ASSESSING THE VALUE OF ACCESS TO URGENT CARE IN AN INTERGATED HEALTH SYSTEM: AN ANALYSIS OF KAISER COLORADO’S URGENT CARE IMPLEMENTATION
by
ANNA RASMUSSEN
B.A., Northern Michigan University, 2001 M.A., University of Denver, 2003
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment
of the requirements for the degree of Doctor of Philosophy
Health Services Research 2019
This thesis for the Doctor of Philosophy degree by Anna Rasmussen
has been approved for the Health Services Research
by
Catherine Battaglia, Chair Richard Lindrooth, Advisor
Andrew Friedson Stan Xu Roberta Capp
Rasmussen, Anna (Ph.D., Health Services Research)
Assessing the Value of Access to Urgent Care in an Integrated Health System: An Analysis of Kaiser Colorado’s Urgent Care Implantation
Thesis directed by Professor Richard Lindrooth ABSTRACT
Statement of the Problem: Use of Emergency Departments (EDs) for non-emergent care has been recognized by the medical community for many decades. Avoidable ED visits are a long-standing problem with an estimated opportunity cost of $4.4 billion. Many avoidable ED visits could potentially be managed in alternative ambulatory care options, but the effect on ED overuse and patient outcomes is not well understood.
Objectives: This thesis explores the demand side of the ambulatory care market within the integrated care delivery system, Kaiser Permanente Colorado (KPCO). To accomplish this goal, the study was comprised of three aims: evaluating patient demand for urgent care centers to establish a potential substitution effect between ambulatory care venues,
determining KPCO members’ willingness to pay for access to KPCO urgent care centers, and evaluating the overall feasibility of operating urgent care centers.
Methods: This study used a discrete choice model as an analytic framework to assess patient demand for ambulatory care, substitution effect, and cost-savings derived from the added convenience. In aim 1, geographic market concentration was used to predict patient flow to a set of ambulatory care venues as well as patient response to convenience. In aim 2, the post estimation results were used to calculate the welfare effect if KPCO urgent care
centers were removed from the choice set. In aim 3, KPCO operational cost difference was calculated with and without urgent care centers.
Results: The demand estimates suggest that KPCO members value convenience, but the degree depends on the type of diagnosis and patient characteristics. Aim 2 results show that KPCO members are willing to pay additional amount to maintain access to all current urgent care locations. Aim 3 results show that Kaiser would accrue an additional annual cost of $8,599,890 if all current urgent care centers were closed. Financial benefit to Kaiser combined with patients’ willingness to pay for ambulatory care services demonstrate the feasibility of maintaining and potentially expanding KPCO’s urgent care network. By offering UC services, Kaiser benefits from a reduced number of avoidable ED visits, and patients benefit from easy access to same day services.
The form and content of this abstract are approved. I recommend its publication. Approved: Richard Lindrooth, PhD
ACKNOWLEDGEMENTS
First and foremost, I would like to express my sincere gratitude to my advisor Prof. Richard Lindrooth for his continuous support and guidance. His steady reassurance and immense knowledge were my northern star throughout the dissertation writing process.
I would also like to thank my committee members, Prof. Catherine Battaglia, Prof. Andrew Friedson, Prof. Stan Xu and Dr. Roberta Capp for their insightful comments and
encouragement. Their diverse perspectives challenged and ultimately enhanced my research.
In addition to my committee, I would like to express my gratitude to Dr. Justin Chang who inspired this dissertation topic and gave me the opportunity to learn more about the urgent care operations. Without his professional guidance and encouragement, I would not have had the insight into the fascinating world of care delivery operations.
Last but not the least, I would like to thank my husband, Brian Rasmussen, for his patience and unwavering belief in me.
This publication is supported by NIH/NCATS Colorado CTSI Grant Number UL1 TR001082. Contents are the author’s sole responsibility and do not necessarily represent official NIH views.
TABLE OF CONTENTS CHAPTER
I. INTRODUCTION ... 1
1.1 HISTORY OF THE PROBLEM... 1
1.2 STATEMENT OF THE PROBLEM ... 2
1.2 PURPOSE OF THE STUDY ... 4
1.3 SIGNIFICANCE OF THE STUDY ... 6
II. BACKGROUND AND LITERATURE REVIEW ... 9
2.1BACKGROUND ... 9
2.1a Urgent Care Centers and Other Ambulatory Care Venues ... 10
2.1b Urgent Care in Colorado ... 12
2.1c Urgent Care in Kaiser Permanente Colorado ... 13
2.1d Urgent Care vs. Primary Care in Kaiser Permanente Colorado ... 14
2.2LITERATURE REVIEW ... 15
2.2a Alternative Ambulatory Care Options ... 17
2.2b Substitutability of ED and Urgent Care by Condition ... 17
2.2c Affordability of Alternative Care Options ... 18
2.2d Substitutability of ED and Primary Care by Conditions ... 20
2.2e Care Coordination for Frequent ED Users ... 21
2.3CONCEPTUAL MODEL LITERATURE REVIEW AND EXAMPLES OF OTHER APPLICATIONS ... 22
2.3a Random Utility Model (RUM) ... 23
2.4OTHER METHODS ... 27
III. METHODS ... 29
3.1RESEARCH AIMS AND HYPOTHESES ... 29
3.1a Aim 1 - Estimate the degree of substitution between the utilization of ED, primary care, and urgent care centers for a set of common diagnoses ... 29
3.1b Aim 2 - Measure the change in consumer welfare related to the loss of access to urgent care center ... 30
3.1c Aim 3 - Calculate the return on investment into urgent care centers ... 30
3.2CONCEPTUAL FRAMEWORK ... 31
3.3RESEARCH APPROACH ... 33
3.3a Aim 1 - Estimate the degree of substitution between the utilization of ED, primary care, and urgent care centers for a set of common diagnoses ... 36
3.3b Aim 2 - Measure the change in consumer welfare related to the loss of access to urgent care center ... 38
3.3c Aim 3 - Calculate the return on investment into urgent care centers ... 40
3.4STUDY DATA ... 41
3.4a Inclusion Criteria ... 41
3.4b Diagnoses Included in the Model ... 43
3.4c Choice Set ... 44
3.5DEFINITION AND CONSTRUCTION OF KEY STUDY VARIABLES ... 46
3.5a Aim 1 - Estimate the degree of substitution between the utilization of ED, primary care, and urgent care centers for a set of common diagnoses ... 47
3.5b Aim 2 - Measure the change in consumer welfare related to the loss of access to an
urgent care center ... 48
3.5c Aim 3 - Calculate the return on investment into urgent care centers ... 48
3.6SUMMARY ... 48
IV. RESULTS... 51
4.1OVERVIEW ... 51
4.2SUMMARY STATISTICS ... 52
4.3AIM RESULTS ... 58
4.3a Aim 1: Estimate the degree of substitution between the utilization of ED, primary care, and urgent care centers for a set of common diagnoses ... 58
4.3b Aim 2 - Measure the change in consumer welfare related to the loss of access to urgent care center ... 64
