University of Southern California Dissertations and Theses

Spatial analysis of vision services of Kaiser Permanente members

(USC Thesis Other)

Spatial analysis of vision services of Kaiser Permanente members

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content i

Spatial Analysis of Vision Services of Kaiser Permanente Members

by

Christine Placencia

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 2018

ii

Copyright ® 2018 by Christine Placencia

iii

To my husband, Michael, for all his patience and support.

iv

Table of Contents
List of Figures ................................................................................................................................ vi
List of Tables ................................................................................................................................ vii
Acknowledgements ...................................................................................................................... viii
List of Abbreviations ..................................................................................................................... ix
Abstract ........................................................................................................................................... x
Chapter 1 Introduction .................................................................................................................... 1
Motivation ............................................................................................................................4
Objectives ............................................................................................................................6
Thesis Organization .............................................................................................................6
Chapter 2 Related Work.................................................................................................................. 8
Kaiser Permanente ...............................................................................................................8
2.1.1. Vision Essentials by Kaiser Permanente....................................................................9
2.1.2. Preventative Care in Eye Health ..............................................................................10
Healthcare Accessibility ....................................................................................................10
2.2.1. Spatial Accessibility.................................................................................................11
Spatial Analysis of Accessibility in Health Care ...............................................................12
2.3.1. Gravity-based Models ..............................................................................................13
2.3.2. Floating Catchment Area Methods ..........................................................................16
2.3.3. ZIP Codes v Census Tracts ......................................................................................20
Chapter 3 Data Sources and Methodology ................................................................................... 21
Research Design .................................................................................................................21
Data Sources ......................................................................................................................21
3.2.1. Patient Data ..............................................................................................................22
3.2.2. Vision Essentials Optical Centers ............................................................................23
The Enhanced 2 Step Floating Catchment Area (E2SFCA) Method.................................26
v
3.3.1. Zip Codes and Census Tracts ...................................................................................26
3.3.2. Catchment Areas .....................................................................................................28
3.3.3. Data Analysis ...........................................................................................................29
Chapter 4 Results .......................................................................................................................... 31
Membership and FTEs (Exam Room Availability) ...........................................................31
Drive Time Zones ..............................................................................................................34
The E2SFCA Method Predictions .....................................................................................36
4.3.1. ZIP Codes.................................................................................................................36
4.3.2. Census Tracts ...........................................................................................................39
Chapter 5 Discussion and Conclusions ......................................................................................... 42
Census Tracts v ZIP Codes ................................................................................................42
The E2SFCA Method ........................................................................................................44
Final Remarks ....................................................................................................................46
References ..................................................................................................................................... 48

vi
List of Figures
1 Southern California Kaiser Permanente MSAs ........................................................................... 2
2 “Demand” – Potential accessibility to general practitioners .................................................... 15
3 Example of floating catchment area showing the catchment created around the centroids for
census tracts 2 and 3 using Euclidian distance ...................................................................... 17
4 2SFCA method example showing catchments around physician points a and b using drive-time
of 30 minutes ......................................................................................................................... 19
5 VE Optometry departments located within the Riverside MSA ................................................ 22
6 ZIP Code boundaries.................................................................................................................. 24
7 Census tract boundaries ............................................................................................................ 25
8 ZIP code centroids ..................................................................................................................... 27
9 Census tract centroids ................................................................................................................ 28
10 ZIP code centroids and the weights used to represent VE membership .................................. 32
11 Census tract centroids and the weights used to represent VE membership ............................. 32
12 Display of how the census tract is mismatched with the MSA boundary and ZIP code
polygons ................................................................................................................................ 33
13 Drive-time zones of 10, 20, and 30 minutes around VE providers.......................................... 34
14 Drive-time zones of 10, 20, and 30 minutes around ZIP code centroids ................................ 35
15 Drive-time zones of 10, 20, and 30 minutes around census tract centroids ............................ 35
16 Spatial accessibility index of VE members by ZIP code ........................................................ 38
17 Spatial accessibility index of VE members by census tract .................................................... 41

vii
List of Tables
1 Example attributes of U.S. Department of Housing and Urban Development’s USPS ZIP code
crosswalk file ........................................................................................................................ 25
2 Calculations used to estimate FTEs per VE location ................................................................. 34
3 Step 1 – Supply availability, ZIP code centroids within provider drive-time zones ................. 37
4 Step 2 – Demand, ZIP polygons containing VE locations within Member drive-time zones ... 38
5 Step 1 – Supply availability, census tract centroids within provider drive-time zones ............. 39
6 Step 2 – Demand, census tract polygons containing VE locations within member drive-time
zones ...................................................................................................................................... 40

viii
Acknowledgements
I would like to acknowledge and share my deepest gratitude to my advisor, Dr. John
Wilson, for his patience and guidance on this project, and the two members of my thesis
guidance committee, Drs. Laura Loyola and An-Min Wu, for their honesty and guidance as well.
I would also like to acknowledge the Vision Essentials department of the southern California
region, especially Mitch Rutledge, of Kaiser Permanente for allowing me the use of their data for
this project.

ix
List of Abbreviations
E2SFCA Enhanced 2 Step Floating Catchment Area
FCA Floating Catchment Area
FTE Full-Time Equivalent
HSPA Health Service Planning Area
HUD U.S. Department of Housing and Urban Development
IRB Institutional Review Board
KP Kaiser Permanente
MRN Medical Record Number
MSA Medical Service Area
PHI Patient Health Information
SDS Specialist Data Solutions
VE Vision Essentials by Kaiser Permanente

x
Abstract
Research has often examined geographical barriers to healthcare accessibility. These
examinations, however, are usually focused on primary care and urgent or specialty care. This
study focuses on access to vision care and services with the goal of bridging the gap in research
for this category of healthcare. Spatial accessibility for Kaiser Permanente members was
examined using the Enhanced 2 Step Floating Catchment Area (E2SFCA) method. This method
has been used in previous studies to examine spatial accessibility of patients to healthcare
services. It examines both supply (the amount of services or providers available to provide
services) and demand (patients who may or have used such services). This study also examined
the differences between using ZIP codes and Census tracts as the base geography and for
understanding how this choice is likely to affect the performance of the E2SFCA method and the
final outputs. The analysis showed that the southern region of the Riverside Medical Services
Area (MSA) has a shortage of optical services and that members must travel longer distances for
these services. Future research should further analyze the accessibility of the members living
within the Riverside MSA to vision services offered by Vision Essentials of Kaiser Permanente.

1

Chapter 1 Introduction
Kaiser Permanente is known for the integrated, high quality healthcare that they provide,
as well as their superior work environments which aids in their efficiency (McHugh, Aiken,
Eckenhoff & Burns, 2016). One part of this healthcare includes optical services provided by
Vision Essentials, a department within Kaiser Permanente. All Kaiser Permanente members have
access to basic eye exams as a part of their medical benefits. There are supplementary optical
benefits available to members that may be purchased in addition to their medical insurance
which may, in turn, provide additional funds for eyeglasses or contact lenses. The optometry
departments that offer these services, however, are not evenly dispersed and therefore not as
easily accessible to all Kaiser Permanente members.
This study examines the spatial accessibility to Vision Essentials optometry department
services in the Riverside Medical Service Area (MSA) of the southern California region. The
objective is to determine areas of low spatial accessibility to Vision Essential optometry
departments using the Enhanced 2 Step Floating Catchment Area (E2SFCA) method. An
analysis for the department has already been completed, focused on this area, examining the
members that live in this area and which medical centers they are visiting. A comparison
between these findings and the findings from this study will be provided as well.
Kaiser Permanente is different from some other medical service providers in that it is
both the provider and insurer. Members will most likely only go to Kaiser Permanente facilities,
as receiving service from a different provider is considered outside of the insurance network,
which may mean that the patient would incur increased out-of-pocket costs. Due to this factor, it
is important to identify areas of low accessibility in hopes that new centers can be developed to
provide a more even distribution of access to the members.
2

Vision Essentials is the branch of Kaiser Permanente that provides vision care services to
its members. There are 47 optometry departments available to members throughout the southern
California region. This is an area that stretches from Kern County to San Diego County; which is
divided into 12 MSAs (see Figure 1). The optometry departments are not dispersed evenly
through the region, and they are also not available at every Kaiser medical center.

