Is testing at home sufficient? Self tests and official statistics

The question of how many new COVID-19 cases per week continues to navigate our political and personal decisions. How we assess the risk of infection, the methods we use to protect ourselves, and how we handle the pandemic in the future all depend on the answer to this question.

For several weeks now, Germany has been discussing a potential underreporting of its COVID-19 incidence rate. While more and more testing centers are closing due to lower summer demand, wastewater sample analysis and respiratory infection monitoring systems indicate increasing case rates. According to the current case definition, the reported 7-day incidence is based exclusively on PCR test results, while rapid antigen tests, whether performed professionally or at home, are not considered. Compared to rapid tests, PCR tests are often more accurate and ensure that local outbreaks can be detected correctly with a high degree of certainty. If the PCR test volume decreases, relative trends will still be visible, but the magnitude of new infections will be much more difficult to estimate.

Various research approaches are attempting to describe this degree of underreporting more precisely. Using questionnaire data from 28,274 data donors, in which we recorded test results and test types (e.g. PCR test, rapid antigen test), we aim to better estimate this situation by adding another piece to the puzzle.

What does our data show?

Our analysis consists of two steps: first, we compare how the official 7-day incidence differs from the “data donation” incidence based on your reported test results. Next, we examine PCR test rates over time to determine if a decline of this number is evident in our population and assess how it affects infection dynamics.

To get our analysis going, we first need to calculate the data donation incidence. For each day, we divide the number of self-reported positive tests via our App by the number of questionnaires submitted that day. From this, we calculate the 7-day average and multiply the result by 100,000. This enables us to estimate the number of cases per 100,000 donors and approximately compare our values to official statistics.

We observe that the two incidences evolve quite similarly until the end of the first quarter in 2022, although their magnitudes do in fact differ. Some of the variation can be attributed to the fact that our data donor population is not necessarily representative of the general German population (more on this later).

Approaching June, however, the trajectories decouple from their former patterns. We hypothesize that this trend may be related to a decline in PCR testing. To investigate this further, we next looked at the ratio of PCR and rapid tests in our data set and compared it to the ratio of the two incidence curves shown above.

The blue line shows by how many times the data donation incidence exceeds the official PCR incidence – we call this the “incidence factor”. In the first quarter, we see that our calculated incidence reflects, on average, about three times the official data. This value remains nearly constant throughout the spring, and may be due to higher than average test rates in our sample (thus resulting in more infections detected) or simply because our data donors are not representative of the German population as a whole.

From June onwards, however, when we observe a change in the previously stable relationship as the two incidences increasingly decouple, the ratio between them grows larger. This means that the donated test data estimates a significantly higher incidence than official statistics. As a result, the data donation incidence factor increases. What is the reason for this?

One explanation could be the assumed decline in PCR testing. If fewer and fewer people confirm their rapid test result by PCR method, these infections are missing from the official statistics. To further assess this assumption, we calculated the ratio of rapid tests to PCR tests for each day (”rapid test ratio"), shown as the green line in the figure above. A value of 0.5 means that half as many rapid antigen tests as PCR tests were reported by our donors on a given day. If the value in contrast is 1, then exactly the same number of rapid tests and PCR tests were recorded. Values above 1 indicate more rapid tests than PCR tests.

We can see that both curves (incidence factor and rapid test ratio) show a similar trend. In the first quarter, the rapid test ratio remains at a similar level except for minor fluctuations. However, from the end of May onwards, we observe a strong growth in both curves — when more infections are reported with rapid tests instead of PCR tests, the data donation incidence differs more and more from the official statistics. This observation allows us to hypothesize a connection between rising incidences and lower PCR test rates.

The figure below confirms that there is indeed a significant decrease in the share of PCR tests in our donor cohort. While at the beginning of the year around 70-80 % of all reported infections were confirmed by PCR test, at the present time this figure amounts to only around 50-60 %.

How can we interpret these results?

Because our donor population is not fully representative of the population of Germany, the groups cannot be directly compared. Differences exist, for example, with regard to age and gender, and presumably also in level of education and socioeconomic status. Therefore, the COVID-19 incidence within the group of data donors likely differs from that in the population as a whole.

Nevertheless, we note that PCR testing in our cohort has declined by about 30 percent so far this year. This may create a blind spot in the estimation of the official incidence, which may suggest that case numbers might be larger than currently reported.

What is important now?

Even if the exact factor behind the number of unreported remains unclear, we see indications in our data that the currently reported 7-day incidence underestimates the pandemic’s dynamics. Given the currently prevalent and highly infectious Omicron subtype BA.5, we would like to draw attention to this issue. Policy makers, society and science need reliable statistics to derive well-informed measures. Therefore, we invite that if you test positive for COVID-19, please also have your infection confirmed with a PCR test.

Any person who can provide a positive rapid test has a right to a PCR test under current testing regulations in Germany, even if the test was carried out at home.

PCR testing also helps to clearly prove the disease. By confirming your infection with a PCR test, you are actively helping to properly estimate an important epidemiological indicator while at the same time caring for your individual future health.