University of Southern California Dissertations and Theses

Providing a new low-cost primary care facility for under-served communities: a site suitability analysis for Service Planning Area 6 in Los Angeles County, California

(USC Thesis Other)

Providing a new low-cost primary care facility for under-served communities: a site suitability analysis for Service Planning Area 6 in Los Angeles County, California

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Providing A New Low-Cost Primary Care Facility for Under-Served Communities: A Site Suitability
Analysis for Service Planning Area 6 in Los Angeles County, California

by

Ada Yue Li Sarain

A Thesis Presented to the
Faculty of the USC Graduate School
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
Master of Science
(Geographic Information Science and Technology)

August 2019

Copyright © 2019 by Ada Yue Li Sarain

To my family

iv
Table of Contents
List of Figures ........................................................................................................................... vi
List of Tables........................................................................................................................... viii
Acknowledgements ................................................................................................................... ix
List of Abbreviations .................................................................................................................. x
Abstract ..................................................................................................................................... xi
Chapter 1 Introduction ................................................................................................................ 1
1.1. Motivation and Study Area ............................................................................................ 2
1.1.1. South Los Angeles................................................................................................. 3
1.1.2. Service Planning Area 6 in Los Angeles County .................................................... 5
1.2. Research Question ......................................................................................................... 7
1.3. Structure of Thesis ......................................................................................................... 8
Chapter 2 Literature Review ....................................................................................................... 9
2.1. Primary Care .................................................................................................................. 9
2.1.1. Primary Care ....................................................................................................... 10
2.1.2. Primary Care and Low-Income Populations ......................................................... 11
2.2. Site Selection for Primary Care Facilities ..................................................................... 14
2.2.1. Site Selection Criteria for Health Care Facilities .................................................. 14
2.2.2. Weighted Overlay and Fuzzy Overlay ................................................................. 20
2.3. Evaluating Accessibility ............................................................................................... 23
2.3.1. Earlier Methods to Measure Accessibility ............................................................ 24
2.3.2. 2-Step Floating Catchment Area .......................................................................... 26
2.3.3. Public Transit in Measuring Accessibility ............................................................ 32
Chapter 3 Data and Methodology ............................................................................................. 34
3.1. Data Sources and Data Preparation ............................................................................... 35
v
3.1.1. SPA Boundaries .................................................................................................. 36
3.1.2. Low-Income Populations ..................................................................................... 37
3.1.3. Zoning and Parcel Dataset ................................................................................... 39
3.1.4. Public Transit Dataset and Road Network Dataset ............................................... 40
3.1.5. Existing Low-Cost Primary Care Facilities .......................................................... 42
3.2. Site Selection Analysis ................................................................................................. 45
3.2.1. Filtering Parcels .................................................................................................. 46
3.2.2. Site Selection Method .......................................................................................... 47
3.2.3. Fuzzy Membership Procedures ............................................................................ 48
3.2.4. Fuzzy Overlay Analysis ....................................................................................... 51
3.3. Candidate Site Evaluation with 2SFCA ........................................................................ 52
Chapter 4 Results ...................................................................................................................... 56
4.1. Initial Filtering ............................................................................................................. 57
4.2. Fuzzy Overlay Analysis ............................................................................................... 58
4.3. 2SFCA Analysis and Recommended Sites ................................................................... 65
4.3.1. Catchment Area ................................................................................................... 65
4.3.2. Accessibility Scores ............................................................................................ 70
Chapter 5 Discussion and Conclusion ....................................................................................... 75
5.1. Discussion.................................................................................................................... 75
5.1.1. Overall Assessment of Methods and Analysis ...................................................... 75
5.1.2. Discussion of the Final Results ............................................................................ 78
5.2. Limitations and Improvements ..................................................................................... 81
5.3. Future Research ........................................................................................................... 83
5.4. Conclusion ................................................................................................................... 85
References ................................................................................................................................ 87
vi

List of Figures
Figure 1. SPA6 and South Los Angeles ...................................................................................... 3
Figure 2. Service Planning Areas of Los Angeles County ........................................................... 5
Figure 3. Low-Income Population by Service Planning Area ...................................................... 6
Figure 4. Primary Care and Low-cost Primary Care Provision by Service Planning Area ............ 7
Figure 5. Illustration of the First Step in 2SFCA ....................................................................... 29
Figure 6. Illustration of the Second Step in 2SFCA ................................................................... 31
Figure 7. Extended Study Area ................................................................................................. 37
Figure 8. Low-Income Populations by Census Tract ................................................................. 39
Figure 9. Parcel Data with Zoning Information ......................................................................... 40
Figure 10. Public Transit Routes and Stops in the Extended Study Area ................................... 41
Figure 11. Road Network in the Extended Study Area .............................................................. 42
Figure 12. Location and Service Capacity of Existing Low-Cost Primary Care Facilities in the
Extended Study Area ................................................................................................................ 45
Figure 13. Workflow for the Proximity to Public Transit Fuzzy Membership Layer .................. 48
Figure 14. Workflow for the Distance to Existing Low-Cost Primary Care Facilities Fuzzy
Membership Layer.................................................................................................................... 49
Figure 15. Workflow for the Land Cost Fuzzy Membership Layer ............................................ 50
Figure 16. Workflow for the Low-Income Population Layer ..................................................... 51
Figure 17. Fuzzy Overlay Workflow ......................................................................................... 52
Figure 18. 2SFCA Workflow .................................................................................................... 53
Figure 19. Candidate Sites After the Initial Filtering Process .................................................... 57
vii
Figure 20. Fuzzy Membership Values Based on Distance to the Closest Existing Low-cost
Primary Care Facility ................................................................................................................ 59
Figure 21. Fuzzy Membership Values Based on Distance to the Closest Public Transit Stop ..... 60
Figure 22. Fuzzy Membership Values Based on Land Value Per Square Foot ........................... 61
Figure 23. Fuzzy Membership Values Based on the Estimated Number of Low-Income Residents
within 1 Mile from Candidate Sites........................................................................................... 62
Figure 24. Final Fuzzy Membership Values for Candidate Sites ............................................... 63
Figure 25. Final Candidate Sites ............................................................................................... 65
Figure 26. 30-Minute Travel Time Areas Via Private Vehicle and Public Transit for Existing
Low-Cost Primary Care Facility 27 .......................................................................................... 67
Figure 27. 30-Minute Travel Time Areas Via Private Vehicle and Public Transit for Final
Candidate Site 12 ...................................................................................................................... 68
Figure 28. 30-Minute Travel Time Areas Via Private Vehicle and Public Transit for Census
Tract 2184 ................................................................................................................................ 69
Figure 29. Catchment Areas for Census Tract 2184 and Final Candidate Site 12 ....................... 70
Figure 30. Current Low-Cost Primary Care Accessibility Scores by Census Tract in SPA6 ...... 72
Figure 31. Large-scale Map of Final Candidate site 12.............................................................. 79
Figure 32. Satellite Image of Final Candidate site 12 ................................................................ 80
Figure 33. Photo of Final Candidate site 12 .............................................................................. 81

viii
List of Tables
Table 1. Disparities in Health care Resources and Other Basic Resources: South Los Angeles
Compared to the Los Angeles County Average ........................................................................... 4
Table 2. Summary of Zoning Regulations ................................................................................. 15
Table 3. Data Sources ............................................................................................................... 35
Table 4. Federal Poverty Thresholds for Households of Specified Sizes.................................... 38
Table 5. Short List of Candidate Sites with Fuzzy Membership Values of 0.9 and Above ......... 64
Table 6. Estimated Weekly Service Provision Capacity of the Final Candidate Sites................. 71
Table 7. Summary Statistics of the Low-Cost Primary Care Accessibility Scores...................... 73

ix
Acknowledgements
I am grateful to my thesis committee members, Drs. Elisabeth J. Sedano, Robert O. Vos, and
Steven D. Fleming, without whose guidance this project would not have been possible.
Moreover, I owe a great debt of gratitude to my doctoral advisor Professor Jefferey M. Sellers
for supporting me to explore interdisciplinary research and to pursue GIS as one of my fields of
concentrations for my Ph.D. studies in Political Science.
Great thanks are owed to my friends, colleagues, and the USC community. I would like
to acknowledge Kendrick Watson and Veridiana Chavarin for making life as a graduate student
pursuing two degrees much easier. I would like to thank Mingmin Yang for encouraging me to
step out of my comfort zone and study spatial sciences. I am also thankful to all my friends and
colleagues for providing a robust network of support during my time at USC.
Lastly, infinite thanks to my family, especially to my wife, for being the anchor in my
life; to my parents, for their unconditional support throughout the years; and to my grandparents,
for believing in me no matter what. My family’s love, support, and sacrifice made this thesis
possible. Words cannot express how grateful I am to them. Special thanks to Serena and Amelia,
who provided a lot of emotional support in the writing of this thesis.

x
List of Abbreviations
2SFCA 2-Step Floating Catchment Area
GIS Geographic information system
GTFS General Transit Feed Specification
MAUP Modified Areal Unit Problem
OECD Organization for Economic Cooperation and Development
SPA Service Planning Area
SPM Supplemental Poverty Measure

xi
Abstract
Primary care is crucial for both individual and public health outcomes. However, access to
primary care remains insufficient for low-income populations even in developed countries like
the United States. Striving to contribute to tackling this problem, this thesis provides a site
suitability analysis for a new low-cost primary care facility in Service Planning Area 6 (SPA6) of
Los Angeles County, a significantly under-served area that largely coincides with South Los
Angeles. This thesis first employs fuzzy overlay analysis to evaluate candidate sites with a series
of criteria including proximity to public transit, distance from existing low-income primary care
facilities, appropriate zoning, an empty or under-utilized parcel, relatively cheap land cost, and
relatively large low-income population. This thesis evaluates a short list of candidate sites
resulting from the fuzzy overlay analysis by calculating their impacts on the primary care
accessibility scores of each census tract in SPA6 using the 2-Step Floating Catchment Area
method (2SFCA). The 2SFCA method quantitatively assesses primary care accessibility using
floating catchment areas calculated based on travel time at population points and primary care
provider points. To account for relatively low car-ownership rates among low-income
populations, this thesis adopts a novel approach that defines the floating catchment area as the
intersection of the two catchment areas: one defined by a 30-minute travel time via public transit
and one defined by a 30-minute travel time via private vehicle. The final results from the fuzzy
overlay and the 2SFCA analyses provide a list of suitable candidate sites with various
geographical and physical attributes ranked by their accessibility scores, providing an
informative, flexible, and intuitive guideline that caters to the different needs of potential
decision makers looking for a site for a new low-cost primary care facility that would improve
primary care accessibility in SPA6.

1
Chapter 1 Introduction
Access to affordable primary medical care should be a universal human right. However, the
reality of primary care provision and accessibility is less than ideal, especially for low-income
communities. The communities collectively known as South Los Angeles face a variety of
unfortunately common inner-city problems including inadequate provision of public services.
Primary health care is the entry point to the health care system; it provides essential curative and
preventive care to the public, and is thus crucial for both individual health and public health
conditions. The Los Angeles County Department of Health Services divides the county into eight
management regions, called Service Planning Areas (SPAs). South Los Angeles largely
coincides with the Department of Health’s SPA6. Department of Health data reveals that SPA6
suffers from a severe shortage of primary care provision in general and low-cost primary care
provision in particular.
Los Angeles County, private health care providers, and community service nonprofits
have been striving to improve affordable primary care provision for the communities within
SPA6. This thesis supports this work by identifying a suitable site for a new low-cost primary
care facility in SPA6. Since low-cost primary care provision usually depends at least partially on
public funding, it is crucial to carefully select a site for a new affordable primary care facility so
that it could serve as many residents in need with as little tax payer money as possible.
Therefore, this thesis is significant as it provides suggestions for selecting a suitable site for a
new primary care facility that takes into consideration the socioeconomic conditions of SPA6
and maximizes the improvement of affordable primary care access with limited resources. This
result of this project could be implemented by Los Angeles County government or private

2
entities that are working to improve affordable primary care for the communities in South Los
Angeles.
In addition to the practical contribution discussed above, this thesis contributes to the
literature on site suitability analysis for primary care facilities with a novel methodological
approach that incorporates a quantitative evaluation step to assess the candidate sites. This thesis
adopts fuzzy overlay analysis to rank all candidate sites based on a set of general site selection
criteria for primary care facilities as the first step of the site suitability analysis. This thesis then
evaluates the top candidate sites by calculating the impacts of each on the primary care
accessibility using the 2-Step Floating Catchment Area (2SFCA) method. The addition of the
2SFCA method adds a level of depth to the site suitability analysis because it verifies the
suitability of the candidate sites by providing intuitive accessibility scores generated from
empirical data, and it makes this approach a novel addition to the scholarly literature.
1.1. Motivation and Study Area
Los Angeles County is a large county both in terms of geographical size and population.
Los Angeles County Department of Health Services divides the county into eight SPAs in order
to better develop and provide health care services based on the needs of residents in different
areas. The study area of this thesis is SPA6 and surrounding communities. SPA6 includes the
communities of Athens, Compton, Crenshaw, Florence, Hyde Park, Lynwood, Paramount, and
Watts, which are also collectively known as South Los Angeles. This section describes the
socioeconomic level of South Los Angeles and then compares the socioeconomics and health
care provision in SPA6.

3
1.1.1. South Los Angeles
Despite the fact every human being should have access to health care, the lack of access
to affordable primary care is severe in SPA6. As one of the eight SPAs of Los Angeles County,
SPA6 geographically coincides with South Los Angeles. Figure 1 below shows SPA6 and South
Los Angeles.

Figure 1. SPA6 and South Los Angeles
In comparison to the county average, South Los Angeles has fewer health care facilities,
a smaller health care workforce, and less health care financing, health care coverage, and
primary care access (Park, Watson, and Galloway-Gilliam 2008). Furthermore, South Los
Angeles is below the county average in terms of nutrition, physical activity options, public
safety, housing, and education (Park, Watson, and Galloway-Gilliam 2008). Table 1 below

4
shows the disparities in health care resources and other basic resources in South Los Angeles in
comparison with the county average.
Table 1. Disparities in Health care Resources and Other Basic Resources: South Los Angeles
Compared to the Los Angeles County Average
Types of
Resources
Resource Items
South LA (Percent Difference Compared to
LA County Average)
General health
care
Health care facilities -28%
Health care workforce -76%
Health care financing -65%
Health care coverage -30%
Primary care Primary care access -34%
Primary care utilization -24%
Percent of adults who reported
having a regular source of care
-7%
Percent of adults who reported easily
obtaining medical care
-18%
Percent of adults who could not
afford dental care at least once in the
past 12 months
-36%
Percent of households with vehicle -68%
Percent of ER hours spent in
diversion a year
-90%
ER visits that leave without being
seen per 1,000 population
-12%
Other Basic
Resources
Nutrition -106%
Physical Activity Options -55%
Public Safety -17%
Housing -40%
Education -43%
Data Source: Park, Watson, and Galloway-Gilliam 2008

5
1.1.2. Service Planning Area 6 in Los Angeles County
Health care provision for communities of South Los Angeles is managed at the county
level by the Los Angeles Department of Health SPA 6. As noted, the study area for this project is
SPA6. The analysis herein also considers the neighboring SPAs to SPA6 to avoid edge effects
where SPA6 residents may prefer to travel across SPA boundaries to seek health care. Figure 2
below shows the eight SPAs in Los Angeles with SPA6 highlighted.

