Title: "Disk Diffusion Breakpoint Determination for Antimicrobial Susceptibility Testing"
Speaker: Glen DePalma, Department of Statistics, Purdue University
Place: PHYS 223
Date: March 20, 2012, Tuesday, 4:30pm


Drug dilution(MIC) and disk diffusion (DIA) are the two antimicrobial susceptibility tests used by hospitals and clinicians to determine a pathogen's susceptibility to various antibiotics. This determination is made in part by comparing the observed test results to assigned antibiotic-specific breakpoints. Because the MIC test deals with concentrations of a drug, breakpoints can be based directly on the pharmacokinetics and pharmacodynamics of the drug. The DIA test breakpoints, on the other hand, must be estimated. To do this, a collection of test pathogens must have both an MIC and DIA test performed and the error rate bounded (ERB) method used to determine the DIA breakpoints. This method finds the set of DIA breakpoints that minimize the classification discrepancies between the pairs of observed results. Craig (2000) showed that this approach is very sample dependent (low precision) and can produce biased results (low accuracy). In this work, we propose a hierarchical modeling approach for breakpoint estimation that provides increased precision and better accuracy. This is done by moving away from calibrating observed test results and instead focusing on making the tests' probability of correct identification curves as similar as possible. Our model accounts for the inherent test variabilities and differing testing properties by linking the observed test pairs to an unobserved (latent) 1-1 function of 'true' test results. This unknown monotonically decreasing function is estimated nonparameterically using I-Splines. Simulations have shown this approach to be more accurate, and will help in correct identification of pathogens that are growing more resistant.

Associated Reading:

B.A. Craig. 2000. Modeling Approach to Diameter Breakpoint Determination. Diagn ostic Microbiology and Infectious Disease 36.3:193-202.

Click here www.stat.purdue.edu/~doerge/BIOINFORM.D/FALL11/sem.html for a full schedule of BIOINFORMATICS SEMINARS, past and present.