4.3c Aim 3 - Calculate the return on investment into urgent care centers ... 68
V. DISCUSSION ... 74
5.1CONCLUSIONS... 74
5.2SUMMARY OF CONTRIBUTIONS AND IMPLICATIONS ... 76
5.2a Potential UC Expansion ... 79
5.2b Zip Codes with High ED Volume ... 80
5.2c Zip Codes with High RADAR Demand ... 81
5.4LIMITATIONS ... 82
5.5FUTURE RESEARCH ... 85
LIST OF TABLES TABLE
1. Alternative-Specific Attributes………....………... 16 2. Urgent Care, ED, and Primary Care Visits between January 1st, 2018 and Aug 1st, 2018... 53 3. UC Visits by Urgent Care Center and Diagnosis Type………... 55 4. Descriptive Statistics……….…….. 57 5. Conditional Logit Coefficients of Demand Model………... 59 6. Average Ambulatory Care Visits and Arc Elasticities, by Diagnosis Type and Drive Time
to UC (Non-Medicare Subset)……….……….. 62 7. Average Ambulatory Care Visits and Arc Elasticities, by Diagnosis Type and Drive Time
to UC (Medicare Part C Subset)……….…………... 63 8. Average Ambulatory Care Visits and Arc Elasticities, by Diagnosis Type and Drive Time
to UC (Pediatric Subset)………...………... 65 9. Aggregate Effect of Urgent Care Center Closure on Patient Welfare (US $1000s)……….…. 67 10. Aggregate Effect of Urgent Care Center Closure on Patient Welfare, Itemized by
Diagnoses Type (US $1,000s)………..….………. 68 11. Aggregate Effect of Urgent Care Center Closure on Visit Volume Across Ambulatory
Care Venues……….……….. 69 12. KP Cost Estimates by Venue Type and Acuity Level……….……….. 71 13. Annualized Estimates of KP Costs Associated with Redistribution of Visits due to
LIST OF FIGURES FIGURE
1. Graph of logit curve……… 34 2. Map of KPCO locations for UC and PC clinics... 46
LIST OF ABBREVIATIONS ACO - accountable care organization
DCE - discrete choice experiment ED – Emergency Department
EEDM - External Encounter Data Mart HMO – Health Maintenance Organization IIA - independence from irrelevant alternatives KP – Kaiser Permanente
KPCO - Kaiser Permanente Colorado PC – primary care clinic
PCMH - primary care medical home
RADAR – regional acute diagnostic and referral center RUM - random utility models
SNF - skilled nursing facility UC - urgent care center RLQ - right upper quadrant WTP - willingness to pay
CHAPTER I INTRODUCTION 1.1 History of the Problem
Use of ED for non-emergent care is a long-standing problem that has been
recognized by the medical community for many decades. In 1958, The New England Journal of Medicine published an article stating that only a minority of ED visits warranted inpatient hospital admission. In fact, authors discovered that only 16 percent of sampled ED patients required hospitalization. Furthermore, researchers reported that hospital administrators expressed a desire to reduce ED utilization by increasing patient access to other ambulatory care options.1
Health care has changed dramatically since 1958. Emergency care has evolved as well. However, non-emergent ED use is still as pertinent today as it was six decades ago. According to the National Hospital Ambulatory Care Survey, in 2014 only 7.9% of ED visits resulted in hospital admission.2 In addition, numerous studies have found that 30% or more of all ED visits are non-emergent and could be treated in other less expansive ambulatory care venues.3-4
Non-emergent ED visits are typically defined as visits for conditions in which a several hours delay in treatment would not lead to an adverse outcome.3 Because the ED focuses on high acuity care, non-emergent ED visits can be categorized as avoidable, at least from a health care provider’s perspective.4 There are several well-researched drivers of avoidable ED utilization, including lack of alternative ED care options, a patient’s
health care services, and complex patients with disproportionately high ED use.5–9 In response, a myriad of interventions have been designed to influence these contributing factors to reduce avoidable ED utilization. These interventions include patient education on appropriate ED use, financial disincentives such as increases to co-payments for ED visits,10 improved coordinated care management for complex patients, and expanded convenient ambulatory care options for patients.
1.2 Statement of the Problem
Non-emergent ED use is expensive and inefficient.4,11 One study projected $4.4 billion in annual savings if less expensive ambulatory care venues could absorb avoidable ED visits.4 As convenience-based ambulatory care continues to expand, it is important to understand how it affects traditional ambulatory venues, especially ED. Urgent care centers represent a significant segment of convenience-based ambulatory care. Urgent care centers offer convenient, affordable and timely care for a wide range of acuity levels. For these reasons, urgent care centers have been long recognized as a plausible solution to overcrowding in other health care sectors, including ED.4
Urgent care expansion is often accompanied by the claim that urgent care centers can absorb a significant portion of avoidable ED visits.4 Although urgent care centers have become more widespread, current literature suggests that we have yet to fully understand their effect on ED utilization. A myriad of factors influences ED utilization, including patient access to episodic ambulatory care, patient education, misdirected referrals, and frequent ED users.11–13 While these are important factors to consider, this study will focus on
assessing the degree of substitution between EDs and urgent care centers for a set of common ED diagnoses.
Numerous studies have evaluated why patients choose to use the ED for
non-emergent care.6-7,11,13-14 Other studies have assessed how many ED visits could be absorbed by less expensive ambulatory care venues.3-4 A few studies have also compared newly emerged ambulatory care options to traditional ambulatory care venues such as primary care offices and ED.16 Most notably, Ho et al. reported on utilization, price per visit, and type of encounters across freestanding EDs, hospital-based EDs, and urgent care centers in Texas.16 Although the study was criticized by the American College of Emergency Physicians, Ho’s findings indicate that there is a significant service overlap in the 20 most common diagnoses between hospital-based EDs and urgent care centers. Furthermore, Ho et al. found a significant price difference between urgent care centers and EDs. Overall, these findings indicate potentially inefficient ED use and raise further questions about whether some of the overlap in services may be absorbed by urgent care centers.
Although ED overuse and urgent care expansion has been extensively covered by both health care researchers and news outlets, there is no definitive conclusion as to whether urgent care centers can decrease ED crowding. Very little is known about the degree of substitution between urgent care centers and EDs. Furthermore, there is very little research on patient demand for additional convenient care venues. This study focused on evaluating how patients choose a care venue for their non-emergent needs and their willingness to pay for additional convenience.
