Reinhard Wagner

I appreciate hints on errors and suggestions for improvement:
r_a_wagner@gmx.de

1.1.2. Discussion by EuroMOMO

EuroMOMO explicitly points out the puzzling facts in its bulletin of week 39/2022:

"Note on observed excess mortality in the age group 0-14 years
during 2021-2022

Since mid-2021, some unusual excess mortality signals have been observed in the age group of 0-14 years. EuroMOMO is looking into the possible explanation for these signals, in consultation with participating countries in the network.

From a methodological perspective, it is important to note that the mortality baseline applied currently in the EuroMOMO excess mortality algorithm does not include any mortality data from 2020, 2021 or 2022, which will be affected by varying excess mortality associated with the overall impact of the COVID-19 pandemic including varying restrictions, which could bias the 'normal' level of mortality.

Furthermore, it should again be stressed that EuroMOMO does not collect any data beyond mortality data to answer questions about causality in observed mortality signals. Therefore, EuroMOMO’s outputs should not, in any way, be cited as 'evidence' for any underlying causes of unusual mortality patterns."¹

The bulletin of week 43/2022 states:

"Cautionary note on cumulated excess mortality reported by EuroMOMO for the pandemic years

The excess mortality is calculated by EuroMOMO every week as the difference between the observed number of deaths and the estimated expected number of deaths (the baseline). The cumulated excess mortality is shown for each calendar year by week and is calculated by adding up the weekly excess number of deaths over the weeks of surveillance.

Because of the unexpected and varying mortality experienced during the COVID-19 pandemic, the calculated baseline applied in the EuroMOMO estimation of excess number of deaths do [sic!] not include data from 2020 and onwards, as these data will distort the estimated level of baseline number of deaths.

Due to the exclusion of data from the COVID-19 pandemic, the trend incorporated in the statistical estimation of the baseline are [sic!] presently forecasted [sic!] beyond the intended duration. This prolonged forecast has introduced an increasing bias over time, which can cause incorrect estimations of the excess mortality, particularly when the numbers are cumulated. (Emphasis mine, R.W.)

Because of this enhanced bias when cumulating data, the cumulated outputs should not be considered reliable at this point and must be interpreted very cautiously. An example is number [sic!] of deaths in the age group 0-14 years, where the EuroMOMO countries in the years preceding the COVID-19 pandemic experienced a decrease in mortality. The model extends this declining baseline mortality in a linear fashion. Before the COVID-19 pandemic, there was a mean of 361 deaths per week in this age group. In week 40, 2022, there is approximately 8.5% bias due to this linear trend meaning that the expected number of deaths (the baseline) is predicted to be 29 deaths lower than in week 1, 2020. This means that even though the number of reported weekly deaths is equal in week 40, 2022 and week 1, 2020, the lower baseline in week 40, 2022 means that there will be 29 deaths higher excess mortality in this week. When cumulating the weekly number of excess deaths, this bias due to the extension of the linear trend in the baseline gets high and may therefore lead to incorrect estimates. During the COVID-19 pandemic the lock-down period (week 1, 2020 to week 21, 2021) had a mean of 326 deaths per week and the following period (week 22, 2021 to week 40, 2022) had a mean of 345 deaths per week which both are lower than the average of 361 deaths per week observed in the period before the COVID-19 pandemic (week 1, 2018 to week 52, 2019). In other words, the EuroMOMO mortality surveillance system does actually not detect more deaths among the 0 to 14-years old during the ongoing COVID-19 pandemic than in the period before the COVID-19 pandemic, even if the cumulated outputs reported on the EuroMOMO website suggest so. (Emphasis mine, R.W.)

In summary, the bias in the weekly estimates of excess number of deaths are still relatively small and considered to be within reasonable and acceptable levels to be used in accordance with its initial purpose, i.e. evaluating weekly changes in mortality.

The EuroMOMO hub, together with the participating countries, are working on solutions to mitigate the bias; however, a revised model is not expected to be ready for use in 2022."²

To summarize: There is a clearly discernible "gradient" from left to right in the model of expected deaths from many years of experiencing a reduction in infant mortality: it was previously assumed from experience that mortality in this age group was constantly decreasing and the calculation method of excess mortality was set up accordingly.

Now, in fact, not more children are dying than before the pandemic, but - with the exception of the first year of the pandemic, when the decline in child deaths was severe - the previously usual decline in child mortality is not continuing now. (If the data from the pandemic were included in the calculation, which so far they are not, there would be an even greater decline in expected child mortality, statistically equivalent to an even greater increase in the current excess mortality of this age group).

It is intended to solve the problem by a new calculation method.

Did it still say in bulletin 2022-39 (See above!): "Since mid-2021, some unusual excess mortality signals have been observed in the age group of 0-14 years. EuroMOMO is looking into the possible explanation for these signals, in consultation with participating countries in the network", now only a statistic-technical explanation of the end of the decrease of infant mortality is being sought: "The EuroMOMO hub, together with the participating countries, are working on solutions to mitigate the bias; however, a revised model (Emphasis mine, R.W.) is not expected to be ready for use in 2022." (See above!)


---> Continue reading: 1.2. Spain: MoMoISCII and Instituto Nacional de Estadística (INE)