Country-level pandemic risk and preparedness classification based on COVID-19 data: A machine learning approach.

Area of Science:

• Machine Learning
• Epidemiology
• Public Health

Background:

• Accurate country-level risk assessment is crucial for managing public health crises like COVID-19.
• Existing methods may not fully leverage geopolitical and demographic data for predictive risk modeling.
• Understanding transmission, mortality, and testing capacity risks is vital for targeted interventions.

Purpose of the Study:

• To develop and evaluate a novel three-stage machine learning strategy for country-level COVID-19 risk classification.
• To predict country risk based on geopolitical and demographic data, independent of real-time epidemiological data.
• To identify the most effective machine learning models for assessing transmission, mortality, and testing inability risks.

Main Methods:

• Utilized K% binning (K=25) to categorize countries into four risk groups (low, medium-low, medium-high, high) based on COVID-19 cases, deaths, and tests per million population.
• Employed leave-one-country-out cross-validation to benchmark various machine learning algorithms.
• Developed ensemble models, including Stacks of Gradient Boosting and Decision Trees, Support Vector Machines and Extra Trees, and standalone Gradient Boosting for different risk predictions.

Main Results:

• Identified distinct machine learning models for predicting transmission, mortality, and testing inability risks.
• Observed that high risk of inability to test often correlates with lower transmission and mortality risks, necessitating a specific interpretation order.
• Found that the model's performance weakened with data from September 2020, suggesting reduced utility as international collaboration increased.

Conclusions:

• The proposed machine learning strategy offers a framework for country-level risk classification using non-epidemiological data.
• The effectiveness of such models may diminish over time due to evolving global dynamics like increased international collaboration.
• Interpreting the risk of inability to test should precede the assessment of transmission and mortality risks for a comprehensive understanding.