Nursing home transitions: a new framework for understanding preferences, barriers and outcomes

PDF

Download

Open document

Flip pages

Summarize

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

NURSING HOME TRANSITIONS: A NEW FRAMEWORK FOR UNDERSTANDING
PREFERENCES, BARRIERS AND OUTCOMES

by

Kathryn Elizabeth Thomas

A Dissertation Presented to the
FACULTY OF THE SCHOOL OF GERONTOLOGY
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(GERONTOLOGY)

August 2009

Copyright 2009 Kathryn Elizabeth Thomas

ii
DEDICATION

This dissertation is dedicated in loving memory to my grandparents -
Harry & Elsie Thomas and James & Kathryn McGowan - and to Dorothy Miller,
my first babysitter and surrogate grandmother. Together they provided my
early, positive exposure to older adults, which planted the seeds of my passion
for gerontology.

iii
ACKNOWLEDGEMENTS

I would like to thank my family and friends for their support during the
Ph.D. program. Mom and Dad, thank you for being the best roommates,
support system and proofreaders a daughter could ask for. Ted, thank you for
making me smile throughout this process and for helping me stay focused on
the big picture.
I would also like to thank my committee for their guidance and feedback.
I would specifically like to thank my mentor and dissertation committee chair,
Kathleen Wilber.
Finally, I would like to thank Zach Gassoumis for all his help, which he
provided so generously and selflessly.

iv
TABLE OF CONTENTS

Dedication ii

Acknowledgements iii

List of Tables v

List of Figures vi

Abstract vii

Chapter I Introduction 1

Chapter II Unit of Analysis - Episodes of Care 22

Chapter III Methods 48

Chapter IV Understanding the Effects of Transition-Specific Enabling 63
Variables, Insurance and Age on Nursing Home Transition

Chapter V Understanding Barriers to Transition Among Nursing 88
Home Residents Who Prefer to Return to the Community

Chapter VI Conclusion 109

Bibliography 127

Appendix SAS Code for Episode Calculation 134

v
LIST OF TABLES

Table 2.1: Scenario Logic to Create Episodes of Care 37

Table 2.2: Distribution of Shortbounces 43

Table 2.3: Bivariate Comparisons of SCAN versus Medicare 44

Table 4.1: Descriptives and Bivariate Comparisons 66
(65+ Sample, n = 4635)

Table 4.2: Logistic Regression, Predicting Community Discharge 69
within 90 days (65+ Sample, n = 4635)

Table 4.3: Descriptives and Bivariate Comparisons 72
(65-85 Sample, n = 2755)

Table 4.4: Descriptives and Bivariate Comparisons 73
(85+ Sample, n = 1880)

Table 4.5: Logistic Regression, Predicting Community Discharge 74
within 90 days (65-85 Sample, n = 2755)

Table 4.6: Logistic Regression, Predicting Community Discharge 75
within 90 days (85+ Sample, n = 1880)

Table 4.7: Logistic Regression, Predicting Community Discharge 77
within 90 days, Controlling for Predisposing, Need and Generic

Table 5.1: Descriptives and Bivariate Comparisons by Q1a 91
(65+ Sample, n = 4635)

Table 5.2: Descriptives and Bivariate Comparisons by Community 93
Discharge (Q1a = Yes, 65+ Sample, n = 2935)

Table 5.3: Logistic Regression, Predicting Community Discharge 95
(Q1a = Yes, 65+ Sample, n = 2935)

Table 5.4: Logistic Regression, Predicting Community Discharge 98
(Q1a = Yes, 65-85 Sample, n = 1806)

vi
LIST OF FIGURES

Figure 1.1: Episode Calculation 16

Figure 1.2: Conceptual Model, Andersen’s Behavioral Model 19
of Health Service Use

Figure 2.1: Episode Calculation 28

Figure 2.2: Stay Length Versus Episode Length 40

Figure 2.3: Cumulative Percentage of Stay and Episode Lengths 41

Figure 2.4: Discharge Destination Distribution 42

Figure 3.1: Conceptual Model, Andersen’s Behavioral Model of 49
Health Service Use

Figure 3.2: Specific Model Predicting Nursing Home Utilization 62

vii
ABSTRACT

This dissertation contributes to the larger policy objective of rebalancing
the long-term care system in the United States. Nursing home transition is one
of several strategies that states are using in order to reduce long-term care
spending, be responsive to consumer preferences and comply with the
Olmstead Decision (Kasper and O’Malley, 2006).
Many policy reports have been written about the outcomes and
implementation issues associated with transition programs, but little empirical
research has been done on the topic. Nursing Home Minimum Data Set (MDS)
data are underutilized in nursing home transition efforts and in empirical studies
of transition outcomes.
Using the Andersen Behavioral Model of Health Service Use as a
framework, this research examines transition outcomes in the context of
predisposing, need and enabling characteristics. Enabling variables, which
speak to the social and structural context surrounding health care utilization, are
more mutable than predisposing and need characteristics and thus are more
responsive to policy changes. Therefore, because of their policy relevance, the
enabling variables are the focus of this research. Specifically, the following
transition-specific enabling variables are the primary variables of interest:
community-living preference, presence of a support person who is positive
toward discharge, predicted time to discharge and receipt of community living
skills training.

viii
Chapter 1 provides a general introduction to long-term care rebalancing,
nursing home transition and the Minimum Data Set. Chapter 2 focuses on the
theoretical and practical issues surrounding the use of aggregated episodes of
care as opposed to unaggregated individual stays in nursing home outcomes
research. Chapter 3 outlines the conceptual model, sample and measures that
are the basis of the analysis in Chapters 4 and 5. Chapter 4 examines the
effect of transition-specific enabling characteristics, insurance and age on
community discharge outcomes. Chapter 5 explores the social and structural
barriers to transition that prevent residents with a preference to return to the
community from doing so. Finally, Chapter 6 provides a discussion of the
dissertation as a whole and provides direction for future nursing home transition
research.
1
CHAPTER I: INTRODUCTION

A. Background and Significance

This dissertation contributes to the larger policy objective of rebalancing
the long-term care system in the United States. Long-term care rebalancing is
defined as serving a larger number of individuals with long-term care needs in
community rather than institutional settings. Rebalancing also involves the
shifting of resources away from institutional care toward home and community-
based services in order to ‘balance’ Medicaid long-term care spending (Kassner
et al., 2008). Nursing home transition is one of several strategies that states
are using as part of their rebalancing efforts.
Among policy makers, practitioners and the general public, there is
increased interest in transitioning nursing home residents back to the
community. This increased interest is driven by three interrelated trends: (1)
state and federal efforts to contain rising long-term care spending associated
with the aging of the population; (2) consumer and family preferences for
consumer choice regarding long-term care settings; and (3) increased
emphasis on administering care in the most integrated setting possible. The
third trend is in response to the Supreme Court’s 1999 Olmstead Decision,
which declared unnecessary institutionalization a violation of the Americans
with Disabilities Act (Chapin et al., 1998; Kasper & O’Malley, 2006).
Population Aging and Escalating Long-term Care Costs
Caring for the rapidly aging older adult population is a significant policy
concern. The number and proportion of older adults in American society is

2
projected to grow significantly in the coming decades. According to the U.S.
Census, 12.4% of the population was age 65 and older in 2001. The number of
people in that age bracket is expected to reach 71 million or 20% of the
population by 2030. Furthermore, the oldest-old, defined as individuals 85
years and older, are the fastest growing segment of the population and the
most likely to need and use long-term care services (Poon et al., 2005).
A significant portion of long-term care services is paid for by public funds.
Specifically, Medicaid finances almost half (49%) of all long-term care expenses
and Medicare pays for 20%, most of which is for rehabilitative and post-acute
services. Families finance 18% out-of-pocket and private insurance pays for
7%. The remaining balance is covered by other private and public sources
(Feder et al., 2007).
According to a 2006 MetLife Market Survey, the average cost of nursing
home care is more than $67,000 per year and can reach $100,000 in some
cities (MetLife Mature Market Institute, 2006). Due to the high cost of nursing
home care, states are trying to reduce unnecessary institutionalization and
provide community-based alternatives to nursing home care. Older adults and
their families generally prefer community-based care and it is potentially more
cost-effective than institutional care (Fischer & Hoffman, 1984).
In 2006, seventy-five percent of all Medicaid funds spent on long-term
care of older adults and adults with disabilities went toward institutional care
(Kassner et al., 2008). There is great variation among states with respect to

3
progress in balancing Medicaid long-term care spending. Seven states,
including California, spend at least forty percent of Medicaid long-term care
funding on home and community-based services. Alternatively, some states
still spend less than five percent of their budget on community-based care
(Kassner et al., 2008).
Consumer choice
Historically, the long-term care system in the United States has had a
strong institutional bias (Batavia, 2002). However, fundamental changes in the
health care system are resulting in higher levels of consumer involvement in the
management of personal health care. Additionally, over the past decade, there
has been a significant increase in the amount of health information available to
consumers (Shaller, 2005). The result is a shift away from the traditional model
of publicly supported, institution-based long-term care and toward more flexible
and consumer-driven delivery and financing models (Kapp, 2000).
Consumer direction is closely related to consumer choice, but there are
important distinctions. Consumer direction refers to control within a specified
care setting while consumer choice operates at a higher level to allow an
individual to choose a model of care that accommodates his/her desired
amount of consumer direction (Batavia, 2002). Consumer choice in the context
of long-term health care is the ability of consumers to exercise autonomy and
control over how, where, when and by whom their assistance and personal care
needs are met (Batavia, 2002).

4
Individuals want control and choice over their daily lives. According to a
recent AARP survey, 84% of people age fifty and older want to “age in place”
and, as needed, receive long-term care in the community (Gibson, 2003). Yet
despite this strong consumer preference, a disproportionate share of Medicaid
resources continues to be allocated to institutional services.
Policy Emphasis on Reducing Unnecessary Institutionalization
For more than 30 years, consumer advocates, academic researchers,
and policy makers have worked to develop effective home and community-
based alternatives to institutionalization. Generally, most efforts targeted at the
older adult population have focused on diversion or delay of nursing facility
entry for community dwelling individuals at risk of placement rather than
transition of those currently residing in facilities. However, over the last decade,
a variety of demonstrations have been initiated to help transition current nursing
facility residents back into a community setting (Kassner et al, 2008).
This new approach reflects the growing consensus that:
There are older adults living in nursing homes who want to return to the
community and could do so cost-effectively;
Some people improve rather than decline in the nursing home; and
Individuals with disabilities and physical limitations can indeed live in the
community (Anderson et al., 2006).
When Medicaid was established in 1965, nursing homes were viewed as
the preferred and in many instances the only long-term care setting. A decade

5
and a half later, in 1981, Congress passed the Section 1915(c) amendments to
the Social Security Act. These amendments granted the Department of Human
Services authority to waive federal rules allowing states that submitted federal
waivers to develop home and community-based alternatives to institutional-
based care, under specified conditions. There are currently 287 active 1915(c)
waiver programs across the country which provide a combination of traditional
medical services and non-medical services such as respite, case management,
homemaker services and adult day health care (Kassner et al., 2008).
In 1999, the Supreme Court's Olmstead Decision declared unnecessary
institutionalization a violation of the Americans with Disabilities Act. This ruling
requires that people with disabilities and long-term care needs be served in the
most integrated and least restrictive setting possible (Williams, 2000).
In response to the Olmstead Decision, President George W. Bush issued
the New Freedom Initiative in 2001. This executive order called for a
comprehensive assessment of all federal policies, regulations and programs to
identify barriers to community living for people of all ages with disabilities and
long-term illness. Under the New Freedom Initiative, the Centers for Medicare
and Medicaid (CMS) began making a series of grants to states and non-profit
agencies for the development of infrastructure and programs to support long-
term care rebalancing. The New Freedom Initiative also clarified that Medicaid
would pay for one-time expenses related to nursing home transition (Kassner et
al, 2008).

6
Between 1998 and 2000, CMS funded Nursing Home Transition
Demonstration Programs in twelve states. Since then, CMS has funded a
number of rebalancing initiatives to transition individuals out of nursing homes
and to promote flexible long-term care funding mechanisms (Reinhard et al.,
2005).
Between 2001 and 2008, CMS awarded 334 Real Choice Systems
Change grants in all fifty states and the District of Columbia. The grants totaled
$270 million dollars over seven different funding cycles (Kassner et al., 2008).
The purpose of these grants was to create enduring systemic change in the
long-term care system. The grantees have used the funds to expand personal
assistance programs, to transition individuals from institutions to the community,
to provide consumer-directed services and to improve access to long-term care
services and supports (Kassner et al., 2008).
In 2001, CMS funded two types of Nursing Facility Transition grants -
one set to state programs and another set to Independent Living Center
Partnerships. The funds were intended to help grantees establish and operate
nursing home transition programs and to identify Medicaid transition candidates
in nursing homes. In 2002, CMS awarded fourteen more Nursing Home
Transition grants (Kassner et al, 2008).
In 2003, CMS awarded grants to nine states to fund the development and
support of Money Follows the Person initiatives. CMS defines Money Follows
the Person as “a system of flexible financing for long-term services and

7
supports that enables available funds to move with the individual to the most
appropriate and preferred setting as the individual’s needs and preferences
change” (Anderson et al., 2006, p. 12). The Money Follows the Person
approach has two major components - a flexible financial system and a nursing
facility transition program (Anderson et al., 2006).
In 2003, CMS partnered with the Administration on Aging and began
funding Aging and Disability Resource Center (ADRC) grants. ADRCs are
physical locations that offer a one-stop, single entry point for information,
referral and access into public and private long-term care services and
supports. As of February 2008, there were 140 pilot ADRCs in 43 states and
territories (Kassner et al., 2008).
In the Deficit Reduction Act of 2005, Congress broadly defined Money
Follows the Person as the “elimination of barriers or mechanisms, whether in
state law, the state Medicaid plan, the state budget or otherwise, that prevent or
restrict the flexible use of Medicaid funds to enable Medicaid-eligible individuals
to receive support for appropriate and necessary long-term care services in the
settings of their choice” (S.1932 Section 6071).
The Deficit Reduction Act of 2005 also authorized a second wave of
Money Follows the Person demonstration grants, which provide enhanced
Medicaid-matched funds to states in order to pay for the first year of home and
community-based services for long-stay individuals who transition from a
nursing home to the qualified community residence. A qualified community

8
residence is defined as a home owned or leased by the individual or the
individual's family member, an apartment with an individual lease, or a
community-based residential setting where four or fewer unrelated individuals
reside. In order to receive the enhanced Medicaid match, the individual must
have resided in a nursing home for at least six months prior to transition.
As of February 2008, CMS had awarded $1.4 billion worth of Money
Follows the Person demonstration grants in 31 states. If states are able to
meet the strict requirements associated with the grants, Money Follows the
Person will be the largest demonstration project in the history of Medicaid
(Kassner et al., 2008).
B. Role of the Minimum Data Set in Nursing Home Transition
Transition programs are intended to help individuals return to the
community who, without such assistance, would remain in nursing homes.
However, transition resources are limited and states must find a way to balance
the judicious use of limited transition resources with equitable access to those
resources (Anderson et al., 2006). Therefore, a critical component of nursing
home transition programs is the targeting residents who have the preference to
return to the community but might have difficulty transitioning without additional
support. In the context of nursing home transition, targeting refers to the
identification of potential transition candidates and giving them access to
specific transition resources.

9
In order to target potential transition candidates, it is necessary to
differentiate between two distinct populations within nursing homes: short-term,
rehabilitation patients who are likely to return to the community without
intervention and residents likely to have a long-stay institutional placement.
The complicated goal of targeting is to identify people who are the cusp
between these two groups; specifically rehabilitation patients at risk of
becoming long-stay and chronic, custodial care residents who could potentially
be cared for in the community (Nishita et al., 2008; Keeler, Kane & Solomon,
1981; Liu & Palesch, 1981). Developing targeting criteria is also difficult to
develop because transition is complex and each individual’s situation is unique.
One tool to identify residents with a community-living preference is the
Nursing Home Minimum Data Set (MDS). The MDS is a federally-mandated
Resident Assessment Instrument that must be administered to all residents in
Medicare or Medicaid certified nursing facilities within fourteen days of
admission and at regular, prescribed intervals during their stay (Morris et al.,
1990). The MDS is the only national database of information collected about
individual nursing home residents. As a key component of the Resident
Assessment Instrument, it provides a comprehensive assessment of the
functional, psychosocial, cognitive and medical status of each resident. The
reliability and validity of the MDS 2.0 instrument and data have been
extensively evaluated in the literature (Phillips et al., 1996; Morris, Hawes, Fries
et al., 1990).