1.2 Purpose of the Study
The purpose of this study was to estimate KPCO patient demand for urgent care centers and to establish a substitution effect between urgent care centers and EDs. There are several reasons to study urgent care centers. Since their inception in the 1970s, urgent care centers experienced rapid growth, which resulted in thousands of urgent care centers across the United States.17 In a market economy, services that do not align with consumers’ preferences quickly disappear. Urgent care expansion is in large part driven by consumers’ preference for convenient, affordable, high quality ambulatory care, and growing numbers of consumers use alternative ambulatory care. In 2016, 27% of Americans reported that they had visited an urgent care center in the last two years. Furthermore, 75% of patients reported their care as excellent or good.17 Convenient care is changing the paradigm of the traditional care delivery model, and it is essential to understand the ramifications.
The value of alternative ambulatory care has become a polarizing topic in health care.18,19 Specifically, urgent care’s rapid expansion has inspired a high volume of literature.4,7,16 There are numerous publications that outline the benefits of urgent care centers.4,18,20,21 Those benefits include patient-oriented convenience, a wide spectrum of services that significantly overlap ED services, and the potential to alleviate crowding in primary care offices.4,16 There are also an equal number of publications that call out urgent care centers’ limitations.19 Limitations include the revenue-oriented business model of urgent care centers, services that may not extend to communities with the highest need, and a concern that urgent care centers generate higher health care utilization without improving crowding in primary care or inefficient ED use.22
However, there are surprisingly few studies that substantiate the claims of either side with comprehensive analytic evaluation. Furthermore, I did not encounter any studies that evaluate the direct substitution effect between alternative and traditional ambulatory care venues. The studies that do exist on urgent care centers mostly discuss the supply-side of the market in an attempt to understand the effects of recent high urgent care center entry rates in the health care industry.16 I did not identify any studies that analyzed the demand-side for alternative ambulatory care, or specifically for urgent care centers.
Given the significance and relevance of alternative ambulatory care to the overall health care industry, it is especially interesting to consider the nature of demand for urgent care centers as well as the degree of substitution between urgent care centers and
traditional ambulatory venues. I used accepted methods for estimating patient venue choice and conducted the estimation using data on Kaiser population. Because convenience is at the forefront of the urgent care center business model, patients should prefer a nearby facility rather than the one farther away, all things being equal. While the level of demand will be affected by many variables (patient characteristics, price, care venue characteristics, geographic location, etc.), the spatial pattern of relatively higher demand for nearby
facilities should hold across all observed care venues. Using logit results, I estimated a substitution matrix based on cross-time elasticity. I also calculated the magnitude of the substitution effect between urgent care centers, primary care clinics, and EDs when all characteristics remain constant and only the urgent care centers location changes. This assessment will help to determine demand elasticity for urgent care centers. Finally, I estimated patient willingness to pay for additional convenience.
1.3 Significance of the Study
The discrete choice framework has been applied by economists in a wide variety of settings. In health care, these analyses have been predominately used to evaluate hospital utilization. By evaluating patients’ choice of care venue, the discrete choice methodology helps to estimate aggregate patient demand for health care services. Due to relatively recent and rapid urgent care expansion, discrete choice methodology has not been applied to evaluate patient demand and elasticity for urgent care centers.
To align with patient demand for convenient ambulatory care, Health Maintenance Organizations (HMO) like Kaiser Permanente (KP) also began to include urgent care centers within their networks. There are no publications that assess the impact of urgent care centers on the care delivery within the HMO framework. Without understanding the demand for urgent care centers, we cannot fully understand how patients choose between available care options. This study focuses on how patients choose between urgent care centers, primary care clinics, and hospital-based EDs for conditions that can be treated in any of these venues. This study also focuses on how patient welfare changes when an urgent care center is removed from the available choice options.
In this study, I developed a discrete choice model of Kaiser Permanente Colorado (KPCO) members’ demand for urgent care centers. Via this discrete choice model, I estimated structural parameters of patient demand function from multinomial logit specifications. I also estimated a substitution matrix based on the cross-time elasticity to explain how patients choose between urgent care centers, KPCO primary care clinics, and EDs. Furthermore, I measured how patient welfare would change if an urgent care center
was removed from patient options. Finally, I evaluated the overall feasibility of expanding the KPCO urgent care network.
To accomplish this, I approached the problem as follows: When patients seek immediate, but non-emergent medical care, they decide which ambulatory option they will use. In the KPCO network, patient options are limited to primary care clinics, urgent care centers, or an ED. Hypothetically, a KP member can also choose non-KP urgent care center; however, because the primary purpose of this study is to evaluate the substitution effect between KPCO urgent care centers and hospital-based EDs, non-KP urgent care centers were not included in the analysis. However, this exclusion should not make much difference for the analyses, as non-KP urgent care use among KPCO HMO members is highly unlikely since Kaiser insurance does not cover these services. This study focuses on how KPCO members select a care venue and how much they value the convenience aspect. I also evaluated how facility characteristics, including geographic location, influence consumer choices. Location was measured with a time-to-travel variable that specifies the distance (in minutes) from each patient’s home address to the service facility. Therefore, my study aims are: 1) estimate patient demand function for ambulatory care services and estimate
ambulatory care substitution effect due to convenience, 2) determine patient willingness to pay for UC service, and 3) assess whether it is financially feasible for Kaiser to maintain its UC network.
These findings provide insight about how KPCO members choose a care venue for non-emergent care needs. Original data from the KPCO HMO network, econometric methods that have not been used to evaluate patient demand for urgent care centers,
KPCO members’ willingness to pay for additional convenience, and financial ramifications for KPCO all contribute to the current knowledge on alternative ambulatory care and its role in the integrated health care system. Furthermore, these findings can be used to inform KPCO leadership about how patients value convenient ambulatory care and whether KPCO should further expand its urgent care network.
