In this article, we report the incidence rates of affective, non-affective and all psychoses in two tightly defined, contrasted catchment areas (rural vs. urban), in France. For both affective and non-affective psychoses, incidence rates were increased in the urban area and in the most urbanized populations of the rural area.
Our discussion will focus first on our methodological choices, strengths and potential limitations and how they could affect results. We then discuss our results in comparison with results from similar studies.
Methodological choices, strengths and potential limitations
First of all, our results are based on incident cases from secondary/tertiary care. Given the severity of these disorders and the fact that in the two areas involved in our study psychiatric services, public and private, are easily accessible, it is probable that this represents a vast majority of the cases of incident psychoses. However, as we did not specifically address the issue of cases that are seen only in primary care, our results should not be generalized to all cases of psychosis.
Urbanicity is probably a proxy for some underlying, as yet unidentified, risk factor. Several hypotheses regarding the exact risk factors have been advanced [
5]: biological, socio-economic or psychological (vitamin D insufficiency, elevated risk for infections, higher levels of everyday stresses, lower social capital and higher social fragmentation, etc.) but none has been validated to date. For this reason, there is debate on how best to define urbanicity and the period of risk associated with it [
12]. For practical reasons, we chose to define urbanicity as a function of place of residence. It has been argued that place of birth (or even the place where the foetal period of development took place) has to be used rather than the place of upbringing or of residence. However, the influence of this choice on the results is limited by the fact that urban birth and urban residence are strongly associated [
13].
The accuracy of the incidence rates depends on our capacity to identify all new cases of psychosis. Cases could be missed if patients are cared for outside the catchment area or if cases from the catchment area are not reported. In France, public psychiatric facilities offer care on a strictly catchment-area basis (“secteur psychiatrique”). The same restrictions do not apply to private practice and thus we cannot exclude that some patients are cared for outside the catchment areas. However the number of private psychiatrists in the two areas was substantial. Therefore, the probability that subjects with first psychotic episodes, who experience important behavioural, cognitive and volitional difficulties, travelled outside the area for psychiatric care seems small.
To ensure adequate reporting of new cases, we made every effort to involve a majority of the psychiatrists from the two catchment areas. Only a small proportion of psychiatrists, all with individual private practice, and all from the urban centre have declined participation to the study. This could result in an underestimation of the incidence rates. However, based on the small number of cases reported by the participating psychiatrists with individual private practice and the proportion of non-participating psychiatrists, we estimate that this could not significantly impact the reported rates. Furthermore, the impact is limited to the urban centre and thus could not affect our conclusions of greater incidence rates compared to the rural centre (or differences between populations with different levels of urbanization in the rural centre).
The procedures to identify and classify cases also deserve some comments. To identify new cases, we used a prospective, standardized methodology involving the reporting of symptoms present and not of specific diagnosis. The anonymous reporting of the cases helped to limit the proportion of patients that opposed the communication of their data. However, this also meant some inherent limitations as the number of details on each case has to be limited (to avoid indirect identification) and a test of inter-rater reliability could not been conducted.
We adopted this method to simplify reporting (and thus limit the risk of not reporting cases), enhance reliability and avoid potential problems with different diagnostic procedures used by different psychiatrists.
We decided to classify cases as affective or non-affective psychosis for several reasons. Firstly, this classification is based on a restricted number of symptoms (hallucinations, delusions, depression or elated mood) and avoids symptoms that tend to show low inter-rater agreement (e.g. formal thought disorder, affect flattening etc.) e.g. [
14,
15]. Second, longitudinal studies (e.g. [
16]) showed that first episode diagnosis could change over time in a sizable proportion of cases. However, a majority of these changes are between different diagnoses of non-affective psychoses (e.g. from brief psychotic disorder to schizophrenia) or between diagnoses of affective psychoses (e.g. from unipolar to bipolar psychosis). Furthermore, the usually short time of observation until cases have been reported would inherently lead to diagnostic uncertainties and potential misclassification had a more specific diagnostic been used (e.g. observation before reporting the case was usually less than the 6 month interval required for a definite schizophrenia diagnosis).
