Living In A Pandemic Changes In Mobility Routines Social Activity And Adherence To Covid 19 Protective Measures
Summary
Living in a pandemic: changes in mobility routines, social activity and adherence to COVID-19 protective measures is a research article in Scientific Reports, (2021) 11:24452, by Lorenzo Lucchini, Simone Centellegher, Luca Pappalardo, Riccardo Gallotti, Filippo Privitera, Bruno Lepri and Marco De Nadai. It uses GPS mobility traces of more than 837,000 anonymous opted-in individuals from 3 January 2020 to 1 September 2020 in Arizona, Oklahoma, Kentucky and New York to study responses to non-pharmaceutical interventions. The authors report that New York had the largest drop in visits to points of interest, 69%, with the other states between 55% and 60%. They find people spent less time in venues, kept simpler routines and kept person-to-person contacts low as visits recovered. The article closes with author contributions and an open access license statement.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Living In A Pandemic Changes In Mobility Routines Social Activity And Adherence To Covid 19 Protective Measures.”
Original at doi.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/living-in-a-pandemic-changes-in-mobility-routines-social-activity-and-cb71acda2e/
, SHA-256 3b227bdf9fbf019eefa9c651bef1ea0ff4d994d25698b69aa834f672b7f03f1f.Full text
www.nature.com/scientificreports
OPEN Living in a pandemic: changes
in mobility routines, social activity
and adherence to COVID‑19
protective measures
Lorenzo Lucchini1,4*, Simone Centellegher1,4, Luca Pappalardo 2
, Riccardo Gallotti1,
Filippo Privitera3, Bruno Lepri1 & Marco De Nadai1*
Non-Pharmaceutical Interventions (NPIs), aimed at reducing the diffusion of the COVID-19 pandemic,
have dramatically influenced our everyday behaviour. In this work, we study how individuals adapted
their daily movements and person-to-person contact patterns over time in response to the NPIs.
We leverage longitudinal GPS mobility data of hundreds of thousands of anonymous individuals to
empirically show and quantify the dramatic disruption in people’s mobility habits and social behaviour.
We find that local interventions did not just impact the number of visits to different venues but also
how people experience them. Individuals spend less time in venues, preferring simpler and more
predictable routines, also reducing person-to-person contacts. Moreover, we find that the individual
patterns of visits are influenced by the strength of the NPIs policies, the local severity of the pandemic
and a risk adaptation factor, which increases the people’s mobility regardless of the stringency of
interventions. Finally, despite the gradual recovery in visit patterns, we find that individuals continue
to keep person-to-person contacts low. This apparent conflict hints that the evolution of policy
adherence should be carefully addressed by policymakers, epidemiologists and mobility experts.
The COVID-19 pandemic has prompted many countries to implement a diverse set of Non-Pharmaceutical
Interventions (NPIs) such as international travel restrictions, physical distancing mandates, closures of business
venues, and stay-at-home orders to prevent the spread of the virus1–7. These policies have had a profound impact
on numerous aspects of human life including employment8,9, economy10–13 and people’s social behaviour14–17.
In this context, mobile phone data offered unprecedented opportunities to capture the effects of the NPIs
and to understand better their impact on the evolution of the epidemic18,19. Different technological and telecom-
munication companies have released analysis and data to help researchers estimate the epidemic spread16,20–25,
mobility reduction16,26–29, physical distancing30,31, physical activity32, and informing on the efficacy of NPIs25,33–36.
However, most of these works are based on aggregated data and focus their analysis on macro mobility indicators
to inform epidemic models16,31,34.
While the short term behavioural adaptation to NPIs and their impact on epidemic models is of paramount
importance to promptly respond to the pandemic threat, an in-depth understanding of how individual behaviour
adapted over time is still lacking. Understanding the long term behaviour in a condition of sustained epidemic
threat represents an essential factor in facilitating policy adherence.
This paper studies the changes in the daily routinary behaviour of people, focusing on how individuals adapted
their pattern of visits and social contacts over time. We combine Point Of Interest (POI) information extracted
from OpenStreetMap (OSM) with privacy-enhanced longitudinal GPS mobility traces of more than 837,000
anonymous opted-in individuals, measured for nine months from 3 January 2020 to 1 September 2020. Our
dataset has an average accuracy of 22 m and covers 16 hours of activity per day. This high precision and coverage
allow us to examine human mobility at a fine spatial and temporal granularity while ensuring users’ privacy (see
SI Section S2-A for additional details). We analyse, model, and compare individual’s mobility in four US states,
including those with the highest and lowest values of daily COVID-19 death rate and NPIs stringency: Arizona
(many deaths and low stringency), Oklahoma (few deaths and low stringency), Kentucky (few deaths and high
stringency), and New York (many deaths and high stringency) (see SI Section S5 for details).
1
Fondazione Bruno Kessler (FBK), Trento, Italy. 2Institute of Information Science and Technologies, National
Research Council (ISTI-CNR), Pisa, Italy. 3Cuebiq Inc., New York, NY, USA. 4These authors contributed equally:
Lorenzo Lucchini and Simone Centellegher. *email: lorenzo.f.lucchini.work@gmail.com; work@marcodena.it
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 1
Vol.:(0123456789)
www.nature.com/scientificreports/
Figure 1. (A) From an individual’s original GPS trajectory, we detect a stop whenever the individual spends
at least 5 minutes within a distance of 65 meters from a given trajectory point. (B) We detect stop locations
through a combination of the Lachesis37 and DBSCAN algorithms38. (C) If present, we associate each stop
location to the nearest Point Of Interest (POI) within a distance of 65 m.
Our results describe the disruption effect of the COVID-19 pandemic on human behaviour over time at an
unprecedented level of detail. Overall, we find that people changed both how they visit places and how they
experience them. They spend less time in venues, preferring simpler and more predictable routines. Individual
patterns of visits to POIs are shaped by the strength of the NPIs policies, the pandemic’s local severity and a risk
adaptation factor, which increases the people’s mobility regardless of the stringency of interventions. Finally,
despite the gradual recovery in visit patterns, person-to-person contacts remained lower than expected given
the increased time spent at venues.
Results
We explore and characterize human mobility from an individual’s set of stop locations, defined as places where
a person stays for at least 5 minutes within a distance of 65 meters. From the original GPS data (see Fig. 1A), we
detect the stop locations of each individual through a combination of the Lachesis37 and DBSCAN algorithms38
(see “Methods” and Fig. 1B). As a result, stop locations are described as tuples (lat, lon, start-time, end-time),
where an individual stays in a particular location with latitude (lat) and longitude (lon) from start-time to end-
time. lat and lon are the mean latitude and longitude values of the GPS points found within the specified distance
of 65 meters (we refer to the “Methods” and SI Section S3 for the details). Then, we focus on the home and work
locations of people. To preserve privacy, the data provider obfuscates users’ precise home and work locations by
transforming it to the centroid of the corresponding Census Block Group. Thus, we identify the home census
block group of users, from now on called Residential area, by looking at the most visited locations during the
nights (from 8 pm to 4 am) with a moving time window of 28 days. Similarly, we also identify the Workplace,
defined as the most visited census block group during the week (from 9 am to 5 pm), which is not marked Resi-
dential. We refer to SI Section S4 for additional details.
