Setting the agenda in research

Comment

EDGAR SU/REUTERS
A migrant worker outside a dormitory in Singapore — such high-rise sites have seen some of the city’s worst clusters of COVID-19.

Cities — try to predict superspreading

hotspots for COVID-19
Roland Bouffanais & Sun Sun Lim

T
Tracking how people move wo months ago in Singapore, walkers, Superspreading sites — where transmission
joggers and cyclists in Bishan-Ang Mo rates are particularly high — can seed tens or
around urban areas can Kio Park got a surprise — a robotic dog hundreds of cases in days.
pinpoint where disease might nipping at their heels. It ‘barked’ at For example, last month in Beijing, more
them to stay metres away from others. than 100 cases were tied to one central food
transmit fastest and farthest. No tactic seems too outlandish to deal with the market. Cases have surged in apartment blocks
COVID-19 pandemic, which has confined bil- in Melbourne, Australia, triggering a fresh
lions of people across the world to their homes. lockdown. In May, more than 200 visitors to
Now, leaders face a conundrum. Reopen Seoul’s Itaewon nightclub district became
facilities too slowly and prolong hardship, infected.
or relax restrictions too quickly and ignite Governments need tools to assess where the
a fresh wave of infections. It’s a precarious riskiest spots in the riskiest places — cities —
balancing act. The spread of the SARS-CoV-2 might be. Big-data studies of human mobility
virus can be lightning fast and surreptitious. need to be combined with epidemiological

352 | Nature | Vol 583 | 16 July 2020

©
2
0
2
0
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 Lorem ipsum

models. And the demographic profiles of SCHOOL NETWORKS

people coming into contact at any particular The movements of almost 6,000 schoolchildren across 33 schools tracked over one day
(12 July 2016) in Singapore reveal how the students mingled. They spent most of their time
location need to be included. in close proximity to their classmates. But they also came into contact with many more
people on public transport, in shopping centres and at after-school activities.
Close contact
Student (coloured by school) Interaction
Researchers understand reasonably well that
the disease spreads in places where people
spend a lot of time face-to-face, such as care
Pupils at central
homes, hospitals and restaurants1–3. Duration schools mingle
of contact, physical proximity and environ- more than those
mental conditions are the main factors. A fleet- at the periphery

ing encounter with a passer-by on a path is less

likely to result in infection than sitting inside
next to a person for 20 minutes. Cramped and
poorly ventilated indoor spaces are riskier
than expansive outdoor ones.
But many more factors have a role, and these
are less understood. Some people do not com-
ply with safety measures, for many reasons.
Inebriated diners and excitable theme-park
visitors, for example, are unlikely to stay
metres apart. The degree of mixing differs. The
SOURCE: R. BOUFFANAIS/S. S. LIM

same group of students might sit together in a

classroom every day. But children in secondary
school mingle with more students and teach- Even children
ers than do those in preschools. And school- from distant
children meet widely outside school hours, schools mix