There are three components that must be met in order to provide spatial accessibility.
This includes the supply of services or providers, demand for supply, and the distance or time
barriers to the healthcare locations (Luo & Wang, 2003; Jamtsho & Corner, 2014; Becker, 2016).
When analyzing accessibility as it pertains to healthcare services, there are two possible types
identified: revealed and potential accessibility (Khan, 1992; Luo & Wang, 2003). Revealed
Figure 1: Southern California Kaiser Permanente MSAs
3

accessibility relates to services that have been identified as being utilized. Potential accessibility
includes the services that may be used but are not guaranteed. This study examined revealed
accessibility to vision services. Access has spatial and non-spatial features (Khan, 1992). Spatial
access takes into account geographic barriers or facilitators, such as a distance variable, and non-
spatial features include barriers and moderators such as income or social class (Khan, 1992; Luo
& Wang, 2003).
As stated earlier, Kaiser Permanente is unique in that it is both the insurer and provider.
Members will typically not access outside medical services unless a referral is received from
their primary care provider. As with other insurance providers, patients are responsible for
medical costs that are obtained outside of their network unless a referral has been issued and
approval received from the insurance provider. This study focuses solely on Kaiser Permanente
members. This simplifies the analysis in that other service providers did not have to be included
in the study, as it is most likely that members will not seek outside care for eye exams. This is
due to two reasons: (1) the basic eye exam is covered under the primary medical insurance and
additional optical coverage is not necessary; and (2) to receive outside provider services would
mean out-of-pocket costs for the member. No other population group was examined for this
particular study.
The E2SFCA method was used to examine spatial accessibility and health plan service
areas (HPSAs) in this study as it is a vector-based method that allowed for the examination of
both spatial and non-spatial variables (Jamtsho & Corner, 2014). ZIP code and census tract areas
were both used as map units in this study. Analysis examining large service areas, such as
counties, are not able to distinguish the detailed spatial variations that may occur when studying
spatial accessibility (Luo & Wang, 2003; Luo, 2014). Only ZIP code data was available for both
4

members and optometrists. Membership numbers were available by ZIP code of patient
residence; optometrists were linked to the ZIP code of the optometry department in which they
provided services. Since only ZIP code data was available the patient membership data was
converted to census tracts using the U.S. Department of Housing and Urban Development
(HUD) crosswalk file. Weighted centroids of population were created for each census tract and
ZIP code. The optometrists and optometry departments were weighted to determine exam room
capacity.
The data obtained from Kaiser Permanente included member demographics, optometry
and medical center locations, type of vision care services, and additional attributes for each
dataset. Once obtained, data was recoded to remove member personal health information to
follow all HIPAA requirements. HIPAA requires that personal health information remain
private. ZIP code data for members was used in this study and is still considered to be personal
health information, although low risk. Approval from Kaiser’s Institutional Review Board (IRB)
was obtained prior to performing the analysis reported in this thesis.
Motivation
There is limited background research examining spatial accessibility to vision services.
This study will work to bridge the gap of research on vision services and to add to the validity of
the E2SFCA method. The importance of filling the gap of research for vision services lies with
the significance of preventative care. One of the Healthy People 2020 initiatives is to improve
the visual health of people through preventative care and early detection and treatment (U.S.
Dept of Health and Human Services, 2017). As people age, bodies begin to change and need to
be monitored for any adverse effects that could affect health. The eyes are important to monitor
as an individual gets older to catch eye diseases at an early stage. There are many types of
5

disease that could occur which include, but are not limited to, presbyopia or far-sightedness,
glaucoma, and cataracts.
As people age, the need for optical services for preventative care and/or disease
maintenance increases. By 2020, it is estimated that 2.3 billion people will be affected by
presbyopia which is but one example of a disease that can develop (NewsRX LLC, 2014).
Although most persons over the age of 45 years suffer from presbyopia, more than half are not
receiving the care they need to correct the issue (NewsRX LLC, 2014). Disparities in access to
vision services prevent individuals from receiving the care they need for correction which
includes the elderly population (Umfress and Brantley, 2016). Access to healthcare is promoted
by having a provider that is identified within a community (Wyn, Teleki, & Brown, 2000).
Riverside County has a population of about 2.4 million with 13.9% being over 65 years old (U.S.
Census Bureau, 2016).
Preventative maintenance in health, such as vision care, is important to catch before
issues become irreparable. Although many of these studies focus on access to healthcare services
such as primary care; there has not been any examining the importance of access to vision
services. It is important to monitor vision health as many individuals suffer from some form of
eye disease or illness. One study using a small sample size of 152 residents found that only 62%
of adults older than 40 years received eye care services in Los Angeles County (Baker,
Bazargan, Bazargan-Hejazi, and Calderón, 2009). They also found that having regular providers
available was significantly and positively associated with vision care utilization. Although this is
a study with a small sample, it is still important to note any consequence that could arise from
lack of availability of services.
6

Of the 47 optometry departments in southern California, three are located within the
Riverside MSA. Many patients within this MSA are required to travel more than 30 minutes to
receive care from a Kaiser Permanente vision center or must visit a non-Kaiser facility. Neither of
these options is convenient for the patient and puts a strain on accessibility for the members.
Objectives
The main objective of this study was to determine the areas of low spatial accessibility to
Vision Essential optometry departments for Kaiser Permanente members in the Riverside MSA.
This study worked to bridge the gap in research on accessibility to vision services, as well as to
provide information to Vision Essentials for future expansion in areas found to be low in spatial
accessibility. This thesis also worked to increase the validity and flexibility of the E2SFCA
method, as well as provide a comparison of the model results at the ZIP code and Census tract
units of analysis.
Through analysis of Kaiser Permanente members and the optical services provided, this
thesis examined the spatial accessibility of members using ratios of patients to optometrists and
optometrists to members for each catchment. Using distance decay to display a more realistic
measure of access, the levels of spatial accessibility per catchment area were created. Without
the distance decay, it is assumed that all members found within each of the three zones would
have the same kind of access to the services. Estimating the decay as the distance increases
mimics the unequal access that the members in different zones would encounter in their everyday
lives.
Thesis Organization
The remainder of this thesis unfolds as follows. Related work on spatial accessibility to
healthcare services, background on Kaiser Permanente, and gravity-based models used to
7

examine spatial accessibility are reviewed in Chapter 2. Chapter 3 describes the methodology
used for this study including the data needs, study area, and data sets that were used. Chapter 4
describes the findings of this study and discusses their significance given the objectives
described earlier. The fifth chapter offers some conclusions and ideas for future work.

8

Chapter 2 Related Work
Accessibility to healthcare services is influenced by both physical and socioeconomic
factors (Joseph & Phillips, 1984). Socioeconomic influences include a person’s ability to afford
services or the institution’s permitted use of them. Physical accessibility means that the service
should be both available for use and easily reached by the individual seeking to use the service.
These factors can either create a barrier or facilitate the utilization of healthcare services. There
is an importance in identifying these influences. Barriers would need to be overcome to increase
accessibility.
There are many different models that have been used to analyze spatial accessibility for
local populations. Each of these approaches, seemingly growing off of one another; examines the
relationship between supply and demand, taking into account spatial variables that may either
impede or facilitate use of services (Jamtsho & Corner, 2014). The development of these
approaches will be discussed later in this chapter. The size and type of study area will likely
determine which approach would work best in analyzing spatial barriers. Research on
accessibility to healthcare services has been examined through regional availability and
accessibility (Joseph & Phillips, 1984). Each of these approaches is discussed in more detail
later.
Using the Enhanced 2 Step Floating Catchment Area method, this study measured the
spatial accessibility of services for Kaiser Permanente members in the southern California
region, specifically the Riverside county area.
Kaiser Permanente
Kaiser Permanente has become one of the largest non-profit health plans in the U.S.
Established in 1945, the organization became the first of its kind allowing affordable health care
9

services to individuals that would otherwise not receive any. It was developed through the
collaboration of Dr. Sidney Garfield and Henry J. Kaiser in which a pre-payment system was
created in exchange for medical services to the employees of Kaiser’s Shipyard. From this
modest beginning, it has grown into an organization that serves more than 11 million members
throughout the U.S. (Kaiser Permanente, 2016).
According to their 2015 Annual Report, the organization has 38 hospitals and about 622
medical office buildings throughout the nation (Kaiser Permanente, 2015). It has continuously
ranked high among all hospitals in care and satisfaction. Both the northern and southern
California regions of Kaiser Permanente have won four-star ratings in overall clinical
effectiveness from California’s Office of the Patient Advocate for the past 10 years (Kaiser
Permanente, 2017). They are the only health plan in California to earn the highest rating
possible.
As stated earlier, Kaiser Permanente is unique in that is both the health plan and the
healthcare provider. Members will typically not access outside medical services unless a referral
is received from their primary care provider. As with other insurance providers, patients are
responsible for medical costs that are obtained outside of their network unless a referral has been
issued.
2.1.1. Vision Essentials by Kaiser Permanente
Vision Essentials is a department within Kaiser Permanente that provides optical services
to its members. There are locations throughout the U.S.; however, this study will focus on the
Riverside MSA of the southern California region of Kaiser Permanente. Within this region, there
are 47 optical centers from Ventura to San Diego counties. Each optical center provides
preventative eye health care through examinations performed by optometrists.
10