Finally, we add semantic meaning to individuals’ mobility trajectories associating each stop location to the
nearest Point of Interest (POI), extracted from OpenStreetMap (OSM)39, whenever a POI lies within 65 m from
the stop location (see Fig. 1C). POIs are commonly described as public locations that people may find interest-
ing, for example, for business or recreational activities40. Since the OSM POIs taxonomy is not hierarchically
organized, we create a human-curated mapping from the OSM tagging system41 to the Foursquare venue category
hierarchy42 (see “Methods” and SI Section S7 for details). SI Section S7-B shows the popularity of POIs in dif-
ferent states, highlighting the variety of the visiting behaviour in the US.
To validate the data provided by Cuebiq and our pre-processing, we compute the correlation of the time-series
of visits to POIs, residential areas and workplace areas between our data, Google data26, and Foursquare data43.
We report an average Pearson correlation in New York state of 0.91 and an average Pearson correlation of 0.84
in all four selected states (see SI Section S8).
Starting from the visits of each individual we analyse how people’s lifestyle adapted during the pandemic.
Changes in POIs visit patterns. We begin our analysis by describing the impact of the COVID-19 pan-
demic on the visits to POIs over time. Overall, we observe that NPIs are associated with fewer visits to POIs,
confirming previous results obtained with similar datasets26,32,34.
Figure 2 shows how the number of visits (panel A) to POIs and the duration of those visits change over time
in the four selected US states. It also highlights the differences across states in which the epidemiological situ-
ation and the stringency of the enacted policies differed most (see Section S5 for more details). For example,
the state of New York experienced the most significant drop in the number of visits to all types of POIs (for
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 2
Vol:.(1234567890)
www.nature.com/scientificreports/
Figure 2. Changes in number and duration of visits to all POIs in the states of Arizona, Kentucky, New York
and Oklahoma. (A) Percentage change over time in the number of visits with respect to the baseline period (3
January 2020–28 February 2020) for all types of venues. (B) Change in percentage of the duration of visits over
time with respect to the baseline period of the median duration of visits to all POIs. For visualization purposes,
the original curves are smoothed using a rolling average of seven days.
details on POIs refer to SI Section S7) with a reduction of 69%. The other states experienced smaller but similar
reductions in POIs visits in a range between 55% and 60% (see SI Section S7-E for a more detailed analysis on
single POI categories). Noteworthy differences are present also in the recovery phases due to the different re-
opening measures. As an example, due to early re-opening phases and the absence of restrictions after the end
of the stay-at-home order on 15 May 2020, the state of Arizona experienced a smaller drop in the number of
visits to POIs with respect to New York state (Fig. 2A). However, in the long run, the state of New York recov-
ers better since Arizona was forced to introduce new closures due to rising COVID-19 cases. Interestingly, we
observe that individuals allocate the everyday time spent similarly across states (Fig. 2B), regardless of the NPIs
implemented, but again we find some differences potentially due to the different NPI strategies. We refer to SI
Section S9 for a set of additional metrics (i.e. individuals’ unique stop locations, diversity of visits, and radius of
gyration) supporting our findings.
A similar change in mobility behaviour is also found at a more fine-grained level, by focusing on the differ-
ent POI categories. As shown in Fig. 3, in the state of New York, POIs belonging to the Food (e.g., restaurants)
and Shop & Service (e.g., book stores and supermarkets) categories experienced low points of −75% and −56%
in the number of visits, respectively. The duration of visits is also severely impacted by COVID-19 reaching low
points of −53% and −27.3% for Food and Shop & Service, respectively. These reductions in the number of visits
are heterogeneous. For example, essential shops such as supermarkets faced a lower reduction in the number
of visits than the non-essential shops (with low points of −38.3% vs −67% shortly after the stay-at-home order)
and even increased the number of visits before the stay-at-home order (see SI Section S7-F).
Since two weeks after the stay-at-home order, we observe an increase in the number of visits, which comes
after the government’s decision of expanding the list of the essential businesses and the progressive relaxation of
government restrictions (see Fig. 3A,B) throughout the various re-opening phases, conveniently named phase-
one, phase-two, and phase-three (see the “Methods” Section). Interestingly, while the number of visits gradually
recover in the period between 25 March 2020 and 20 May 2020 (see Fig. 3A), the duration of visits recover much
slowly. At the end of the period of study, the duration of visits increases by 60% and 51% for Food and Shop &
Service respectively from the pre-pandemic period, while visits recover by 69% and 67% for Food and Shop &
Service respectively from the pre-pandemic period. Thus, individuals changes how they experience POIs. We note
that the number of visits to Food and Shop & Service starts increasing between the stay-at-home order and phase-
one (see Fig. 3A), despite the absence of significant changes in the restrictions. We will later analyse this finding.
Figure 3C shows the distribution of time spent by individuals over time in the state of New York. We group
the time spent into five categories: Residential, Workplace, POI, Other (i.e. stop locations not matched with a POI
and not detected neither as Residential or Workplace), and Moving (i.e. time spent moving from place to place).
After the stay-at-home order, we find that the percentage of time spent in residential areas increases significantly
to the expense of the time spent at Workplace, POI, Other, and Moving. After phase three, we find that the time
spent in residential areas gradually reduces while other categories increase. The time spent in each category,
however, does not return to the pre-pandemic period.
Figure 3D shows the percentage of time spent by people in the venues of the eight first-level categories of
POIs. The time spent at Shop & Service increases at the expense of the other POIs. Then, after phase three, people
significantly increase the time spent in some categories such as Outdoors & Recreation and Travel & Transport,
probably also due to seasonal effects, but other categories such as Arts & Entertainment do not recover to pre-
pandemic levels. We refer the readers to SI Section S1 for a more complete example of the reduction of visits in
New York City.
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 3
Vol.:(0123456789)
www.nature.com/scientificreports/
Figure 3. Changes in number and duration of visits to POIs in the state of New York. (A) Percentage change
over time in the number of visits with respect to the baseline period (3 January 2020–28 February 2020) for
venues in the Food and Shop & Service categories. (B) Change in percentage of the duration of visits over
time with respect to the baseline period of the median duration of visits to Food and Shop & Service POIs. For
visualization purposes, the original curves are smoothed using a rolling average of seven days. Vertical dashed
lines indicate the date of restrictions and orders imposed by the state of New York government. (C) Percentage
of time spent by people at Residential, Work, POIs, Other, and Moving (i.e. people in movement). (D) Percentage
of time spent by people in venues under the eight first-level categories of POIs.