Figure 2. Service Planning Areas of Los Angeles County
SPA6 faces a wide variety of disadvantages in terms of social and economic resources,
including insufficient affordable primary care provision and access. According to the 2016
American Community Survey, SPA6 has the largest proportion of low-income population of the
eight SPAs of Los Angeles County. As indicated in the map on the left in Figure 3, 40% of

6
residents in SPA6 live in households with an annual income lower than $35,000, whereas in
SPA5, to the west, only 23% of residents live in such households. SPA6’s economic
disadvantage remains when using an even stricter definition of low-income population, as
presented in the map on the right in Figure 3 below. More than 30% of the residents in SPA6 live
in households with annual income lower than $25,000. The surfeit of individual economic
resources seen in SPA6 must be taken into consideration when conducting a site suitability
analysis for a new affordable primary care facility that serves SPA6 residents.

Figure 3. Low-Income Population by Service Planning Area
SPA6 features the lowest level of primary care provision amongst all eight SPAs,
according to Los Angeles Department of Health Services. The four maps in Figure 4 present the
status of overall primary care provision and low-cost primary care provision in Los Angeles
County. The map on the top left depicts overall primary care provision normalized by population
in each SPA. This reveals SPA6 has the lowest level of overall primary care provision in Los
Angeles County. The map on the top right shows overall primary care provision normalized by
low-income population whose annual household income is lower than $25,000. As shown in this

7
map, SPA6’s low-income population has the lowest level of primary care provision. Similarly,
the map on the bottom left indicates the low-income population whose annual household income
is lower than $35,000 also has the lowest level of primary care provision. The map on the bottom
left corner shows affordable primary care provision normalized by low-income population,
which reveals SPA6 is also among the areas with the lowest level of affordable primary care
provision. Therefore, it is crucial to find a way to better provide affordable primary care in SPA6
to ensure SPA6’s residents’ right to medical care.

Figure 4. Primary Care and Low-cost Primary Care Provision by Service Planning Area
1.2. Research Question
The primary objective of this thesis is to identify a suitable site for a new primary care
facility in SPA6. In order to best promote low-cost primary care accessibility in SPA6 with new
primary care facilities, it is important to select sites that maximize the improvement of affordable

8
primary care accessibility with the available resources. This project achieves this goal by first
identifying suitable parcels in SPA6 and second measuring how much each candidate site would
improve health care accessibility using the 2SFCA method.
This thesis has a two-fold spatial question:
1) What are the parcels in SPA6 that could potentially be a suitable and convenient site
for a new low-cost primary care facility? and
2) Which of these parcels could increase the accessibility to affordable primary care for
communities in SPA6 the most?
1.3. Structure of Thesis
This thesis consists of five chapters. Following this introductory chapter, Chapter 2
reviews related literature on primary care site selection and analytical methods for similar
research topics. Chapter 3 introduces data sources and explains the methodology for this thesis
project. Chapter 4 presents and discusses the results from the analyses described in Chapter 3.
Chapter 5 concludes this thesis with a summary of results and a discussion of research
limitations and future research.

9
Chapter 2 Literature Review
Inequities in primary care access are often severe for less-privileged individuals who neither
have medical insurance nor the financial resources to afford health care. To provide a scholarly
foundation for selecting a suitable site for a new affordable primary care facility to serve SPA6,
this chapter reviews prior research on primary care, primary care facility site selection, and
primary care accessibility.
2.1. Primary Care
In the health care system, four categories differentiate levels of care: 1) primary care
provides the essentials of health care as well as coordinates patients with other levels of care; 2)
secondary care is mostly provided by specialists for conditions that require a higher level of
expertise than what primary care providers can offer; 3) tertiary care provides highly specialized
health care for patients who need hospitalization; and 4) quaternary care is essentially an
extension of tertiary care with unusual and even more highly specialized care (Torrey 2015).
Among these four levels of care, primary care is particularly important because of it includes the
most commonly used and comprehensive health care services. The social welfare and public
health services covered within the rubric of primary care are crucial to broad-scale health
outcomes. Moreover, access to primary care is particularly important for the underprivileged
populations as poverty and undesirable health outcome tend to have a positive correlation (Hales
et al. 1999). The remainder of this section discusses primary care in greater detail and analyzes
the impacts of primary care provision on low-income populations

10
2.1.1. Primary Care
Primary care serves as a patient’s entry point into the health care system. Primary care
should coordinate health care services for patients and provide them long-term and
comprehensive care while taking in consideration their societal context (Macinko et al. 2003).
Furthermore, primary care combines curative and day-to-day preventive care. Overall, evidence
shows that primary care improves individual and population health and is crucial in the health
care system (Macinko et al. 2003).
The importance of primary care is two-fold. First, primary care is crucial for individual
health. Donaldson et al. (1996) identify five benefits of primary for individual health outcomes:
Primary care 1) treats or resolves a wide range of health problems so that patients can avoid
needing higher level care; 2) guides patients through the health care system; 3) facilitates patient
participation in decision-making regarding their own care; 4) prevents worse health outcome for
the patients through health promotion and early detection of health problems; and 5) incorporates
the patient with his/her family and community to achieve better health outcomes.
Second, primary care contributes to society in general by increasing the efficiency of
health care through reducing preventable hospitalizations and by improving public health
conditions (Donaldson et al. 1996). Preventable hospitalizations are those cases where outpatient
care such as primary care can potentially prevent the need for hospitalization or prevent further
complications and more severe medical conditions (Pezzin et al. 2018). Numerous studies
identify a strong negative association between primary care provision and preventable
hospitalizations; Namely, the number of preventable hospitalizations in a community is lower
when there is higher primary care provision and vice versa (Parchman and Culler 1994; Bindman
et al. 1995; Starfield 1995; Donaldson 1996). This relationship is especially true for conditions
such as asthma, hypertension, congestive heart failure, chronic obstructive pulmonary diseases,

11
and diabetes. In 1982, Medi-Cal benefits were terminated for 270,000 indigent residents of
California. A number of researchers saw that this deprivation of primary care service created an
opportunity to assess the relationship between access to primary care and health outcomes. Lurie
and coauthors (1984, 1986) studied those who lost access to primary care and found their
average health status worsened. Two decades later, Bindman et al. (2005) linked primary care
required by Medicaid and lower hospitalization rates in California. Furthermore, primary care
improves the cost efficiency of health care (Bindman et al. 2005). Adequate primary care not
only improves the efficiency of health care by reducing preventable hospitalizations and the high
cost associated with it (Bindman et al. 2005), but also saves costly specialized care by preventing
or resolving a wide variety of health problems at early stages (Friedberg, Hussey, and Schneider
2010).
2.1.2. Primary Care and Low-Income Populations
In addition to the overall societal benefits discussed in the previous section, primary care
is particularly important for the low-income population for two reasons. Firstly, low-income
populations are more prone to health issues than populations with more financial resources. This
is due to a wide range of reasons, such as limited access to healthy food and the lack of health
information. Secondly, access to primary care for the low-income population can significantly
mitigate the disadvantages caused by lack of financial resources. These two reasons are explored
in the remainder of this section.
Researchers find strong evidence that supports the link between poverty and undesirable
health outcomes across the world. Poverty is highly associated with infant mortality, a key
indicator of public health. Hales and coauthors (1999) identify a negative relationship between
national infant mortality rates and gross national product. Macinko and coauthors (2004)

12
establish the same relation with empirical data from member countries of the Organization for
Economic Cooperation and Development. Shi and coauthors (1999) find a similar negative
correlation between state-level infant mortality rates and economic inequality using empirical
data from the United States. Poverty is also linked to shorter life expectancy. DeVogli (2005)
confirms this link with evidence from Italy. Marmot and Bobak (2000) use multi-national data to
show this link both on the national and subnational levels. Poverty is also correlated with a series
of undesirable health outcomes. Shi and coauthors (1999) show that citizens who live under the
poverty line in the United States are more likely to suffer from several public health indicator
diseases such as arthritis, cancer, cardiovascular diseases, diabetes, obesity, etc. Pickett and
coauthors (2005) find similarly positive correlations between poverty and obesity in a cross-
national study among industrialized countries. These research findings indicate that low-income
populations tend to suffer more from health issues.
Scholars also demonstrate the causal relationship between sufficient primary care
provision and better health outcomes among the low-income population. Using data from 26
health service areas in Pennsylvania, Parchman and Culler (1994) argue that preventable
hospitalizations are significantly fewer where there is sufficient primary care provision and
access even after controlling for socioeconomic status. Moreover, Casanova and Starfield (1995)
provide evidence in a cross-national comparative study between Spain and the United States that
a variation in the preventable hospitalization rate disappears once universal primary care access
for children is introduced. Sood and coauthors (2014) examine the effects of a government
insurance program in India that provides free health care access to citizens below the poverty
line in half of the villages in the state of Karnataka. Since the villages receiving free health care
access are randomly selected, the authors take this opportunity and design a quasi-experiment

13
where the villages implementing the new government insurance program are the treatment group
and the other half of the villages in Karnataka are the control group. They find that the average
health outcomes are significantly better in the treatment group than the control group,
demonstrating the causal relationship between increased health care access and more desirable
health outcomes for low-income citizens.
Primary care has also been shown to reduce the cost of health care at the individual and
societal levels (Donaldson 1996). First, primary care providers on average are less expensive
than specialized care providers and other higher-level care providers (Donaldson 1996). Second,
as discussed previously in this section, access to primary care significantly decreases the need for
specialized care and hospitalization, which are considerably more expensive (Donaldson 1996).
Improving access to primary care not only reduces the financial burden of health care for the
low-income population, but also saves government health care expenditures by reducing
avoidable high-level care and increasing health care efficiency. Dor and Holahan (1990)
demonstrate that total Medicare expenditures per beneficiary decreases by one percentage point
per ten percentage points increase in primary care provision. Moreover, geographical variation in
primary care access also affects macro health care expenditures. Mark et al. (1996) and Welch et
al. (1993) respectively demonstrate that increased primary care reduces overall Medicare
expenditures and that insufficient primary care increases overall Medicare expenditures by
examining local primary care provision in the United States.
In sum, improving access to primary care for a low-income population greatly benefits
both the low-income population and the society as a whole. Primary care contributes to
improving the health outcome of low-income communities and overall public health as well as to
reducing the financial burden for both indigent individuals and the society. As the SPA with the

14
highest percentage of low-income residents and the lowest level of primary care provision in Los
Angeles County, SPA6 and its residents could benefit from additional primary care provision in
terms of both economic conditions and health outcomes.
2.2. Site Selection for Primary Care Facilities
Site selection traditionally consists of two steps: The first step generates a small number
of candidate sites by applying the predefined selection criteria, and the second step evaluates the
resulting shortlist of candidate sites to identify the most suitable one (Chang et al. 2008). This
section focuses on reviewing the site selection methods and criteria adopted in prior research.
The rest of this section reviews the site selection criteria for health care facilities in previous
research and the two commonly used methods for site selection – weighted overlay and fuzzy
overlay.
2.2.1. Site Selection Criteria for Health Care Facilities
Selecting a suitable site is important for any new facility, because the facility in question
could only function as intended if the site meets its needs. GIS provides powerful tools to
analyze the needs of the facility and the attributes of candidate sites. Previous research conducted
by both medical authorities and researchers using GIS has proposed a wide range of site
selection criteria for health care facilities of different sizes and purposes. This section reviews
the general site selection criteria applicable for most health care facilities including primary care
facilities suggested in prior research.
2.2.1.1. Initial Filtering Criteria
Complex site selection analysis often requires an initial filtering step in order to increase
computational efficiency (Estill & Associates 2006). For health care site selection analyses, three

15
initial filtering criteria – consistency with zoning regulations, vacancy status, and parcel size –
are frequently used to reduce the number of candidate sites (i.e. Estill & Associates 2006, Soltani
and Marandi 2011, University of California San Francisco 2011).
Firstly, candidate sites for new primary care facilities should comply with zoning
regulations (University of California San Francisco 2011). Specifically, a desirable candidate site
should be a parcel with a zoning code that allows medical practices according to local zoning
regulations. This criterion contributes to the computational efficiency of the site selection
analysis by filtering out a large number of candidate sites with zoning codes inappropriate for a
health care facility. Moreover, this criterion ensures the feasibility of the final site selection in
the real world. According to the regulations and policies of Los Angeles County Department of
Regional Planning, suitable zoning for a low-cost primary care facility falls in the commercial
zoning category. Within the commercial zoning category, there are eight zoning codes: C-1, C-2,
C-3, C-H, C-M, C-R, C-RU, and CPD. Table 2 below summarizes the requirements for each
commercial zoning code with a focus on requirements for medical facilities.
Table 2. Summary of Zoning Regulations
Zoning Permitted Uses Minimum
Required
Area
Minimum
Required
Parking
Maximum
Lot
Coverage
Outside Storage
C-1 Zone C-H uses,
commercial services,
retail sales of new
goods and genuine
antiques
No
minimum
required
area
1 parking
space for each
250 sq. ft. of
floor space for
medical
offices
90% of net
area of lot
Not permitted
C-2 Zone C-1 uses,
rentals, outdoor
advertising, tailor
shops
No
minimum
required
area
1 parking
space for each
250 sq. ft. of
floor space for
medical
offices
90% of net
area of lot
Not permitted

16
Zoning Permitted Uses Minimum
Required
Area
Minimum
Required
Parking
Maximum
Lot
Coverage
Outside Storage
C-3 Zone C-2 uses,
secondhand stores
No
minimum
required
area
1 parking
space for each
250 sq. ft. of
floor space for
medical
offices
90% of net
area of lot
Permitted at the
rear of a parcel
when incidental
to the permitted
use in the front
of the parcel
C-H Community and
financial services,
parks and
playgrounds,
business and
professional offices,
no retail sales
No
minimum
required
area
1 parking
space for each
250 sq. ft. of
floor space for
medical
offices
90% of net
area of lot
Not permitted
C-M Zone-3 uses, limited
manufacture and
assembly
No
minimum
required
area
1 parking
space for each
250 sq. ft. of
floor space for
medical
offices
90% of net
area of lot
Permitted at the
rear of a parcel
when incidental
to the permitted
use in the front
of the parcel
C-R Amusement parks,
campgrounds, tennis
courts, golf courses,
limited agriculture
5 acres N/A No
Maximum
coverage
N/A
C-RU Limited, low-
intensity commercial
uses that are
compatible with rural
and agricultural
activities
N/A N/A 50% N/A
CPD Single-family
residences, crops,
non-residential C-1
uses
5000 sq. ft. 1 parking
space for each
250 sq. ft. of
floor space for
medical
offices
40% N/A
Source: Los Angeles County Department of Regional Planning
Secondly, candidate sites for new primary care facilities should be vacant to ensure the
availability for construction (Estill & Associates 2006, Soltani and Marandi 2011). This criterion
also contributes to the computational efficiency of the site selection analysis by eliminating