Regular preventative care eye exams are covered under the patient’s medical insurance
which may help some patients overcome the barrier to services caused by financial stress. The
patient would only be responsible for paying the co-pay at the time of service. The co-pays can
range from $0 to $60 per visit depending on the type of insurance coverage the patient has. The
co-pay may actually be a barrier for some patients who may find it hard to pay any amount;
however, this study will focus on spatial accessibility.
2.1.2. Preventative Care in Eye Health
As people age, bodies begin to change and need to be monitored for any adverse effects
that could affect health. As an individual gets older, catching eye diseases in their early stages
could help in preventing further deterioration from occurring. There are many different types of
diseases that could occur which include but are not limited to presbyopia or far-sightedness,
glaucoma, and cataracts.
Many individuals are not receiving the preventative care that they need in order to
prevent further damage to their eyes. For example, although most persons over the age of 45
years suffer from presbyopia, more than half are not receiving the care they need to correct the
issue (News RX LLC, 2014). Disparities in access to vision services prevent individuals from
receiving the care that they need for correction (Umfress & Brantley, 2016).
Healthcare Accessibility
Access to healthcare services is reliant on different variables that include availability,
affordability, and geographical accessibility (Gao et. al., 2016). Access has both spatial and non-
spatial elements (Khan, 1992) and has been measured by the closeness between the provider and
the patient (Rosero-Bixy, 2004). Spatial access means that a patient has overcome barriers such
as distance and traffic congestion to access services. The non-spatial element refers to barriers
11

such as economic and/or behavioral variables. Even when all of these needs are met, one cannot
automatically assume that such services will be utilized. Geographical accessibility has been
found to be both a predisposing and enabling factor to whether or not individuals will receive the
care they need (Arcury et al., 2015).
There are two different types of accessibility: potential and revealed accessibility (Joseph
& Phillips, 1984; Khan, 1992; Gao et. al., 2016). Potential accessibility refers to services that are
available for use but does not automatically mean they are used. There would be probable use of
services if barriers are overcome. Revealed or realized accessibility are services that have been
utilized which means that any barriers deterring the use have been overcome.
2.2.1. Spatial Accessibility
There are three major factors that play into spatial accessibility which include: (1) the
supply of available healthcare services; (2) the demand by patients to use these services; and (3)
the geographical location of these services (Joseph & Phillips, 1984; Khan, 1992; Jamtsho &
Corner, 2014; Becker, 2015). Geographical location, in this sense, refers to how easily the
individual is able to get to the services as they may be impeded by time or distance. Supply of
healthcare services can be interpreted as supply of healthcare providers. The availability of these
resources influences the accessibility and utilization of them (Joseph & Phillips, 1984). The
demand refers to the utilization of services by patients.
The approach to be taken to analyze spatial accessibility typically depends on the level of
aggregation to be studied (Joseph & Phillips, 1984). Two common approaches are regional
availability and accessibility. Regional availability is the simpler of the two approaches and
examines the distribution of supply and demand of healthcare services throughout a region.
Regional accessibility also examines supply and demand, however, in more detail by looking at
12

the interactions between them. These interactions are analyzed spatially to determine
accessibility.
There are problems that arise when using either approach to analyze spatial barriers so
these must be taken into account before analysis to determine which model would be the best fit.
The drawback of regional availability is that it assumes patients will not visit facilities or obtain
services outside of a designated region (Joseph & Phillips, 1984; Luo & Wang, 2003). In other
words, it assumes that the boundaries are not permeable. It also makes the assumption that all
individuals within a region have equal access to services (Luo & Wang, 2003). Also, with larger
levels of aggregation, it is difficult to identify any variations that are occurring at smaller levels.
There are also limitations associated with using the regional accessibility approach. This
approach examines regional supply and demand through analysis of centroids in smaller regions;
therefore, accessibility is centered at these points within the region (Joseph & Phillips, 1984).
Although the regions are much smaller than in the regional availability method, the same
problem arises in which it assumes that all individuals within the region have the same access to
services. Many models have been proposed to try to rectify this issue including the E2SFCA
method which incorporates smaller zones within larger catchments.
Spatial Analysis of Accessibility in Health Care
Different methods have been developed to measure special accessibility to healthcare
services such as computation of ratios and distances, gravity-based methods, space-time
accessibility techniques, and kernel density methods (Guagliardo, 2004; Jamtsho & Corner,
2014;). Geographical Information Systems (GIS) have been a tool used to analyze need and
accessibility in healthcare allowing for researchers to combine both spatial and aspatial variables
(McLafferty, 2003). This study used the Enhanced 2 Step Floating Catchment Area Method to
13

examine the spatial accessibility of Kaiser Permanente patients to Vision Essential locations.
Many models have been used to examine the spatial accessibility of healthcare, as briefly
described below.
2.3.1. Gravity-based Models
The gravity-based method is one method that can be considered the basic formula used in
spatial accessibility examining regional accessibility (Joseph & Phillips, 1984; Jamtsho &
Corner, 2014). Regional availability recognizes the interactions between supply and demand.
Gravity model approaches are used to examine these interactions, taking into account the
location of the supply and demand, as well as distance. These two items must be specified when
utilizing this approach (Joseph & Phillips, 1984).
The regional accessibility approach differs, and is more complex. Regional accessibility
examines supply and demand by creating a ratio between the two variables. This simple
examination of accessibility, however, makes a few assumptions: (1) patients only access
healthcare services within the boundaries created; and (2) the method does not identify spatial
variations that could be occurring in smaller areas (Joseph & Phillips, 1984; Khan, 1992).
A simple gravity-based model was created by Walter G. Hansen when trying to develop a
method to determine a pattern between accessibility and residential development in city areas.
This model states that the accessibility of location A is influenced by the size of activity in
location B, as well as the distance between them (Hansen, 1959). In other words, the more
activity around location B and the closer they are, the higher the accessibility of location A:
𝐴 1,2
=
𝑆 2
𝑇 1−2
𝑥 (1)
where A
1,2
is the accessibility measure in zone 1 to an activity in zone 2, S
2
is the activity size
(such as number of people), and T 1-2 is the travel time between the two zones. The exponent is
14

supposed to explain or describe the effects of the travel time. This early model by Hansen only
considered the supply side of the equation and not demand. Other models did not take into
account the diversity of availability of supply or healthcare providers; for example, some
locations may have more providers to offer services or a certain location may offer different
types of services.
Joseph and Bantock (1982) applied the gravity model to healthcare accessibility and also
tried to capture this diversity by estimating the demand relative to supply by using assumed
population utilization for that area. They also used an index of potential physical access to the
practitioners. They believed that there are two approaches when analyzing healthcare
accessibility, specifically to general practitioners: (1) measures of revealed accessibility through
utilization data; and (2) the measurement of potential accessibility which they based off where
the patient lives in relation to the services. The formula they used to examine the potential
accessibility of patients was:
𝐴 𝑖 = ∑ 𝐺 𝑗 𝑃 𝑗 𝑑 𝑖𝑗
𝑏 ⁄ (2)
where A i is the potential accessibility of location i to the providers, GP j is the general practitioner
at location j within range of location i, d ij is the distance between locations i and j, and b is the
distance exponent. They also estimated the demand on doctors since there is a variability of
availability of providers in areas with differing population numbers:
𝐷 𝑗 = ∑ 𝑃 𝑖 𝑑 𝑗𝑖
𝑏 ⁄
𝑖 (3)
where D j is the demand on the provider at location j, P i is the population at location i within the
range of location j, and d ji reflects the distance between locations j and i (Figure 2).

15

Figure 2: “Demand” – Potential accessibility to general practitioners (Joseph & Bantock,
1982). The scores associated with the four maps refer to the potential physical accessibility
measurement with 0.0 representing zero accessibility.
16

.
Although accessibility was being examined from both a supply and demand perspective,
it still did not capture the full picture of spatial accessibility. Distance decay and distance ranges
were added to imitate the mobility of the population (Khan, 1992). In this gravity model,
weighted estimates of potential availability of providers were also introduced.
2.3.2. Floating Catchment Area Methods
The floating catchment area (FCA) methods have been used to measure accessibility as
well, but not necessarily in healthcare. Peng (1997) used a version of the FCA to examine jobs
and housing for example. In the FCA, boundaries are created around a specific area or point
creating catchments. For example, in Peng’s (1997) study, catchments were created around each
traffic zone, extending out 5 miles. Then jobs and housing were aggregated within each of these
catchments to determine the ratio of accessibility. In doing so, it was assumed that each resident
would have access to all of the employment opportunities within each of the catchments,
however, that is not always the case (Luo & Wang, 2003). When applying this method to
healthcare, it does not consider that providers may provide services outside of the proposed area
as well and therefore they may not be providing full services to only those residents in the
designated catchment areas (Figure 3).
The Two Step Floating Catchment Area (2SFCA) method proposed by Luo and Wang
(2003) took the FCA further. It examined both supply and demand in the same analysis instead
of just being one sided (i.e., considering supply only) such as in the gravity and early FCA
models. In this method, catchments were created around each census tract centroid, as well as the
provider location. Travel time was used instead of Euclidian distance to create catchments
around each of the centroids. A provider-to-population ratio is used to relate supply to demand.
17

The providers and residential populations are examined within boundaries with the numerator
being the supply such as provider capacity and the denominator being demand which could be
the population living near the facility or services (Luo & Wang, 2003; Guagliardo, 2004; Wang
& Luo, 2005; Dewulf, Neutens, De Weerdt, & Van de Weghe, 2013). Ngui and Apparicio (2011)
even added weights to the provider locations by using the number of providers available at each
location.

