Skip to main content
Deutsches Ärzteblatt International logoLink to Deutsches Ärzteblatt International
letter
. 2023 May 12;120(19):337–344. doi: 10.3238/arztebl.m2023.0072

Immunity Against SARS-CoV-2 in the German Population

Kai Schulze-Wundling 1,2, Patrick Frank Ottensmeyer 1,2, Kristin Maria Meyer-Schlinkmann 3, Marek Deckena 3, Stefan Krüger 4, Simon Schlinkert 4, Axel Budde 1,2, Dieter Münstermann 3, Nicole Töpfner 5, Astrid Petersmann 6, Matthias Nauck 7, André Karch 8, Berit Lange 1, Sabine Blaschke 9, Carsten Tiemann 3, Hendrik Streeck 1,2,*
PMCID: PMC10408282  PMID: 37155224

Abstract

Background

Early during the SARS-CoV-2 pandemic, national population-based seroprevalence surveys were conducted in some countries; however, this was not done in Germany. In particular, no seroprevalence surveys were planned for the summer of 2022. In the context of the IMMUNEBRIDGE project, the GUIDE study was carried out to estimate seroprevalence on the national and regional levels.

Methods

To obtain an overview of the population-wide immunity against SARS-CoV-2 among adults in Germany that would be as statistically robust as possible, serological tests were carried out using self-sampling dried blood spot cards in conjunction with surveys, one by telephone and one online. Blood samples were analyzed for the presence of antibodies to the S and N antigens of SARS-CoV-2.

Results

Among the 15 932 participants, antibodies to the S antigen were detected in 95.7%, and to the N antigen in 44.4%. In the higher-risk age groups of persons aged 65 and above and persons aged 80 and above, anti-S antibodies were found in 97,4% and 98.8%, respectively. Distinct regional differences in the distribution of anti-S and anti-N antibodies emerged. Immunity gaps were found both regionally and in particular subgroups of the population. High anti-N antibody levels were especially common in eastern German states, and high anti-S antibody levels in western German states.

Conclusion

These findings indicate that a large percentage of the adult German population has formed antibodies against the SARS-CoV-2 virus. This will markedly lower the probability of an overburdening of the health care system by hospitalization and high occupancy of intensive care units due to future SARS-CoV-2 waves, depending on the viral characteristics of then prevailing variants.


By November 2022, there were around 36 million officially registered SARS-CoV-2 infections in Germany (1). The proportion of undetected infections is estimated to be 1.5 to four times the number of recorded cases (2, 3). With the spread of the Omicron variant, it is likely that the proportion of unreported cases has increased even more.

According to information from the Robert Koch Institute (RKI), around 65 million people (77.9%) have received at least one dose of vaccine against SARS-CoV-2, 63.5 million people (76.3%) at least two doses, and 51.9 million people (62.4%) have had three doses (4). There are also uncertainties about the vaccination rate, since surveys indicate that there may be under-reporting of about 5% (5). In order to assess the pandemic risk situation, it is important to estimate the number of persons exposed to SARS-CoV-2 after vaccination or infection.

During the transition phase from a pandemic situation to recurrent seasonal waves, it is important to determine the level of underlying population immunity, not least to be able to estimate the need for population-based infection control measures. Seroprevalence studies conducted so far create a picture of the population’s antibody responses to SARS-CoV-2 up until February 2022, but include neither the Omicron waves nor the various stages of immune response, including hybrid immunity (6). Assessment of immunity in the population is a complex matter because many people were vaccinated at different times during the pandemic in a different order. They also suffered an infection – sometimes with different variants of the SARS-CoV-2 virus.

Furthermore, regional and demographic differences with regard to immunity in Germany and risk-specific differences in vaccination recommendations also need to be considered. The main focus here is on the elderly as well as on vulnerable patients who have a high risk of suffering severe disease due to their poor immune protection (7).

Seroprevalence of antibodies against the S and N antigens throughout Germany was therefore determined in a random sample of adults in order to gain a statistically robust overview as possible of the anti-SARS-CoV-2 immunity status of the adult population in Germany and to be able to assess regional and demographic differences. The presence of antibodies against the N antigen is an indication of previous infection, regardless of the person’s vaccination status, whereas presence of antibodies against the S antigen is a sign of either earlier infection or vaccination. The present results provide an overview of the humoral immunity status in the German-speaking adult population living in private households until September 2022, exposing immunity gaps both regionally and demographically.

Methods

Sample survey

The German-speaking population aged 18 years and over, living in private households in Germany, was defined as the study population.

Participants were identified using an online access panel, the PAYBACK Panel, and a telephone interview (CATI, computer assisted telephone interview). Based on predefined factors relating to gender, age groups, education, federal state, and regional distribution, a random selection was taken from more than 130 000 persons (PAYBACK), and 28 965 invitations were sent out (efigure 1). In order to more fully cover the study population of the German-speaking community aged 65 years and over living in private households, the data of 1500 people aged 65 years and over were also gathered in addition by telephone (CATI). The survey results were verified using a random sample from the residents’ registration office of Berlin, taken as an example, in order to classify and validate the combined sample. All three samples were weighted using different methods. More detailed information on the sample survey and weighting may be found in the eMethods section and in the eTables 1, 2). The University Hospital of Bonn granted a positive ethics opinion to conduct the study (application 202/22). The study is registered under DRKS00029693.

eFigure 1.

eFigure 1

Presentation of the sample survey process

The questionnaire employed to survey the participants is attached in the online section of the present article (eQuestionnaire), where more detailed information on the serological laboratory analysis is also to be found.

Statistical data analysis

The characteristics of the study population and the number of declared vaccinations and infections were analyzed and estimated, stratified by age according to an appropriate weight variable. Stratification by age was performed for the groups 18–29, 30–34, 35–39, 40–49, 50–59, 60–64, 65–79, and over 80 years. Age-stratified estimates of the prevalence of antibodies to the S and N antigens were made using a second weight variable especially created for laboratory parameters, and the geographical distribution was presented according to NUTS2 and NUTS3 regions (NUTS, nomenclature of territorial units for statistics; administrative districts level as well as the rural and urban district level).

Antibody titers above 35.2 BAU/mL against the anti-S protein and an antibody ratio greater than, or equal to, 1.1 for the anti-N antigen were defined as positive. Another endpoint were those individuals without any immunity who stated in their questionnaire that they were neither vaccinated nor had been infected, and whose antibody tests were both negative. The proportion of persons in the sample with two booster vaccinations according to the current STIKO (German Standing Committee on Vaccination) recommendation was combined as one variable and presented by age groups (8).

All statistical analyses were performed using Stata 17.1. Estimates of total prevalences, mean values, stratified results, and corresponding 95% confidence intervals were implemented with the aid of the Survey Data [SVY] command and the two weight variables (9). Two-tailed t-tests were used to compare the mean values between groups in the sample. The geographic presentations (Figures 1, 2, eFigures 25) according to NUTS regions are based on map data from Eurostat (10) and were visualized with the Stata package SpMap. The detailed report on the methods of the sample survey and the code used for preparing and analyzing the data in the form of STATA do-files may be found at www.github.com/kaischulzewundling/guide.

Figure 1.

