Advantages and Difficulties of using Spatial Enablement to Support

Public Health in Cities: The PULSE Case Study

Daniele Pala

, Marica Teresa Rocca

and Vittorio Casella

Department of Electrical, Computer and Biomedical Engineering, University of Pavia, Pavia, Italy

Department of Civil Engineering and Architecture, University of Pavia, Pavia, Italy

Keywords: Public Health, Spatial Enablement, Asthma, Regression, Big Data.

Abstract: Big cities are heterogeneous environments in which socioeconomic and environmental differences among the

neighborhoods are pronounced, therefore research projects that aim at informing public health policies at a

single city level are being developed. Since most of public health data is referred to some geography, spatial

enablement plays a fundamental role when it comes to analysis and visualization of urban health data. The

PULSE project, part of the EU Horizon 2020 framework, involves five cities to transform public health from

a reactive to a predictive system, and promote wellbeing by developing an integrated data ecosystem based

on continuous large-scale collection of information, leading to better-informed data-driven health policy. One

of the goals of PULSE is to apply spatial enablement to generate statistics useful to asses public health at a

high spatial resolution, allowing to organize interventions at a neighborhood level. In this paper, we present

a preliminary spatial enablement study carried out in this context, in which we show opposite sides of its

application: while the results are promising, the lack of standardization and protocols in the data collection

and representation processes make spatial enablement very difficult to apply to open data.

1 INTRODUCTION

The percentage of the world’s population living in

urban areas is projected to increase from 54% in 2015

to 60% in 2030 and to 66% by 2050 (United Nations

and Department of Economic and Social Affairs,

2014). It is important to acknowledge that big cities

are perfect labs for innovation aiming at managing

demographic and epidemiological transitions (WHO,

2016). Big cities are heterogeneous environments

where social, environmental and demographic

conditions can vary significantly within relatively

small distances. For this reason, studies aiming at

improving health and wellbeing in the urban areas

have to address the problem at a neighborhood level,

taking into account the underlying spatial variability.

In line with this principle, the international project

named Participatory Urban Living for Sustainable

Environments (PULSE) has been funded by the EU

Commission under the Horizon 2020 framework to

undertake research and innovation in big cities in

Europe, the United States and Asia. PULSE is

partnering with five important cities – Barcelona,

Birmingham, New York, Paris and Singapore – and

has two main focuses: the link between air quality and

asthma, and the one between physical inactivity and

type 2 diabetes. PULSE aims at providing effective

solutions to prevent and treat these diseases through

an innovative data integration platform, where data

will be collected directly from the users/citizens,

enrolled in each of the five cities, through a mobile

App, and from open data sources and air quality

sensors. PULSE aims also at involving public health

authorities directly. Besides the App, the PULSE

system features also an innovative WebGIS that

allows data visualization, a Decision Support System

that allows to analyze the data, runs predictive models

and sends notifications and advice directly to the

users, and dashboards to help public health authorities

visualize the situation in the city and design proper

interventions.

One of the key features of the PULSE system, is

the so-called spatial enablement, i.e. the addition of a

spatial description to a dataset and/or an analysis

procedure. Most of the data regarding public health

has a natural spatial reference, since demographic

data is collected considering areas of residence of the

population, and environmental measurements clearly

depend on the geographical zone they are referred to.

Although the concept of spatial enablement is not

322

Pala, D., Rocca, M. and Casella, V.

Advantages and Difﬁculties of using Spatial Enablement to Support Public Health in Cities: The PULSE Case Study.

DOI: 10.5220/0007900003220329

In Proceedings of the 5th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2019), pages 322-329

ISBN: 978-989-758-371-1

new, studies that address this topic in urban

environments, breaking down the problem to a

neighborhood wellbeing study, are not common in

literature. Some studies focusing on urban areas have

been carried out, but they generally analyze the whole

city or use broad spatial subdivisions. In the PULSE

context, we seek to study public health problems in

urban contexts at a fine spatial resolution, considering

all the characteristics of each single neighborhood.