10
Resident Assessment Protocols (RAP) are also part of the Resident
Assessment Instrument. RAPs are used in conjunction with the MDS 2.0.
Specific answers to MDS items can trigger the initiation of a RAP. The purpose
of RAPs is to identify when follow-up action is required of the nursing home
staff, to assist in decision-making and to support the development of
individualized care plans.
CMS collect approximately ten million MDS records each year (Mor,
2004) on the roughly three million individuals who receive care in a nursing
facility each year (Tucker & Decker, 2004). Despite its rich and relevant data,
the MDS is rarely used to systematically identify potential transition candidates.
A 2006 report found that, out of 33 grantees, only two states used the MDS to
generate a list of potential transition candidates (Reinhard & Hendrickson,
2006).
There are a number of reasons that transition programs do not use the
MDS. These reasons include difficulty getting a MDS Data Use Agreement in a
timely manner, difficulty making useful queries of the complex dataset, other
sources of referrals, limited resources and reservations about the validity and
reliability of the variables which address transition potential in Section Q:
Discharge Potential (Reinhard & Hendrickson, 2006).
MDS 2.0 Section Q
In the past five years, advocates have become aware of Section Q, an
isolated and rarely analyzed part of the MDS 2.0. Although the MDS is

11
administered quarterly, Section Q is only asked upon admission and during
annual assessments. Section Q1 of the MDS 2.0 contains the following three
questions about discharge potential:
(Q1a) Resident expresses/indicates preference to return to the community
[no/yes];
(Q1b) Resident has a support person who is positive toward discharge
[no/yes];
(Q1c) Stay projected to be of a short duration – discharge projected within
90 days?
0 - discharge not projected within 90 days
1 - discharge projected within 30 days
2 - discharge projected within 31 – 90 days
3 - discharge status uncertain
(Centers for Medicare and Medicaid, 2002).
CMS now includes Section Q variables online in the quarterly Active
Resident Information Report. This report contains information collected on
residents currently in nursing homes in the United States, which can be broken
down by state and county. According to the fourth quarter 2008 CMS MDS
Active Resident Information Report, 22.4% of the MDS assessments nationally
had a yes response to question Q1a (“resident expresses/indicates a
preference to return to the community”) and approximately 19% of assessments
had an affirmative answer to question Q1b (“resident has a support person who

12
is positive towards discharge) (Centers for Medicare and Medicaid, 2008). It is
important to note that, due to the population mix in nursing homes, this
snapshot view tends to over represent long-stay residents. This is because, on
any given day, 70% of nursing home residents have been there more than six
months (Alecxih, 2007).
In cases where MDS data are used in transition efforts, the preference
question (Q1a) is often the only variable used. According to CMS, the purpose
of the question is to identify potential candidates for discharge. Given the high
percentage of people who prefer to return to the community, simply using Q1a
is unlikely to be a sufficient targeting strategy. For example, each year 3,000 to
4,000 nursing home residents in Kansas answer yes to Q1a. As of 2006,
Kansas was actively pursuing ways to further filter this list (Reinhard and
Hendrickson, 2006). States can use question Q1a as a first step in identifying
potential transition candidates. However, additional information contained in
the MDS is necessary in order to further refine the list.
Using the recently created California Nursing Facility Transition Screen,
Nishita and colleagues found that 46% of respondents expressed a preference
to return to the community and that 23% of them believed a successful
transition was feasible (Nishita et al., 2008). After the residents received
information about housing and community-based services, the percentage who
felt that a successful transition was possible went up to 33%. These findings
suggest that transition decisions are complex and that the receipt of structured

13
information changed the respondents’ perspectives on transitioning. Another
important finding of the study was that, compared to question Q1a on the MDS
2.0, the transition screen identified a higher proportion of residents with a desire
to transition back to the community.
C. Dissertation Research
Although long-term care rebalancing and nursing home transition are
timely topics with significant policy and practice implications, empirical analysis
of the predictors of community discharge is lacking. In Chapters 4 and 5 of this
dissertation, the relationship between transition-specific enabling variables and
transition outcomes are examined. The transition-specific enabling variables
include presence of a support person who is positive toward discharge,
discharge prediction timeframe, and receipt of community living skills training.
However, before the relationship between these variables and transition
outcomes could be examined, the raw data had to be transformed from
individual nursing home stays to episodes of care in order to address
methodological issues that arise when looking at nursing home outcomes. The
episode calculation is the topic of Chapter 2 and provides the foundation of the
analysis done in Chapters 4 and 5.
Episodes of Care
The need to create episodes of care is driven by two inherent issues in
nursing home research: sampling disparities and transitions across care

14
settings. Methodological differences in how these issues are handled result in
substantially different results, which have implications for research and policy.
Sampling is problematic in nursing home research because nursing
homes contain two distinct populations: a smaller short-stay group with high
turnover and a larger long-stay group with a low discharge rate (Keeler, Kane &
Solomon, 1981; Liu & Palesch, 1981). Due to the population mix, prevalence or
cross-sectional sampling over-represents long-stay residents while sampling of
admission or discharge cohorts over-represents short-stay residents (Wayne,
Rhyne, Thompson, & Davis, 1991).
Transitions across care settings also complicate nursing home research,
particularly outcomes research. Lewis, Cretin & Kane (1985) found that half of
all transfers in and out of the nursing home were between an acute hospital and
a nursing home, in either direction. The authors dubbed this a “ping-pong”
pattern and argued that transfers between hospitals and nursing homes ought
to be viewed as components of an episode of long term care.
Studies using traditional length of stay calculations tend to classify short
visits to the hospital as acute discharges rather than brief interruptions of a
longer nursing home stay. Doing so overestimates the number of acute
discharges and short stay residents and obscures much of the richness of the
MDS.
Improving the definition of an episode of care by more carefully defining
the episode of stay calculation allows researchers to examine transition

15
outcomes, which are different from discharge outcomes in subtle but important
ways. Discharge outcomes are described from the perspective of a single care
setting, in this case the nursing home. Admission and discharge assessments
keep track of an individual’s comings and goings at a particular nursing home.
Transition outcomes require a broader perspective. For example, if an
individual is discharged home but returns to the nursing home three days later,
he or she did not have a stable community transition. Episodes of care provide
this broader perspective by allowing researchers to follow individuals outside
the nursing home to see what happens after the discharge. Episodes also
enable researchers to examine transition outcomes without the over
representation of acute discharges.
Building on the work of Mehr et al. (1997), Fisher et al. (2003) and Nishita
et al. (2008), date information from admission, discharge and reentry forms
were used to create episode beginning and end dates. For the purpose of this
dissertation, an episode is operationally defined as a nursing home stay without
an intervening discharge of more than thirty days. Subsequent stays separated
by less than thirty days are concatenated to create an aggregated episode of
care. An episode ends when the resident dies or when the resident remains
outside the nursing home for more than thirty days.
For example, as seen in Figure 1.1 below, the resident was in the nursing
home for 20 days and then discharged to home. Ten days later the individual
was admitted to the acute care hospital. After 10 days, the individual was

16
discharged from the hospital to the nursing home, where he/she stayed for 10
additional days. Upon discharge, the individual was readmitted to the nursing
facility, where he/she died 20 days later.

HOME NH ACUTE ACUTE NH NH NH
100 Day Episode
FIGURE 1.1: EPISODE CALCULATION

Using the traditional single stay calculation this individual’s MDS records
would show 4 discrete nursing home stays: 20-day stay with discharge to home,
10-day stay with discharge to acute, 20-day stay with discharge to acute and
20-day stay with discharge to death. However, using the episode definition in
this study, the four nursing home stays and the intervening bounces are
concatenated to form a single episode of care. This is done because the
resident returned to the nursing home within 30 days of each discharge. In this
example, the episode length for this individual would be 100 days with a single
discharge to death. Additional, detailed information about the episode
calculation can be found in Chapter 2.
Sample
This study is part of a larger study of the SCAN Health Plan. SCAN,
located in Southern California, was one of the original sites of the National
Social Health Maintenance Organization (S/HMO) Demonstration Project. The

17
Health Care Financing Administration funded the S/HMO demonstration project
from 1985 to 2007 (Wooldridge et al., 2001).
The Social HMO is a community-based model of managed care for the
elderly. As a Social HMO, SCAN provided the full range of Medicare benefits
offered by standard Medicare HMOs plus additional services such as care
coordination, prescription drug coverage, chronic care benefits and home and
community-based services such as homemaker services, personal care
services, adult day care, respite care and medical transportation. The extra
services were intended to help members stay in the community and avoid
nursing home admission (Fischer et al., 2003; Newcomer et al., 1986; Leutz et
al., 1985).
The data for this study are from Nursing Home Minimum Data Set (MDS)
version 2.0 records for adults aged 65 and older from SCAN and from a
Medicare Fee-for-Service 5% sample of individuals who entered a nursing
home in one of four Southern California counties (Los Angeles, Orange,
Riverside and San Bernardino) between January 1, 2001 and December 31,
2003. During the time the data were collected, SCAN was a Social HMO.
Conceptual Framework
The conceptual framework for the empirical analysis is Andersen’s
Behavioral Model of Health Services Use (Andersen, 1995; Andersen &
Newman, 1973), which has been widely used to explain older adults’ health
service utilization related to nursing home placement (Miller & Weissert, 2000).

18
As seen in Figure 1.2 below, health behaviors are modeled as a function
of predisposing, need and enabling characteristics in Andersen’s model.
Predisposing characteristics, typically socio-demographic variables, represent
the individual’s propensity to use health care services. These factors are
exogenous and independent of the individual’s need for care or ability to control
service use. The need characteristics represent the most immediate causes of
health service use and include both perceived and evaluated symptoms,
diagnoses and abilities (Andersen & Newman, 1973). Enabling characteristics
describe an individual’s available means for obtaining or avoiding long-term
care services (Wolinsky, 1990). Enabling characteristics represent available
resources or barriers that influence health care utilization such as health
insurance, social support, family and community resources. These
characteristics can often be manipulated through policy changes. Therefore,
due to their policy relevance, the enabling variables are the focus of this
dissertation.

19

D. Organization of the Dissertation
Chapter 1 provided a general introduction to long-term care rebalancing,
nursing home transition and the Minimum Data Set. Chapter 2 focuses on the
theoretical and practical issues surrounding the use of aggregated episodes of
care as opposed to unaggregated individual stays in nursing home outcomes
research. Chapter 3 outlines the conceptual model, sample and measures that
are the basis of the analysis in Chapters 4 and 5. Chapter 4 examines the
effect of transition-specific enabling characteristics, insurance and age on
community discharge outcomes. Chapter 5 explores the social and structural
barriers to transition that prevent residents with a preference to return to the

Predisposing Factors
Individual’s Propensity to use
health care services
Includes demographic and
social structure variables
Need Factors
Individual’s most immediate
causes of service use
Includes both perceived and
evaluated symptoms and
diagnosis-related variables
Need Factors
Individual’s available means for
obtaining or avoiding service use
Includes family and community
resources variables
Health Service Utilization
FIGURE 1.2: CONCEPTUAL MODEL, ANDERSEN’S BEHAVIORAL MODEL
OF HEALTH SERVICE USE

20
community from doing so. Finally, Chapter 6 provides a discussion of the
practice and policy implications of the findings and provides direction for future
nursing home transition research.
E. Contribution to the Literature
This dissertation contributes to the scientific literature in three
interrelated ways. The first is the methodological contribution regarding
episodes of care. Using episodes of care rather than individual stays
addresses some of the inherent issues in long-term care research such as
transitions across care settings and sampling biases. These issues make it
difficult for researchers and practitioners to make comparisons across studies
or to draw meaningful conclusions. Additionally, one of the reasons that the
MDS is underutilized in transition research and practice is because it is difficult
to extract meaningful data from the large and complicated administrative
database. Creating episodes of care from raw MDS data is a difficult task,
especially given the size, complexity and anomalies of the administrative
dataset. By providing detailed information about the episode calculation
process and showing the comparisons of stay versus episode data, the method
can be leveraged by subsequent researchers. This will make the process of
creating episodes more efficient and consistent, and thus will make the MDS
more attractive to researchers and will facilitate comparisons across studies.

21
Second, the results of this dissertation support the argument that the
MDS 2.0 needs to be revised in order to better support the New Freedom
Initiative and provides some suggestions of how to do so.
Third, the predictors of community discharge found in Chapters 4 and 5
can be used to target transition candidates, to design tiered transition strategies
and to shape long-term care rebalancing policy. The policy and practice
implications of the results are described in detail in Chapter 6.

22
CHAPTER II: UNIT OF ANALYSIS - EPISODES OF CARE

A. Introduction

Given the focus on long-term care rebalancing, nursing home utilization
and discharge outcomes are increasingly important topics in long-term care
research, practice and policy. In nursing home outcomes research, there are
theoretical and practical issues surrounding the measurement of length of stay.
This chapter describes the theoretical rationale for using aggregated episodes
of care instead of single nursing home stays; provides a detailed description of
how to create episodes of care from Minimum Data Set (MDS) records; and
compares usage patterns and discharge outcomes using episode versus stay
as the unit of analysis.
One of the primary reasons that the MDS is underutilized in transition
research and practice is because it is difficult to extract meaningful data from
the large and complicated administrative database. Creating episodes of care
from raw MDS data is a difficult task, especially given the size, complexity and
anomalies of the administrative dataset.
Inconsistent length of stay definitions and sampling techniques make it
difficult for researchers to compare results across studies. Using episodes of
care rather than individual stays addresses some of the inherent issues in
nursing home outcomes research, which are described in more detail below.
Furthermore, providing detailed information about the episode calculation

23
method will make the process of creating episodes more efficient and
consistent.
Data Issues
The need to create episodes is driven by two issues inherent in nursing
home research and MDS data: sampling disparities and transitions across care
settings. Methodological differences in how these issues are handled can result
in significantly different results, which have implications for research and policy.
Sampling
Sampling is problematic in nursing home research because nursing
homes contain two distinct populations: a smaller short-stay group with high
turnover and a larger long-stay group with a low discharge rate (Nishita et al.,
2008, Keeler, Kane & Solomon, 1981; Liu and Palesch, 1981).
Differentiating between the two types of residents is difficult given the
limitations of the data collected about nursing facility residents. CMS collect
approximately ten million MDS records each year (Mor, 2004) on the roughly
three million individuals who receive care in a nursing facility each year (Tucker
& Decker, 2004). The MDS, which includes admission and discharge
information, over-represents short-stayers because they come and go more
often and therefore have more MDS records. Seventy-five percent of the 2.5
million annual discharges were for stays of less than six months.
Alternatively, the Active Resident Information (RAI) dataset shows
information about residents in the nursing home on a particular day in time. The

24
RAI captures more long-term residents because, on any particular day, 70% of
nursing home residents have been there more than six months (Alecxih, 2007).
Finally, distribution of residents is very different than distribution of
nursing home days. Half of all nursing home residents have a length of stay
less than six months, but these individuals make up only 5% of all nursing home
days. Similarly, 40% of nursing home days are over the 3-year threshold, but
only 20% of residents spend more than 3 years in a nursing home (Spence &
Wiener, 1990).
Due to the population mix and length of stay patterns, prevalence or
cross-sectional sampling over-represents long-stay residents while sampling of
admission or discharge cohorts over-represents short-stay residents (Wayne,
Rhyne, Thompson, & Davis, 1991). Long-stay residents may be missed entirely
by surveys of discharged residents (Liu & Palesch, 1981). Most of the studies
of nursing home outcomes in the 1980’s and 1990’s used either admission or
discharge cohorts (Mehr, Williams, & Fries, 1997).
There have been several studies in the past quarter century that
examine discharge outcomes of nursing home residents. Traditional research
on nursing home usage patterns has customarily measured the length of a
single nursing home stay. By simply using admission and discharge dates from
each individual stay, these calculations do not take immediately previous or
subsequent nursing homes stays into account and stays interrupted by brief
visits to the hospital are treated as two separate stays. Because of resident

25
transition patterns, traditional length of stay calculations tend to underestimate
total length of stay, thereby overestimating short-stay residents and
underestimating long-stay residents. Using episodes of care as the unit of
analysis and following individuals over time and across settings addresses
some of the shortcomings of traditional sampling techniques.
Transition Patterns
Discharge data can be misleading because many nursing home
discharges are actually brief visits to acute hospitals or transfers to other
nursing homes (Wayne, Rhyne, Thompson & Davis, 1991). Over two decades
ago, Lewis, Cretin & Kane (1985) found that half of all transfers among their
sample were between a nursing home and an acute hospital, in either direction.
The authors dubbed this a “ping-pong” pattern and argued that it is fostered by
reimbursement policies.
A majority of the studies of nursing home discharge outcomes done in
the 1980’s and 1990’s considered acute-care hospitalization to be a final
outcome (Mehr, Williams, & Fries, 1997). However, studies have shown that
most nursing home residents who are admitted to an acute care hospital either
subsequently die or return to a nursing home (Hing et al., 1989; Narin et al.,
1988). Therefore, studies using traditional length of stay calculations tend to
classify short visits to the hospital as acute discharges rather than brief
interruptions of a longer nursing home stay. Doing so overestimates the
number of acute discharges and short stay residents and underestimates the

26
proportion of residents who die at the end of a nursing home stay (Spence &
Wiener, 1990).
Hospitals and nursing homes share the responsibility for the care of older
adults, but the interplay has not been formally recognized in the way data are
collected and analyzed or in current reimbursement policies. For many
decades, some researchers have advocated for using episodes of care rather
than separate encounters when analyzing ambulatory and hospital services
(Hornbrook et al., 1985; Barer & Stoddart, 1981). Similarly, Lewis et al. (1985)
argued that transfers between hospitals and nursing homes need to be viewed
as components of an episode of long term care.
Episodes of care are essential when studying nursing home transitions
because transition outcomes are conceptually different than discharge
outcomes. Traditional length of stay and nursing home utilization studies are
narrowly focused on whether or not an individual is in a particular nursing home.
Transition research requires a broader perspective. For example, if an
individual is discharged home but returns to the nursing home three days later,
he or she had a discharge but did not have a stable community transition. The
episode calculation allows researchers to track an individual after they leave the
nursing home to see if it was a stable discharge or just a “ping pong” bounce.
The episode calculation used in this dissertation and described below
helps mitigate these issues with the raw MDS dataset and facilitates the
examination of transition outcomes as opposed to discharge outcomes.

27
B. Episode Definition
In this dissertation, rather than looking at individual nursing home stays
separately, previous and subsequent stays that meet defined criteria are
concatenated to form episodes of care. Building on the work of Mehr et al.
(1997), Fisher et al. (2003) and Nishita et al. (2008), date information from
admission, discharge and reentry forms are used to create episode beginning
and end dates. For this analysis, an episode is operationally defined as any
consecutive nursing home stays without an intervening discharge of more than
30 days. Consecutive stays separated by less than 30 days are concatenated
to create an aggregated episode of care. An episode ends when the resident
dies or when the resident stabilizes outside the nursing home for more than 30
days (i.e., no nursing home admissions in the 30 days following discharge).
The following examples illustrate various episodes of nursing home care:
• Admitted to nursing home Discharged to acute care setting Outside
nursing home for less than 30 days Re-entered the nursing home
Death
• Multiple transitions separated by less than 30 days each Community
discharge for more than 30 days
• Admitted to nursing home In nursing home at the end of the study
• Admitted to nursing home Death
Figure 2.1 below provides an illustration of the episode calculation. The
figure shows an individual bouncing in and out of the nursing facility; however,

28
the periods of time outside the nursing home are never more than 30 days.
Therefore, the four stays are concatenated to create a single 100-day episode.

HOME NH ACUTE ACUTE NH NH NH
100 Day Episode
FIGURE 2.1: EPISODE CALCULATION

C. Method
Sample
This study is part of a larger study of the SCAN Health Plan. SCAN,
located in Southern California, was one of the original sites of the National
Social Health Maintenance Organization (S/HMO) Demonstration Project. Data
for this study are from Nursing Home Minimum Data Set (MDS) records for
adults aged 65 who were either enrolled in SCAN Health Plan’s Social HMO or
in the Medicare Fee-for-Service 5% sample. The data set includes individuals
who entered a nursing home in one of four Southern California counties (Los
Angeles, Orange, Riverside and San Bernardino) between January 1, 2001 and
December 31, 2003.
For this research, each episode needed to start with a full admission
assessment. The admission assessment is necessary because the transition-
specific enabling variables of interest for this dissertation are only collected on
full assessments.

