The Iranian Drug-related Deaths with Time Series Modeling and Forecasting: 2014–2016

Abstract

Abstract Background: Investigating the temporal variations and forecasting the trends in drug-related deaths can help prevent health problems and develop intervention programs. Iran's recent policy is strongly focused on deterring drug use and replacing illicit drugs with legal drugs. This study was conducted to investigate drug-related deaths in 2014-2016 in Iran and forecast the death toll by 2019.Methods: This longitudinal study used a Box-Jenkins time series analysis to forecast referrals with a diagnosis of drug-related death. The number of referrals was extracted from March 2014 to March 2017 by month. After using appropriate data transformation methods to create a stationary time series and carrying out a more accurate examination of the stationary assumption by the Dickey-Fuller test, the ARIMA model parameters were determined using Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) graphs. Comparing the Akaike statistics of the proposed models led to the selection of ARIMA (0,1,2) as the best model for fit to the data. In the final stage, the number of deaths was forecast in Iran until 2019 using the ARIMA (0,1,2) model. The final extracted data were analyzed in R software, Minitab and SPSS-23.Results: The death toll in Iran during the entire study period of three years was 8883 according to the Iranian Ministry of Health and the Legal Medicine Organization. The number of deaths varied by year and was 2840 in 2014, 2810 in 2015 and 3233 in 2016. According to the time-series findings, data on the number of deaths showed an increasing trend and was not under any regular seasonal effects. In addition, the mean number of deaths per month in Iran until 2019 was forecast as 245.8.Conclusion: This study shows that the trend of drug-related deaths rose during the study period, and the modeling process for their forecasting suggests that this trend shall continue until 2019 if proper interventions are not instituted.