In line with our research aims, we used reported COVID-19 case data at the neighborhood level for the Neukölln district in Berlin. This case study represented a good candidate for such a detailed analysis because of the high temporal and spatial dynamics in the reported COVID-19 cases and the very diverse characteristics of the district in terms of population and urban structure.

Neukölln, one of the 12 districts in Berlin, is located in the southern part of the capital city of Germany (Figure 1). With a population of 327,100 inhabitants in an area of 44.9 km², Neukölln was, in 2021, the third most densely populated district in Berlin (7,284 persons/km²) and, in terms of population, the 20th largest city in Germany (43). The district exhibits distinct spatial differences in regard to built-up structures and environmental features as well as the socioeconomic and demographic characteristics of the population (44) (Figures 1A–D). The greater part of the north has a high population density, with residential buildings primarily consisting of dense housing structures (e.g., multi-story block developments and social housing estates; Figures 1A,B). Several areas in the north register low to very low on the index of socioeconomically disadvantaged areas (Figure 1C). The district has a rather young population with a low proportion of elderly people (Figure 1D). In contrast, large parts of the southern district are dominated by more dispersed housing structures with primarily single-family or semi-detached houses, resulting in a comparatively low population density. In the center of the southern district, however, there is a sub-district that differs from its surroundings, having denser housing structures and a higher population density. Most areas of the south register medium on the socioeconomic status index, whereas the area in the southern center registers low to very low on the index. The population is considerably older in the south than in the north of the district, having many areas where those over 65 years make up nearly one-third of the population. As a unit of analysis, we used the 46 planning units of the so-called small Lebenswelt-oriented areas (‘lebensweltlich orientierte Räume’), which were created by the city of Berlin as relatively homogeneous spatial entities within a district as a basis for the planning and prediction of demographic and socioeconomic development (45).

For the investigated time period from the first positive cases in early March until the end of the fourth wave (late December 2021), we observed a highly dynamic development of the overall and age group-specific incidence for the Neukölln district (Figure 2). We identified a large number of measures that were implemented during this time. Figure 2 depicts only a subset of implemented measures and influencing factors (for detailed descriptions see Supplementary Table 2). Further details on Berlin-wide COVID-19 regulations and the measure-index are presented in Supplementary Figure 1.

The first cases in Neukölln were recorded in early March 2020, 1 month after the virus was reported in Germany. From then on, incidence increased in Neukölln and reached its first peak in mid-March 2020. During this time, far-reaching measures for contact reduction were put in place (see Figure 2). These measures included the closure of schools, childcare centers, businesses, institutions, and services, plus additional restrictions on public events and private gatherings. Shortly thereafter, a lockdown was declared to reduce contact, which added further restrictions. The measures succeeded in decreasing incidence in Neukölln in the following weeks. At the end of April 2020, the measures were eased for the first time, allowing schools, childcare centers, and businesses to open again under specific hygiene conditions. Incidence remained mainly low in Neukölln during the summer months, with one exception in mid-June 2020 when incidence rose to a sharp peak owing to a large outbreak of cases in a residential complex in the north of the district. Measures within the residential complex, such as large-scale testing and strict quarantine rules, were implemented by the Neukölln department of health to prevent any further spread of the illness.

The second wave started at the end of September 2020, with a steep rise in incidence until a temporary peak in mid-November 2020. Around that time, a partial lockdown was introduced to control the spread of the virus. Businesses, institutions, and services had to close, and restrictions were imposed on public events and private gatherings. However, the measures did not achieve the desired effect, leading to an extended lockdown in mid-December 2020, when schools and childcare centers were closed, and enhanced restrictions were imposed on private gatherings. Although a decrease in incidence was already evident in Neukölln after the introduction of the partial lockdown, the extended lockdown ultimately contributed to a more significant easing of the situation. Incidence started to decrease, reaching a temporary low during the Christmas break in 2020, followed by a short period of increase during the first weeks of January 2021 and a further fall, with substantially fewer case numbers in mid-February 2021. The vaccination campaign, which focused initially on people over 80 as well as people working in medical facilities and nursing staff, also started during the second wave.

The third pandemic wave followed directly on from the second wave without any relaxation of measures. The more contagious virus variant Alpha (B.1.1.7) became the dominant virus variant in Berlin at the end of February 2021, resulting in a large increase in case numbers until the end of March 2021. In March, rapid antigen tests became available, which led to the opening of many centers where persons could get tested once a week free of charge. During this time, test-related access requirements for businesses and services were also introduced for the first time and then gradually extended over the following weeks. The Easter holidays at the beginning of April 2021 led to a temporary drop in incidence, followed, however, by a rise that lasted until mid-April. Rising case numbers could not be controlled by the measures already in place, which resulted in an expansion of measures at the end of April 2021 (“Bundesnotbremse,” literally “federal emergency brake”). The measures were connected for the first time with the numbers of regional cases. At the time the new regulation took effect, for Berlin as a state, the measures included the reduction of contact in private and business-related settings and the closure of schools and childcare centers. The result was a fall to a low level of cases by the end of June 2021. The data show a temporary peak in early June 2021, which may be attributed to a change of the software used by the Neukölln department of health to record cases. Many cases had to be updated in the new software, resulting in a brief time lag of new cases and thus leading to an apparent temporary increase in numbers. The decrease in incidence continued in Neukölln after the short peak in June 2021. Already by early June, however, interventions were partly revoked for Berlin, which permitted the opening of schools and fewer restrictions on private gatherings.

