To determine the variation in prevalence of common mental disorders and general practitioner (GP) responses across Australian general practices, and to identify practice characteristics that predict these rates. Cross-sectional national audit of general practices throughout Australia in 1998-1999. 46515 ambulatory care patients attending 386 GPs. Practice-based prevalence of mental disorders (based on patient-reported symptoms) and GP-reported rates of psychological diagnoses and treatment (median and range, excluding the upper and lower 10% of practices); practice characteristics (patient, doctor and organisational) that predict prevalences and rates, determined by multiple regression analysis. Even after omitting the upper and lower 10% of practices, there were large variations between practices in prevalence of common mental disorders (range, 39% to 59% of patients; median, 48%), and substance misuse (range, 3%-13%; median, 7%). There were also large variations between practices in rates at which GPs made psychological diagnoses in each practice (range, 12%-51%; median, 27%), judged patients to be at risk to self or others (range, 6%-54%; median, 23%), provided psychological treatments (range, 8%-41%; median, 22%) and referred patients to specialist services (range, 1%-10%; median, 4%). Practice-based rates of disorders and GP responses were predicted not only by sociodemographic characteristics of patients in each practice (eg, mean age or proportion of unemployed people), but also by doctor characteristics (eg, age and sex) and practice organisation characteristics (eg, urban versus regional or rural location). We identified patient, GP and practice characteristics that predict rates of mental disorder and treatments provided. These could be used to guide mental health service reform in general practice and assist with targeting relevant education and practice support programs.