Figure 1

Proportions of persons with proven positive S-antigen immunity in %, stratified by rural and urban districts (NUTS-3) and weighted by population data

eFigure 2.

eFigure 2

Proportions of persons without declared vaccination, infection, or proof of immunity in %, stratified by rural and urban districts (NUTS-3) and weighted by population variables

eFigure 5.

eFigure 5

Proportion of persons 60+ with two booster vaccinations according to the STIKO recommendation in %, stratified by rural and urban districts (NUTS-3)

Results

Included in the study were a total of n = 15 932 participants who completed the questionnaire between June and September 2022, either online or by telephone, and returned the dried blood spot test kits to the laboratory (study population characteristics in Table 1). It was not possible to evaluate 1203 dried blood spot cards as they had not been provided with enough blood. The participants were 52.0 years old on average (95% confidence interval: [51.7; 52.3]), of whom 52.3 % [51.5; 53.2] were female. 22.5% [21.9; 23.2] of the study population had a university degree and 78.1% [77.2; 78.9] were employed. 47.1% of participants were working full-time and 15.3% part-time, while 27.8% were retired. About 40% of the study participants were active or former smokers. The most common pre-existing conditions were hypertension (30.6%), lung diseases (12.1%), and diabetes (9.9%).

A positive antibody test against the S antigen was found in 95.7% [95.3; 96.1] of the study participants and a positive antibody test against the N antigen in 44.4% [43.5; 45.3] (table 2). Positive antibodies (to S or N antigen) were detected in a total of 14 398 study participants with evaluable dried blood spot samples. Prevalences were similar across age groups 18–64 for anti-S antigen, but significantly higher for the age groups of over 65-year-olds and above all for the over 80-year-olds with 98.8 % [97.2; 99.5]. For the N antigen, however, the estimated N antibody positivity decreased continuously with increasing age group (here, the age group 80+ had the lowest value of 28.5 % [24.4; 33.0]. The majority of participants were at least triple vaccinated. 57.4 % of the participants stated that they had not yet been infected [56.6; 58.3].

Table 2. Proportions of S and N antigen detections and declared vaccinations and infections, stratified according to age and weighted according to population variables (in % [95% confidence interval]).

Total 18–29 years 30–34 years 35–39 years 40–49 years 50–59 years 60–64 years 65–79 years 80 years
S antigen N 14 464 1297 1144 1146 2291 3397 1430 3291 468
 positive 95.69
[95.29; 96.05]
96.15
[94.75; 97.19]
94.42
[92.64; 95.79]
93.51
[91.62; 94.99]
94.09
[92.92; 95.07]
95.37
[94.52; 96.10]
95.02
[93.53; 96.19]
97.35
[96.61; 97.93]
98.84
[97.21; 99.52]
 negative 4.31
[3.95; 4.71]
3.85
[2.81; 5.25]
5.58
[4.21; 7.36]
6.49
[5.01; 8.38]
5.91
[4.93; 7.08]
4.63
[3.90; 5.48]
4.98
[3.81; 6.47]
2.65
[2.07; 3.39]
1.16
[0.48; 2.79]
N antigen N 14 465 1297 1144 1146 2291 3397 1430 3291 469
 positive 44.38
[43.47; 45.29]
58.62
[55.73; 61.45]
53.97
[50.90; 57.01]
52.42
[49.36; 55.47]
50.01
[47.87; 52.15]
41.77
[40.05; 43.50]
37.32
[34.76; 39.95]
35.31
[33.55; 37.10]
28.49
[24.35; 33.02]
 negative 55.62
[54.71; 56.53]
41.36
[38.55; 44.27]
46.03
[42.99; 49.10]
47.58
[44.32; 50.28]
49.99
[47.85; 52.13]
58.23
[56.50; 59.95]
62.68
[60.05; 65.24]
64.69
[62.90; 66.45]
71.51
[66.98; 75.65]
Vaccinations 15 640 1382 1206 1217 2412 3612 1540 3700 571
 0 5.09
[4.73; 5.48]
5.85
[4.68; 7.30]
8.31
[6.78; 10.16]
8.60
[7.04; 10.46]
7.47
[6.46; 8.63]
4.78
[4.11; 5.55]
3.84
[2.97; 4.96]
2.10
[1.68; 2.63]
1.61
[0.75; 3.42]
 1 0.93
[0.78; 1.11]
1.37
[0.83; 2.24]
1.39
[0.83; 2.31]
1.46
[0.91; 2.43]
1.55
[1.12; 2.14]
0.84
[0.56; 1.26]
0.72
[0.40; 1.27]
0.26
[0.14; 0.49]
0
 2 9.25
[8.76; 9.76]
16.57
[14.59; 18.76]
14.63
[12.67; 16.84]
14.63
[12.69; 16.81]
10.83
[9.61; 12.19]
8.24
[7.36; 9.22]
5.78
[4.70; 7.08]
3.77
[3.17; 4.48]
2.68
[1.53; 4.67]
 3 67.99
[67.17; 68.79]
73.70
[71.17; 76.08]
73.04
[70.35; 75.58]
71.52
[68.8; 74.09]
74.79
[72.97; 76.53]
76.68
[75.24; 78.07]
73.23
[70.9; 75.43]
54.59
[52.87; 56.29]
34.02
[29.80; 38.52]
 4+ 16.65
[15.99; 17.33]
2.5
[1.73; .3.47]
2.54
[1.82; 3.57]
3.76
[2.83; 4.99]
5.36
[4.50; 6.21]
9.39
[8.48; 10.39]
16.44
[14.63; 18.43]
39.28
[37.50; 40.97]
61.69
[56.92; 66.87]
Infections N 15 580 1377 1197 1213 2403 3599 1536 3689 566
 0 57.41
[56.56; 58.25]
40.55
[37.84; 43.31]
42.62
[39.77; 45.51]
43.63
[40.13; 46.52]
46.28
[43.97; 48.19]
59.01
[57.35; 60.65]
66.30
[63.84; 68.67]
74.57
[73.05; 76.04]
81.68
[77.86; 84.96]
 1 39.34
[38.52; 40.18]
53.17
[50.38; 55.94]
52.64
[49.73; 55.54]
51.15
[48.25; 54.05]
49.60
[47.54; 51.65]
38.54
[36.92; 40.19]
31.51
[29.19; 33.94]
24.34
[22.90; 25.85]
16.84
[13.67; 20.58]
 2+ 3.25
[2.95; 3.57]
6.28
[5.07; 7.76]
4.74
[3.64; 6.15]
5.21
[4.02; 6.74]
4.12
[3.37; 5.03]
2.45
[1.98; 3.02]
2.19
[1.55; 3.08]
1.09
[0.79; 1.50]
1.48
[0.74; 2.94]

According to the STIKO recommendation for COVID vaccination, a second booster vaccination is recommended for all persons over 60 and all persons of the age group 18–59 with one or more risk factors. Altogether, 46.0% [45.2; 46.9] of the sample had one or more risk factors according to the recommendation or were over 60 years old (table 3). In the age groups 18–29, 30–34, 35–39, and 40–49, between 16 and 32% of the persons had one or more risk factors for a more severe disease, depending on their group. The proportion of participants who had stated they had neither a positive antibody test against the S or N antigen nor declared in their questionnaire that they had been vaccinated or suffered infection was 1.2% of all study participants [1.1; 1.4] (efigure 2). This proportion was the lowest at 0.3% [0.0; 2.0] in the age group of those over 80 years, while it was the highest in the age group 40–49 years at 1.67 % [1.2; 2.3]. Altogether, 27.0% [25.9; 28.1] of the participants fulfilled the vaccination recommendation for a second booster. The proportion of persons with a second booster vaccination according to the STIKO recommendation was the highest in the age group 80+ at 61.7% [56.9; 66.9] and the lowest in the age group of the 30 to 34-year-olds at 3.8 %.

