Corona Monitoring Nationwide (RKI-SOEP-2): Seroepidemiological Study on the Spread of SARS-CoV-2 Across Germany

2.1 Study Design

Corona Monitoring Nationwide (RKI-SOEP-2) is a population-based, epidemiological cohort study. Its units of observation are participants in the 2021 annual SOEP wave and their household members aged 14 and older. The SOEP is a longitudinal survey of private households in Germany and all individuals living in them, conducted annually since 1984 (Göbel et al. 2019). The survey covers a wide range of topics, from demographics, income, the labor market, education, and health to general attitudes, concerns, and life satisfaction.

The SOEP consists of different sub-samples, which are either random samples of the total population or subsamples of selected population groups, including families, high-income earners, and refugees and other immigrants. The IAB-SOEP sample is one of the SOEP’s migration samples. It consists of immigrants who entered Germany between 1995 and 2013 as well as second-generation immigrants (Kroh et al. 2015). The IAB-BAMF-SOEP study is comprised of refugees who entered Germany between 2013 and 2020 and applied for asylum (Kühne et al. 2019; Steinhauer et al. 2022), most of whom are from Syria, Iraq, and Afghanistan. The exact sample compositions, participation propensities, and (non-)response rates for the various SOEP samples since it began in 1984 are described in detail in Siegers et al. (2022). Overall, the SOEP had a high repeat participation rate of 85% up to 2020, which fell to just under 70% in 2021. The CAPI-by-Phone survey mode, introduced after the start of the pandemic (the standard is CAPI), and the change of survey institute from Kantar Public to infas were probably the two main factors behind this drop.^[4] The advantage of a panel study like the SOEP is that rich information is available from previous waves to compensate for selective dropouts in statistical analyses of the entire population. Based on the extensive information available in previous waves, the SOEP therefore provides methodologically high-quality weighting factors for each annual survey wave (for more details, see Siegers et al. 2022). In addition, SOEP counters panel attrition through regular, large-scale refreshers (the next one in 2022).

In the Corona Monitoring Nationwide (RKI-SOEP-2) study, individuals were invited who:

1. lived in a SOEP household that participated in the 2021 annual SOEP survey

2. were 14 years or older (at the time of the survey)

3. gave their written consent to participate in the study

4. were able to take a capillary blood sample on their own (self-sampling)

5. were able to read the study instructions and information in German, Arabic, Farsi, English, Polish, Bulgarian, or Romanian.

2.2 Study Process

2.2.2 Data Collection

The study consisted of a questionnaire part and a blood test part. In the questionnaire part, respondents completed a survey asking for information about previous infections, vaccination against the virus, attitudes towards vaccination, and health issues. The questionnaire could either be completed in written form or using the computer-assisted web interviewing (CAWI) survey technique. In the blood test part, respondents collected a capillary blood sample by pricking a fingertip and allowing the blood to dry on a collection card. We used the CE-certified sample collection and submission kits produced by the manufacturer Euroimmun AG (Lübeck, Germany) (Figure 1).

Figure 1:

Self-collection of capillary blood from a Fingertip. Source: Robert Koch Institute.

The blood collection kit contained detailed illustrated instructions with written explanations under each picture, a blood collection card, a compress, two adhesive bandages, two alcohol swabs, two sterile lancets, and a sealable plastic bag with desiccant. There were five circles on the blood collection card as markers for the blood samples. The illustrated instructions included with the sample collection kit also provided a web link and QR code leading to videos showing how sample collection was to be done. In addition, respondents were directed to a study-specific webpage with frequently asked questions (FAQ) and answers. Respondents were cautioned not to collect a blood sample if there were any issues or acute health conditions that might skew the results or make it painful to collect a sample.

Respondents were asked to send their blood collection cards by mail to the RKI, if possible on the day of sample collection, and to send the completed questionnaire and signed consent form to infas. The materials that respondents returned to the RKI and infas (blood collection cards, questionnaires, and consent forms) had a machine-readable bar code printed on them to enable linkage of laboratory and survey results at a later point in time.

