Focusing a viral risk ranking tool on prediction

Area of Science:

• Epidemiology
• Infectious Disease Dynamics
• Public Health Preparedness

Background:

• Emerging infectious diseases pose a significant global threat, necessitating tools for early risk assessment.
• The SpillOver: Viral Risk Ranking tool was developed to predict zoonotic spillover risk from wildlife to humans.
• Current risk factors in the tool rely on post-spillover data, limiting its predictive power for novel viruses.

Purpose of the Study:

• To re-evaluate the SpillOver tool's risk ranking performance after removing spillover-dependent risk factors.
• To assess the impact of excluding post-spillover knowledge on the tool's predictive accuracy for novel viruses.
• To identify more suitable, non-spillover-dependent risk factors for future iterations.

Main Methods:

• Reanalyzed the SpillOver tool's risk rankings after removing 8 of 31 spillover-dependent risk factors.
• Compared the predictive performance (Area Under the Receiver Operating Characteristic Curve) of original vs. adjusted risk scores.
• Analyzed mean and standard deviation of risk scores for human vs. non-human viruses at the factor level.

Main Results:

• Predictive accuracy for classifying viruses as human pathogens significantly decreased (AUC from 0.94 to 0.73) after removing spillover-dependent factors.
• Excluded spillover-dependent factors showed distinct means between human and non-human virus classifications.
• Non-spillover-dependent factors often exhibited similar means across both classifications, indicating limited discriminatory power.

Conclusions:

• The original SpillOver tool's reliance on post-spillover data hinders its ability to predict risk for novel viruses.
• Excluding spillover-dependent factors significantly reduces the tool's predictive capability.
• Future versions should prioritize non-spillover-dependent risk factors to enhance prediction accuracy for novel zoonotic threats.