Figure 3: Example of floating catchment area showing the
catchment created around the centroids for census tracts 2 and 3
using Euclidian distance (adopted from Luo & Wang, 2003).
18

For the first step, a catchment is created around each healthcare provider centroid using a
specified distance or drive time. The total population found within each catchment is summed up
to create the provider-to-population ratio (supply):
𝑅 𝑗 =
𝑆 𝑗 ∑ 𝑃 𝑘 𝑘 ∈{𝑑 𝑘𝑗
≤𝑑 0
}
(4)
where R j is the provider-to-population ratio at location or catchment j, S j is the number of
providers at location j, d kj is the distance between locations k and j, and P k is the population
within the spatial area such as a census tract or ZIP code.
For the second step, a catchment is created around each population centroid using a
specified distance or drive time. The previous ratio of provider-to-population found within each
catchment is summed up to create the accessibility index (demand):
A
i
F
= ∑
j∈{d
ij
≤d
0
}
(5)
𝑅 𝑗 = ∑
𝑆 𝑗 ∑ 𝑃 𝑘 𝑘 ∈{𝑑 𝑘𝑗
≤𝑑 0
}
𝑗 ∈{𝑑 𝑖𝑗
≤ 𝑑 0
}
(6)
where 𝐴 𝑖 𝐹 is the accessibility at location i (residential), R j is the provider-to-population ratio at
location j who fell within the catchment of i, and d ij is the distance between locations i and j. See
Figure 4 for a visual example of the 2SFCA method. The disadvantage of this method is that it
assumes that all persons within the catchments have equal access to service providers and that all
service providers have equal access to the residential population (McGrail & Humphreys, 2009).
For example, a person living the closest to a boundary line has the same access as the person
living near the centroid of the catchment. Also, the measure is dichotomous, meaning that access
is either gained or not. There is no variation in the level of access. The Enhanced 2 Step Floating
Catchment Area (E2SFCA) method was created in order to combat these problems.
19

The E2SFCA proposed by Luo & Qi (2009) creates travel time zones within each
catchment and assigns different weights to each of these zones in an attempt to address the
disadvantage of the 2SFCA method. These zones and weights are to account for the distance
decay that occurs as the supply moves away from the population centroid or vice versa. A
Gaussian weight was used for each time zone in their study, however, different weights can be
used depending on the type of accessibility that is being examined.

Figure 4: 2SFCA method (Luo & Wang, 2003) showing the catchments
created around physician points a and b using a drive-time of 30 minutes.
20

The difference between the 2SFCA and the E2SFCA method are the zones that are
created within the catchment. These three extra zones are created on both the supply and demand
sides of the analysis:
𝑅 𝑗 =
𝑆 𝑗 ∑ 𝑃 𝑘 𝑊 𝑟 𝑘 ∈{𝑑 𝑘𝑗
∈𝐷 𝑟 }
=
𝑆 𝑗 ∑ 𝑃 𝑘 𝑊 1
+∑ 𝑃 𝑘 𝑊 2
+∑ 𝑃 𝑘 𝑊 3
𝑘 ∈{𝑑 𝑘𝑗
∈𝐷 3
} 𝑘 ∈{𝑑 𝑘𝑗
∈𝐷 2
} 𝑘 ∈{𝑑 𝑘𝑗
∈𝐷 1
}
(7)
where the difference lies in the weighted zones that have been added with the addition of the W r
term. Equation 8, on the other hand, incorportes three zones to simulate the distance decay effect:
𝐴 𝑖 𝐹 = ∑ 𝑅 𝑗 𝑊 𝑟 𝑗 ∈{𝑑 𝑖𝑗
∈𝐷 𝑟 }
= ∑ 𝑅 𝑗 𝑊 1 𝑗 ∈{𝑑 𝑖𝑗
∈𝐷 1
}
+ ∑ 𝑅 𝑗 𝑊 2 𝑗 ∈{𝑑 𝑖𝑗
∈𝐷 2
}
+ ∑ 𝑅 𝑗 𝑊 3 𝑗 ∈{𝑑 𝑖𝑗
∈𝐷 3
}
(8)
The Huff model has also been used in within the FCA method as a third step in order to
account for the probability that a patient may go somewhere else for services (Luo, 2014).
However, for this study this will not be considered as most patients will stay within the Kaiser
Permanente network.
2.3.3. ZIP Codes v Census Tracts
The use of ZIP codes for spatial and socio-economic analysis has increased through the
years, however, they must be used with caution (Grubesic, 2015). The size of ZIP codes, as
spatial units, changes depending on the area under examination. For example, in rural areas the
size of ZIP code areas may be larger than those used within urban areas. Riverside is considered
a rural area and both ZIP codes and census tracts were examined in this thesis project. There are
few studies examining spatial accessibility using ZIP codes spatial units. As the E2SFCA method
uses intervals within the catchments, census tract and ZIP codes were both examined to compare
any differences that may arise from the choice of spatial unit.

21

Chapter 3 Data Sources and Methodology
This project aimed to identify the spatial accessibility levels of areas in the Riverside
MSA for Kaiser Permanente members to Vision Essentials by Kaiser Permanente (VE)
optometry departments using the Enhanced 2 Step Floating Catchment Area method. The
research design, data sources, area of study, and method used for this of analysis are described
below.
Research Design
The service areas for Kaiser Permanente are delineated using the nearest medical service
centers. Patients living within the Riverside MSA were the focus of this project. Currently, there
are three VE optical centers located within the Riverside MSA in the cities of Corona, Moreno
Valley, and Riverside (Figure 5). The E2SFCA method was used to examine spatial accessibility
of Kaiser’s patients living within the Riverside MSA boundary. The MSA boundary has been
predetermined by Kaiser Permanente.
Data Sources
Data on patient visits were gathered through the point-of-sale system called Specialist
Data Solutions (SDS) provided by VE by Kaiser Permanente. Appropriate steps were taken to
obtain Institutional Review Board (IRB) approval from Kaiser Permanente for the use of
personal health information (PHI). The steps taken to ensure protection of PHI will be discussed
later. The residential location data was obtained through spreadsheets maintained and made
available by VE.

22

3.2.1. Patient Data
The patient visit details were obtained in .xls format and pulled from the VE point-of-sale
system called SDS. IRB approval was required before data could be used in this study. The
patient’s name and medical record number (MRN) are considered unique identifiers and were
removed before analysis to protect PHI, as required by law. A preliminary study outline was
submitted as a part of the IRB approval process. Additional steps were required to be taken to
ensure patient privacy for the protection of the patient’s identity which included a second review
Figure 5: VE Optometry Departments located within the Riverside
MSA
23

of data to ensure PHI was removed. ZIP codes and visit details are still considered risky,
although minimal, and therefore required IRB approval for usage. This information remained for
analysis after approval was granted.
The patient visit detail is a compilation of data that includes date of visit, location of visit,
home medical service area, home medical office building or MOB, type of procedure, length of
procedure, and the patients’ ZIP code in 2016 (Figure 6). This dataset displayed Kaiser
Permanente members that have already visited an optical center and therefore represents revealed
accessibility. A listing of all members by ZIP code was not approved for use in this study.
Therefore, only the members that had already received services at one of the optometry
departments were counted. As the patient detail will only provide ZIP codes for members, each
was proportionally assigned to census tracts using the ZIP code tabulation area cross-walk
provided by the U.S. Department of Housing and Urban Development (HUD) which will be
discussed later (Figure 7). The patient visit detail data was provided in .xls format only and
therefore was formatted to be added and used in ArcGIS Business Analyst.
3.2.2. Vision Essentials Optical Centers
The addresses for each optical center were provided by VE in .xls format. This dataset
includes number of exam rooms for each optical center, as well as the addresses itself. All
addresses were geocoded in Esri’s Business Analyst, although only the optical centers located in
the Riverside MSA were examined for this study (see Figure 2). Optometrist FTEs were used to
weight each location.

