We used classification accuracy as a simple, effective evaluation metric that is consistent with a naive bayes approach. Most of the dataset's attributes colud be selecyed as the target attribute; we focused on length of hospital stay; as a potential proxy for bote heatlth care quality and cost, and discharge status, for its obvious humanitarian implications.We also considered the number of diagnostic codes as another potential proxy for cost and severity of illness We also considered the number of diagnostic codes, as another potential proxy for cost and severity of illness. We divided the dataset into training and testing data, using 10-fold cross-validation. Cross validation is a method by which the data is randomly subdivided into subsets, in this case, ten. one subset is isolated for use as testing data and the remaining serve as training data. In successive iterations, the modl is run with each subset serving as testing data,and the results averaged. In this way, actual data values are available to calculate classification accuracy but bias is minimized.