Disruption of individual mobility routines. By looking at the aggregated mobility, we only have a par-
tial view of the disruption in people’s lives due to COVID-19. Individual’s behaviour exhibit mobility motifs and
routines that characterize the chronological sequence of where they go, spend money and meet people44. For
this reason, we now focus on the chronological sequences of visits, which help us understand the complexity of
changes in mobility.
We transform the individual’s chronological sequence of visits to places into a sequence of symbols (e.g.,
Food, Residential, Workplace). Then, we apply the Sequitur algorithm44–46 to generate a hierarchical representa-
tion of the original sequence, compressing repeated occurrences. Due to computational constraints, we focus on
two 4-week periods, before (from 1 February 2020 to 28 February 2020) and during pandemic (from 21 March
2020 to 17 April 2020) and focus on significant routines of each user. To detect the significant routines, we
generate 1000 randomized sequences with the same number of visits as the original sequence. Then, we define
an individual’s sequence as significant if it has a z-score < 2 between the occurrence of the real routines with
respect to the randomized ones. Finally, we filter out all the non-significant routines. We refer to the “Methods”
for additional details.
The significant routines represent meaningful sub-sequences of an individual’s mobility and allow to better
understand, at a micro-level, how human mobility preferences changed with COVID-19. To that end, we model
the ordered sequences of visits to POIs as a weighted undirected network in which the nodes represent the
POI categories, and a link between category c1 and category c2 exists if there is at least a chronological sequence
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 4
Vol:.(1234567890)
www.nature.com/scientificreports/
Figure 4. Changes in routine behaviour before and during the pandemic periods. (A, B) Network of the
subsequent movements between POI categories for all users in the state of New York before and during the
pandemic. The thickness of the links is proportional to the square root of the number of movements between
the two POI categories. For visualization purposes we are showing only links for which the average number of
daily movements is greater than 1/10 000. Pre-pandemic (C) and during-pandemic (D) distributions of these
intensities excluding self-loops (e.g., Residential ↔ Residential, Food ↔ Food connections). Jaccard similarity
matrix between individuals’ routines before the pandemic (E) and during the pandemic (F).
where c1 immediately follows c2 or vice-versa. The weight of the link represents the daily average proportion of
sequences containing [c1 , c2 ] or [c2 , c1 ]. Figure 4A,B shows the weighted network of the POI categories in the
state of New York, where the size of the links is proportional to the intensity of the relationship between the two
POI categories. We observe that all links reduce their intensity (on average − 79%) with the exception of the
Residential ↔ Residential, which increased by 5%.
Figure 4C and D shows the distributions of the network weights in the pre-pandemic (from 1 February 2020
to 28 February 2020) and during pandemic (from 21 March 2020 to 17 April 2020) periods in New York state,
from which we excluded all self-loops (e.g., Residential ↔ Residential, Food ↔ Food connections). We observe
the mildest reduction for Shop & Service ↔ Residential (− 40% visits), Shop & Service ↔ Workplace (− 61%),
and Outdoors & Recreation ↔ Residential (− 63%), which might be interpreted as essential routines. Higher
levels of reduction are found in those POIs connections which include sectors strongly afflicted by NPIs: Arts &
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 5
Vol.:(0123456789)
www.nature.com/scientificreports/
Baseline Weather Full
NPIs stringency −0.148 ± 0.004 −0.204 ± 0.004 −0.242 ± 0.009
Death ratio −0.072 ± 0.004 −0.053 ± 0.003 −0.040 ± 0.007
Max temperature – 0.009 ± 0.004 −0.001 ± 0.099
Precipitations – −0.002 ± 0.003 0.001 ± 0.100
Risk adaptation – – 0.118 ± 0.004
R2 0.67 ± 0.02 0.72 ± 0.02 0.78 ± 0.02
PSIS-LOO 610.60 ± 31.94 747.44 ± 28.78 846.76 ± 29.59
Table 1. Quantitative results of the Bayesian multivariate linear mixed model to explain the daily number of
visits to POIs. We report the mean and 95% confidence intervals of all the β coefficients. We report the mean
and standard deviation for R2 and PSIS-LOO.
Entertainment ↔ College & University (− 95%), College & University ↔ Nightlife Spot (− 95%), College & Univer-
sity ↔ Residential (− 93%), and Arts & Entertainment ↔ Food (− 92%) (for further details refer to SI Fig. S20).
The dramatic change of people’s behaviour also emerges from the similarity between the characteristic routine
of different individuals. We represent each individual’s significant routine behaviour regarding the presence or
absence of two-elements sequences between POI categories. Then, we compute the Jaccard similarity between
a randomly selected sample of 10,000 individuals. Finally, we apply agglomerative hierarchical clustering47 to
find relevant groups of individuals with similar routine behaviour. By comparing Fig. 4E and F, we observe that
Fig. 4F contains larger clusters, which means that mobility routines simplify and people’s behaviour gets more
homogeneous. This result is also quantitatively confirmed by the larger silhouette score48 of all individuals during-
pandemic period compared with the pre-pandemic period (see SI Section S10-D for details). By inspecting the
everyday routines in the two biggest clusters, which include almost 34% of users, we observe that individuals
limit their mobility to Residential ↔ Residential, Residential ↔Shops & Services and Shops & Services ↔ Shops &
Services routines. We refer to SI Section S10-E for a comparison between the six biggest clusters.
During the pandemic period, we also find that individuals tend to favour simpler and more redundant
sequences. We measure the compression ratio44 defined as the length of the original sequence divided by the
length of the Sequitur compressed sequence. On average, in the state of New York, the individual’s sequences
before the pandemic have a compression ratio of 2.75, while during the COVID-19 pandemic, the compres-
sion ratio increases by 40%, reaching 3.79 (see SI Fig. S16). Similar results apply in all the other states (see SI
Fig. S10B).
Evolution of policy adherence and risk adaptation in visit patterns. In Fig. 3A, we observe that
the number of visits starts increasing between the stay-at-home order and phase-one, despite the absence of
significant changes in the restrictions. To explain this behaviour, we hypothesise the presence of progressive
behavioural relaxation and adaptation to the epidemic risk, also observed by previous literature30,49 (and some-
times called “pandemic fatigue"50).
We use a multivariate Bayesian linear mixed model to examine the daily number of visits to POIs in each
state through the strength of the NPIs, daily death ratio, and weather conditions (i.e. daily max temperature and
precipitations). NPIs’ strength and daily death ratio are used as proxies to capture the local socio-epidemiological
condition. More precisely, the first include information about the intensity of restrictions, such as closures and
gathering restrictions, while the second quantify the severity of the condition and the local burden of the disease
over time. Similarly, weather conditions account for environmental confounding variables which might have an
impact in human mobility behaviour. We complement the model by adding a factor that tests the risk adaptation
hypothesis by means of a time-dependent sigmoid function.
