How did we make our selection?

From the information companies Infotorg and Retriever, we obtained basic data on all companies in the elderly care industry. The industry's demarcation was made up of the Swedish Industry Division (SNI) and the specific code 87.301, which relates to "Care and care in special housing for older people". A lot of data about the companies was collected. Accounting information, addresses, information on the number of employees, group affiliation and more. An important task that was collected was the companies' workplaces (individual elderly homes).

In this first selection, there were a large number of workplaces that belonged to the companies, many of which belonged to other related businesses that were included in the companies' other activities. These irrelevant workplaces were eliminated by excluding those who did not have the SNI code 87.301. Subsequently, additional items were sold away, such as companies that lacked operations, missing address or were obviously incorrectly coded.

Remaining was a file with 2 957 addresses for workplaces (elderly housing) that we sent to Ratsit.

The file we got back from Ratsit contained 1,985 elderly homes with information on the number of deceased persons and dates during the period 1 January to 14 May 2017, 2018, 2019 and 2020. This became our final basic file on which our analyzes are based. The 1,985 elderly homes are run by 389 principals (companies or municipalities).

We then further delimited the material by selecting the period week 10-19 this year and compared the data with an average for the corresponding weeks 2017–2019. It therefore refers to the period 2 March to 10 May.

How can we say that our data gives a correct picture?

When we looked more closely at the data we obtained and compared with data from other sources, we saw that there were differences in level. Our numbers were lower than other sources. To investigate whether our figures nevertheless followed the same pattern, we compared a number of different registers, such as the National Board of Health's statistics on deceased persons on special housing and the Swedish palliative register's information on deceased persons on special housing. We also looked at Statistics Sweden's preliminary data on the number of deceased. You can roughly say that our numbers were at a level of 55-60 percent of the National Board of Health and 75-80 percent of the Palliative Register's figures.

One likely reason for the drop in our numbers may be that we only have information about people who have died and been registered in the address of the elderly person's home. Some deceased persons may be registered in their former home address and these will not be included in our documentation. We also found some elderly homes that had the wrong address in Ratsit, and there may be several elderly homes with incorrect addresses.

The question was whether the drop in our numbers was skewed or whether it was evenly distributed geographically and between the time periods. We could see that the dropout rate was fairly evenly distributed between the various time intervals Jan-May and between the years 2017-2020. When we looked at the geographical differences in the lapse between our numbers and the Palliative register, the differences were slightly larger, but still quite

evenly distributed between regions. Our overall assessment was that the dropout was sufficiently evenly distributed, both between the time periods and geographically.

Are there other problems with data?

A problem in analyzes is when breaking down figures at lower levels, for example municipalities and elderly housing. There will be very small numbers on the individual items and then analyzing the percentage changes will be difficult. A death toll on a particular dwelling or a smaller municipality with small numbers does not necessarily have to depend on the pandemic. But the numbers that stand out may be a reason to investigate further and find out if the pandemic is the cause. Then it is important to remember that it applies to numbers of deceased in general and not in covid-19.

We have contacted a large number of residents to verify the patterns we see in our data. And although the figures are sometimes not accurate, the patterns we see match the details of the residents. That is, there has been an increase in the number of dead 2020 compared to previous years.

We have also selected housing and municipalities with small numbers in our analysis. That is to say a doubling from two people who have died one year to four people next year, we believe are too small numbers and too dependent on chance and such an increase we have therefore not included in our analysis.

What did we find?

In our material, we see an increase of 100 per cent for elderly housing in the Stockholm region - and an increase of 30 per cent in the nation. We have been able to identify some 40 homes that stand out in our analysis where the number of deceased persons has at least doubled from an average over 2017–19 compared with 2020. This for the period we have chosen to analyze. These accommodations are found mainly in Stockholm but also in, for example, Borlänge, Linköping and Gothenburg.

But we also found a number of municipalities and housing units where in our analysis we can see no difference between the death rates for 2017-19 and 2020.

We have then linked our basic material with other data and done analyzes to see if there are different connections. Geographical links with the most affected regions and municipalities. Regions and municipalities that do not show any excess. Differences between private and public actors. Analysis of the major groups in the industry.