Despite the increasing prevalence of multiple-drug-resistant (MDR) Pseudomonas aeruginosa, the factors predictive of MDR have not been extensively explored. We sought to examine factors predictive of MDR among patients with P. aeruginosa respiratory tract infections and to develop a tool to estimate the probability of MDR among such high-risk patients. This was a single-site, case-control study of patients with P. aeruginosa respiratory tract infections. Multiple-drug resistance was defined as resistance to four or more antipseudomonal antimicrobial classes. Clinical data on demographics, antibiotic history, and microbiology were collected. Classification and regression tree analysis (CART) was used to identify the duration of antibiotic exposure associated with MDR P. aeruginosa. Log-binomial regression was used to model the probability of MDR P. aeruginosa. Among 351 P. aeruginosa-infected patients, the proportion of MDR P. aeruginosa was 35%. A significant relationship between prior antibiotic exposure and MDR P. aeruginosa was found for all of the antipseudomonal antibiotics studied, but the duration of prior exposure associated with MDR varied between antibiotic classes; the shortest prior exposure duration was observed for carbapenems and fluoroquinolones, and the longest duration was noted for cefepime and piperacillin-tazobactam. Within the final model, the predicted MDR P. aeruginosa likelihood was most dependent upon length of hospital stay, prior culture sample collection, and number of CART-derived prior antibiotic exposures. A history of a prolonged hospital stay and exposure to antipseudomonal antibiotics predicts multidrug resistance among patients with P. aeruginosa respiratory tract infections at our institution. Identifying these risk factors enabled us to develop a prediction tool to assess the risk of resistance and thus guide empirical antibiotic therapy.