Comment on paper by Cohen, Patel, Thara, and Gureje.

I agree with the authors’ criticism of the use of the dichotomy between developed and developing countries, partly because of the difficulty in defining these terms and partly due to the myriad different social, cultural, and economic factors subsumed by them. While ‘‘low income’’ and ‘‘middle income’’ can be reasonably accurately defined, they also encompass a great diversity of factors, both within and between countries. In addition, the authors have aggregated 23 studies including prevalence and incidence samples and prospective and retrospective designs. They acknowledge that a meta-analysis is ruled out by this diversity of sampling procedures and methods but nevertheless proceed to treat these studies as providing equally informative findings. An incidence study is likely to miss a small proportion of individuals fulfilling the selection criteria—11% in the AESOP study which used case finding procedures based on those in the International Studies of Schizophrenia (ISoS) research. However, a prevalence study will fail to include a high proportion of people who experience an acute first onset of schizophrenia from which they recover completely, thus introducing a bias toward chronicity. The International Pilot Study of Schizophrenia was based on prevalence samples because its aim was to determine whether it was possible to train psychiatrists from different countries to use assessment instruments in a reliable way, to establish whether schizophrenia exists in all the cultures studied, and to determine whether an international collaborative study in psychiatry was achievable. The success of this venture paved the way for the Determinants of Outcome of Severe Mental Disorders (DOSMeD), the main strength of its design being the collection of an epidemiologically based incidence sample followed prospectively in each center using the same instruments. Cohen and colleagues have taken a step backwards in conflating results from both incidence and prevalence studies. These authors state that ‘‘Except for the China ISoS site, sampling in all the WHO studies relied on a variety of help-seeking agencies to identify potential subjects.’’ I am particularly familiar with the Chandigarh site from the DOSMeD study because I visited it several times and went on field trips to the rural areas with the researchers. The city of Chandigarh has a highly literate population, 70% during the period of the study, and a Postgraduate Medical Institute of considerable sophistication in which the psychiatric facility was sited. The proportion of incident cases derived from help-seeking agencies would consequently be minimal. By contrast, the rural areas around the city have populations with a low level of literacy, 30% at the time, and limited access to medical facilities. To deal with this problem, Professor Wig, the director of the center, established a mobile team of psychiatric professionals who made regular circuits of the rural areas, holding outpatient clinics to identify and treat potential subjects for the study. This procedure increased the likelihood that all incident cases were identified. It is noteworthy that the data from the Nigerian center in Ibadan were given less weight than those from other centers in the ‘‘developing’’ countries because the casefinding procedures, compared with those in Chandigarh, were not considered to be sufficiently comprehensive.