when commuting, playing sports or attending

tutorials (see ‘School networks’).
The details of such everyday interactions
are documented too poorly to model risk fac- skyrocketed to more than 40,000, or more building’ reforms across ventilation, drainage,
tors accurately, as experiences with COVID-19 than 12% of that population, compared with refuse and building maintenance4. Air purifi-
show. Resorts, conferences, religious gath- fewer than 2,600 infections elsewhere in the cation systems were installed in hospitals. A
erings and workplaces have all experienced compact city-state of 5.3 million people. research programme looked into the impacts
notable outbreaks. In late May, the South On 18 June, 700 residents of a tower block of the built environment on health. Studies
Korean city of Bucheon saw hundreds of cases in the German city of Göttingen were quaran- showed that residents were affected by, for
at an e-commerce distribution warehouse. tined, with some 100 cases detected. Some example, the orientation of the building
Meat-packing plants in the United States, residents clashed with police as they tried they lived in, the height of their apartment
Canada, Germany, Spain, Ireland, Brazil and above ground and their immediate outdoor
Australia have seen surges. Some of the first “Very little is known about environment4.
superspreading events took place on cruise
ships — more than 700 people on board the
how the disease spreads in Flows of people
British-registered Diamond Princess tested dynamic spaces.” But people don’t stay put. Cities are bustling
positive in February. places — open systems in which many people
Care homes with many elderly or from many different locations come and go.
immune-compromised residents are especially to flee. Hong Kong saw something similar in An urbanite might cross paths with thousands
vulnerable. Long-term care facilities have seen 2003, at the start of its epidemic of severe of people each day, during long commutes in
almost 60% of the COVID-19 deaths in Washing- acute respiratory syndrome (SARS) — more packed trains and buses and in stations in rush
ton state, 45% in Sweden and almost one-third than 300 people living in block E of Amoy hour5,6. Even car drivers encounter lots of peo-
in the United Kingdom. Health-care workers Gardens became infected, almost overnight. ple, at petrol stations or while running errands.
are, of course, highly exposed, with more than Indeed, many of the same factors that lie Any of these interactions could expose them
92,000 infections and more than 500 deaths behind superspreading of COVID-19 today — to COVID-19. Yet very little is known about how
among US health-care personnel alone. high population densities, poor accommo- the disease spreads in dynamic spaces, such as
Groups living in close proximity are at dation and narrow streets — contributed to airports, stations, bars, restaurants, cinemas
risk. For example, almost 93% of Singapore’s some of the worst pockets of past outbreaks, and aeroplanes.
COVID-19 cases in the first 48 days occurred in including those of cholera, typhoid, tubercu- It’s clear that long-distance travel extends
dormitories for migrant workers (see ‘Hous- losis and SARS (see ‘Cleaner cities’). the reach of an outbreak. Contact tracing
ing risks’). Each block houses thousands of Much stands to be learnt. For example, after tied a nightclub visitor in Seoul’s Itaewon
workers, sometimes 20 per room. Cases there SARS in 2003, Hong Kong adopted ‘healthy district to a staggering 1,300 others — he had

