No 7-8 - 2011
Monitoring of mortality in Denmark
Monitoring of mortality in Denmark
Disease surveillance is an important basis for targeted prevention. As a supplement, surveillance of the general population’s mortality may contribute to the surveillance of infectious diseases. Since 2009, the Department of Epidemiology has continually received data on all-cause deaths reported to the Danish Civil Registry System (CRS) and used these as part of the surveillance of infectious diseases. To determine the expected number of deaths (baseline), historical data have also been collected.
Modelling excess mortality
In countries with a temperate climate, mortality displays a distinct seasonal variation with more deaths in the winter months and fewer in the summer, Figure 1. Therefore, the expected weekly number of deaths (baseline) may be estimated using a relatively simple sinus transformation. Excess mortality in the population may then be calculated as the observed number of deaths in a given time period and age group minus the expected number of deaths in the same age group.
This simple model may be refined in several ways:
- Data are reported to the CRS after a small delay (median time from date of death to registration is three days). It is possible to adjust for this delay.
- Statistical methods are used to determine the uncertainty of the estimates. Figure 1 shows a threshold for significant excess mortality equivalent to three standard deviations above baseline.
- It is possible to immediately assess any minor, but sustained tendencies
in mortality by studying cumulated deviations between the observed
and the expected results.
- In multivariate models, additional data, e.g. influenza activity and deviations in temperature conditions, may be included with a view to explaining any excess mortality.
Mortality frequently increases immediately after Christmas and New Year's Eve, which may have a number of causes.
In winters when influenza activity does not coincide with Christmas, an M-shaped mortality pattern is observed, e.g. in 2001/2, 2002/3, 2004/5, Figure 1.
The initial peak is associated with the Christmas week.
In the 2000-2011 period, the observed mortality in the winters of 2002/3, 2003/4 and 2008/9 clearly exceeded the expected mortality by 1700, 2100 and 1700 additional deaths, respectively.
However, in the winter of 2005/6 and in the influenza pandemic year of 2009/10, no substantial excess mortality was seen.
In the beginning of 2010, however, mortality increased during a period with cold weather. This period did not coincide with influenza activity.
Furthermore, towards the end of 2010, an increase in mortality was observed, and once again it coincided with a cold weather period at the end of year, but also with a mycoplasma epidemic, EPI-NEWS 48/10. No substantial excess mortality was seen in connection with the current influenza season in which influenza activity is estimated to have peaked in Week 5, 2011.
Timely monitoring of population mortality is an instrument that may be used in particular to assess if outbreaks of infectious diseases or other events are associated with excess mortality, but it also assists in documenting situations with no excess mortality.
In rare cases, it is conceivable that surveillance may identify events that warrant further scrutiny. For instance, the epidemic in the USA could, theoretically, have been identified earlier if efficient surveillance including age and gender analyses had been in place in the early phases of the epidemic to detect specific changes in mortality.
Infectious as well as non-infectious diseases may cause excess mortality. It is well-described that some population groups are at special risk of seasonal influenza, and the excess mortality seen in the winters of 2002/3, 2003/4 and 2008/9 may, in particular, be attributed to the occurrence of influenza; in these years excess mortality primarily occurred among persons aged > 75 years.
Owing to natural immunity, elderly citizens have enjoyed a low risk of pandemic influenza A (H1N1), which explains why a substantial excess mortality was not seen in these age-groups during the latest influenza seasons.
Data on population mortality may be followed at www.ssi.dk (Danish language) and the equivalent data of a number of European countries are available at www.euromomo.eu.
(J. Nielsen, A. Mazick, K. Mølbak, Department of Epidemiology)
23 February 2011