Multivariate analysis of antibiograms for typing Pseudomonas aeruginosa

A method for typing Pseudomonas aeruginosa using antibiotic susceptibility patterns is presented, which allows recognition of clusters of the same strain among clinical isolates from different patients, thus indicating whether cross infection has occurred. An index of similarity (the euclidean or the oblique distance), which includes all the differences of disk zone sizes among isolates, is computed and then elaborated by a clustering algorithm that successively groups all the isolates in larger clusters. The results of clustering are presented as dendrograms, whose terminal branches are pruned down to a level below which differences are casual; isolates that still appear on a common branch are considered identical. The reliability of this technique for detecting nosocomial cross infections was assessed by comparing its results with that of serotyping and pyocin typing. Only 2 of 31 (6.4%) clusters detected by multivariate analysis were not confirmed, while 4 of 33 (12.1%) clusters were recognized by serotyping and pyocin typing, but not by multivariate analysis. In at least two instances the differences in susceptibility patterns were due to cytoplasmic R factors. The routine use of antibiogram data for typing purposes should be considered an essential part of nosocomial infection control.