BACKGROUND: Clozapine is the treatment of choice for medication refractory psychosis, but it does not benefit half of those put on it. There are numerous studies of potential post-clozapine strategies, but little data to guide the order of such treatment in this common clinical challenge. We describe a naturalistic observational study in 153 patients treated by a specialist psychosis service to identify optimal pharmacotherapy practice, based on outcomes. METHODS: Medication and clinical data, based on the OPCRIT tool, were examined on admission and discharge from the national psychosis service. The primary outcome measure was the percentage change in mental state examination symptoms between admission and discharge and the association with medication on discharge. Exploratory analyses evaluated the specificity of individual medication effects on symptom clusters. RESULTS: There were fewer drugs prescribed at discharge relative to admission, suggesting an optimisation of medication, and a doubling of the number of patients treated with clozapine. Treatment with clozapine on discharge was associated with maximal decrease in symptoms from admission. In the group of patients that did not respond to clozapine monotherapy, the most effective drug combinations were clozapine augmentation with 1) sodium valproate, 2) lithium, 3) amisulpride, and 4) quetiapine. There was no support for a dose-response relationship for any drug combination. CONCLUSIONS: Clozapine monotherapy is clearly the optimal medication in medication refractory schizophrenia and it is possible to maximise its use. In patients unresponsive to clozapine monotherapy, augmentation with sodium valproate, lithium, amisulpride and quetiapine, in that order, is a reasonable treatment algorithm. Reducing the number of ineffective drugs is possible without a detrimental effect on symptoms. Exploratory data indicated that clozapine was beneficial across a range of symptoms domains, whereas olanzapine was beneficial specifically for hallucinations and lamotrigine for comorbid affective symptoms.