We account for the different mobility behaviour of people across states and day of the week, by including a
random effect for the state and a random effect for the day of the week. More information on the data sources and
models is provided in “Methods”. Alternative model specifications, including different risk adaptation functional
forms, are provided in the SI Section S12.
We select, as a baseline, a model that includes as fixed effects only the NPIs stringency and the death ratio
over the state population. We evaluate the model through the well-established Bayesian R251 and the PSIS-LOO
information criterion52. Table 1 shows that this simple model achieves R2 = 0.67 and PSIS-LOO = 610.60 and,
as expected, shows that the NPIs stringency correlates negatively with the mobility of people. Interestingly, we
also find that the death ratio influences the number of visits to POIs.
Then, we account for local weather conditions that might influence the visits to POIs. The Weather model
adds the daily precipitations and maximum temperature to the baseline model. Table 1 shows that these two
variables significantly increase the model’s performance ( R2 = 0.72, PSIS-LOO = 747.44) that grow by 7.46%
and 22.41%, respectively.
Table 1 shows that the inclusion of the risk adaptation factor (see Full model) achieve the highest performance
( R2 = 0.78, PSIS-LOO = 846.76). We observe that in this model, the risk adaptation factor represents the sec-
ond most important factor for understanding the daily visits to POIs. This result holds even when we model the
time spent outside the home and when we hypothesise that the risk adaptation depends on the cumulative NPIs
stringency, which varies from state to state (see SI Section S12).
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 6
Vol:.(1234567890)
www.nature.com/scientificreports/
Figure 5. Co-location events. (A) Percentage change of co-location events from the baseline (until 29 February
2020) in New York state. We measure the change in co-location events for Residential areas (only one of the
two individuals is in proximity of their residential area), POI (both individuals are at the same POI), Workplace
(both individuals have the same work location) and Other co-location events. (B, C) We compare the difference
between the expected and observed number and duration of co-location events. During the pandemic,
individuals tend to have fewer and shorter co-locations than expected.
Protective behaviour in co‑location events. The risk adaptation factor can be interpreted as a change
in adherence to COVID-19 protective behaviours and policies. However, it is not clear whether this adaptation
only affect people’s tendency to visit more venues or if it is also reflected on the physical distancing patterns.
As a proxy to understand how much people engage in physical and social activities, we empirically define a
co-location event as when two individuals stop for at least fifteen minutes and are at most 50 m apart from each
other. These events include, among others, people going to the same grocery store, residential area and work-
place. We note that our definition of co-location is more robust than previous literature30,31 since we represent a
social contact from stop locations, thus reporting a co-location event between two people when they share the
same stationary point.
We aggregate co-location events in four different categories depending on where the possible social contact
took place: (i) Residential, a co-location event where only one of the two individuals have the stop marked as
Residential location; (ii) Workplace, a co-location event that happened in a venue labeled as a workplace for both
the individuals; (iii) POI, a co-location event where both the individuals are in the same POI; and (iv) Other,
a co-location event in which the two individuals meet in a place that it is neither a Residential nor a Workplace
nor a POI (see SI Section S11 for additional details). We here note that Residential and Workplace co-locations
are defined at a coarse level, due to the anonymisation process performed by Cuebiq.
Figure 5A shows the abrupt change of the co-location events in the New York state, starting at the school
closure day on 15 March and reaching low points of −95%, −88% and −90% for POI, Workplace and Other
co-location events respectively. Interestingly, we observe that Residential co-location events, namely between
people who do not live together, experience the smallest reduction, decreasing at most by 53% from the pre-
pandemic levels. During the strictest measures put in place in New York state, we notice that people maintain
their co-location events inside other’s people residential areas and in places which are not marked as POIs (see
Residential and Other in Fig. 5A). This presumably happens because of the impossibility of having co-location
events in venues such as pubs and restaurants (see Food and Nightlife Spot categories in SI Section S11-A) due
to the NPI interventions including physical distancing measures and closures of POIs such as Arts & Entertain-
ment and College & University.
We obtain similar results in the other states (see SI Section S11), although with some differences. For example,
in Oklahoma, Kentucky and Arizona, the Residential co-locations events experience a lower reduction, with a
low point around −33% in Arizona and around −40% in both Oklahoma and Kentucky. Notably, we observe a
decrease in co-locations events from the partial reversal of the reopening from 29 June 2020 in Arizona.
Again, we find some differences between the number and the duration of co-location events. First, the dura-
tion decrease less than the number of co-location events, reaching low points of around −10%, −48%, −10%,
−30% for Residential, POI, Workplace and Other co-location events respectively (see SI Section S11-B).
Overall, we find a strong negative Spearman correlation between the daily number of co-location events
and the NPIs stringency in each state (−0.83 with p-value p < 0.001). Similarly, we find that the daily number
of co-location events is also negatively correlated with the daily number of new cases and deceased (−0.58 and
−0.67 respectively, with p-values p < 0.001).
The co-location events and the visits to places are inherently connected. Therefore, as soon as the number
and the duration of visits decrease, it becomes less probable to have co-location events. Thus, we estimate the
daily expected number of co-location events through a null model and compare it with the observed co-locations.
The number of co-location events ei,d occurring at a POI i on a specific day d can be estimated from both the
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 7
Vol.:(0123456789)
www.nature.com/scientificreports/
number of individuals visiting the POI ni,d , and the median
duration
of their stops there d̄i,d . We can then have
ni,d
an estimate of the co-location events following ei,d = pi,d , where pi,d is the probability of having a co-
2
location event given two individuals visiting POI i on day d and pi,d is computed assuming a uniform distribution
for the time-interval of visit of two individuals potentially having a co-location event (see SI Section S11-C for
additional details). We follow a similar reasoning to model the expected duration of the individual’s co-location
events.
Figure 5B shows that, during the pandemic, the observed number of co-location events at POIs are lower
than expected. Similarly, Fig. 5C shows that the duration of co-location events is slightly lower than expected.
We investigate the deviation between the expected and observed number of co-location events and find a high
correlation with the daily new cases and deaths (0.66 p < 0.001 and 0.48 p < 0.001, respectively), and a lower
correlation with the NPIs stringency (0.28 p < 0.001). Thus, individuals’ discrepancy between the theoretical
and observed number of co-location events appear to be largely driven by the epidemic burden (i.e. daily new
cases and deaths).
Discussion
This paper investigates the changes in human mobility behaviour induced by the COVID-19 pandemic in four
US states. We exploit a privacy-enhanced longitudinal GPS dataset, counting more than 837,000 anonymous
opted-in individuals, to show how individuals changed their patterns of visits, their routinary behaviour and
their person-to-person contact activity over time.
