Suicide rates and income in São Paulo and Brazil: a temporal and spatial epidemiologic analysis from 1996 to 2008

Suicide rates vary widely between and within countries, since it is a complex phenomenon, related to several singularities. Some general demographic risk factors are known as sex (men), age-strata (young, elderly) and ethnicity (European). Other factors contribute to suicide and include: genetic loading, personality characteristics (impulsivity, aggression), psychiatric and physical disorders (pain, incapacity), life events (loss, trauma), social isolation, availability of means, substance abuse, economic condition [1, 2].

The highest suicide rates are in Eastern Europe (former USSR countries) and the lowest rates, in some Latin American countries [1, 3]. Disparate geographic distribution of suicide has been recognized for the past two centuries, with the first seminal observations of Morselli [4] and Durkheim [5], who acknowledged an endemic pattern of suicide in the late XIX century in Europe. Morselli noted a higher suicide rate in Denmark and central Germany, while Durkheim observed the same in northern France, a finding confirmed recently by Baller & Richardson [6] in an analysis using the same data.

The first population-based suicide studies date from the early nineteenth century Europe, highlighted through the work of Jean-Pierre Falret, Esquirol disciple [7, 8]. Later, many other studies emerged as led by Morselli, Masaryk, Guerry, Tarde, Winslow, Wagner [4, 5, 7, 9]. Methodologically, the most consistent work belonged to Emile Durkheim, which combined the available empirical data with a well-defined sociological theory [5]. "The suicide”, Durkheim's masterpiece, inaugurated the modern sociology and was one of the first ecological studies, a major influence in epidemiology. Durkheim's theory is based on two concepts: social integration and social regulation. Suicidal behavior is common in societies where there is a low degree of social integration, culminating in the egoistic suicide. The individual is protected from egoism by religions with strong group ties (e.g. Catholic Church) and family ties. Suicidal behavior is also common in societies where there is a low degree of social regulation, culminating in the anomic suicide. Social regulation can be understood as external regulatory forces on the individual. Economic cycles (depression or prosperity) and income level are examples of factors that could modulate anomic suicide. Durkheim is commonly criticized for not providing a specific explicit definition of these social variables. Some derivations, extensions and reinterpretations of his theory were attempts to overcome such omission [10‐13]. Despite the criticism, he remains one of the most well-known names of suicidology [14]. Sociologists, psychologists, epidemiologists, psychiatrists have used the same basic methodology established by Durkheim [3]. In the late XIX century, France history was marked by great economic development, which was accompanied by one of the highest suicide rates ever observed. When Durkheim claimed that "poverty protects against suicide," he based his observation on two area-based comparisons. He noted higher suicide rates in France compared to poorer countries, and also that suicide rates were higher in regions of France with greater wealth concentration. This finding raised the hypothesis that economic development could be related to individualism and, ergo, to social isolation and suicide. We aimed to evaluate the same relationship in Brazil and São Paulo regions using Geographic Information System (GIS) and statistical technics. However, the relationship between wealth and suicide is not straightforward; on the contrary, it is complex and changes throughout time and space. In Europe, for instance, different patterns were observed at the beginning of the twentieth century [15].

In this context, several methodological aspects hinder analysis of this phenomenon. For instance, different study designs have been used to evaluate suicide rates in relation to socioeconomic factors, although most have been ecological. There are also multiple indexes that can be used to evaluate socioeconomic characteristics (e.g. poverty/deprivation, unemployment, Gross Domestic Product (GDP), average income, etc.). Also, most studies were performed in developed countries. Partially because of these questions, results have been heterogeneous.

