By Ente Rood, KIT Epidemiologist
Geographic analysis of COVID-19 surveillance data can help us better understand how the virus spreads and design more effective mitigation measures for future waves of the pandemic.
As COVID-19 spreads around the world, politicians and public health authorities are confronted with a devilish dilemma as they try to balance the pandemic’s public health impact with the economic fallout of their mitigation measures. Given the weight of these choices, many decision-makers have embraced mathematical modelling to predict the possible effect of different interventions on the virus’s progression over time.
But to truly understand how the virus spreads – and appropriately gauge our mitigation efforts – we also need to understand how fast it spreads, by what routes and which populations are most affected. This knowledge will enable policymakers to tailor mitigation measures to local contexts, which could help to relieve the burden on health systems and ease the social and economic pressures faced by communities grappling with the pandemic.
Spatial analysis of COVID-19 complements mathematical models
Using local, national and international health surveillance statistics, conventional mathematical models evaluate the way different mitigation measures – such as social distancing or stay-at-home orders – impact virus transmission. Such models do not explicitly use information about the geographic spread of the disease. Instead, they assume that infection risk does not change between different areas.
This is problematic for two reasons.
First, because COVID-19 is transmitted by droplets in the air and through contact, the risk of transmission is inevitably tied to areas where infected people are located.
And second, spatial epidemiology has taught us that a range of geographic factors – such as socio-economic conditions, demographics, or behavioural patterns – affect the risk of certain populations, and therefore places, falling victim to widespread disease transmission.
This means that transmission risk will never be the same across a country or across populations, and that it will change over time. Spatial modelling can complement the insight generated from mathematical models by enabling us to model how factors such as an individual’s proximity to infection hotspots and geographic variations in socio-economic and environmental conditions affect the spread of COVID-19.
For example, using publicly available data, epidemiologists at KIT’s Centre for Applied Spatial Epidemiology conducted a spatial trend analysis of the daily number of COVID-19 hospitalisations per municipality in the Netherlands (pictured above). These data are routinely collected and made available by the Dutch public health authorities, which allows for real-time monitoring of COVID 19’s spread throughout the country.
The model predicts the geographic direction and speed of the epidemic’s spread between municipalities. It does so by studying the relationship between the time at which the first cases were reported in a certain municipality with the timing and distance of spread into surrounding municipalities.
What can we learn from this visualisation?
At the start of the epidemic in the Netherlands, COVID-19 transmission was most severe in the centre and south of the country. Local clusters were identified in the municipalities of Tilburg, Utrecht, Zaltbommel, Loon-op-Zand and Altena.
In the following weeks, the epidemic leap-frogged throughout the country, establishing new localised clusters far from the original hotspots. These included some of the most densely populated areas, including Den Haag, Rotterdam and Amsterdam.
As more areas became affected, the velocity of the spread increased, and the epidemic started to move into less populated areas in the north-west and east of the country. It is important to note that after three weeks – when the national government decided to enforce a partial lockdown and closing of schools – the epidemic had already spread to almost 80% of the country.
Four weeks after the first cases were reported, most Dutch municipalities were reporting cases and hospitalisation rates continued to rise steadily. From this point on, localised transmission, which could have occurred two weeks earlier due to the virus’s lengthy incubation period, was most likely the main driver of the epidemic.
These analyses highlight how quickly the virus established several, disjointed footholds throughout the country, from which it rapidly diffused to nearby areas. The introduction of the virus – often between non-adjacent municipalities – was likely caused by the movement of people as they commuted to work, visited friends or returned home from abroad.
Preparing for a second wave
As we look to the end of the epidemic’s first wave, and consider moves to prevent a second, it is clear that effective control relies on early identification and isolation of local transmission hotspots. Within this context, it is particularly important that we closely monitor COVID-19 cases in cities and other densely populated and highly connected areas, which appear to play an outsized role in the disease’s spread.
The geographically disjointed pattern which we observed in the Netherlands also makes it clear that human movement and connectivity play a key role. Interventions such as social distancing and lockdown measures should be enforced in areas of high mobility, such as commuter belts, at an early stage to halt the spread. Indeed, many cities – and countries – implemented aggressive lockdowns in an effort to halt the spread of COVID-19, but in most cases, these measures came far too late.
Given the social and economic fallout of these actions, local epidemics should be closely monitored on a case-by-case basis to decide when mitigation measures can be relaxed and to detect a possible resurgence due to an influx of infected people.
By keeping an eye on the disease’s geographic spread, while closely monitoring its development at the community level, we can encourage locally-tailored actions that create space for economies to recover and lessen the pressure on those most vulnerable to COVID-19.
About the author: Ente Rood is an epidemiologist at KIT Royal Tropical Institute and one of the founders of KIT’s Centre for Applied Spatial Epidemiology. Ente’s work focuses on the development of methods and statistical approaches to predict geographic distributions of disease and to compare spatial patterns of public health needs to geographic patterns of health service delivery.