To restrict or not to restrict? Use of artificial neural network to evaluate the effectiveness of mitigation policies: A case study of Turkey

Abstract

Outbreaks, epidemics or pandemics have increased over the last years, increasing the morbidity and mortality over large geographical areas, as well as causing financial crises and irreversible social changes. Coping with emerging infectious diseases such as Covid-19, different mitigation policies are developed by countries. However, the benefit of each mitigation policy is still not well-explored due to the considerable difference between implementations of policies in each country. The question is which policies play a significant role in controlling Covid-19 transmission. Developing two models used in Artificial Neural Network, this study investigates the impact of mitigation policies or strategies (a combination of policies) by considering different vaccination and mutation scenarios. The former model requires the prediction of reproduction number based on the number of cases reported in previous days; whereas, the latter model is constructed based on the number of people impacted by a mitigation policy or strategy. Although the first model yields more accurate results, it requires the use of historical data; hence, the passage of time during a critical period of fighting against Covid-19. The benefit of the second model is that it can be implemented more quickly by determining a coefficient for each policy or strategy based on the restricted population and/or limited mobility. Testing different scenarios through a real-world example from Turkey, we find mitigation policies or strategies play a significant role in controlling Covid-19; as well as vaccination and mutation scenarios. Our results suggest continuous and predetermined mitigation policies or strategies should be implemented to control the spread of infectious diseases in addition to a successful vaccination program.