29
To address right censoring issues, any episode that starts within 120
days of the end of the study is excluded. This is done in order to ensure that all
residents in the nursing home at the end of the study have had an episode
length of at least 90 days.
Nursing Home Minimum Data Set
The MDS is a federally mandated resident assessment instrument that
must be administered to all residents in Medicare or Medicaid-certified nursing
facilities within fourteen days of admission and at regular, prescribed intervals
during their stay (Morris et al., 1990). The MDS is the only national database of
information collected about individual nursing home residents and it provides a
comprehensive assessment of the functional, psychosocial, cognitive and
medical status of each resident. There are various types of assessments that
make up the MDS database. These include admission, quarterly, annual,
significant change in status, discharge, reentry and correction assessments.
There are three types of discharge tracking forms: discharged - return not
anticipated, discharged - return anticipated and discharged prior to completing
initial assessment. The data used in this study were collected using MDS 2.0.
The implementation of the significantly revised MDS 3.0 is scheduled for
October 2010.
Episode Calculation
Although one might assume that there would be a simple pattern of
admissions followed by discharges in the MDS, this is not the case. There are

30
many instances when two admission assessments or discharge assessments
are completed in a row, sometimes on the same day. The discharge
assessment that is filled out when a resident is discharged prior to completing
the initial assessment causes the most issues in the episode calculation.
Consequently, calculating episodes of care from the raw MDS records is difficult
and requires flexible and sophisticated programming logic to account for the
various exceptions and anomalies in the large, complex administrative dataset.
Creating meaningful episodes requires researchers to make
assumptions and decisions about how to address several data issues.
Methodological differences in how these issues are handled can produce
significantly different results. Therefore, it is critically important to be
transparent about the nuances of the calculation and to share code among
researchers so that results can be replicated and appropriate comparisons and
conclusions can be drawn across studies. An overview of the steps required to
concatenate individual stays to create episodes is described below. A portion
of the SAS code is included in the Appendix and the complete code is available
from the author upon request.
Process
The Process described below is illustrated in Table 2.1.
Step 1: Select Admission, Discharge & Reentry Assessments
The first step in the episode calculation is to select just the admission,
discharge and reentry assessments. Based on the definition of an episode of

31
care in this dissertation, the other assessment forms are not needed. Other
researchers may need data from other assessments. If so, the data could be
merged in Step 5 after the episode beginning and end dates are calculated.
Step 2: Create Date of Interest and Assessment Type Variables
Each assessment type has different entry and exit date fields. For
example, entry date on the admission assessment is AB1 and on the reentry
assessment the reentry date is A4a. Therefore, it is necessary to create
uniform entry date and/or exit date variable for each type of assessment type.
Having uniform date variables allows the assessments to be sorted by date as
described below.
In this step, each assessment type is also assigned a number: discharge
(return not anticipated) and discharge (return anticipated) assessments = 1,
discharged prior to completing initial assessment = 2, full admission
assessment = 3 and reentry assessment = 4. To determine this ordering, the
assumption was made that patients who have both a discharge and an entry on
the same day would have discharged (e.g. to an acute care facility) and
returned on the same day, instead of having entered and immediately
discharged. This ranking allows the data to be sorted by person, by date and
by assessment type, which is important for the next step.
Step 3: Preparing the Dataset for Episode Calculation
At this point, for reasons of programming simplicity and processor
efficiency, it is also worthwhile to discard all data not related to date or type of

32
assessment for each individual. Once the episode dates have been
established, these data will be merged back into the dataset using the unique
MDS identification number. Finally, the dataset should be sorted by person
(MDS ID), then by date and then by assessment type. This sorting gets the
assessments in the right order so that the scenarios are stepped through
properly. Before going into the scenarios, episode beginning, episode end and
bounce reference date are each set to blank.
Bounce reference date (shortbdate) is a placeholder variable, which is
used to keep track of whether or not an episode is ‘open’ or ‘closed.’ An open
episode is an episode that already has an episode beginning and the most
recent record was either an admission or reentry assessment. A ‘closed’
episode is one where the previous record was a discharge assessment.
Specifically, bounce reference date keeps track of the date that an individual
bounced back into the nursing home after a brief discharge. If the bounce
reference date is more recent than the current episode end date, it means the
individual has bounced back in and the episode is ‘open.’ If the episode end
date is more recent than the current bounce reference date, it means a
discharge has occurred since the person bounced back into the nursing home.
In this case the episode is ‘closed.’ It is important to note that just because an
episode is closed does not mean it is necessarily the end of the episode. If the
individual bounces back in within thirty days of the current episode end date,
the episode is ‘open’ again.

33
Step 4: Run Scenario Logic to Create Episode Beginning and End Dates
The next step is to calculate episode beginning and end dates for each
of the eight possible scenarios. The scenarios are based on two criteria: (1) the
current values for episode beginning date, episode end date and bounce
reference date and (2) the type of assessment in the current data record. It is
important to note that the order of these scenarios is important and intentional.
When the code is run, it works its way chronologically through the scenarios.
The order of the scenarios should not be changed. The logic is presented in
Table 2.1 and described in detail below.
Scenario 1 is initiated if episode beginning date is blank and the current
record is a full admission assessment. If the entry date on the admission
assessment is more than 120 days before the end of the study period, then
episode beginning date and bounce reference date are both set equal to the
entry date on the admission assessment. If the entry date is too close to the
end of the study then no new episode is created.
Scenario 2 is initiated if the episode has a valid beginning date and the
current record is a discharge assessment. If the episode end date is blank or
the episode is open (bounce reference date is after the episode end date), then
episode end date is set equal to the exit date on the discharge form.
In scenario 3, the episode has valid beginning and end dates and the
current record is a discharged before completing admission assessment’ form.
If the exit date on the discharge assessment form is later than the episode end

34
date then the length of the gap between the episode end date and the
assessment exit date is checked. If the gap is less than or equal to 30 days,
then the episode end date is set equal to the exit date from the discharge form
and the short bounce counter is increased by 1. If the gap is greater than 30
days, then the episode ends with the current episode end date. If applicable,
the next episode will start with the next full admissions assessment for the
individual.
In scenario 4, if the episode has a valid beginning date but does not have
an episode end date and the current record is the ‘discharged before
completing admission assessment’ form, then episode end date is set equal to
the exit date on the discharge form.
Scenario 5 is initiated when the episode has a valid beginning and end
date and the current record is a reentry assessment. If the episode is ‘closed’
(i.e. bounce reference date is less than episode end date) then the gap length
between the episode end date and the reentry date is checked. If the gap is
less than or equal to 30 days, then the individual bounced back into the nursing
home. In this situation, bounce reference date is set to equal the reentry date
and the short bounce counter is increased by 1. If the gap is greater than 30
days, then the episode ends with the current episode end date and the episode
beginning date, episode end date and short bounce count variables are reset.
However, if the episode is ‘open’ (i.e. bounce reference date is greater than the

35
episode end date) then the bounce reference date is set equal to the entry date
on the reentry form.
Scenario 6 is similar to scenario 5 except that the current record is a full
admission assessment. All of the actions are the same except in the case
where the episode is ‘closed’ and the gap is greater than 30 days. In this case,
the previous episode is ended and, since the current record is a full admission
assessment, a new episode can be started. Episode beginning date is set
equal to the entry date on the admission assessment. However, similar to
scenario 1, if the entry date on the admission assessment is within 120 days of
the end of the study, then no new episode is created. In this case, episode
entry date is set to blank and bounce reference date is set equal to the entry
date on the admission assessment.
Scenario 7 is initiated when the episode has a valid episode beginning
date, the current record is either an admission assessment or a reentry form
and it is the last record for the individual. In this case, episode end date is set
equal to January 1, 2004. January 1, 2004 is the day after the end of the study
period. This episode end date signifies that the individual was in the nursing
home at the end of the study.
Scenario 8 is initiated when the episode has a valid episode end date,
the current assessment is a discharge assessment and it is the last record for
the individual. In this case, any duplicate discharge data are disregarded and
the episode is output with the current beginning and end dates.

36
Step 5: Merging Episode Dates with Assessment Data
After the episode scenario code is run, the file contains only the following
variables: MDS ID, episode beginning date, episode end date and short bounce
count. The next step is to merge the admission data that corresponds to each
individual’s episode beginning date and the discharge data that corresponds to
each episode end date.
Assumptions
The following assumptions are made in the above scenarios:
If an individual is in the nursing home and his/her next assessment is an
admissions or reentry assessment, nothing is done. The episode
beginning date remains the same and calculation continues until the next
discharge assessment or until the end of the study.
If an individual has two consecutive discharges, the first one is used for
the episode end date and the second discharge is ignored.
If an individual has a discharge assessment and an entry assessment on
the same day, the assumption was made that he/she was discharged
(e.g. to an acute care facility) and returned on the same day, as opposed
to being admitted and then immediately discharged the same day.
These assumptions are important to keep in mind when interpreting the results.
It is also important that other researchers be explicit about how these situations
are handled in their own research.

37
TABLE 2.1: SCENARIO LOGIC TO CREATE EPISODES OF CARE
# Status Current Record Action
1

No episode
beginning date

Admission
assessment (3)
If entry date is more than 120 days from
the end of the study, then:
Episode beginning date = entry date.
Bounce reference date = entry date
2

Have episode
beginning date

Discharge
assessment (1)
If episode is open (either episode end
date is blank or bounce reference date is
after the episode end date), then
Set episode end date = exit date
from discharge assessment.
3

Have episode
beginning date
Have episode
end date.

Discharged
before
admission
assessment
completed (2)
If exit date is more recent than episode
end date, then check gap:
If gap is ≤ 30 then episode end date
is set equal to the exit date on the
current discharge form. Short
bounce + 1
If gap is > 30 then the episode is
over as of the previous episode end
date.
4

Have episode
beginning date
No episode
end date
Discharged
before prior to
completing
admission
assessment (2)
Set episode end date = exit date
5

Have episode
beginning date
Have episode
end date

Reentry
assessment (4)
If episode is closed (bounce reference
date is before episode end date), then
check gap length between episode end
date and re-entry date:
If gap ≤ 30 then set bounce
reference date = reentry date.
Increase shortbounce count by 1
If gap > 30 days then the episode
has ended as of the previous
episode end date and episode
beginning date, episode end date
and the shortbounce count values
are reset.
If episode is open (bounce reference
date > episode end date), then:
Set shortbdate = entrydate

38

TABLE 2.1, CONTINUED: SCENARIO LOGIC TO CREATE EPISODES OF CARE
# Status Current Record Action
6

Have episode
beginning date
Have episode
end date

Full admission
assessment (3)
If episode is closed (bounce reference
date < episode end date) then check gap
length:
If gap ≤ 30 then set bounce
reference date = entry date.
Increase shortbounce count by 1
If gap > 30 then start a new episode.
o Set episode beginning date =
entry date.
o If entry date < 120 days from
the end of the study then set
episode beginning date =
blank. Set bounce reference
date = entry date.
If episode is open (bounce reference
date > episode end date) then:
Set shortbdate = entrydate
7

Have episode
beginning date
Last record is
another full
admission
assessment or
reentry form
(3,4)
If it’s the last record for the individual,
then
Set episode end date = Jan 1, 2004
(which means the individual was in
the NH at the end of the study).
8

Have episode
end date
Discharge
assessment
(1,2)
This represents a discharge after a
discharge. Do nothing, simply output
episode beginning and end date.

Episode Data Set
The final dataset contains 5618 episodes. There is one record for each
episode and an individual can have more than one episode. By definition, if an
individual has multiple episodes, each episode is separated by at least thirty
days outside the nursing facility and begins with a new full admission
assessment. Each episode entry includes the admissions assessment
variables and discharge assessment variables that correspond to the beginning

39
and end of the episode. Each episode record also includes a count of the
number of short bounces that occurred during that episode.
Comparable Stays
After creating the episodes, the next objective is to compare the episode
sample to the stay sample. However, the episode sample cannot simply be
compared to the raw stay data because, due to the inclusion criteria and
episode definition, not all stays were included in episodes.
In order to make valid comparisons between the episode and single stay
data, only the individual stays that were aggregated into episodes could be
used in the analysis. Working backwards, episode beginning and end dates
were used to bound the individual assessments that fell within each episode for
each individual. Next, scenarios similar to the scenarios in the episode
calculation were run. These scenarios dealt with the same exceptions and
anomalies as the episode calculation. However, there are two key differences
between the episode code and the stay code: (1) the stay code does not check
for gaps between stays or concatenate stays, and (2) re-entry assessments are
used to begin stays but not episodes.
D. Results
After the episode and stay calculations above were run, 9642 stays were
extracted from 5618 episodes. Comparisons of these two samples, each with a
different unit of analysis, are detailed below.

40
Number of Stays vs. Number of Episodes per Person
The average number of stays per person is 1.72 with a standard
deviation of 1.25. The maximum number of stays by any one individual during
the study period is 13 individual stays. Using the episode data, the average
number of episodes per person is 1.16 with a standard deviation of 0.44. The
maximum number of episodes per individual during the study period is 7.
Length of Stay vs. Episode Length
The average length of stay is 126 days compared to the average episode
length of 169 days. The distribution of stay length and episode length are
shown in Figures 2.2 and 2.3. As seen in Figure 2.2, almost 18% of stays are
less than 14 days but just 10% of episodes are less than two weeks. According
to the stay data 20% of individual had stays of more than 180 days. However,
almost 26% of episodes are greater than 180 days.

FIGURE 2.2: STAY LENGTH VERSUS EPISODE LENGTH

0
5
10
15
20
25
30
35
0-14 15-30 31-60 61-90 91-120 121-150 151-180 180+
Days
Stays Episodes

41
Looking at cumulative percent in Figure 2.3, the stay data show 47% of
residents being discharged within thirty days compared to 39% when using
episode data.

Discharge Destination
Figure 2.4 shows a comparison of discharge destinations using stay
versus episode data. Episode data show a higher percentage of residents
being discharged to home without home health (10.2% vs. 7.7%) and to home
with home health (27.9% vs. 20.6%). Additionally, the episode data show a
much lower percentage of individuals being discharged to acute (13.4% vs.
31.9%). It is important to note that the same number of individuals are
discharged to death in both scenarios; but because there are fewer episodes
than stays the percentage discharged to death is higher using episode data.

FIGURE 2.3: CUMULATIVE PERCENTAGE OF STAY AND EPISODE LENGTHS

0
10
20
30
40
50
60
70
80
90
100
0-14 15-30 31-60 61-90 91-120 121-150 151-180 180+
Days
Stays Episodes

42

“Ping-Pong” Bounces
Table 2.2 shows the distribution of short bounces within the episode
data. A short bounce occurs when an individual is discharged from a nursing
home but returns within thirty days. The short bounce count keeps track of the
number of short bounces per episode. Seventy-two percent of the episodes
had zero short bounces, which means the episode was one uninterrupted stay.
Eighteen percent of the episodes had one short bounce, which means two
individual stays were concatenated into a single episode. Five percent of
individuals had two short bounces or three concatenated stays. Two percent
had four stays (three short bounces) joined into a single episode. The
individual with the most bounces in the study had eleven bounces, which
means twelve stays were strung together into a single episode. This data
represent just the bounces that occurred within episodes during the study
0
5
10
15
20
25
30
35
Home, no
home
health
Home with
home
heaalth
Assisted
living /
Board and
care
Acute Death In NH at
End of
Study
Other
Destination
Stays Episodes
FIGURE 2.4: DISCHARGE DESTINATION DISTRIBUTION

43
period. Many of the individuals who were bouncing during the study period
were presumably also bouncing before and/or after the study period.

S/HMO vs. Medicare Comparison
Proponents of Social Health Maintenance Organizations (S/HMO) argue
that prospective payment reimbursement policies discourage nursing homes
from providing intermittent acute care services to residents. The S/HMO
proponents contend that capitated care models encourage more rational
allocation of services across care settings (Wooldridge et al., 2001; Angelelli,
Wilber & Myrtle, 2000). Taking advantage of the unique structure of the dataset
used in this dissertation, some basic comparisons between SCAN and
Medicare were run using the stay data and the episode data.
As seen in Table 2.3, t-tests show that there is no significant difference
between the average length of single stay for SCAN members versus traditional
TABLE 2.2: DISTRIBUTION OF SHORTBOUNCES
# Frequency % Cumm %
0 4668 71.88 71.88
1 1194 18.39 90.27
2 341 5.25 95.52
3 143 2.20 97.72
4 75 1.15 98.88
5 30 0.46 99.34
6 19 0.29 99.63
7 7 0.11 99.74
8 8 0.12 99.86
9 5 0.08 99.94
10 3 0.05 99.98
11 1 0.02 100

44
Medicare beneficiaries. However, using the episode calculation, SCAN
members are shown to have significantly shorter episode lengths compared to
Medicare. SCAN members also have significantly fewer stays per person, fewer
episodes per person and fewer “ping-pong” bounces per episode.

TABLE 2.3: BIVARIATE COMPARISONS OF SCAN VERSUS MEDICARE

t
Mean (SD) Mean (SD) Mean (SD)
Stay Length 126.07 (208.01) 125.59 (215.27) 126.46 (201.90) 0.20 0.8395
Episode Length 168.71 (208.01) 155.33 (244.35) 181.19 (260.96) 4.12 <.0001 ***
Stays/Person 1.72 (1.25) 1.58 (1.04) 1.85 (1.41) 8.38 <.0001 ***
Episodes/Person 1.16 (0.44) 1.14 (0.40) 1.17 (0.48) 2.98 0.0029 **
Shortbounces/Person 0.47 (1.02) 0.36 (0.82) 0.58 (1.16) 8.63 <.0001 ***
* p < 0.05, ** p < 0.01, *** p < 0.001
p Total SCAN Medicare

The lack of a significant difference with respect to single stay length and
the highly significant difference with respect to episode length show how the
data can tell a different story based on the unit of analysis. In the stay data,
SCAN and Medicare have similar lengths of stay, but using the episode data
SCAN members on average have shorter episodes of care by 25 days.
E. Discussion
This chapter describes the theoretical rationale for using aggregated
episodes of care instead of single nursing home stays; provides a detailed
description of how to create episodes of care from MDS records; and compares
usage patterns and discharge outcomes using episode versus stay as the unit
of analysis. The research and policy implications are described below.
Research Implications
The first section brings to light the inherent issues in nursing home
outcomes research and highlights the various sampling techniques and

45
discharge definitions. Depending on how these methodological issues are
handled, different biases will be introduced and the results will vary. Even
apparently simple calculations are likely based on multiple assumptions,
especially when using large administrative databases such as the MDS. It is
critically important that researchers be explicit about sampling methods,
definitions and assumptions. Without doing so, it is virtually impossible to make
comparisons across studies, draw conclusions or replicate results.
In the second section, the episode calculation used in this dissertation is
described in detail and sample SAS code is provided in Appendix A. In most
published studies, the description of the sample selection and analysis method
is limited to just a few paragraphs. This is not sufficient given the complexity of
administrative databases.
It is inefficient and impedes comparison when individual researchers
create unique approaches to complex logic required to measure length of stay.
Making best practices for extracting meaningful data from the large and
complicated MDS available will increase consistency and encourage more
researchers to explore this rich dataset. Ideally there would be consistent
protocols building on a repository of code that all MDS researchers could
reference. This would increase consistency across studies and save
researchers considerable time. Beyond a repository, it would be ideal if there
were a program that researchers could run on their raw MDS data. For
example, the researcher could input episode parameters such as the inclusion

46
criteria (admissions assessment), short bounce length (30 days) and exclusion
criteria (120 days from the end of the study) and the program would return the
episode beginning and end dates according to an agreed upon method.
The final section shows comparisons between the stay sample and the
episode sample. These differences highlight how different the results look
when based on different units of analysis. This is particularly important for
policy relevant data such as the MDS, because different manipulation
techniques can result in different policy implications.
Taken together, this chapter highlights the importance of consistency
and transparency in research and specifically with complex administrative
databases such as the MDS. Considerable time, money and effort are spent
collecting these data, but the value cannot be fully realized without efficient and
consistent analysis methods.
Policy Implications
The growing demand for long-term care combined with the high cost of
these services increases the importance of accurate measurement of nursing
home use. Accurate measurement is particularly important for the design and
pricing of public and private long-term care insurance. Misestimating length of
stay can cause publicly funded insurance programs to be substantially under
funded and private insurance companies to lose considerable amounts of
money (Spence & Weiner, 1990).

