Estimating the prevalence of mental illness within the general health care population has proven problematic as the result of significant variations in methodological approach such as the use of measurement tools (Mayou and Sharpe 1991). Prevalence rates have frequently been estimated based on mental illness symptoms (self- reported or observed), rather than formal diagnoses. Symptoms of depression may be confounded with physical illness symptoms (Parker et al 2001). Selection bias effects may also lead to an underestimation of the prevalence of psychiatric co-morbidity. Estimates based on discharge diagnoses or consultation files are particularly susceptible, because: a) the psychiatric disorders are not recognised by non-psychiatric staff; b) some physicians avoid a psychiatric diagnosis for fear of patient stigmatisation; and, c) often only the most serious cases are attended to (Wancata et al 2001). Overall, these methodological shortcomings are likely to lead to an underestimation of true prevalence rates.