2.3 Laboratory Analytics

The analysis of the dried blood samples was carried out in the Central Epidemiological Laboratory at the RKI. Dried blood spots were punched out^[7] and extracted from the filter paper of the blood collection card according to the manufacturer’s instructions. Analyses were performed automatically on a high-throughput analyzer “EUROLab Workstation ELISA” from the manufacturer Euroimmun. Blood samples were tested for immunoglobulin G (IgG) antibodies against two different antigenic structures of SARS-CoV-2 using two commercial enzyme-linked immunosorbent assays (ELISA) from Euroimmun: The “Anti-SARS-CoV-2-NCP ELISA” was used to test for the presence of antibodies against the nucleocapsid protein (NCP) (manufacturer’s specifications from serum samples, dated November 9, 2021: sensitivity 94.6%: > 10 days after symptom onset or direct pathogen detection by PCR analysis; specificity 99.8%). Measurement results were semi-quantitative, given as ratios, and divided into the three categories “positive” (≥ 1.1 ratio), “borderline” (≥ 0.8 to < 1.1 ratio) or “negative” (< 0.8 ratio) according to the manufacturer’s specifications for serum samples.

The “Anti-SARS-CoV-2-QuantiVac ELISA” was used for the quantitative detection of antibodies against the S1 domain of the virus’s spike protein. This test was used for the indirect detection of either a previous infection with SARS-CoV-2 or a vaccination with an S1/RBD-based vaccine (manufacturer’s specification from serum samples, dated November 11, 2021: sensitivity 90.3%: > 10 days after symptom onset or positive direct pathogen detection; specificity 99.8%). Measurement results were given in relative units (RU/mL) and classified according to the manufacturer’s specifications for serum samples into the three categories: “positive” (≥ 11 RU/mL), “borderline” (≥ 25.6 to < 35.2 RU/mL), or “negative” (< 25.6 RU/mL). For better comparability with the results of test systems from other manufacturers, the results were converted into “Binding Antibody Units per mL” (BAU/mL) according to the First WHO International Standard for anti-SARS-CoV-2 immunoglobulin, NIBSC code 20/136. For this purpose, the relative units were multiplied by 3.2 according to the manufacturer’s instructions. Blood samples that tested above the upper detection limit of the test (120 RE/mL or 384 BAU/mL) were diluted tenfold and quantified with the “Anti-SARS-CoV-2-QuantiVac-ELISA 1010.”

For both the “Anti-SARS-CoV-2-NCP ELISA” and the “Anti-SARS-CoV-2-QuantiVac ELISA,” the manufacturer reports very high levels of agreement between results from capillary dry blood samples and venous serum blood samples: 99.5%; (positive percent agreement: 98.4%; negative percent agreement: 100%; n = 215), and 100% (positive percent agreement: 100%; negative percent agreement: 100%; n = 26 and n = 85 in a second study), respectively.^[8] Borderline results were not considered for either test.

2.4 Questionnaire Content

In the questionnaire part of the study, respondents completed a 12-page questionnaire. Table 1 summarizes the topics covered:

Wherever possible, survey items were designed such that comparability with other studies and commonly used items, including those from regular SOEP waves, was ensured. Study participants whose blood sample was not valid were asked to repeat their blood test. In this case, they also received another short two-page questionnaire so that the most time-sensitive survey information was current at the date of their newly taken blood sample. The survey items in this short version covered the topics of SARS-CoV-2 infection and vaccination.

2.5 Data Protection and Ethics

Similar to other studies conducted by RKI, IAB, BAMF, and SOEP, the Corona Monitoring Nationwide (RKI-SOEP-2) study was subject to strict data protection provisions set out in the EU General Data Protection Regulation (GDPR) and the Federal Data Protection Act (BDSG). The ethics committee of the Berlin Chamber of Physicians assessed the ethics of Corona Monitoring Nationwide (RKI-SOEP-2) study and provided approval (Eth-33/20). Participation in the study was voluntary. The participants were informed about the aims and contents of the study, how their data would be handled, and when the data would be deleted. Respondents provided written informed consent.

3.1 Gross Sample

The gross sample of the Corona Monitoring Nationwide (RKI-SOEP-2) study consisted of 12,101 households (containing 21,456 target respondents) in which at least one household member had completed a household or personal questionnaire in the regular 2021 wave of the main SOEP study (SOEP-Core). Of the 12,101 invitation letters sent out for RKI-SOEP-2, 11,785 reached households in the gross sample.^[9] Within these households that were invited, a total of 20,774 persons aged 14 and older were asked to participate in the study. Of these, 1737 were under 18 years of age. The maximum age of a person invited to participate was 102. Of those invited, 12,407 individuals were assigned to the first tranche, 1984 to the second tranche, and the remaining 6383 to the third tranche. 1950 persons (including 217 minors) had a refugee background and 4203 were born abroad (with 280 minors). 51.4% of all individuals in the gross sample were female and 48.6% male, and 23.0% lived in East Germany (including Berlin). Overall, the (invited) sample covers 400 counties and county-level cities (NUTS-3), ranging from NUTS-3 areas with 1–944 target respondents. On average, a NUTS-3 area contains 52 target persons with a standard deviation of 67 persons. A household in the RKI-SOEP-2 gross sample includes between one and nine individuals, with an average of 1.8 individuals aged 14 and older and a standard deviation of 0.9.

