Epidemiology, the study of factors governing the frequency of disease in the community, can be the first step in understanding why people fall ill. Clinical samples are unrepresentative of what illnesses look like in the community, whereas community surveys tell one how frequent disorders are and examine their correlates in detail.

In psychiatric epidemiology, many large and expensive studies have been conducted over the past 25 years, particularly in the USA. Research began with the Epidemiological Catchment Area Study (Robins ' Regier, 1991), expanded in the National Comorbidity Studies (Kessler et al, 2005a,b), and was augmented by large-scale studies of substance misuse (Grant et al, 2004). Yet many of the conclusions that have been drawn by these studies remain doubtful. Some of the methodological problems include an almost complete reliance on self-report, the use of cross-sectional assessments without follow-up, and the use of partially trained research assistants (Newson et al, 2010). But the most worrying issue is that so much of psychiatric epidemiology has depended on diagnosis as defined by our current classification systems. No one knows whether diagnoses, such as those derived from the DSM or the ICD, are valid in the same way as medical conditions. When research has moved beyond these diagnoses, it has had to assess the frequency of symptoms considered to be their subclinical manifestations (e.g. Merikangas et al, 2007). In the absence of any gold standard for determining the boundary between normality and pathology, this approach leads to a consistent inflation in estimates of prevalence.

Inflated prevalence

Grant reviewers have to read a large number of submissions and easily get tired. That is why researchers learn to write the first paragraph of their applications to grab the reader's attention. The most common strategy is to support the importance of research by describing the prevalence of the condition under study and its impact on public health. For example, if the grant concerns alcoholism, reviewers can be told that this condition affects 10% of all men (Robins ' Regier, 1991). If the research concerns major depression, reviewers can be informed that surveys find that 13% of all men and 21% of all women will meet criteria sometime during their lifetime (Kessler et al, 1993).