CHAPTER II
BACKGROUND AND LITERATURE REVIEW 2.1 Background
Emergency departments (EDs) are an important part of the US health care system. EDs offer a full range of medical services available at any time and to anyone regardless of their ability to pay.23 According to the Healthcare Cost and Utilization Project (H-CUP) report, ED visits have been rising over the past two decades.23 In 2014, close to 20% of US adults used EDs for their healthcare needs.12 It is often believed that uninsured patients drive the overall numbers of high ED utilization. However, several published reports indicate high ED utilization across different payer groups.10-11,13 The 2014 National Statistics Survey reported that among adults aged 18 to 64, 14% of privately-insured, 35% of Medicaid-insured, and 17% of uninsured US adults had visited ED one or more times in the past 12 months.12 These data reinforce the fact that ED utilization continues to increase across all populations, regardless of age and insurance type.11
Access to an ED can provide peace of mind, but it is an expensive place to get health care if there are more appropriate care options. The ED is not an appropriate venue for all healthcare needs. Many ailments treated in EDs can be safely and effectively treated in other less expensive ambulatory settings.11 Nationally, an estimated 14% to 27% of ED visits are deemed avoidable and can be managed in a primary care setting.4 Solving ED overuse has been on the forefront of the healthcare industry for the last decade, and yet ED
utilization has remained largely unchanged.4,24 There are many reasons why people choose EDs for their health care needs. For example, EDs represent a convenient care venue that is
open 24 hours a day. There is no need to schedule an appointment, and service is guaranteed. Additionally, EDs can address troublesome symptoms, offer immediate
reassurance, and provide timely feedback. Finally, EDs are equipped to offer a wide range of readily-available services. However, ED overuse may lead to unintended consequences and can diminish the quality of care for people with acute emergencies due to crowding, long waits, and reduction in access.11 ED overuse also has an adverse effect on the continuity of care and cost.4,25
Overall, avoidable ED visits represent an inefficient use of healthcare resources and contribute to high health care costs. ED visits cost roughly ten times more than the same treatment in a primary care setting. If avoidable ED visits across the US were managed through other less costly care channels, the substitution would result in $4.4 billion in annual savings.4
2.1a Urgent Care Centers and Other Ambulatory Care Venues
In addition to increasing ED utilization, there has been an unprecedented growth in ambulatory care venues for urgent and emergent medical needs in the past decade.26 Consumers are looking for quicker and more accessible care options, forcing the health care industry to adjust their care delivery models.27 According to the 2015 Merritt Hawkins’ report, an increasing number of healthcare organizations are moving away from a traditional transaction-based care delivery model to an experience-based care delivery model that prioritizes customer service, convenience, and price transparency.20 In this new convenience-oriented climate, healthcare is shifting to an outpatient care delivery model with a wide menu of care options geared to meet a variety of customer preferences. Many
healthcare organizations have expanded their care options to include urgent care centers, ambulatory surgery centers, retail clinics, freestanding EDs, and virtual care services.26
Although all the newly developed amenities deserve closer inspection, this study will focus on urgent care centers. Urgent care is one of the largest components of this new convenience-driven trend.17,20 Urgent care centers serve as a bridge between primary care and emergency care services. They are designed to provide episodic care for a range of acuity levels, and are typically staffed with emergency and primary care physicians and nurse practitioners.18,26 There are currently 7,500 urgent care centers across the United States and are growing at a rate of approximately 300 to 600 new centers per year.17,28 Due to patients’ preference for fast, convenient, and affordable ambulatory care, further urgent care expansion is likely to continue.26
Favorable patient response and low regulatory barriers are the main drivers behind urgent care’s rapid expansion. Urgent care centers do not require additional licensing other than that of a primary care clinic.26 Urgent care providers are not required to have
additional emergency medicine training.20 In addition, urgent care offers ambulatory, episodic care in addition to being easily accessible, convenient, affordable, and offering short waiting times. Most urgent care centers treat illness and injuries that are not life-threatening but that are too pressing to wait for an appointment at a doctor’s office. Many are open in the evenings, on weekends, and on holidays.21 Most urgent care centers maintain a wait time of 30 minutes or less, with total patient experience at 60 minutes or less.15 Finally, patients pay less than $150 on average for an urgent care visit compared to an average of $1354 for an ED visit.29
2.1b Urgent Care in Colorado
Colorado has experienced rapid population growth of 11 percent from 2009 to 2016, and some believe that urgent care centers play an important role in improving access to medical services in growing areas. Furthermore, some research has shown that urgent care centers help alleviate pressure on overcrowded EDs by treating patients with non-emergent and non-life-threatening illnesses. According to the 2017 Colorado Health Access Survey, 6.4% of Coloradans used an urgent care center for their health care needs, which is twice as many as did in 2009.15 Although ED utilization across Colorado has remained relatively steady, ED use among the uninsured dropped from 18.3% in 2009 to 10.9% in 2017.15 In addition, avoidable ED visits in Colorado have been on a steady decline since 2009. In 2017, non-emergency ED use dropped from 43.7% to 36.4%.15 Nevertheless, many Coloradans reported using the ED for ailments that could have been treated in a primary care setting, including 72.4% who needed non-emergency care outside of normal business hours and 61.4% who reported non-emergency ED use due to convenience.15 Another 59% reported non-emergency ED use because they were not able to schedule an appointment in a primary care setting to address their health concerns as quickly as they preferred.
In line with national trends, the Colorado market has seen a marked increase in the number of urgent care centers. All major healthcare providers, including Banner Medical Group, Centura Health, Denver Health, and UCHealth, are building urgent care centers or partnering with companies that run urgent care centers. There are currently 128 urgent care centers across Colorado and this number continues to grow.30
2.1c Urgent Care in Kaiser Permanente Colorado
Many insured patients use EDs as primary care venues, even though they are not designed to provide this type of care.11 Even within the HMO framework, avoidable ED visits are prevalent. KPCO) insured members reflect the national trend, with a significant
percentage using ED services for their non-emergency needs. KPCO survey data indicates that only 20% of members who call the Medical Advice triage line before choosing a care venue end up using ED services. However, because only 30% of Kaiser members call the Medical Advice triage line before going to the ED, there is a potentially significant
opportunity to reduce avoidable ED visits. KPCO internal survey findings indicate that many Kaiser members choose ED services due to convenience, and this reflects the trend
reported by National Health statistics.6-7,11,13-14
In recent years, KPCO introduced, expanded, and enhanced several urgent care centers in the metro Denver area to mitigate avoidable ED visits and improve patient access to same-day services. Currently, there are four urgent care centers within the KPCO metro Denver network. Kaiser patient satisfaction surveys indicate an overall positive response to the newly-enhanced urgent care centers. However, it is not known whether urgent care clinics affect avoidable ED utilization among KPCO members.
In 2017, Kaiser observed a reduction in ED utilization, but it is not clear whether urgent care centers influenced this reduction and whether the effect warrants further urgent care expansion. The purpose of this study is to analyze KPCO members’ responses to the KPCO urgent care clinics and to understand whether they chose urgent care as a
urgent care center is removed from the Kaiser network. Finally, this research will estimate the overall change in KP operating income31 associated with urgent care implementation. 2.1d Urgent Care vs. Primary Care in Kaiser Permanente Colorado
Although urgent care centers represent one care venue that may absorb avoidable ED visits, traditional ambulatory settings such as primary care should not be discounted. Many primary care offices accommodate after-hours and walk-in appointments and can treat low acuity conditions. This begs the question: to reduce avoidable ED visits, why not simply expand primary care services? The answer is that several factors make urgent care more efficient in providing episodic ambulatory care. First, the KPCO urgent care staffing model differs considerably from the primary care staffing model. An urgent care staffing model allows care for a wide range of medical conditions, and urgent care includes ED physicians as well as primary care physicians who can treat moderate to high acuity cases. Many urgent care nurses have ED or critical care experience and are trained to follow the ED workflow. Second, the urgent care model is designed to handle a high volume of patients. Immediate triage allows the urgent care provider to efficiently manage UC visits. To maximize efficiency, KPCO urgent care centers use the ASAP ED application. ASAP is an Epic module that helps facilitate ED workflow. ASAP allows staff to track occupied rooms and displays room and bed status on monitors in nursing stations. On average, KPCO urgent care clinics see 600-700 patients per month, and the numbers continue to grow. If all these patients were redirected to primary care, KPCO primary clinics would be overwhelmed.