Table 3. Proportions of persons with increased risk of severe disease*1, stratified by age and weighted by population variables (in % [95% CI]).

Total 18–29 years 30–34 years 35–39 years 40–49 years 50–59 years 60–64 years 65–79 years 80 years
Risk of severe disease N 15 643 1382 1206 1217 2412 3612 1540 3703 571
 yes 46.03
[45.18; 46.88]
16.30
[14.32; 18.48]
19.80
[17.58; 22.24]
22.59
[20.26; 25.10]
32.14
[30.25; 34.09]
48.16
[46.49; 49.83]
100*2 100 100
 no 53.97
[53.12; 54.82]
83.70
[81.52; 85.68]
80.20
[77.76; 82.42]
77.41
[74.90; 79.74]
67.86
[65.91; 69.75]
51.84
[50.17; 53.51]
Immunity N 15 643 1382 1206 1217 2412 3612 1540 3703 571
 immunity 98.78
[98.58; 98.94]
98.85
[98.11; 99.30]
98.60
[97.69; 99.15]
98.97
[98.16; 99.42]
98.33
[97.71; 98.78]
98.65
[89.81; 93.70]
98.52
[97.77; 99.02]
99.00
[98.61; 99.28]
99.72
[98.05; 99.96]
 no immunity 1.22
[1.06; 1.42]
1.15
[0.70; 1.89]
1.40
[0.85; 2.31]
1.03
[0.58; 1.84]
1.67
[97.71; 98.78]
1.35
[1.02; 1.80]
1.48
[0.98; 2.23]
1.00
[0.72; 1.39]
0.28
[0.04; 1.95]
Second booster vaccination N 9047 219 234 278 775 1740 1540 3692 569
 yes 26.98
[25.92; 28.06]
4.63
[2.46; 8.55]
3.76
[2.02; 6.88]
4.60
[2.67; 7.81]
8.04
[0.63; 10.19]
12.76
[11.26; 14.42]
16.44
[14.63; 18.43]
39.28
[37.50; 40.97]
61.69
[56.92; 66.87]
 no 73.02
[71.94; 74.08]
95.37
[91.45; 97.54]
96.24
[93.12; 97.98]
95.40
[92.19; 97.33]
91.96
[89.81; 93.70]
87.24
[85.58; 88.74]
83.56
[81.57; 85.37]
60.86
[59.17; 62.53]
38.41
[34.03; 42.99]

*1 According to the STIKO recommendation of November 17, 2022 (persons aged 60 years and over, persons aged 18 to 59 years with chronic cardiovascular, lung, cancer, or neurological diseases and diabetes); proportions of persons with any type of immunity (defined as: declared vaccination or infection or positive S or N antigen test); proportion of persons with an increased risk according to the STIKO recommendation with double booster vaccination.

*2 According to the recommendation, all persons over 60 years have an increased risk.

CI, confidence interval

The regional distribution of underlying population immunity at the rural or urban level is an important parameter (Figures 1, 2 and eFigure 2). The highest proportions of antibody detection against the S antigen was seen, amongst others, in the districts of Coburg, Fürth, Hof, Potsdam, Ravensburg, Rostock, and Tuttlingen at about 100%, whereas antibody detection against the S antigen was below 80% in the districts of Hohenlohe, Hildburghausen, Kempten, Memmingen, and Suhl.

Figure 2.

Figure 2

Proportions of persons with proven positive N-antigen immunity in %, stratified by rural and urban districts (NUTS-3) and weighted by population data

There were also similar regional differences for antibodies to the N antigen. While antibody prevalence of more than 75% against the N antigen was found in districts such as Deggendorf, Hof, Memmingen, Rhein-Lahn, and Saale-Orla, the antibody prevalence in Ansbach, Baden-Baden, Bremerhaven, Emden, Osterholz, and Wolfsburg, for example, was comparatively low at below 20%. The regional differences become even more pronounced when presented by NUTS2 regions (eTable 3, eFigures 3, 4). There is a tendency to observe the highest proportions of positive anti-S antigen antibodies in the western federal states and the highest proportions of anti-N antigen antibodies in the eastern states.

eTable 3 **. Proportions of persons without vaccination, infection, or immunity, stratified by administrtive district (NUTS-2), weighted by population variables.

Vaccination, infection or immunity not specified
NUTS2 N Proportion (%) [95% CI]
DE11 Stuttgart 824 1.38 [0.67; 2.79]
DE12 Karlsruhe 536 0.66 [0.24; 1.85]
DE13 Freiburg 554 1.54 [0.71; 3.33]
DE14 Tübingen 435 0.29 [0.00; 2.11]
DE21 Upper Bavaria 1 055 0.92 [0.47; 1.77]
DE22 Lower Bavaria 355 1.99 [0.74; 5.29]
DE23 Upper Palatinate 336 0.52 [0.00; 3.59]
DE24 Upper Franconia 309 0.93 [0.22; 3.75]
DE25 Middle Franconia 422 0.49 [0.12; 1.94]
DE26 Lower Franconia 395 1.28 [0.04; 3.95]
DE27 Swabia 409 1.10 [0.04; 2.99]
DE30 Berlin 613 1.80 [0.99; 3.27]
DE40 Brandenburg 548 1.26 [0.59; 2.68]
DE50 Bremen 120 3.72 [1.33; 9.94]
DE60 Hamburg 336 0.59 [0.15; 2.53]
DE71 Darmstadt 524 1.16 [0.58; 2.29]
DE72 Giessen 231 1.17 [0.25; 5.24]
DE73 Kassel 293 1.15 [0.33; 3.86]
DE80 Mecklenburg-Western Pomerania 375 1.88 [0.84; 4.14]
DE91 Brunswick 326 0.28 [0.00; 1.94]
DE92 Hanover 472 2.42 [1.33; 4.36]
DE93 Luneburg 456 1.69 [0.75; 3.78]
DE94 Weser-Ems 536 1.42 [0.68; 2.92]
DEA1 Düsseldorf 731 1.52 [0.87; 2.64]
DEA2 Cologne 815 0.46 [0.17; 1.22]
DEA3 Münster 385 1.05 [0.27; 4.01]
DEA4 Detmold 382 0.67 [0.15; 2.90]
DEA5 Arnsberg 438 0.41 [0.12; 1.35]
DEB1 Koblenz 383 0.00 [0.00; 0.00]
DEB2 Trier 166 0.00 [0.00; 0.00]
DEB3 Rhenish Hesse-Palatinate 485 0.45 [0.11; 1.86]
DEC0 Saarland 182 0.84 [0.21; 3.39]
DED2 Dresden 354 2.63 [1.36; 5.02]
DED4 Chemnitz 343 0.77 [0.19; 3.07]
DED5 Leipzig 221 1.62 [0.06; 4.32]
DEE0 Saxony-Anhalt 432 1.95 [0.95; 3.97]
DEF0 Schleswig-Holstein 715 1.74 [0.89; 3.36]
DEG0 Thuringia 482 2.46 [1.21; 4.92]
Total 1.20 [1.04; 1.40]

N, number of study participants per district; CI, confidence interval

eFigure 3.

eFigure 3

Proportions of persons with proof of immunity to S antigen in %, stratified by administrative district (NUTS-2) and weighted by population data

The analysis of vaccination history conducted regionally at the level of rural or urban districts based on the STIKO recommendations also revealed large regional differences (presented in eFigure 5 for the age group 60+). There was a tendency for less vaccinations to be declared in the southeast and more vaccinations in the northwest. While less than 5% of the participants in the districts of Bautzen, Darmstadt, Frankfurt/Oder, and Stendal, amongst other places, reported having had a second booster vaccination at the time of the survey, this figure was more than 75% of the over 80-year-olds in Delmenhorst, Offenbach am Main, Rosenheim, Wilhelmshaven, and Worms.

