Statistical model of seasonal forecasting the completed suicides number in the regions of Ukraine

Abstract

Suicide prevention efforts require conscious coordination and close collaboration between health agencies. They should be based on an understanding of the true picture of the prevalence of this phenomenon in a particular area, the characteristics of the dynamics of changes in the frequency of suicides, and high-risk factors, namely age, gender, climatic and social components. The purpose of this study was an attempt to create for each region of Ukraine a statistical model of the dynamics of the frequency of completed suicides depending on the time component (months of the year) and to build on its basis a forecast of the dynamics of the indicator of the number of deaths due to intentional self-harm. For this, the autocorrelation of absolute indicators was carried out and correlograms of time series of indicators of the deaths' number due to intentional self-harm were constructed. The obtained correlograms had sufficiently pronounced features, which made it possible to structure them into 4 separate groups. For further analysis, we used the time series of the areas that made up the first two groups, characterized by a trend and seasonality. For further analysis, only models of exponential smoothing of the time series of areas were used, whose indicators of Ljung-Box Q-statistics, coefficient of determination, mean modulus of error, and smoothing of the mean were in an acceptable range. Based on the created time series model, it is possible to assume that the period from August 2021 to September 2022, will increase in the absolute indicator of the number of deaths due to intentional self-harm in the spring months and, for most regions, in January. For the autumn period, on the contrary, a decrease in the number of completed suicides is characteristic. the characteristics of the time series models for a whole group of regions did not allow us to use them to build a forecast. These regions are represented by two different geographical groups – a group of regions of Western Ukraine and three Black Sea regions. Created for each region of Ukraine, a statistical model of the frequency dynamics of the completed suicides depending on the time component (months of the year) made it possible to build an annual forecast for the number of deaths dynamics due to intentional self-harm. Longer-term forecasts are possible by analyzing more data.