17
unavailable candidate sites and the real-world accuracy by ensuring the final selected site is
available for the new primary care facility.
Thirdly, candidate sites should be suitably sized (Estill & Associates 2006). Parcels that
do not meet the basic spatial needs for parking and building of primary care facilities are too
small, while parcels that exceed the spatial needs for primary care facilities are likely to result in
waste in financial resources. Since all parcels with suitable zoning codes allow multiple-story
buildings, previous research tends to analyze the size of candidate sites without considering the
height of the facilities. Basic primary care facilities with one physician typically contain two to
three exam rooms, a consultation room, and a reception room, which together require at least
1,200 square feet (Freedman 2007). Freedman (2007) suggests a method to estimate the size of a
medical facility based on the number of physicians: 1,200-1,500 square feet for the first
physician and 1,000-1,200 square feet for each additional physician. In order to catch as many
otherwise suitable parcels as possible, this project defines the lower cutoff point for floor space
based on the space required by the most basic one-physician facility: 1,200 square feet. Los
Angeles County zoning regulations require one parking space per 250 square feet of floor space.
According to the Transportation Engineering Online Lab Manual (2003), one parking space
requires 310-330 square feet in the United States, including the driveway to access the parking
space. This project takes the lowest required space, 310 square feet, to calculate the minimum
space for the parking lot. Using this metric, the most basic primary care facility with one
physician and 1,200 square feet of floor space will require five parking spots, which requires an
additional 1,550 square feet, making the total minimum parcel size 2,750 square feet.
Unlike the minimum parcel size, the cutoff for parcels too large for a new low-cost
facility is defined relatively flexibly because it is possible to have large-scale medical facilities

18
and because a larger parcel size can allow for future expansion (Soltani and Marandi 2011). Los
Angeles County Department of Health Services data shows that the number of physicians at
existing low-cost primary care facilities ranges from zero (nurse practitioner or physician
assistant-led facilities) to 45. The largest existing low-cost primary care facility in Los Angeles
County, the LAC + USC Medical Center occupies a parcel of over 100,000 square feet.
However, the LAC + USC Medical Center also provides secondary and tertiary care in addition
to primary care. The largest parcel containing a facility that only provides low-cost primary care
is around 25,000 square feet.
2.2.1.2. Key Site Selection Criteria
Prior research on site selection for health care facilities suggests four key site selection
criteria including proximity to public transit, distance to existing similar facilities, land cost, and
proximity to targeted service recipients.
Firstly, as a facility that provides primary care services to the public, a primary care
facility should be close to public transit (Soltani and Marandi 2011). Proximity to the public
transit system is particularly important for health care facilities intended for disadvantaged
groups such as elderly or low-income residents as they might not have access to private vehicles
(Wu et al. 2007). Prior studies often employ proximity to public transit stops as a measurement
of this criteria (Soltani and Marandi 2011), namely, the closer a candidate site is to the nearest
public transit stop, the more desirable it is.
Secondly, a new health care facility should not be in the close vicinity of an existing
health care that provides similar services to avoid the waste of resources due to overlapping
service provision (Estill & Associates 2006, Soltani and Marandi 2011). There are two common
ways to measure the distance from a candidate site to existing health care facilities that provide

19
similar services. The first way defines the distance from a candidate site to existing health care
facilities that provide similar services in a binary way where candidate sites within a certain
distance of any existing similar facilities are considered unqualified while other candidate sites
remain in the candidate pool (Estill & Associates 2006). While this method is convenient and
easy to compute, the binary separation of suitable and unsuitable candidate sites might exclude
potentially satisfactory sites that are right outside the pre-defined distance. The second way
measures this criterion with the actual distance from a candidate site to existing health care
facilities that provide similar services (Soltani and Marandi 2011). This method provides a more
accurate measurement for this criterion but increases analytical complexity for the site selection
analysis.
Thirdly, land cost is a criterion for primary care facility site selection. This criterion is
important because land cost can potentially vary significantly while most of the other costs
related to a new primary care facility are relatively fixed (Vahidnia, Alesheikh, and
Alimohammadi 2008). For instance, for a primary care facility that has three physicians, two
nurse practitioners, five registered nurses, and five exam rooms, the labor cost, site construction
cost, medical service cost, and operational cost are relatively stable regardless of the location of
the new facility in a given neighborhood, city, or county. However, land cost can vary
considerably across a study area, especially in and around a metropolitan area such as Los
Angeles. Therefore, evaluating land cost for candidate sites is crucial if one wants to select the
cheapest suitable site for a new primary care facility. The most financially efficient way to use a
fixed amount of financial resources on a new primary care facility that aims at increasing
primary care provision is to choose sites with lower land cost, all else being equal. The less
money spent on site purchase, the more money there will be for primary care provision.

20
Lastly, proximity to targeted service recipients is one of the most common criteria for
health care facility site selection (Schuurman et al. 2006). Ideally, a health care facility should be
close to as many potential clients as possible. There are two aspects to this assessment: first the
population of potential clients must be identified, and second, the proximity must be measured.
Population density and local socio-demographics are usually used as proxies to measure the
number of potential clients close to the facility. For general health care facilities such as primary
care providers, population density alone constitutes a regularly used criterion since everyone can
be a potential recipient of their service (e.g. Schuurman et al. 2006; Wu, Lin, and Chen 2007 ;
Vahidnia, Alesheikh, and Alimohammadi 2008). A site in an area with a larger population is
more suitable for a primary care facility intended for the general public, ceteris paribus.
Similarly, an ideal site for a primary care facility that focuses on a specific group of potential
clients should take into consideration its socio-demographic characteristics. For instance, a clinic
that aims at providing health care to elderly residents should consider the population of adults
over 65 years of age (Kim et al. 2015). Therefore, this thesis includes the size of low-income
populations within one mile of candidate sites as one of the site selection criteria.
2.2.2. Weighted Overlay and Fuzzy Overlay
The previous section discusses key criteria for selecting suitable sites for primary care
facilities and this section reviews methods commonly used to integrate these criteria in the site
suitability analysis. Weighted overlay and fuzzy overlay are two commonly used methods to rate
suitable locations in site suitability analyses (Mitchell 2012).
In weighted overlay, analysts bring together data layers of their chosen criteria and
weight them relative to each other in terms of their impact on suitability. For instance, if a site
suitability analysis includes three criteria – land cost, slope, and aspect – analysts would first

21
gather data and create a source layer for each of these three criteria. Second, analysts must create
classes within each source layer and assign values to those classes using a scale of their own
choosing. Higher values are given to those classes that are more suitable. For instance, if analysts
are using land cost as a factor and want to keep land costs down, they will break up parcels into
classes based on cost and assign higher values to those parcels with lower land costs. Third,
analysts assign weights to each data layer according to its relative importance in the overall
analysis: the criteria that are deemed more important to the outcome are given higher weights.
Finally, analysts overlay the layers by adding up the weighted values of all the suitability criteria
for each location.
The fuzzy overlay method provides a tool for site suitability analyses where site selection
criteria are defined by continuous data without clear cut-off points between suitable values
(Mitchell 2012). Similar to the weighted overlay method, the fuzzy overlay method requires
analysts to first define a set of criteria for the site selection and create corresponding data layers
for further analysis.
Instead of assigning values on a scale of their own choosing to the observed data, the
fuzzy overlay method requires analysts to reclassify the observed data to values on a common
continuous scale of zero to one which represent the probability of candidate sites belonging to
sets of site selection criteria (Baidya et al. 2014). This value is called the fuzzy membership
value. A fuzzy membership value of zero indicates non-membership and a fuzzy membership of
one indicates full membership (Mitchell 2012). This contrasts from values in a weighted overlay
analysis, in which higher values represent more favorable sites. Furthermore, the fuzzy overlay
method allows analysts to transform observed data to fuzzy membership values with different
functions, giving the method more analytical flexibility. For instance, the Fuzzy Gaussian

22
function transforms observed data into a normal distribution, the Fuzzy Large function
transforms the observed data in a way so that larger input values are more likely to be a member
of the set, and the Fuzzy Near function transforms observed data by assigning full membership to
the midpoint data and decreasing the fuzzy membership value as values move away from the
midpoint data (Esri 2018a).
The fuzzy overlay method provides a variety of ways to assess the fuzzy membership
values for each data layer against each other, again providing more analytical flexibility than the
weighted overlay method, which only employs addition of values. Esri’s GIS products allow two
different logical and three different mathematical fuzzy overlay methods. The Fuzzy And
overlay type returns the minimum value of all input fuzzy membership values at each cell and
the Fuzzy Or overlay type returns the maximum value (Esri 2018b). Among the three
mathematical overlay types in the fuzzy overlay method, the Fuzzy Product overlay type
multiplies all input fuzzy membership values at each cell, the Fuzzy Sum overlay type sums all
input fuzzy membership values at each cell, and the Fuzzy Gamma overlay type combines the
Fuzzy Product and Fuzzy Sum overlay types by raising both to the power of gamma (Esri
2018b).
These two site selection methods each have their own advantages and disadvantages for
different site suitability analyses. The weighted overlay method is convenient and intuitive but it
requires well-defined, quantifiable criteria and lacks computational flexibility when combining
different criteria. The fuzzy overlay method is more flexible in terms of data requirements and
methods to combine different criteria even though it tends to be more complicated. As Mitchell
(2012) argues, the weighted overlay method is desirable for rating suitable locations when the
site selection criteria are defined by distinct categories or class ranges with clear cut-off points.

23
The fuzzy overlay method, in contrast, is desirable when criteria are defined by continuous data
with no clear cut-off points between suitable and unsuitable sites (Mitchell 2012). For instance,
the weighted overlay method is desirable when the analyst clearly knows candidate sites that cost
lower than $300 dollars per square foot are suitable. But the fuzzy overlay method is more
preferable than the weighted overlay method when the analyst only has a general idea that
cheaper sites are more suitable. In addition, the weighted overlay method is ideal when the
relationship between the overall suitability of a site and site selection criteria is linear while the
fuzzy overlay method is suitable when the relationship is more complicated. If the suitability
level of candidate sites is not only related to the weighted sum of all site selection criteria, the
weighted overlay method is likely to be inappropriate. Consider the selection of a suitable site for
a giant panda reserve as an example. The three criteria are coverage of bamboos, site size above
10,000 hectares, and access to fresh water. Among the three criteria, coverage of bamboos is a
necessary condition. If an analyst adopts the weighted overlay method for this site suitability
analysis, there is a chance that a site with no coverage of bamboos but extremely high ratings on
the other two criteria would be considered as suitable. Therefore, in this case, the weighted
overlay method is not appropriate.
2.3. Evaluating Accessibility
After reviewing methods and criteria for the first stage of site selection in the previous
section, this section focuses on research related to the second stage, namely, the evaluation stage.
Since the goal of this project is to identify a suitable site for a new low-cost primary care facility
in order to increase the primary care accessibility of low-income residents in SPA6, the
evaluation stage is centered on accessibility. Therefore, this section reviews methods to evaluate
and measure accessibility.

24
2.3.1. Earlier Methods to Measure Accessibility
One of the most commonly used conventional methods to measure accessibility is the
provider-to-population ratios (Guagliardo 2004). This method is intuitive and easy to compute:
analysts simply have to define areal units and divide the number of health care providers by the
number of residents within a given areal unit. The provider-to-population ratio is suitable for
comparisons of health care supply between large geopolitical units such as states and countries
(Connor, Hillson, and Krawelski 1995; Fortney et al. 1995). However, this method is
problematic because it does not account for a variety of issues that undermine the stability of
values across and between the chosen areal units such as patient border crossing, variation in
accessibility within areal units, or travel impedance (Guagliardo 2004). Thus, it suffers from
what is known as the modifiable areal unit problem (MAUP). Since this thesis aims to evaluate
primary care accessibility of SPA6, a relatively small geographical area bordering other urban
areas, the provider-to-population ratio method is not suitable.
Building on the concept of provider-to-population ratios while striving to address the
MAUP, gravity models provide more valid measures of spatial accessibility (Guagliardo 2004).
Gravity models aim to reflect the potential interactions between all population points and all
provider points within a certain distance while accounting for travel impedance. Instead of
calculating one provider-to-population ratio for each areal unit, gravity models account for all
potential providers for a given population point. Moreover, Guagliardo (2004) introduces the
concept of health care service capacity to represent the supply of primary care and uses the
number of physicians as the indicator for health care service capacity. The basic form of gravity
models can be summarized in the formula below:
𝐴
"
= %
𝑆
'
𝑑
"'
)
'

25
where accessibility (A) for population (i) is the sum for all provider locations (j) of the ratio of
service capacity S at provider location j to the travel impedance d between i and j, modified by
the gravity decay coefficient 𝛽. Travel impedance can be either travel distance or travel time.
Gravity models successfully account for the MAUP, yet, the results from gravity models tend to
be less intuitive in comparison to the conventional physician-to-population ratio. More
importantly, gravity models omit an important aspect of accessibility: the demand. Gravity
models only use population in terms of a location point from which to measure travel impedance,
such as the centroid of a census tract; The size of the population does not affect the accessibility
results. For instance, according to the gravity model of accessibility, the primary care
accessibility of a town with two primary care providers and 1,000 residents would be the same as
an otherwise identical town 10,000 residents. This is problematic.
The method by which gravity models assess provision capacity – a simple count of the
number of physicians – is also problematic. Firstly, using the number of physicians as the
indicator for primary care provision capacity assumes all physicians provide the same amount of
primary care, which is unrealistic. Even assuming each physician provides approximately the
same amount of primary care per unit time, the working hours for physicians at different primary
care facilities differ significantly. Secondly, using the number of physicians as the indicator for
primary care provision capacity neglects the primary care provided by other medical
professionals. Evidence from previous research shows that non-physician clinicians such as
nurse practitioners and physician assistants can also provide primary care when they have their
own panel of patients and supporting teams (Altschuler et al. 2012, Dill et al. 2013).

26
2.3.2. 2-Step Floating Catchment Area
The 2SFCA method builds upon conventional gravity models of accessibility by
including a measurement of demand. First introduced by Luo (2004) as the floating catchment
area (FCA), this method thus considers both the supply and the demand of a resource. Moreover,
it also provides a flexible way to assess primary care accessibility that is not constricted by the
MAUP and border effects.
Using the FCA method (Luo 2004), analysts first construct catchment areas for the
demand locations by defining the centers and radius. Population points, which can be a home
address if data allows or the centroid of an areal unit of population such as a census tract, are the
centers of catchment area. The radius of catchment areas is the distance that the user of a
resource is willing or able to travel to access that resource. Secondly, analysts add population
and provider data to each catchment areas and calculate the provider-to-population ratio. The
result is the measurement of accessibility. For instance, in an FCA analysis where catchment
areas are defined as circles centered on census tract centroids with a 5-mile radius, if the
catchment area centered on the centroid of census tract A contains four other census tract
centroids and three points representing service provider locations, the provider-to-population
ratio for census tract A would be ratio of the total number of providers at the three provider
locations and the sum of population in the five census tracts.
One limitation of the FCA method is that it neglects the possibility that providers at the
periphery of a catchment could also provide service to potential patients in nearby catchments.
Moreover, the FCA method assumes equal accessibility to all providers in a catchment for all
potential patients in that catchment. In order to address these issues, Wang and Luo (2005)
proposed the 2SFCA method.

