We analyzed factors associated with the discharge diagnosis of delirium among 1,285 patients admitted to a major teaching hospital during a 2-year period, developed a model to classify the risk of developing delirium on the basis of clinical and diagnostic data, and tested the model on 471 patients admitted during the subsequent year. Using the multivariate technique of recursive partitioning, we identified four factors that distinguished 80% of all cases of delirium: 1) a urinary tract infection at any time during the hospital stay (odds ratio = 3.1; 95% confidence interval = 2.02-4.58); 2) no urinary tract infection, but low serum albumin on admission (odds ratio = 2.4; 95% confidence interval = 1.43-3.99); 3) neither urinary tract infection nor low serum albumin, but elevated white blood cell count on admission (odds ratio = 1.99; 95% confidence interval = 1.18-3.37); 4) none of these risk factors, but proteinuria on admission (odds ratio = 1.82; 95% confidence interval = 1.25-2.66). Patients without any of these four risk factors had the lowest probability of developing delirium during their hospital stay. Among individuals with delirium, in-hospital mortality and hospital charges were higher. The model developed accurately characterized the risk of delirium when it was tested on patients admitted to the same hospital during the subsequent year.