Artikel
An Endemic-Epidemic Beta Model for Time Series of Infectious Disease Proportions
Suche in Medline nach
Autoren
-
Junyi Lu
- Institute of Medical Informatics, Biometry, and Epidemiology, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
-
Sebastian Meyer
- Institute of Medical Informatics, Biometry, and Epidemiology, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
| Veröffentlicht: | 26. Februar 2021 |
|---|
Gliederung
Text
Background: Statistical models for infectious disease occurrence help to understand the mechanism of disease spread and accurate forecasts enable health officials to plan disease prevention and allocate treatment resources. The so-called endemic-epidemic regression approach (HHH) for times series of case counts is frequently adopted and particularly suitable for notifiable diseases with a long history of public health surveillance.
In some applications, however, the proportion of infected individuals is of more direct interest than their absolute number.
Methods: We propose an endemic-epidemic beta model as an extension of the well-established HHH framework.
Built on beta regression and simplicial geometry for compositional data, our model accommodates the asymmetric shape and heteroskedasticity of proportion distributions and is consistent for complementary proportions. Coefficients can be interpreted in terms of odds ratios.
Typically, public health surveillance gives rise to multivariate time series with stratification by region or age group. We extended the beta model for such areal time series by borrowing the structure of the multivariate HHH model.
Results: Using U.S. national influenza-like illness surveillance data over 18 seasons, we assessed probabilistic forecasts of this univariate new beta model with proper scoring rules. Other readily available forecasting tools were used for comparison, including Prophet, (S)ARIMA and kernel conditional density estimation (KCDE). Furthermore, we evaluate the short-term forecast performance of the multivariate beta model using U.S. influenza-like illness surveillance data for 10 Health & Human Service (HHS) regions.
Conclusion: We conclude that the univariate endemic-epidemic beta model is a performant and easy-to-implement tool to forecast flu activity a few weeks ahead. The performance evaluation of the multivariate beta model is still in progress。
The authors declare that they have no competing interests.
The authors declare that an ethics committee vote is not required.