Nature | Vol 583 | 16 July 2020 | 353

©
2
0
2
0
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 Comment
travelled 42 kilometres to get there from the where tracing has not happened at scale. Con-
city of Yongin, and visited other districts and tact-tracing apps and the data they generate
provinces before testing positive. Air travel are potentially valuable, but surveillance
spread COVID-19 quickly to the world’s most
connected metropolises — from Wuhan to
fears, among other concerns, have stymied
their adoption.
Cleaner
Singapore, Hong Kong, London and Milan, or
Modelling flows
cities
through Europe to New York and Boston.
We don’t know how the virus was carried A model of disease spread can be built and How changes to the built environment
and which measures might have prevented or refined as data and knowledge improve on have thwarted infection.
slowed its transmission. For example, would human flows on the following three levels.
more mask-wearing, social distancing or clean- Underground sewers and pipes led to
ing at airports have helped? Or was it the time City-wide. A base map is needed to represent straighter and wider streets in London, after
spent on the plane breathing in others’ air or the main flows of people throughout a city. a severe cholera outbreak was linked to raw
touching surfaces that was crucial? Or did the This should cover the numbers and timings sewage in drinking water in the 1850s.
disease spread in the streets and restaurants of people travelling between key sites, such Wipe-down surfaces made homes easier
after passengers had disembarked? We need as schools, shopping centres, train stations or to clean after ‘germ theory’ arrived in the
to learn more. care homes. Such data might be drawn from nineteenth century. Dusty draperies made
Likewise, little is known about how people anonymized mobile-phone records or infor- way for tile and linoleum.
move through cities. Some municipalities, mation about trips on public transport — local Bright light and fresh air were
including Paris, Singapore and Taipei, release authorities need to be encouraged to release championed by modernist architects to
real-time data on land transport, including pas- the information to researchers. Travel into and promote health. People with tuberculosis
senger volumes at stops or stations by origin out of the city should be estimated at major were put in airy sanatoriums with
and destination. These data sets are made avail- stations or airports, where a high volume of whitewashed walls, large windows and
able to companies, third-party developers and mixing occurs. The availability and quality of balconies. Old tenements were torn down
researchers to develop route-planning apps, data, as well as overcoming privacy concerns, to avoid contagion. And green spaces,
for example. But most cities don’t supply, and are the main limitations. Individuals in cars are such as New York’s Central Park, were
might not even collect, such data. They lack difficult to monitor, for example. The mod- protected as urban ‘lungs’.
the technology and regulatory frameworks elling frameworks are well established in the Medicines such as vaccines, antibiotics
needed to share sensitive personal informa- field of transportation and mobility research and antiviral drugs now deflect attention
tion on where and when someone has travelled. for planning and development purposes. For from the state of our streets. Yet poor
Privacy must be assured. example, in 2016 in Singapore, a National people in urban settings still experience
Also, trip data cannot tell you how much Science Experiment tracked the movements ill health in overcrowded and cramped
time someone spent in a certain location, or of 43,000 students from 128 schools using a housing. Diseases such as malaria are
with whom or for how long. Little is known custom-built pocket sensor. spreading fast in rapidly developing urban
about people entering or leaving a city, includ- fringes in Africa. Severe acute respiratory
ing tourists. And detailed information on the Bottlenecks. The mixing indoors at transport syndrome (SARS) and COVID-19 highlight
locations of COVID-19 cases is largely missing. hubs, airports, large shopping centres or the risks of outbreaks in densely populated
Some countries have comprehensive test and libraries should be added to the base map, and high-throughput locations.
trace systems; most do not. GPS data from for example by using Wi-Fi signals from smart
mobile phones have been used in Singapore devices5. Locations can then be ranked accord- destinations7. Typical flow patterns can be
and South Korea as part of contact tracing, ing to the intensity of incoming and outgoing extrapolated from previous studies. For exam-
but not in countries such as the United States, flows and the diversity of journey origins and ple, shopping centres and parks are known to
be busiest at weekends, train stations at rush
HOUSING RISKS hours on weekdays, schools at the beginning
Almost 95% of Singapore’s cases of COVID-19 have occurred in dormitories that house migrant workers. So far, and end of the day and airports during long
more than 42,000 workers have been infected — 12.5% of the people living in these high-density blocks. In the
rest of the population of around 5.3 million people, just 2,600 cases have been recorded. weekends. The proximities of people and the
durations of interaction can be estimated,
Rest of Singapore (excluding migrant workers) with assumptions such as geographer Waldo
Dormitories (migrant workers and related cases)
Tobler’s ‘first law of geography’ — “everything
1,200
is related to everything else, but near things
are more related than distant things”. More
Number of COVID-19 cases

accurate data should then be sought at key

points of convergence, such as stadiums, bars,
800 nightclubs and theatres, and the model refined
accordingly.
SOURCE: SINGAPORE MINISTRY OF HEALTH

Epidemiology. All known factors affecting

disease transmission, in terms of demo-
400 graphics and human interactions, should be
included across a representative range of
human activities. For example, preschoolers
mingling in playgrounds, co-workers interact-
ing in factories and offices or elderly residents
0
1 April 8 15 22 29 6 May 13 20 27 3 June 10 17 24 1 July 8 shopping and socializing. Environmental fac-
2020 2020 2020 2020 tors must also be considered, such as risk of

354 | Nature | Vol 583 | 16 July 2020

©
2
0
2
0
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
EDGAR SU/REUTERS

A robot dog reminds park users in Singapore to stay apart.