47
The episode of care perspective is particularly important for the planning,
organization and delivery of care for the frail elderly, whose healthcare is
predominantly publicly funded by Medicare and Medicaid. This group is cared
for across multiple settings but the interplay has not been formally recognized in
the way the data are collected or analyzed or in reimbursement policies.
The episode calculation used in this study, which tracks residents for
thirty days post-discharge, is just the tip of the iceberg. Ideally, in order to get a
fuller picture of older adults’ health care utilization, researchers need to be able
to track individuals across all settings and for longer periods of time. However,
lessons learned from this calculation of episodes of nursing home care can
inform research methods for tracking older adult utilization across other
healthcare settings.
Limitations
The episode calculation used in this dissertation has some limitations.
Due to the criteria that all episodes must start with a full admissions
assessment, the episode calculation does not necessarily capture people who
were already in the nursing home when the study began. Similarly the episode
calculation underestimates the total episode length of individuals who may have
been bouncing in and out right before the study began. The choice to use 30
days as the gap distance between stays was based on the work of Spence and
Wiener (1990), Nishita et al. (2008) and others. However, future research
should be done to test how other lengths of time affect the results.

48
CHAPTER III: Methods

A. Introduction

This chapter describes the conceptual model, sample, measures and
analyses used in Chapters 4 and 5.
B. Conceptual model
The conceptual framework for the empirical analysis in Chapters 4 and 5
is the Anderson Behavioral Model of Health Service Use (Andersen & Newman,
1973; Andersen, 1995), which has been widely used to explain older adults’
health service utilization (Miller & Weissert, 2000). The health service focus of
this study is the risk of remaining in the nursing facility versus returning to the
community.
According to the model, shown in Figure 3.1 below, the use of health
services is a function of predisposing, need and enabling characteristics.
Predisposing characteristics, typically socio-demographic variables, represent
the individual’s propensity to use health care services. These factors are
exogenous and independent of the individual’s need for care or ability to control
service use. The need characteristics represent the most immediate causes of
health service use and include both perceived and evaluated symptoms,
diagnoses and abilities (Andersen & Newman, 1973). Enabling characteristics
describe an individual’s available means for obtaining or avoiding long-term
care services (Wolinsky, 1990). Enabling characteristics represent available

49
resources or barriers that influence health care utilization such as health
insurance, social support, family and community resources.

The enabling factors are generally the most mutable factors contributing
to health care utilization and can often be manipulated through policy changes.
Therefore, due to their policy relevance, the enabling variables are the focus of
this dissertation. Moreover, due to the transition focus of this dissertation, the
enabling characteristics are divided into two groups: generic and transition-
specific. The transition-specific enabling variables include community living
preference, presence of a support person who is positive toward discharge,
discharge prediction timeframe and receipt of community living skills training.

Predisposing Factors
Individual’s Propensity to use
health care services
Includes demographic and
social structure variables
Need Factors
Individual’s most immediate
causes of service use
Includes both perceived and
evaluated symptoms and
diagnosis-related variables
Need Factors
Individual’s available means for
obtaining or avoiding service use
Includes family and community
resources variables
Health Service Utilization
FIGURE 3.1: CONCEPTUAL MODEL, ANDERSEN’S BEHAVIORAL MODEL
OF HEALTH SERVICE USE

50
These specific MDS variables, which should theoretically be related to
transition, have not been used in predictive models of community discharge.
B. Sample
This dissertation research is part of a larger study of the SCAN Health
Plan. SCAN, located in Southern California, was one of the original sites of the
National Social Health Maintenance Organization (S/HMO) Demonstration
Project. The Social HMO, a model of managed care for the elderly, was
developed in the 1980s. As a Social HMO, SCAN provided the full range of
Medicare benefits offered by standard Medicare HMOs plus additional services
such as care coordination, prescription drug coverage, chronic care benefits
and home and community-based services such as homemaker services,
personal care services, adult day care, respite care and medical transportation.
The extra services were intended to help members stay in the community and
avoid nursing home admission. The primary objective of the original study was
to analyze SCAN’s effectiveness at diverting members from nursing home
placement. The focus of this dissertation research is on transition as opposed
to diversion.
During the time the data were collected, SCAN was a Social HMO.
However, evaluations of the S/HMO demonstration were unable to prove that
the model was worth the substantial additional cost to Medicare (Wooldridge et
al., 2001). Therefore, at the end of the demonstration, the programs were
converted to traditional Medicare + Choice plans. Today, SCAN serves over

51
100,000 older adults in the southwestern United States using a Medicare
managed care model.
Data for this study are from de-identified Nursing Home Minimum Data
Set (MDS) version 2.0 records for adults age 65 and older from SCAN and from
a Medicare Fee-for-Service 5% sample of individuals who entered a nursing
home in one of four Southern California counties (Los Angeles, Orange,
Riverside and San Bernardino) between January 1, 2001 and December 31,
2003. The University of Southern California Institutional Review Board
approved the study as exempt.
Individuals with mental retardation/developmental disabilities and those
in a permanent vegetative state were excluded from the analysis. In order to
remove short-term respite residents and to keep the focus on individuals at risk
of becoming long-stay residents, individuals with episodes of less than two
weeks were also excluded. Studies have shown that most nursing home
residents who are admitted to an acute care hospital either die or return to a
nursing home (Hing et al., 1989; Narin et al., 1988). Therefore, residents who
were discharged to acute and did not return to the nursing home within thirty
days and residents who died before ninety days were excluded from the
analysis.
Episode is the unit of analysis for this dissertation research. After
making the above exclusions, 88% of individuals in this study had just one
episode during the study period. The remaining 12% had more than one

52
episode. By definition, if an individual has multiple episodes, each episode is
separated by at least 30 days outside the nursing home and begins with a new
full admission assessment.
Among the unduplicated sample (n = 4143), the average age is 83 and
two thirds is female. The sample is approximately 75% Caucasian, 11%
Hispanic and 8% African American. Three quarters of the sample has at least a
high school education.
C. Measures
Episode Calculation
As described in detail in Chapter 2, the unit of analysis for this study is
an episode of care as opposed to an individual nursing home stay.
Consecutive stays separated by less than 30 days are concatenated to create
an aggregated episode of care. An episode ends when the resident dies or
when the resident stabilizes outside the nursing home for more than 30 days
(i.e., no nursing home admissions in the 30 days following discharge).
In order to meet the inclusion criteria for this research, all episodes must
start with each individual’s first full admission assessment during the study
period. The admission assessment is necessary because community living
preference information is only collected on full assessments.
To address right censoring issues, any resident whose episode started
within 120 days of the end of the study was excluded from the study. This was

53
done in order to ensure that all residents in the nursing home at the end of the
study have accrued an episode length of at least 90 days.
Transition Outcomes versus Discharge Outcomes
Episodes of care are essential when studying nursing home transitions
because transition outcomes are different from discharge outcomes. Traditional
length of stay and nursing home utilization studies are narrowly focused on
whether or not an individual is in a particular nursing facility. Transition research
requires a broader perspective. For example, if an individual is discharged
home but returns to the nursing home three days later, he or she did not have a
stable community transition. The episode calculation allows researchers to
track an individual once they leave the nursing home to see if it is a stable
discharge in contrast to a “ping pong” bounce.
Dependent Variable
The dependent variable used in Chapters 4 and 5 is a dichotomous
variable where yes (1) represents a community discharge within ninety days
and no (0) represents a nursing home episode of more than ninety days.
Community discharge is defined as a discharge at the end of an episode to one
of the following destinations: home without home health, home with home
health or board and care/assisted living facility.
Independent Variables
The independent variables are either taken directly or derived from the
MDS 2.0 Full Assessment Admission Form. Following Andersen’s behavioral

54
model (Andersen, 1995), the independent variables are classified below as
predisposing, need or enabling characteristics.
The predisposing variables included are: age at admission, gender,
marital status, race, and education. The need variables are cognitive
impairment, depression, comorbidities, social engagement, behavioral
problems, activities of daily living limitations, bowel incontinence, bladder
incontinence, recent fracture, recent fall and admitted-from location.
The enabling variables, the focus of this dissertation, are divided into two
groups: generic and transition-specific. The generic enabling variables include
living situation prior to admission, legal responsibility, payment source and type
of insurance. The transition-specific enabling variables include community
living preference, the presence of a support person who is positive towards
discharge, discharge prediction timeframe and receipt of community living skills
training.
Predisposing Characteristics
The five demographic variables included in the analysis are age at
admission, gender, marital status, race, and education. Age at admission is
calculated by subtracting birth date from date of entry. Age is a continuous
variable and includes fractions of years. Gender is coded as a dichotomous
variable - female or male. Marital status is re-coded as a dichotomous variable
- married or not married. Not married includes individuals with the following
statuses: never married, widowed, separated or divorced. Race/ethnicity is

55
coded into four dummy variables: White (not of Hispanic origin), Hispanic,
African American and Other (Asian/Pacific Islander and American
Indian/Alaskan Native). For education, the eight possible education levels are
recoded into three dummy variables: below average, average (completed high
school) and above average.
Need Characteristics
According to Andersen and Newman (1973), need indicators include
both perceived and evaluated symptoms, diagnoses and abilities.
To create a cognitive measure, items were selected and scored
according to the Cognitive Performance Scale (Morris et al, 1994). The validity
of the MDS 2.0 Cognitive Performance Scale (CPS) has been established in
relation to two standard cognitive assessment tools - the Mini-Mental State
Examination and the Test for Severe Impairment. Items used in the hierarchical
scale are comatose, short-term memory, cognitive skills for decision-making,
making self-understood, and self-performance eating. The range of CPS
scores is 0 to 6, with higher scores representing worse cognitive impairment.
Per convention, in order to dichotomize the index, the three highest categories
(4,5,6) were combined to create the high impairment/low cognitive performance
group and the four lowest categories (0-3) were combined into the low
impairment/high cognitive performance group (Mor et al., 1995)
The Depression Rating Scale (Burrows et al. 2000) was used to create a
depression measure. The additive scale uses the following items from the

56
Mood and Behavior Patterns section of the MDS 2.0: resident made negative
statements; persistent anger with self or others; expression of what appears to
be unrealistic fears; repetitive health complaints; repetitive anxious
complaints/concerns (non health related); sad, pained, worried facial
expression; and crying, tearfulness. Each item is scored from zero to two. Zero
means the indicator was not exhibited in the last 30 days; one means the
indicator was exhibited up to five days a week; and two means the indicator
was exhibited daily or almost daily (six or seven days a week). The scores
have shown good interrater reliability (Morris et al., 1997). The range for the
additive scale is 0 to14 with higher scores corresponding to more depressive
symptoms. The scores were dichotomized using a cut point of 3, which has
been found to differentiate between residents with or without easily observable
depressive symptoms (Burrows et al., 2000).
To measure each resident’s health, a comorbidity scale was created.
The new scale is based on the Charlson Comorbidity Index (CCI), one of the
most widely used comorbidity scoring systems. The CCI score is the sum of the
weights assigned to 19 predefined medical conditions (Charlson et al., 1987).
Unfortunately, not all 19 items used in the CCI are found on the MDS 2.0.
Using the subset items from the CCI index that are available in the MDS 2.0, a
comorbidity index was created by summing the following available disease
diagnoses: diabetes, congestive heart failure, dementia (including Alzheimer’s
disease), cerebrovascular event (stroke), chronic obstructive pulmonary

57
disease, cancer and renal failure. Using the weights from the original Charlson
index as a guide, each disease was given a weight of 1, except cancer and
renal failure, which each received a weight of 2. The range of the scale is 0 to
9, with higher scores representing more comorbidity.
The Index of Social Engagement (ISE) was created by summing the
following six dichotomous MDS 2.0 items: at ease interacting with others, at
ease doing planned or structured events, at ease doing self initiated activities,
establishes own goals, pursues involvement in life of the facility, and accepts
initiations to most group activities (Mor et al., 1995). The items have acceptable
interrater reliability (Sgadari et al., 1997). The index ranges from 0 through 6,
with higher scores corresponding to higher levels of social engagement.
The behavioral symptom variable was created using the five items in the
behavioral symptom section of the MDS 2.0: wandering, verbally abusive,
physically abusive, socially inappropriate or disruptive behavioral symptoms
and resisting care. Each item is scored based on the symptom’s alterability in
the previous 7 days. A score of zero is given if the behavior is not present or if
the behavior was present but easily altered. A score of 1 is given if the
behavior was present and not easily altered. If the individual exhibited at least
one behavior that was not easily altered, the dichotomous behavioral symptom
variable is coded as a 1; otherwise it is coded as 0.
Physical functioning was scored using the MDS Activities of Daily Living
(ADL) Self-Performance Hierarchy. The scale is constructed using the following

58
four individual ADLs: personal hygiene, toileting, locomotion and eating. The
seven-category scale ranges from independent (0) to total dependence (7).
The benefit of the hierarchical scale over other additive scales is that each
category represents a specific performance level and shifts between each level
have substantive meaning (Morris, Fries & Morris, 1999).
Bowel and bladder continence measures are ordered on a scale from 0
to 4, where 0 represents complete control/continence and 4 represents severe
incontinence. Per conventions, individuals who scored a 3 or a 4 on either
measure were recoded as incontinent.
The dichotomous variable, fracture in the last 180 days, was created
using hip fracture in the past 180 days and other fracture in the last 180 days. If
the resident experienced a hip and/or other fracture in the past 180 days then
fracture = 1, otherwise fracture = 0. Similarly, if the resident experienced a fall
in the past 180 days then fall = 1, otherwise fall = 0.
Eight potential admitted-from locations were recoded into three dummy
variables: home without home health, acute hospital or other. The “other”
category includes anyone admitted from a supportive setting (e.g. home with
home health, assisted living/board and care, group home, rehabilitation
hospital, psychiatric hospital, or another nursing home).
Enabling Characteristics
Enabling variables represent the community and support context
surrounding the individual (Anderson & Newman, 1973). Due to the policy

59
relevance of enabling characteristics and the transition focus of this
dissertation, the enabling characteristics are divided into two groups: general
and transition-specific. Surprisingly, these transition-specific variables, which
should theoretically be related to transition, have heretofore not been used in
predictive models of community discharge.
Enabling: General
If the resident lived alone prior to entry, the lived alone variable is coded
as 1. If the individual did not live alone or lived “in another facility” prior to
admission, the lived alone variable was coded as 0. For the legal responsibility
variable, residents who are legally responsible for them self are coded as 1; all
other responsible parties are coded as 0. Previous research has shown that
self-consenting residents are more likely to express a stable transition
preference (Nishita et al., 2008).
An individual in the nursing facility’s billing office checks all that apply
from a list of ten potential payment sources. For the purpose of this study, the
payment source variable is recoded as Medicaid per diem = 1, no Medicaid per
diem = 0. As an indication of type of insurance, subjects are coded as either
SCAN members (1) or as traditional fee-for-service Medicare (0).
Enabling: Transition-Specific
Section Q1 of the MDS 2.0 contains questions regarding the resident’s
discharge preference and potential. Question Q1a asks if the resident
expresses/indicates a preference to return to the community. An affirmative

60
response is coded as 1 and a negative response is coded as 0. According to
the Centers on Medicare and Medicaid, the purpose of the question is to
identify potential candidates for discharge. Question Q1b asks whether or not
the resident has a support person who is positive towards discharge (yes = 1).
The final question in Section Q1 asks the MDS assessor whether, in
their opinion, the resident is predicted to discharge within 90 days. The four
response categories are no - discharge not expected within ninety days,
discharge expected within 30 days, discharge expected within 30 to 90 days or
discharge uncertain.
Section P of the MDS 2.0 asks whether or not the resident is involved in
various special programs. One of the programs is training in skills required to
return to the community. The training includes skills such as taking medication,
activities of daily living, housework, shopping and transportation. This training
may also include training of family or other caregivers. Whether or not a
resident received training in skills required to return to the community is a
dichotomous variable where 1 means that the resident is regularly involved in
activities with a licensed skilled professional to attain the skills necessary for
community living
How long an individual is predicted to stay and whether or not he/she
receives community living skills training are complex variables. They are a
function of the individual, but they also reflect the assessor’s beliefs about the

61
individual’s potential for discharge. These variables may represent ‘tracks’ that
residents are intentionally or unintentionally categorized into upon admission.
D. Analysis
In Chapters 4 and 5, bivariate statistics are used to compare
characteristics across groups. For example t-tests and chi-sq tests are used to
compare residents who transition against those who do not transition.
Both chapters also use logistic regression in order to identify predictors
of transition among the various sub-populations. Three logistic regression
models, which are described in Chapters 4 and 5, are run for each sub-sample.
This step-wise approach makes it possible to isolate and examine the effects of
the transition-specific enabling variables. In each regression model, community
discharge within 90 days is the dichotomous dependent variable; the
predisposing, need and enabling measures, described above and shown in
Figure 3.2 below, are included as the independent variables. In the results
described in Chapters 4 and 5, all differences are highly significant (p < 0.001)
unless otherwise noted.

62

Predisposing Factors
Age
Female
Married
Race
Education
Need Factors
Cognitive impairment
Depression
Cormorbidity
Social engagement
Behavioral problems
ADL limitations
Incontinence (bladder)
Incontinence (bowel)
Recent fracture
Recent fall
Need Factors
Generic
Living situation prior to admission
Legal responsibility
Medicaid
SCAN
Transition-Specific
Community living preference
Support person positive toward discharge
Discharge prediction
Receipt of community living skills training
Nursing Home Utilization
(Community discharge within
90 days versus long-stay in
nursing home)
FIGURE 3.2: SPECIFIC MODEL PREDICTING NURSING HOME UTILIZATION

63
CHAPTER IV: UNDERSTANDING THE EFFECTS OF
TRANSITION-SPECIFIC ENABLING VARIABLES, INSURANCE
AND AGE ON NURSING HOME TRANSITION

A. Introduction

This study investigates nursing facility transition outcomes in the context
of the following transition-specific enabling characteristics: community living
preference, the presence of a support person who is positive towards
discharge, discharge prediction and community living skills training. These
specific MDS 2.0 variables, which should theoretically be related to transition,
have not been used in predictive models of community discharge.
Given the history of the S/HMO model and the unique structure of this
SCAN/Medicare sample, the effect of SCAN membership on nursing home
transition outcomes is also investigated as part of this study.
In almost all industrial countries, including the United States, the 85 plus
group is the fastest growing segment of the population. This group, the “oldest-
old”, are also the most likely segment of the population to need and use long-
term care services. Despite this anticipated growth, most aging research has
focused on the entire 65 plus population, which obscures much of the diversity
encompassed within this large age span. The oldest-old are an increasingly
important group to understand and for whom to prepare. This is the rationale for
looking specifically at the role of preference among the oldest-old.

64
Research Questions
This study seeks to answer the following research questions:
1. Do transition-specific enabling variables affect transition outcomes?
2. Do these factors matter differentially by age?
3. Does SCAN membership affect transition outcomes? Does the effect of
SCAN matter differentially by age?
Analysis
The conceptual model, sample and measures used in this analysis are
described in detail in Chapter 3.
The following analytical steps were taken in order to answer the above
research questions. First, bivariate analyses are used to compare the
characteristics of residents with a community discharge within 90 days to
residents who had episode lengths greater than 90 days. Next, the following
three logistic regression models were run to identify predictors of community
discharge:
Model 1 (base model) = predisposing, need and generic enabling
characteristics.
Model 2 = Model 1 + community living preference + presence of a
support person who is positive towards discharge.
Model 3 (full model) = Model 2 + discharge prediction + receipt of
community living skills training.

65
In each model, community discharge within 90 days is the dichotomous
dependent variable. The predisposing, need and enabling measures described
in Chapter 3 are included as the independent variables. To examine the age
differences, the same analyses were done separately on sub-samples of
individuals 65 to 85 and 85 and older. In the results below, all differences are
highly significant (p < 0.001) unless otherwise noted.
C. Results
Table 4.1 (below) shows the characteristics of the total sample (n = 4635),
which includes all residents age 65 and older who have an episode length
greater than 14 days. The average age of the sample is 83 years old. The
sample is 67% female, 29% married and 74% Caucasian. Seventy-five percent
of the sample has at least a high school education.

