Re-examination of the seasonality of suicide in Taiwan during 1991–2008: A population-based study

Abstract

The purpose of the present study was to re-examine the monthly distribution of suicide death rates and the change of the suicide seasonality in Taiwan during 1991–2008. The monthly suicide death rates of the total, male and female populations during January 1991–December 2008 were obtained from a population-based database. There was an upward trend. The means of the monthly suicide rates after adjusting the calendar effect and the upward trend were compared. The suicide data were further analysed in three 6-year periods to explore the change of seasonality. During 1991–2008 in Taiwan, suicide rates had significant peaks in April and May, and nadirs in January and December. The monthly distributions varied during 1991–1996, 1997–2002 and 2003–2008 in the patterns and the statistical significance. There was no linear trend for the change of the seasonality of suicide death rate.

Introduction

Suicide, a major threat to public health, is a complicated phenomenon. Among many other factors such as psychiatric disorders, gender, age, socioeconomic conditions, life events and region, season/climate also has been considered to play a role in triggering suicidal behaviour (Petridou et al., 2002). However, studies on the seasonality of suicide showed discrepancies between their findings. Some researchers reported a diminishing trend (Yip et al., 2000, Yip & Yang, 2004, Ajdacic-Gross et al., 2005a, Ajdacic-Gross et al., 2005b), while others addressed increasing seasonality (Voracek et al., 2002, Rock et al., 2003). Different statistical methods for analysing the seasonality of suicide might be one of the reasons (Hakko et al., 2002). Taking Taiwan for example, both Lee et al., 2006, Lin et al., 2008, using autoregressive integrated moving average (ARIMA)/seasonal ARIMA model, found a statistically significant peak in the spring, while Ho et al. (1997), using harmonic analysis, found a statistically significant nadir in the winter months. Although some researchers suggested that the delineation of seasonal variation in suicide rate should rely on different approaches (Altamura et al., 1999), speculation whether the preference of using certain statistical methods is due to the fact that they may lead to easier publication of study findings existed (Hakko et al., 2002). Then, which statistical method is more relevant for suicidal seasonality? If the aim was to explore the unknown periodic rhythm or complex temporal structures or to predict the secular trend, ARIMA was suitable (Munoz-Tuduri et al., 2006), while harmonic analysis was preferred if the periodic frequencies of the seasonality, say annual cycle and its harmonics, were predetermined (Hakko et al., 2002). However, Altamura et al. (1999) suggested that if the goal is simply “to find the significant differences in suicide between months,” analysis of variance (ANOVA) is appropriate.

The purpose of the present study was to re-examine the monthly distribution of suicide death rates in a longer period than the previous studies in Taiwan mentioned above (Ho et al., 1997, Lee et al., 2006, Lin et al., 2008) since longer time series could easily distinguish random fluctuations from systematic fluctuations. In addition, as the trend of the suicide seasonality remained controversial, the change of the seasonality was also of interest.