Summing up all these arguments, we are confident that even if an underestimation of cases cannot be ruled out, it would not significantly influence the incidence rates and, more importantly, would not bias the results between centres. A more accurate estimation of potentially missed cases could be achieved by a leakage study. Such a study is planned at end of the data collection period of the EU-GEI study, which is scheduled for mid-2014.
Comparison of our results with data from the literature
Comparing our data with previous data from France seems difficult. To our knowledge, the last published data are more than 30 years old [
17]. In the cited article, for the 1973–1982 period, first-time hospitalizations for schizophrenia were, at national level, around 10/100000 at risk population. Several important methodological differences with the present study (diagnostic criteria used, categories of diagnoses reported, population at risk, etc.) limit the interest of this comparison.
In sharp contrast with the lack of epidemiological data in France, a recent review [
3] found more than 80 reports on incidence of psychosis in England between 1950 and 2009.
Compared with data from this review, our data show several similarities: affective psychoses show lower incidences relative to non-affective psychoses, non-affective psychoses rates are elevated in men compared to women and more so before mid-life. For affective psychoses, in the cited review, there are less gender differences in incidence rates (when they exist, they are, as in our study, in the direction of greater incidence in women). The only notable difference is that our data did not show a clear peak in the twenties and second peak in the late forties for non-affective psychoses in women. More research is needed to confirm or infirm this difference in pattern.
In the cited review of studies of incidence of psychosis in England [
3] there is no study that assessed the incidence of affective and non-affective psychoses in both a rural and an urban setting. Thus, we chose for comparison data from two recent studies which showed the greatest contrast in terms of urbanicity: one from East London [
18] and one from Northumberland [
19]. With the exception of affective psychoses in the urban centre which rate is similar to ours (13.5), rates in England are higher than those we observed in France (non-affective psychoses urban 36.8, rural 17.8 and affective psychoses in the rural area 8.6) but there is a similar general pattern of more non-affective than affective psychoses and greater incidence rates in the urban as opposed to the rural site.
Studies that concomitantly explored incidence of psychoses, using the same categories as we did (i.e. affective, non-affective and all psychoses), in both rural and urban settings are more useful as a comparison for the pattern of urban–rural differences observed in our study. To our knowledge, there are only 5 such studies in the literature. Four of them, originating from north European countries, are based on national psychiatric registries and compare rates of non-affective psychoses [
20,
21] or both non-affective and affective psychoses [
8,
9] according to the degree of urbanization. With the exception of the study by Marcelis et al., that included subjects between 14 and 50 years old, the other studies had a restricted range of age for at risk population (16 to 25 for one study, over 25 for another, less than 22 for the third) thus preventing any generalization of their results. However, despite this limitation and different definitions of urban/rural exposures (either at birth or at the time of diagnosis) and levels of urbanicity the four studies suggest that incidence of non-affective psychoses (including schizophrenia) is increased by the degree of urbanization. In the Eaton et al. study [
9], for affective psychoses (including bipolar psychoses), there was not a clear trend (both extremes of urbanicity showing a small excess of cases and the middle category showing the smallest incidence rate) but this study was restricted only to cases with a first diagnosis before age 22. In the Marcelis et al., study [
8] affective psychoses showed a similar trend to non-affective psychoses (higher risk in more urban areas) but with a lower relative risk.
Because it is similar to our study in many methodological aspects, a recent study from Ireland [
5] deserves more comments. The authors used a similar design, prospectively collecting data on subjects at their first contact with psychiatric services for psychotic symptoms, in two catchment areas: one urban and one rural. A major difference from our study is that they did not use exclusion criteria based on age or aetiology (due to medical general conditions or use of toxic substances).
The results from this study are detailed by diagnostic category, gender and catchment area (urban vs. rural). They are at odds with our results and most of the results from the literature as they show, as a whole, larger incidence rates for psychoses (with the exception of schizophrenia) in the rural areas. These differences (in affective, non-affective and overall psychoses) are essentially driven by very high incidence rates in females in the rural area (36.7 per 100000). Almost half of the total cases of psychoses in this category of subjects (i.e. 18.3 per 100000) are non-affective psychoses other than schizophrenia raising the question about the contribution of secondary psychoses (due to medical conditions, to dementia or substance induced) to the data.