Forecasting the Epidemic Process of HIV Infection using ARIMA and Microsoft Excel Tools

Abstract

Relevance: The epidemiological situation on HIV infection in the Russian Federation, in particular in the Tyumen Region, remains very tense. Due to the uncertainty of the situation, modeling and forecasting the evolution of the epidemic occupy a special place in the epidemiology of the disease. Objective: To forecast the epidemic process for the years 2021 to 2025 based on the results of a retrospective analysis of HIV incidence in the Tyumen Region in 1993–2020 using two different tools (ARIMA and Microsoft Excel) and to compare predictions with the rates actually observed in 2021–2023. Materials and methods: The study was conducted in March to May 2021 by applying a complex of epidemiological, statistical, and mathematical research methods. For predictive modeling, hierarchical models of the ARIMA family were used with the inclusion of two levels of hierarchy (district and city). Microsoft Excel was used to construct a polynomial trend line. Results: Based on the output of the ARIMA model, the incidence forecast cartogram was created. According to statistics for the Tyumen Region, HIV incidence rates in the years 2021 and 2022 were 67.9 and 71.4 per 100,000 population, respectively. The predicted rate in 2021 was 52 per 100,000 population (80 % prediction interval (PI): 29 to 76 per 100,000; 95 % PI: 16 to 89 per 100,000 population). When constructing a polynomial trend line (approximation coefficient R² = 0.7497), an increase in HIV incidence was predicted in the region. Conclusions: The comparison of two predictive methods shows that ARIMA results are more consistent with actual data, which means that the ARIMA model has an advantage in forecasting.