24

The FTEs were calculated using exam room availability (see Table 1). For example, an
exam room available for 58 hours per week (10 hours per weekday and 8 hours on Saturday)
would yield 1.45 FTEs (58 hours / 40 hours for the optometrist = 1.45 FTE per exam room).
From this calculation, only 90% will be counted to account for any discrepancies that may occur
because of patient no-shows and other problems, leaving the yield at 1.31 FTEs per exam room.
This approach was used to generate realistic and conservative outputs in this thesis project. The
FTEs per exam room was then multiplied by the number of exam rooms to yield total availability
of provider for that location. This calculation was added to the attributes of the provider locations
as its own field.

Figure 6: ZIP code boundaries
25

Table 1: Example attributes of U.S. Department of Housing and Urban Development USPS ZIP
code crosswalk file

ZIP Tract RES_RATIO
92501 030100 0.35244300000
92501 030200 0.28240400000
92501 030300 0.23990500000
92501 030700 0.00330994000
Figure 7: Census tract boundaries
26

The Enhanced 2 Step Floating Catchment Area (E2SFCA) Method
The Enhanced 2 Step Floating Catchment Area (E2SFCA) Method is a type of gravity-
based model (Luo & Wang, 2003). It has been used to examine the level of spatial accessibility
that patients have to healthcare services. It takes both supply and demand into consideration
when determining accessibility. For this study, the calculated FTEs per location will represent
supply and the number of Kaiser Permanente patients that have received care through each VE
per ZIP code or census tract will represent the demand.
3.3.1. Zip Codes and Census Tracts
The membership data was available through the patient visit details. This dataset
provided ZIP codes for each member which were summed and joined to the ZIP code polygons
used by the U.S. Postal Service and obtained from ArcGIS Online. The ZIP codes with
membership totals were clipped to the areas within the Riverside MSA. The centroids within
each of the ZIP code polygons were created (Figure 8). They were weighted by the VE
membership.
Census tract information for membership was not available and had to be estimated using
the HUD USPS ZIP code crosswalk file. There were multiple census tracts found within each
ZIP code creating a many-to-one relationship. Table 1 showed a snippet of the crosswalk file
which shows the corresponding census tract areas for a single ZIP code, along with the ratio of
residential population that should be distributed. In order to obtain membership for the census
tracts, first the crosswalk file and the patient visit detail files had to be merged. The already
clipped ZIP code membership data was exported and then matched to the crosswalk file using
MATCH and VLOOKUP. Membership was then calculated for each census tract area using the
residential ratio or RES_RATIO using Microsoft Excel.
27

Once the breakdown for membership was obtained it was then pulled into Esri’s Business
Analyst and joined to the census tract shapefile obtained from ArcGIS Online which was then
clipped to the area contained within the Riverside MSA. Centroids were created within each census
tract (Figure 9). These centroids were also weighted by the membership.

Figure 8: ZIP code centroids
28

3.3.2. Catchment Areas
According to a study of 5,000 adults, people are willing to travel up to about 30 minutes
or about 20.4 miles for future non-urgent health care (Yen, 2013). Therefore, zones of 10, 20,
and 30 minutes were created around each of the population centroids and optometry offices. For
the E2SFCA method, catchments were created for both members and optometrists.
The E2FCA required sub-group catchments representing different time thresholds within
the larger catchments. These zones were created in both the ZIP code and census tract analyses.
They were split into 10 minute distance increments: 0-10 minutes for Zone 1, 10-20 minutes for
Zone 2, and 20-30 minutes for Zone 3. This was to allow for a more accurate examination of spatial
Figure 9: Census tract centroids
29

variations occurring in spatial accessibility (Luo & Qi, 2009). Each zone was weighted to account
for distance decay using the Gaussian function (Luo & Qi, 2009). The Guassian function has a
smoothing property where the exponent for distance equals 2 which takes into account the idea of
space as well as the movement of populations (de Smith, Goodchild & Longley, 2015; Salze et al.,
2011). The time buffers were created using the Trade Area Tool within Esri’s Business Analyst.
3.3.3. Data Analysis
The measurement for accessibility is created using the ratio of optometrist to member
within each catchment using the 2 step approach described below. The ratio of optometrist to
membership for the first step was calculated by summing up all points found within each zone or
buffer around the optometrist locations using:
𝑅 𝑗 =
𝑆 𝑗 ∑ 𝑃 𝑖 𝑊 𝑟 𝑖 ∈(𝑑 𝑖𝑗 ∈𝐷 1)
=
𝑆 𝑗 ∑ 𝑃 𝑖 𝑊 2
+∑ 𝑃 𝑖 𝑊 2
+∑ 𝑃 𝑖 𝑊 3 𝑖 ∈(𝑑 𝑖𝑗 ∈𝐷 3
) 𝑖 ∈(𝑑 𝑖𝑗 ∈𝐷 2)
𝑖 ∈(𝑑 𝑖𝑗 ∈𝐷 1
)
(9)

where R j represents the provider-to-membership ratio found within catchment locations j, P i is the
population total within each census tract or ZIP code that is found within catchment j, S j is the
total providers/calculated FTEs at each location j, d ij is the distance between locations i and j,
and W
r
represents the weight of each zone using the Gaussian function.
Once the zones were created around each of the provider locations, the provider-to-
population ratio was calculated by completing a spatial join between the zones and the centroids
of both census tracts and ZIP codes, separately. In the tool, a one-to-many relationship was
established in order to pull each individual centroid that falls within each of the zones. A new
field was created to perform the calculation of FTE with distance decay divided by the
membership to give us the provider-to-population ratio.
30

The data were next joined back to the VE locations using Join and Relate creating a new
shapefile containing the ratio. The second step was used to sum up the number of providers
found within each catchment around the membership centroids as shown in the following
equation:
𝐴 𝑖 𝐹 = ∑ 𝑅 𝑗 𝑊 𝑟 𝑗 ∈(𝑑 𝑖𝑗
∈𝐷 1)
= ∑ 𝑅 𝑗 𝑊 1
+ ∑ 𝑅 𝑗 𝑊 2
+ ∑ 𝑅 𝑗 𝑊 3 𝑗 ∈(𝑑 𝑖𝑗
∈𝐷 3
) 𝑗 ∈(𝑑 𝑖𝑗
∈𝐷 2)
𝐾 ∈(𝑑 𝑖𝑗
∈𝐷 1
)
(10)
where 𝐴 𝑖 𝐹 is the accessibility to providers for members found within location or catchment i, and
d ij is the distance between locations i and j. The Gaussian weights, represented by W r, were used
in step 2 as well. For this step, after the zones were created around each of the membership
centroids (census tracts and ZIP codes were handled separately), a spatial join was used to
connect with the new shapefile created in the previous step. A many-to-one relationship was
created in order to pull in each centroid that falls within the zones. Once joined, the spatial
accessibility index was calculated by summing up the provider-to-member ratio that was
obtained in the previous step.

31

Chapter 4 Results
This chapter reviews the results obtained through the examination of spatial accessibility
using the Enhanced 2 Step Floating Catchment Area Method in the Riverside MSA. The spatial
accessibility of Kaiser Permanente (KP) members to optometry departments were analyzed using
tabulations of members by ZIP code and census tract.
Membership and FTEs (Exam Room Availability)
Both the centroids for the census tracts and ZIP codes and the optometry locations were
weighted in order to create provider-to-population ratios that would become the spatial
accessibility index. The patient visit detail data contains a total of 707,216 KP members that
have received VE services within the southern California region. There are 2,029 patients who
have visited this region but live outside of the area. Of all the patients, 78,741 members live in
the Riverside MSA. There were 63 ZIP codes in the Riverside MSA and of these, 62 were
occupied by patients that had visited at least one of the Riverside MSA VE locations during
2016. The membership data was joined to the ZIP code shapefile to create the weighted
centroids (Figure 10).
Membership was not available by census tracts and therefore was redistributed using the
U.S. Department of HUD’s fourth quarter 2016 USPS ZIP crosswalk file as described earlier.
The redistribution was then joined to the census tract shapefile for which weighted centroids
were then created (Figure 11). All 78,741 members were able to be distributed into the census
tracts provided by the crosswalk. However, once clipped, the census tract membership lost
19,308 members due to those that lived outside of the Riverside MSA boundary. Figure 12
32

shows how the census tracts matched up to the Riverside MSA. There was a total of 59,434
members that were retained within the MSA, and a total of 333 census tracts that were counted.
Figure 11: Census tract centroids and the weights used
to represent VE membership
Figure 10: ZIP code centroids and the weights used to
represent VE membership
33

The membership was joined to the clipped census tract boundaries to create the weighted
centroids.
The weights for the optometry locations were created by calculating FTEs through exam
availability. A full time optometrist works 40 hours per week which is equal to one FTE. The
hours of operation determined the number of hours per week an exam room was open. The
weighted FTE for each location was calculated at a 90% utilization rate for each exam room.
Table 2 shows the breakdown of room numbers for each location, as well as the number of hours
they are open per week. The Corona location had a weighted yield of 9 FTEs, the Moreno Valley
location had 4.5 FTEs, and the Riverside location had 13.1 FTEs available.
Figure 12: Display of how the census tract is mismatched with the
MSA boundary and ZIP code polygons.
34

Table 2: Calculations used to estimate FTEs per VE location
Drive Time Zones
Drive time zones in 10 minute increments (i.e. 0-10, 10-20, and 20-30 minute intervals)
were created around the optometry locations and the membership centroids of the census tracts
and ZIP codes and clipped to the Riverside MSA boundary (Figures 13-15).