Third, urgent care’s mission is to deliver ambulatory, low- to high-acuity episodic care. Unlike traditional primary care, urgent care centers offer a more comprehensive menu
of lab work, advanced diagnostics such as cardiac monitoring and CT, and dedicated and trained urgent care staff. Relative to primary care, urgent care centers can diagnose and treat a wider range of non-life and limb threatening illnesses and injuries.
Primary care, on the other hand, focuses on a continuity of care. When urgent care centers absorb urgent visits, they allow primary care clinics to focus on delivering better access for patients with chronic conditions and general health maintenance needs. Finally, KPCO urgent care cost per visit is comparable to primary care visits. Because KPCO does not build freestanding urgent care centers, KPCO is able to leverage existing clinic buildings to include urgent care centers. In an integrated healthcare system, urgent centers enhance primary care services by referring unbonded KPCO patients to clinics and by allowing primary care to focus on continuity of care. Table 1 captures the differences in service across KPCO urgent care centers, KPCO primary care clinics, and hospital-based EDs. A literature review of this topic will be discussed in section 2.2.
2.2 Literature Review
Between 2004 and 2014, ED visits increased by 21%.32 Patient misperception of medical need, misdirected referrals, convenience, and financial disincentives were the main drivers behind this trend. Nevertheless, prevalent, non-emergent ED use is expensive, inefficient, and avoidable.3 Studies show that two main factors contribute to the growing volume of avoidable ED visits. First, primary care-treatable ED visits represent a significant portion of non-emergent ED volume. These avoidable ED visits are primarily due to a lack of accessible primary care alternatives or because patients see the ED as the only care option that can effectively meet their needs.15 Second, a small fraction of frequent ED users makes
up a significant portion of avoidable EDs. These complex patients tend to have multiple physical and mental comorbidities that require comprehensive, multi-disciplinary care coordination, including ED.13
Table 1. Alternative-Specific Attributes
Characteristics ED UC PC
7-day access Yes Yes No
24-hours access Yes No No
Emergency Medicine physicians Yes Yes No
ED workflow Yes Yes No
KP network No Yes Yes
Financial counseling No Yes No
On-site pharmacy Yes Yes Yes
CT scan Yes Yes* No
Cardiac monitoring Yes Yes No
Hospital access Yes No No
Triage to PC/UC No Yes Yes
High acuity treatment Yes Yes* No
Low acuity treatment Yes Yes Yes
*Urgent Care Centers with RADAR
As the number of ED visits continues to grow at an alarming rate, debates about possible solutions have ensued with greater frequency. This literature review will focus on alternative ambulatory care options that may serve as substitutes for avoidable ED visits, care coordination programs for high ED users, and the overall social determinants that influence avoidable ED utilization. This review includes articles published after 2010 in order to capture the most recent health care trends, newly-emerged care solutions, and
2.2a Alternative Ambulatory Care Options
With a recent expansion of ambulatory care options, there are numerous
publications that review visit volume, clinical effectiveness, affordability, and access across emerging ED alternatives. Although there is a wide spectrum of ambulatory care options, this literature review predominantly focuses on urgent care centers, retail clinics, and freestanding EDs.
Proponents of alternative ambulatory care options state that these care venues offer high quality, affordable and timely care, and alleviate crowding at hospital-based EDs.4,18,33 On the other hand, critics of convenient ambulatory care venues argue that these care options serve as supplements to a traditional ED but have very limited impact on ED utilization and further exacerbate national health care costs. This following section focuses on recent publications that evaluate alternative ambulatory care options and their ability to influence avoidable ED visits.
2.2b Substitutability of ED and Urgent Care by Condition
Studies show that 30 percent or more of ED visits are deemed primary care treatable, creating a significant overlap in the clinical conditions that can be treated in a variety of care venues.3-4 Every year 1.2 billion encounters take place in ambulatory care settings.34 Although many of these visits take place in traditional primary care clinics and EDs, an increasing number of patients choose new alternative ambulatory care venues such as retail clinics and urgent care centers. Although similar in their convenience factor, retail clinics and urgent care centers have different functions. Retail clinics provide low-acuity
episodic care, immunizations, and limited preventive care. Urgent care centers offer episodic care for a wider range of acuity levels.18
Collectively, urgent care centers and retail clinics treat a wide range of acuity levels. If 27% of avoidable ED visits could be treated in other less expensive venues, it is reasonable to suggest that urgent care centers and retail clinics can absorb some of this volume. Some publications find that convenient ambulatory options are appropriate ED substitutes for low-acuity, non-life-threatening conditions.4 Moreover, one publication reports a 60% overlap in the 20 most common diagnoses between hospital-based EDs and urgent care centers.16
2.2c Affordability of Alternative Care Options
Affordability has been a key argument in favor of convenient ambulatory care settings. If all avoidable ED visits were managed in other less costly ambulatory care, the substitution would result in $4.4 billion in annual savings.4 Urgent care centers and retail clinics leverage less expensive staffing models and often incur lower fixed costs relative to traditional ambulatory settings. In fact, several publications have found urgent care centers to be much more affordable than primary care and EDs.4,16,18,35
However, despite a significant difference in cost per visit and a significant service overlap, total health care cost may increase if urgent care centers and retail clinics generate additional visit volume instead of replacing visits to primary care offices and EDs. Other studies have noted that conclusions about inefficient ED use are premature. Health care industry experts define primary care-treatable ED visits as avoidable, but this may not be true from a patient perspective.19 Many patients cannot accurately self-assess the severity
of their health concern and therefore defer to the ED. Furthermore, many primary care providers refer their patients to the ED.3
Physician organizations such as the American Academy of Pediatrics and the American Academy of Family Physicians have expressed concerns that ambulatory care venues may disrupt continuity of care and provide lower quality of care.18,19 This literature review outlines key factors that contribute to this ongoing debate.
Quality assessment in episodic ambulatory care can be challenging, primarily because quality indicators historically have focused on evaluating patient care over time. However, there is enough preliminary evidence to suggest that urgent care centers provide at least equal quality of care as compared to traditional ambulatory care options.35,36 Close to one-third of urgent care centers regularly measure their quality of care via the National Committee for Quality Assurance’s Healthcare Effectiveness Data and Information Set (HEDIS).36 In comparison only 20% of all physicians track quality of care metrics on their patient panels.36 Furthermore, one study found that retail clinics achieved a ranking above the 90th percentile for several HEDIS measures.37 Finally, there is evidence to suggest that vast majority of patients are satisfied with their convenience care options.18,21
Some have questioned whether business feasibility determines the location of these acute care options, which are predominantly available in areas with a better patient payer mix.22,38 The result is to create redundant services in more affluent areas but not address health care needs in low-income communities.
Despite several study limitations, two main findings remain. First, there is a
centers. Second, there is a significant overlap in the clinical services offered in EDs and urgent care centers. These findings warrant further investigation of care venue substitution between the ED and other ambulatory care options.