27
The 2SFCA method, as its name indicates, consists of two steps: conducting the floating
catchment area calculation once from provider points and again from population points. The first
step calculates the provider-to-population ratios for the catchments centered on provider points.
Analysts define the radius of catchments just as in the FCA method, but the 2SFCA method
allows the radius to be measured as travel time via transportation networks in addition to simple
Euclidean distance. Previous research using the 2SFCA method to determine primary care
accessibility tends to define the radius of catchment areas by travel time via private vehicle. For
instance, Lee (1991) proposes 30 minutes as a reasonable driving time for primary care in rural
areas. Luo and Wang (2005) tests the 2SFCA method using driving time thresholds from 20 to
50 minutes. Travel time by private vehicle on a road network provides a more realistic
measurement of spatial accessibility than Euclidean distance because it reflects the cost to access
primary care facilities for potential clients. Moreover, road network datasets for travel time
analysis have become widely available in the past decade. Therefore, most previous research
research regarding 2SFCA uses travel time by private vehicle to calculate travel time. However,
while travel time by private vehicle on a road network already provides a more desirable
measurement than Euclidean distance, it still does not fully capture the travel cost for potential
clients due to the underlying assumption that all potential clients have access to private vehicles.
For instance, some potential clients, especially the low-income clients, may reply on public
transit to access primary care services. Yet, few previous research projects consider travel time
through public transit as a measurement of the catchment area radius due to the lack of road
network datasets constructed for public transit.
After the catchments for each provider point are drawn, a provider-to-population ratio is
calculated by dividing the total capacity of primary care provision by the total population within

28
the catchment. Similar to the earlier accessibility measures such as the provider-to-population
ratio and Floating Catchment Area methods, Wang and Luo (2005) also use the number of
physicians as the indicator for primary care provision capacity. For instance, Figure 5 below
provides an example area that consists of six rectangles to represent six census tracts: Census
Tract 1 with 2,000 residents, Census Tract 2 with 3,000 residents, Census Tract 3 with 1,000
residents, Census Tract 4 with 1,500 residents, Census Tract 5 with 1,000 residents, Census Tract
6 with 1,000 residents. The small blue circles are the centroids of each census tract. In this
example, they are also the population centers. The small blue triangles represent the primary care
provider locations: Provider Location 1 with five physicians, Provider Location 2 with three
physicians, and Provider Location 3 with four physicians. The three blue polygons are the
catchment areas of the three provider locations based on travel time. The catchment area for
Provider Location 1 includes the centroids of Census Tracts 1 and 4, and thus the provider-to-
population ratio for Provider Location 1 is 5:(2,000 + 1,500), namely, 1:700. Similarly, the
provider-to-population ratios for Provider Location 2 and 3 are 3:(1,000 + 1,000) and 4:(1,500 +
1,000 + 1,000), namely, 3:2000 and 1:875.

29

Figure 5. Illustration of the First Step in 2SFCA

30
The second step begins by drawing catchment areas for each population point. Next, the
provider-to-population ratios for each provider location that within each population catchment
are summed. For instance, if a given population point sits within the catchment areas of two
provider location points, the provider-to-population ratios for those two provider location points
are summed to generate the accessibility score for the population point. For instance, as shown in
Figure 6, the red polygon is the catchment area for Census Tract 4. This catchment area contains
two provider locations, namely, Provider Location 1 with a provider-to-population ratio of 1:700
and Provider Location 3 with a provider-to-population ratio of 1:875. Therefore, the accessibility
score for Census Tract 4 is the sum of these two provider-to-population ratios, which is
approximately 0.00257.

31

Figure 6. Illustration of the Second Step in 2SFCA

32

The two steps in 2SFCA discussed above can be summarized with the following two
formulas:
𝑅
'
=
𝑆
'
∑ -𝑑
"'
≤ 𝑑
/
0𝑃
2 2∈

𝐴
"
4
= % 𝑅
'
'∈{6
78
9 6
:
}

For each provider point j, Sj is the service provision capacity, Pk is the population that
falls into the catchment area of j, Rj is the provider-to-population ratio for point j, and dij is the
travel time through road network between the provider location j and the population location i
and d0 is the cutoff travel time for the catchments. 𝐴
"
4
is the accessibility score for location i,
which is the sum of the provider-to-population ratios for each provider point j inside of the
catchment area based on location i.
2.3.3. Public Transit in Measuring Accessibility
Paez, Scott, and Morency (2012) define accessibility as the potential for reaching
spatially distributed opportunities. They measure this using the cost of travel and the quantity of
opportunities. Previous accessibility research on health care facilities tends to measure the cost of
travel with travel time or distance to service by private transport, which includes indicators such
as the distance from a candidate site to major roads and driving time to a candidate site via a road
network (i.e. Wu et al. 2007; Brabyn and Beere 2006). However, potential patients’ ability to
afford transportation could undermine primary care accessibility for low-income populations
since their access to private vehicles might be limited. Martin et al. (2008) demonstrates that
spatial access to health care services by public transport is significantly different from access by
private transport, which raises the need to incorporate public transport accessibility into low-cost

33
primary care facility site selection. Therefore, it is not sufficient to simply calculate travel time
based on automobile in site selection for a low-cost primary care facility. The analysis should
also consider travel time via public transit in order to more accurately measure geographical
accessibility.
In prior research that considers the public transit factor, proximity to public transit stops
is often used as a measurement of public transport accessibility (Soltani and Marandi 2011).
However, this method does not account for travel time via public transit accurately or in detail.
Martin et at. (2008) proposed to take advantage of public transit timetable data and used
Microsoft Visual Basic to calculate travel time via public transit for the Derriford Hospital in
Devon, England. While effective, Microsoft Visual Basic is more demanding in terms of coding
in comparison to ArcGIS. The Add GTFS (General Transit Feed Specification) to a Network
Dataset tool in ArcMap provides a powerful solution in network analysis to perform schedule-
ware analysis, which is ideal for calculating travel time via public transit (Esri 2018c).

34
Chapter 3 Data and Methodology
The key goals of this project were to identify potential suitable sites for a new low-cost primary
care facility in SPA6 of Los Angeles County and to evaluate those potential new sites for their
ability to improve accessibility to primary health care for residents of SPA6. Chapter 2 reviews
prior research on site selection and site evaluation for primary care facilities. Building upon this
prior work, this chapter discusses the data and methods used in this thesis project.
This chapter discusses data used in this project and the data preparation process in the
first section. This section first defines the study area for this thesis before introducing datasets
used in this study, including existing low-cost primary care facilities, service planning areas,
census tracts, demographic data, parcels and zoning information, public transit, and road network
in SPA 6 and Los Angeles County.
The other two sections in this chapter focus on the methods adopted in this study. This
project employed the fuzzy overlay and 2SFCA methods in two analytical stages: site selection
and site evaluation. This project first generated a shortlist of candidate sites through the fuzzy
overlay method. Instead of simply evaluating the suitability of the candidate sites with the fuzzy
membership values like conventional site suitability analyses, this project employed the 2SFCA
method, a novel quantitative measurement of accessibility, to evaluate how each candidate site
on the shortlist affects the primary care accessibility for low-income populations in SPA6. In the
site evaluation stage, this project calculated low-cost primary care accessibility scores for each
low-income population point in SPA6 with existing low-cost primary care facilities as the
baseline accessibility and then calculated the accessibility scores with the addition of each
candidate site to compare with the baseline.

35
3.1. Data Sources and Data Preparation
This section introduces the data used in this thesis for the fuzzy overlay site selection
analysis and the site evaluation analysis with 2SFCA. Table 3 below summarizes the data
sources used in this thesis. The datasets for low-cost primary care facilities, SPAs, census tracts,
demographics, zoning, parcel boundaries, and road network were downloaded directly from the
sources. The dataset for the provision capacity of each low-cost primary care facility in Los
Angeles County was constructed by the author for this project by combining data on the number
of primary care providers and business hours at each existing low-income primary care facility
from the Los Angeles County Department of Health Service and the websites of the low-cost
primary care facilities. A public transit road network dataset was constructed by the author for
this project using data from the United States EPA Smart Location Database on routes, stops,
and schedules of public transit in Los Angeles County. The rest of this section discusses each
data source in detail.
Table 3. Data Sources
Dataset Source File Type Purpose
Service Planning Areas Los Angeles County
Department of Health Service
Vector Polygon
Shapefile
Fuzzy
Overlay;
2SFCA
Census Tracts United States Census Bureau Vector Polygon
Shapefile
Fuzzy
Overlay;
2SFCA
Demographics American Community Survey
by Census Bureau
Table Fuzzy
Overlay;
2SFCA
Parcels and Zoning Los Angeles County GIS Portal Vector Polygon
Shapefile
Fuzzy
Overlay
Low-Cost Primary Care
Facilities
Los Angeles County
Department of Health Services
Vector Point
Shapefile
Fuzzy
Overlay;
2SFCA
Low-cost Primary Care
Provision Capacity
Los Angeles County
Department of Health Service;
Table 2SFCA

36
Websites of Low-cost Primary
Care Facilities
Public Transit United States EPA Smart
Location Database
Vector Point
Shapefile
Fuzzy
Overlay;
2SFCA
Los Angeles County
Road Network
UCLA Geoportal Vector Line
Shapefile
Fuzzy
Overlay;
2SFCA
3.1.1. SPA Boundaries
As introduced in Chapter 1, the study area of this thesis is SPA6 of Los Angeles County.
This section further elucidates the geographical extent and spatial units of the analysis. The
boundary data of SPA6 was acquired from the Los Angeles County Department of Health
Services along with the boundaries of the other seven SPAs in the county.
In order to account for patients crossing SPA boundaries to seek primary care services
and outside SPA6, this analysis employed a three-mile buffer around SPA6 as the extended study
area. Figure 7 below shows both the SPA6 and the extended study area.

37

Figure 7. Extended Study Area

3.1.2. Low-Income Populations
The American Communities Survey offers data on demographic and income information
by census tracts. The 2016 American Communities Survey was the most updated version with
detailed data on the census tract level during the data collection phase of this project.
The U.S Census Bureau sets the poverty thresholds for households of specified sizes
annually on the federal level. However, the federal poverty thresholds might not accurately
reflect the poverty status in Los Angeles because they do not account for difference in the cost of
living or government poverty relief programs across the country. Aiming to address this
problem, the U.S. Census Bureau also provides the Supplemental Poverty Measure (SPM) as an

38
alternative poverty threshold guideline. Table 4 below presents both the federal poverty
thresholds and the SPM for the Los Angeles-Long Beach-Anaheim metropolitan area, where
SPA6 is located. Since the demographic data acquired for this project is from the 2016 American
Community Survey, this project also adopts poverty data in 2016.
Table 4. Federal Poverty Thresholds for Households of Specified Sizes
Poverty Guideline
Household Size
1 2 3 4 5 6 7 8
Federal 12,228 15,569 19,105 24,563 29,111 32,928 37,458 41,781
SPM for Renters 15,900 22,420 34,308 41,962 49,056 55,734 62,084 68,167
SPM for
Homeowners with
Mortgage
15,954

22,495

34,424

42,104

49,222

55,922

62,294

68,397

SPM for
Homeowners without
Mortgage
13,224

18,645

28,532

34,897

40,797

46,350

51,632

56,691

Data Source: U.S. Census Bureau
The average size of household in the United States is 2.58 according to the 2016
American Community Survey. Therefore, this project chose the threshold for low-income
populations for this analysis between the SPM thresholds for households of two and three. As
Table 4 indicates, the poverty guideline is $22,420 for households of two renters, $22,495 for
households of two homeowners with mortgage, $18,645 for households of two homeowners
without mortgage, $34,308 for households of three renters, $34,424 for households of three
homeowners with mortgage, $28,532 for households of three homeowners without mortgage. It
would be ideal if this project could calculate the poverty guideline for households of 2.58 people
and use it to define low-income populations. However, the American Community Survey does
not report household income as a ratio variable. Instead, it aggregates annual household income
data into categories including less than $10,000, $10,000 - $15,000, $15,000 - $25,000, $25,000
- $ 35,000, and so on. Therefore, this project selected $25,000, a number between the SPMs for

39
households of two and three that is also available in the American Community Survey, as the
cutoff for low-income populations. Figure 8 below shows the number of low-income residents by
census tract in the extended study area:

Figure 8. Low-Income Populations by Census Tract
3.1.3. Zoning and Parcel Dataset
In this thesis, parcel data in the extended study area is the candidate pool for site
selection. This project acquired parcel data from the Los Angeles GIS Portal. The parcel dataset
includes zoning information, parcel size, land value, address, and other attributes of the parcels.
This dataset provides information for three site selection criteria discussed in Chapter 2.
Firstly, the parcel size data can be used to filter out parcels that are too big or too small for a new
low-cost primary care facility. Secondly, the zoning data allows this analysis to select only

40
parcels with commercial zoning that is proper for medical facilities. Thirdly, the land value data
allows the fuzzy overlay analysis to use it as a factor. Figure 9 below presents the parcel data
with zoning information in the extended study area:

Figure 9. Parcel Data with Zoning Information
3.1.4. Public Transit Dataset and Road Network Dataset
As discussed in Chapter 2, public transit is a key factor both in the site selection for a
new low-cost primary care facility that serves low-income populations and in the 2SFCA site
evaluation. A desirable site should be easily accessed via public transit. Moreover, travel time
via public transit is an important factor for defining catchment areas for the 2SFCA analysis that
evaluates the impact on primary care accessibility brings by each candidate site.

41
This project acquired public transit data from the U.S. Environmental Protection Agency
Smart Location Database. This dataset contains public transit routes, stops, 0.25 and 0.5-mile
Euclidean buffers from public transit stops, number of trips per hour, and maximum wait time by
time periods throughout the day. Figure 10 shows the public transit routes and stops:

Figure 10. Public Transit Routes and Stops in the Extended Study Area
In order to define catchment areas in the 2SFCA analysis, this project needed to generate
travel time for both private vehicles and public transit. Since road network datasets for public
transit is not widely available, this project acquired public transit data such as bus routes, bus
schedule, subway routes, and subway schedule from the U.S. EPA Smart Location Database in
order to create a road network dataset for public transit. For the private vehicle travel-time
calculation, this project acquired the road network dataset for Los Angeles County from the

42
UCLA Geoportal. This road network dataset was generated based on the street layer from Esri
data and map collection. Figure 11 below shows the road network in the extended study area:

Figure 11. Road Network in the Extended Study Area
3.1.5. Existing Low-Cost Primary Care Facilities
Los Angeles County Department of Health Services maintains a dataset of primary care
facilities that provide low-cost primary care. This dataset was used in the fuzzy overlay analysis
to ensure candidate sites do not overlap with existing facilities in terms of service area and in the
2SFCA analysis to calculate accessibility scores. However, the Los Angeles County Department
of Health Services only includes the name, address, contact information, service information, and
business hours of the low-cost primary care facilities.