Location
# of Exam
Rooms
# of
Hours/Week
Open
Full-Time
Optometrist
(Hrs)
Utilization
(%)
FTE
Weighted
FTE
Corona 8 50 40 90 1.13 9.0
Moreno
Valley
4 50 40 90 1.13 4.5
Riverside 10 58 40 90 1.31 13.1
Figure 13: Drive-time zones of 10, 20, and 30 minutes around
VE providers

35

Figure 15: Drive-time zones of 10, 20, and 30 minutes around census
tract centroids
Figure 14: Drive-time zones of 10, 20, and 30 minutes around ZIP
code centroids
36

The zones were created using the Trade Area tool within Esri’s Business Analyst. Weights
were given for each drive zone (zone 1: 1.0; zone 2: 0.6; and zone 3: 0.2) in order to account for
distance decay which was based off the Gaussian decay function. These weights were used as they
created a slower distance decay (Luo & Qi, 2009). Riverside has more rural areas throughout the
MSA and therefore distance may not be as much of a hindrance for patients as they will need to
do some traveling for services anyway. So the slower distance decay would not create drastic
changes in accessibility as the distance increases away from the provider locations.
The E2SFCA Method Predictions
The E2SFCA model runs and the subsequent analysis for ZIP codes and census tracts
were completed separately.
4.3.1. ZIP Codes
The first step of the E2SFCA method, after the creation of the catchments and drive-
times around each of the service providers and population centroid locations, is to find all the
population centroids that fall within each of the service provider drive-time zones. A spatial join
was used to “catch” all of the population centroids. There were centroids that were counted more
than once when they fell into multiple zones. For instance, all the centroids that fell within the 0-
10 minute drive-time zone also were counted again in the 10-20 and 20-30 minute drive-time
zones. Table 3 displays the total numbers of centroids counted per zone, as well as the number of
centroids when removing the overlaps. Similarly, the number of members served per drive-time
also were double counted, however, the totals in Table 3 reflect the actual numbers per drive-
time zone without overlap. There was a combined estimated total of 76,825 members served and
their distribution across each of the drive-time zones is reported in the last column of Table 3.
37

Table 3: Step 1 – Supply Availability, ZIP Code Centroids within Provider Drive Time Zones
Step 1: Vision Essentials - Supply Availability
Drive-Time Zones
No. of Zip
Code
Centroids
Within
Single Zone
(No
Overlap)
Estimated
No. of
Members
Served
0-10 Minutes 7 7 15,518
10-20 Minutes 27 20 37,014
20-30 Minutes 39 12 24,293
Totals 73 39 76,825

Once the membership was added to the provider zones, the provider-to-population ratio
had to be calculated. A new field called Prov_POP was created and calculated by dividing the
provider availability, which reflects the distance decay, by the membership that fell within that
drive-time zone. The spatial accessibility ratio ranges from 0 to 1 with 0 meaning no access and
1 being full access. When looking at the spatial accessibility ratio, however, the actual totals are
far less than 1. This is due to the fact that there is a small number of providers that provide
service to a large number of members. To obtain a 1 would mean that there is exactly 1 provider
that is available for services for every 1 member. The provider-to-population was then joined to
the provider location centroids.
In step 2, the provider-to-population ratio (Prov_POP) was joined back to the VE location
centroids. A spatial join relationship was then created to capture all provider locations within each
of the ZIP code drive-time zones. Table 4 displays the number of ZIP code polygon drive-time
zones in which a VE location was captured. There was a total of 81 times that a VE location landed
within a zone without the overlap of drive-times. The provider-to-population ratio calculations
were summed for each ZIP code boundary that a provider fell into. This ratio was then joined to
the ZIP code boundary shapefile for display. Figure 16 shows the spatial accessibility for members
38

Table 4: Step 2 – Demand, ZIP Polygons containing VE location within Member Drive-Time
Zones
Step 2: KP Members -Demand (3 Locations)
Drive-Time Zones
Total No. of
Times VE
locations
Caught in
Zone
Within Single
Drive-Time
Zone (No
Overlap)
0-10 Minutes 8 8
10-20 Minutes 41 33
20-30 Minutes 81 40
Totals 130 81
Figure 16: Spatial accessibility index of VE members by ZIP code
39

within each ZIP code boundary. Members located in the southern region of the MSA have no
spatial accessibility to any VE location while members living in the northern half of the Riverside
MSA have greater but varying degrees of spatial accessibility.
4.3.2. Census Tracts
The analysis for the census tracts followed the same steps used when analyzing the ZIP
code spatial accessibility using the E2SCFA method. The first step was to count all census tract
centroids that fell within each of the provider drive-time zones. Table 5 displays the number of
centroids that was captured within each zone. Similar to the ZIP code analysis, census tract
centroids fell into more than one of the drive-time zone when they were closer to the provider
location. For example, a centroid that fell within zone 1 (0-10 minutes) would fall into every
drive time zone. There were a total of 240 centroids, without overlap, that fell into any one of the
drive-time zones. There was an estimated number of 50,456 members that were served within 30
minutes of drive times. This estimated number does not include members that may have landed
in more than one drive-time zone as they are only counted once.
Table 5: Step 1 – Supply Availability,
Census Tract Centroids within Provider Drive Time Zones

Step 1: Vision Essentials -Supply Availability
Drive-Time Zones
No. of
Census
Tract
Centroids
Within
Single Zone
(No
Overlap)
Estimated
No. of
Members
Served
0-10 Minutes 80 80 16,004
10-20 Minutes 186 106 25,129
20-30 Minutes 240 54 9,322
Totals 506 240 50,456

40

Step two was completed by joining the provider-to-population ratios to the VE location
points and then catching all of provider locations that fell within any one of the three drive-time
zones for the census tract centroids. Table 6 displays the number of census tract polygons that
the VE locations fell within the different drive-time zones. Without overlap, there were 507
census tracts that were touched by the VE locations.
Table 6: Demand, Census Tract Polygons containing VE location within
Member Drive-Time Zones

Step 2: KP Members -Demand (3 Locations)
Drive-Time Zones
Total No. of
Times VE
locations
Caught in
Zone
Within Single
Drive-Time
Zone (No
Overlap)
0-10 Minutes 93 93
10-20 Minutes 286 193
20-30 Minutes 507 221
Totals 886 507

The spatial accessibility index was again created by summing up the provider-to-member
ratios during the second step. As mentioned earlier, the ratio between provider and members will
be small given that there are few providers compared to the thousands of members. The final
map was then created by joining together the census tract boundaries and the spatial accessibility
ratios (Figure 17). Members residing in the southern region of the MSA have no spatial
accessibliity to any VE services located within the Riverside MSA boundary and northern
residents once again have varying degrees of accessiblity.

41

Figure 17: Spatial accessibility index of VE members by census tract

42

Chapter 5 Discussion and Conclusions
The spatial accessibility of the KP members is not evenly distributed among KP members
throughout the Riverside MSA. It is evident that the northern region of the MSA has much
greater access to VE providers than the central and southern parts. The results from this analysis
coincided with some analysis completed within the VE department. This MSA has been an issue
for the department because many of the KP members lack access to VE services and have to
travel to other MSAs to receive needed care.
It was estimated during VE in-house analysis that 16.3% of the patients residing in the
Riverside MSA seek services outside of the MSA. Many of these patients live within the
southern region. This region contains 20% of the total patient population in this MSA. The
MSAs that patients are traveling to for services include the Fontana, San Diego, and Orange
County MSAs; with a small percentage traveling to various other MSAs. This poses difficulties
and issues for the providers since it could mean a cut in the funds and/or services provided. This
study was completed in order to support the original findings completed by the department in a
preliminary study and although it did not examine the patient’s path to where they are receiving
services, it does identify those areas with service deficits.
Census Tracts v ZIP Codes
This study compared the results obtained from examining both census tracts versus ZIP
codes as the unit of analysis. Although ZIP codes are not a preferred unit of analysis, it was
important to examine them in this study because the patient data collected by Kaiser Permanente
is available by ZIP code only. Currently, Kaiser Permanente does not keep census tract or block
group information for its membership database. Therefore, exploring the differences between the
43