A systematic review of ecological studies in North America, Europe, Australia, and New Zealand, dating from 1897 to 2004, reported an inverse relation between socioeconomic characteristics and suicide [16]. Other recent ecological studies performed in the United States [17], Japan [18], Taiwan [19], Australia [20], England [21], Finland [22, 23] and Italy [24] also demonstrated an inverse relationship. Moreover, an ecological study that grouped data from the G7 countries in 2007 observed an inverse relation between income and male suicides [25]. However, other ecological studies found different results. One worldwide, cross-sectional study performed in the late 1990s included data from 52 countries and identified a direct association between per capita GDP and male suicide rates in all regions except former socialist economy countries, which had abnormally high suicide rates [3]. Another study using World Health Organization (WHO) data focused on countries with a medium Human Development Index (HDI) and found that education and telephone density was directly related to suicide while a high Gini index was inversely related to suicide [26].

Longitudinal studies also present interesting findings. One study using data from 1960 to 1999 included data from 21 countries grouped into three categories according to GDP per capita; results showed notably that countries with higher income presented higher suicide rates than poorer countries [27]. Other longitudinal studies also observed this trend [28‐30].

Therefore, the relationship between income and suicide varies, although it seems that it is mediated by socioeconomic development and income level. In this context, Brazil covers a wide area with almost 200 million inhabitants with very distinct cultural and socioeconomic characteristics, as shown by indexes such as the Gini coefficient and the HDI. In addition, although the overall suicide rate in Brazil is low, in Brazilian urban centers the rates vary from 5 to 15 per 100,000 [31‐33]. Few studies have evaluated temporal trends of suicide in Brazil [34, 35] and none has explored its relationship with income including spatial distribution.

Considering that (I) suicide and income are related, (II) Brazil presents important socioeconomic disadvantage, and (III) suicide rates in Brazil vary among location, the purpose of our study was to evaluate the relationship between suicide and income in Brazil and its regions. We also verified whether such relationship is related to socioeconomic characteristics.

Our study used ecological data to explore whether income and suicide were related in Brazil. Although ecological studies are considered methodologically problematic since they are based on aggregate data, ours has the strength of performing not only descriptive but also temporal and spatial analyses. While descriptive analysis assesses the magnitude of association between suicide and income in the different regions, the temporal analysis takes into account all these attributes over time. Finally, spatial analysis surveys for clusters of high suicide risk within regions. Taken together, these are useful tools for the evaluation and implementation of suicide interventions.

The descriptive approach showed that for Brazil and the city of SP, suicide rates were higher in wealthier areas; the opposite relation was found for the State of SP, i.e. the wealthiest area had the lowest suicide rate. Remarkably, this was confirmed by spatial analysis showing similar associations for all regions – i.e., a direct relation between income and suicide in Brazil and the city of SP, and an inverse relation for the State of SP.

Temporal analysis also showed interesting findings – one being that trends for different areas in Brazil and the State of SP converged at the end of the study period. This could be related to general improvements in health care and citizenship observed in Brazil in the last decade. On the other hand, the city of SP showed no such convergence. This could be related to the fact that the city maintained its high income inequality over the study period. Spatial analysis identified clusters with high and low suicide rates in Brazil, State of SP, and city of SP. Brazil had the highest relative risk of suicide, and also the lowest relative risk, comparing the south versus east/northeast regions. Considering the clusters of high suicide rates, the State of SP showed the lowest relative risk.

One important finding was that income and suicide directly related in Brazil and in the city of SP, but not in the State of SP, where suicide was inversely related with income. Interestingly, the State of SP presents a lower Gini index than the other regions. A systematic review of ecological studies from 1897 to 2004 showed that most studies (70%) reported an inverse association between income and suicide; most (88%) were also performed in developed countries [16]. Recent ecological studies conducted in developed countries show similar patterns, the majority finding an inverse relationship between suicide and income [17, 18, 20‐23, 25, 44‐46].