43
As discussed in Chapter 2, the way previous research uses the number of physicians as
the indicator of primary care provision capacity is problematic since it neglects non-physician
primary care providers and assumes equal amount of primary care provision per physician. This
project employed a revised indicator that represents primary care provision capacity more
realistically by accounting for the primary care provided by non-physician clinicians and the
various working hours of different primary care providers. Since physicians are usually the core
component of medical care, this study used their primary care provision capacity as the baseline
supply unit following the path of previous research (e.g. Guagliardo 2004). However, unlike
previous research which measure capacity by the number of physicians, this thesis defined the
baseline unit of supply as the amount of primary care one physician can provide in an hour. This
is to account for the various working hours of different primary care providers. For instance, the
new indicator can account for the different amount of primary care provided by a facility with
two physicians that opens for 40 hours a week and a facility with two physicians that opens for
20 hours a week.
This methodology also considered the work of physician assistants and nurse
practitioners in its measure of provision capacity, since both are capable of diagnosing medical
conditions, performing health examinations, treating illnesses, etc. (California Code, Business
and Professions Code, BPC § 12714). However, physician assistants and nurse practitioners are
legally required to work in collaboration with physicians and are usually under the supervision of
physicians (California Code, Business and Professions Code, BPC § 12714). Thus, this study
sets the primary care provision capacity of each physician assistant or nurse practitioner provides
per hour as 0.5 unit of provision capacity, acknowledging both their important roles in primary

44
care supply and their professional limitations. The calculation of the primary care service
capacity S of each existing low-cost primary care facility j is defined below:
𝑆
'
=(𝑀𝐷
'
∗1+ 𝑃𝐴
'
∗ 0.5+ 𝑁𝑃
'
∗0.5)∗𝐻
'

Where MDj is the number of physicians at facility j, PAj is the number of physician assistants at
location j, NPj is the number of nurse practitioners at location j, and Hj is the total business hours
per week at location j. The primary care capacity each physician provides per hour is calculated
as 1 unit of provision capacity.
The author identified the business hours of each low-cost primary care facility in the
extended study area as well as the number of each type of primary care provider from the
websites of the Los Angeles County Department of Health Services and low-cost primary care
facilities. Figure 12 presents the existing low-cost primary care facilities within the extended
study area and their service capacity according to the data provided by the Los Angeles County
Department of Health Services.

45

Figure 12. Location and Service Capacity of Existing Low-Cost Primary Care Facilities in the
Extended Study Area
3.2. Site Selection Analysis
This section describes the methods used in the site selection stage of this analysis. Firstly,
this section discusses the filtering process of all candidate sites that aims at increasing
computational efficiency prior to site selection. This section then explains the reason for
choosing fuzzy overlay as the site selection method. Lastly, this section details the work flow in
the fuzzy overlay analysis.

46
3.2.1. Filtering Parcels
A first step of filtering out candidate sites that are definitely not suitable for the new low-
cost primary care facility contributes to computational efficiency. This thesis adopted the three
filtering criteria of zoning designation, vacancy status, and parcel size.
If a candidate parcel is not zoned for commercial uses, it would not be suitable for the purpose of
this project regardless of its other attributes. Moreover, there are 502,889 parcels within the
extended study area, most of which are not zoned for commercial uses. It would require a lot of
computational power if this project conducted the fuzzy overlay analysis without first
eliminating parcels with inappropriate zoning. As discussed in Chapter 2, a site for a primary
care facility should have one of the eight commercial zoning designations. Since the study area is
SPA6 of Los Angeles County, this project eliminated parcels zoned C-R (large-scale recreational
uses), C-RU (agricultural activities), and CPD (maximum of 40% lot coverage), as these are
unsuitable for a primary care facility in an urban study area.
Moreover, since the goal of this project is to select a suitable site for a new low-cost
primary facility, it is also important to ensure site availability. Therefore, this project filtered out
all non-vacant parcels.
Parcel size is another essential requirement for candidate sites. If an otherwise suitable
parcel is too small to build a primary care facility, the parcel would be unfeasible for the goal of
this project. If an otherwise suitable parcel is too large, too many financial resources would be
spent on purchasing the parcel rather than primary care provision. But for a parcel that is neither
too large or too small, the size does not matter for site suitability because the parcel size is
positively correlated with primary care provision capacity, which can be used to calculate the
impact of the new low-cost primary care facility on primary care accessibility using 2SFCA.
Therefore, this project only eliminated parcels too large or too small in this filtering stage. As

47
discussed in Chapter 2, a basic primary care facility with only one physician requires at least
2,750 square feet. Therefore, this analysis eliminated any candidate sites smaller than 2,750
square feet as they cannot accommodate the spatial needs of a primary care facility. Also
discussed in Chapter 2, the largest parcel containing a facility that only provides low-cost
primary care in Los Angeles County is around 25,000 square feet. In order to allow for more
flexibility in the fuzzy overlay analysis, this project eliminated parcels over 30,000 square feet.
3.2.2. Site Selection Method
Chapter 2 reviews the two commonly used site selection methods, weighted overlay and
fuzzy overlay. This thesis adopted the fuzzy overly method for the following two reasons.
Firstly, the fuzzy overlay method is suitable for this thesis because the four key site selection
criteria, proximity to public transit, distance from existing low-cost primary care facilities, land
cost, and density of low-income populations, are all defined by continuous values with no clear
cut-off points between suitable and unsuitable sites. Previous research has only suggested
general directions of site suitability for these criteria instead of clearly distinguishable cut-off
points of suitability, making the weighted overlay method an undesirable option.
Secondly, the fuzzy overly method is ideal for this thesis because the relationship
between the overlay suitability and the four key site selection criteria is not linear. A desirable
site for the new low-cost primary care facility should meet the requirements of all four key
criteria simultaneously, which requires more flexible options to analyze the site selection criteria.
The weighted overlay method is limited in this case because it can only generate overall site
suitability as the weighted sum of all site selection criteria. By summing up the weighted values
of each criteria, a site with extremely desirable values on some of the four key site selection
criteria and undesirable values on others might have a high overall suitability if the weighted

48
overlay method is applied. The fuzzy overlay method, however, allows both logical and
mathematical operators in the analysis of different site selection criteria. The Fuzzy And method
offered in the fuzzy overlay analysis is ideal to combine the four key site selection criteria since
it returns the minimum value of all criteria, which ensures the final high-ranking sites are likely
to meet all site selection criteria.
3.2.3. Fuzzy Membership Procedures
After determining the site selection method, this thesis assigned functions to the
remaining four criteria to create fuzzy membership layers for the fuzzy overlay analysis.
3.2.3.1. Proximity to Public Transit
The closer to public transit a parcel is, the more suitable it is for a new low-cost primary
care facility. Therefore, this project used the Fuzzy Small transformation function for the
proximity to public transit criterion. With the Fuzzy Small transformation function, values larger
than the midpoint have a lower possibility of being a member and values smaller than the
midpoint have a higher possibility of being a member (Esri 2018a). Figure 13 shows a diagram
summarizing the workflow for creating this fuzzy membership layer:

Figure 13. Workflow for the Proximity to Public Transit Fuzzy Membership Layer

49
3.2.3.2. Distance to Existing Low-Cost Primary Care Facilities
A desirable candidate site should be as far away from the closest existing low-cost
primary care facility as possible. This project therefore employed the Fuzzy Large
transformation function for the distance to existing low-cost primary care facilities criterion. This
function allows values larger than the midpoint to have a higher possibility of being a member
and values smaller than the midpoint to have a lower possibility of being a member (Esri 2018a).
Figure 14 shows a diagram summarizing the workflow of this distance to existing low-cost
primary care facilities layer:

Figure 14. Workflow for the Distance to Existing Low-Cost Primary Care Facilities Fuzzy
Membership Layer
3.2.3.3. Land Cost
A desirable site should have as low per unit land cost as possible. Since the area of
candidate parcels varies, the total values of parcels are not comparable. This project first
calculated the land price per square foot for all candidate parcels and chooses the Fuzzy Small
transformation function for the land price per square foot. This function allows values larger than
the midpoint to have a higher possibility to be a member and values smaller than the midpoint to
have a lower possibility to be a member (Esri 2018a), namely, parcels with cheaper per square
foot land price are more likely to be members. Figure 15 shows a diagram summarizing the
workflow of this land cost layer:

50

Figure 15. Workflow for the Land Cost Fuzzy Membership Layer
3.2.3.4. Low-Income Populations
The last criterion for the site selection process is the proximity to targeted service
recipients, namely, low-income residents in SPA6. A desirable site for the new low-cost primary
care facility should be close to as many low-income residents as possible. This project chose the
Fuzzy Large transformation function for the low-income population criterion because this
function allows values larger than the midpoint to have a higher possibility of being a member
and values smaller than the midpoint to have a lower possibility of being a member (Esri 2018a).
In order to measure the number of low-income residents close to a candidate site, this project
created a one-mile buffer around all candidate parcels and calculated the number of low-income
residents within each buffer. Since the American Community Survey data used in this project
aggregates demographic data at the census tract level, this project assumed even distribution of
population within each census tract for this step. The number of targeted service recipients for
each candidate site was calculated by intersecting the 1-mile buffer with census tracts and then
calculating the number of low-income residents in the overlapping area with the Tabulate
Intersection tool offered in ArcGIS Pro. For instance, with the Tabulate Intersection tool, if a 1-
mile buffer contains 100% of census tract 1, 80% of census tract 2, and 30% of census tract 3,
the total population of the 1-mile buffer is the sum of 100% of the population in census tract 1,
80% of the population in census tract 2, and 30% of the population in census tract 3. While
assuming even population distribution within census tracts is not the most ideal way to reflect
population information accurately, it still has advantages over the alternative method that uses

51
census tract centroids as population points. This step also better accounted for primary care
accessibility for potential service recipients who have no access to either private vehicles or
public transit, as one mile is a reasonable distance for an average person to walk. Figure 16
shows a diagram summarizing the workflow of this low-income population layer:

Figure 16. Workflow for the Low-Income Population Layer
3.2.4. Fuzzy Overlay Analysis
This section explains the fuzzy overlay process used to combine the fuzzy membership
layers for the site selection.
This project chose the AND operator for the fuzzy membership analysis because a
desirable site for the new low-cost primary care facility should satisfy all four site selection
criteria to the greatest extent possible. The AND overlay method returns the minimum value
among all input fuzzy membership layers as the result, and thus the cells with high output values
are more likely to meet all site selection criteria (Mitchell 2012). The diagram in Figure 17
below presents the workflow for the fuzzy overlay analysis:

52

Figure 17. Fuzzy Overlay Workflow
The site selection analysis with the fuzzy overlay method is likely to have results where
more than one candidate site has the highest possibility or several parcels have similar high
possibility. Therefore, instead of simply comparing the final fuzzy membership values, this
project included a novel evaluation stage to further assess the suitability of candidate sites with
high fuzzy membership values with the 2SFCA method. The following section discusses how
this project employed the 2SFCA method to evaluate how the candidate sites with high scores
affect health care accessibility for low-income residents of SPA6.
3.3. Candidate Site Evaluation with 2SFCA
This section describes the process for evaluating whether the addition of a low-cost
primary care at each candidate site would improve health care accessibility for residents of
SPA6. To set a baseline from which to judge candidate sites, the health care accessibility for
each census tract in SPA6 given existing conditions was calculated. Then, the impact on
accessibility for each census tract was assessed with the addition of each candidate site. The
candidate site that brings the largest increase in the sum of accessibility scores in all census tracts

53
within SPA6 was the most suitable site for the new low-cost primary care facility. Figure 18
below summarizes the workflow to calculate the baseline accessibility scores and the updated
accessibility scores with the addition of each candidate site for the census tracts in SPA using the
2SFCA method:

Figure 18. 2SFCA Workflow
The first step was to define catchment areas. Conventionally, catchment areas for 2SFCA
analyses are defined by driving time via private vehicle as discussed in Chapter 2. This project
strived to account for the lower car ownership among the low-income residents; it therefore
defined catchment areas with both travel time via both private vehicle and public transit. In terms
of travel time via private vehicle, this project chose the 30-minute threshold, proposed by Lee
(1999) and accepted by Wang and Luo (2005). The travel time via private vehicle from each
population point, namely, the census tract centroids, can be calculated through with the Create
Service Area tool using the Los Angeles road network dataset acquired previously.
Public Transit
Data
Network Analyst:
New Network
Dataset
Public Transit
Road Network
Dataset
Road Network
Dataset
Existing Low-Cost Primary Care Facilities
OR Existing Low-Cost Primary Care
Facilities And One Candidate Site
Create Service
Area
Create Service
Area
30-Minute Travel
Time Areas via
Public Transit
30-Minute Travel
Time Areas via
Personal Vehicles
Intersect
Catchment Areas
Census Tract with
Demographic Data
Tabulate
Intersection
Existing Low-Cost Primary Care Facilities
OR Existing Low-Cost Primary Care
Facilities And One Candidate Site
Targeted Service
Recipients by
Catchment Areas
Summarize Within
Low-Cost Primary
Care Capacity by
Catchment Areas
Calculate
Provider-to-Population
Ratios for Each
Provider Location
Calculate
Census Tract with
Demographic Data
Accessibility
Scores by Census
Tract
Buffer
30-Minute Travel
Time Areas via
Public Transit with
A 0.5-Mile Buffer

54
In addition to travel time via private vehicle, this project also addressed travel time via
public transit. While there is no up-to-date road network dataset available for public transit in
Los Angeles County, it is possible to construct a road network dataset with the public transit data
acquired from the U.S. EPA Smart Location Dataset with ArcMap. Similar as travel time via
private vehicle, this project selected 30 minutes as the threshold for travel time via public transit.
Considering the lower density of public transit routes in comparison to road network, this project
also included any area within 0.5 mile, a reasonable walking distance, from public transit stops
as part of the service areas. Therefore, this project defined catchment area of a given low-cost
primary care facility or population center as the intersection of the 30-minute public transit travel
time polygon with a 0.5-mile buffer and the 30-minute private vehicle travel time polygon to
ensure a low-income resident can access this site within a reasonable amount of time in Los
Angeles traffic regardless of her car ownership status.
Since the population data for this project is aggregated at census tract level, it is
necessary to define the calculation of targeted service recipients for the candidate sites. As
discussed previously in this Chapter, the Tabulate Intersection tool provides a more realistic
result than assuming all residents concentrate on the centroid of a census tract. Therefore, this
project also adopted the Tabulate Intersection tool to estimate the number of low-income
residents in catchment areas.
Calculating accessibility scores also requires calculating the low-cost primary care
provision capacity for the candidate sites. As discussed in Chapter 2, the parcel size is positively
correlated with the number of primary care providers a candidate site can host. This project first
calculated the number of physicians each candidate site can host based on its size and then
calculated the primary care provision capacity with the assumption that the facility is open for 40

55
hours a week. The assumption for business hours per week is based on the average weekly
business hours of the 34 existing low-cost primary care facilities in the extended study area. The
actually average business hours per week for the 34 existing facilities is 42.69 and this project
adopted 40 for easy computation.
Since the accessibility scores calculated with the 2SFCA method are quantitative and
intuitive, it is relatively easy to compare the impacts on low-cost primary care accessibility
brought by each of the candidate site using the sum of accessibility scores in SPA6. This thesis
employed the 2SFCA method to calculate accessibility scores for each census tract in SPA6 first
with only the existing low-cost primary care facilities and used the sum of the accessibility
scores as the baseline for evaluating the shortlist of candidate sites generated from the fuzzy
overlay analysis. This project then calculated accessibility scores for each census tract
respectively with each final candidate site added. Next, this thesis compared the sum of the
primary care accessibility scores with the addition of each final candidate to the baseline, the
candidate site that leaded to the highest increase in the sum of primary care accessibility scores
was the overall most suitable site for a new low-cost primary care facility for SPA6. For
instance, if the sum of the baseline accessibility scores with only existing low-cost primary care
facilities in SPA6 is 100, the sum of the accessibility scores with the addition of candidate site A
is 103, the sum of the accessibility scores with the addition of candidate site B is 108, the sum of
the accessibility scores with the addition of candidate site C is 101, candidate site B would be the
most suitable site for a new low-cost primary care facility for improving primary care
accessibility for low-income populations in SPA6.
The following chapter presents the results of this thesis.

