This article follows up on two older articles by RGA associates: “Seasonality of Mortality,” by Kyle Nobbe, published in the January 2017 edition of ReFlections,1 and “Seasonal Flu and the Impact on Mortality,” by Dr. Dave Rengachary.16 These articles established that with the exception of cancer and non-natural deaths, almost all major causes of mortality are seasonal and more deaths are seen in winter. Also established was that influenza mortality exhibits significant variation from season to season, with a greater than tenfold difference in mortality between the least and most severe seasons. Finally, data from U.S. and Japan demonstrates a statistically significant positive correlation between excess deaths due to influenza and pneumonia and excess deaths due to all other medical causes.
Several theories attempt to rationalize why all-cause mortality is seasonal. The severity of influenza during a given season might offer insight into how severe overall mortality during that season might be. Additionally, although influenza and pneumonia deaths are only a small proportion of overall seasonal mortality, they could serve as valuable indicators of overall mortality during a given influenza season. Knowledge of an influenza season’s potential severity, either in advance or at onset, might also help insurance companies improve their financial planning and explain their earning patterns. Also important to consider is that age is the most significant risk factor for influenza mortality: greater than 90% of influenza deaths in recent decades have occurred in people age 65 and older.
RGA is partnered with BlueDot, a Toronto-based company that studies how infectious diseases disperse worldwide through analysis of big data, and brings together predictive modeling and data visualization to deliver timely evidence to decision-makers. This article presents the RGA/BlueDot partnership’s most recent, focused research on seasonal influenza mortality variation.
Although influenza epidemics occur every year, the severity of these epidemics, as measured by physician visits, hospitalizations, and influenza-attributable mortality, varies from season to season. The ability to predict the severity of a particular influenza season improves as the season progresses. There is a continuous stream of influenza data collected globally, which when examined collectively, offers insights into how severe the upcoming influenza season might be.
Wading through the various data sources and understanding how all of these different metrics relate to influenza season severity is a complex task. RGA partnered with BlueDot to i) describe the process by which data about influenza are collected globally, highlighting the significance of these data with respect to influenza-attributable mortality, and ii) develop a statistical model, using 15 years of historical mortality data from the United States, to assess if and when selected indicators could predict excess influenza-attributable mortality.
An example of how influenza-attributable mortality varies across seasons is shown in Figure 1 (below). Here, data from the Centers for Disease Control and Prevention (CDC) WONDER databases from the 1999-2000 to the 2014-2015 seasons for the U.S. population age 65 and older was used to estimate excess deaths per month over that 15-year span. The baseline came from a Serfling periodic regression model, the standard CDC algorithm for calculating excess influenza-attributable mortality. This model controls for linear and quadratic trends as well as annual cyclical patterns.
Influenza Data Generation Process
In the northern hemisphere, influenza season normally starts in October and ends in April, while in the southern hemisphere, influenza season typically occurs between May and October. The different timing of influenza seasons in the northern and southern hemispheres means that data on influenza activity are generated throughout the year. These data are gathered by local, national, and international organizations, and their use has the potential to improve the ability to predict not only influenza and pneumonia mortality for a given season, but potentially overall mortality as well.
Before investigating how different indicators can help anticipate the severity of an influenza season, it is important to understand the various sources and types of data that are collected. A large volume of data are generated relating to both influenza activity and the characteristics of circulating influenza viruses. We sought to describe the complex processes by which influenza data are collected and to identify indicators that might be used prior to or during the early stages of influenza season to better predict influenza-associated mortality
Two important collectors and disseminators of influenza data are the World Health Organization (WHO) and the CDC. Each uses a variety of methodologies to monitor and collect data on the severity of influenza seasons. WHO’s FluNet, a global tool for influenza virological surveillance, reports weekly, by country, the number of specimens processed that test positive for different influenza subtypes. The CDC’s FluView provides similar information for the U.S. only, reported at the national and regional level.
In late September to early October, prior to the onset of the U.S. influenza season, the CDC’s Morbidity and Mortality Weekly Report (MMWR) publishes a summary of
influenza activity that occurred during the summer. The report includes the results of antigenic match assays performed in CDC laboratories, comparing the match between circulating viruses isolated in the U.S. and internationally to the viruses included in the seasonal influenza vaccine. The Worldwide Influenza Centre at the Francis Crick Institute in the U.K. also releases its summer report around September, although this report is more technical in nature and is not as easily interpretable as the MMWR report. Another source of antigenic match data comes from the Chinese National Influenza Center, which issues weekly reports of the proportion of influenza isolates matching the vaccine strain cumulatively since March. (The vaccine used in China is generally the same as the northern hemisphere vaccine used in the U.S.).
Starting in October and lasting throughout the U.S.’s influenza season, the CDC’s FluView reports on the cumulative proportion of isolates that match the vaccine strain, by influenza type.
Review of Indicators Associated with Influenza Mortality
These different data sources provide a number of potential indicators that may provide early warning signs as to the severity of an upcoming influenza season. BlueDot reviewed and synthesized the literature on different potential indicators and evaluated them on four criteria:
Read More +