Spatial enablement methods are promising both

for analysis and visualization matters, but their

application is generally quite complicated due to a

diffuse lack of standards and regularization in the data

collection process. Data regarding demographic,

socioeconomics, environment and air pollution, when

available, is often collected by different public and

private entities, that apply different collection

procedures and storage standards, making it long and

uneasy to retrieve and process all the data.

In this paper, we present as case study a series of

analyses we carried out within the PULSE project

using almost uniquely open data, in particular we

applied a spatial enabled method called

Geographically Weighted Regression (GWR) to a

combination of datasets referred to New York City,

with the aim of investigating the link between asthma

hospitalizations and several socioeconomic and

environmental factors.

After a brief presentation of the methodology and

the results, explained in detail in another paper that is

currently under review, we focus also on the

difficulties that we encountered during our analyses,

highlighting the need of a better-defined system in the

data collection and storage processes in the public

health environment.

2 MATERIALS AND DATASETS

PULSE is characterized by a complex architecture

that allows an intense data flow through several

different integrated systems. The main components of

this architecture are:

 The Pulsair App for smartphone, through which

users can send their data and position, and receive

personalized feedbacks concerning their

condition in relation to the situation in their city;

 Backend analytics and a Decision Support

System, that apply big data methods to analyze the

input data and use predictive risk models, in order

to eventually generate feedbacks for the users;

 Dashboards that allow the public health policy

makers to inspect the situation in different neigh-

borhoods and organize proper interventions;

 A large and innovative WebGIS that allows to

visualize all the data on maps and quickly spot the

main features and criticalities in the studied cities.

Since the geographical description of health-related

phenomena is at the base of PULSE, the WebGIS

could be considered the most interesting architecture

element in the project, as it collects and integrates a

large wealth of spatially-enabled data.

In line with the PULSE principle, and to start

investigating its applications and extensions, we

carried out a preliminary spatial enablement study

using some open data currently integrated in the

PULSE WebGIS.

2.1 A Data Integration Example: New

York City

While the PULSE system is still in a development

phase and the WebGIS is expected to be complete by

the spring of 2019, a lot of data integration, modeling

and analysis is already being carried out with data

coming from the five cities. In particular, thanks to its

peculiar data availability, we developed a large

WebGIS prototype of New York City, and performed

some preliminary analyses on it, in order to

demonstrate the importance of spatial enablement in

studying public health in cities and the usefulness and

innovation of PULSE.

Several sources of data have been used to carry

out the analyses reported in this paper. Most of the

data has been kindly provided to the PULSE

consortium by The New York Academy of Medicine.

We used socioeconomic data freely available in the

NYC Neighborhood Health Atlas website (“New

York City Neighborhood Health Atlas,” n.d.), from

which it has been downloaded. The hospitalization

and ED visit rates data, as well as the PM2.5 historical

data, has been downloaded from the NYC

Environmental & Health Data Portal(“Environment

& Health Data Portal,” n.d.). Information regarding

age and race of hospitalized people has been acquired

from the SPARCS(“Statewide Planning and Research

Cooperative System,” n.d.) limited 2014 dataset.

2.2 Geographically Weighted

Regression

The collected datasets were analyzed through

Geographically Weighted Regression (GWR)

(McMillen, 2004), that is a linear regression model

with the addition of a weight that provides a spatial

description.

Advantages and Difﬁculties of using Spatial Enablement to Support Public Health in Cities: The PULSE Case Study

323

Given a dependent variable  and one or more

explanatory one, 



, known for an adequate number 

of observations, which can be represented by points





,



, a regular linear regression is characterized by

a set of equations that can be represented in the vector

formalism







(1)

where 



contains the actual measurements of  and

the vector  is composed by the unknown

coefficients 



. Its estimation is usually performed by

a minimum problem; the minimized quantity is the

squared norm of the difference between the observed

values of 



and those given by the model:

≔





















(2)

The depicted solution can be generalized by

introducing a weight, thus giving each observation a

different relevance

≔























(3)

GWR uses the above defined method to take into

consideration spatial variability. In a common GIS

layer with polygon representation of an environment,

each polygon corresponds to an observation, located

in its centroid. The studied area can be overlapped

with a set of regularly-spaced dots. For each dot, a

distinct regression is calculated, in which the

observed values for the dependent and explanatory

variables are the same, but the weights change. In our

case, the weight function is



















(4)

where 



is the distance between the considered dot

and the i-th centroid and s is a threshold,

corresponding in our case to 5 km. In our study, we

overlapped to the NYC map a grid of points distant 1

km from each other.