56
Chapter 4 Results
This chapter first describes the results from the initial filtering process before explaining the
fuzzy overlay analysis and the short list of candidate sites generated from it. Then, this chapter
demonstrates the evaluation of the short list of candidate sites using 2SFCA analysis and presents
the final results.
The initial filtering process reduced the number of candidate sites from 502,889 to 2,096
based on the three filtering criteria of zoning designation, vacancy status, and parcel size. The
fuzzy overlay analysis examined the 2,096 candidate sites using the four criteria of distance to
existing low-cost primary care facilities, distance to public transit, land cost, and number of low-
income residents in close proximity. This analysis generated 13 candidate sites with membership
values above 0.9 as a short list of candidate sites to be further evaluated with the 2SFCA
analysis.
The 2SFCA analysis first generated the service area for each of the 13 final candidate
sites based on travel time via private vehicles and via public transit. After calculating the service
capacity at each final candidate site based on the land available, the 2SFCA analysis calculated
the low-cost primary care accessibility scores for each census tract in SPA6 with only existing
low-cost primary care facilities as baseline and with the addition of each final candidate site in
order to evaluate their impacts on low-cost primary care accessibility for low-income populations
in SPA6. The 2SFCA analysis indicated the candidate site at 13910 Wilmington Avenue,
Compton, CA, 90250 as the most suitable site for a new low-cost primary care facility that serves
SPA6. The 2SFCA analysis also demonstrated how other final candidate sites affect the low-cost
primary care accessibility for the reference of potential decision makers.

57
4.1. Initial Filtering
The initial filtering process significantly reduced the number of candidate sites with the
three filter criteria. Among the total 502,889 parcels in the extended study area, 34,345 parcels
are zoned for commercial uses, and of these, 10,591 have zoning designations that are suitable
for primary care facilities such as C-1, C-2, C-3, C-H, or C-M. Among the 10,591 parcels with
suitable zoning, 2,980 parcels are vacant. After excluding sites smaller than 2,750 square feet
and larger than 30,000 square feet, the number of candidate sites reduced to 2,096. Figure 19
below presents the 2,096 candidate sites after the filtering process:

Figure 19. Candidate Sites After the Initial Filtering Process

58
4.2. Fuzzy Overlay Analysis
As discussed in Chapter 3, this thesis used fuzzy overlay analysis to select suitable sites
for the new low-cost primary care facility for low-income populations in SPA6 based on four
criteria after filtering out the sites that were not eligible because of zoning regulations and parcel
size. The fuzzy overlay analysis first assigned fuzzy membership values to the 2,096 candidate
sites based on the four criteria, including distance to the closest existing low-cost primary care
facility, distance to the closet public transit stop, land cost per square foot, and number of low-
income residents within a 1-mile radius. Figures 18-21 below summarize the fuzzy membership
values assigned to the candidate sites for each of the four criteria. For the sake of better
visualization, this thesis converted the fuzzy membership raster layers to vector point layers.
Figure 20 below presents the fuzzy membership values for all candidate sites based on
their distance to the closest existing low-cost primary care facility. As discussed in Chapter 2, a
desirable site for a new low-cost primary care facility should not be too close to an existing one
in order to avoid overlapping service areas. The distance from a candidate site to the closest
existing low-cost primary care facility ranges from 75 feet to 4.5 miles. This thesis used the
Fuzzy Large transformation function to assign fuzzy membership to the candidate sites, which
allowed values larger than the midpoint to have higher fuzzy membership values, and generated
fuzzy membership values range from 1.43e-11 to 0.97.

59

Figure 20. Fuzzy Membership Values Based on Distance to the Closest Existing Low-cost
Primary Care Facility
Figure 21 below shows the fuzzy membership values for all candidate sites based on their
distance to the closest public transit stop. As discussed in Chapter 2, a desirable site for a new
low-cost primary care facility should as close to public transit stops as possible in order to ensure
accessibility via public transit. This criterion is particularly important for the selection of a
primary care facility for the low-income populations as the car ownership rate is significantly
lower among them. The distance from a candidate site to the closest public transit stop ranges
from 13 feet to 1.02 miles. This thesis used the Fuzzy Small transformation function to assign
fuzzy membership to the candidate sites, which allowed values smaller than the midpoint to have
higher fuzzy membership values, and generated fuzzy membership values range from 0.03 to 1.

60

Figure 21. Fuzzy Membership Values Based on Distance to the Closest Public Transit Stop
Figure 22 below shows the fuzzy membership values for all candidate sites based on their
land value per square foot. As discussed in Chapter 2, a desirable site for a new low-cost primary
care facility should be as cheap as possible. Since the area of candidate sites varies, land value
per square foot is more comparable in comparison to total land value. The land value per square
foot for all candidate sites ranges from $23 to $97. This thesis used the Fuzzy Small
transformation function to assign fuzzy membership to the candidate sites, which allowed values
smaller than the midpoint to have higher fuzzy membership values, and generated fuzzy
membership values range from 0.87 to 1.

61

Figure 22. Fuzzy Membership Values Based on Land Value Per Square Foot
Figure 23 below shows the fuzzy membership values for all candidate sites based on the
estimated number of low-income residents within a 1-mile radius from each candidate site. As
discussed in Chapter 2, a desirable site for a new low-cost primary care facility should be as
cheap as possible. Instead of using the centroids of census tract as population points like some of
the previous research discussed in Chapter 2 does, this thesis assumed even distribution of
populations in census tracts and used the Tabulate Intersection tool offered in ArcGIS Pro to
estimate the number of low-income residents within each 1-mile buffer around each candidate
sites. While the assumption that the population density within each census tract is still not
accurate, it provides a more realistic estimate than treating the census centroids as population
points. The number of low-income residents within a 1-mile buffer around each candidate site

62
ranges from 2,164 to 74,515. This thesis used the Fuzzy Large transformation function to assign
fuzzy membership to the candidate sites, which allowed values greater than the midpoint to have
higher fuzzy membership values, and generated fuzzy membership values range from 0.33 to
0.94.

Figure 23. Fuzzy Membership Values Based on the Estimated Number of Low-Income Residents
within 1 Mile from Candidate Sites
After assigning fuzzy membership based on the four site selection criteria, this thesis
conducted the fuzzy overlay analysis using the Fuzzy AND overlay type and generated final
fuzzy membership values for the candidate sites. Fuzzy AND provides the minimum fuzzy
membership value from all fuzzy overly criteria, which ensures the candidate sites that have high
final fuzzy overlay membership scores satisfy all site selection criteria. Final fuzzy membership

63
scores range from 1.434E-11 to 0.969, with an average score of 0.168 and a standard deviation of
0.244. Among all candidate sites in the fuzzy overlay analysis, 13 candidate sites have final
fuzzy membership scores higher than 0.9. Figure 24 below presents the final fuzzy membership
values for the candidate sites.

Figure 24. Final Fuzzy Membership Values for Candidate Sites
The 13 candidate sites with final fuzzy membership values higher than 0.9 are the final
candidate sites in this project. Table 5 presents the shortlist of the final candidate sites to be
further evaluated with the 2SFCA method. These final candidate sites are arranged by their final
fuzzy membership scores in a descending order.

64
Table 5. Short List of Candidate Sites with Fuzzy Membership Values of 0.9 and Above
ID Address Size (Sq. Ft.)
Final Fuzzy
Membership Value
1
1532 FIRESTONE BLVD, LOS ANGELES, CA,
90001 4870.179248 0.969193935
2 4130 FLORENCE AVE, BELL, CA, 90201 16323.30486 0.945967317
3
8001 SANTA FE AVE, HUNTINGTON PARK, CA,
90255 21470.54491 0.923763096
4
2603 GRAND AVE, HUNTINGTON PARK, CA,
90255 8135.013437 0.921251476
5
2184 FIRESTONE BLVD, LOS ANGELES, CA,
90002 3372.569554 0.920206487
6
2876 E FLORENCE AVE, HUNTINGTON PARK,
CA, 90255 11653.9327 0.919949174
7 7721 COMPTON AVE, LOS ANGELES, CA, 90001 5539.405809 0.919698536
8
2182 FIRESTONE BLVD, LOS ANGELES, CA,
90002 3141.668533 0.918887079
9 7718 WILCOX AVE, CUDAHY, CA, 90201 22860.88452 0.915940046
10 1447 E 73RD ST, LOS ANGELES, CA, 90001 3756.327908 0.914339483
11
1754 FIRESTONE BLVD, LOS ANGELES, CA,
90001 4687.389735 0.91266644
12 13910 WILMINGTON AVE, COMPTON, CA, 90250 24312.85478 0.91241771
13 306 N. CENTRAL AVE, COMPTON, CA, 90220 8858.312761 0.905847728

Figure 25 below shows the locations of the candidate sites with fuzzy membership values
of 0.9 and above as final candidate sites in the map context.

65

Figure 25. Final Candidate Sites
4.3. 2SFCA Analysis and Recommended Sites
This section discusses the results of the 2SFCA analysis for the 13 candidate sites with
the highest final fuzzy membership values from the fuzzy overlay analysis. This section first
presents the catchment areas for existing low-cost primary care facilities and the 13 final
candidate sites based on travel time via both private vehicles and public transit and then shows
the impacts of each final candidate site on the accessibility scores for the census tracts in SPA6.
4.3.1. Catchment Area
As discussed in Chapter 3, this thesis defines catchment areas by travel time via both
private vehicle and public transit in order to account for the lower care ownership among the

66
low-income residents. The catchment area for a given low-cost primary care facility or
population center is the intersection of its 30-minute service area via private vehicle and its 30-
minute service area via public transit with a 0.5-mile buffer. The 0.5-mile buffer around the 20-
minute service area via public transit is to include areas within walking distance of public transit.
This sub-section presents the catchment areas for existing low-cost primary care facilities, final
candidate sites, and the census tracts in SPA6.
This project created 30-minute travel time areas via private vehicle and the 30-minute
travel time areas plus a 0.5-mile buffer via public transit for existing low-cost primary care
facilities in preparation of creating the catchment areas of these facilities. Since the travel areas
overlap with each other to an extent where it is difficult to visualize all travel time areas in one
map, Figure 26 only presents the 30-minute travel time area via private vehicle and the 30-
minute travel time area plus a 0.5-mile buffer via public transit for existing low-cost primary care
facility 27, Martin Luther King, Jr. Outpatient Center, as an example. As indicated in the map,
the two sets of travel time areas are significantly different, which supports this thesis’s choice of
incorporating both travel methods in defining catchment areas for a new low-cost primary care
facility that aims at increasing affordable primary care accessibility for low-income residents in
SPA6. The final catchment area for a given existing low-cost primary care facility is the
intersection of its 30-minute travel time areas via private vehicle and its 30-minute travel time
areas plus a 0.5-mile buffer via public transit.

67

Figure 26. 30-Minute Travel Time Areas Via Private Vehicle and Public Transit for Existing
Low-Cost Primary Care Facility 27
This project also created 30-minute travel time areas via private vehicle and the 30-
minute travel time areas plus a 0.5-mile buffer via public transit for the 13 final candidate sites.
Due to the similar problem of the overlapping of travel time areas, Figure 27 only presents the
30-minute travel areas via private vehicle and public transit for final candidate site 12 as an
example.

27
SPA6
Extended Study Area: 3-Mile Buffer Around SPA6
30-Minute Travel Time Areas and 1/2-Mile Buffers Via Public Transit for Existing Low-Cost Primary Care Facility 27
Existing Low-Cost Primary Care Facility 27
30-Minute Travel Areas Via Private Vehicles for Existing Low-Cost Primary Care Facility 27
0 10 20 5 Miles
±

68

Figure 27. 30-Minute Travel Time Areas Via Private Vehicle and Public Transit for Final
Candidate Site 12
Similarly, this project created 30-minute travel time areas via private vehicle and the 30-
minute travel time areas plus a 0.5-mile buffer via public transit for the census tract centroids in
SPA6. For the sake of better visualization, Figure 28 below only shows the 30-minute travel
areas via private vehicle and public transit for Census Tract 2184 (GEOID:
1400000US06037218400) as an example.

2
SPA6
Extended Study Area: 3-Mile Buffer Around SPA6
Final Candidate Site 12
30-Minute Travel Areas Via Private Vehicles for Final Candidate Site 12
30-Minute Travel Time Areas and 1/2-Mile Buffers Via Public Transit for Final Candidate Site 12
0 10 20 5 Miles
±

69

Figure 28. 30-Minute Travel Time Areas Via Private Vehicle and Public Transit for Census
Tract 2184
The catchment area for a given census tract or primary care facility is defined as the
intersection of its 30-minute travel time areas vis private vehicle and public transit. As discussed
in Chapter 3, the first step of the 2SFCA analysis uses the catchment areas for existing low-cost
primary care facilities and the final candidate sites to calculate the physician-to-population ratios
for each low-cost primary care facility and the second step of the 2SFCA analysis uses the
catchment areas for census tract centroids in SPA6 to sum up the physician-to-population ratio
for each census tract as its accessibility score. Like the 30-minute travel time areas presented
above, catchment areas of census tracts and primary care facilities also overlap to the point
2184
SPA6
Extended Study Area: 3-Mile Buffer Around SPA6
Census Tract 2184
30-Minute Travel Time Areas and 1/2-Mile Buffers Via Public Transit from the Centroid of Census Tract 2184
30-Minute Travel Time Areas Via Private Vehicle for Census Tract 2184
0 10 20 5 Miles
±

70
where visualizing all catchment areas in one map would be difficult. Therefore, Figure 29 below
presents the catchment areas of Census Tract 2184 and final candidate site 12 as an example.