However, the trend of low incidence did not last long during the summer months of 2021, and it began to rise rapidly again in July, owing to the new Delta (B.1.617.2) variant, which had become the most dominant virus variant in Berlin by then. This new and very contagious virus variant led to an unprecedented surge in cases in a short period of time. In Neukölln, a small, wave-like trend was visible from August until October. During that time, new regulations for access restrictions were implemented, meaning that persons needed to be vaccinated, recovered, or tested negative in order to access businesses, institutions, and services. Shortly thereafter, this regulation was tightened, and access was allowed only to vaccinated and recovered persons. Nevertheless, from mid-October 2021 onward, incidence again increased strongly, reaching a peak at the end of November. From the beginning of December, a sharp drop occurred, with another small peak 2 weeks before Christmas 2021. Although there has been an unprecedented increase in cases, only a few of the social and physical distancing measures were implemented during the fourth pandemic wave.

The first wave was characterized by high incidence among the working population (18–44 and 45–65 years), whereas the younger population (< 18 years) exhibited a very low incidence (Table 1). Incidence across all age groups decreased similarly after the implementation of the first set of measures (Figure 2). With the first easing of measures at the end of April 2020, the elderly (>65 years) were the most affected age group for a short time. During the summer months, incidence among the elderly decreased again and remained at a low level. In contrast, the young population, who had hitherto been the least affected, showed two short-term increases, one related to the outbreak of cases in a residential complex and the other occurring shortly after the summer holidays.

At the beginning of the second wave, incidence increased across all age groups, the most rapid increment and highest overall incidence being among the age group 18–44. Incidence among the elderly and the young was relatively low compared to the other age groups until mid-November 2020. However, after the implementation of measures in late November, the rate among the elderly increased, while it decreased steadily among the young, such that the elderly were much more affected than the young by the end of the year. Toward the end of the second wave, incidence among all age groups was back at a similar level.

The beginning of the third wave was characterized by a marked increase in cases among the young and working population. In contrast, incidence among the elderly remained low. A temporary decrease was observed in all age groups over the Easter holidays, the most significant decline being among the young. However, the rate increased again shortly thereafter across all age groups. Compared to the previous waves, the younger population was by then much more affected than the elderly. Again, the implementation of extensive control measures in April 2021 led to a decline in cases across all age groups. During the summer plateau of 2021, the incidence was similar across age groups, whereas, from early July 2021 onward, the incidence in the 18–44 age group again rose sharply.

With the emergence of the Delta (B.1.617.2) variant, the age group-specific incidence underwent a drastic change. Incidence among the young population rose sharply, remaining highest among all age groups over the entire fourth wave, although the pattern was similar among adults aged 18–44, whereas the incidence among those aged 45–64 and the elderly was comparatively low until the end of 2021.

The spatial distribution showed that cumulative incidence rates were heterogeneously spread over the district during the first wave (Figure 3A). Areas with high incidence were mainly in the southwestern and the northern part of the district, the lowest incidence being in the southeast. Only one significant high-risk cluster was detected for the first wave (Figure 4A). During the summer plateau of 2020, higher incidence was concentrated in the north, resulting in two high-risk clusters with a high relative risk value. In contrast, larger parts of the southeast and southwest exhibited comparably low incidence and low-risk clusters (Figures 3B,4B). The second wave was again characterized by a more heterogeneous distribution of cases. High incidence occurred partly in the northern and south-central parts of Neukölln, while planning units with low incidence were distributed variably across the whole district (Figure 3C). Clusters of high and low relative risk were both found in the northern and southern parts (Figure 4C). Planning units with high incidence reappeared in the third wave increasingly in the north, whereas areas with low incidence were observed more often in the south (Figure 3D). High-risk clusters were observed in the northeast and northwest, while large areas of the south were detected as low-risk clusters (Figure 4D). The summer plateau of 2021 and the fourth wave showed similar spatial patterns compared to the third wave, with high incidence and high-risk clusters more prevalent in the north and lower incidence and low-risk clusters more predominantly in the south (Figures 3E,F, 4E,F).

More details of the results can be explored via the web-based interactive visualization tool.