3 RESULTS

As mentioned in section 1, the complete results of our

spatial enablement study are presented in another

paper, currently under final review. In this section

however, we present a brief extract of the results to

show the power of spatial enablement and how

promising these methods are.

In our study, we investigated the relations among

asthma hospitalizations and several socioeconomic,

demographic and environmental factors in the

different neighborhoods of New York City. The

factors we considered were PM2.5 and ozone

concentration, percentage of land used for industrial

activities, poverty rate, race, age, medical insurance

and garbage recycling. Both univariate and

multivariate analyses have been carried out.

3.1 Univariate Analysis

In the first step of our analysis, we considered each

variable individually and ran the geographic

regression model to inspect its relation with the

hospitalization rate, and how this varies in the

different zones of the city.

As an example, figure 1 reports the result of this

algorithm applied using poverty rate as covariate,

showing a map of the 



coefficients (left side) and

one of the 



scores (right side) in each point in

which the regression is calculated. It can be easily

noticed that, although there is general positive

relation between high poverty and higher

hospitalization rates, this relation changes throughout

the city, sometimes importantly. In this example, as

in most of the poorer areas of the city (i.e, the Bronx

and the inner border between Brooklyn and Queens)

the hospitalization rate is high, there are some areas

in which the relation is not as strong, as it becomes

even unreliable in south-west Brooklyn and East

Staten island. This highlights the importance of

studying public health problems at a granular spatial

resolution, in order to spot all the possible relations.

We obtained similar results testing also all the

other covariates. As a general rule, socioeconomic

and demographic variables appear to be more related

to hospitalizations than air pollution, but this could be

due to the lack of measurements taken in a proper

time frame, that forced us to use averaged data. All

factors anyway showed a different level of influence

in different areas of the city.

3.2 Multivariate Analysis

We also performed GWR testing several covariates at

the same time in a multivariate model, that has the

advantage of being able to ease the comparison of the

effects of different variables, allowing to spot

outliers, confounders and relations between different

factors. In this section, we show an example of

multivariate GWR that combines poverty rate and

percentage of people identifying as Black and

Hispanic. The underlying model can be described by

the following equation, valid for each point where the

GWR is performed:

 







∗



∗%



∗%

HGIS 2019 - Special Session on GIS in Healthcare

324

Figure 1: Results of the GWR algorithm using the poverty rate as covariate. On the left side a map of the 



coefficients

calculated in all the points in which the regression is performed is visible; on the right side, a map of the 



scores is shown.

Where  stands for “hospitalizations”, 

stands for “poverty” and  and  for “Black” and

“Hispanic” respectively.

Race/ethnicity is known to be related to asthma

(Litonjua et al., 1999). Several studies suggest that

Black and Hispanic residents in the USA tend to

suffer from higher rates of asthma as they live in

poorer areas close to industrial sites and large

highways (Clark et al., 2017).

Results are visible in figure 2. On the left side of

the image, maps of the β coefficients are shown,

whereas panels in the right side show correspondent

significance maps based on the t-statistic values. In

detail, we created 3 significance levels: Non-

Significant (NS), Partially Significant (PS),

Significant (S), based on the t-statistic threshold

values 1.96 (5% confidence level) and 2.58 (1%

confidence level). Several interesting phenomena can

be noticed:

 



is extremely high in all the region (> 0.7, image

not shown), therefore the linear model is globally

reliable;

 Considering poverty and percentage of Black

population, the correspondent  are always

positive, indicating a positive correlation between

either of these factors and the hospitalization rate;

 In general, the higher the , the higher the level of

significance. Therefore, in the neighborhoods in

which we found that high variables’ levels lead to

high hospitalization rates, the found relations are

significant;

