Spatio-Temporal Analysis of Surveillance Data

TL;DR

This paper reviews and critiques space-time models for surveillance count data, focusing on discretized time models like TSIR and epidemic/endemic models, and demonstrates their application on measles data using Stan.

Contribution

It provides a comparative analysis of TSIR and epidemic/endemic models implemented in Stan for space-time surveillance data.

Findings

Both models effectively analyze measles spread.

Discretized time models are suitable for infectious disease data.

Stan implementation facilitates model comparison and analysis.

Abstract

In this chapter, we consider space-time analysis of surveillance count data. Such data are ubiquitous and a number of approaches have been proposed for their analysis. We first describe the aims of a surveillance endeavor, before reviewing and critiquing a number of common models. We focus on models in which time is discretized to the time scale of the latent and infectious periods of the disease under study. In particular, we focus on the time series SIR (TSIR) models originally described by Finkenstadt and Grenfell in their 2000 paper and the epidemic/endemic models first proposed by Held, Hohle, and Hofmann in their 2005 paper. We implement both of these models in the Stan software and illustrate their performance via analyses of measles data collected over a 2-year period in 17 regions in the Weser-Ems region of Lower Saxony, Germany.