2.2d Substitutability of ED and Primary Care by Conditions
As discussed in section 2.1, primary care should be considered one possible solution to avoidable ED visits. As a result, research must consider whether primary care can
mitigate low acuity ED overuse. Several studies evaluated what primary care can do better to fill the gap and why some patients with low acuity conditions continue to choose ED over primary care.5,7,12,39,40 One New York-based study identified several factors that can improve primary care and reduce reliance on EDs,7 including:
1. Increasing after-hours and weekend hours 2. Increasing telephone consultation capacity
3. Providing more education to patients with chronic conditions such as asthma and diabetes
4. Adjusting the care delivery model to meet patient needs
5. Applying federal and state subsidies to expand primary care services
Although many of the proposed enhancements seem practical, they may not address the core reasons why patients choose ED. There are several published studies that explore the patient venue selection process from a consumer perspective.5,12,39 These studies found there are several key factors that explain why patients with low acuity conditions chose ED over primary care, even when the same day and after-hours appointments were available.
Although access to primary care is a factor in avoidable ED utilization,40 other important patient-reported reasons explain why primary care may not be able to absorb low acuity avoidable ED visits. Perceived seriousness of the medical problem is one of the most common reasons patients chose ED instead of primary care.9,39 Easy access to
advanced imaging and lab tests is another commonly-cited reason.39 Negative perceptions of primary care providers also contributes to avoidable ED use.3,39 Finally, misdirected primary care referrals contribute to a number of low acuity avoidable ED visits.3 Combined, these factors continue to drive avoidable ED utilization even when primary care is
accessible. These findings suggest that patients may have a different interpretation of urgent care services. Do patients view urgent care centers as a reasonable substitute to the ED for time-sensitive needs? This question will be explored in the next chapter of the dissertation.
2.2e Care Coordination for Frequent ED Users
A small percentage of frequent ED users represent a significant portion of avoidable ED visits. Several publications have proposed and evaluated care coordination programs designed to mitigate avoidable ED visits among frequent ED users. Many proposed programs incorporate a multidisciplinary approach with care coordination across primary care, behavioral health, and community resources. The premise of these programs is to preemptively connect patients with the right heath care resources to avoid the “revolving door” process between ED, hospital admission, and skilled nursing facilities (SNFs).
Although the comprehensive nature of care coordination is both time and labor intensive, these programs are effective in addressing the long view of care for vulnerable
high-ED users. Published results indicate a promising reduction in ED visits and
hospitalization among program participants. Some studies report that intervention effects last for at least a year post-intervention.13,41 In addition to reducing ED visits and hospital admissions, these programs also fill gaps in chronic disease management and behavioral health.13 Another study reported a promising 5-8% reduction in ED utilization among chronically ill patients when traditional primary care was transformed into a primary care medical home (PCMH) with an emphasis on care coordination.42
2.3 Conceptual Model Literature Review and Examples of Other Applications “Nothing can have value without being an object of utility.”
-Karl Marx, Economic Manuscripts
Undoubtedly, Karl Marx had a very specific definition of utility in mind. However, the statement in its absolute form holds true: nothing can have value without providing utility, whether tangible or intrinsic. In this study, one key economic assumption is that consumers are rational decision makers who maximize their utility. Therefore, if an object or service is chosen from a set of options, it must maximize the decision maker’s utility, within the context of imperfect information. However, consumers are not homogeneous and as a result may have different definitions of utility, introducing, from the researcher’s perspective, a random factor in a decision-making process. A fundamental goal of economics is to understand how people make decisions. Discrete choice models built on principles of economics encapsulate an aggregate choice behavior based on the
distributions of individual choices. Individual choice reflects the unobserved utility derived by a consumer from a good or service. In this context, utility itself is a composite of a
systematic component and a random component. The latter could be due to unobserved preference variation or measurement error, for example. Because of utility maximization, discrete choice models provide a theoretical framework to capture aggregate demand for differentiated products. Specified discrete choice models inform how individual consumers choose mutually exclusive options from a set of substitutable goods. For example, the discrete choice model can be used to explain how patients choose different ambulatory care venues for their non-emergent needs. This section will provide a brief review of key economic concepts that inform the conceptual model discussed in chapter 3.
2.3a Random Utility Model (RUM)
The random utility model (RUM) is a type of probabilistic choice model that was first developed by psychologists.43 In 1927 Louis Thurstone applied probabilistic choice models to his work in psychometrics. Thurstone was the first to use choice models to account for inconsistencies in patterns of individual behavior.43 Similarly, economists apply the scientific method to explain people’s behavior in the context of a market environment. The
economics field leverages methodology that captures unobserved consumer choice
patterns among discrete alternatives, making the random utility theoretical framework very useful in the area of applied economics.
RUMs start from the assumption that each consumer makes one selection out of a set of available choice options that yields the highest utility. In this context, utility is defined as a random variable that reflects unobservable taste preferences.44,45 Random utility does not necessarily mean that consumer choice is stochastic; it only indicates that from the provider’s viewpoint, the factors influencing individual preferences are unknown.
One of the most frequently applied RUMs is a logistic model (logit). This model was originally introduced by R. Duncan Luce in 1959. The logistic model assumes that the random component of utility follows an extreme value distribution. This assumption yields specific substitution patterns across choices. Specially, the choice probabilities exhibit “independence from irrelevant alternatives,” (IIA) which means that the probability of making one selection over another from a set of alternatives is not affected by the presence or absence of any other alternatives in the set.46 IIA implies that each alternative is distinct enough to reflect appropriately specified choice probabilities. For example, patient
preference between urgent care centers and hospital-based EDs cannot be affected by the presence of other ambulatory care options such as primary care clinics or freestanding EDs. IIA implies that choice alternatives are distinct with specific patterns of substitution. Collectively, choice probabilities add up to 1, so if a probability of choosing one alternative increases, then the probability of selecting other alternatives must decrease. If
substitutional patterns are subtle, i.e., alternatives are close substitutes, then more flexible models are needed, and logit would not be a good fit.
Although both multinomial logit and conditional logit can be used in the discrete choice analyses, conditional logit model, developed by McFadden, has been used in many demand studies.44 The main difference between the two logits is that the multinomial logit focuses on the individual as the unit of analysis and uses the individual characteristics as explanatory variables. On the other hand, conditional logit focuses on the choice set for each individual and the explanatory variables are characteristics of those alternatives.