We show that, in line with precious research18,26,32, the COVID-19 pandemic dramatically reduced the number
of visits to POIs, which only partially recovered the pre-pandemic levels (−28% in the state of New York at the
end of the period of study). Additionally, while the duration of visits to POIs decreases less than the number of
visits (−23% at the end of the period), its recovery after the reopening phases is slower. This finding suggests that
people are less willing to spend time in POIs, reasonably to minimise social contacts in public venues. This result
is strengthened by the analysis of co-location events used as a proxy of social contacts. Since the number of co-
location events depends on the number and duration of visits to places, we expect that it becomes less probable
to have social contact as soon as these quantities decrease. Nevertheless, we find that individuals have fewer and
shorter contacts even after the reopening phases. Despite increasing the number of visits to POIs, individuals
display persistent protective behaviours, limiting unnecessary social contacts.
We also find that changes in co-location events are often place-dependent. For example, on average, co-
locations events at other’s people residential areas, namely between people who do not live together, reduced
by 42%, while co-locations in POIs and workplaces reduce on average by 94% and 73% after the stay-at-home
order. While individuals limit unnecessary social contacts, this behaviour shows a limited impact on in-house
gatherings. Here, our results reveal a potentially unsafe behaviour, as the literature reports in-house contacts
among the ones with a higher risk of infection53,54, highlighting the importance of considering all co-location
places when modelling optimal intervention strategies.
Overall, human routines during the COVID-19 pandemic got shorter and more predictable. The statis-
tical patterns that characterise these new routines (e.g., distribution of travel distance, typical predictability
of an individual’s whereabouts) may differ significantly from those observed ubiquitously for pre-pandemic
mobility55,56,56–65. Further research is needed to understand these differences.
Finally, taking into account the stringency of the NPIs, the local severity of the pandemic, local weather condi-
tions, and a risk adaptation factor, people adapt to the pandemic risk over time, showing an increased tendency
in visiting POIs and spending time outside their home.
Multiple reasons may explain the risk adaptation effect. One hypothesis is that risk adaptation arises from
a change in the risk assessment of individuals. As the pandemic unfolds, people become less influenced by
the number of cases and deaths, therefore paying less attention to protective behaviours. A similar hypothesis
sees the risk adaptation as a consequence of the psychological burden, which reduces the ability or motivation
to perform self-protective behaviour66,67. Another hypothesis considers the sustained economic burden as an
important factor, which may force people to decrease policy adherence to go back to work. In all cases, our
results well align with previous self-reported evidence49 and warn that adherence should not be dismissed and
ignored. Long and sustained restrictions to human activities might build up the so-called pandemic fatigue and
make NPIs less effective49.
Analysing everyday activities from GPS data does not come without limitations. First, the analysis from
smartphone data might be biased towards younger adults and fail to capture the mobility of those who do not
carry their phones while visiting places. Second, our Residential and Workplace are just an estimate of the actual
home and work locations, which are obfuscated for privacy reasons by the data provider. Third, we acknowledge
that our co-location events are a loose proxy of social interactions, and people might share the same location
even without knowing each other (i.e., familiar strangers68).
COVID-19 disrupted the lives of millions of people. Our work shows that the pandemic-induced changes
are not just about how much people stay at home and visit POIs. Instead, the COVID-19 risk and policy inter-
ventions reshaped people’s routines and habits, changing how individuals experience POIs, places and social
interactions during the pandemic.
Future work might clarify how people adapt to the epidemic risk and the medium- and long-term effects of
the COVID-19 pandemic and restriction measures on human behaviour.
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 8
Vol:.(1234567890)
www.nature.com/scientificreports/
Methods
Selection of the important dates. We identify several important dates that help readers with the inter-
pretation of our results. In the state of New York, on March 1, the first positive cases were broadly discussed by
public opinion. On March 15, Governor Andrew Cuomo announced that New York City schools would have
closed from the following day®. On March 22, Andrew Cuomo announced the statewide stay-at-home order,
also known as the NYS on Pause Program, with a mandate that all non-essential workers work from home begin-
ning at 8 p.m. and that only businesses declared as essential were allowed to remain open”™”!.
On May 15, Andrew Cuomo announced a gradual plan of reopening, also called Phase I. From May 28, the
New York State Department of Health released the guidelines for the reopenings of Phase I, which included
the reopening of professional services including finance and insurance, retail, administrative support, and real
estate/rental leasing”. We note that New York City met the criteria on June 22. On June 15, Andrew Cuomo
announced that regions upon entry of Phase 3 will be allowed non-essential gatherings of up to 25 people and
that on-location restaurants would open’*. New York City entered this phase on July 6. We highlight some of
these dates in the plots, and we describe the important days for all the states in SI Section S5-A.
Stop location detection. We use GPS location data provided by Cuebig, a location intelligence company
that shared a dataset consisting of anonymised GPS locations from users that opted-in to share the data anony-
mously for research purposes through a CCPA (California Consumer Privacy Act) compliant framework. To
further preserve privacy, the data provider obfuscates users’ precise home and work locations by transforming it
to the centroid of the corresponding Census Block Group.
The dataset span a period of 9 months, from January 2020 to September 2020 (details in SI Section S2). The
data is provided through the Cuebiq Data for Good COVID-19 Collaborative program, which provides access
to de-identified and privacy-enhanced mobility data for academic research and humanitarian initiatives only.
To ensure the data describes people’s mobility throughout the pandemic, we filter out all users with less than
one month of data before declaring a national emergency (March 13, 2020) and less than four months after it.
We also require users have 5 hours per day covered by at least one GPS location. The resulting dataset includes
more than 837,000 anonymous, opted-in individuals.
For all users, we extract their stop events with an algorithm based on Hariharan and Toyama’’. A stop event
is defined as a temporal sequence of GPS coordinates in a radius of A, meters where a user stayed for at least A;
minutes. The algorithm, its optimisation, and its computational complexity are explained in detail in SI Section
S3. To define a stop event, we used A, = 65 meters and A; = 5 minutes due to the distribution of accuracy of
the underlying data (see SI Section $3).
For each user, we then define their stop locations as the sequences of stop events that can be considered
as part of the same place. To determine a stop location from a sequence of stop events we use the DBSCAN
algorithm**’. With DBSCAN, we group points within a distance of e = A, — 5 meters to form a cluster with at
least minPoints = 1 stop event (see SI Section S3 for more details).
Point of Interest (POls) association. We extract all POIs from OpenStreetMap (OSM) (https://www.
openstreetmap.org/) and then, due to the lack of structure in OSM POIs, we map each OSM POI to the cor-
responding Foursquare Venue Category Hierarchy (https://developer.foursquare.com/docs/build-with-fours
quare/categories/). After the association, each OSM POI is mapped to the Foursquare categorisation with 8 first-
level categories and 178 second-level categories. Further details are described in the SI Section S7.
Then, we associate each stop location to its nearest POI for all users whenever the Haversine distance is less
or equal to 65 meters. Since in OpenStreetMap POIs can be represented as Points or Polygons, sometimes nested,
we assign POIs to stop locations with the heuristic described in SI Section $7.