The RKI-SOEP-2 sample originated from a complex study design aimed at sampling households within many regional units distributed across Germany and at oversampling certain population groups (e.g., immigrants, refugees, families with many children). The willingness of these households to participate in subsequent SOEP waves, and thus to be included in the RKI-SOEP-2 sample, varied across population groups. For example, people with an immigrant background and those in full-time employment have been shown to have a lower willingness to participate in subsequent waves than people who do not have an immigrant background or are not in full-time employment (Siegers et al. 2022). This study therefore has a higher sampling error than a pure random sample in which households or individuals are drawn from a nationwide list and all participate. Adjustment factors (sampling weights) are an appropriate means of compensating for the resulting deviations in population statistics. The design effect quantifies the increase in sampling error caused by doing so (Kish 1965). For the SOEP weights, the estimated design effect is between 3 and 3.5. This means that the gross sample size of 20,774 persons corresponds to an effective sample size of 5935–6925 persons. That is, in terms of sampling error, this is the sample size we would have gotten from a pure random sample without special design features such as oversampling or attrition and non-response over the course of the panel.

3.2 Response Rates and Weighted Sample Statistics

A total of 11,162 persons aged 14 years or older participated in the Corona Monitoring Nationwide (RKI-SOEP-2) study. Using the standards of the American Association for Public Opinion Research (AAPOR 2016), the calculated response rate was 53.7% for the total sample and 51.9% for the 14–69 age group. Table 2 shows the detailed response rates and characteristics of the RKI-SOEP-2 net sample. The response rate of female participants was generally slightly higher (56.3%) than that of male participants (51.1%). With the exception of those over 80 years of age, the response rate increased with age from 41.0% in the youngest age group (14–17 years) to 68.9% in the 65–79 age group. Using the CASMIN (Comparative Analyses of Social Mobility in Industrial Nations) educational classification, the highest response rate was observed among highly educated respondents (68.5%) followed by the middle education group (59.1%) and the low education group (44.4%). Willingness to participate was greater in households with two to four persons (55.5%) than in one-person households (50.8%) or in households with more than five persons (44.3%). Furthermore, response rates differed by migration status. The response rate among German-born respondents was significantly higher (60.5%) than among foreign-born respondents without a refugee background (37.3%) and also significantly higher than among foreign-born respondents with a refugee background (17.1%). Response rates varied by federal state between 45.5% (Saarland) and 60.9% (Saxony).

In the weighted sample (see Section 3.3, Weighting), 50.8% of the respondents were female and 49.2% were male. About quarter (25.5%) of respondents were between 50 and 64 years old, and a slightly lower percentage were in the 18–34 age group (22.2%) and in the 35–49 age group (20.5%). Significantly fewer respondents were in the youngest (14–17) and oldest (80 years and older) age groups. The majority belonged to the middle education group (33.9%); about a quarter to the high education group (26.2%); and slightly fewer to the low education group (20.8%). Around 63% of respondents lived in a household with two to four persons and over one third (35.7%) in one-person households. Only a very small minority of respondents (2.4%) were foreign-born with a refugee background. The vast majority (84.4%) was German-born, and 13.2% were foreign-born without a refugee background. The regional coverage of the sample largely corresponded to the expected distribution across the different states. Of the 11,162 respondents, more than three quarters (78.5%) took part in the study in the period November to December 2021.

3.3 Weighting

For the Corona Monitoring Nationwide (RKI-SOEP-2) study, appropriate weights were created. The weighting strategy considered the different selection processes (contactability and participation) from the gross sample to the net sample of respondents at both the household and the individual level separately. For each selection process at each level (e.g., the process of determining who qualified for selection), a model was estimated to derive the corresponding probabilities used for the adjustment of weights. The variables that may enter these models covered topics such as socio-demographics, household composition, regional characteristics, health status, worries and attitudes regarding political and societal issues, and others. Variables best describing the selection processes were identified using lasso regression (Tibshirani 1996). On the household as well as the individual level, weights were adjusted to conform to known population distributions provided by the Federal Statistical Office. On the household level, these included household typology, size, home ownership, and federal state, and on the individual level, they included age, sex, and citizenship (German vs. non-German). A more detailed version of the weighting strategy is provided for the Corona Monitoring Nationwide (RKI-SOEP-2) study by Danne et al. (2022).