Since logit relies on IIA assumptions, IIA assumptions should be validated to ensure the correct analytical framework. In 1978 Jerry Hausman developed a methodology to test IIA. If the IIA assumption is valid then the parameter estimates obtained on the subset of alternatives will not be significantly different than the parameter estimate obtained on the full set of options. The Hausman test is often used to validate IIA.47
2.3b Practical Applications of RUMs
The RUM theoretical framework has been applied across a wide spectrum of industries, including health care.47–49 In the context of health care, the discrete choice methodology has been applied to traditional care venues most frequently to estimate demand for hospitals. There are numerous publications that use discrete choice models to assess patient experience in the context of health care services such as welfare derived from ambulatory surgery services, hospital network expansion, hospital closure, and hospital quality.50–52 While Small and Rosen (1981) provided the general theoretical
framework of the discrete choice analyses used in this study, there are several studies that demonstrate the practical application of this framework: Capps et al.,52 Tay,51 Capps et al.,50 and Weber.53 These studies serve as a reference guide for the structural model discussed in the next chapter.
In the above studies, consumer demand for healthcare facilities was modeled as a discrete choice of one facility from a well-defined choice set. In the context of health care services, the patient is the consumer who selects one option from a defined set of care options. Patients have idiosyncratic preferences and choose a health care facility in accordance with their personal characteristics as well as the health care venue’s
characteristics. For example, high acuity patients may prefer hospitals to prevent an adverse outcome, while healthier patients who require less complex treatment may be more willing to choose an outpatient facility. Younger patients may be willing to travel longer distances than older patients for a given procedure. The interaction between patient and facility characteristics in an empirical model allows the coefficients that reflect
preferences to differ across consumers.
Several studies used conditional logit as a discrete choice model to estimate their demand for hospitals.51,53–55 A conditional logit specification implies that utility function parameters are fixed across all patients who make a hospital selection. This may become problematic if unobserved patient characteristics cause certain care options to be close substitutes. When this occurs the IIA assumption is violated, and logit is no longer an appropriate option. To relax this assumption, many studies defer to a mixed logit model that allows for random coefficients. Mixed logit accounts for patient variation in patient preferences and allows researchers to identify a degree of trade-offs. For example, one of the trade-offs could be between hospital services and patient travel time.
To estimate demand for hospital or any other medical services, the underlying patient utility function must be determined. As previously discussed, patients choose an alternative that generates maximum utility. Utility function is a mathematical
representation of the key explained and random factors that influence patient choice. Utility functions often include individual patient characteristics, care venue characteristics (i.e., hospital features), distance, and interactions across these variables. The list of variables depends on the hypothesis being tested. The reviewed literature indicates that
there are several examples that demonstrate practical applications of the discrete choice model. For example, allowing care venue characteristics and distance to interact with patient characteristics acknowledges that trade-offs between distance and hospital characteristics vary by patient type. Furthermore, allowing age to interact with distance allows researchers to test whether younger patients are willing to travel further than older patients. Overall, the intent is to establish patient demand and assess elasticities across available options with respect to price and other characteristics such as travel distance. In this study, I used logit methodology to estimate demand and substitution effect between urgent care centers and EDs.
2.4 Other Methods
Although not used in this study, in recent years, discrete choice experiments (DCEs) represent another technique for applying RUM to a variety of health care questions. DCEs represent a composite of several theoretical frameworks including RUM,44,45 welfare
economics,56–58 and Lancaster’s theory of demand.59 Lancaster’s theory of demand is based on the notion that the utility of goods or services is determined by different attributes associated with a good or service. Attributes vary by levels.59 DCEs leverage different service attributes to ascertain consumer preferences. A key DCE characteristic is a reliance on surveys to gather data about consumers’ choices.60 In contrast, this study focuses on methodology that uses observed patient choices.
The World Health Organization defines the social determinants of health as “conditions in which people are born, grow, live, work, and age. These circumstances are shaped by the distribution of money, power and resources at global, national and local
levels.”61 The Aday-Andersen behavioral model of health services use is a conceptual framework that is often used to explain an individual’s use of health services. The Aday-Andersen model can be applied to describe how patients use ambulatory care and care delivery outcomes. The model applies social and individual determinants of health that influence the likelihood of illness, then describes care-seeking decisions, care delivery settings, care integration, clinical outcomes, and health care costs. The model suggests that health care needs occur in the context of social and individual health determinants. These determinants include socio-environmental determinants, public health determinants, and individual health determinants. Combined, these factors influence the likelihood that an individual will experience an illness or injury, and subsequent care delivery and clinical outcomes.
Covariates selected in the discrete choice model for this dissertation are aligned with Anderson’s demographics, social, and health beliefs components.54,62 According to the Anderson model, the enabling resource and need constructs will explain most variability in health services use.63 Through a discrete choice framework, I analyzed enabling resources via newly introduced urgent care centers. The discrete choice framework also accounted for patient need via a utility function that captured patient characteristics and unobserved preferences.
CHAPTER III METHODS 3.1 Research Aims and Hypotheses
This study sought to determine how KPCO members decide between urgent care centers and EDs for most common ED chief complaints,64 and measure the value KP
members put on additional convenience of care. HMOs and accountable care organizations (ACOs) pursue efficient and effective ways to deliver care through integrated care
coordination. Determining how HMO members utilize in-network urgent care centers provides insight on whether HMOs like Kaiser or ACOs should expand their in-house urgent care networks.
As a result, it is important to understand KPCO member venue choices. A specific application of RUM, a logit-demand model, was used to analyze the behavior of
heterogenous decision makers who may chose mutually exclusive care options from a set of substitutable care venues. KPCO urgent care centers, KPCO primary care clinics, and
frequently utilized EDs were the substitutable options that I analyzed in the model. The overall objective was to evaluate KPCO member choice patterns for their same day care needs, and to measure the substitution effect between three different care venue types. 3.1a Aim 1 - Estimate the degree of substitution between the utilization of ED, primary care, and urgent care centers for a set of common diagnoses
To understand KP members’ utility associated with additional urgent care sites, first I established the baseline venue choice pattern. Given the current urgent care clinic
clinic vs. ED? By limiting the data set to the encounters with the most common diagnoses, I was able calculate the probability of selecting different care venues based on KPCO member individual choices. These probabilities were then aggregated to measure the expected number of visits at each facility. To evaluate the substitution effect, I included reduced distance to urgent care centers in the model and evaluated whether this “increase in
access” changed the probability of selecting urgent care centers and the expected change in visits.
3.1b Aim 2 - Measure the change in consumer welfare related to the loss of access to urgent care center
The intent of this aim was to examine whether KPCO members were more likely to use an urgent care if it was more accessible. In addition, I examined whether an urgent care with enhanced diagnostic capabilities makes a difference in patient venue selection.
Welfare change was measured in terms of KPCO members’ willingness to pay (WTP) for the improved convenience of the KPCO urgent care network. This reflects the additional
amount a member would pay for a plan that included access to a specific urgent care site. 3.1c Aim 3 - Calculate the return on investment into urgent care centers
I measured the KPCO net revenue change prior to additional urgent care expansion investments. Through the statistical modeling applied in the previous two aims, I
extrapolated the KPCO financial ramifications associated with the urgent care expansion. The overall intent was to align KPCO member venue choices with operational investments.