two could determine whether it is appropriate to use ZIP codes for this type of analysis of spatial
accessibility in future Kaiser Permanente studies.
One thing observed in this study was the difficulty of redistributing KP patients into
census tracts from ZIP codes using the HUD crosswalk file. The results in the previous chapter
showed that KP members were lost, notwithstanding the use of 2016 membership data, ZIP code
boundaries, census tract boundaries, and the HUD crosswalk file. An accurate join between the
members within the MSA and the ZIP code boundaries was obtained, however, the ZIP codes,
once clipped, did not match the clipped census tract boundaries perfectly. There was a total of
19,308 members that were lost due to this incompatibility. Figure 12 shows the distribution of
census tracts that hold the same numbers of membership that was distributed among the ZIP
codes but fall outside of the Riverside MSA boundary. The loss of large numbers of patients
affected the comparison of the two units of analysis. However, the analysis was still completed
to compare the distribution of the population throughout the MSA, as well as the distribution of
spatial accessibility as follows.
In both the ZIP code and census tract membership distribution, it is evident that the
majority of the population resides in the northern region of the Riverside MSA. There are a few
more clusters of membership in smaller areas of the southern region when looking at the census
tract distribution because denser settlement patterns will produce smaller census tracts.
When examining spatial accessibility, both maps show that the members living in the
southern regions of the Riverside MSA have no spatial accessibility to VE services provided
within the Riverside MSA boundary. However, the census tract boundaries provide a little more
detail in showing the distinctions between levels of access. This could have been caused by a few
factors: (1) the units are small enough to provide the extra detail or (2) the loss of membership
44

from the distribution could have altered the data and created larger provider-to-population ratios
in selected areas. The latter is likely the better explanation since the spatial accessibility index
values for the census tract analysis were larger on average because of the loss of KP members
due to the problems with the crosswalk noted earlier.
With the E2SFCA method, using a Gaussian distance decay as suggested by previous
studies (Luo & Wang, 2009) did have a modest effect on the calculation of spatial accessibility
after joining with both the census tract and ZIP code polygons. Due to the distance decay,
looking at both census tracts and ZIP codes, a gradual decrease in spatial accessibility can be
seen as distance increases between the member locations and the VE service locations. Although
it does allow for a gradual decline in spatial accessibility, and avoids designating access or not,
the approach still gives all individuals within a polygon the same spatial accessibility. This is
more evident in the ZIP code analysis than in the census tract analysis because the latter units
were much smaller.
The E2SFCA Method
Since total membership data was not available, realized data was used which means that
the data only containted patients that had already been seen by a service provider. For future
studies, total membership could be used along with realized accessibility in order to calculate a
more specific distance decay surface to support the analysis. For this study, slower decay weights
were used for distance decay under the assumption that KP members living in rural areas would
still need to travel for service despite the distances involved. However, there may be underlying
reasons or situations that change the decay speed. The development of distance decay
relationships using the data available in-house to Kaiser Permanente could help to reduce the
guess work incorporated with this step.
45

This study also did not take into consideration that patients may leave the Riverside MSA
to receive services elsewhere, whether with Kaiser in a different MSA or with an outside
provider. Per the analysis done within the department, it is known that patient’s from the
southern parts of the Riverside MSA, such as Temecula, do go outside of the Riverside MSA for
services. However, examining the potential of using services outside of a MSA is not considered
during analysis conducted by Kaiser Permanente as each MSA functions almost as its own
entity. Further analysis using a buffer that reaches outside of the Riverside MSA; as well as
including the VE locations within a 30 minute drive-time could provide a more accurate
assessment of spatial accessibility of VE services in the future because members are able to leave
the MSA to receive care.
The total Kaiser Permanente membership was not used in this study as it was not permitted.
Only a small portion of the membership was able to be used. With the total membership, a more
realistic idea of distribution of membership throughout the Riverside MSA would be obtained.
From the use of total membership, a better distance decay could be calculated to be used for future
analysis.
Another thing noticed during this study was the overlapping catchments which meant some
patients were captured multiple times. This topic was not discussed in the previous studies. Luo
and Wang (2003) vaguely skimmed the topic when discussing the transition from the FCA method
to the 2SFCA method in which Euclidian distances were used for catchments. In order to account
for this overlap, separate drive-time zones for both providers and membership centroids were
created. This allowed for the breakdown and analysis of each zone separately. These special steps
had to be taken in order to avoid counting centroids more than once in the totals. Any future
exploration will need to take extra steps to avoid this outcome as well.
46

Final Remarks
As stated earlier, ZIP codes are not the preferred unit of analysis. Many researchers have
reviewed the units and found many issues when using them. They are not discretely bounded areas
and could lead to modifiable area unit problems in which data and patterns may be distorted
(Grubesic, 2008). However, when using Kaiser Permanente data, ZIP code distributions are what
is available for analysis. Finding a proper way to redistribute this membership to census tracts or
other small units of analysis would be beneficial to any future spatial analysis. It may be beneficial
for healthcare providers to obtain census tract data on patients from the outset so no redistribution
needs to take place. More steps added to any procedure increases the risk of corruption and/or
error. A better analysis can be performed on the smaller census tract spatial units and is thought
to be more manageable for estimating travel time and analyzing accessibility (Luo, Wang, &
Douglass, 2004). This type of analysis may be something Kaiser would want to explore in the
future.
For future analysis, the research should examine all VE locations within a certain distance
of the KP members living in a specific MSA. This may reduce the loss of patients that occurred
when patients were redistributed to census tracts; as well as provide a better idea of the true
accessibility of the members since they are not physically bound by the MSA in which they reside.
The full membership, including patients that have used services as well as those that have not,
should also be used for future analysis to determine an accurate distance decay that would reflect
the member behavior and provide a more complete and accurate picture of spatial accessibility as
well.
Examining the distribution of membership and spatial accessibility suggested broad
similarities in the outputs for both ZIP codes and census tracts. Although this study was a
47

preliminary analysis to test the effectiveness of using ArcGIS for spatial analysis, it helps to
point to areas in need of additional services. The visualization of the service accessibility using
GIS can assist in making the case for additional funds and services to underserved areas.

48

References

Arcury, Thomas, Wilbert M. Gesler, John S. Preisser, Jill Sherman, John Spencer, and Jamie
Perin. 2005. The effects of geography and spatial behavior on health care utilization among
the residents of a rural region. HSR: Health Services Research 40:1 (February): 135-156.

Baker, Richard S., Mohsen Bazargan, Shahrzad Bazargan-Hejazi and José L. Calderón. 2004.
Access to Vision Care in an Urban Low-Income Multiethnic Population. Ophthalmic 12,
1: 1-12.

Becker, Charles. 2016. A spatial analysis of Veteran healthcare accessibility. Master’s thesis,
University of South California. In Blackboard Gist MS Theses,
https://blackboard.usc.edu/webapps/blackboard/execute/announcement?method=search&
context=course_entry&course_id=_68810_1&handle=announcements_entry&mode=vie
w (accessed October 1, 2017)

De Smith, Michael, Michael Goodchild and Paul Longley, 2015. Geospatial analysis: A
comprehensive guide to principles, techniques, and software tools. Rev. ed. Leicester,
UK: Matador on behalf of The Winchelsea Press.
http://www.spatialanalysisonline.com/HTML/index.html?distance_decay_models.htm
(accessed January 2018).

Dewulf, Bart, Tijs Neutens, Yves De Weerdt, and Nico Van de Weghe. 2013. Accessibility to
primary health care in Belgium: An evaluation of policies awarding financial assistance in
shortage areas. BMC Family Practice 14(122).

Gao, Fei, Wahida Kihal, Nolwenn Le Meur, Marc Souris, and Séverine Deguen. 2016. Assessment
of the spatial accessibility to health professionals at French census block level.
International Journal of Equity Health 15: 125.

Grubesic, Tony. 2015. Zip Codes and Spatial Analysis: Problems and Prosepcts. Socio-Economic
Planning Sciences 42: 129-149.

Guagliardo, Mark. 2004. Spatial Accessibility of Primary Care: Concepts, Methods and
Challenges. International Journal of Health Geographics 3(3). http://www.ij-
healthgeographics.com/content/3/1/3

Hansen, Walter. 1959. How Accessibility Shapes Land Use. Journal of the American Institute of
Planners, 25:2, 73-76, doi: 10.1080/01944365908978307

Jamtsho, Sonam and R. J. Corner. 2014. Evaluation of Spatial Accessibility to Primary Healthcare
Using GIS, ISPRS Annals of the Photogammetry, Remote Sensing and Spatial Information
Sciences Volume 11-2, 79-86. Doi: 10.5194/sprsannals-ii-2-79-2014.

Joseph, Alun, and Peter Bantock. 1982. Measuring Potential Physical Accessibility to General
49

Practitioners in Rural Areas: A Method and Case Study. Social Science & Medicine vol
16:85-90.

Joseph, Alun and David Phillips. 1984. Accessibility & Utilization: Geographical perspective on
health care delivery. New York: Harper & Row, Publishers.

