Geography of suicide in Taiwan: Spatial patterning and socioeconomic correlates

Abstract

In industrialised Western nations suicide rates tend to be high in inner city areas and socially fragmented neighbourhoods. Few studies have investigated spatial variations in suicide in non-Western settings. We estimated smoothed standardised mortality ratios (1999–2007) for suicide for each of the 358 Taiwanese districts (median population aged 15+: 27,000) and investigated their associations with area characteristics using Bayesian hierarchical models. The geographic distribution of suicide was similar in men and women; young people showed the greatest spatial variation in rates. Rates were highest in East Taiwan, a mostly mountainous rural area. There was no evidence of above average rates in large cities. Spatial patterns of method-specific suicide rates varied markedly, with solids/liquids poisonings showing the greatest geographic variation and hangings the least. Factors most strongly associated with area suicide rates were median household income, population density and lone-parent households. Spatial patterning of suicide in Taiwan differed from that observed in Western nations. Suicide prevention strategies should take into account unique local patterns.

Introduction

There are striking geographic variations in the incidence of suicide. Across Europe, national suicide rates differ more than 10-fold (Levi et al., 2003). Within the same country, the incidence of suicide also varies between regions and has distinct features in its geographic distribution (Gunnell, 2005). Previous studies have investigated associations of area characteristics with suicide rates within a country or region, particularly differences between rural and urban areas (Hirsch, 2006), but few have examined the spatial patterning of suicide. In a recent study which mapped variation across small areas in England and Wales, higher suicide rates were found in central parts of cities and remote/coastal areas (Middleton et al., 2008). Such spatial variation is associated with area indicators of social fragmentation and deprivation (Congdon, 1996, Middleton et al., 2004, Rezaeian et al., 2005). A recent review, mostly of studies from Western countries, shows that area suicide rates are higher in areas characterised by high levels of poverty, unemployment, low educational achievement and lower occupational social class (Rehkopf and Buka, 2006).

Considerable intra-country variations in suicide rates are also seen in some non-Western countries. For example, in China rural suicide rates are three times unban rates (Phillips et al., 2002). However, few studies have investigated the spatial patterning of suicide and its possible contributors in non-Western settings and most of them have used data for rather large geographic units. Spatial analyses at a fine scale are most useful for investigating the geographic pattern of suicide (Middleton et al., 2008) and its association with area socioeconomic characteristics (Rehkopf and Buka, 2006). Small geographic units have more internal homogeneity than large units and their aggregate socioeconomic characteristics are thus more likely to reflect the nature of social environment where people live. Appropriate statistical approaches such as Bayesian hierarchical models can be used to tackle the problem of uncertainty in estimates in small-area analyses, as well as potential spatial autocorrelation in the data (i.e. rates in geographically close areas are likely to be related) (Besag et al., 1991, Lawson et al., 2000); however, they have only been used in a limited number of studies of suicide (Congdon, 1997, Middleton et al., 2008, Pirkola et al., 2009).

We used Bayesian hierarchical models to investigate the spatial patterning of suicide by sex, age group and suicide method across 358 districts in Taiwan, an East Asian country with a population of 23 million, over the period 1999–2007. We also investigated associations with a range of area socioeconomic characteristics.