Mortality monitoring (MOMO)

| Weekly update, Tuesday at 2 p.m.

image – experimental statistics

Background

Through its mortality monitoring, the FSO observes excess mortality, i.e. the weekly number of deaths above the usual value for the season. The mortality rates are one of the main indicators in health statistics and play a fundamental role in public health. This is why mortality monitoring is so important during an influenza or other pandemic. During a severe pandemic, monitoring is a robust means of observing the pandemic’s progression and its impact on public health. Mortality monitoring is also invaluable in identifying the impact of extreme environmental conditions such as heatwaves or cold spells on human life.

Switzerland has joined the European EuroMomo network, which analyses and presents Europe-wide data supplied by the different countries using a standardised method. The aim is to “Detect and measure excess deaths related to seasonal flu, pandemics and other public health threats. Official national mortality statistics are provided weekly from the 24 European countries in the EuroMOMO collaborative network supported by the European Centre for Disease Prevention and Control (ECDC) and the World Health Organisation (WHO).”

Results


The excess mortality observed at the nationwide level from 16 March to 19 April 2020 (calendar weeks 12–16) developed differently in the regions NUTS2 and the cantons (NUTS3).
In Ticino (region CH07), excess mortality became apparent as early as 9 March 2020 (week 11) and, at its peak in week 14, reached more than three times the usual number of deaths. In the Lake Geneva region (CH01), excess mortality began on 16 March (week 12) and lasted longer, until 5 May (week 18). In the week from 30 March to 5 April, twice as many people died there as would normally be expected. This makes it the second most affected region in terms of excess mortality. Relative to the significantly larger population, this is the region with the highest number of additional deaths in absolute terms. All three cantons in the region (VD, VS, GE) are affected. In this region, the under-65s are most affected.
In the region of north-western Switzerland (CH03), Basel-Stadt and -Landschaft were affected, and by calendar week 14 the number of deaths was clearly above the average. In Aargau, no increased values were observed. In the Espace Mittelland (CH02), excess mortality is visible in the cantons of Fribourg and Neuchâtel, but not in Bern and Solothurn; the data for the canton of Jura are not shown.
In Eastern Switzerland (CH05), only Graubünden showed a slight excess mortality, while in Central Switzerland (CH06) the canton of Schwyz showed a slight excess mortality. In Zurich (CH04) the values are in the upper expectation range.
At a certain point in time, a greatly increased local mortality rate may be concealed in the national average. No results are presented for the eight smallest cantons.

Objectives

To monitor mortality in Switzerland, the number of deaths in a given year are estimated based on the trends seen in the previous five years; distribution across individual weeks is estimated on the basis of the median value for each individual calendar week of the previous ten years. Separate estimates are carried out for those under the age of 65 and for those aged above 65. The mortality monitoring figures are based on the daily civil registry notifications, which are sent to the FSO for its Vital Statistics (BEVNAT) by the civil registry offices. The processing of notifications takes time. A sufficiently large percentage (> 85%) of deaths are usually registered after nine days. This allows an estimate to be made of the actual number of deaths based on a wide data base. The level of excess mortality is calculated based on the difference between the estimated and expected number of deaths. It is therefore an estimate.

Since 28 April 2020, mortality monitoring has also been providing the number of deaths in Switzerland’s seven major regions (as defined by the FSO): Lake Geneva region, Espace Mittelland, Northwest Switzerland, Zurich region, Easter Switzerland, Central Switzerland and Ticino. Deaths in the seven major regions are counted using similar processes as at national level. As of 15 May 2020, the Federal Statistical Office has also been publishing mortality monitoring results at cantonal level. Mortality monitoring is usually updated on Tuesdays.

Methodology

Summary

Deaths are reported to the relevant civil registry office and registered in a centralised database (BEVNAT). The FSO estimates the number of cases based on the assumption that the flow of reported deaths is constant. The number of deaths usually expected is calculated based on the number of deaths occurring in the previous five years. Lastly, a range is calculated for each expected value within which fluctuations must be considered random. The calculation of expected deaths, therefore, is not just an average value, but also takes into account changes in the population from year to year as well as random fluctuations. For each expected value, a range is calculated within which fluctuations are considered to be random.

The methodology used at national level has been extended to the major regions and cantons with more than 100 000 inhabitants, by making an individual extrapolation for each canton or major region and calculating the expected values according to the same principles.

Data source

The mortality monitoring is based on the day-to-day reporting of deaths from the electronic civil register (computerised civil status register, Infostar). Deaths must be confirmed by a doctor’s death certificate and reported to the cantonal civil registry office concerned.
The monitoring includes all deaths occurring in Switzerland of people resident in Switzerland. It does not include the deaths of Swiss residents occurring abroad.

Data flow

The mortality monitoring figures are based on the daily civil registry notifications, which are sent to the FSO for its Vital Statistics by the communes (civil registry offices). After nine days there is usually a sufficiently large database. The reporting flow to the civil registry offices is assumed to be constant. After 40 days, 97.5% of all deaths have been reported and numbers can be shown without the need for extrapolation. The remaining 2.5% of reports arrive within the course of the year and are once received are included in the figures.

Calculation

Calculations for mortality monitoring are made in two stages. At the start of every year, extrapolation weightings for reporting delays (A) are calculated as well as the expected number of deaths with an upper and lower limit (range) (B). This is the basis for the calculations currently being made every Tuesday (C).

A) Number of deaths (extrapolation)

Based on the previous year’s figures, the distribution of reporting delays (time between day of death and inclusion in register) is estimated. This distribution is used to calculate weighting factors to correct the current year’s number of deaths in respect of reporting delays. Separate weighting factors are calculated for each canton and major region to take into account individual reporting characteristics.

B) upper / lower limit for expected numbers

  • Step 1: using a regression approach, an estimate of the number of deaths is established based on figures for the previous five years (after prior smoothing). (The number of deaths is considered to be a Poisson distributed variable. Estimates are made for several age groups).
  • Step 2: based on the previous 10 years, a median value is calculated for the number of deaths.
  • Step 3: The weekly figures from step 2 are smoothed.
  • Step 4: The figures from step 3 are scaled up to give the same total as the total of the estimate made in step 1.
  • Step 5: Based on the assumption that the number of deaths can be considered to be Poisson-distributed and on the figures from step 4, prediction intervals are calculated for the number of deaths (upper / lower limits).

C) Weekly analysis

In the weekly analysis, the reported deaths for each day are supplemented using extrapolation factors. This results in the time series shown as the number of deaths (extrapolation). The expected numbers from step B are added with their range.