Discussion

As part of the IMMUNEBRIDGE project, the GUIDE study was able, for the first time, to identify within in a very short time the exposure status to SARS-CoV-2 in the German speaking adult population using a self-sampling method. Anti-S seroprevalence in the general population was very high at 95.7 %. There were regional differences, however, indicating on the one hand a regional variation in the incidence of infection as well as different vaccination rates, which were particularly lower in the Eastern federal states. There was also an anti-N seropositivity of 44.4%, which is an indication of having been infected with SARS-CoV-2.

A series of seroprevalence studies of anti-SARS-CoV-2 antibodies has so far been conducted in Germany (6). To date, the majority of these, however, were restricted only to a certain region, so there has been no sample survey covering all areas of Germany. They were also conducted during the earlier times of the pandemic (and so do not take into consideration the largest proportion of the infections which have developed as the result of the appearance of the Omicron variant). Furthermore, these studies did not distinguish between antibodies to the S or N antigen, or there was no further reaching information on risk factors or number of vaccinations to be able to assess immunity in detail. The concept of subject acquisition and sampling by means of self-sampling and dried blood spot cards proved itself to be extraordinarily efficient and robust during the present study.

The GUIDE study, with its design, its sampling method, specimen collection modality, and its extensive questionnaire allowed a large group of participants to be mobilized within a time span of five months with a very high return rate. To the best of our knowledge, the GUIDE study is so far the only available study in Germany which has examined at this speed robust generalizations about the entire population with regard to seroprevalence of anti-SARS-CoV-2 antibodies. The combination of questionnaire with anti-S and anti-N antibody tests allowed a good assessment of the exposure status of the German speaking adult population and integration into current scenario modeling (11, 12).

Limitations

The study was limited to adults aged 18 years and over, while SARS-CoV-2-specific T-cell and memory-B-cell responses, neutralization assays, and other critical components of acquired immunity were not assessed. Even if antibody levels show that there is a certain underlying protection, they do not allow the degree and quality of the immunity to be fully assessed. Indeed, some people with detected antibodies can still suffer severe COVID-19 disease, especially if they belong to a high-risk group. Apart from antibody immunity, the number of vaccinations, in addition to laboratory tests for T and B cells, can most likely be regarded as a further surrogate of immunity and protection from severe disease. This way, complex endpoints can be formed which correlate well with markers of immunological protection, such as neutralizing antibodies (12, 13). Furthermore, it is also known from various systematic reviews and meta-analyses of epidemiological and vaccine effectiveness studies that immunity against (re-)infection decreases within a few months, yet is sustained against severe disease (14, 15). A recent meta-analysis established that a high level of lasting protection against severe disease exists from previous infection and hybrid immunity through vaccination and infection (but not against re-infection) (16).

Furthermore, high-risk groups in particular were underestimated due to the study design: Surveys on comorbidities are based on independent statements by the participants and allow only a rough classification of concomitant diseases since these data were not recorded in medical terms. Furthermore, unvaccinated individuals were possibly underrepresented in the study. According to data from the RKI, 86% of the adult population had received basic immunization, while this was 94% in the GUIDE study (4). Publication of the 7th report of the COVIMO study revealed that data from RKI’s digital vaccination rate monitoring possibly underestimates the actual vaccination rate by about 5%, while the COVIMO survey possibly over-estimates vaccination rates (5, 17).

Most recent estimates from the COVIMO study are very similar to those of the GUIDE study (COVIMO versus GUIDE 18–39 years: 86% versus ˜90%; 40–59 years: 94% versus 95%; 60+ years: 96% versus ˜97%) (17). Accordingly, a slight overestimation of the vaccination rate and underreporting of the unvaccinated may also be assumed in the GUIDE study. Qualitatively speaking, this underestimation does not change the conclusions, however, given the probably low absolute difference from actual values. Future studies would also benefit from data acquisition from the cohort of high-risk groups in the form of a prospective clinical trial, in addition to the population-based approaches of a representative analysis. Apart from strictly registering main and secondary diagnoses, serological examinations to gather data on humoral immunity, together with the necessary analysis of cellular immunity, could also be conducted.

It should also be noted that, since the end of the survey period, a further 3.5 million infections have been officially reported which are not reflected in the data and for which reason the anti-N-antibody prevalence is under-estimated.

Conclusion

Our findings show that a large proportion of the general population in Germany have humoral immunity to SARS-CoV-2. Depending on the characteristics of the particular SARS-CoV-2 variant, there will be a significant reduction in the likelihood of healthcare system overload scenarios due to hospitalizations and the need for intensive medical care for patients with COVID-19 in the next waves of the disease, compared with a situation without this immunity status in the population.

Protection against severe disease is due to a considerable level of cellular immunity, which is also further reinforced robustly against variant evolution (18, 19) and by repeated exposition. Although the GUIDE study does not measure cellular immunity directly, measurement of antibody levels and details about vaccinations and infections allow conclusions to be drawn about ongoing protection against severe disease (20). This is also suggested by the more moderate burden on intensive care units during the last two Omicron waves in the summer and fall of 2022 as compared with previous waves of infection. Predictions about the impact of future waves of infection alone, based on immunity status, are nevertheless not possible.

Supplementary Material

eMethods

Sample survey

Preliminary planning

The aim of the study was to gain an overview that is as statistically robust as possible about antibody prevalence of the adult population in Germany. The study design was structured with a mixed-mode approach to include different options and data collection methods. A survey in residential homes and care facilities would only have been possible—if at all—with considerably more effort, so it was rejected for practical research reasons. The German-speaking population aged 18 and over, living in private households with primary residence in Germany, was therefore defined as the study population.

Survey sample drawn from the PAYBACK Panel

Sampling

By far the largest portion of the study participants was recruited from an online access panel, the PAYBACK panel. For years, dimap and its subsidiary institute Infratest dimap have been working with PAYBACK on behalf of various clients. This has been of advantage both for implementing a reliable sampling method and for managing projects. The online panel comprises more than 130 000 active panelists. So it was possible to include not only marginal distributions but also some cross ratios. From this, a gross sample comprising 45 000 panelists was selected separately for West and East German states.

The first step in conducting the study was to randomly draw a gross sample from the PAYBACK panel with due consideration of specified marginal distributions.

