City of Valencia between the years 2012-2020 we built a synthetic population, the population dynamics were obtained from the hospital flow of patients and from the epidemiological information of the periods 2016-2019 and 2015-2018 respectively. We included characteristics of the infection (community, hospital, origin of infection, hospital stay) and different states in which the patient could be found (susceptible, sensitive, resistant). We assumed that every person infected with Acinetobacter baumannii was hospitalized if they were sensitive or resistant to colistin. Among the parameters included in the model were the daily probability of change in the clinical status of new patients and that of hospitalized patients without infection or with nosocomial infection. Using the Particle Swarm Optimization (PSO) algorithm, we carry out the calibration process through simulation and optimization.

RESULTS: With the built and calibrated model, we maintained the probability of being admitted daily (18.4%) and of nosocomial infection (2.6%) by Acinetobacter baumannii. We found that the increase in the number of cases resistant to Acinetobacter baumannii to colistin is related to age, the largest increase being in the ages> 60 years, followed by the age group 30 to 59 years and finally the group of <30 years old.

CONCLUSIONS: The calibrated model adjusted the estimated prevalence to the real prevalence, which allowed us to make predictions of the increase in cases of Acinetobacter baumannii resistant to colistin by age groups, temporalizing the loss of effectiveness of colistin.