Percentage change. Throughout the analysis we compute the change with the same methodology of the
Google mobility reports”*. Specifically, we compute the percentage change as:
-_ bay
p, = i= 5 100
by (i)
where 1 is the original value at day i and by is the median value at day of the week w(i), going from 0 to 6, com-
puted during the baseline period (i.e. before the pandemic).
Sequitur, significant routine patterns, and user-user similarity. Sequitur is a compression algo-
rithm that reduces a sequence size by introducing new symbols/words when in the original sequence appear
repetitions of short sub-sequences and motifs*’. We represent an individual's mobility through a sequence of
symbols that maps a stop location into a category (e.g., Residential, Workplace, Food). From these sequences
of symbols we extracted recurrent patterns of visits consisting of sub-sequences of length / > 2 following Di
Clemente et al.**.
We focus our attention on two 4-week periods. The first one starts on February 1, 2020 and excludes January,
which might display unusual patterns due to seasonal effects (e.g., the end of the holiday period). The second one
starts at the beginning of each state emergency response orders, i.e. the “stay-at-home” order (in Kentucky, we
selected the “healthy-at-home” order since no “stay-at-home” was ever issued). This period terminates before the
first reopenings to include the most stringent early regulatory phase for each state (for more details on relevant
dates see SI Table S1).
Scientific Reports |
(2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 nature portfolio
www.nature.com/scientificreports/
For each individual sequence of visits, we randomly shuffle the sequence 1000 times and apply the sequi-
tur algorithm on each one of them*’. For each symbol s (a visit in our case), we compute the mean number of
occurrences, j/; and its standard deviation, o;, across all random synthetic sequences. We use this quantity to
compute the standard score, z; = “—“S, where o, is the number of occurrences of each symbol s within the
original sequences. Significant routines are selected from all the sub-sequences if z; is greater than 2. We refer
to significantly recurrent sub-sequences as “routines”.
Modeling expected number and duration of co-location events. The expected number of co-loca-
tion events is assumed to depend on both the number of individuals visiting a POI, i on day d, and the average
duration of their stop there, dj. The first quantity is used to compute the combinations of possible co-location
between different individuals. The second is used to estimate the probability p;,q of having two temporal interval
of duration d;,q overlapping by at least 15 minutes over the entire time span « or a day. Combined together, our
estimate of the number of co-location events, e;,g, can be written as: ej,q = ma Pid:
Following similar reasoning, the average duration of co-location events is estimated conditioning on the
occurrence of the event and assuming that the temporal overlap of two individuals, 0;,g, depends on their average
permanence at POT i, di, qd. Thus, the average overlap area can be straightforwardly computed as: 0j,4 = i.
More information about the formulation and computations in support of these models can be found in SI
Section S11-C.
Modeling visits to Points of Interest. We model the daily average number of visits to POIs Y with a
Bayesian linear regression formulated for each day of year d, state s and weekday w as:
Ya,sw X Ms + BpolicySd + BaeathsDa—1 + BtempTa+
Badapt (1)
T+ e7Vs(d—$s) + Pw + &
BprecP ‘d+
where ¢ is the residual error, a, is the state-specific intercept, Bpolicy, Bdeaths» Btemp> Aprec aNd Badapt are the B
coefficients for each independent variable in the regression, while p,, is the week-day random effect controlled
by the variable w, the day of the week of (from 0 to 6 where 5 is Saturday and 6 is Sunday). Sq is the value of the
Stringency Index, measuring local enacted regulations and preventive and informative campaigns (see”* for
more details), Dg_, is the death ratio at the previous day (measured over a population of 100k people), Tg is the
maximum temperature in Celsius degrees and Pg represents the millimeters of precipitations. Tg and Py account
for the seasonal effects of POI visits. Then, era is a standard sigmoid function that models the collective
behavioural adaptation of people to the perceived epidemic risk. The sigmoid function depends on time where
os and y; model the location and the sharpness of the sigmoid, respectively. All the independent variables are
z-score standardised.
We extract the daily temperature and precipitation from the PRISM Climate Group”, which provides the
maximum temperature and precipitations with a 4km grid. Then, we compute the average maximum temperature
and precipitations for each state. Finally, the average is weighted with the population of each county to account
for the number of people exposed to the measured temperature and precipitations.
We assess the out of sample predictive accuracy through the Pareto-smoothed importance sampling Leave-
One-Out cross-validation (PSIS-LOO)*. This metric overcome the issues of the Deviance Information Criterion
(DIC)”* such as its lack of consistency and the fact that is not a proper predictive criterion*””, and it has rapidly
become state of the art for evaluating Bayesian models. The PSIS-LOO is defined in the log score as:
n S s :
SH(y.|0
PSIS-LOO = $7 In 2st MPU) °)
i=1 yye1 W;
where n is the number of data points, 0° are draws from the full posterior p(@|y), s = 1,...,S represent the S
draws, and w; is a vector of weights that are the Pareto Smoothed importance ratios built through an algorithm
described in the PSIS-LOO original paper”. The best model is associated with the highest PSIS-LOO value. We
also report Bayesian R®°! as an additional and easy-to-interpret measure of goodness of fit.
Data availability
Replication code is available on GitHub at https://github.com/denadai2/living-the-pandemic. All the data sources
are freely available on the Internet while the mobility data from Cuebigq can be accessed only through the Data
for Good initiative of the company (https://www.cuebiq.com/about/data-for-good/). Limitations apply to the
availability of this data, due to the rigorous anonymity constraints.
Received: 17 September 2021; Accepted: 8 December 2021
Published online: 27 December 2021
References
1. Brauner, J. M. et al. Inferring the effectiveness of government interventions against covid-19. Science 371, 1-10 (2021).
2. Chinazzi, M. et al. The effect of travel restrictions on the spread of the 2019 novel coronavirus (covid-19) outbreak. Science 368,
395-400 (2020).
3. Dehning, J. et al. Inferring change points in the spread of covid-19 reveals the effectiveness of interventions. Science 369, 1-10
(2020).
Scientific Reports |
(2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 nature portfolio
www.nature.com/scientificreports/
4. Flaxman, S. et al. Estimating the effects of non-pharmaceutical interventions on covid-19 in europe. Nature 584, 257–261 (2020).
5. Haug, N. et al. Ranking the effectiveness of worldwide covid-19 government interventions. Nat. Hum. Behav. 4, 1303–1312 (2020).
6. Haushofer, J. & Metcalf, C. J. E. Which interventions work best in a pandemic?. Science 368, 1063–1065 (2020).
7. Lai, S. et al. Effect of non-pharmaceutical interventions to contain COVID-19 in China. Nature 585, 410–413. https://doi.org/10.
1038/s41586-020-2293-x (2020).
8. Baek, C., McCrory, P. B., Messer, T. & Mui, P. Unemployment effects of stay-at-home orders: Evidence from high frequency claims
data. Rev. Econ. Stat. 1, 1–72 (2020).
