Interpretability of Epidemiological Models : The Curse of Non-Identifiability

TL;DR

This paper explores how non-identifiability affects the interpretability of epidemiological models, defining three types of identifiability issues and proposing a framework to understand and mitigate them.

Contribution

It introduces a comprehensive framework for understanding non-identifiability in epidemiological models, linking model definition, data, and fitting methods.

Findings

Identifiability issues are linked to model, data, and methodology.

Three notions of identifiability are defined and analyzed.

Framework highlights strategies to mitigate non-identifiability.

Abstract

Interpretability of epidemiological models is a key consideration, especially when these models are used in a public health setting. Interpretability is strongly linked to the identifiability of the underlying model parameters, i.e., the ability to estimate parameter values with high confidence given observations. In this paper, we define three separate notions of identifiability that explore the different roles played by the model definition, the loss function, the fitting methodology, and the quality and quantity of data. We define an epidemiological compartmental model framework in which we highlight these non-identifiability issues and their mitigation.