English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Background Seasonally-adjusted linear models (‘Serfling’ models) serve as an important surveillance measure to estimate influenza (flu) attributable deaths for resource allocation to public health programs (e.g., vaccination campaigns). We compared performance of traditional time-series and viral activity-based models to a novel open-source R-package ‘Prophet’ for estimating the number of deaths attributable to influenza per season. Methods We evaluated deaths from the 122-Cities Mortality Reporting System which reports the total number of death certificates where pneumonia or flu was listed as a contributing cause of death. Models were fitted to 2010–2015 influenza seasons. The first Serfling model (M1) used baseline periods of low-endemicity (summer months) to estimate attributable deaths during a flu season (OCT-MAY), the second Serfling model (M2) incorporated both baseline and virology data (count laboratory proven flu cases in a given period). The Prophet model (M3) uses generalized additive models incorporating annual, seasonal terms and viral activity data. The difference between observed deaths, and those predicted by each model in the absence of flu were ‘attributable death.” Epidemic weeks exceeded the 95% upper prediction interval. Model performance was assessed by Root Mean Square Error (RMSE). Results From 2010 to 2015, the average deaths due to pneumonia and influenza per season numbered 824 per week (total 198,692). Compared with the traditional Serfling model (M1), the Prophet model estimated 52% more influenza-attributable deaths (13,443 vs. 8,800) and more epidemic weeks (25 vs. 10) with lower RMSE (75.9 vs. 95.3 [lower is better]). Compared with the viral activity-based model (M2), the Prophet model estimated 6% fewer attributable deaths (13,443 vs. 14,326), with more epidemic weeks (25 vs. 19) and lower RMSE (RMSE 75.9 vs. 92.6). Conclusion Generalized additive models, implemented through the R-package Prophet, are superior in terms of reducing model prediction error for influenza mortality vs. traditional models. Based on superior model performance, the attributable mortality estimated by these novel models may be preferred over traditional regression models. This study was funded by Sanofi Pasteur. Disclosures All authors: No reported disclosures.

2730. Estimating Deaths Attributable to Influenza Mortality Using Traditional and Novel Forecasting Methods