The pandemic phases exhibited great differences in terms of the total number of reported cases as well as in the dynamics of the rise and fall of cases. Each phase was characterized by a set of influencing factors that changed and interacted with one another constantly. For example, the two new virus variants, Alpha (B.1.1.7) and Delta (B.1.617.2), drastically changed incidence patterns during the third and fourth waves. Both variants were more transmissible (59, 60), which likely led to strong incidence increases at the beginning of the third and throughout the fourth wave in Neukölln. Interestingly, although the number of cases rose to unprecedented levels in the fourth wave, only a few social and physical distancing measures were in place at that time. One possible explanation might be the greater immunity among the population (either via previous infection or vaccination), which most often prevented severe illness (61, 62) and therefore reduced the need for stringent control measures.

Previous studies have also shown that various measures like social and physical distancing, mask-wearing, or testing strategies have been effective in reducing case numbers (2, 3, 52). However, the quantification of their direct impact on COVID-19 incidence is not always feasible, since interventions often came in packages and were influenced by a number of other factors (63). In our study, we are also hesitant to derive general statements. It may be observed that the far-reaching measures during the first, second, and third waves significantly decreased incidence in Neukölln, as also in Germany as a whole (14, 49). However, further analyses on neighborhood level, using, for example, spatiotemporal pandemic simulation models (64–67) could provide more insights to better understand the influence of policy interventions on case numbers.

Moreover, behavioral factors, e.g., the individual adaptations to daily routines to reduce potential exposure to the virus, might have influenced case transmissions as shown in previous studies (8, 68). Due to the absence of any individual behavioral data, we cannot derive any initial hypothesis on the impact of behavior on case numbers in this study.

It is relevant to note that the number of tests performed most likely influenced the case numbers and thereby the incidence. For example, the number of PCR tests doubled between early August and early November 2020 in Berlin (from 40,000 to 80,000 PCR tests per week) (69), which could also have contributed to the strong increases in the number of recorded cases at the beginning of the second wave. Conversely, test statistics revealed a substantial decrease in testing over vacation periods such as Christmas or Easter, which may also be reflected in the short-term decreases in cases over these periods. The impact of testing on the occurrence of cases has also been shown in other regions and in other pandemic phases (14, 70). Still, frequent and large-scale rapid testing was evidently an effective means of decreasing case numbers in Germany during the third pandemic wave (71). The introduction of rapid antigen tests has made it possible to identify cases that would otherwise remain undetected and to isolate infected individuals in time before the infection was passed on to others.

The exploration of cases also showed that not only the numerous measures in Berlin as a whole (Supplementary Figure 1) but also local events and measures influenced case development. This can be especially observed with overall low incidences. In Neukölln, the influence of local events was illustrated by the major outbreak in a residential complex during the summer plateau of 2020. A sharp increase in incidence occurred, which quickly leveled off again after quarantine and testing measures were implemented locally. This demonstrates how important it is to understand local events and interventions when interpreting changes in incidence in a study area.

Finally, the results showed that seasonality may have an effect in slowing the spread of infections. Incidence was substantially lower in the summer compared to the winter months in Neukölln, which is in line with previous research that found that infectiousness increased during winter, owing to more frequent indoor gatherings and weakened immune systems (10, 11, 71).

Our findings show that single measures focused specifically on the reduction of transmissions within certain age groups. Measures relating to schools and childcare centers were mainly implemented to reduce incidence among the young (<18 years). For Neukölln, school and childcare center closures during the first and second waves were consistent with an incidence reduction among young people. Similarly, however, the number of cases decreased among the young during school holidays (e.g., summer holidays 2020 and 2021, autumn holidays 2021). Testing in schools may also have influenced the number of cases in this population. For example, schoolchildren had to test themselves several times a week after the summer holidays in 2021, which also likely contributed to an increase in the incidence of recorded cases among the young at the beginning of August 2021.

In contrast, the vaccination campaign was initially a measure specifically targeted at the elderly. During the third wave, when the elderly were prioritized for vaccination, the number of cases among the elderly was, for the first time, significantly lower than in other age groups. This trend continued in the fourth wave, with the lowest incidence being found among the elderly. The high rates of vaccination of the elderly in Berlin during the third and fourth waves (72) are likely to have prevented numerous cases among this group. Still, it has been shown for Germany that, especially in the 6 months after the start of the vaccination campaign in December 2021, the effects of vaccinations on the total number of infections were probably minimal, as the vaccination rates of the whole population were rather low compared to other countries (71, 73).

