Limits...
Forecasting Influenza Epidemics in Hong Kong.

Yang W, Cowling BJ, Lau EH, Shaman J - PLoS Comput. Biol. (2015)

Bottom Line: Average forecast accuracies were 37% (for both peak timing and magnitude) at 1-3 week leads, and 51% (peak timing) and 50% (peak magnitude) at 0 lead.Forecast accuracy increased as the spread of a given forecast ensemble decreased; the forecast accuracy for peak timing (peak magnitude) increased up to 43% (45%) for H1N1, 93% (89%) for H3N2, and 53% (68%) for influenza B at 1-3 week leads.These findings suggest that accurate forecasts can be made at least 3 weeks in advance for subtropical and tropical regions.

View Article: PubMed Central - PubMed

Affiliation: Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, New York, United States of America.

ABSTRACT
Recent advances in mathematical modeling and inference methodologies have enabled development of systems capable of forecasting seasonal influenza epidemics in temperate regions in real-time. However, in subtropical and tropical regions, influenza epidemics can occur throughout the year, making routine forecast of influenza more challenging. Here we develop and report forecast systems that are able to predict irregular non-seasonal influenza epidemics, using either the ensemble adjustment Kalman filter or a modified particle filter in conjunction with a susceptible-infected-recovered (SIR) model. We applied these model-filter systems to retrospectively forecast influenza epidemics in Hong Kong from January 1998 to December 2013, including the 2009 pandemic. The forecast systems were able to forecast both the peak timing and peak magnitude for 44 epidemics in 16 years caused by individual influenza strains (i.e., seasonal influenza A(H1N1), pandemic A(H1N1), A(H3N2), and B), as well as 19 aggregate epidemics caused by one or more of these influenza strains. Average forecast accuracies were 37% (for both peak timing and magnitude) at 1-3 week leads, and 51% (peak timing) and 50% (peak magnitude) at 0 lead. Forecast accuracy increased as the spread of a given forecast ensemble decreased; the forecast accuracy for peak timing (peak magnitude) increased up to 43% (45%) for H1N1, 93% (89%) for H3N2, and 53% (68%) for influenza B at 1-3 week leads. These findings suggest that accurate forecasts can be made at least 3 weeks in advance for subtropical and tropical regions.

No MeSH data available.


Related in: MedlinePlus

Accuracy predicting outbreak peak timing (A), peak magnitude (B), onset (C), and duration (D).Accuracy was calculated over all forecasts (332,400 for each setting of the forecast system). This analysis includes the forecasts for seasonal H1N1, the 2009 pandemic H1N1, H3N2, B and all strains combined. Results are shown for the EAKF (red) and the PF (blue), evaluated using two standards (solid vs. dashed lines, as specified in the parentheses). On the x-axis, positive leads indicate that a peak is forecast in the future; negative leads indicate that a peak is forecast in the past; a 0 week lead indicates that a peak is forecast as the same week of forecast. Leads are relative to the predicted peak for forecasts of the peak timing, peak magnitude, and duration, and relative to the predicted onset for forecasts of onset timing.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4520691&req=5

pcbi.1004383.g003: Accuracy predicting outbreak peak timing (A), peak magnitude (B), onset (C), and duration (D).Accuracy was calculated over all forecasts (332,400 for each setting of the forecast system). This analysis includes the forecasts for seasonal H1N1, the 2009 pandemic H1N1, H3N2, B and all strains combined. Results are shown for the EAKF (red) and the PF (blue), evaluated using two standards (solid vs. dashed lines, as specified in the parentheses). On the x-axis, positive leads indicate that a peak is forecast in the future; negative leads indicate that a peak is forecast in the past; a 0 week lead indicates that a peak is forecast as the same week of forecast. Leads are relative to the predicted peak for forecasts of the peak timing, peak magnitude, and duration, and relative to the predicted onset for forecasts of onset timing.

Mentions: Fig 3 presents prediction accuracy for epidemic onset, duration, peak timing, and peak ILI+ magnitude for both forecast systems. These are tallied for all forecasts—individual strain and all strains—a total of 332,400 weekly forecasts (i.e., 831 weekly forecasts for each strain × 4 strains × 100 runs). Here we focus our analysis on predicted lead weeks ranging from -10 to 10 weeks; a positive lead (e.g. 2 wk) indicates the event (e.g., the epidemic peak or onset) is predicted to occur 2 weeks in the future from the time of forecast initiation; a 0 wk lead indicates the event is predicted to occur at the time of forecast initiation; and a negative lead, say -3 wk, indicates the event is predicted to have occurred 3 weeks prior to the forecast initiation. Forecasts with negative lead times may appear counterintuitive; however, accurate prediction that an event has passed is an important capability of a forecast system. In regions experiencing year-round influenza transmission, such as Hong Kong, multimodal epidemics, i.e. epidemics with multiple crests, are common (Fig 1). A forecast initiated after a smaller crest but preceding the overall peak may mistakenly identify that smaller crest as the peak and predict that the peak has passed, i.e. an inaccurate forecast with negative lead. Conversely, an accurate forecast with negative lead indicates that no spurious future increase in incidence is predicted. Therefore, forecast accuracy at negative leads also represents the ability of the forecast system to predict future epidemic trajectories.


