Deep Spatial Q-Learning for Infectious Disease Control

Zhishuai Liu, Jesse Clifton, Eric B. Laber (), John Drake and Ethan X. Fang
Additional contact information
Zhishuai Liu: Duke University
Jesse Clifton: NC State University
Eric B. Laber: Duke University
John Drake: University of Georgia
Ethan X. Fang: Duke University

Journal of Agricultural, Biological and Environmental Statistics, 2023, vol. 28, issue 4, No 10, 749-773

Abstract: Abstract Infectious diseases are a cause of humanitarian and economic crises across the world. In developing regions, a severe epidemic can result in the collapse of healthcare infrastructure or even the failure of an affected state. The most recent 2013–2015 outbreak of Ebola virus disease in West Africa is an example of such an epidemic. The economic, infrastructural, and human costs of this outbreak provide strong motivation for the examination of adaptive treatment strategies that allocate resources in response to and anticipation of the evolution of an epidemic. We formalize adaptive management of an emerging infectious disease spreading across a set of locations as a treatment regime that maps up-to-date information on the epidemic to a subset of locations identified as high-priority for treatment. An optimal treatment regime in this context is defined as maximizing the expectation of a pre-specified cumulative utility measure, e.g., the number of disease-free individuals or the estimated reduction in morbidity or mortality relative to a baseline intervention strategy. Because the disease dynamics are not known at the beginning of an outbreak, an optimal treatment regime must be estimated online, i.e., as data accumulate; thus, an effective estimation algorithm must balance choosing interventions that lead to information gain and thereby model improvement with interventions that appear to be optimal under the current estimated model. We develop a novel model-free algorithm for the online management of an infectious disease spreading over a finite set of locations and an indefinite or infinite time horizon. The proposed algorithm balances exploration and exploitation using a semi-parametric variant of Thompson sampling. We also introduce a graph neural network-based estimator in order to improve the performance of this class of algorithms. Simulations, including those mimicking the spread of the 2013–2015 Ebola outbreak, suggest that an adaptive treatment strategy has the potential to significantly reduce mortality relative to ad hoc management strategies. Supplementary materials accompanying this paper appear online.

Keywords: Infectious diseases; Reinforcement learning; Graph neural networks (search for similar items in EconPapers)
Date: 2023
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s13253-023-00551-4 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:jagbes:v:28:y:2023:i:4:d:10.1007_s13253-023-00551-4

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/13253

DOI: 10.1007/s13253-023-00551-4

Access Statistics for this article

Journal of Agricultural, Biological and Environmental Statistics is currently edited by Stephen Buckland

More articles in Journal of Agricultural, Biological and Environmental Statistics from Springer, The International Biometric Society, American Statistical Association
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().