Further, the results highlight that, even when certain measures target all age groups similarly, they still can lead to different incidence between age groups. For example, after the partial lockdown in mid-November 2020 was imposed, incidence among the young and working population decreased, while it increased among the elderly. The same development was also identified in other regions in Germany, where incidence among the elderly revealed a different infection dynamic than other age groups after the implementation of measures in mid-November (14). It was hypothesized, based on the findings of Thurner et al. (74) that infection dynamics among the elderly might be characterized by a linear or diffuse growth dynamic, owing to lower connectivity in their social networks. In contrast, the young and working populations may show higher connectivity in their networks, resulting in an exponential growth dynamic, where in particular super-spreading events may play a major role. Strict measures for contact reduction mainly prevent exponential growth but have only little effect on networks with low connectivity (75, 76). This observation may also apply to Neukölln. On the one hand, it may be observed that far-reaching measures had a major impact on the young and working population by preventing further exponential growth. On the other hand, the measures showed only minor effects on incidence among the elderly, which could lead to the hypothesis that, in Neukölln, too, incidence among this group tends to exhibit a linear or diffuse growth dynamic.

The results of our spatial analysis highlight that the distribution of COVID-19 incidence is most often heterogeneous, with changing spatial patterns across the different pandemic phases. Other intra-urban studies have shown similarly heterogeneous patterns, with spatial distributions also changing over the course of different phases (25, 26, 35). A number of studies have tried to understand how these spatial patterns emerge on a small scale by analyzing the relationship between COVID-19 cases and socioeconomic and demographic characteristics or the degree to which an area is built up. These studies indicated that socioeconomically disadvantaged areas were more susceptible to the spread of COVID-19 on a small scale. COVID-19 cases were, for example, negatively related to income (25, 26, 77) or indices based on socioeconomic data (78–82) and positively related to population density, average household size (77, 83, 84), and the proportion of young people in the population (26, 77). These analyses are still pending for Neukölln and no general statements can be made on the basis of our initial exploration. However, our results allow us to draw initial hypotheses about the association between the abovementioned characteristics of the planning units and COVID-19 cases. The detection of clusters during the second, third, and fourth waves reveals that high-risk areas were located in neighborhoods with a low to very low socioeconomic status index, while low-risk areas were mainly characterized by a medium socioeconomic status index. Further, high incidence was mostly found in areas with dense housing and a high population density. On the contrary, the less affected areas were mainly situated in areas with dispersed housing and a lower population density. With regard to the age distribution in the district, the results showed that low-risk clusters were concentrated in areas with a high percentage of elderly people. These initial visual comparisons may indicate an association between factors such as socioeconomic status, residential density, or age and COVID-19 incidence for Neukölln. However, such exploratory results require examination in further in-depth modeling analyses as for example conducted (77, 85).

Initial explanations may also be given for the changes in the patterns over time. We observed that in Neukölln, the first wave differed in its spatial patterns compared to the consecutive phases. Similar patterns were also found at the national level (86, 87) and in other intra-urban settings (26). One possible explanation might relate to the varying effects of the pandemic on different types of employment. During the first wave, higher incidence might be attributed to employed people with higher mobility (e.g., business travel, transregional commuting, or holidays), whereas high incidence during the consecutive waves could have occurred mainly among people in less advantaged occupations, who were not able to work from home (86). In addition, it has been shown that the impact of far-reaching measures can affect socioeconomic groups differently (88), which also could have resulted in different spatial patterns over time. In Neukölln, for example, we see that, during the second and third wave, when extensive control measures were implemented, socioeconomically disadvantaged areas experienced a higher risk compared to other areas. Still, analyses with a higher temporal resolution are needed to investigate this relationship in more detail.

Our study is subject to some limitations. First, we used reported cases, which have been shown to deviate from the number of true cases, as shown in previous studies for Germany (89, 90). Test capacities and test strategies add a bias to the reported case data. Not only did the test capacity increase over time in Germany, but there were also changes in the testing strategy and availability and accessibility of tests. At the beginning of the pandemic, only symptomatic persons were tested, which likely led to many asymptomatic cases not being detected. Later on, travel returnees or contacts of positive-tested persons were also tested, leading to more actual cases being detected. However, reported cases are still one of the main indicators to trace transmissions in disease surveillance systems, and have been proven to be useful in other studies to explore the relationship between incidence and interventions (3). Most importantly, there are no other comprehensive data available at the neighborhood level (e.g., testing capacities, tests conducted, or severity of the disease).

The second limitation is due to the aggregated analysis of COVID-19 cases at the planning unit level, which influences the results by the different shapes and sizes of the underlying unit. Thus, our analysis is—like other spatial analyses—subject to the modifiable areal unit problem (MAUP), which describes the bias when aggregating spatial point data into larger spatial features (91). However, we believe that the planning units are a suitable unit of analysis because they reveal a high spatial resolution and are, unlike zip codes, quite homogeneous within themselves in terms of built structures and socio-demographic characteristics. This enables us to identify high-risk areas on a small scale and thus to provide support for developing locally specific measures that target vulnerable locations and groups.