66

Research Question #1: Do transition-specific enabling variables affect transition
outcomes?
Bivariate Analysis
The total sample is broken down into two groups: those who return to the
community within 90 days and those who are in the nursing home for at least 90
days (Table 4.1). As expected, the two groups differ in many ways. The

TABLE 4.1: DESCRIPTIVES AND BIVARIATE COMPARISONS (65+ SAMPLE, N = 4635)

Total No Transition Yes, Transition
N = 4635 N = 2317 N = 2318
Mean SD mean SD mean SD t chi sq p sig
Predisposing
Age 83.01 7.22 83.21 7.37 82.80 7.07 1.94 0.053
Female 0.67 0.65 0.69 7.09 0.0077 **
Married 0.29 0.27 0.31 9.17 0.0025 **
Caucasian 0.74 0.69 0.80 73.16 <.0001 ***
Hispanic 0.11 0.12 0.09 14.29 0.0002 ***
African American 0.08 0.10 0.05 38.92 <.0001 ***
Race - Other 0.07 0.08 0.06 11.96 0.0005 ***
Low Education 0.25 0.30 0.20 64.05 <.0001 ***
Average Education 0.48 0.47 0.49 2.06 0.1512
High Education 0.27 0.23 0.31 37.85 <.0001 ***
Need
Cognitively Impaired 0.19 0.28 0.10 252.93 <.0001 ***
Depressed 0.04 0.05 0.03 10.78 0.001 **
Comorbidities 1.42 1.21 1.57 1.21 1.27 1.20 8.22 <.0001 ***
Socially Engaged 0.08 0.08 0.09 1.35 0.2458
Behavioral Problems 0.10 0.12 0.07 44.46 <.0001 ***
ADL Score 3.75 1.44 3.96 1.46 3.53 1.38 10.3 <.0001 ***
Incontinence - Bladder 0.40 0.49 0.30 161.78 <.0001 ***
Incontinence - Bowel 0.42 0.55 0.30 293.08 <.0001 ***
Recent Fracture 0.17 0.11 0.23 117.44 <.0001 ***
Recent Fall 0.42 0.37 0.47 43.80 <.0001 ***
Admitted from Acute 0.81 0.72 0.91 269.66 <.0001 ***
Admitted from Home 0.08 0.10 0.05 48.76 <.0001 ***
Admitted from Other 0.11 0.18 0.04 211.34 <.0001 ***
Enabling: Generic
Lived Alone 0.25 0.19 0.31 95.88 <.0001 ***
Legally Responsible for Self 0.45 0.32 0.59 332.38 <.0001 ***
Medicaid 0.17 0.26 0.08 276.92 <.0001 ***
SCAN 0.48 0.44 0.52 27.54 <.0001 ***
Enabling: Transition-Specific
Q1a) Community Living Preference 0.63 0.41 0.86 997.75 <.0001 ***
Q1b) Support Person 0.64 0.40 0.88 1140.16 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.27 0.11 0.43 608.78 <.0001 ***
Predicted 30-90 days 0.10 0.07 0.13 50.98 <.0001 ***
Prediction Uncertain 0.37 0.36 0.38 1.94 0.1641
Predicted long-stay 0.26 0.46 0.06 985.10 <.0001 ***
P1ar) Community Living Skills Training 0.45 0.30 0.60 442.70 <.0001 ***
Episode Length 224.53 280.88 416.56 289.40 32.57 18.21 63.74 <.0001 ***
* p < 0.05, ** p < 0.01, *** p < 0.001

67
average length of stay of individuals who are discharged to the community is
almost 33 days (SD 18.21), compared to over 400 days (SD 289.4) for the long-
stay group.
Looking first at the predisposing characteristics, the community discharge
group is significantly more likely to be female, married, Caucasian, and well
educated. Among the need characteristics, individuals who reenter the
community have less cognitive impairment, fewer depressive symptoms (p <
0.01), less comorbidity, fewer behavioral problems, fewer ADL limitations, and
less incontinence. Those who transition are more likely to have had a recent
fracture or fall, more likely to have been admitted from an acute care hospital
and less likely to have been admitted from home or from another supportive
setting.
The two groups also have many significant differences among the generic
enabling characteristics. Those who transition are significantly more likely to
have lived alone prior to admission and to be legally responsible for
themselves. They are also significantly less likely to be on Medicaid and
significantly more likely to have the SCAN S/HMO Health Plan.
Transition-Specific Enabling Characteristics
Looking at the transition-specific enabling variables in Table 4.1, 86% of
those who return to the community expressed a preference to return to the
community compared to 41% of those who did not return to the community.
Those who return to the community are also significantly more likely to have a

68
support person who is positive toward discharge (88% vs. 40%). The transition
group is more likely to have a predicted discharge of less than 30 days (43% vs.
11%) or to be predicted to be discharged between 30 and 90 days (13% vs.
7%). Alternatively, residents who do not reenter the community are significantly
more likely to have a predicted stay of at least 90 days (46% vs. 6%).
Residents who transition and residents who do not are equally likely to have an
uncertain discharge prediction. Lastly, residents who return to the community
are significantly more likely to receive training in skills required to return to the
community (60% vs. 30%).
Regression Results
Table 4.2 presents the results of three logistic regressions predicting
community discharge among the total sample. The full model has the lowest
(best) Akaike information criterion (AIC) and an adjusted R
2
of 0.44. The full
model results are described below.

69

TABLE 4.2: LOGISTIC REGRESSION, PREDICTING COMMUNITY DISCHARGE WITHIN 90 DAYS
(65+ SAMPLE, N = 4635)

B Exp(B) B Exp(B) B Exp(B)
Predisposing
Age 0.00 1.00 0.9168 0.00 1.00 0.766 0.00 1.00 0.621
Female 0.31 1.36 <.0001 *** 0.28 1.33 0.001 *** 0.30 1.35 0.000 ***
Married 0.42 1.52 <.0001 *** 0.23 1.26 0.009 ** 0.25 1.28 0.007 **
Caucasian 0.00 . . 0.00 . . 0.00 . .
Hispanic 0.00 1.00 0.9774 0.01 1.01 0.946 0.01 1.01 0.933
African American -0.22 0.81 0.1072 -0.15 0.86 0.309 -0.09 0.91 0.549
Race - Other 0.05 1.05 0.7295 0.10 1.10 0.524 0.19 1.21 0.237
Average Education 0.00 . . 0.00 . . 0.00 . .
High Education 0.14 1.15 0.0927 0.16 1.17 0.075 0.21 1.22 0.023 *
Low Education -0.25 0.78 0.0048 ** -0.21 0.81 0.030 * -0.16 0.85 0.095
Need
Cognitively Impaired -0.47 0.63 <.0001 *** -0.14 0.87 0.233 -0.23 0.79 0.049 *
Depressed -0.44 0.65 0.0146 * -0.31 0.74 0.110 -0.28 0.75 0.153
Comorbidities -0.07 0.93 0.0104 * -0.04 0.97 0.253 -0.03 0.97 0.320
Socially Engaged -0.12 0.89 0.3396 -0.16 0.85 0.215 -0.18 0.84 0.186
Behavioral Problems -0.14 0.87 0.2425 0.14 1.15 0.283 0.17 1.18 0.217
ADL Score -0.10 0.91 0.0006 *** -0.11 0.90 0.000 *** -0.11 0.90 0.001 **
Incontinence - Bladder -0.02 0.98 0.8074 0.05 1.05 0.612 0.09 1.09 0.369
Incontinence - Bowel -0.70 0.50 <.0001 *** -0.58 0.56 <.0001 *** -0.57 0.57 <.0001 ***
Recent Fracture 0.58 1.78 <.0001 *** 0.42 1.52 0.000 *** 0.39 1.48 0.000 ***
Recent Fall -0.02 0.00 0.8287 -0.06 0.94 0.429 -0.10 0.90 0.224
Admitted from Acute 0.00 . . 0.00 . . 0.00 . .
From Home -0.71 0.49 <.0001 *** -0.22 0.81 0.145 -0.16 0.85 0.297
From Other -1.38 0.25 <.0001 *** -0.85 0.43 <.0001 *** -0.65 0.52 <.0001 ***
Enabling: Generic
Lived Alone 0.08 1.08 0.3402 0.11 1.11 0.248 0.09 1.09 0.352
Legally Responsible for Self 0.56 1.76 <.0001 *** 0.40 1.49 <.0001 *** 0.27 1.31 0.001 **
Medicaid -1.01 0.37 <.0001 *** -0.61 0.55 <.0001 *** -0.48 0.62 0.000 ***
SCAN 0.25 1.29 0.0004 *** 0.34 1.40 <.0001 *** 0.41 1.50 <.0001 ***
Enabling: Transition-Specific
Q1a) Community Living Preference 0.60 1.82 <.0001 *** 0.24 1.28 0.035 *
Q1b) Support Person 1.56 4.76 <.0001 *** 0.91 2.50 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.00 . .
Predicted 30-90 days -0.56 0.57 <.0001 ***
Prediction Uncertain -0.83 0.44 <.0001 ***
Predicted long-stay -1.81 0.16 <.0001 ***
P1ar) Community Living Skills Training 0.35 1.42 <.0001 ***
Intercept 0.41 1.51 0.3356 -1.07 0.34 0.024 0.26 1.29 0.609
R
2
0.21 0.31 0.33
Adjusted R
2
0.28 0.41 0.44
AIC 5272.01 4686.56 4520.45
* p < 0.05, ** p < 0.01, *** p < 0.001
Model 3 (full)
Pr > ChiSq
Model 1 (Base)
Pr > ChiSq
Model 2 (Q1a & Q1b)
Pr > ChiSq

In Model 3, gender, marital status and education are the only significant
predisposing characteristics. Being female increases the likelihood of
community discharge by 35% and being married increases the likelihood by
28% (p < 0.01). Having an above average education level increases the
likelihood of community discharge by 22% (p < 0.05).
The significant predictors among the need variables are cognitive
impairment, ADL limitations, bowel incontinence, recent fracture and admitted-
from location. Cognitive impairment is modestly significant and decreases the
likelihood of community discharge by 21% (p < 0.05). ADL limitations and

70
bowel incontinence decrease the likelihood of community discharge by 10% (p
< 0.01) and 43% respectively. However, residents who’ve had a recent fracture
are 48% more likely to have a community discharge within 90 days. Finally,
compared to individuals admitted from acute hospitals, residents who are
admitted from another supportive setting (e.g. assisted living, board and care,
group home, rehabilitation hospital, psychiatric hospital, or nursing home) are
48% less likely to transition.
All of the generic enabling characteristics are significant except having
lived alone prior to admission. Retaining one’s own legal responsibility
improves the likelihood of transition by approximately 30% (p < 0.01). Payment
source and insurance are also highly significant enabling factors. Medicaid-
funded residents are 38% less likely to be discharged to the community
compared to non-Medicaid residents. Alternatively, SCAN members are 50%
more likely to be discharged to the community compared to individuals with
traditional Medicare fee-for-service coverage.
Transition-Specific Enabling Characteristics
Each of the transition-specific enabling variables is significant in the full
model. Individuals with a preference to return to the community are 28% (p <
0.05) more likely to have a community discharge than residents without a
preference. Even more significant, individuals with a support person who is
positive towards discharge are 2.5 times more likely to have a community
discharge compared to individuals without a support person.

71
The predicted discharge variables are significant predictors of community
discharge. Compared to individuals who are predicted to be discharged within
30 days, individuals in each of the other time frames are significantly less likely
to transition to the community. Specifically, individuals who are predicted to
stay 30-90 days are 43% less likely, individuals whose discharge prediction is
uncertain are 56% less likely and individuals predicted to stay longer than 90
days are 84% less likely to have a community discharge compared to
individuals who are predicted to discharge within 30 days. Receipt of
community living skills training is also highly significant. Individuals who receive
training are 42% more likely to transition to the community within 90 days.
Research Question #2: Do the effects of the transition-specific enabling
variables vary differentially by age?
In order to answer the second research question, the sample was divided
into the following two groups based on age:
Young-old, which includes individuals age 65 to 85;
Oldest-old, which includes those 85 years and older.
Bivariate Analysis
The bivariate analyses were done to compare the characteristics of those
who discharge to the community within 90 days and those who remain in the
nursing home. The comparison results for the young-old sample are shown in
Table 4.3 and the results for the oldest-old are presented in Table 4.4.
Although the mean values change slightly with age, the differences between

72
residents who transition and those who do not are generally similar to those of
the total sample as described above and seen in Table 4.1.

TABLE 4.3: DESCRIPTIVES AND BIVARIATE COMPARISONS (65-85 SAMPLE, N = 2755)

Total No Transition Yes, Transition
N = 2755 N = 1324 N = 1431
Mean SD mean SD mean SD t chi sq p sig
Predisposing
Age 78.22 4.70 78.03 4.71 78.39 4.69 -1.97 0.049 *
Female 0.62 0.58 0.66 16.83 <.0001 ***
Married 0.35 0.33 0.37 5.36 0.021 *
Caucasian 0.72 0.67 0.77 36.78 <.0001 ***
Hispanic 0.12 0.14 0.11 3.46 0.063
African American 0.08 0.11 0.06 27.43 <.0001 ***
Race - Other 0.07 0.09 0.06 6.49 0.011 *
Low Education 0.24 0.28 0.20 20.81 <.0001 ***
Average Education 0.49 0.49 0.49 0.05 0.817
High Education 0.27 0.23 0.31 21.29 <.0001 ***
Need
Cognitively Impaired 0.17 0.27 0.07 188.71 <.0001 ***
Depressed 0.04 0.05 0.04 3.43 0.064 ***
Comorbidities 1.48 1.26 1.67 1.25 1.31 1.24 7.56 <.0001 ***
Socially Engaged 0.09 0.08 0.10 2.02 0.155
Behavioral Problems 0.10 0.13 0.07 27.70 <.0001 ***
ADL Score 3.63 1.47 3.89 1.52 3.40 1.39 8.76 <.0001
Incontinence - Bladder 0.35 0.45 0.26 113.35 <.0001 ***
Incontinence - Bowel 0.38 0.52 0.25 210.71 <.0001
Recent Fracture 0.15 0.08 0.20 78.07 <.0001 ***
Recent Fall 0.39 0.34 0.43 24.83 <.0001 ***
Admitted from Acute 0.82 0.73 0.91 154.18 <.0001 ***
Admitted from Home 0.07 0.10 0.05 30.78 <.0001 ***
Admitted from Other 0.10 0.17 0.04 116.48 <.0001 ***
Enabling: Generic
Lived Alone 0.25 0.18 0.31 60.64 <.0001 ***
Legally Responsible for Self 0.50 0.37 0.62 178.78 <.0001 ***
Medicaid 0.17 0.28 0.08 189.11 <.0001 ***
SCAN 0.47 0.42 0.52 25.66 <.0001 ***
Enabling: Transition-Specific
Q1a) Community Living Preference 0.66 0.42 0.87 630.59 <.0001 ***
Q1b) Support Person 0.65 0.41 0.87 660.72 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.28 0.11 0.44 353.97 <.0001 ***
Predicted 30-90 days 0.10 0.06 0.13 40.39 <.0001 ***
Prediction Uncertain 0.36 0.36 0.37 0.23 0.633 ***
Predicted long-stay 0.26 0.47 0.06 589.33 <.0001 ***
P1ar) Community Living Skills Training 0.45 0.28 0.61 293.49 <.0001 ***
Episode Length 223.01 284.40 428.33 294.57 33.03 18.53 48.74 <.0001 ***
* p < 0.05, ** p < 0.01, *** p < 0.001

73

TABLE 4.4: DESCRIPTIVES AND BIVARIATE COMPARISONS (85+ SAMPLE, N = 1880)

Total No Transition Yes, Transition
N = 1880 N = 993 N = 887
Mean SD mean SD mean SD t chi sq p sig
Predisposing
Age 90.02 3.65 90.11 3.69 89.92 3.61 1.14 0.254
Female 0.73 0.74 0.73 0.14 0.709
Married 0.21 0.20 0.22 1.45 0.228
Caucasian 0.78 0.72 0.85 42.75 <.0001 ***
Hispanic 0.08 0.11 0.05 18.85 <.0001 ***
African American 0.07 0.09 0.05 12.54 0.000 ***
Race - Other 0.07 0.08 0.05 5.87 0.015 *
Low Education 0.26 0.33 0.19 47.21 <.0001 ***
Average Education 0.47 0.45 0.50 6.03 0.014 *
High Education 0.27 0.23 0.31 16.41 <.0001 ***
Need
Cognitively Impaired 0.22 0.30 0.14 69.08 <.0001 ***
Depressed 0.03 0.05 0.02 10.05 0.002 **
Comorbidities 1.33 1.14 1.43 1.13 1.22 1.15 4 <.0001 ***
Socially Engaged 0.07 0.07 0.07 0.01 0.918
Behavioral Problems 0.09 0.12 0.06 17.05 <.0001 ***
ADL Score 3.91 1.37 4.06 1.37 3.74 1.34 5.06 <.0001 ***
Incontinence - Bladder 0.46 0.53 0.38 45.35 <.0001 ***
Incontinence - Bowel 0.48 0.58 0.37 81.52 <.0001 ***
Recent Fracture 0.21 0.15 0.27 47.09 <.0001 ***
Recent Fall 0.47 0.42 0.53 23.19 <.0001 ***
Admitted from Acute 0.80 0.71 0.91 113.56 <.0001 ***
Admitted from Home 0.08 0.11 0.06 17.53 <.0001 ***
Admitted from Other 0.11 0.18 0.04 93.97 <.0001 ***
Enabling: Generic
Lived Alone 0.25 0.20 0.32 35.90 <.0001 ***
Legally Responsible for Self 0.39 0.26 0.53 146.77 <.0001 ***
Medicaid 0.16 0.23 0.07 91.90 <.0001 ***
SCAN 0.50 0.47 0.53 5.10 0.024 *
Enabling: Transition-Specific
Q1a) Community Living Preference 0.60 0.40 0.83 364.23 <.0001 ***
Q1b) Support Person 0.62 0.39 0.88 476.94 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.26 0.11 0.43 252.89 <.0001 ***
Predicted 30-90 days 0.10 0.08 0.13 12.76 0.000 ***
Prediction Uncertain 0.37 0.35 0.39 2.66 0.103
Predicted long-stay 0.27 0.46 0.06 395.00 <.0001 ***
P1ar) Community Living Skills Training 0.45 0.31 0.60 151.52 <.0001 ***
Episode Length 226.75 275.70 400.86 281.75 31.84 17.66 41.18 <.0001 ***
* p < 0.05, ** p < 0.01, *** p < 0.001

Regression Results
The results of the three logistic regression equations for the young-old
group are presented in Table 4.5 and the results for the oldest-old are shown in
Table 4.6.