3.2 Conceptual Framework
Discrete choice models are built on the assumption that decisionmakers make choices that maximize their utility. In 1960 Jacob Marschak mathematically confirmed that consumer choices are based on utility maximization, a concept that resulted in a branch of discrete choice models commonly referred to as RUMs.65 The fundamental analytical framework applied in this study is grounded in RUM, which captures individual preferences for differentiated products. Formally, a decision maker i obtains certain utility, Uij, from an
available alternative j (j = 1,2,….,J). The decision maker (consumer) then would choose an alternative with the highest utility: alternative k if and only if Uik > Uij, ∀ ≠ .
As the random utility framework implies, the researcher does not know the
consumer’s utility function. However, the researcher has some information on the decision maker’s Xi attributes as well as on the characteristics of the set of alternatives Aj, ∀ .
Therefore, a researcher can specify a representative utility function Vij that connects
observable characteristics to the decision maker’s actual utility. However, representative utility does not equal consumer’s utility, ≠ . Instead the representative utility function equals consumer utility plus unobserved components that influence the decision maker’s choice. = + . represents the unexplained difference between representative utility and a consumer’s actual utility. Because the researcher does not know ∀ , it is treated as random.
= ( > , ∀ ≠ ) 3.1 = ( + > + , ∀ ≠ ) 3.2
= ( − > − , ∀ ≠ ) 3.3
Different assumptions about the distribution of the unobserved portion of utility, and density function f( ), determine the type of RUM appropriate for the choice behavior analyses. The logit is most commonly used due to its computational simplicity that are due to closed form expressions for this integral.66 The logit model also invokes the assumption that the unobserved portion of utility, , is independently and identically distributed (iid) extreme value for all j alternatives.
= ! "#$ ! (density function)
% = "#$ ! (cumulative distribution)
Formally, logit choice probability is a closed form expression, and the probability of choosing option k out of all available options J is expressed as follows:
=
∑ ""& '& !! 3.4
Because representative utility is linear in parameters, then = )′+ where + is a vector of observed variables relating to alternative j. Then probability function under the logit model becomes:
=
∑ "",-. !,-. !! 3.5
The above formula indicates that probability of decision maker i choosing an alternative k is always between zero and one. If alternative k’s attributes become more desirable, rises, then will approach one. If representative utility for alternative k
decreases then would be closer to zero. The sum of the choice probabilities of all alternatives is equal to one; meaning that a decision maker must choose one of the alternatives.
∑456 = ∑ exp( ) / ∑ exp = 1 3.6
The sigmoid shape (S shape), illustrated in Figure 1, indicates the relationship between the logit probability and representative utility. The nature of the shape suggests that if representative utility of an alternative is low, then incremental increases do very little to change the probability of this alternative’s selection. When representative utility is very high, again, any incremental changes in utility do very little to change the probability of this alternative’s selection. However as representative utility approaches probability of 0.5 then any incremental changes in representative utility results in a significant probability of changing a decision maker’s selection. In other words, if the value of a particular alternative is low then small improvements do not make a significant difference in consumer selection because other alternatives are still much better. Similarly, if the value of an alternative is very high, then any further improvements will not make a significant difference in the selection process. However, if the value of an alternative is closer to the 50-50 mark, then marginal improvements in that alternative will significantly influence consumer choice. 3.3 Research Approach
This study applied a discrete choice framework to a specific multinomial choice scenario. In recent years, KPCO introduced, expanded, and enhanced several urgent care centers in the metro Denver area to mitigate avoidable ED visits and to improve patient
Figure 1. Graph of Logit Curve.
access to same day services. Currently, there are four urgent care centers within the KPCO network. In 2017, KPCO merged three out of its four urgent care clinics with Regional Acute Diagnostic and Reference Centers (RADAR) to enhance urgent care services. Urgent Care WEST is the only urgent care that does not have a RADAR component as of spring 2019, which means it lacks access to advanced imaging (CT scan). This limitation is due to the physical space constraints specific to this clinic. Otherwise workflow, operating hours, and ASAP ED application are consistent across all four locations. New urgent care centers have several added features including:
1. Seven days per week access, with 8 am to 8 pm operating hours
2. Emergency Medicine physicians and Family Medicine physicians with the experience to address high acuity cases
3. Nurses trained to support ED workflow 4. Bedside check-ins and financial counseling
5. Onsite access to pharmacy, lab, X-ray, CT scan, IV, IM, observation unit and cardiac monitoring
6. Triage to primary care for continuity care
This study investigated the decision-making process that leads KPCO members to select a care venue for their non-emergent medical needs. Specific observable and unobservable factors contribute to venue selection. When a KPCO member seeks non-emergent, time sensitive medical care, they have several options. A member may choose to schedule an appointment with their primary care provider, seek help at KPCO urgent care center, or visit an ED. Hypothetically, a KPCO member may also choose to visit a non-KP urgent care center, but because this study was focused on the substitution effect between EDs and KPCO urgent care centers, non-KPCO urgent care venues were not included in the analyses.
Formally, a KPCO member i derives utility, Uij, from using an available ambulatory
care venue j (j = urgent care center, or primary care office, or ED). The KPCO member then would choose a care venue with the highest utility: alternative k if and only if Uik > Uij, ∀ j
≠k. As a researcher, I did not know KP members’ preferences, and could not directly observe the true utility derived from choosing each care option. However, using information about observable attributes (Xi ) of KPCO members, including primary diagnosis, gender, age, day
and month of the visit, insurance plan, home address, and characteristics of the set of care venues (Aj) such as type of venue, location, and menu of services, I could obtain an estimate
of each member’s utility. Combined, the vectors of observed and unobserved characteristics contain all elements of the care venue and personal preference that influence KPCO
member choice. A representative utility function that captures the value that KPCO member i derives from choosing care venue j is as follows:
+ , 7 , 8 = 9: + +′ Г7 + <6= + <>= 7 + <?= : − @(A , B ) (B ) + Ɛ 3.7
Xi[Yi, Di] is patient i’s attributes that include KPCO member’s socioeconomic factors Yi as
well as primary diagnosis Di that may influence services required for patient care.
Aj[Rj, Sj] is a function of observable attributes of the care venue alternatives. Rj includes
venue features that are common across all patient conditions. Sj includes venue features
that depend on the patient’s diagnosis, such as cardiac monitoring.
Li is the geographical location of a patient’s home and Tij = Tj(Li) is the approximate travel
time from a KPCO member’s zip code to the venue j.
Pj(Di) is the price that KPCO member i with primary diagnoses Di pays for the services at the
chosen venue (ED, primary care clinic, or urgent care center)
The representative utility of a specific venue choice in a logit demand model where venue option ∈ {1, … . , H} is:
(H/A , B , 8 , ) = JKLM4∈4N + , A,B,8 + O 3.8
= P[R MS ( (+ , T
∈4 , 8 ))]
3.3a Aim 1 - Estimate the degree of substitution between the utilization of ED, primary care, and urgent care centers for a set of common diagnoses