9. Forsythe, E., Kahn, L. B., Lange, F. & Wiczer, D. Labor demand in the time of covid-19: Evidence from vacancy postings and ui
claims. J. Public Econ. 189, 104238 (2020).
10. Dunn, A., Hood, K. & Driessen, A. Measuring the effects of the covid-19 pandemic on consumer spending using card transaction
data. In: US Bureau of Economic Analysis Working Paper WP2020-5 (2020).
11. Chetty, R., Friedman, J. N., Hendren, N., Stepner, M. & Team, T. O. I. The Economic Impacts of COVID-19: Evidence from a New
Public Database Built using Private Sector data. w27431 (National Bureau of Economic Research, 2020).
12. Bonaccorsi, G. et al. Economic and social consequences of human mobility restrictions under covid-19. Proc. Natl. Acad. Sci. 117,
15530–15535 (2020).
13. Brodeur, A., Gray, D. M., Islam, A. & Bhuiyan, S. A literature review of the economics of covid-19. IZA Discussion Paper (2020).
14. Feehan, D. M. & Mahmud, A. S. Quantifying population contact patterns in the united states during the covid-19 pandemic. Nat.
Commun. 12, 1–9 (2021).
15. Zhang, J. et al. The impact of relaxing interventions on human contact patterns and sars-cov-2 transmission in china. Sci. Adv. 7,
eabe2584 (2021).
16. Wellenius, G. A. et al. Impacts of social distancing policies on mobility and COVID-19 case growth in the US. Nat. Commun. 12,
3118. https://doi.org/10.1038/s41467-021-23404-5 (2021).
17. De Nadai, M., Roomp, K., Lepri, B. & Oliver, N. The impact of control and mitigation strategies during the second wave of covid-19
infections in spain and italy. medRxiv (2021).
18. Perra, N. Non-pharmaceutical interventions during the covid-19 pandemic: A review. Phys. Rep. 913, 1–52 (2021).
19. Oliver, N. et al. Mobile phone data for informing public health actions across the covid-19 pandemic life cycle. Sci. Adv.https://
doi.org/10.1126/sciadv.abc0764 (2020).
20. Huang, B. et al. Integrated vaccination and physical distancing interventions to prevent future covid-19 waves in Chinese cities.
Nat. Hum. Behav. 5, 695–705 (2021).
21. Xiong, C., Hu, S., Yang, M., Luo, W. & Zhang, L. Mobile device data reveal the dynamics in a positive relationship between human
mobility and covid-19 infections. Proc. Natl. Acad. Sci. 117, 27087–27089 (2020).
22. Hâncean, M.-G., Slavinec, M. & Perc, M. The impact of human mobility networks on the global spread of covid-19. J. Compl. Netw.
8, 041 (2020).
23. Ribeiro, H. V., Sunahara, A. S., Sutton, J., Perc, M. & Hanley, Q. S. City size and the spreading of covid-19 in Brazil. PLoS ONE 15,
e0239699 (2020).
24. Mazzoli, M. et al. Interplay between mobility, multi-seeding and lockdowns shapes covid-19 local impact. PLoS Comput. Biol. 17,
e1009326 (2021).
25. Lemey, P. et al. Untangling introductions and persistence in covid-19 resurgence in Europe. Nature 595, 713–717 (2021).
26. Google. Mobility Reports. https://www.google.com/covid19/mobility/ (2015).
27. Fraiberger, S. P. et al. Uncovering socioeconomic gaps in mobility reduction during the covid-19 pandemic using location data.
In KDD Humanitarian Mapping 2020 (2020).
28. Schlosser, F. et al. Covid-19 lockdown induces disease-mitigating structural changes in mobility networks. Proc. Natl. Acad. Sci.
117, 32883–32890 (2020).
29. Woskie, L. et al. Early social distancing policies in europe, changes in mobility & covid-19 case trajectories: Insights from spring
2020. PLoS ONE 16, e02530e0253071 (2021).
30. Klein, B. et al. Reshaping a nation: Mobility, commuting, and contact patterns during the covid-19 outbreak. Northeastern Uni-
versity-Network Science Institute Report (2020).
31. Pepe, E. et al. Covid-19 outbreak response, a dataset to assess mobility changes in Italy following national lockdown. Sci. Data 7,
1–7 (2020).
32. Hunter, R. F. et al. Effect of covid-19 response policies on walking behavior in US cities. Nat. Commun. 12(1), 1–9 (2021).
33. Aleta, A. et al. Modelling the impact of testing, contact tracing and household quarantine on second waves of covid-19. Nat. Hum.
Behav. 4, 964–971 (2020).
34. Chang, S. et al. Mobility network models of covid-19 explain inequities and inform reopening. Nature 12, 1–8 (2020).
35. Benzell, S. G., Collis, A. & Nicolaides, C. Rationing social contact during the covid-19 pandemic: Transmission risk and social
benefits of us locations. Proc. Natl. Acad. Sci. 117, 14642–14644 (2020).
36. Cintia, P. et al. The relationship between human mobility and viral transmissibility during the covid-19 epidemics in italy (2020).
arxiv:2006.03141.
37. Hariharan, R. & Toyama, K. Project lachesis: Parsing and modeling location histories. In Geographic Information Science (eds
Egenhofer, M. J. et al.) 106–124 (Springer, 2004). https://doi.org/10.1007/978-3-540-30231-5_8.
38. Ester, M., Kriegel, H.-P., Sander, J. & Xu, X. A density-based algorithm for discovering clusters a density-based algorithm for
discovering clusters in large spatial databases with noise. In Proceedings of the Second International Conference on Knowledge
Discovery and Data Mining, KDD’96, 226–231 (AAAI Press, 1996).
39. OpenStreetMap Contributors. Planet dump. https://planet.osm.org. https://www.openstreetmap.org (2017).
40. Point of interest. (Accessed 21 July 2021); https://en.wikipedia.org/wiki/Point_of_interest.
41. Map features. (Accessed 21 July 2021); https://wiki.openstreetmap.org/wiki/Map_features.
42. Venues categories: Foursquare. (Accessed 21 July 2021); https://developer.foursquare.com/docs/build-with-foursquare/categories/.
43. Foursquare. VisitData. https://visitdata.org/index.html (2017).
44. Di Clemente, R. et al. Sequences of purchases in credit card data reveal lifestyles in urban populations. Nat. Commun. 9, 1–8 (2018).
45. Nevill-Manning, C. G. & Witten, I. H. Identifying hierarchical structure in sequences: A linear-time algorithm. J. Artif. Intell. Res.
7, 67–82 (1997).
46. Vaitla, B. et al. Big Data and the Well-being of women and girls: Applications on the Social Scientific Frontier (Springer, 2017).