Forecasting Influenza Epidemics in Hong Kong.

Yang W, Cowling BJ, Lau EH, Shaman J - PLoS Comput. Biol. (2015)

Accuracy predicting outbreak peak timing (A), peak magnitude (B), onset (C), and duration (D).Accuracy was calculated over all forecasts (332,400 for each setting of the forecast system). This analysis includes the forecasts for seasonal H1N1, the 2009 pandemic H1N1, H3N2, B and all strains combined. Results are shown for the EAKF (red) and the PF (blue), evaluated using two standards (solid vs. dashed lines, as specified in the parentheses). On the x-axis, positive leads indicate that a peak is forecast in the future; negative leads indicate that a peak is forecast in the past; a 0 week lead indicates that a peak is forecast as the same week of forecast. Leads are relative to the predicted peak for forecasts of the peak timing, peak magnitude, and duration, and relative to the predicted onset for forecasts of onset timing.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4520691&req=5

pcbi.1004383.g003: Accuracy predicting outbreak peak timing (A), peak magnitude (B), onset (C), and duration (D).Accuracy was calculated over all forecasts (332,400 for each setting of the forecast system). This analysis includes the forecasts for seasonal H1N1, the 2009 pandemic H1N1, H3N2, B and all strains combined. Results are shown for the EAKF (red) and the PF (blue), evaluated using two standards (solid vs. dashed lines, as specified in the parentheses). On the x-axis, positive leads indicate that a peak is forecast in the future; negative leads indicate that a peak is forecast in the past; a 0 week lead indicates that a peak is forecast as the same week of forecast. Leads are relative to the predicted peak for forecasts of the peak timing, peak magnitude, and duration, and relative to the predicted onset for forecasts of onset timing.
Mentions: Fig 3 presents prediction accuracy for epidemic onset, duration, peak timing, and peak ILI+ magnitude for both forecast systems. These are tallied for all forecasts—individual strain and all strains—a total of 332,400 weekly forecasts (i.e., 831 weekly forecasts for each strain × 4 strains × 100 runs). Here we focus our analysis on predicted lead weeks ranging from -10 to 10 weeks; a positive lead (e.g. 2 wk) indicates the event (e.g., the epidemic peak or onset) is predicted to occur 2 weeks in the future from the time of forecast initiation; a 0 wk lead indicates the event is predicted to occur at the time of forecast initiation; and a negative lead, say -3 wk, indicates the event is predicted to have occurred 3 weeks prior to the forecast initiation. Forecasts with negative lead times may appear counterintuitive; however, accurate prediction that an event has passed is an important capability of a forecast system. In regions experiencing year-round influenza transmission, such as Hong Kong, multimodal epidemics, i.e. epidemics with multiple crests, are common (Fig 1). A forecast initiated after a smaller crest but preceding the overall peak may mistakenly identify that smaller crest as the peak and predict that the peak has passed, i.e. an inaccurate forecast with negative lead. Conversely, an accurate forecast with negative lead indicates that no spurious future increase in incidence is predicted. Therefore, forecast accuracy at negative leads also represents the ability of the forecast system to predict future epidemic trajectories.

Bottom Line: Average forecast accuracies were 37% (for both peak timing and magnitude) at 1-3 week leads, and 51% (peak timing) and 50% (peak magnitude) at 0 lead.Forecast accuracy increased as the spread of a given forecast ensemble decreased; the forecast accuracy for peak timing (peak magnitude) increased up to 43% (45%) for H1N1, 93% (89%) for H3N2, and 53% (68%) for influenza B at 1-3 week leads.These findings suggest that accurate forecasts can be made at least 3 weeks in advance for subtropical and tropical regions.

View Article: PubMed Central - PubMed

Affiliation: Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, New York, United States of America.

ABSTRACT
Recent advances in mathematical modeling and inference methodologies have enabled development of systems capable of forecasting seasonal influenza epidemics in temperate regions in real-time. However, in subtropical and tropical regions, influenza epidemics can occur throughout the year, making routine forecast of influenza more challenging. Here we develop and report forecast systems that are able to predict irregular non-seasonal influenza epidemics, using either the ensemble adjustment Kalman filter or a modified particle filter in conjunction with a susceptible-infected-recovered (SIR) model. We applied these model-filter systems to retrospectively forecast influenza epidemics in Hong Kong from January 1998 to December 2013, including the 2009 pandemic. The forecast systems were able to forecast both the peak timing and peak magnitude for 44 epidemics in 16 years caused by individual influenza strains (i.e., seasonal influenza A(H1N1), pandemic A(H1N1), A(H3N2), and B), as well as 19 aggregate epidemics caused by one or more of these influenza strains. Average forecast accuracies were 37% (for both peak timing and magnitude) at 1-3 week leads, and 51% (peak timing) and 50% (peak magnitude) at 0 lead. Forecast accuracy increased as the spread of a given forecast ensemble decreased; the forecast accuracy for peak timing (peak magnitude) increased up to 43% (45%) for H1N1, 93% (89%) for H3N2, and 53% (68%) for influenza B at 1-3 week leads. These findings suggest that accurate forecasts can be made at least 3 weeks in advance for subtropical and tropical regions.

No MeSH data available.


Related in: MedlinePlus