Low significance levels could be due to the effect of

other confounding variables and to a smaller quantity

of data available. For instance, looking at the

percentage of Hispanic people (image not shown), it

can be noticed that most of the Hispanic population is

concentrated in the Bronx, Upper Manhattan

(Harlem, East Harlem and Washington Heights), in

central and west Queens and some areas of east

Brooklyn (Bushwick and south of Highland Park),

plus some isolated spots in west Brooklyn (Sunset

Park) and north Staten Island. Apart from these last

isolated spots, in the same areas in which the

concentration is higher, also the significance of the

correspondent beta is high. Hence lower significance

corresponds to higher scarcity of data.

These results show that even multivariate

geographical analysis can be helpful to describe and

visualize important public health phenomena and

discover the relations among different factors.

4 COMPLICATIONS

As promising as spatial enablement is, its application

to an urban health problem is not easy. Large cities

such as New York are sources of enormous quantities

of heterogeneous data, that are by definition difficult

to process as they differ in typology, dimension,

scale, collection method etc.

Moreover, public health agencies have not yet

defined standards and harmonization methods to ease

the process of analyzing the huge quantity of data

collected. Within PULSE, we are experiencing longer

than expected processing times for all the cities due

to the same problem. In this section we present a case

study upon NYC data used for our GWR study, that

shows some examples of the typical problems

encountered during these kinds of protocols.

Advantages and Difﬁculties of using Spatial Enablement to Support Public Health in Cities: The PULSE Case Study

325

Figure 2: Results of the multivariate model. On the left side, the regression coefficients; on the right side, the correspondent

significance level, based on the confidence intervals.

4.1 Conventional Spatial Subdivisions

New York City has a very large amount of open data

available regarding all different aspects of the

population and the environment. Other data is

available upon request and/or purchase to the State’s

public health departments.

Public health data is collected with reference to

several different spatial subdivisions in NYC, that

are:

 Boroughs: 5 polygons correspondent to the main

HGIS 2019 - Special Session on GIS in Healthcare

326

districts of the city (Bronx, Brooklyn, Manhattan,

Queens, Staten Island);

 UHF34 (United Hospital Fund), with 34

polygons;

 UHF42, 42 polygons;

 CD55 (Community Districts), with 55 polygons;

 CD59, 59 polygons;

 CD71, 71 polygons;

 PUMA (Public Use Microdata Areas) 55

polygons not overlapping with the CD55

subdivision;

 NTA (Neighborhood Tabulation Areas), 195

polygons;

 ZIP Codes, 262 polygons.

Once imported in a GIS software and visualized on

maps, none of the polygons in one subdivision has

vertices that can be overlapped to the ones of any

other subdivision. This poses a problem when it

comes to analyze data from multiple sources, since

different kinds of data are collected referred to

different subdivisions. A harmonization algorithm is

therefore necessary. For instance, in our GWR

application we decided to adopt the UHF42

subdivision, as most of the data was already available

in this spatial description. All the data available in

another tessellation was transformed through the

following algorithm: let’s consider a polygonal

subdivision for which a certain variant is available,

for each polygon 



, the variant value v



is known. If

we consider another polygon 



, belonging a

different subdivision, in general it won’t coincide

with any 



and, instead, will overlap to several of

them. The estimated v



can be obtained by the

weighted sum







∑









∑





(5)

where 



is the area of the intersection between 



and 



and the summation is only performed over the

polygons for which the intersection is non-empty.

This means that the value of a certain phenomenon in

any spatial subdivision can be represented in the

UHF42 one constructing each polygon as the sum of

the same value in the other subdivision, weighted for

the overlapping area.

4.2 Geometric Consistency

Most of the public health data, both openly available

and provided by the public health agencies is

generally in tabular form rather than in shapefiles,

making it necessary to create the shapefiles by joining

the tables with pre-existing polygons. In the NYC

open data, however, several indicators are already

represented in shapefiles. In our study, and during the

acquisition of data to be integrated in the PULSE

WebGIS and system, we used several shapefiles, and

found that lots of them contain errors or imprecisions.