Sources of the sample specifications

Given that there are no official structural figures for the German-speaking population, certain figures were used as a substitute for the German population, namely:

The voting behavior at the German Bundestag elections of 2021 for a subset of panelists was collected by Infratest dimap. A corresponding target was used for this subset, taken from the official election results. The marginal distribution for the random sample took the following parameters into consideration:

Recruitment of participants

The invitation to the panelists referred, on the one hand, to the health status survey and, on the other hand, to the option that, after providing their own address, participation in a self-test to check COVID antibody status would be possible. At the end of the questionnaire, a separate window was available to insert an address, after which it was immediately and exclusively stored on the laboratory’s server, together with a link-ID for later matching of questionnaire and test result. The laboratory sent the test kits together with instructions for use and a declaration of consent for signing (both after receiving a positive opinion from the ethics council) and a stamped addressed return envelope. The design of the test kit and the description of how to collect blood were very easy to understand and supported by a link to a description video.

Two reminders and two follow-ups were sent: PAYBACK reminded non-participants after three and ten days by email of their study participation or of the previously sent invitation. PAYBACK also conducted follow-ups of belated test kits by email. Two weeks before the completion of the study, the participants were informed of the closing date, and they were still given the opportunity to return their test kit to the laboratory for evaluation. The laboratory sent all participants information of their laboratory results by post. Over 90% of the dispatched test kits were returned to the laboratory.

Utilization

A total of 28 965 invitations were sent out. Altogether, 19 044 participants completed the questionnaire, of whom 16 415 provided their address for the dispatch of a test kit (etable 1).

Telephone interview

In order to back up the sample from the PAYBACK panel, an additional 1500 persons (target sample) aged 65 years and over were contacted by telephone (CATI, computer assisted telephone interview). Interviews were only considered complete if an address had been submitted.

For this sample, the pool of telephone numbers was drawn upon for persons who had in the past given their consent to be re-contacted for a telephone survey and were of the same age as the target group. In addition, it was to be expected that willingness to participate in a telephone survey on health matters as well as to provide an address would be considerably greater in persons of the pool than during an initial telephone contact. In this respect, this was a practical research decision.

The German-speaking population aged 65 years and over, living in private households in Germany, was taken as the study population. Persons prepared to participate formed the telephone number pool for the current survey. That means that random samples were taken from this pool, stratified according to the combination of Federal state and BIK district type. The selection was limited to telephone numbers connected with a household with at least one target person aged 65 or older.

Utilization

A total of 15 151 telephone numbers were used, and at least one call attempt was made (etable 2).

Addresses were collected according to the same pattern as used for the PAYBACK panel: The address was immediately stored on the laboratory server, together with a link ID previously generated by the laboratory. That means that, although the telephone interviewer heard the address and entered it into the system, it was only stored on the laboratory server.

Survey sample drawn from a residents’ registration office

The aim of this part of the study was to use a high-quality sample of the adult German-speaking population of Berlin as a basis for classifying and validating the Berlin questionnaire results of the large-scale nationwide CATI study collected using the PAYBACK panel.

Samples from the Residents’ Registration Office (EWMA) are of higher quality than samples from an online access panel or from telephone surveys. They are used in many best-practice surveys for population studies. Apart from the advantage of clearly defined options with reliable selection probabilities and failure probabilities, they also have the benefit that (predominantly valid) addresses are provided by the registration offices and do not first need to be enquired from the respective participants. On the other hand, there is the drawback that considerable organizational challenges and research costs arise if the registration offices need to be contacted in larger numbers when there is no central resident register.

Conducting a study based on registration office data is easier for a town rather than a survey covering several regions. It is also possible within the specified time scale. Furthermore, it helps to classify the results of the online survey appropriately. For these reasons, we conducted the survey in Berlin using random registration office samples. Berlin cannot claim to be a scaled-down representation of the whole of Germany, but given that Berlin, as the largest city with a population of almost four million, comprises around 4% of the study population, a correspondingly large proportion of the sample formed from the PAYBACK panel is also located in Berlin.

The study population comprised all persons aged 18 years and older living in Berlin. The sample was obtained from a selection of individuals registered with their primary place of residence in Berlin and born before August 01, 2004. The samples were drawn from the register for the city of Berlin as part of a group information request according to Section 46 of the Federal Registration Law. The area of examination was the city of Berlin (district municipality code 11000000).

The study adopted a mixed-mode design as a written survey using a self-administered questionnaire (PAPI, paper and pencil interviewing) or, alternatively, as an online survey (CAWI, computer-assisted web interviewing). The selected target persons could therefore choose individually their preferred mode of participation. The usable participation channels were thus extended, and at the same time sampling disadvantages of a purely online survey (undercoverage for target persons with no internet connection) were avoided.

Data was collected online and transferred to dimap whose staff checked it, as with the results from the PAYBACK panel. The written questionnaires were gathered, checked internally, and the data sets were then combined.

Sampling frame

Respondents were selected by purely random sampling. The sample size comprised 1500 persons. Five hundred and eleven interview data sets were evaluated.

To be on the safe side, the number of addresses requested from the city of Berlin was considerably higher than the aspired gross case number in order to compensate, if necessary, for the non-response rate of those contacted.

Supplement to the weighting concept

Step 1: Structural weighting of the online survey (PAYBACK) on infection status

Although sampling was conducted as a random sample from the access panel population, it was not possible to calculate a selection probability for obtaining the total panel population. Purely structural weighting was therefore applied during the first step, that means, the distribution of the realized net sample (in this case, completed questionnaires on vaccination status) was adjusted to target structures of the study population according to the micro census and our own calculations. The following characteristics were included:

Not only were the marginal distributions adjusted, but also a number of cross ratios. Thus, a margin comprises the combination of federal state × administrative district × BIKtype of region.

Since no official target figures are available for the German-speaking population, the same sources were applied for the target figures that were used for the sampling, i.e., in particular, figures from the German population in private households. Combinations with lower numbers (n ≤3) were grouped together, and factors were further limited to the interval of 0.33–3.0.

Step 2: Weighting for non-response to the online survey (PAYBACK) on antibody status

Use of an evaluable test kit represented a selection process that resulted in non-random failures. Apart from failures for technical reasons, for example, wrong or incomplete addresses, experience has shown that certain groups of people have their reservations about tests and so tend to participate in them less often. Such failures resulted in bias, and the subsample with a test kit needed to be compensated for.

The failure was modeled using the realized sample with vaccination status information via logistic regression. This allowed estimation of different participation probabilities. The failure correction factor is calculated from the reciprocal value of the participation probability. All the characteristics of the total (partial) sample were available as independent variables. The following characteristics were included:

Missing values were excluded when generating the logit model and replaced by the mode when calculating the participation probability.

In a further step, the default-weighted sub-sample with an evaluable test kit was calibrated again to distributions from the official statistics, as with the total sample. The person factor of the total (partial) sample and the failure factor were used as input factors in the calibration of the subsample with test kit. The same characteristic were again used. However, the smaller number of cases in the subsample with test kit required, in part, stronger combinations of low-population cells.

Separate weighting of the telephone interview

Separate weighting for those persons participating in the CATI was performed in a similar fashion. The target figures for the structural weighting of this subsample of the target persons (German-speaking individuals age 65 years and older living in Germany) were also based on the results of the micro census and our own calculations. Cross quotes were also used, albeit with fewer cells due to the lower number of cases. Thus, the regional margin comprises a combination of federal states and BIK type of region. Administrative districts were no longer included. Independent of the logit analysis for the online cases, a logit analysis was conducted for the CATI cases, since the failure mechanisms were different due to the different interview situation.

Combining the partial samples

Both partial samples (PAYBACK and CATI) were then combined. This was done for both the overall vaccination status samples and the two subgroups with test kits.

