The aim of this study was to determine the feasibility and suitability of implementing consensually derived, evidence based, medication algorithms in the care of patients treated in the public mental health sector. METHOD: Each of 40 physicians at 16 sites treated 5-15 patients for up to four months with one of the algorithms (schizophrenia [SCZ], bipolar disorder [BPD}, or major depressive disorder [MDD]). Process measures included visit frequency, staff time, algorithm stages used, time on treatment stage, and medications used. Outcome measures included the Brief Psychiatric Rating Scale (BPRS), Inventory of Depressive Symptoms (IDS), Clinical Global Impression (CGI), and the Multnomah Community Abilities Scale (MCAS). RESULTS: 222 patients entered the study. Survival rate at 90 days was 75%. Physician time was approximately 40 minutes at the initial visit and 30 minutes at follow-up visits; visits were approximately 3 weeks apart. 52% of patients with SCZ and 55% of patients with BPD had > 30% decrease in BPRS score. 38% of patients with MDD had > 50% decrease in IDS score. All patient groups except outpatient SCZ had significant improvement on the MCAS. Physician satisfaction was high with nearly 80% of the physicians stating that they would continue using the algorithms. CONCLUSIONS: Medication algorithms were implemented in the population, and in general, were associated with good improvement in patient symptoms and function. Major obstacles to widespread implementation are adequate staffing, staff support, and availability of certain medications. Clinical and economic impact of the algorithms are currently being evaluated in a comparative study.