Practical Time Series Analysis

Chapter 13. Healthcare Applications

In this chapter we will look at time series analysis in the healthcare context with two case studies: Flu forecasting and nowcasting and Blood glucose forecasting. >These are both important healthcare applications for common health problems. Also, in both cases these are not solved problems but rather topics of ongoing research in academia and in the healthcare industry.

Predicting the Flu

Predicting the flu rate from week to week in a given geographic area is a longstanding and ongoing problem. Infectious disease specialists and global security professionals alike agree that infectious diseases pose a significant risk to human welfare. This is particularly the case for the flu, which strikes the vulnerable worldwide, inflicting hundreds of fatalities every year, mostly among the very young and very old. It’s crucial from both a health and national security standpoint to develop accurate models of how the flu will run its course in a given season. Flu prediction models are useful both to predict the virus specifically and also to help researchers explore general theories for how infectious diseases travel geographically.

A Case Study of Flu in One Metropolitan Area

We’ll look at a data set of weekly flu reports for a variety of administrative regions in France for the years of 2004 through 2013. We will predict the flu rate for Île de France, the Paris metropolitan region. The data can be downloaded from Kaggle¹ and is also available in the ...