infection in an outdoor stadium versus gyms, Wider sources of data on human mobility residential blocks, such as those for migrant
hair salons and theatres. Such knowledge is need to be tapped8. For example, ‘smart’ cities workers, should be examined.
rapidly evolving for SARS-CoV-2. Where there such as Singapore have networks of cameras Longer term, the design and management of
are gaps, information on other viral diseases on lamp posts to track traffic flows. These cities should be altered to minimize the spread
with similar transmission modes (such as could be reconfigured to track the density of disease and the chance of future pandem-
rubella, SARS, pertussis, smallpox and influ- and mixing of people (anonymously). Built-in ics. Failing that, we’ll have to rely on packs of
enza) can inform preliminary strategies. barometers in some smartphones could fol- robotic dogs to shepherd us.
By combining all these insights, govern- low vertical movement, such as the flows of
ments will be better able to anticipate super- workers in office towers and the circulation
spreading locations and target precautionary of people in high-rise shopping centres and The authors
measures, such as delaying reopening busi- residences. Data from geolocation and con-
nesses, quarantining arrivals, tightening tact-tracing apps can provide insight into Roland Bouffanais is an associate professor
crowd control and intensifying cleaning and where people go, with whom they interact, at Singapore University of Technology
disinfection in particular places. what they do and even how they feel9. For and Design, Singapore. Sun Sun Lim is a
instance, the GeoCoV19 study analysed 524 professor and head of humanities, arts and
Next steps million geolocated tweets in 62 languages, social sciences at Singapore University of
Funding agencies should support accelerated posted over 90 days since 1 February to eval- Technology and Design, Singapore.
studies of human movement and interac- uate public sentiments, emergency needs, e-mails: [email protected];
tions in key superspreading locations such knowledge gaps and misinformation trails10. [email protected]
as transport hubs. Scientists need to rank Such data would have to be anonymized to
superspreading potential and compute the protect privacy.
effect of measures such as social distancing Governments should use these data and 1. Manivannan, A., Yow, W. Q., Bouffanais, R. & Barrat, A.
EPJ Data Sci. 7, 34 (2018).
or mask-wearing. For example, might primary models to hone their public-health strategies. 2. Isella, L. et al. J. Theoret. Biol. 271, 166–180 (2011).
schools with a small catchment area reopen More effective targeting of measures will help 3. Stehlé, J. et al. PLoS ONE 6, e23176 (2011).
at full capacity if all students and educators to avoid ‘virus fatigue’ among the public and 4. Baldwin, A. N. Proc. Inst. Civil Eng. Municipal Eng. 159,
37–42 (2006).
wear a mask? What about secondary schools, help education and the economy by allowing 5. Del Valle, S. Y., Hyman, J. M., Hethcote, H. W. &
which have a more diverse student body that places to minimize the risks of some kinds of Eubank, S. G. Soc. Networks 29, 539–554 (2007).
is likely to engage in riskier behaviours, such reopening. 6. Jiang, L. et al. J. Hosp. Infect. 98, 404–411 (2018).
7. Chin, W. C. B. & Bouffanais, R. Preprint at https://arxiv.org/
as dating or drug-taking? Urban planners should also re-examine abs/2005.05063 (2020).
Urban analysts and modellers need to disease spread in the built environment. For 8. Willemse, E. J., Tunçer, B. & Bouffanais, R. Proc. 24th Int.
Conf. Assoc. Computer-Aided Architect. Design Res. Asia
understand the dynamics of face-to-face inter- example, they might integrate safe-distancing
2, 805–814 (2019).
actions, networks and crowd mixing. Key ques- measures into street and building designs. 9. Ling, R., Fortunati, L., Goggin, G., Lim, S. S. & Li, Y. in The
tions include: how often are we in close contact More walkways and vertical parks might sep- Oxford Handbook of Mobile Communication and Society
(eds Ling, R., Fortunati, L., Goggin, G., Lim, S. S. & Li, Y.)
with people? For how long? What places in our arate flows of people, as at the Kampung Admi-
3–12 (Oxford Univ. Press, 2020).
normal daily movements put us in close con- ralty development in Singapore (see go.nature. 10. Qazi, U., Imran, M. & Ofli, F. SIGSPATIAL Spec. 12, 6–15
tact with the greatest number of new people? com/2z9pxmz). The health impacts of dense (2020).

Nature | Vol 583 | 16 July 2020 | 355

©
2
0
2
0
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.