Projecting Course COVID-19 Turkey

Projecting Course COVID-19 Turkey projects the course of the COVID-19 pandemic in Turkey using a Bayesian negative binomial multilevel model with mixed effects to quantify uncertainty in SARS-CoV-2 case trajectories.

Key Features:

• Bayesian Negative Binomial Multilevel Model: Employs a Bayesian negative binomial multilevel model with mixed effects to model count data while incorporating uncertainty and overdispersion.
• Mixed Effects Modeling: Accounts for fixed and random variations across different levels of data aggregation to capture hierarchical heterogeneity.
• Prediction Intervals (PI): Produces 80%, 95%, and 99% prediction intervals to express uncertainty in short-term projections.
• Predictive Validity under Policy Stability: Demonstrates approximately 95% predictive accuracy within the first 12 days contingent on stable diagnostic and management protocols, and notes sensitivity to significant deviations in policy or epidemic dynamics.

Scientific Applications:

• Epidemiological Forecasting: Projects near-term COVID-19 case trajectories for Turkey to inform surveillance and trend analysis of SARS-CoV-2 transmission.
• Policy Evaluation: Assesses potential impacts of varying compliance with containment measures on projected case counts.
• Resource Allocation: Provides short-term estimates useful for anticipating healthcare resource needs such as ICU capacity.

Methodology:

Fit all available observations up to a specified date and project outcomes 20 days into the future, with dynamic updates as new data become available.