47. Tan, P.-N., Steinbach, M. & Kumar, V. Introduction to Data Mining (Addison Wesley, 2005).
48. Rousseeuw, P. J. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. J. Comput. Appl. Math. 20, 53–65
(1987).
49. Petherick, A. et al. A worldwide assessment of covid-19 pandemic-policy fatigue. SSRN 3774252 (2021).
50. Covid-19 caution fatigue: why it happens, and 3 ways to prevent it. Accessed 22 Aug 2021.
51. Gelman, A., Goodrich, B., Gabry, J. & Vehtari, A. R-squared for bayesian regression models. The American Statistician (2019).
52. Vehtari, A., Gelman, A. & Gabry, J. Practical Bayesian model evaluation using leave-one-out cross-validation and waic. Stat.
Comput. 27, 1413–1432 (2017).
53. Fisher, K. A. et al. Community and close contact exposures associated with covid-19 among symptomatic adults 18 years in 11
outpatient health care facilities-united states, July 2020. Morbidity Mortal. Wkly Rep. 69, 1258 (2020).
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 11
Vol.:(0123456789)
www.nature.com/scientificreports/
54. Galmiche, S. et al. Exposures associated with sars-cov-2 infection in France: A nationwide online case-control study. Lancet Reg.
Health Eur. 7, 100148 (2021).
55. Gonzalez, M. C., Hidalgo, C. A. & Barabasi, A.-L. Understanding individual human mobility patterns. Nature 453, 779–782 (2008).
56. Song, C., Qu, Z., Blumm, N. & Barabási, A.-L. Limits of predictability in human mobility. Science 327, 1018–1021 (2010).
57. Pappalardo, L. et al. Returners and explorers dichotomy in human mobility. Nat. Commun. 6, 1–8 (2015).
58. Alessandretti, L., Sapiezynski, P., Sekara, V., Lehmann, S. & Baronchelli, A. Evidence for a conserved quantity in human mobility.
Nat. Hum. Behav. 2, 485–491 (2018).
59. Alessandretti, L., Aslak, U. & Lehmann, S. The scales of human mobility. Nature 587, 402–407 (2020).
60. Schläpfer, M. et al. The universal visitation law of human mobility. Nature 593, 522–527. https://doi.org/10.1038/s41586-021-
03480-9 (2021).
61. Deville, P. et al. Scaling identity connects human mobility and social interactions. Proc. Natl. Acad. Sci. 113, 7047–7052 (2016).
62. Pan, W., Ghoshal, G., Krumme, C., Cebrian, M. & Pentland, A. Urban characteristics attributable to density-driven tie formation.
Nat. Commun. 4, 1–7 (2013).
63. Cuttone, A., Lehmann, S. & González, M. C. Understanding predictability and exploration in human mobility. EPJ Data Sci. 7,
1–17 (2018).
64. Barbosa, H. et al. Human mobility: Models and applications. Phys. Rep. 734, 1–74 (2018).
65. Luca, M., Barlacchi, G., Lepri, B. & Pappalardo, L. A survey on deep learning for human mobility. ACM Comput. Surv. (CSUR) 55,
1–44 (2021).
66. Harvey, N. Behavioral fatigue: Real phenomenon, naïve construct, or policy contrivance?. Front. Psychol. 11, 10 (2020).
67. Michie, S., West, R. & Harvey, N. The concept of “fatigue’’ in tackling covid-19. BMJ 371, 1–10 (2020).
68. Milgram, S. The familiar stranger: An aspect of urban anonymity. The individual in a social world 51–53 (1977).
69. Shapiro, E. New York city public schools to close to slow spread of coronavirus. https://www.nytimes.com/2020/03/15/nyregion/
nyc-schools-closed.html (2020).
70. New York State on Pause. https://coronavirus.health.ny.gov/new-york-state-pause (2020).
71. Cuomo, G. A. M. Governor cuomo issues guidance on essential services under the ‘New York state on pause’ executive order.
https://www.governor.ny.gov/news/governor-cuomo-issues-guidance-essential-services-under-new-york-state-pause-executive-
order (2020).
72. Goettig, M. Phase 2 guidance for employers reopening in New York. https://www.dwt.com/blogs/employment-labor-and-benef
its/2020/06/new-york-state-phase-2-reopening-guidelines (2020).
73. Cuomo, G. A. M. Governor cuomo announces gatherings of up to 25 people will be allowed in phase three of reopening. https://
www.governor.ny.gov/news/governor-cuomo-announces-gatherings-25-people-will-be-allowed-phase-three-reopening-0 (2020).
74. Hale, T. et al. A global panel database of pandemic policies (Oxford covid-19 government response tracker). Nat. Hum. Behav. 5,
529–538 (2021).
75. Di Luzio, M., Johnson, G. L., Daly, C., Eischeid, J. K. & Arnold, J. G. Constructing retrospective gridded daily precipitation and
temperature datasets for the conterminous united states. J. Appl. Meteorol. Climatol. 47, 475–497 (2008).
76. Spiegelhalter, D. J., Best, N. G., Carlin, B. P. & Van Der Linde, A. Bayesian measures of model complexity and fit. J. R. Stat. Soc. B
64, 583–639. https://doi.org/10.1111/1467-9868.00353 (2002).
77. Spiegelhalter, D. J., Best, N. G., Carlin, B. P. & Van der Linde, A. The deviance information criterion: 12 years on. J. R. Stat. Soc. B
76, 485–493 (2014).
Acknowledgements
We would like to thank Cuebiq for allowing the use of anonymized data through their COVID19 Collaborative
program. L.L., S.C. and M.D.N. thank Matteo Saloni for the help in scaling the algorithms seamlessly. L.L., S.C.
and M.D.N. also thank Brennan Lake for the insightful comments on the article. L.P. has been supported by
H2020 SoBigData++ #871042.
Author contributions
S.C., M.D.N., L.L., and L.P. conceived the original idea and planned the experiments. F.P. provided the mobility
data. S.C., M.D.N., and L.L. pre-processed the mobility data, carried out the experiments and made the Figures.
S.C., M.D.N., L.L., and L.P. contributed to the interpretation of the results and wrote the manuscript. S.C., M.D.N.,
L.L., L.P., R.G. and B.L. provided critical feedback, helped shape the manuscript and substantively revised it.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https://doi.org/
10.1038/s41598-021-04139-1.
Correspondence and requests for materials should be addressed to L.L. or M.N.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
© The Author(s) 2021
Scientific Reports | (2021) 11:24452 | https://doi.org/10.1038/s41598-021-04139-1 12
Vol:.(1234567890)
File and source
- File
- living-in-a-pandemic-changes-in-mobility-routines-social-activity-and-adherence-to-covid-19-protective-measures.pdf
- Size
- 2,798,638 bytes
- SHA-256
- 3b227bdf9fbf019eefa9c651bef1ea0ff4d994d25698b69aa834f672b7f03f1f
- Original
- doi.org