Stopping the SuperSpreader Epidemic, Part III: Prediction

TL;DR

This paper explores prediction methods for super-spreader driven epidemics, highlighting biases that cause false negatives and proposing improved data collection and techniques to enable early warnings and better public health responses.

Contribution

It demonstrates that despite biases, early prediction of super-spreader epidemics is feasible with better data and methods, allowing timely intervention.

Findings

Prediction bias leads to false negatives in epidemic forecasting.

Improved data gathering reduces bias and enhances early warning capabilities.

Modeling can provide early alerts long before pandemic onset.

Abstract

In two previous papers, I introduced SuperSpreader (SS) epidemic models, offered some theoretical discussion of prevention issues, and fitted some models to data derived from published accounts of the ongoing MERS epidemic (concluding that a pandemic is likely). Continuing on this theme, here I discuss prediction: whether, in a disease outbreak driven by superspreader events, a rigorous decision point---meaning a declaration that a pandemic is imminent---can be defined. I show that all sources of prediction bias contribute to generating false negatives (i.e., discounting the chance of a pandemic when it is looming or has already started). Nevertheless, the statistical difficulties can be overcome by improved data gathering and use of known techniques that decrease bias. One peculiarity of the SS epidemic is that the prediction can sometimes be made long before the actual pandemic onset, generating lead time to alert the medical community and the public. Thus modeling is useful to overcome a false sense of security arising from the long "kindling times" characteristic of SS epidemics and certain political/psychological factors, as well as improve the public health response.