Since the CATI sample only included persons aged 65 years and over, there was disproportionality in the age distribution. This disproportionality had to be corrected when the two partial samples were combined. An adjustment of the factor limits to the interval 0.3–3.3 was required to obtain equivalent adjustments in comparison with the subsamples.

Input factor for proportionalization was, in each case, the personal weighting of the total or partial samples. Weighting for non-response for the subsamples with test kits had already been performed in the previous steps and did not need to be repeated for the overall dataset.

The total data set was once again calibrated after proportionalization of the age groups. The larger number of cases of persons over the age of 65 years once again allowed a somewhat more differentiated weighting in this age group.

It was not possible to adjust the offline proportion in the course of weighting, as this proportion is also not realistically represented in the CATI sample due to the type of incentive offered (online voucher).

Weighting for the sample from the residents’ registration office

A separate weighting was performed for the residents’ registration office sample. On completion of the data check and creation of a harmonized data set from the PAPI and CAWI data sets, a weighting factor was calculated for each target person. During population surveys it is usual to align the structure of the participants in a study with the structure of the study population using this type of weighting. In the present case, the distribution of the characteristics age, sex, and urban district was used in the gross sample taken from the registration office as a target weighting parameter for the study population.

Serological laboratory analysis

The study participants were sent an individually assigned blood collection set—comprising instructions, informed consent declaration, sampling material, including a dried blood spot card, and a return envelope. With due consideration of the required hygiene measures, the blood sample was taken by the participants themselves using a lancet (Owen Mumford). At least two spots of blood were to be dropped on the dried blood spot card, dried for two to four hours, and then sent back to the laboratory in the enclosed return envelope.

The return envelopes were first checked in the laboratory for a correctly completed consent form. Then two disks each were punched out of the dried blood spot card, which had been completely soaked in the blood, measuring 4.7 mm in diameter (Panthera-PuncherTM 9, Perkin Elmer). The specimen was discarded if it was not possible to extract two punch disks soaked with blood.

The punch disks were eluted in 500 µL sample buffer (Sample Buffer Blue, Euroimmun) in microtiter dilution plates (Deep-well plate 1 mL, Euroimmun). The microtiter dilution plates were shaken for 30 seconds at 1000 rpm and then incubated for 1 hour at 37 °C. After centrifuging for 5 minutes at 1207 g at room temperature, DBS holders (Euroimmun) were inserted in the wells. Finally, the samples were analyzed for SARS-CoV-2 antibodies (against spike antigen [S antigen]: EUROIMMUN Anti-SARS-CoV-2 QuantiVac-ELISA [IgG]; against nucleocapsid protein [N antigen]: NEUROIMMUN Anti-SARS-CoV-2 NCP-ELISA [IgG]) using the ELISA technique (EUROLabWorkstation, Euroimmun).

  • Population statistics as of 31 December 2019 based on the latest territorial boundaries of 31 December 2020

  • Education statistics from the 2020 Micro Census

  • Other general population figures (sex × age) taken from updated data, based on Census 2011.

  • Sex

  • Age groups

  • Education

  • Federal state

  • BIK regions (BIK, regional classification system for Germany in 10 categories).

  • Federal state

  • Administrative region

  • BIK type of region

  • Number of people in household

  • Highest educational qualification

  • Age

  • Sex.

  • Federal state

  • BIK type of region

  • Age

  • Sex

  • Highest school leaving qualification

  • Number of pre-existing conditions

  • Number of COVID infections

  • Symptomatic COVID infection

  • Emergency hospital admission

  • Vaccination status

  • Willingness to be vaccinated.

Table 1. Participant characteristics, weighted by population variables.

N total 15 932
Age Proportion in % [95% CI]
 18–29 years 12.88 [12.24; 13.55]
 30–34 years 7.34 [6.93; 7.76]
 35–39 years 8.01 [7.57; 8.46]
 40–49 years 14.91 [14.35; 15.49]
 50–59 years 22.17 [21.51; 22.85]
 60–64 years 9.80 [9.33; 10.29]
 65–79 years 17.55 [16.99; 18.13]
 ≥≥80 years 7.34 [6.73; 8.00]
 Age (average in years) 51.99 [51.66; 52.32]
Sex (% female) 52.34 [51.47; 53.21]
Highest educational achievement (in %)
 Lower secondary school leaving exam 14.98 [14.37; 15.61]
 Middle secondary school leaving exam 46.75 [45.91; 47.60]
 University entrance exam 15.34 [14.75; 15.94]
 University degree 22.53 [21.86; 23.20]
Employment status (in %)
 Full-time 47.08 [46.24; 47.92]
 Part-time 15.31 [14.71; 15.91]
 Temporarily without employment 1.28 [1.11; 1.49]
 In training 3.17 [2.83; 3.56]
 Without employment 2.47 [2.22; 2.74]
 Retired 27.83 [27.05; 28.62]
 Other employment status/no details 2.87 [2.54; 3.27]
Occupational status
 Employee 78.06 [77.21; 78.90]
 Civil servant 7.35 [6.84; 7.89]
 Self-employed 5.38 [4.95; 5.84]
 Craftsman 7.96 [7.42; 8.54]
 Other occupational status/no details 1.25 [1.02; 1.44]
Proportion smokers
 active 17.89 [17.27; 18.53]
 former 23.03 [22.33; 23.74]
 non-smoker 59.08 [58.25; 59.90]
Lung disease 12.07 [11.53; 12.64]
Cardiovascular disease 8.08 [7.60; 8.59]
Hypertension 30.56 [29.79; 31.35]
Stroke/neurological disorder 3.74 [3.41; 4.10]
Cancer disease 5.77 [5.38; 6.19]
Diabetes 9.91 [9.43; 10.42]
Chronic viral disease 0.59 [0.48; 0.73]

CI, confidence interval

eFigure 4.

eFigure 4

Proportions of persons with proof of immunity to N antigen in %, stratified by administrative district (NUTS-2) and weighted by population data

eTable 1. Utilization of the PAYBACK sample.

Utilization overview: Absolute
Invitations 28 965
 – undelivered invitations −202
Delivered invitations 28 763
Participation (click on survey link) 22 005
 – filtered out/no consent −706
 – dropouts −2255
Completed questionnaires 19 044
 including address details 16 415
 without address details 2629

eTable 2. Utilization of telephone questionnaire (CATI sample).

Parameter Value/Number Proportion (%)
Gross sample: used telephone numbers 15 151 100
 Telephone number not connected/non-existent 3063 20
 Fax/modem 82 1
 Not a private household (business telephone number or TAM) 56 0
 Double addressing according to CP/TP 6 0
 Blacklist 1 0
Total unusable numbers 3208 21
 private TAM/mobile mailbox (resubmit) 2000 13
 technical errors 14 0
 line engaged 283 2
 no contact despite maximum number of attempts 3127 21
 respondent hangs up before interview starts 164 1
 total non-contacts established during field time 5588 37
Contact made 6355 100
Dropouts after CP/TP contact
 no communication with CP/TP due to language problems 83 1
 TP unable/ill 24 0
 no TP in the HH 173 3
 CP/TP refuses to provide information/no interest/hangs up 3698 58
 CP/TP busy/no appointment possible or kept 152 2
 call stopped during interview 316 5
 number of dropouts after CP/TP contact 4446 70
Completed interviews 1909
 denominator: contact made 30.0
 denominator: gross sample 12.6
Interviews with address details 1500
 denominator: contact made 23.6
 denominator: gross sample 9.9

TAM, telephone answering machine; CP, contact person; TP, target person; HH, household

Acknowledgments

Translated from the original German by Dr Grahame Larkin MD.