74

TABLE 4.5: LOGISTIC REGRESSION, PREDICTING COMMUNITY DISCHARGE WITHIN 90 DAYS
(65-85 SAMPLE, N = 2755)

B Exp(B) B Exp(B) B Exp(B)
Predisposing
Age 0.02 1.02 0.0325 0.02 1.02 0.063 0.02 1.02 0.042 *
Female 0.44 1.55 <.0001 *** 0.37 1.45 0.000 *** 0.38 1.47 0.000 ***
Married 0.55 1.74 <.0001 *** 0.37 1.44 0.001 ** 0.38 1.46 0.001 **
Caucasian 0.00 . . 0.00 . . 0.00 . .
Hispanic 0.16 1.18 0.2882 0.15 1.17 0.350 0.14 1.15 0.401
African American -0.23 0.79 0.1748 -0.13 0.88 0.506 -0.07 0.93 0.711
Race - Other 0.04 1.04 0.8351 0.02 1.02 0.915 0.08 1.09 0.678
Average Education 0.00 . . 0.00 . .
High Education 0.22 1.25 0.0394 * 0.25 1.28 0.037 * 0.30 1.35 0.01 *
Low Education -0.09 0.92 0.4786 -0.11 0.89 0.381 -0.04 0.96 0.75
Need
Cognitively Impaired -0.65 0.52 <.0001 *** -0.32 0.73 0.053 -0.44 0.65 0.010 **
Depressed -0.21 0.81 0.3563 -0.08 0.92 0.729 -0.03 0.97 0.894
Comorbidities -0.10 0.90 0.0047 ** -0.08 0.93 0.051 -0.07 0.94 0.106
Socially Engaged -0.10 0.91 0.5236 -0.12 0.89 0.472 -0.13 0.87 0.438
Behavioral Problems -0.04 0.96 0.7848 0.29 1.34 0.097 0.32 1.37 0.076
ADL Score -0.12 0.89 0.0011 ** -0.15 0.86 0.000 *** -0.15 0.86 0.001 ***
Incontinence - Bladder -0.04 0.96 0.7481 0.03 1.03 0.812 0.09 1.09 0.514
Incontinence - Bowel -0.79 0.45 <.0001 *** -0.60 0.55 <.0001 *** -0.60 0.55 <.0001 ***
Recent Fracture 0.70 2.02 <.0001 *** 0.53 1.69 0.001 *** 0.45 1.58 0.004 **
Recent Fall -0.02 0.98 0.8034 -0.06 0.94 0.563 -0.09 0.91 0.406
Admitted from Acute 0.00 . . 0.00 . . 0.00 . .
From Home -0.85 0.43 <.0001 *** -0.31 0.73 0.117 -0.25 0.78 0.229
From Other -1.28 0.28 <.0001 *** -0.77 0.46 <.0001 *** -0.57 0.57 0.003 **
Enabling: Generic
Lived Alone 0.19 1.21 0.093 0.22 1.24 0.074 0.17 1.19 0.172
Legally Responsible for Self 0.47 1.61 <.0001 *** 0.28 1.33 0.008 ** 0.16 1.17 0.158
Medicaid -1.11 0.33 <.0001 *** -0.71 0.49 <.0001 *** -0.56 0.57 0.000 ***
SCAN 0.32 1.38 0.0007 *** 0.37 1.44 0.000 *** 0.45 1.57 <.0001 ***
Enabling: Transition-Specific
Q1a) Community Living Preference 0.81 2.25 <.0001 *** 0.38 1.46 0.015 *
Q1b) Support Person 1.41 4.09 <.0001 *** 0.80 2.24 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.00 . .
Predicted 30-90 days -0.39 0.68 0.025 *
Prediction Uncertain -0.84 0.43 <.0001 ***
Predicted long-stay -1.78 0.17 <.0001 ***
P1ar) Community Living Skills Training 0.45 1.57 <.0001 ***
Intercept -1.20 0.30 0.124 -2.67 0.07 0.002 ** -1.68 0.19 0.057
R
2
0.24 0.33 0.35
Adjusted R
2
0.31 0.43 0.47
AIC 3069.60 2734.83 2632.13
* p < 0.05, ** p < 0.01, *** p < 0.001
Model 3 (Full)
Pr > ChiSq
Model 1 (Base)
Pr > ChiSq
Model 2 (Q1a & Q1b)
Pr > ChiSq

75

TABLE 4.6: LOGISTIC REGRESSION, PREDICTING COMMUNITY DISCHARGE WITHIN 90 DAYS
(85+ SAMPLE, N = 1880)

B Exp(B) B Exp(B) B Exp(B)
Predisposing
Age 0.00 1.00 0.799 -0.01 0.99 0.664 -0.01 0.99 0.519
Female 0.09 1.09 0.500 0.16 1.18 0.237 0.17 1.18 0.239
Married 0.21 1.24 0.131 0.07 1.07 0.650 0.06 1.06 0.714
Caucasian 0.00 . . 0.00 . . 0.00 . .
Hispanic -0.26 0.77 0.232 -0.16 0.85 0.489 -0.13 0.88 0.592
African American -0.17 0.84 0.426 -0.18 0.83 0.441 -0.12 0.89 0.620
Race - Other 0.06 1.06 0.789 0.25 1.28 0.336 0.34 1.40 0.187
Average Education 0.00 . . 0.00 . . 0.00 . .
High Education 0.02 1.02 0.897 0.03 1.03 0.814 0.07 1.07 0.636
Low Education -0.46 0.63 0.001 *** -0.34 0.71 0.023 * -0.33 0.72 0.032 *
Need
Cognitively Impaired -0.32 0.73 0.033 * 0.02 1.02 0.923 -0.07 0.93 0.684
Depressed -0.78 0.46 0.013 * -0.73 0.48 0.030 * -0.70 0.50 0.044 *
Comorbidities -0.03 0.97 0.562 0.03 1.03 0.537 0.02 1.02 0.669
Socially Engaged -0.16 0.85 0.431 -0.26 0.77 0.244 -0.27 0.76 0.233
Behavioral Problems -0.28 0.75 0.141 -0.06 0.94 0.766 -0.07 0.94 0.760
ADL Score -0.06 0.94 0.194 -0.04 0.96 0.408 -0.04 0.96 0.485
Incontinence - Bladder -0.01 0.99 0.936 0.08 1.08 0.587 0.09 1.09 0.530
Incontinence - Bowel -0.60 0.55 <.0001 *** -0.58 0.56 0.000 *** -0.56 0.57 0.000 ***
Recent Fracture 0.48 1.61 0.001 ** 0.33 1.40 0.036 * 0.34 1.40 0.037 *
Recent Fall -0.02 0.98 0.861 -0.08 0.92 0.517 -0.12 0.88 0.350
Admitted from Acute 0.00 . . 0.00 . . 0.00 . .
From Home -0.54 0.58 0.009 ** -0.14 0.87 0.535 -0.13 0.88 0.592
From Other -1.51 0.22 <.0001 *** -0.92 0.40 <.0001 *** -0.73 0.48 0.002 **
Enabling: Generic
Lived Alone -0.07 0.93 0.590 -0.05 0.95 0.721 -0.03 0.97 0.829
Legally Responsible for Self 0.70 2.01 <.0001 *** 0.59 1.80 <.0001 *** 0.44 1.56 0.001 ***
Medicaid -0.85 0.43 <.0001 *** -0.43 0.65 0.030 * -0.35 0.71 0.096
SCAN 0.13 1.14 0.236 0.27 1.32 0.021 * 0.31 1.36 0.011 *
Enabling: Transition-Specific
Q1a) Community Living Preference 0.32 1.38 0.058 0.08 1.09 0.643
Q1b) Support Person 1.81 6.11 <.0001 *** 1.10 3.00 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.00 . .
Predicted 30-90 days -0.77 0.46 <.0001 ***
Prediction Uncertain -0.82 0.44 <.0001 ***
Predicted long-stay -1.84 0.16 <.0001 ***
P1ar) Community Living Skills Training 0.20 1.22 0.109
Intercept 0.8349 2.305 0.533 -0.79 0.45 0.587 0.87 2.38 0.564
R
2
0.19 0.30 0.32
Adjusted R
2
0.26 0.39 0.43
AIC 2203.88 1956.55 1900.80
* p < 0.05, ** p < 0.01, *** p < 0.001
Model 3 (Full)
Pr > ChiSq
Model 1 (Base)
Pr > ChiSq
Model 2 (Q1a & Q1b)
Pr > ChiSq

In tables 4.5 and 4.6, compared to the other models, the full models each
have the lowest (best) Akaike Information criterion (AIC) and the highest
adjusted R
2
. The full model results for the oldest-old (Table 4.6) are described
below.
Among the oldest-old the only significant predisposing variable is low
education. Compared to individuals with an average education level (high
school graduate), individuals with below average education are 28% (p < 0.05)
less likely to be discharged to the community.

76
Among the need characteristics, depression (OR = 0.50, p < 0.05) and
bowel incontinence (OR = 0.57) are barriers to community discharge for the
oldest-old. However, having had a recent fracture increases the likelihood of
community discharge by 40% (p < 0.05). Compared to residents who are
admitted from the hospital, individuals who are admitted from another
supportive setting are 52% less likely to transition to the community (p < 0.01).
Focusing on the generic enabling characteristics of the oldest-old,
individuals who are legally responsible for themselves are 56% more likely to
transition to the community than residents who are no longer legally responsible
for themselves. SCAN membership increases the likelihood of transition by
36% (p < 0.05).
Transition-Specific Enabling Characteristics
To answer the second research question about the differential effect of the
transition-specific variables with age, the regression results of these variables
from the full models for each age group are presented in Table 4.7. Preference
is modestly significant for the young-old (OR = 1.46, p < 0.05), but not
significant for the oldest-old. Surprisingly, controlling for other factors, oldest-old
residents with a preference to return to the community are no more likely to
transition than their age peers without a preference.

77

TABLE 4.7: LOGISTIC REGRESSION, PREDICTING COMMUNITY DISCHARGE WITHIN 90 DAYS,
CONTROLLING FOR PREDISPOSING, NEED AND GENERIC.
Enabling Variables
Variables B Exp(B) B Exp(B) B Exp(B)
Q1a) Community Living Preference 0.24 1.28 0.035 * 0.38 1.46 0.015 * 0.08 1.09 0.643
Q1b) Support Person 0.91 2.50 <.0001 *** 0.80 2.24 <.0001 *** 1.10 3.00 <.0001 ***
Q1c) Predicted Discharge < 30 days 0.00 . . 0.00 . . 0.00 . .
Predicted 30-90 days -0.56 0.57 <.0001 *** -0.39 0.68 0.025 * -0.77 0.46 <.0001 ***
Prediction Uncertain -0.83 0.44 <.0001 *** -0.84 0.43 <.0001 *** -0.82 0.44 <.0001 ***
Predicted long-stay -1.81 0.16 <.0001 *** -1.78 0.17 <.0001 *** -1.84 0.16 <.0001 ***
P1ar) Community Living Training 0.35 1.42 <.0001 *** 0.45 1.57 <.0001 *** 0.20 1.22 0.109
SCAN 0.41 1.50 <.0001 *** 0.45 1.57 <.0001 *** 0.31 1.36 0.011 *
* p < 0.05, ** p < 0.01, *** p < 0.001
Total Sample (n = 4635) 65-85 Sample (n = 2755) 85+ Sample (n = 1880)
Pr > ChiSq Pr > ChiSq Pr > ChiSq

Although preference is not significant for the oldest-old, the presence of a
support person who is positive towards discharge is highly significant for both
age groups, especially the oldest-old. Young-old residents with a preference to
return to the community are 2.24 times more likely to return to the community
than their age peers without a preference. Among the oldest-old, residents who
have a support person who is positive towards discharge are three times more
likely to transition.
The predicted discharge variables are significant predictors of community
discharge for both age groups. Among the young-old, compared to individuals
who are predicted to be discharged within 30 days, individuals in each of the
other time frames are significantly less likely to transition to the community.
Specifically, individuals who are predicted to stay 30 to 90 days are 32% (p <
0.05) less likely, individuals whose discharge prediction is uncertain are 57%
less likely and individuals predicted to stay longer than 90 days are 83% less
likely to have a community discharge.

78
Among the oldest-old, each of the predicted discharge variables are highly
significant. The odds ratios are similar to the young-old for the predicted long-
stay and discharge uncertain groups. However, the odds ratio and significance
level for individuals predicted to stay 30 to 90 days are quite different between
the age groups. The odds ratio for young-old individuals predicted to be
discharged in the 30 to 90 day time frame is 0.68 (p < 0.05), compared to 0.46
for the oldest-old. Compared to their age peers who are predicted to be
discharged within 30 days, young-old individuals who are predicted to discharge
within 30 to 90 days are 1.47 times more likely to become long-stay residents.
Oldest-old individuals with the same discharge prediction (30 to 90 days) are
2.17 times more likely to become long-stay residents compared to their age
peers with predicted discharge within 30 days. .
Among the young-old population, receiving training in community living
skills increases the likelihood of transition by 57% (p < 0.001). However receipt
of training is not significant for the oldest-old.
Research Question #3: Does SCAN membership matter?
The answer to the final research question regarding the effect of SCAN
membership on community discharge outcomes can be found in Table 4.7.
Among the total sample, SCAN increases the likelihood of community discharge
by 50% in the full model. When broken down by age, the positive effect of
SCAN is stronger for the young-old population (OR = 1.57), but it remains
positive and significant for the oldest-old as well (OR = 1.36, p < 0.05).

79
D. Discussion
The results above confirm that transition-specific enabling characteristics
and SCAN membership are important predictors of community discharge and
that their effects vary with age.
Transition-Specific Enabling Characteristics
Preference
One of the key findings of this study is that community living preference
matters for individuals 65 to 85 years old, but that it is not a significant predictor
of community discharge for the oldest-old. Individuals in the young-old group
who have a preference to return to the community are 46% more likely to
transition than their age peers without a preference. However, among the
oldest-old, preference does not increase the likelihood of community discharge.
This disconcerting finding underscores the unique challenges facing the oldest-
old and suggests that this group could benefit from transition efforts tailored
specifically to the particular circumstances of this later life stage.
Support Person
Having a support person who is positive toward discharge is the most
significant predictor of community re-entry for all residents in the study,
especially those over the age of 85. Among the sample of 65 to 85 year olds,
individuals who have a support person who is positive towards discharge are
more than 224% more likely to transition than individuals without a support
person. For residents over age 85, the positive impact is even greater. Among

80
the oldest-old, having a support person increases the likelihood of transition by
over 300%. This finding is positive for individuals with a support person, but
highlights a significant disadvantage for individuals without such a person.
Whether or not an individual has a support person will affect the type and
amount of transition assistance and community support the individual needs. A
support person is likely to help not only with the one-time transition but he/she
is also likely to be called upon to provide informal support to the individual in the
community. As individuals age, the size and nature of their social networks
change due to changes in health and as a consequence of outliving spouses,
relatives and friends (Poon et al., 2000). As a result, the oldest-old often have a
smaller number of potential caregivers (Chapin et al., 1998). Considering the
significant impact of a support person, specific transition strategies are needed
in order to further assess residents’ available support systems and to develop
services to compensate for the lack of informal support, especially for the
oldest-old.
Some Money Follows the Person grantees have created ‘Transition
Counselors’ whose job it is to locate affordable housing, coordinate community
services, and assist with transition needs (e.g., furnishing the home, transferring
prescriptions, stocking the pantry). Transition counselors could be of particular
assistance to oldest-old individuals who do not have a support person to help
with discharge planning and to potentially provide care after discharge.

81
Discharge Prediction
Discharge prediction is a significant predictor of community discharge for
the young-old and the oldest-old. Compared to residents who are predicted to
stay less than 30 days, individuals predicted to be discharged during each of
the other three time frames are significantly less likely to transition to the
community within 90 days. The effect size for uncertain discharge prediction
and predicted long-stay are similar for both age groups - approximately 56%
and 83% decreased likelihood respectively. However, it is notable that among
the young-old, individuals who are predicted to be discharged within 30 to 90
days are 32% less likely to have a community discharge and among the oldest-
old, individuals with the same predicted discharge window are 54% less likely to
have a community discharge. This finding suggests that the oldest-old who are
predicted to discharge within 30 to 90 days are at a higher risk of becoming
long-stay than younger residents with the same discharge prediction.
Although discharge prediction is a highly significant predictor of
community discharge, it is a complex variable to interpret because it is a
summary variable based on the assessor’s judgment. The discharge prediction
variables are correlated with length of stay and discharge outcomes, but the
causal direction is unclear.
Community Living Skills Training
In the full models, receipt of community living training is significant for the
young-old but not for the oldest-old. Even though residents within each age

82
group were equally likely to receive community living training (45%), the training
increased the likelihood almost 60% for the young-old but had no effect on
discharge outcomes of the oldest-old.
Relatively little is known about how residents are selected for such
training, what the training entails or how consistent the training is across
different facilities. Regardless, the differential effect on the young-old versus
the oldest-old is interesting and warrants further examination. Additionally,
although the training is associated with community discharge, the causal
direction is again unclear. If the only people who receive training are those who
have been identified by the nursing home staff as likely to transition, then the
association would be the result of selection bias. In order to fully understand
the effect on community living training on discharge outcomes more research is
needed, ideally in the form of a more controlled experiment.
Generic Enabling Characteristics
Among the generic enabling characteristics, maintaining legal
responsibility, receipt of Medicaid funding and SCAN membership are
significant factors.
Retaining legal responsibility for oneself is significant for the total
sample, but when broken down by age it is only significant for the 85 and older
group. Among the oldest-old, those who are legally responsible for themselves
are 56% more likely to transition than their peers who have not retained legal
responsibility. Previous research has also shown that self-consenting residents

83
are more likely to express a stable transition preference (Nishita et al., 2008). If
cognitively intact, residents should be encouraged to retain legal responsibility
for decision-making. Establishing one’s own goals and taking a proactive role
in health care decisions, allows an individual to actively work toward his/her
desired outcomes, such as transition.
Payment source is an important enabling factor for individuals age 65 to
85, but not for the oldest-old. Among the young-old population, Medicaid is
highly significant and decreases the likelihood of transition by 43%. The finding
that Medicaid is a significant barrier to nursing home transition is consistent with
the literature. There are a number of explanations for why and how Medicaid
inhibits nursing home transition including the requirement to liquidate assets in
exchange for Medicaid coverage, uncertainty about eligibility for home and
community-based services, waiting lists for community-based care and the
potential loss of spousal impoverishment protection.
Meanwhile, membership in the SCAN Health S/HMO Plan significantly
increases the likelihood of community discharge for both age groups. The
effect of SCAN is larger and more significant in the 65 to 85 population
compared to the oldest-old. SCAN membership increases the likelihood of
community discharge by 57% among the young-old and by 36% (p <0.05)
among the oldest-old. These findings may be attributable to the care
coordination and expanded community services that were available to SCAN
members when it was a S/HMO. The consistent positive SCAN findings in this

84
study, suggest that additional research on these models could prove
worthwhile.
Other Factors
In addition to a support person, other factors are particularly important for
the oldest-old including education level and depression. Education about
available services can be particularly beneficial to individuals with low levels of
education. Nishita et al (2008) found that community living preference and
beliefs about feasibility can be manipulated through education about available
home and community based services.
Finally, depression is a significant barrier to community discharge for
individuals over age 85, but not for those under age 85. Depression decreases
the likelihood of transition by 50% among the oldest-old. It is important that
residents, family members and staff are familiar with the symptoms of
depression, are aware that depression is not a part of normal aging and know
that effective treatments are available.
Changes Between Models
It is interesting to note the changes that occur between Models 2 and 3
in Tables 4.5 and 4.6. The addition of discharge prediction and community
living skills training has a dampening effect on two important enabling variables:
legal responsibility for self and presence of a support person positive towards
discharge. Among the 65 to 85 year old population (Table 4.5), maintaining
one’s legal responsibility increases the likelihood of transition by 33% (p < 0.01)

85
and the presence of a support person increases the likelihood by 409%.
However, after discharge prediction or training variables are added (Model 3),
the effect of legal responsibility is no longer significant and the effect of having a
support person decreases to 224%. This same pattern can be seen between
Models 2 and 3 among the oldest-old (Table 4.6).
The dampening effect of discharge prediction and training can be
interpreted in two ways. It may reflect specific tracks that individuals get put on
based on the staff’s beliefs about the residents potential for transition. Based
on staff perception, residents may receive different levels of care such that
discharge prediction becomes a self-fulfilling prophecy. Alternatively, the
dampening effect of discharge prediction and training may reflect the tempering
of unrealistic resident transition expectations.
The California Nursing Facility Transition Screen (CNFTS) is a potentially
useful tool for sorting out realistic versus unrealistic transition expectations for
residents and their families (California Pathways, 2005). The CNFTS includes
27 open and closed-ended questions about the resident’s preference and ability
to move to the community. The screen also explores potential living
arrangements and services in the community (Nishita et al., 2008).
Additional practice and policy implications of these findings are described
in Chapter 6.

