Early detection of disease outbreaks and non-outbreaks using incidence data: A framework using feature-based time series classification and machine learning

Forecasting the occurrence and absence of novel disease outbreaks is essential for disease management, yet existing methods are often context-specific, require a long preparation time, and non-outbreak prediction remains understudied. To address this gap, we propose a novel framework using a feature-based time series classification (TSC) method to forecast outbreaks and non-outbreaks. We tested our methods on synthetic data from a Susceptible–Infected–Recovered (SIR) model for slowly changing, noisy disease dynamics. Outbreak sequences give a transcritical bifurcation within a specified future time window, whereas non-outbreak (null bifurcation) sequences do not. We identified incipient differences, reflected in 22 statistical features and 5 early warning signal indicators, in time series of infectives leading to future outbreaks and non-outbreaks. Classifier performance, given by the area under the receiver-operating curve (AUC), ranged from 0 . 99 for large expanding windows of training data to 0 . 7 for small rolling windows. The framework is further evaluated on four empirical datasets: COVID-19 incidence data from Singapore, 18 other countries, and Edmonton, Canada, as well as SARS data from Hong Kong, with two classifiers exhibiting consistently high accuracy. Our results highlight detectable statistical features distinguishing outbreak and non-outbreak sequences well before potential occurrence, in both synthetic and real-world datasets presented in this study.Author summary: Timely prediction of disease outbreaks and non-outbreaks is crucial for effectively implementing preventative measures and for avoiding unnecessary overreaction. While early warning signals (EWSs) and mathematical modeling are often used for such predictions, the former may fail in systems involving stochasticity, and the latter requires underlying mechanisms that are unclear at the early stage. Here, we propose a novel framework using a feature-based time series classification method and training classifiers on these features for prediction. Within this framework, we develop a general model, with no real-world training data, that accurately forecasts outbreaks and non-outbreaks. Our results show that statistical features exhibit different distributions for outbreak and non-outbreak sequences long before outbreaks occur. We can detect these differences, evident in both synthetic and real-world datasets, to anticipate the occurrence of outbreaks and non-outbreaks. Our work would contribute to the early detection of disease outbreaks and non-outbreaks with a less sophisticated approach, providing valuable insights for mitigation strategies of novel diseases.