4.1 Research Potential and Strengths of the Study

The second wave of the Corona Monitoring Nationwide (RKI-SOEP-2) study provides data on topics relevant to public health after two years of the SARS-CoV-2 pandemic. The main aim of the study was to estimate the percentage of the population that had been either infected by SARS-CoV-2 or vaccinated by February 2022, differentiated by social groups (e.g., age, sex, educational level). The study data thus allow for identification of groups at risk of SARS-CoV-2 infection. The prevalence of vaccination, attitudes towards vaccination, health information-seeking behavior, and conditions reported that influenced decision-making will also provide a better understanding of the epidemiological status quo. Appraising the expected percentage of long COVID cases and analyzing the medium- to long-term effects of the pandemic on health conditions, health behavior, and individual socio-economic situations will ultimately make it possible to identify areas and groups requiring special policy measures in the future.

Because of the extensive measures undertaken to include SOEP respondents from across Germany as well as a wide variety of socio-economic living conditions, the study data provide the basis for differentiated analysis and identification of specific subgroups that are more severely affected by the pandemic. The fact that all of the information from this study can be linked to respondent data from earlier or later waves of the SOEP offers extensive potential for analysis of the medium- to long-term consequences of the COVID-19 pandemic.

In sum, the results of this study will provide important information to government bodies and other stakeholders that will be useful in evaluating the management of the pandemic thus far and in shaping future responses to this and other pandemics.

4.2 Limitations

Besides its strengths, the Corona Monitoring Nationwide (RKI-SOEP-2) study also has some limitations. First, individuals were only invited to participate if they had taken part in the regular 2021 wave of the SOEP-Core survey. Since fieldwork was still ongoing for the SOEP and particularly for the IAB-BAMF-SOEP Refugee Study during the fieldwork for this study, only a limited number of refugees could be included. This might to some extent result in selection bias, particularly for participants from refugee subsample. Second, response rates among immigrants and refugees were lower than among non-immigrants. These lower response rates were probably driven regular SOEP respondents’ lack of experience with written surveys that they have to return by mail: The regular waves of the SOEP are mainly carried out as face-to-face interviews. Another limitation results from the fact that panel studies are usually subject to healthy volunteer bias, meaning that people who are suffering from a severe disease are less likely to participate (e.g., Schnell and Heller 2000; Zheng et al. 2020), potentially leading to an underestimation of the severely diseased population. Although separate weighting was conducted for this study, taking into account a wide variety of selection processes, it is possible that some amount of bias due to unobserved variables such as aversion to the study topic may not have been fully corrected. Moreover, data from the current 2021 SOEP wave were not yet available for the weighting process.

Limitations are also to be expected on the laboratory side. An inherent limitation of IgG antibody studies as a basis for prevalence estimates of SARS-CoV-2 immunity is that the presence of IgG antibodies does not mean full immunity, and a lack of IgG antibodies does not mean a complete lack of protection. However, serological studies may provide an estimate of the proportion of the population that has been in contact with the virus antigen(s) and may have developed at least some degree of immunity or be less likely to become severely diseased after infection. In this regard, a few factors need consideration. One of these is that study participation may depend on infection and vaccination status. This is addressed through careful non-response analyses and appropriate weighting algorithms.

Antibody development (seroconversion) depends on the severity of the disease or the type of vaccine, on age, and on preexisting health conditions (e.g., Wei et al. 2021). Antibody tests and questionnaire data do not have perfect sensitivity and specificity (Scheinlauer et al. 2022), and there is some antibody decay over time (Achiron et al. 2021). This is also of relevance to this study, because the fieldwork period lasted for about four months, and during this time, the infection process across Germany was highly dynamic. This impairs accurate determination of seroprevalence. Furthermore, the previously used estimation methods (e.g., Hoebel et al. 2021) may require further adjustment since they apply to pre-Omicron variants. Data on the kinetics of antibody response after SARS-CoV-2 Omicron infection is still scarce. First results of neutralization profiles against the Omicron BA.1 variant in comparison to former SARS-CoV-2 variants indicate different antibody profiles/kinetics depending on preexisting SARS-CoV-2 (vaccine- or infection-induced) immunity after infection with Omicron (Rössler et al. 2022).