New Approaches to Assessing and Forecasting Morbidity With Enterovirus (Non-Polio) Infection in the Russian Federation Using Mathematical Models

Abstract

At present it is impossible to develop epidemiologic surveillance and control over any infection regarding studies on dynamics of morbidity, seasonality and periodicity without using mathematical modeling techniques. Our research goal was to study regularities in manifestations of epidemic process for enterovirus (non-polio) infection (EVnI) in the Russian Federation over 14 years (2006–2019) using mathematical models (linear, logarithmic, power, and exponential approximation).An optimal mathematical model was selected using three statistical parameters, namely determination coefficient, Fischer’s exact test, and standard error. Periodicity of rises and falls in morbidity was calculated with Fourier one- dimensional spectral analysis. Intra-year dynamics of morbidity with EVnI was estimated basing on monthly spread of the disease cases on the RF territory. Classic seasonal decomposition, Census I technique, was applied to analyze time series of monthly morbidity. It was determined that EVnI epidemic process was unevenly spread over years in the RF in the examined period of time (2006–2019) and there were two opposite trends in it; the first one lasted from 2006 to 2010 when morbidity was declining and the second was from 2010 to 2019 when it was growing. Having analyzed manifestations of EVnI epidemi- ologic process in long-term dynamics given its uneven spread as per years, we established that it was advisable to use mathematical models approximated as per separate time periods. Average long-term morbidity with EVnI amounted to 8.09 0/0000 in the RF in 2010–2019 with growth rate being equal to 17.7 %. Maximum value was registered in 2017 (16.32 0/0000). An unfavorable prediction for further epidemic situation development was revealed for the examined pe- riod. The epidemic process was characterized with 4-year periodicity and summer-autumn seasonality with peaks usually occurring in August and September. Rates that characterized intensity of the trends in long-term morbidity dynamics and were calculated with mathematical models differed authentically from those obtained via conventional calculations of average values (χ=11.08; d.f.=1; p=0.0009).