Kaiser Permanente. 2015. 2015 Annual Report. Oakland, CA: Kaiser Permanente, 2016.
https://share.kaiserpermanente.org/static/kp_annualreport_2015/ (accessed September 17,
2016)
Kaiser Permanente. 2017. Kaiser Permanente Earns California’s Highest Ratings for Clincal Care.
https://share.kaiserpermanente.org/article/kaiser-permanente-earns-californias-highest-
ratings-clinical-care/ (accessed February 20, 2018)

Kaiser Permanente. 2016. Fast Facts About Kaiser Permanente. Kaiser Permanente, 2016.
https://share.kaiserpermanente.org/article/fast-facts-about-kaiser-permanente/ Accessed
September 17 2016

Khan, Abdullah. 1992. An Integrated Approach to Measuring Potential Spatial Access to
Health Care Services. Socio-Economic Planning Science 26(4): 275-287.

Luo, J. 2014. Integrating the Huff Model and Floating Catchment Area Methods to Analyze
Spatial Access to Healthcare Services, Transactions in GIS, 18(3): 436-448. Doi:
10.1111/tgis.12096

Luo, Wei, and Yi Qi. 2009. An Enhanced Two-Step Floating Catchment Area (E2SFCA)
Method for Measuring Spatial Accessibility to Primary Care Phycisions. Health & Place
15, 1100-1107. doi:10.1016/j.healthplace.2009.06.002

Luo, Wei, and Fahui Wang. 2003. Measures of Spatial Accessibility to Health Care in a GIS
Environment: Synthesis and a Case Study in the Chicago Region. Environment and
Planning B: Planning and Design 30: 865-884. DOI: 10.1068/b29120

Luo, Wei, Fahui Wang and Carolinda Douglass. Temporal changes of access to primary care health
care in Illinois (1990-2000) and policy implications. Journal of Medical Systems, 28(3):
287-299.

McGrail, Matthew and John Humphreys. 2009. Measuring spatial accessibility to primary care in
rural areas: Improving the effectiveness of the two-step floating catchment area method.
Applied Geography, 29: 533-541.

McHugh, Matthew D. Linda H. Aiken, Myra E. Eckenhoff, and Lawton R. Burns. 2016.
Achieving Kaiser Permanente Quality. Health Care Manage Rev 41, vol 3 (July-
September): 178-188

McLafferty, Sara. 2003. GIS and Health Care. Annual Reviews of Public Health 24: 25-42.

50

NewsRX, LLC. Eye Health and Aging. Health & Medicine Week 21 Nov. 2014: 1207. Academic
OneFile. Web. 18 Sept. 2016. Accessed September 17 2016.

Ngui, André, and Phillipe Apparicio. 2011. Optimizing the Two-Step Floating Catchment
Area method for Measuring Spatial Accessibility to Medical Clinic in Montreal. BMC
Health Services Research 11: 166.

Peng, Zhong-Ren. 1997. The job-housing balance and urban commuting. Urban Studies 34(8):
1215-1235.

Rosero-Bixby, Luis. 2004. Spatial Access to Health Care in Costa Rica and its Equity: a GIS
Based Study. Social Science and Medicine 58: 1271-1284.

Salze, Paul, Arnaud Banos, Jean-Michel Oppert, Helene Charreire, Romain Casey, Chantal Simon,
Basile Chaix, Dominique Badariotti, and Christiane Weber. 2011. Estimating spatial
accessibility to facilities on the regional scale: an extended commuting-based interaction
potential model. International Journal of Health Geographics 10(2) doi: 10.1186/1476-
072X-10-2

Umfress, Allison C. and Milam A. Brantley Jr. 2016. Eye Care Disparities and Health
Related Consequences in Elderly Patients with Age-Related Eye Disease. Seminars in
Ophthalmology 4: 432-438.

U.S. Census Bureau. 2016. TIGER/Line Shapefiles: Census Tracts, 2016,
https://www.census.gov/cgi-
bin/geo/shapefiles/index.php?year=2016&layergroup=Census+Tracts (accessed April
2018).

U.S. Census Bureau, 2016. Quickfacts: Riverside County, California.
https://www.census.gov/quickfacts/fact/table/riversidecountycalifornia,riversidecitycalifo
rnia,US/PST045216 (accessed December 2017)

U.S. Department of Health and Human Services, 2017. Healthy People 2020, Vision. Washington
D.C.: https://www.healthypeople.gov/2020/topics-objectives/topic/vision (accessed April
9, 2017)

Wang, Fahui and Wei Luo. 2005. Assessing spatial and nonspatial factors for healthcare access:
Towards an integrated approach to defining health professional shortage areas. Health &
Place 11: 131-146.

Wyn, R., Teleki, S., & Brown, E. (2000). Differences in Access to Health Care Among the
Moderate- and Low-Income Population across Urban Areas. UCLA: Center for Health
Policy Research. Retrieved from https://escholarship.org/uc/item/6sn4d9tz

Yen, Wei. 2013. How Long and How Far Do Adults Travel and Will Adults Travel for
51

Primary Care? Washington State Office of Financial Management. Research Brief No. 70.
Accessed Feb 12, 2017. http://www.ofm.wa.gov/researchbriefs/2013/brief070.pdf

Abstract (if available)

Abstract Research has often examined geographical barriers to healthcare accessibility. These examinations, however, are usually focused on primary care and urgent or specialty care. This study focuses on access to vision care and services with the goal of bridging the gap in research for this category of healthcare. Spatial accessibility for Kaiser Permanente members was examined using the Enhanced 2-Step Floating Catchment Area (E2SFCA) method. This method has been used in previous studies to examine spatial accessibility of patients to healthcare services. It examines both supply (the amount of services or providers available to provide services) and demand (patients who may or have used such services). This study also examined the differences between using ZIP codes and Census tracts as the base geography and for understanding how this choice is likely to affect the performance of the E2SFCA method and the final outputs. The analysis showed that the southern region of the Riverside Medical Services Area (MSA) has a shortage of optical services and that members must travel longer distances for these services. Future research should further analyze the accessibility of the members living within the Riverside MSA to vision services offered by Vision Essentials of Kaiser Permanente.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Spatial analysis of veteran access to healthcare in Los Angeles County

PDF

A spatial analysis of veteran healthcare accessibility

PDF

An exploratory spatial analysis of fire service and EMS accessibility in northeastern Illinois communities

PDF

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

The role of amenities in measuring park accessibility: a case study of Downey, California

PDF

Spatial distribution of the endangered Pacific pocket mouse (Perognathus ssp. pacificus) within coastal sage scrub habitat at Dana Point Headlands Conservation Area

PDF

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

PDF

A comparison of GLM, GAM, and GWR modeling of fish distribution and abundance in Lake Ontario

PDF

Beyond visual line of sight commercial unmanned aircraft operations: site suitability for landing zone locations

PDF

Wetland mapping and restoration decision making using remote sensing and spatial analysis: a case study at the Kawainui Marsh

PDF

Evaluating predator prey dynamics and site utilization patterns of golden eagles using resource selection modeling and spatiotemporal pattern mining

PDF

3D fossil visualization and mapping of the La Brea Tar Pits, Los Angeles, California

PDF

Species distribution modeling to predict the spread of Spartium junceum in the Angeles National Forest

PDF

Building better species distribution models with machine learning: assessing the role of covariate scale and tuning in Maxent models

PDF

Radio frequency identification queuing & geo-location (RAQGEO): a spatial solution to inventory management at XYZ Logistics, Inc.

PDF

Spatial and temporal patterns of long-term temperature change in Southern California from 1935 to 2014

PDF

Spatial analysis of human activities and wildfires in the Willamette National Forest

PDF

A spatial and temporal exploration of how satellite communication devices impact mountain search and rescue missions in California’s Sierra Nevada mountain range

PDF

Using aerial imagery to assess tropical forest cover surrounding restoration sites in Costa Rica

PDF

That sinking feeling: predicting land subsidence in California’s San Joaquin Valley with a spatial regression model

Asset Metadata

Creator Placencia, Christine (author)

Core Title Spatial analysis of vision services of Kaiser Permanente members

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 07/17/2018

Defense Date 07/16/2018

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

Tag census tracts,E2SFCA,enhanced 2-step floating catchment area,geographic accessibility,geographical barriers,healthcare,healthcare accessibility,Kaiser Permanente,medical service Area,OAI-PMH Harvest,Riverside,Southern California,supply and demand,vision,ZIP codes

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Wilson, John P (committee chair), Loyola, Laura (committee member), Wu, An-min (committee member)

Creator Email placenciachristine@gmail.com,schurawe@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-17913

Unique identifier UC11671922

Identifier etd-PlacenciaC-6408.pdf (filename),usctheses-c89-17913 (legacy record id)

Legacy Identifier etd-PlacenciaC-6408.pdf

Dmrecord 17913

Document Type Thesis

Format application/pdf (imt)

Rights Placencia, Christine

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