When to include polymyxins in the empirical antibiotic regimen in critically ill patients with fever? A decision analysis approach.

We sought to approach a practical question: Should polymyxins be used in the initial empirical antibiotic regimen in the intensive care unit (ICU) patient with fever that is thought to be due to infection? By retrieving data from the literature and the WHONET Greece, we formulated a mathematical model to estimate the probability (Ptotal) that a gram-negative bacterium susceptible only to polymyxins is isolated from an ICU patient. Ptotal = P1 * P2 * P3 * P4, where Ptotal = total probability; P1 = probability that fever is due to infection; P2 = probability that infection is due to gram-negative bacteria; P3 = probability that the gram-negative bacterium is Acinetobacter baumannii, Pseudomonas aeruginosa, or Klebsiella pneumoniae; and P4 = probability that A. baumannii, P. aeruginosa, or K. pneumoniae is susceptible only to polymyxins. Using the information from our data sources, we estimated that P1 (before physician input in differential diagnosis) = 0.5, P2 = 0.523, P3 = 0.79, and P4 = 0.567, thus Ptotal = P1 * P2 * P3 * P4 = 0.5 * 0.523 * 0.79 * 0.567 = 0.117 = 11.7%. Based on this information and combining it with data regarding the attributable mortality of inappropriate empirical antimicrobial treatment, 4 to 5 patients in every 100 ICU patients will die if physicians do not include polymyxins in the initial empirical regimen in the ICU setting for an episode of fever due to infection. Polymyxins should probably be included in the empirical antibiotic regimen in the ICU setting in hospitals, where the observed probability that a gram-negative bacterium (A. baumannii, P. aeruginosa, or K. pneumoniae) is polymyxin-only-susceptible is close to that (50%) used in our model (based on the individual hospital data).