Provider's practice behaviors before and after physician and staff training in the use of a schizophrenia medication algorithm and the effects of education on physician adherence to the algorithm were evaluated. Medical records of 30 patients admitted between September 1 and November 30, 1999, and 30 patients admitted from September 1 to November 30, 2000, with an admitting and discharge diagnosis of schizophrenia and a minimum length of stay of 14 days were randomly selected and analyzed. Clinical data, including prescribed psychotropic medications and dosages, documentation of target symptoms and severity, adverse drug effects, appropriate clinical ratings, patient's response to treatment, and reason for medication change, were collected and compared with the recommendations in the schizophrenia medication algorithm. Efforts to implement the schizophrenia algorithm included staff education and uniform documentation. Progress notes were evaluated before and after training. After physician and staff training, only 5 of 359 progress notes were written using the recommended documentation form. The number of progress notes containing no documentation of symptoms decreased from 66 to 41, and those documenting three to five target symptoms increased from 74 to 140. Documentation of physician assessment of the presence or absence of adverse effects and their severity decreased from 35.2% to 18.7% and from 22.3% to 17.0%, respectively. Physicians increased the documentation of their clinical global impressions from 12.1% to 20.3%. The recording of medication changes increased twofold, but the difference was not significant. Physician and staff education alone did not significantly alter providers' practice behavior. Inadequate and inconsistent documentation of clinical outcomes made it difficult to assess physician adherence to the treatment algorithm.