86
Limitations
The preference data and all of the independent variables are taken from
each resident’s admission assessment. This is a limitation because, after
admission, any number of health, personal or financial changes could occur in
an individual’s life that could affect discharge preferences and potential.
Additionally, Section Q is only on the admission assessment and subsequent
full annual assessment forms, not on quarterly assessments. Therefore,
information about a resident’s preference may be up to one year old.
Understanding how discharge preferences and predicted discharge potential
can change over time would be a valuable topic for future research. Adding the
Section Q questions to the quarterly assessment would also greatly improve
targeting and transition efforts.
The receipt of community living training and discharge prediction variables
have not been extensively used in nursing home transition studies. Although
they were found to be significant predictors in this study, especially discharge
prediction, additional research is necessary. Discharge prediction is a complex
variable to interpret because it is a summary variable based on the assessor’s
judgment. The discharge prediction variables are correlated with length of stay
and discharge outcomes, but the causal direction is unclear. Based on the
discharge prediction and the staff’s beliefs about the individual’s potential for
discharge, it is possible that residents may receive different levels of care such
that the prediction becomes a self-fulfilling prophecy. These limitations should

87
be kept in mind when interpreting the full model regression results. However, in
spite of the limitations, these variables are promising tools for transition practice
and policy.
Although these findings are based on a diverse sample from four large
Southern California counties, researchers and practitioners must use caution
generalizing from these results. Additional research is needed to identify ways
to systematically use the rich MDS data to increase the effectiveness and
efficiency of nursing home transition programs.

88
CHAPTER V: UNDERSTANDING BARRIERS TO TRANSITION
AMONG NURSING HOME RESIDENTS WHO PREFER TO
RETURN TO THE COMMUNITY

A. Introduction

Despite the richness and relevance of the Nursing Home Minimum Data
Set (MDS), recent studies of Nursing Home Transition projects have found that
very few of them use it to identify potential transition candidates (Siebenaler,
O'Keeffe, Brown & O’Keefe, 2005; Reinhard & Hendrickson, 2006). Moreover,
in cases where MDS 2.0 data are used in transition efforts, the preference
variable (Q1a) is often the only variable used. Given the high percentage of
people who prefer to return to the community and the results found in Chapter
4, simply using Q1a is unlikely to be a sufficient targeting strategy. For
example, each year 3,000 to 4,000 nursing home residents in Kansas answer
yes to Q1a. As of 2006, Kansas was actively pursuing ways to further filter this
list (Reinhard & Hendrickson, 2006).
For transition purposes, researchers and practitioners need to know not
just who has a preference to return to the community, but who has a preference
to leave and needs assistance in order to do so. Using scarce transition
resources to assist residents who would have transitioned on their own is
inefficient. States can use question Q1a as a first step in identifying potential
transition candidates, but additional information contained in the MDS is
necessary in order to sufficiently refine the list. Therefore, the purpose of this
study is to investigate the social and structural factors that affect transition

89
outcomes among the subset of residents with a preference to return to the
community.
Analysis
The conceptual model, sample and measures used in this analysis are
described in detail in Chapter 3.
For the following analysis, steps are taken to examine which factors
facilitate and which factors impede transition. First, bivariate analyses are used
to compare the characteristics of residents with a preference to return to the
community versus those without a community living preference. Second,
looking only at the subsample of residents with a preference to return to the
community, residents with a community discharge within 90 days are compared
to residents who had episode lengths greater than 90 days.
Next, among residents with a preference to return to the community,
logistic regression is used to identify predictors of community discharge within
90 days. Based on the Chapter 4 finding that preference is a significant
predictor of transition for the young-old but not the oldest-old, the logistic
regression models were re-run on just the subset of individuals ages 65 to 85
who have a preference to return to the community. The following three logistic
regression models were run for each age group:
Model 1 (base model) = predisposing, need and generic enabling
characteristics.

90
Model 2 = Model 1 + presence of a support person who is positive
towards discharge.
Model 3 (full model) = Model 2 + discharge prediction + receipt of
community living skills training.
In each model, community discharge within 90 days is the dichotomous
dependent variable; the predisposing, need and enabling measures described
in Chapter 3 are included as the independent variables. In the results below, all
differences are highly significant (p < 0.001) unless otherwise noted.
B. Results
Total Preference Sample
Bivariate Analyses
Among the total sample (n = 4635), 63% of residents expressed or
indicated a preference to return to the community (n = 2935). Results of the
bivariate analysis indicate that individuals with a preference to return to the
community are significantly different than individuals without a community living
preference (Table 5.1). With respect to the predisposing characteristics,
residents with a preference to return to the community are, on average,
significantly younger, more likely to be Caucasian and more likely to have an
above average education level.

91

TABLE 5.1: DESCRIPTIVES AND BIVARIATE COMPARISONS BY Q1A (65+ SAMPLE, N = 4635)

Total (N = 4635) Q1a = 0 (n = 1700) Q1a = 1 (n = 2935)
Mean (SD) Mean (SD) Mean (SD) t chi sq p
Predisposing
Age 83.01 7.22 83.63 7.15 82.64 7.24 4.50 <.0001 ***
Female 0.67 0.67 0.67 0.04 0.846
Married 0.29 0.28 0.30 2.78 0.095
Caucasian 0.74 0.66 0.79 95.12 <.0001 ***
Hispanic 0.11 0.13 0.09 17.17 <.0001 ***
African American 0.08 0.12 0.06 49.81 <.0001 ***
Race - Other 0.07 0.09 0.06 17.66 <.0001 ***
Average Education 0.48 0.47 0.49 0.83 0.362
High Education 0.27 0.22 0.30 32.40 <.0001 ***
Low Education 0.25 0.30 0.21 47.92 <.0001 ***
Need
Cognitively Impaired 0.19 0.39 0.07 701.02 <.0001 ***
Depressed 0.04 0.04 0.04 0.28 0.594
Comorbidities 1.42 1.21 1.67 1.18 1.28 1.21 10.85 <.0001 ***
Socially Engaged 0.08 0.07 0.09 8.56 0.003 **
Behavioral Problems 0.10 0.16 0.06 126.61 <.0001 ***
ADL Score 3.75 1.44 4.12 1.49 3.53 1.36 13.28 <.0001 ***
Incontinence - Bladder 0.40 0.55 0.31 252.07 <.0001 ***
Incontinence - Bowel 0.42 0.62 0.31 426.23 <.0001 ***
Recent Fracture 0.17 0.10 0.21 99.23 <.0001 ***
Recent Fall 0.42 0.34 0.47 66.85 <.0001 ***
Admitted from Acute 0.81 0.66 0.90 403.27 <.0001 ***
From Home 0.08 0.13 0.05 94.65 <.0001 ***
From Other 0.11 0.21 0.05 280.76 <.0001 ***
Enabling: Generic
Lived Alone 0.25 0.14 0.32 186.87 <.0001 ***
Legally Responsible for Self 0.45 0.22 0.59 573.77 <.0001 ***
Medicaid 0.17 0.30 0.09 324.00 <.0001 ***
SCAN 0.48 0.47 0.49 1.92 0.165
Enabling: Transition-Specific
Q1b) Support Person 0.64 0.19 0.90 2360.89 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.27 0.05 0.40 641.38 <.0001 ***
Predicted 30-90 days 0.10 0.02 0.15 180.17 <.0001 ***
Prediction Uncertain 0.37 0.28 0.41 77.41 <.0001 ***
Predicted long-stay 0.26 0.64 0.04 1976.09 <.0001 ***
P1ar) Community Living Skills Training 0.45 0.20 0.60 685.64 <.0001 ***
Episode Length 224.53 280.88 358.24 304.40 147.08 233.64 24.70 <.0001 ***
90+ days (y/n) 0.50 0.80 0.32 997.75 <.0001 ***
* p < 0.05, ** p < 0.01, *** p < 0.001

Among the need variables, residents with a community living preference
have significantly less cognitive impairment, less comorbidity, higher levels of
social engagement (p < 0.01), fewer behavioral symptoms, fewer ADL
limitations and less incontinence. Those with a preference are also more likely
to have had a recent fall or fracture, more likely to have been admitted from an
acute-care hospital and less likely to have been admitted from home or another
supportive setting.
With respect to the generic enabling characteristics, residents with a
preference to return to the community are significantly more likely to have lived

92
alone prior to admission and to be legally responsible for themselves. They are
also significantly less likely to be on Medicaid at the time of admission.
The most striking difference between the two groups is the presence of a
support person who is positive toward discharge; 90% of residents with a
preference to return to the community have a support person that is positive
toward discharge compared to just 19% of residents who do not express a
community living preference. Residents with a preference are significantly
more likely to be predicted to be discharged within 90 days or to have an
uncertain discharge prediction and significantly less likely to be predicted to be
a long-stay resident. Finally, individuals with a community living preference are
more likely to receive community living skills training.
Thirty-two percent of individuals with a preference to return to the
community end up having an episode longer than 90 days. Among those who
expressed/indicated a preference to return to the community, individuals who
have a community discharge within 90 days (n = 1986) are compared to
individuals who had an episode of more than 90 days (n = 949) in Table 5.2.

93

TABLE 5.2: DESCRIPTIVES AND BIVARIATE COMPARISONS BY COMMUNITY DISCHARGE
(Q1A = YES, 65+ SAMPLE, N = 2935)

Mean (SD) Mean (SD) Mean (SD) t chi sq p
Predisposing
Age 82.64 7.24 82.81 7.57 82.57 7.08 0.82 0.413
Female 0.67 0.63 0.69 12.24 0.001 ***
Married 0.30 0.28 0.31 3.87 0.049 *
Caucasian 0.79 0.75 0.81 13.31 0.000 ***
Hispanic 0.09 0.10 0.09 1.15 0.283
African American 0.06 0.08 0.05 14.99 0.000 ***
Race - Other 0.06 0.07 0.06 1.32 0.251
Average Education 0.49 0.48 0.49 0.69 0.407
High Education 0.30 0.26 0.32 8.65 0.003 **
Low Education 0.21 0.26 0.19 18.49 <.0001 ***
Need
Cognitively Impaired 0.07 0.11 0.05 33.59 <.0001 ***
Depressed 0.04 0.06 0.03 13.12 0.000 ***
Comorbidities 1.28 1.21 1.41 1.26 1.21 1.19 4.04 <.0001 ***
Socially Engaged 0.09 0.09 0.10 0.59 0.443
Behavioral Problems 0.06 0.07 0.05 3.89 0.049 *
ADL Score 3.53 1.36 3.72 1.34 3.44 1.36 5.35 <.0001 ***
Incontinence - Bladder 0.31 0.39 0.27 44.46 <.0001 ***
Incontinence - Bowel 0.31 0.43 0.25 94.58 <.0001 ***
Recent Fracture 0.21 0.15 0.24 31.25 <.0001 ***
Recent Fall 0.47 0.45 0.48 2.49 0.115
Admitted from Acute 0.90 0.85 0.93 44.13 <.0001 ***
From Home 0.05 0.07 0.04 13.65 0.000 ***
From Other 0.05 0.08 0.04 29.56 <.0001 ***
Enabling: Generic
Lived Alone 0.32 0.28 0.33 8.04 0.005 **
Legally Responsible for Self 0.59 0.49 0.64 60.85 <.0001 ***
Medicaid 0.09 0.15 0.06 61.11 <.0001 ***
SCAN 0.49 0.45 0.51 9.95 0.002 **
Enabling: Transition-Specific
Q1b) Support Person 0.90 0.80 0.95 153.97 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.40 0.24 0.47 146.17 <.0001 ***
Predicted 30-90 days 0.15 0.15 0.14 0.35 0.556
Prediction Uncertain 0.41 0.52 0.36 60.17 <.0001 ***
Predicted long-stay 0.04 0.09 0.02 86.44 <.0001 ***
P1ar) Community Living Skills Training 0.60 0.50 0.64 56.81 <.0001 ***
Episode Length 147.08 233.64 388.36 286.64 31.78 17.67 38.29 <.0001 ***
* p < 0.05, ** p < 0.01, *** p < 0.001
Total Q1a Sample
(N = 2935) (n = 949)
Yes, Short Stay /
Community Discharge
(n = 1986)
No, Short Stay /
Community Discharge

Among residents with a preference to return to the community, residents
who transition to the community within 90 days have many significantly different
characteristics than residents with episodes longer than 90 days. On average,
the transition group is more likely to be female, married (p < 0.05), Caucasian,
highly educated (p < 0.01), to have had a recent fracture and to have been
admitted from an acute hospital. They are also more likely to have lived alone
prior to admission, to be legally responsible for self (p < 0.01) and have the
SCAN health plan (p < 0.01).

94
By comparison, long-stay residents are significantly more likely to be
African American and to have below average education. The long-stay group
has significantly more severe cognitive impairment, depressive symptoms,
comorbidities, behavioral symptoms (p < 0.05), ADL limitations, and
incontinence. They are also more likely to have been admitted from home with
no home health services or from a supportive setting and to be on Medicaid.
Among the transition-specific enabling characteristics, the transition
group is significantly more likely to have a support person who is positive
toward discharge. They are more likely to be predicted to discharge within 30
days and significantly less likely to have an uncertain discharge prediction or to
be predicted to stay more than 90 days. Individuals in both groups are equally
likely to be predicted to discharge within 30 to 90 days. Finally, residents
discharged to the community within 90 days are more likely to receive
community living skills training.
Regression Results
The results of the three logistic regressions are displayed in Table 5.3.
The full model has the lowest (best) Akaike information criterion (AIC) and an
adjusted R
2
of 0.22.

95

TABLE 5.3: LOGISTIC REGRESSION, PREDICTING COMMUNITY DISCHARGE
(Q1A = YES, 65+ SAMPLE, N = 2935)

B Exp(B) B Exp(B) B Exp(B)
Predisposing
Age 0.00 1.00 0.859 0.00 1.00 0.490 -0.01 1.00 0.426
Female 0.40 1.50 <.0001 *** 0.34 1.41 0.001 *** 0.35 1.42 0.000 ***
Married 0.36 1.43 0.001 *** 0.28 1.33 0.008 ** 0.29 1.33 0.009 **
Caucasian 0.00 . . 0.00 . . 0.00 . .
Hispanic 0.11 1.12 0.487 0.06 1.06 0.716 0.05 1.06 0.742
African American -0.28 0.76 0.123 -0.29 0.75 0.109 -0.22 0.81 0.248
Race - Other 0.08 1.08 0.674 0.03 1.03 0.863 0.13 1.14 0.502
Average Education 0.00 . . 0.00 . . 0.00 . .
High Education 0.14 1.15 0.180 0.13 1.14 0.194 0.19 1.21 0.071
Low Education -0.22 0.80 0.050 -0.20 0.82 0.086 -0.15 0.86 0.215
Need
Cognitively Impaired -0.21 0.81 0.196 -0.21 0.82 0.212 -0.31 0.73 0.061
Depressed -0.60 0.55 0.005 ** -0.39 0.68 0.081 -0.35 0.70 0.122
Comorbidities -0.04 0.96 0.213 -0.05 0.95 0.146 -0.05 0.95 0.161
Socially Engaged -0.03 0.97 0.842 -0.07 0.93 0.648 -0.11 0.90 0.494
Behavioral Problems 0.03 1.03 0.886 0.18 1.19 0.339 0.22 1.25 0.233
ADL Score -0.11 0.90 0.004 ** -0.12 0.88 0.001 *** -0.11 0.90 0.004 **
Incontinence - Bladder 0.00 1.00 0.977 0.06 1.07 0.582 0.10 1.10 0.391
Incontinence - Bowel -0.68 0.51 <.0001 *** -0.69 0.50 <.0001 *** -0.67 0.51 <.0001 ***
Recent Fracture 0.54 1.71 <.0001 *** 0.47 1.60 0.000 *** 0.43 1.54 0.001 ***
Recent Fall -0.14 0.87 0.143 -0.16 0.85 0.100 -0.18 0.83 0.062
Admitted from Acute 0.00 . . 0.00 . . 0.00 . .
From Home -0.54 0.58 0.006 ** -0.34 0.71 0.091 -0.33 0.72 0.117
From Other -0.76 0.47 <.0001 *** -0.64 0.53 0.001 *** -0.46 0.63 0.019 *
Enabling: Generic
Lived Alone -0.02 0.98 0.845 0.05 1.05 0.656 0.03 1.03 0.759
Legally Responsible for Self 0.36 1.44 <.0001 *** 0.38 1.47 <.0001 *** 0.24 1.28 0.010 *
Medicaid -0.78 0.46 <.0001 *** -0.66 0.52 <.0001 *** -0.56 0.57 0.000 ***
SCAN 0.27 1.31 0.002 ** 0.29 1.34 0.001 ** 0.36 1.44 <.0001 ***
Enabling: Transition-Specific
Q1b) Support Person 1.37 3.92 <.0001 *** 0.99 2.69 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.00 . .
Predicted 30-90 days -0.56 0.57 <.0001 ***
Prediction Uncertain -0.78 0.46 <.0001 ***
Predicted long-stay -1.35 0.26 <.0001 ***
P1ar) Community Living Skills Training 0.29 1.33 0.002 **
0.96 2.60 0.076 0.08 1.08 0.886 0.74 2.09 0.201
Intercept
0.10 0.13 0.15
R
2
0.14 0.18 0.22
Adjusted R
2
3386.48 3291.86 3215.02
AIC
* p < 0.05, ** p < 0.01, *** p < 0.001
Model 3 (Full)
Pr > ChiSq
Model 1 (Basic)
Pr > ChiSq
Model 2 (Q1b)
Pr > ChiSq

In the full model, the only significant predisposing variables are gender
and marital status. Being female or married increase the odds of having a
community discharge within 90 days by 42% and 33% (p < 0.01) respectively.
Among the need variables, ADL limitations and bowel incontinence
decrease the likelihood of a community discharge by 10% (p < 0.01) and almost
50% respectively. Alternatively, a recent fracture increases the likelihood of
community discharge by 54%. Residents admitted from a supportive
environment (e.g. assisted living, board and care, group home, rehabilitation

96
hospital, psychiatric hospital, or nursing home) are 37% (p < 0.05) less likely to
have a community discharge before 90 days compared to residents admitted
from acute care.
Among the generic enabling characteristics, individuals who are legally
responsible for themselves are almost 30% more likely to transition compared
to residents who no longer have legal responsibility (p < 0.05). Controlling for
other factors, individuals on Medicaid are 43% less likely than non-Medicaid
residents to have a community discharge within 90 days. Compared to
traditional fee-for-service Medicare, SCAN members are 1.44 times more likely
to have a community discharge within 90 days.
The transition-specific enabling variables, the focus of this paper, are
indeed important predictors of discharge outcomes among residents with a
preference to return to the community. Residents with a support person who is
positive towards discharge are 2.69 times as likely to have a community
discharge within 90 days compared to residents without a support person who
is positive toward discharge. Compared to residents whose predicted time to
discharge is less than 30 days, residents who are predicted to discharge in 30
to 90 days are 43% less likely to transition. Residents with an uncertain
discharge prediction and residents predicted to stay more than 90 days are
even less likely to transition. Respectively, these two groups are 54% and 74%
less likely to have a community discharge within 90 days, compared to
residents with a predicted discharge within 30 days. Residents who receive

97
community living training are 33% (p < 0.01) more likely to have a community
discharge than residents who do not receive such training.
Young-Old Preference Sample
The logistic regression results for the young-old sample are presented in
Table 5.4. Among the predisposing variables age, female gender, marital
status and high education each significantly increase the likelihood of transition
by 3% (P < 0.05), 58%, 54% and 36% (p < 0.05) respectively. The following
four need variables significantly decrease the likelihood of transition among the
young-old sample: comorbidities (9%, p < 0.05), ADL limitations (15%), bowel
incontinence (53%) and admitted-from home (42%, p < 0.05). Having had a
recent fracture is the only significant need variable that increases the likelihood
of community transition (68%, p < 0.01).