For example, taking two different shapefiles

representing data on the same spatial subdivision and

overlapping them, we could notice that some of the

polygons’ border didn’t overlap properly, showing an

offset between the two shapes in some areas of the

map. Figures 3 and 4 show some examples.

Figure 3: Detail of two shapefiles representing the same

area of the city. It can be noticed that some polygons have

a visible offset with the same polygons of other shapefiles.

Figure 4: Another clear example of two shapefiles not

properly overlapping. In this case, some little islands in the

Rockaway area are entirely absent from one of the two

shapes.

A similar problem was found in relation to some

Advantages and Difﬁculties of using Spatial Enablement to Support Public Health in Cities: The PULSE Case Study

327

uninhabited areas of the city, such as Central Park and

JFK and La Guardia airports, that were marked as

holes in some shapefiles, and represented as polygons

in others. To overcome these multiple problems, it

was necessary to manually modify the shape of the

polygons, creating a serious waste of time and

resources. This phenomenon is shown in Figure 5.

Figure 5: In this overlap of two shapefiles, it can be clearly

seen that Central Park and the areas with the two airports

within the city border (JFK and La Guardia) are treated as

holes in one of the shapes, but not in the other.

4.3 Tabular Data Problems

Besides all the difficulties encountered with the

shapefiles, also the more common tabular data,

typically found in xls or csv formats, had some issues

we had to deal with. Figure 6 shows an extract of a

typical table representing health data in NYC. Several

things can be noticed, for example the key “1”, used

to indicate the geographical area of reference and

necessary to match this table with other tables, is used

both referred to the whole city and to the borough of

the Bronx. Moreover, in this table the comma sign is

used in multiple different ways, as it simultaneously

represents the thousands separator, the column

delimiter and the item delimiter inside the brackets

showing confidence intervals.

Figure 6: Example of tabular data found in public health

open data sources. The way data is represented makes the

import and processing part very complicated.

Such features are confusing and make it very

difficult to import the data and analyze the table with

a software. The difficulties rise considering that in the

integration of several health data sources, a researcher

usually has to deal with a large number of tables not

different from the one shown in this example.

5 DISCUSSION

With the continuous development of sophisticated big

data analytics and the technological progress, both the

quantity of data that can be collected and the

collection velocity are rapidly increasing in several

applications of the medical research. One of the most

rapidly developing fields of research is indubitably

exposomics, i.e. the study of the combination of

factors to which an individual is exposed during a

certain amount of time that can cause a change in

his/her phenotype regarding a specific pathology.

During our life, we are exposed to an enormous

quantity of phenomena that can influence our health.

Since exposure depends on location, spatial

enablement is gradually becoming necessary in

public health analyses.

In this paper, we briefly presented the PULSE

project and explained how it intends to apply spatial

enablement in its most innovative form, then we

showed a case study, carried out within PULSE, in

which we applied a spatial analysis technique that

allowed us to show how public health could benefit

from spatial enablement in big cities. In our study we

considered asthma hospitalizations as an easy to

monitor outcome of a series of exposures which every

citizen is exposed to involuntarily, and we aimed at

looking at how these factors could influence the

impact of a common disease such as asthma. The

application of these techniques in the public health

field allows to increase the awareness of the health

problems in cities and therefore to take better

informed decisions on how to intervene, making

interventions more targeted and leading to a cost

reduction for public improvements and healthcare

assistance.

The final aim of PULSE is to create a protocol,

tested in five cities but eventually extendable to

potentially any city, in which public health policy

makers are supported by specific entities, called

Public Health Observatories (PHO), that help them in

the data interpretation and decision-making

processes.

PULSE has been funded by the European

Commission, but involves several institutions from

all the world with the common aim of promoting

HGIS 2019 - Special Session on GIS in Healthcare

328

public health and wellbeing by creating a system that

can be extended in ideally all the big cities. The

system is being tested in 7 cities – Barcelona,

Birmingham, Paris, Pavia, New York, Keelung and

Singapore – and data is being collected in those cities

both from the system itself and from outside sources,

thanks to the cooperation with the local public health

authorities, that will help providing data and

deploying the system in a way that makes it adaptable

to each city’s environment and legal limitations.