The same was noted in longitudinal approach studies [47‐49] and using GIS techniques [19, 50]. The spatial clusters of suicide in Japan were assessed using data from 1985 and 1995 for prefectures. Researchers used GIS techniques and models showed that suicide rates were higher in less populated areas, areas with more elderly persons, and in more religious areas. They suggested an influence of traditional Japanese culture that endorses suicide [51]. A study identified a marked change in the spatial epidemiology of suicide in young men from 2001 to 2005. Suicide rates in London had declined while they had increased in Wales [52]. In London neighborhoods, from 1996 to 1998, a study identified spatial dependency and a direct relationship between deprivation and suicide in males aged 30–49 years [53].

Some studies found a direct relation between suicide and income [3, 26]. The lack of studies in developing countries still persists; in China (Shandong) an inverse relationship between suicide and income was identified [54]. Conversely, a study focusing on developing countries revealed a direct association between income (indexed by high education levels, high telephone density, and high cigarette consumption) and suicide. One possible explanation for these findings is that, in developing countries, social deprivation might be more accepted by the population, while in developed countries notions of self-worth and equality are valued [26]. Another interesting explanation is proposed by Paugam [55], after Castel [56]: authors who distinguish between disqualifying and integrated poverty. The latter, which best suits Brazil and the city of São Paulo (higher Gini coefficient), is observed in societies in which poverty is compensated by a solidarity response within family, neighborhood, and region. Integrated poverty was, in fact, the social reality that Durkheim referred when he claimed that “poverty protects” [3]. On the other hand, disqualifying poverty, which is more common in developed countries, might be related to demotion and loss of social status. Becoming poor in a rich society is, in fact, different than being poor in a poor society, and such loss of social status might lead to frustration and later to suicide [13].

The present study has some limitations. Different procedures and cultural and social practices and values probably have various effects on death records and lead to misclassification of suicide [1, 2]. Another aspect is related with data quality. It is possible that the quality of the information have influenced suicide trends in Brazil. A study based on vital information provided by the Ministry of Health Brazilian system identified more deficient information in the poorest macro-regions [57], although this issue has improved in recent years. So, it is possible that the significant suicide increase in area 3 in Brazil could be an artifact. For the nation-wide comparison, we did a sensitivity analysis, based on two periods: 1996–2002 and 2003 onwards. We focused on area 3 because it was the poorest and presented the highest variation on suicide rates. We also analyzed the age-adjusted death rates by "Ill-defined and unknown causes of mortality" (code R99 of ICD-10). For the first period (1996–2002) suicide remained stable and death by “R99” showed a significant increase (APC = +12.2). For the second period (2003–2008) suicide increased significantly (APC = +3.7) and death by “R99” decreased significantly (APC = −15.2). There was an improvement in data quality from 2003 onwards, and also a slightly and significant increase on suicide rates. Thus, partially, we can assume that the significant suicide increase in area 3 in Brazil was due to an artifact, but suicide is a rare event, the APC was lower. Further research must be conducted to clarify these patterns. Spatial analysis using satscan has a limitation. For each area unit, the software considers the respective centroid and its radius, according to the suicide rate. This can bias the cluster detection, mostly in large areas with spatial variations in the population density. Despite this limitation in a recent review of software for space-time disease surveillance (include ClusterSeer, SaTScan, GeoSurveillance, Surveillance package for R), satscan was highlighted as the most developed and robust software for cluster detection [58].

In conclusion, we performed a descriptive, temporal, and spatial analysis regarding income level and suicide in different regions of Brazil. Results showed that the direction of association varied, albeit resembling similar patterns observed worldwide, with a direct and inverse association in less and more unequal areas, respectively. Taken together, these findings suggest that income and suicide in Brazil are related, although socioeconomic regional characteristics might act as a moderator variable. Future studies comparing suicide rates in other Brazilian States and developing countries should take into account these results by performing separate analyses by region. Also, these results highlight that the association between suicide and income in Brazil varies importantly according to the considered region. The result of the present study is not conclusive. It helps to understand the phenomenon of suicide, generates new hypotheses for other study designs and can be useful for prevention strategies.