Footnotes

Funding

The study was funded via the Network of University Medicine (NUM) as part of the IMMUNEBRIDGE project by the Federal Ministry of Education and Research (BMBF) (Research number (FKZ 01KX2121).

Conflict of interest statement

NT is a member of the Board of Directors of the DGPI [German Society for Pediatric Infectious Diseases].

AP is spokesperson of the specialist and organ-specific working groups (FOSA) of Laboratory Medicine and the UAC use and access committee of the German National Pandemic Cohort Network “NAPKON” (both are committees of the Network of University Medicine).

MN is spokesperson of the FOSA of Laboratory Medicine, spokesperson of the advisory committee, and member of the control group of the NUM Research Information System (NUM-FIS). He is also responsible for the NUM-LIMS (all named institutions are committees of the Network of University Medicine).

AK received financial support from the German Research Foundation (DFG), the BMBF, the Federal Ministry of Health (BMG), the RKI, the Institute for Public Health North Rhine Westphalia, the Volkswagen Foundation, and the Innovation Fund of the Federal Joint Committee.

BL was supported by the Helmholtz Association, the Horizon 2020 Research and Innovation Program of the European Union, the IMMUNEBRIDGE project, and the Federal Ministry of Education and Research (BMBF) via the projects RESPINOW (Modeling Platform on Respiratory Infections) and OptimAgent, a standardized model-based framework to support public health decision-making processes. She is spokesperson of the Modeling Network for Serious Infectious Diseases, Deputy President of the German Society for Epidemiology, member of the Internal Advisory Board of the German Center for Infection Research, and member of the Steering Committee of TBNet (Tuberculosis Network).

HS received a fee for participation in the discussion round at a Johnson & Johnson Open House event. He was member of the Scientific Advisory Boards of AstraZeneca and Seqirus on an honorary basis. Johnson & Johnson also supports a study led by HS.

The other authors declare that no conflict of interest exists.

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

eMethods

Sample survey

Preliminary planning

The aim of the study was to gain an overview that is as statistically robust as possible about antibody prevalence of the adult population in Germany. The study design was structured with a mixed-mode approach to include different options and data collection methods. A survey in residential homes and care facilities would only have been possible—if at all—with considerably more effort, so it was rejected for practical research reasons. The German-speaking population aged 18 and over, living in private households with primary residence in Germany, was therefore defined as the study population.

Survey sample drawn from the PAYBACK Panel

Sampling

By far the largest portion of the study participants was recruited from an online access panel, the PAYBACK panel. For years, dimap and its subsidiary institute Infratest dimap have been working with PAYBACK on behalf of various clients. This has been of advantage both for implementing a reliable sampling method and for managing projects. The online panel comprises more than 130 000 active panelists. So it was possible to include not only marginal distributions but also some cross ratios. From this, a gross sample comprising 45 000 panelists was selected separately for West and East German states.

The first step in conducting the study was to randomly draw a gross sample from the PAYBACK panel with due consideration of specified marginal distributions.

Sources of the sample specifications

Given that there are no official structural figures for the German-speaking population, certain figures were used as a substitute for the German population, namely:

The voting behavior at the German Bundestag elections of 2021 for a subset of panelists was collected by Infratest dimap. A corresponding target was used for this subset, taken from the official election results. The marginal distribution for the random sample took the following parameters into consideration:

Recruitment of participants

The invitation to the panelists referred, on the one hand, to the health status survey and, on the other hand, to the option that, after providing their own address, participation in a self-test to check COVID antibody status would be possible. At the end of the questionnaire, a separate window was available to insert an address, after which it was immediately and exclusively stored on the laboratory’s server, together with a link-ID for later matching of questionnaire and test result. The laboratory sent the test kits together with instructions for use and a declaration of consent for signing (both after receiving a positive opinion from the ethics council) and a stamped addressed return envelope. The design of the test kit and the description of how to collect blood were very easy to understand and supported by a link to a description video.

Two reminders and two follow-ups were sent: PAYBACK reminded non-participants after three and ten days by email of their study participation or of the previously sent invitation. PAYBACK also conducted follow-ups of belated test kits by email. Two weeks before the completion of the study, the participants were informed of the closing date, and they were still given the opportunity to return their test kit to the laboratory for evaluation. The laboratory sent all participants information of their laboratory results by post. Over 90% of the dispatched test kits were returned to the laboratory.

Utilization

A total of 28 965 invitations were sent out. Altogether, 19 044 participants completed the questionnaire, of whom 16 415 provided their address for the dispatch of a test kit (etable 1).

Telephone interview

In order to back up the sample from the PAYBACK panel, an additional 1500 persons (target sample) aged 65 years and over were contacted by telephone (CATI, computer assisted telephone interview). Interviews were only considered complete if an address had been submitted.

For this sample, the pool of telephone numbers was drawn upon for persons who had in the past given their consent to be re-contacted for a telephone survey and were of the same age as the target group. In addition, it was to be expected that willingness to participate in a telephone survey on health matters as well as to provide an address would be considerably greater in persons of the pool than during an initial telephone contact. In this respect, this was a practical research decision.

The German-speaking population aged 65 years and over, living in private households in Germany, was taken as the study population. Persons prepared to participate formed the telephone number pool for the current survey. That means that random samples were taken from this pool, stratified according to the combination of Federal state and BIK district type. The selection was limited to telephone numbers connected with a household with at least one target person aged 65 or older.

Utilization

A total of 15 151 telephone numbers were used, and at least one call attempt was made (etable 2).

Addresses were collected according to the same pattern as used for the PAYBACK panel: The address was immediately stored on the laboratory server, together with a link ID previously generated by the laboratory. That means that, although the telephone interviewer heard the address and entered it into the system, it was only stored on the laboratory server.

Survey sample drawn from a residents’ registration office

The aim of this part of the study was to use a high-quality sample of the adult German-speaking population of Berlin as a basis for classifying and validating the Berlin questionnaire results of the large-scale nationwide CATI study collected using the PAYBACK panel.

Samples from the Residents’ Registration Office (EWMA) are of higher quality than samples from an online access panel or from telephone surveys. They are used in many best-practice surveys for population studies. Apart from the advantage of clearly defined options with reliable selection probabilities and failure probabilities, they also have the benefit that (predominantly valid) addresses are provided by the registration offices and do not first need to be enquired from the respective participants. On the other hand, there is the drawback that considerable organizational challenges and research costs arise if the registration offices need to be contacted in larger numbers when there is no central resident register.

Conducting a study based on registration office data is easier for a town rather than a survey covering several regions. It is also possible within the specified time scale. Furthermore, it helps to classify the results of the online survey appropriately. For these reasons, we conducted the survey in Berlin using random registration office samples. Berlin cannot claim to be a scaled-down representation of the whole of Germany, but given that Berlin, as the largest city with a population of almost four million, comprises around 4% of the study population, a correspondingly large proportion of the sample formed from the PAYBACK panel is also located in Berlin.