Figure 29. Catchment Areas for Census Tract 2184 and Final Candidate Site 12
4.3.2. Accessibility Scores
This sub-section presents the accessibility scores for census tracts in SPA6 with existing
low-cost primary facilities as baseline and the accessibility scores for census tracts in SPA6 with
the addition of each of the final candidate sites. It then compares the impacts of each candidate
site on the overall low-cost primary care accessibility for low-income residents in SPA6.
As discussed in Chapter 3, the size of a candidate site determines its maximum service
provision capacity because the need for space is a function of the number of physicians. Thus,
2
2184
SPA6
Extended Study Area: 3-Mile Buffer Around SPA6
Census Tract 2184
Final Candidate Site 12
Catchment Area for Census Tract 2184
Catchment Area for Final Candidate Site 12
0 6.5 13 3.25 Miles
±

71
this project estimated the maximum number of physicians working at a given final candidate site
and the weekly service provision capacity of each final candidate site based on its size as
presented in Table 6 below.
Table 6. Estimated Weekly Service Provision Capacity of the Final Candidate Sites
ID Address Size (Sq. Ft.)
Estimated
Number of
Physicians
Estimated
Weekly
Service
Provision
Capacity
1
1532 Firestone Blvd, Los Angeles, CA,
90001 4870.179248 2
80
2 4130 Florence Ave, Bell, CA, 90201 16323.30486 7 280
3
8001 Santa Fe Ave, Huntington Park, CA,
90255 21470.54491 9
360
4
2603 Grand Ave, Huntington Park, CA,
90255 8135.013437 2
120
5
2184 Firestone Blvd, Los Angeles, CA,
90002 3372.569554 1
40
6
2876 E Florence Ave, Huntington Park, CA,
90255 11653.9327 5
200
7 7721 Compton Ave, Los Angeles, CA, 90001 5539.405809 2 80
8
2182 Firestone Blvd, Los Angeles, CA,
90002 3141.668533 1
40
9 7718 Wilcox Ave, Cudahy, CA, 90201 22860.88452 10 400
10 1447 E 73rd St., Los Angeles, CA, 90001 3756.327908 1 40
11
1754 Firestone Blvd, Los Angeles, CA,
90001 4687.389735 1
40
12
13910 Wilmington Ave, Compton, CA,
90220 24312.85478 10
400
13 306 N. Central Ave, Compton, CA, 90220 8858.312761 3 120

In order to set a baseline to evaluate which final candidate site best increases low-cost
primary care accessibility among low-income residents in SPA6, this project first calculated the
low-cost primary care accessibility scores for the census tracts within SPA6 using only existing
low-cost primary care facilities. With only existing low-cost primary care facilities, the mean
accessibility score for the census tracts in SPA6 is 0.0210 with a standard deviation of 0.0172.

72
The sum of accessibility scores of all census tracts within SPA6 is 6.0909. Figure 30 below
presents the low-cost primary care accessibility score for each census tract in SPA6 with
currently existing low-cost primary care facilities.

Figure 30. Current Low-Cost Primary Care Accessibility Scores by Census Tract in SPA6
After calculating the current low-cost primary care accessibility scores for census tracts
in SPA6 as the baseline, this project calculated the low-cost primary care accessibility scores for
census tracts in SPA6 with the addition of each of the 13 candidate sites. Table 7 below shows
the mean, standard deviation, and sum of accessibility scores with the addition of each final
candidate site as well as the baseline accessibility scores with only existing low-cost primary
care facilities.

73
Table 7. Summary Statistics of the Low-Cost Primary Care Accessibility Scores
Mean
Accessibility
Score
Standard
Deviation
Sum of Accessibility Scores of
All Census Tracts within
SPA6
Existing Facilities
Only
0.0210 0.0172 6.0909
Existing Facilities &
Final Candidate Site 1
0.0212 0.0173 6.1578
Existing Facilities &
Final Candidate Site 2
0.0215 0.0182 6.2261
Existing Facilities &
Final Candidate Site 3
0.0220 0.0175 6.3787
Existing Facilities &
Final Candidate Site 4
0.0213 0.0175 6.1755
Existing Facilities &
Final Candidate Site 5
0.0211 0.0173 6.1137
Existing Facilities &
Final Candidate Site 6
0.0217 0.0176 6.2840
Existing Facilities &
Final Candidate Site 7
0.0212 0.0173 6.1484
Existing Facilities &
Final Candidate Site 8
0.0211 0.0173 6.1141
Existing Facilities &
Final Candidate Site 9
0.0215 0.0176 6.2457
Existing Facilities &
Final Candidate Site
10
0.0211 0.0173 6.1157
Existing Facilities &
Final Candidate Site
11
0.0211 0.0173 6.1213
Existing Facilities &
Final Candidate site
12
0.0226 0.0174 6.5604
Existing Facilities &
Final Candidate Site
13
0.0214 0.0172 6.1960

Table 7 above shows that final candidate site 12 increases the mean and sum of low-cost
primary care accessibility scores for census tracts in SPA6 the most, followed by final candidate
sites 3 and 6. Therefore, final candidate site 12 is the most suitable site for a new low-cost

74
primary care facility to achieve the goal of improving low-cost primary care accessibility in for
low-income populations in SPA6 if there are no other conditions, concerns, or restrictions. If
financial resources for purchasing the land for the new low-cost primary care facility is limited,
final candidate sites 13 and 4 are the most suitable parcels under 20,000 square feet and final
candidate sites 1 and 7 are the most suitable parcels under 10,000 square feet.

75
Chapter 5 Discussion and Conclusion
This thesis provided both practical and methodological contributions to the GIS field and public
health. In terms of practical contribution, this project provided site selection suggestions for a
new low-cost primary care facility that could better increase affordable primary care accessibility
for low-income residents in SPA6 than sites with similar size and cost. This is helpful for
allocating resources more efficiently in order to provide more primary care to low-income
residents in SPA6. Moreover, this project offered a list of final candidate sites with various sizes
and land cost instead of one most suitable site, which provided flexible guidance to potential
decision-makers. The first section discusses the methods used in this thesis and the results
presented in the previous chapter. The second section analyzes the limitation of this project.
Lastly, the third section discusses future research.
5.1. Discussion
This section discusses the methods used in this thesis and analyzes the suitable sites from
the results in greater detail.
5.1.1. Overall Assessment of Methods and Analysis
The fuzzy overlay analysis assessed the 2,096 candidate sites based on four criteria,
namely, distance to the closest existing low-cost primary care facility, distance to the closet
public transit stop, land cost per square foot, and number of low-income residents within a 1-
mile radius. This process generated final fuzzy membership values for each candidate sites,
among which 13 candidate sites have a final membership value of 0.9 or above.
The fuzzy overlay analysis was successful because it evaluated candidate sites flexibly
and holistically. The final fuzzy membership value reflected which sites best fit all four site

76
selection criteria without allowing the high rating of one criterion to overly compensate for the
low rating of another criterion. The fuzzy overlay analysis is also easily adjustable to suit
different needs of future potential decision-makers’ who are interested in other types of site
selection. Firstly, potential decision-makers can use the result from the fuzzy overlay analysis
itself as a reference for site selection if they do not require quantitative accessibility analysis
because the fuzzy overlay analysis already provides a lot of decision-making support. Secondly,
it is convenient to slightly alter the fuzzy membership assignment in the fuzzy overlay analysis
to suit potential decision-makers’ requirements. Moreover, potential decision-makers can choose
to keep any number of sites from the fuzzy overlay analysis for further 2SFCA analysis based on
their needs.
In this project, the 2SFCA method provided a quantitative and flexible assessment of
low-cost primary care accessibility for the final candidate sites resulting from the fuzzy overlay
analysis that not only accounted for both the supply and demand of low-cost primary care, but
also avoided MAUP and border effects. Aiming to assess low-cost primary care accessibility in
SPA6 more accurately and realistically, this project calculated catchment areas in the 2SFCA
analysis with both travel time via private vehicle and public transit. As the maps in Chapter 4
indicate, the 30-minute travel time area via public transit with a 0.5-mile buffer is significantly
smaller than the 30-minute travel time area via private vehicle. Considering the relatively low car
ownership rate among low-income populations, it was crucial to incorporate travel time via
public transit when defining the catchment areas for both candidate sites and population points.
Due to the lack of available network dataset for public transit, this project created one using
point and line features of public transit stops and routes in Los Angeles County. This project also

77
added a 0.5-mile buffer, a reasonable walking distance, to the 30-minute travel time area via
public transit to better reflect the reality.
The 2SFCA analysis generated accessibility scores of the 13 candidate sites with final
fuzzy membership values above 0.9, providing a quantifiable criterion for potential decision-
makers to select the most suitable site for a new low-cost primary care facility. The innovative
use of the 2SFCA method to further evaluate the final candidate sites from the fuzzy overlay
analysis provides insightful results. The fuzzy overlay analysis ranked candidate sites by how
likely it is feasible and suitable to build a new low-cost primary care on them. All candidate sites
with high final membership values could be suitable for this project. However, the fuzzy overlay
analysis cannot provide a direct measurement of how these candidate sites contribute to the goal
of this study, namely, increasing primary care accessibility for low-income residents in SPA6.
Thus, this project adopted the 2SFCA method to build upon the fuzzy overlay analysis and to
directly evaluate how much each final candidate site could contribute to the goal of this study.
The results of the 2SFCA analysis indicate that the candidate sites with the highest final fuzzy
membership value do not lead to the highest increase in low-cost primary care accessibility in
SPA6. In fact, final candidate site 1, which has the highest final fuzzy membership value, ranks
the 8
th
in improving the overall low-cost primary care accessibility among the 13 final candidate
sites. Moreover, accessibility scores generated in the 2SFCA analysis can help entities that are
interested in investing for a new low-cost primary care facility in SPA6 better present their case
to the public and the local governments by intuitively demonstrating the positive impacts of the
added affordable primary care provision. Furthermore, if potential decision-makers desire, the
2SFCA analysis can be easily expanded to more candidate sites or limited to fewer candidate
sites. This provides the flexibility required in real-world decision making.

78
5.1.2. Discussion of the Final Results
This analysis shows that final candidate site 12 is the most suitable site for a new low-
cost primary care facility that best increases primary care accessibility for low-income
populations in SPA6. This sub-section presents detailed information on this site and discusses the
potential of building a new low-cost primary care facility on the site.
Final candidate site 12 locates in the southern part of SPA6 surrounded by the West 139
th

Street, North Wilmington Avenue, North Kemp Avenue, and West Cressey Street. The size of
this site is 24312.85 square feet or roughly 0.56 acre. It was last sold as vacant land in 2014 for
$345,000 (Redfin.com), which matches the land value data used in this thesis. Figure 31 below
presents a large-scale map for final candidate site 12 and an inset map that contextualizes its
location in SPA6.

79

Figure 31. Large-scale Map of Final Candidate site 12
In order to get a better idea of final candidate site 12, Figure 32 below presents a satellite
image of final candidate site 12 (circled in red) acquired from Google Maps and Figure 33 shows
a photo of the site taken in 2014. These two figures show that final candidate site 12 is located in
the middle of residential communities.

80

Figure 32. Satellite Image of Final Candidate site 12. Source: Google Maps. Accessed: March
22nd, 2019.

81

Figure 33. Photo of Final Candidate site 12. Source: Redfin.com. Accessed: March 22nd, 2019.
5.2. Limitations and Improvements
This section discusses the limitations of this project and potential improvements that
could be made for similar future analyses. Most of the limitations of this project are due to data
quality and data availability. In terms of data quality, this project strived to obtain the most
accurate and up-to-date data available. However, due to data accessibility and availability issues,
this project had to compromise and use the best-available data within in the scope of this thesis.
The first limitation of this project is that the demographic data regarding the distribution
of low-income populations is aggregated on the census tract level. While this project strived to
model low-income populations more accurately in the first step of the 2SFCA analysis by
assuming even distribution of low-income populations across SPA6 instead of assuming all low-
income populations concentrate on the census tract centroids as conventional research does, it

82
still inevitably introduced errors to the analysis. Moreover, due to the lack of reliable data, this
project did not consider the primary care need of homeless populations in the study area, which
could potentially introduce inaccuracies to the results. While it is unlikely to completely solve
this data availability problem, a potential improvement for this limitation could be made if future
projects with more resources could acquire household-level or up-to-date census block
demographic data.
The second limitation of this project is the road network dataset for public transit. Due to
the lack of publicly available public transit road network dataset for the study area, this project
created one using vector public transit route and stop data acquired from United States EPA
Smart Location Database. This public transit road network dataset may not contain the most
current or complete public transit information for the study area. Moreover, it also does not
include data for accurately modeling traffic friction, wait time, or delay status. A potential
improvement for this limitation could be made if future analyses could acquire a more detailed,
accurate, and complete road network dataset for public transit in their study areas. If research
resources are limited, a project that solely focuses on creating a road network dataset to model
real-life public transit accurately and realistically would also have significant contribution to
future studies.
The third limitation of this project is the land value data in the Parcel and Zoning Dataset.
This project used land value per square foot as one of the criteria to evaluate the suitability of a
candidate site. However, the land value data in the Parcel and Zoning Dataset is acquired from
the Los Angeles County GIS Portal, which is based property tax data from 2015. Estimated land
value for taxation purposes may not accurately reflect a property’s market value and could be not
up-to-date. However, since the Los Angeles County property tax assessor uses the same

83
methodology to estimate the value of all properties, the estimation error is likely to be consistent
for all candidate sites. Therefore, the relative land cost of a given candidate site in comparison to
other candidate sites in the property tax data is still informative and accurate enough for this
project to evaluate candidate sites. A potential improvement for this limitation is to use more
accurate and up-to-date land cost data in the analysis. For instance, real estate listing websites
such as Zillow.com and Redfin.com have algorithms developed to estimate the market price for
properties based on a great number of publicly available data and user-submitted data. The
results might be more accurate if future analyses have the resources to scrape these real estate
listing websites and use their estimated land cost for the candidate sites in the site suitability
analysis.
The fourth limitation of this project is the low-cost primary care provision capacity of
each facility used in the 2SFCA analysis. Due to the lack of precise information that documents
the actual primary care provided at each facility, this project estimates the service provision
capacity based on the business hours and the number of providers at each facility. A potential
improvement for this limitation could be made if similar future analyses had the authorization
and budget to gather the actual service provision statistics of existing low-cost primary care
facilities.
5.3. Future Research
This section discusses future research directions building upon this project. Firstly, this
project can serve as a framework for similar future analyses. Potential decision-makers and GIS
analysts can adopt the analytical methods and work flow used in this thesis and replicate the
analysis using data from other study areas. A lot of communities, both domestic and
international, desperately need more easily accessible affordable primary care. Future research

84
replicating this project on other study areas could potentially help decision-makers to better
allocate resources to benefit these communities. This type of future research could be particularly
useful for local public health authorities such as Los Angeles Department of Health Services and
nonprofit health care providers with government contracts. The analytical framework of this
project can be tailored to the specific needs of the aforementioned entities, such as budget,
facility size, specific candidate sites, and targeted clients, to help them select the most suitable
site for a new primary care facility. If local public health authorities could provide more accurate
data such as household-level demographic data and the actual workload of existing primary care
facilities, the analytical framework from this thesis could also produce a more accurate analysis
for the site selection.
Moreover, future research can also adapt the methods used in this project to address
primary care provision issues for other disadvantaged groups, health care provision for patients
with different health care needs, and other public service provision. For instance, future research
that is interested in increasing affordable primary care for low-income senior citizens in a given
study area can acquire demographic data to estimate the distribution of targeted service
recipients and adjust the site selection criteria to account for the characteristics of senior citizens.
In addition, future research can also adapt the analytical framework of this project to optimize
the provision of other types of health care. Future research could collect relevant data on certain
health care needs, especially the kind of health care needs that have clear geographic pattern. For
instance, if local health authorities need to better allocate health care resources to serve residents
with lead pollution-related diseases, future research could adapt the methods used in this project
by including the spatial distribution of lead pollution and the targeted service recipients to
address this new problem. Moreover, the methods used in this project can also be adapted and

85
used in site selection analyses for other public service provision. For instance, future research
can employ the same Fuzzy Overlay Analysis and 2SFCA methods to select a suitable site for a
new soup kitchen or a low-budget food store to increase food accessibility for low-income
populations.
Furthermore, future research can build upon this project and improve the modeling of
travel time via public transit. Future research could build more realistic and accurate road
network datasets using public transit routes, stops, wait time, delay and other data for more study
areas. If resources permit, future research could even use real time public transit data to analyze
travel time and catchment areas.
5.4. Conclusion
This thesis provides site selection suggestions for a new low-cost primary care facility
that best improves the affordable primary care accessibility for low-income populations in SPA6
using fuzzy overlay analysis and 2SFCA analysis. The results of this project could help private
or public entities that are interested in providing low-cost primary care in SPA6 better allocate
their resources so that low-income residents could benefit more from the investment.
In terms of contribution, this project analyzes site suitability while taking into
consideration the economic characteristics of the targeted service recipients and assesses low-
cost primary care accessibility quantitatively. This thesis accounts for the relatively low car
ownership rate among low-income populations and calculates service area with travel time via
both private vehicle and public transit in order to more accurate model the suitability and
geographical accessibility of candidate sites. Moreover, this thesis quantitatively assesses the
impact on low-cost primary care accessibility in SPA6 of each final candidate sites using the
2SFCA method, providing more accurate information for potential decision-makers.