Unfortunately, the capability of quickly collect

and analyze rising quantities of data is developing

without enough regularization and awareness from

the public health authorities, leading to a constantly

increasing quantity of unorganized, chaotic data,

difficult to import and analyze without accurate

supervision. What we showed in this paper is just a

sample use case of a diffuse problem in big data

studies with public health focuses. Data

harmonization and preprocessing are often time-

consuming tasks, and this could be changed with the

introduction of standardization protocols in the data

collection within the same areas.

Despite municipalities and public organizations

in large cities are moving forward in the right

direction in increasing the quantities of spatially

enabled data collected and making it more easily

available to the public and the scientific community,

more effort should be put in defining protocols that

can allow an easier use of the data.

6 CONCLUSIONS

Spatial enablement is the development of techniques

that add a spatial description to datasets and analysis

tools. Thanks to these techniques, public health

problems can be better addressed thanks to the

detailed knowledge of the location of environmental

and social exposures, reducing time and costs of

interventions. Unfortunately, the use and spreading of

these methods is slowed down by a lack of efficient

data collection and representation protocols, that

leads to a dilation of processing and analyses time,

and a consequent waste of resources.

ACKNOWLEDGEMENTS

The authors wish to acknowledge Elisa Fisher, Foram

Jasani, Kumbie Madondo and José Pagàn from The

New York Academy of Medicine for providing part

of the data and supporting the analyses; and the

College of Global Public Health at the New York

University for allowing international students

exchange that facilitated the workflow.

The PULSE project has been funded by the

European Commission Horizon 2020 framework

program under grant GA-727816.

REFERENCES

Clark, L.P., Millet, D.B., Marshall, J.D., 2017. Changes in

Transportation-Related Air Pollution Exposures by

Race-Ethnicity and Socioeconomic Status: Outdoor

Nitrogen Dioxide in the United States in 2000 and 2010.

Environ. Health Perspect. 125, 097012.

https://doi.org/10.1289/EHP959

Environment & Health Data Portal [WWW Document],

n.d. URL http://a816-dohbesp.nyc.gov/IndicatorPublic/

PublicTracking.aspx?theme_code=2,3&subtopic_id=1

1 (accessed 2.28.18).

Exposomics: mathematics meets biology. - PubMed - NCBI

[WWW Document], n.d. URL https://www.ncbi.

nlm.nih.gov/pubmed/26371206 (accessed 3.7.19).

Litonjua, A.A., Carey, V.J., Weiss, S.T., Gold, D.R., 1999.

Race, socioeconomic factors, and area of residence are

associated with asthma prevalence. Pediatr. Pulmonol.

28, 394–401. https://doi.org/10.1002/(SICI)1099-

0496(199912)28:6<394::AID-PPUL2>3.0.CO;2-6

McMillen, D.P., 2004. Geographically Weighted

Regression: The Analysis of Spatially Varying

Relationships. Am. J. Agric. Econ. 86, 554–556.

https://doi.org/10.1111/j.0002-9092.2004.600_2.x

New York City Neighborhood Health Atlas [WWW

Document], n.d.. Tableau Softw. URL

https://public.tableau.com/views/NewYorkCityNeighb

orhoodHealthAtlas/Home?%3Aembed=y&%3Ashow

VizHome=no&%3Adisplay_count=y&%3Adisplay_st

atic_image=y&%3AbootstrapWhenNotified=true

(accessed 2.27.18).

Statewide Planning and Research Cooperative System

[WWW Document], n.d. URL https://www.health.ny.

gov/statistics/sparcs/ (accessed 11.13.18).

United Nations, Department of Economic and Social

Affairs, 2014. World urbanization prospects, the 2014

revision: highlights.

World Health Organization, UN-Habitat, 2016. Global

report on urban health: equitable healthier cities for

sustainable development. World Health Organization.

Advantages and Difﬁculties of using Spatial Enablement to Support Public Health in Cities: The PULSE Case Study

329