The study population comprised all persons aged 18 years and older living in Berlin. The sample was obtained from a selection of individuals registered with their primary place of residence in Berlin and born before August 01, 2004. The samples were drawn from the register for the city of Berlin as part of a group information request according to Section 46 of the Federal Registration Law. The area of examination was the city of Berlin (district municipality code 11000000).

The study adopted a mixed-mode design as a written survey using a self-administered questionnaire (PAPI, paper and pencil interviewing) or, alternatively, as an online survey (CAWI, computer-assisted web interviewing). The selected target persons could therefore choose individually their preferred mode of participation. The usable participation channels were thus extended, and at the same time sampling disadvantages of a purely online survey (undercoverage for target persons with no internet connection) were avoided.

Data was collected online and transferred to dimap whose staff checked it, as with the results from the PAYBACK panel. The written questionnaires were gathered, checked internally, and the data sets were then combined.

Sampling frame

Respondents were selected by purely random sampling. The sample size comprised 1500 persons. Five hundred and eleven interview data sets were evaluated.

To be on the safe side, the number of addresses requested from the city of Berlin was considerably higher than the aspired gross case number in order to compensate, if necessary, for the non-response rate of those contacted.

Supplement to the weighting concept

Step 1: Structural weighting of the online survey (PAYBACK) on infection status

Although sampling was conducted as a random sample from the access panel population, it was not possible to calculate a selection probability for obtaining the total panel population. Purely structural weighting was therefore applied during the first step, that means, the distribution of the realized net sample (in this case, completed questionnaires on vaccination status) was adjusted to target structures of the study population according to the micro census and our own calculations. The following characteristics were included:

Not only were the marginal distributions adjusted, but also a number of cross ratios. Thus, a margin comprises the combination of federal state × administrative district × BIKtype of region.

Since no official target figures are available for the German-speaking population, the same sources were applied for the target figures that were used for the sampling, i.e., in particular, figures from the German population in private households. Combinations with lower numbers (n ≤3) were grouped together, and factors were further limited to the interval of 0.33–3.0.

Step 2: Weighting for non-response to the online survey (PAYBACK) on antibody status

Use of an evaluable test kit represented a selection process that resulted in non-random failures. Apart from failures for technical reasons, for example, wrong or incomplete addresses, experience has shown that certain groups of people have their reservations about tests and so tend to participate in them less often. Such failures resulted in bias, and the subsample with a test kit needed to be compensated for.

The failure was modeled using the realized sample with vaccination status information via logistic regression. This allowed estimation of different participation probabilities. The failure correction factor is calculated from the reciprocal value of the participation probability. All the characteristics of the total (partial) sample were available as independent variables. The following characteristics were included:

Missing values were excluded when generating the logit model and replaced by the mode when calculating the participation probability.

In a further step, the default-weighted sub-sample with an evaluable test kit was calibrated again to distributions from the official statistics, as with the total sample. The person factor of the total (partial) sample and the failure factor were used as input factors in the calibration of the subsample with test kit. The same characteristic were again used. However, the smaller number of cases in the subsample with test kit required, in part, stronger combinations of low-population cells.

Separate weighting of the telephone interview

Separate weighting for those persons participating in the CATI was performed in a similar fashion. The target figures for the structural weighting of this subsample of the target persons (German-speaking individuals age 65 years and older living in Germany) were also based on the results of the micro census and our own calculations. Cross quotes were also used, albeit with fewer cells due to the lower number of cases. Thus, the regional margin comprises a combination of federal states and BIK type of region. Administrative districts were no longer included. Independent of the logit analysis for the online cases, a logit analysis was conducted for the CATI cases, since the failure mechanisms were different due to the different interview situation.

Combining the partial samples

Both partial samples (PAYBACK and CATI) were then combined. This was done for both the overall vaccination status samples and the two subgroups with test kits.

Since the CATI sample only included persons aged 65 years and over, there was disproportionality in the age distribution. This disproportionality had to be corrected when the two partial samples were combined. An adjustment of the factor limits to the interval 0.3–3.3 was required to obtain equivalent adjustments in comparison with the subsamples.

Input factor for proportionalization was, in each case, the personal weighting of the total or partial samples. Weighting for non-response for the subsamples with test kits had already been performed in the previous steps and did not need to be repeated for the overall dataset.

The total data set was once again calibrated after proportionalization of the age groups. The larger number of cases of persons over the age of 65 years once again allowed a somewhat more differentiated weighting in this age group.

It was not possible to adjust the offline proportion in the course of weighting, as this proportion is also not realistically represented in the CATI sample due to the type of incentive offered (online voucher).

Weighting for the sample from the residents’ registration office

A separate weighting was performed for the residents’ registration office sample. On completion of the data check and creation of a harmonized data set from the PAPI and CAWI data sets, a weighting factor was calculated for each target person. During population surveys it is usual to align the structure of the participants in a study with the structure of the study population using this type of weighting. In the present case, the distribution of the characteristics age, sex, and urban district was used in the gross sample taken from the registration office as a target weighting parameter for the study population.

Serological laboratory analysis

The study participants were sent an individually assigned blood collection set—comprising instructions, informed consent declaration, sampling material, including a dried blood spot card, and a return envelope. With due consideration of the required hygiene measures, the blood sample was taken by the participants themselves using a lancet (Owen Mumford). At least two spots of blood were to be dropped on the dried blood spot card, dried for two to four hours, and then sent back to the laboratory in the enclosed return envelope.

The return envelopes were first checked in the laboratory for a correctly completed consent form. Then two disks each were punched out of the dried blood spot card, which had been completely soaked in the blood, measuring 4.7 mm in diameter (Panthera-PuncherTM 9, Perkin Elmer). The specimen was discarded if it was not possible to extract two punch disks soaked with blood.

The punch disks were eluted in 500 µL sample buffer (Sample Buffer Blue, Euroimmun) in microtiter dilution plates (Deep-well plate 1 mL, Euroimmun). The microtiter dilution plates were shaken for 30 seconds at 1000 rpm and then incubated for 1 hour at 37 °C. After centrifuging for 5 minutes at 1207 g at room temperature, DBS holders (Euroimmun) were inserted in the wells. Finally, the samples were analyzed for SARS-CoV-2 antibodies (against spike antigen [S antigen]: EUROIMMUN Anti-SARS-CoV-2 QuantiVac-ELISA [IgG]; against nucleocapsid protein [N antigen]: NEUROIMMUN Anti-SARS-CoV-2 NCP-ELISA [IgG]) using the ELISA technique (EUROLabWorkstation, Euroimmun).

  • Population statistics as of 31 December 2019 based on the latest territorial boundaries of 31 December 2020

  • Education statistics from the 2020 Micro Census

  • Other general population figures (sex × age) taken from updated data, based on Census 2011.

  • Sex

  • Age groups

  • Education

  • Federal state

  • BIK regions (BIK, regional classification system for Germany in 10 categories).

  • Federal state

  • Administrative region

  • BIK type of region

  • Number of people in household

  • Highest educational qualification

  • Age

  • Sex.

  • Federal state

  • BIK type of region

  • Age

  • Sex

  • Highest school leaving qualification

  • Number of pre-existing conditions

  • Number of COVID infections

  • Symptomatic COVID infection

  • Emergency hospital admission

  • Vaccination status

  • Willingness to be vaccinated.


Articles from Deutsches Ärzteblatt International are provided here courtesy of Deutscher Arzte-Verlag GmbH

RESOURCES