Determining whether a class of random graphs is consistent with an observed contact network.

Area of Science:

• Epidemiology
• Network Science
• Mathematical Modeling

Background:

• Understanding infectious disease transmission relies heavily on network structure.
• Previous models often simplify complex contact patterns.
• Sensitivity analysis is crucial for assessing model robustness.

Purpose of the Study:

• To develop a general method for analyzing network structure's impact on epidemic attack rates.
• To quantify the epidemic potential of networks using network reliability statistics.
• To assess the generalizability of simulation results across different network structures.

Main Methods:

• Employed Moore and Shannon's network reliability statistic to measure epidemic potential.
• Generated synthetic networks using exponential random graph models based on Add Health survey data.
• Compared epidemic attack rates and temporal dynamics between original and synthetic networks.

Main Results:

• Significant differences observed in expected infections between the original Add Health network and generated models.
• Re-calibration of transmissibility could match attack rates but not temporal outbreak behavior.
• Small network perturbations disrupted re-calibration, highlighting model sensitivity.

Conclusions:

• Epidemic simulations on one network structure are unreliable for predicting outcomes on another.
• Network structure and individual transmissibility are not easily separable from population attack rate data.
• Reliable estimation of dynamical processes requires accurate specification of contact network models.