86
This thesis offers both practical guidance for decision-makers to address the real-world
problem in SPA6 and an analytical framework for GIS analysts to solve similar problems
elsewhere.

87
References
Altschuler, Justin, David Margolius, Thomas Bodenheimer, and Kevin Grumbach. 2012.
“Estimating a Reasonable Patient Panel Size for Primary Care Physician with Team-
based Tasked Delegation.” Annals of Family Medicine 10(5): 396-400.
Baidya, Priyanka, Dibyajyoti Chutia, Singuluri Sudhakar, Chandan Goswami, Jonali Goswami,
Victor Saikhom, Puyam S. Singh, and Kamini Kanta Sarma. 2014. “Effectiveness of
Fuzzy Overlay Function for Multi-Criteria Spatial Modeling: A Case Study on
Preparation of Land Resources Map for Mawsynram Block of East Khasi Hills District of
Meghalaya, India.” Journal of Geographic Information System 6: 605-612.
Bindman, Andrew B., Kevin Grumbach, Dennis Osmond, Miriam Komaromy, Karen Vranizan,
Nicole Lurie, and John Billings. 1995. “Preventable Hospitalizations and Access to
Health Care.” Journal of the American Medical Association 274: 305-311.
Bindman, Andrew B., Arpita Chattopadhyay, Dennis H. Osmond, William Huen, and Peter
Bacchetti. 2005. “The Impact of Medicaid Managed Care on Hospitalizations for
Ambulatory Care Sensitive Conditions.” Health Services Research 40: 19-38.
Brabyn, Lars and Beere, Paul. 2006. “Population Access to Hospital Emergency Departments
and the Impacts of Health Reform in New Zealand.” Health Information Journal 12:
227–237.
California Code, Business and Professions Code, BPC § 12714
Casanova, Carmen and Barbara Starfield. 1995. “Hospitalizations of Children and Access to
Primary Care: A Cross-National Comparison.” International Journal of Health Services
25: 283-294.
Chang, Ni-Bin, G. Parvathinathan, and Jeff B. Breeden. 2008. “Combining GIS with Fuzzy
Multicriteria Decision-making for Landfill Siting in A Fasting-growing Urban Region.”
Journal of Environmental Management 87: 139-153.
Conner, Robert A., Steven D. Hillson, and John E. Krawelski. 1995. “Competition, Professional
Synergism, and the Geographic Distribution of Rural Physicians.” Medical Care 33(11):
1067-1078.
De Vogli, Roberto, Ritesh Mistry, Roberto Gnesotto, and Giovanni Andrea Cornia. 2005.
“Income Inequality and Life Expectancy: Evidence from Italy.” Journal of Epidemiology
and Community Health 59: 158-162.
Dill, Michael J., Stacie Pankow, Clese Erikson, and Scott Shipman. 2013. “Survey Shows
Consumers Open to A Greater Role for Physician Assistants and Nurse Practitioners.”
Health Affairs 32(6). Accessed February 5
th
, 2019.
https://doi.org/10.1377/hlthaff.2012.1150

88

Donaldson, Molla S., Karl D. Yordy, Kathleen N. Lohr, and Neal A. Vanselow. 1996. Primary
Care: America’s Health in A New Era. National Academy Press.
Dor, Avi and John Holahan. 1990. “Urban-Rural Differences in Medicare Physician
Expenditures.” Inquiry 27: 307-318.
Esri. 2018a. “How Fuzzy Membership Works. ” Esri.
http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/how-fuzzy-
membership-works.htm (Accessed January 18th, 2019).
Esri. 2018b. “How Fuzzy Overlay Works. ” Esri.
http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/how-fuzzy-
membership-works.htm (Accessed January 18th, 2019).
Esri. 2018c. “Add GTFS to a Network Dataset.” Esri. https://esri.github.io/public-transit-
tools/AddGTFStoaNetworkDataset.html (Accessed June 17, 2019)
Estill & Associates. 2006. Site Selection Community Consultation Report. Washington:
Department of Health.
Freedman, Robert. 2007. “Planning a New Medical Office Space.” Medispcape. Accessed Nov
29
th
, 2018. https://www.medscape.com/viewarticle/554115
Fortney, John C., Brenda M. Booth, Frederic C. Blow, Janice Y. Bunn, and Cynthia A. Loveland
Cook. 1995. “The Effects of Travel Barriers and Age on the Utilization of Alcoholism
Treatment Aftercare.” The American Journal of Drug and Alcohol Abuse 21(3): 391-406.
Friedberg, Mark W., Peter S. Hussey, and Eric C. Schneider. 2010. “Primary Care: A Critical
Review of the Evidence on Quality and Costs of Health Care. ” Health Affairs 29: 766-
772.
Guagliardo, Mark F. 2004. “Spatial Accessibility of Primary Care: Concepts, Methods, and
Challenges. ” International Journal of Health Geographics 3: 1-13.
Hales, Simon, Philippa Howden-Chapman, Clare Salmond, Alistair Woodward, and Johan
Mackenbach. 1999. “National Infant Mortality Rates in Relation to Gross National
Product and Distribution of Income.” The Lancet 354: 2047-2047.
Healey, Michael and Brian Ilbery. 1990. Location and Change: Perspectives on Economic
Geography. Oxford University Press.
University of California San Francisco. 2011. “Hospital Replacement Site, Selection Evaluation
Criteria. (n.d.).” Accessed Nov 15
th
, 2018.
https://www.ucsf.edu/sites/default/files/legacy_files/LRDP-Appendices-I.pdf

89
Kim, Jung In, Devini Manouri Senaratna, Jacobo Ruza, Calvin Kam, and Sandy Ng. 2015.
“Feasibility Study on An Evidence-Based Decision-Support System for Hospital Site
Selection for An Aging Population.” Sustainability 7: 2730-2744.
Lee, Richard C. 1991. “Current Approaches to Shortage Area Designation.” The Journal of
Rural Health 7: 437 – 450.
Luo, Wei. 2014. “Using a GIS-based Floating Catchment Method to Assess Areas with Shortage
of Physicians.” Health & Place 10: 1-11.
Lurie, Nicole, Nancy B. Ward, Martin F. Shapiro, and Robert H. Brook. 1984. “Termination
from Medi-Cal – Does It Affect Health? ” New England Journal of Medicine 311: 480-
484.
Lurie, Nicole, Nancy B. Ward, Martin F. Shapiro, Claudio Gallego, Rati Vaghaiwalla, and
Robert H. Brook. 1986. “Special Report.Termination from Medi-Cal Benefits – A
Follow-Up Study One Year Later.” New England Journal of Medicine 314: 1266-1268.
Macinko, James, Barbara Starfield, and Leiyu Shi. 2003. “The Contribution of Primary Care
Systems to Health Outcomes Within Organization for Economic Cooperation and
Development (OECD) Countries, 1970 – 1998.” Health Services Research 38： 831-865.
Macinko, James, Leiyu Shi, and Barbara Starfield. 2004. “Wage Inequality, the Health System,
and Infant Mortality in Wealthy Industrialized Countries, 1970-1996.” Social Science &
Medicine 58: 279-292.
Mark, David H, Mark S. Gottlieb, Bruce B. Zellner, John E. Chetty, and David H. Midtling.
1996. “Medicare Costs and the Supply of Primary Care Physicians.” Journal of Family
Practice 43: 33-39.
Marmot, Michale and Martin Bobak. 2000. “International Comparators and Poverty and Health
in Europe.” British Medical Journal 321: 1124-1128.
Martin, David, Hannah Jordan, and Paul Roderick. 2008. “Taking the Bus: Incorporating Public
Transport Timetable Data into Health care Accessibility Modelling.” Environment and
Planning A: Economy and Space 40: 2510–2525.
Mitchell, Andy. 2012. The Esri Guide to GIS Analysis Volume 3: Modeling Suitability,
Movement, and Interaction. Esri Press.
Parchman, Michael L. and Steven D. Culler. 1994. “Primary Care Physicians and Avoidable
Hospitalizations.” The Journal of Family Practice 39: 123-128.
Park, Annie, Nancy Watson, and Lark Galloway-Gilliam. 2008. “South Los Angeles Health
Equity Scorecard.” Community Health Councils.

90
Paez Antonio, Darren M. Scott, and Catherine Morency. 2012. “Measuring Accessibility:
Positive and Normative Implementations of Various Accessibility Indicators.” Journal of
Transport Geography 25: 141-153.
Pezzin, Liliana E., Hillary Bogner, Jibby Kurichi, Pui Kwong, Joel Streim, Dawei Xie, Ling Na,
Sean Hennessy. 2018. “Preventable Hospitalizations, Barriers to Care, and Disability.”
Medicine 97: 1-8.
Pickett, Kate E., Shona Kelly, Eric Brunner, Tim Lobstein, and Richard G. Wilkinson. 2005.
“Wider Income Gaps, Wider Waistbands? An Ecological Study of Income Inequality and
Obesity.” Journal of Epidemiology and Community Health 59: 670-670.
Prentice, Julia C. 2005. “Neighborhood Effects on Primary Care Access in Los Angeles.” Social
Science & Medicine 62: 1291-1303.
Qiu, Fang, Bryan Chastain, Yuhong Zhou, Caiyun Zhang, and Harini Sridharan. 2014.
“Modeling Land Suitability/Capability Using Fuzzy Evaluation.” GeoJournal 79(2): 167-
182.
Schuurman, Nadine, Robert Fielder, Stefan Grzybowski, and Darrin Grund. 2006. “Defining
Rational Hospital Catchments for Non-urban Areas Based on Travel-time.” International
Journal of Health Geographics 5(43): 1-11.
Shi, Leiyu., Barbara Starfield, Bruce Kennedy, and Ichiro Kawachi. 1999. “Income Inequality,
Primary Care, and Health Indicators. ” The Journal of Family Practice 48: 275-284.
Soltani, Ali and Ebrahim Zargari Marandi. 2011. “Hospital Site Selection Using Two-Stage
Fuzzy Multi-Criteria Decision Making Process.” Journal of Urban and Environmental
Engineering 5: 32-43.
Sood, Neeraj, Eran Bendavid, Arnab Mukherji, Zachary Wagner, and Somil Nagpal. 2014.
“Government Health Insurance for People Below Poverty Line in India: Quasi-
experimental Evaluation of Insurance and Health Outcomes.” British Medical Journal
349: 511-541.
Starfield, Barbara. 1995. “Access-Perceived or Real, and to What?” Journal of the American
Medical Association 274: 346-347
Transportation Engineering Lab. 2003. “Parking Lot Design: Theory and Concepts.”
Transportation Engineering Online Lab Manual. Accessed Nov 29
th
.
http://www.webpages.uidaho.edu/niatt_labmanual/chapters/parkinglotdesign/theoryandco
ncepts/parkingstalllayoutconsiderations.htm
Torrey, Trisha. 2018. “Levels of Medical Care: Primary, Secondary, Tertiary and Quaternary
Care.” About Health. Accessed October 20, 2018.
https://www.verywellhealth.com/primary-secondary-tertiary-and-quaternary-care-
2615354

91
Wallace, Stephen P. 1990. “The No-Care Zone: Availability, Accessibility, and Acceptability in
Community-Based Long-Term Care.” The Gerontologist 30: 254-261.
Welch, W. Pete, Mark E. Miller, Gilbert H. Welch, Elliott S. Fisher, and John E. Wennberg.
1993. “Geographic Variation in Expenditures for Physicians’ Services in the United
States.” New England Journal of Medicine 328: 621-627.
Wu, Cheng-Ru, Chin-Tsai Lin, and Huang-Chu Chen. “Optimal Selection of Location for
Taiwanese Hospitals to Ensure A Competitive Advantage by Using the Analytic Hierarcy
Process and Sensitivity Analysis.” Building and Environment 42: 1431-1444.
Vahidnia, Mohammad H., Ali A. Alesheikh, and Abbas Alimohammadi. 2008. “Hospital Site
Selection Using Fuzzy AHP and Its Derivatives.” Journal of Environmental Management
90: 3048-3056.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A spatial analysis of veteran healthcare accessibility

PDF

An accessibility analysis of the homeless populations' potential access to healthcare facilities in the Los Angeles Continuum of Care

PDF

Suitability analysis for wave energy farms off the coast of Southern California: an integrated site selection methodology

PDF

Selection of bridge location over the Merrimack River in southern New Hampshire: a comparison of site suitability assessments

PDF

Spatial analysis of veteran access to healthcare in Los Angeles County

PDF

A site suitability analysis for an inland port to service the ports of Los Angeles and Long Beach

PDF

Suitable California Opportunity Zone (COZ) locations for affordable housing development in the cities of Bakersfield, Los Angeles, and Palmdale

PDF

Assessing the connectivity for the jaguar (Panthera onca) in the United States-Mexico border ecoregions using species distribution modeling and factorial least cost path analysis

PDF

Use of GIS for analysis of community health worker patient registries from Chongwe district, a rural low-resource setting, in Lusaka Province, Zambia

PDF

A model for emergency logistical resource requirements: supporting socially vulnerable populations affected by the (M) 7.8 San Andreas earthquake scenario in Los Angeles County, California

Asset Metadata

Creator Li Sarain, Ada Yue (author)

Core Title Providing a new low-cost primary care facility for under-served communities: a site suitability analysis for Service Planning Area 6 in Los Angeles County, California

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 07/19/2019

Defense Date 03/29/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag 2SFCA,fuzzy overlay,low-cost primary care,OAI-PMH Harvest,primary care accessibility,site suitability analysis

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sedano, Elisabeth (committee chair), Fleming, Steven (committee member), Vos, Robert (committee member)

Creator Email adalisarain@gmail.com,yli257@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-183876

Unique identifier UC11660419

Identifier etd-LiSarainAd-7564.pdf (filename),usctheses-c89-183876 (legacy record id)

Legacy Identifier etd-LiSarainAd-7564.pdf

Dmrecord 183876

Document Type Thesis

Format application/pdf (imt)

Rights Li Sarain, Ada Yue

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA