A Bayesian Monte Carlo approach for predicting the spread of infectious diseases

In this paper, a simple yet interpretable, probabilistic model is proposed for the prediction of reported case counts of infectious diseases. A spatio-temporal kernel is derived from training data to capture the typical interaction effects of reported infections across time and space, which provides insight into the dynamics of the spread of infectious diseases. Testing the model on a one-week-ahead prediction task for campylobacteriosis and rotavirus infections across Germany, as well as Lyme borreliosis across the federal state of Bavaria, shows that the proposed model performs on-par with the state-of-the-art hhh4 model. However, it provides a full posterior distribution over parameters in addition to model predictions, which aides in the assessment of the model. The employed Bayesian Monte Carlo regression framework is easily extensible and allows for incorporating prior domain knowledge, which makes it suitable for use on limited, yet complex datasets as often encountered in epidemiology. Author summary Why was this study done? Statistical modeling is invaluable to public-health policy as it helps understand and anticipate the dynamics of the spread of infectious diseases. The available training data is often limited and reported with a low spatial and temporal resolution. This poses a challenge and makes it particularly important to incorporate domain knowledge and prior assumptions to guide the modeling process. In order to evaluate the trustworthiness and reliability of a model’s predictions, it is crucial to be able to interpret the model and quantify the model uncertainty. To address this, we develop an interpretable model that uses Bayesian inference (rather than commonly used maximum likelihood estimation) and provides a probability distribution over inferred parameters. What did the researchers do and find? We develop and test a single probabilistic model that learns to predict the number of weekly case counts for three different diseases (campylobacteriosis, rotaviral enteritis and Lyme borreliosis) at the county level one week ahead of time. We employ a Bayesian Monte Carlo regression approach that provides an estimate of the full probability distribution over inferred parameters as well as model predictions. The model learns an interpretable spatio-temporal kernel that captures typical interactions between infection cases of the tested diseases. The predictive performance of our model compares favorably with a contemporary reference model for all diseases tested. What do these findings mean? Interpretable predictive models can be applied to surveillance data to gain insights into the dynamics of infectious diseases. Probabilistic modeling approaches provide a suitable framework for many challenges of working with epidemiological data.