98

TABLE 5.4: LOGISTIC REGRESSION, PREDICTING COMMUNITY DISCHARGE
(Q1A = YES, 65-85 SAMPLE, N = 1806)

B Exp(B) B Exp(B) B Exp(B)
Predisposing
Age 0.03 1.03 0.014 * 0.03 1.03 0.026 * 0.03 1.03 0.016 *
Female 0.54 1.72 <.0001 *** 0.46 1.58 0.000 *** 0.45 1.58 0.000 ***
Married 0.49 1.63 0.000 *** 0.41 1.51 0.002 ** 0.43 1.54 0.002 **
Caucasian 0.00 . . 0.00 . . 0.00 . .
Hispanic 0.31 1.37 0.111 0.26 1.30 0.188 0.24 1.27 0.238
African American -0.29 0.75 0.196 -0.34 0.71 0.144 -0.29 0.75 0.231
Race - Other -0.04 0.96 0.853 -0.05 0.95 0.825 -0.01 0.99 0.951
Average Education 0.00 . . 0.00 . . 0.00 . .
High Education 0.24 1.27 0.082 0.24 1.27 0.081 0.31 1.36 0.029 *
Low Education -0.19 0.83 0.214 -0.17 0.84 0.254 -0.09 0.92 0.581
Need
Cognitively Impaired -0.17 0.84 0.467 -0.22 0.81 0.360 -0.38 0.69 0.118
Depressed -0.50 0.61 0.060 -0.25 0.78 0.378 -0.15 0.86 0.608
Comorbidities -0.09 0.92 0.046 * -0.10 0.90 0.023 * -0.10 0.91 0.038 *
Socially Engaged 0.20 1.23 0.301 0.19 1.21 0.349 0.14 1.15 0.507
Behavioral Problems 0.17 1.19 0.482 0.29 1.33 0.248 0.33 1.40 0.183
ADL Score -0.15 0.86 0.001 ** -0.18 0.84 0.000 *** -0.17 0.85 0.001 ***
Incontinence - Bladder 0.14 1.15 0.373 0.18 1.20 0.255 0.23 1.26 0.152
Incontinence - Bowel -0.84 0.43 <.0001 *** -0.80 0.45 <.0001 *** -0.75 0.47 <.0001 ***
Recent Fracture 0.68 1.97 0.000 *** 0.60 1.82 0.001 *** 0.52 1.68 0.004 **
Recent Fall -0.22 0.81 0.082 -0.22 0.80 0.080 -0.23 0.79 0.071
Admitted from Acute 0.00 . . 0.00 . . 0.00 . .
From Home -0.78 0.46 0.003 ** -0.57 0.57 0.034 * -0.54 0.58 0.049 *
From Other -0.57 0.56 0.017 * -0.41 0.66 0.093 -0.18 0.83 0.472
Enabling: Generic
Lived Alone 0.08 1.09 0.535 0.15 1.17 0.266 0.12 1.12 0.410
Legally Responsible for Self 0.36 1.43 0.003 ** 0.36 1.43 0.004 ** 0.21 1.24 0.094
Medicaid -0.89 0.41 <.0001 *** -0.78 0.46 <.0001 *** -0.63 0.53 0.001 **
SCAN 0.30 1.35 0.011 * 0.30 1.36 0.011 * 0.38 1.46 0.002 **
Enabling: Transition-Specific
Q1b) Support Person 1.24 3.47 <.0001 *** 0.87 2.39 <.0001 ***
Q1c) Discharge Prediction < 30 days 0.00 . .
Predicted 30-90 days -0.44 0.64 0.016 *
Prediction Uncertain -0.86 0.42 <.0001 ***
Predicted long-stay -1.42 0.24 <.0001 ***
P1ar) Community Living Skills Training 0.35 1.42 0.004 **
-1.38 0.25 0.151 -2.17 0.11 0.027 -1.81 0.16 0.073
Intercept
0.12 0.15 0.18
R
2
0.18 0.21 0.25
Adjusted R
2
2006.32 1958.29 1906.50
AIC
* p < 0.05, ** p < 0.01, *** p < 0.001
Model 3 (Full)
Pr > ChiSq
Model 1 (Basic)
Pr > ChiSq
Model 2 (Q1b)
Pr > ChiSq

Among the generic enabling characteristics, Medicaid and SCAN are
significant predictors of transition among the young-old. Medicaid decreases
the likelihood of transition by 47% (p < 0.01). However, SCAN members are
almost 1.5 times more likely to transition than traditional fee-for-service
Medicare residents (p < 0.01).
Similar to the total sample, each of the transition-specific enabling
variables is significant for the young-old sample. Individuals with a support
person who is positive towards discharge are 2.39 times more likely to transition
than individuals without such a person. Compared to individuals who are

99
predicted to discharge within 30 days, individuals in the other three prediction
categories are significantly less likely to transition. Individuals who are predicted
to discharge within 30 to 90 days are 36% less likely (p < 0.05), those with an
uncertain discharge prediction are 58% less likely and those predicted to be
long-stay are 76% less likely to transition. Finally, individuals who receive
training in community living skills are more than 40% (p < 0.01) more likely to
transition than their peers who do not receive training.
C. Discussion
Among individuals with a preference to return to the community,
understanding which factors increase and which factors decrease the likelihood
of nursing home transition is useful for targeting transition candidates and
shaping transition policy.
The results among the predisposing and need variables are in line with
previous studies of nursing home utilization and discharge. Being female,
married, admitted from an acute hospital or having had a recent fracture
increase the likelihood of transition. ADL limitations and bowel incontinence
significantly decrease the likelihood of community discharge within 90 days.
The enabling factors, the focus of this study, prove to be significant
factors with respect to transition outcomes and fulfillment of community living
preferences. The results for the transition-specific and generic enabling
characteristics are discussed in detail below.

100
Transition Specific Enabling
The presence of a support person is the most valuable factor among the
transition-specific enabling variables. Among the total sample of residents with
a transition preference, the presence of a support person increases the odds of
a community discharge almost 270%. This is a positive finding for individuals
who have a support person, but highlights a significant disadvantage facing
individuals without such a person. Whether or not an individual has a support
person will affect the type and amount of transition assistance and community
support the individual requires. A support person is likely to help not only with
the transition, but is also likely to be called upon to provide informal support to
the individual in the community.
Among older adults who live in the community and need assistance with
everyday activities, 76% receive only informal help from family members and
unpaid caregivers. In 2007, the economic value of informal caregiving was
valued at $375 billion per year, which exceeded total Medicare or Medicaid
spending (Houser and Gibson, 2008). Supporting family caregivers is a
critically important component of long-term care rebalancing.
Some Money Follows the Person grantees have created ‘Transition
Counselors’ whose job it is to locate affordable housing, coordinate community
services, and assist with transition needs (e.g., furnishing the home, transferring
prescriptions, stocking the pantry). Transition counselors could be of particular
assistance to individuals who have a preference to return to the community but

101
do not have a support person who can help with discharge planning and
potentially provide care after discharge. Transition counselors could also
provide support to the transitioning individual’s support person.
Although discharge prediction is a highly significant predictor of
community discharge, it is a complex variable to interpret because it is a
summary variable based on the assessor’s judgment. The discharge prediction
variables are correlated with length of stay and discharge outcomes, but the
causal direction is unclear.
Individuals who receive training in community living skills are 33% more
likely to return to the community compared to residents who do not receive such
training. This rather intuitive finding deserves future examination. Little is
known about how residents are selected for the training, what the training
entails or how consistent it is across different facilities. Additionally, although
the training is associated with community discharge, the causal direction is
again unclear. If the only people who receive training are those who have been
identified by the nursing home staff as likely to transition, then the association
would be the result of selection bias. In order to fully understand the effect of
community living training on discharge outcomes more research is needed,
ideally in the form of a more controlled experiment.
It is interesting to note the changes that occur between Models 2 and 3
in Tables 5.3 and 5.4. The addition of discharge prediction and community
living skills training has a dampening effect on two important enabling variables:

102
legal responsibility for self and presence of a support person positive towards
discharge. In Table 5.3, before the discharge prediction or training variables
are added (Model 2), maintaining legal responsibility is highly significant and
increases the likelihood of transition by 47%; the presence of a support person
increases the likelihood by 392%. However, once the additional variables are
added in Model 3, the effect of legal responsibility decreases to 28% and
becomes only modestly significant (p < 0.05). Similarly, the effect of having a
support person decreases to approximately 270%. A similar effect can be seen
among the age 65 to 85 sample in Table 5.4. Legal responsibility for self goes
from significant (OR 1.43, p < 0.01) to insignificant and the effect size of having
a support person remains highly significant but decreases from 3.47 to 2.39.
The dampening effect of discharge prediction and training may reflect
specific tracks on which individuals get put based on the staff’s beliefs about
their potential for transition. Based on staff perception, residents may receive
different levels of care such that discharge prediction becomes a self-fulfilling
prophecy. Alternatively, the dampening effect of discharge prediction and
training on preference may reflect unrealistic transition expectations on behalf
of the individual. The California Nursing Facility Transition Screen (CNFTS)
may be a useful tool for sorting out realistic versus unrealistic transition
expectations of residents and their families (Kane, 2008; Nishita, 2008)
The CNFTS includes 27 open and closed-ended questions about the
resident’s preference and perceived ability to move to the community. The

103
screen also explores potential living arrangements and services in the
community (Nishita et al., 2008). Using the CNFTS, Nishita et al (2008) found
that the percent of nursing facility residents who endorsed community living
could be increased through education about available home and community-
based services.
Generic Enabling
Among the generic enabling characteristics, legal responsibility for
oneself and membership in the SCAN Health Plan increase the likelihood of
transition while Medicaid funding and admission from another care setting are
significant barriers to transition.
Among the full sample, residents who are legally responsible for
themselves are almost 30% (p < 0.05) more likely have a community discharge
compared to residents who are no longer legal responsible for themselves.
Previous research has also shown that self-consenting residents are more likely
to express a stable transition preference (Nishita et al., 2008). If cognitively
intact, older adults should be encouraged to retain legal responsibility for
decision-making for as long as it is feasible and prudent. Establishing one’s own
goals and taking a proactive role in health care decisions, allows individuals to
actively work toward their desired outcomes, such as transition.
Medicaid is the most significant barrier to community transition.
Medicaid coverage decreases the likelihood of community discharge within 90
days by 43%. This finding is particularly notable because the Medicaid status

104
was taken from each individual’s admission assessment. Therefore the
analysis does not include individuals who spend down and enroll in Medicaid
during the course of their stay. Therefore, the negative effect of Medicaid on
discharge in this study is a conservative estimate.
This Medicaid finding is not unexpected and highlights the institutional
bias of the long-term care system, which persists ten years after the Olmstead
Decision. Home and Community Based Services waivers, Real Choice
Systems Change grants, Money Follows the Person grants and Nursing Home
Transition programs have been chipping away at the institutional bias but there
remains much progress to be made in order to balance the long-term care
system in America. Adding to the difficulty is the fact that each state has
considerable flexibility regarding how state Medicaid programs are designed.
This flexibility is important and necessary, but also makes it difficult to make
systemic change across the country.
The lack of affordable, accessible housing in the community presents an
additional barrier to nursing home residents, especially Medicaid recipients.
Long waiting lists and low vacancy rates for Section 8 and other public housing
are problems in many states. Additionally, Medicaid’s strict asset and income
limits make it difficult for it’s beneficiaries to afford to live in the community.
Many people, even homeowners, are at risk of losing their housing during a
nursing home stay.

105
Transition counselors, mentioned above, could also be of particular
assistance to Medicaid residents with a preference to return to the community.
This population is potentially eligible for a host of home and community based
services such as Multipurpose Senior Services Program and In-Home
Supportive Services but may need help navigating the complex eligibility and
enrollment process.
Previous evaluations of the S/HMO plans were unable to identify
sufficient benefits to justify the additional 5% capitated rate per person
compared to Medicare Managed Care. Specifically, the evaluations provided
no consistent evidence that the S/HMO benefit affected physical health or
service utilization, including nursing home admission. Based on the evaluation
results, it was recommended that the S/HMOs be converted to Medicare +
Choice plans at the end of the demonstration in 2007 (Wooldridge et al., 2001).
However, in 2003, Fischer and colleagues found that the Social HMO model
delayed long-term nursing home placement among at-risk older adults (Fischer
et al., 2003). The current study also found a favorable outcome among S/HMO
members. Specifically, SCAN membership increases the likelihood of
community discharge by approximately 45%.
The consistent findings across various analyses in the present study that
SCAN membership positively affects transition have potentially significant policy
implications. Although the demonstration evaluations concluded that the
S/HMOs did not keep individuals out of nursing home, findings in the present

106
study suggest that SCAN helps individuals transition back to the community.
Compared to the extensive body of research on diversion, transition, while
equally important has received little systematic research attention. Given that
approximately 40% of older adults will spend at least some time in a nursing
home prior to death (Liu, 1994), transition back to community living is an
important part of the equation. This consistent finding is likely due to the
availability of community-based long-term care services to SCAN members.
These services included care coordination, home nursing visits, transportation,
adult day care and homemaking services (Fischer et al., 2003)
The findings that SCAN decreases the likelihood of becoming a long-stay
nursing home resident have significant implications for the design and funding
of long-term care services. The findings indicate that people do not just need
help getting out of nursing facilities, they also need integrated health and long-
term care services once they are back in the community. As the new
administration examines health care reform options, these findings suggest that
the Social HMO model should be thoughtfully reconsidered.
Additional practice and policy implications of these findings are described
in Chapter 6.
Limitations
The preference data and all of the independent variables are taken from
each resident’s admission assessment. This is a limitation because, after
admission, any number of health, personal or financial changes could occur in

107
an individual’s life that affect discharge preferences and potential. Additionally,
Section Q is only on the admission assessment and subsequent full annual
assessment forms of MDS 2.0, not on quarterly assessments. Therefore,
information about a resident’s preference may be up to one year old.
Understanding how discharge preferences and predicted discharge potential
can change over time would be a valuable topic for future research. Adding the
Section Q questions to the quarterly assessment would also greatly improve
targeting and transition efforts.
The receipt of community living training and discharge prediction variables
have not been extensively used in nursing home transition studies. Although
they were found to be significant predictors in this study, especially discharge
prediction, additional research is necessary. Discharge prediction is a complex
variable to interpret because it is a summary variable based on the assessor’s
judgment. The discharge prediction variables are correlated with length of stay
and discharge outcomes, but the causal direction is unclear. Based on the
discharge prediction and the staff’s beliefs about the individual’s potential for
discharge, it is possible that residents may receive different levels of care such
that the prediction becomes a self-fulfilling prophecy. These limitations should
be kept in mind when interpreting the full model regression results. However, in
spite of the limitations, these variables are promising tools for transition practice
and policy.

108
Although these findings are based on a diverse sample from four large
Southern California counties, researchers and practitioners must use caution
generalizing from these results. Additional research is needed to identify ways
to systematically use the rich MDS data to increase the effectiveness and
efficiency of nursing home transition programs.

109
CHAPTER VI: CONCLUSION

These dissertation findings have significant implications for practice and
policy. The findings support the argument that the Nursing Home Minimum
Data Set (MDS) version 2.0 assessment process and forms need to be
modernized in order to better support the New Freedom Initiative. Additionally,
the predictors of community discharge can and should be used to target
transition candidates, develop tiered transition interventions and inform
rebalancing policy.
A. The Need to Modernize the MDS
For some years, Section Q of the MDS has been a topic of discussion
among advocates, researchers and practitioners. Most agree that the MDS 2.0
needs to be modernized in order to better support the policies of the New
Freedom Initiative. The results of this dissertation support the argument for
modernization of the MDS assessment forms specifically and the overall
assessment process generally.
Assessment Form
Many states and advocacy organizations find Section Q of MDS 2.0 to
be of little value given the current structure of the questions. There are issues
with the current wording of some of the Section Q items as well as the
placement of Section Q.
Although the preference question (Q1a) may have some utility for
identifying potential transition candidates, it has empirical and practical

110
limitations. The CMS MDS User’s Manual instructs assessors to pose the
question differently depending on how long the person has resided in the
nursing facility. Assessors are instructed to ask new and recent admissions the
question directly. But, when evaluating individuals who have resided in the
facility for a long time, assessors are advised to use indirect questions to avoid
creating unrealistic expectations. These instructions are viewed as paternalistic
by advocacy organizations and are likely to introduce the assessor’s bias into
the responses (Kane, 2008).
Question Q1b asks whether the individual has a support person who is
positive toward discharge. The potential answers are simply yes or no. This
leaves ambiguity with respect to what a negative response means. It could
mean that there is no support person or it could mean that there is a support
person but that he/she is not positive towards discharge. These are two very
different scenarios with very different implications for transition.
Answers to the question regarding predicted discharge timeframe (Q1c)
have not been reported on in descriptions of transition programs (Ribar &
O’Keeffe, 2005). In light of the Deficit Reduction Act’s focus on residents who
have been in the nursing home at least six months, some advocates have
suggested adding answers to capture the possibility of longer stays such as 90
to 120, 120 to 150 and 180 or more days.
Section Q is one of the last sections of the admissions and annual
assessment forms. This placement implies that discharge planning is less

111
important than other assessed topics. Moving the discussion of discharge
preferences to the beginning of the MDS assessment interview would increase
the visibility of and emphasis on transition for residents, families and the nursing
home staff.
Furthermore, Section Q is only on the full assessment forms and is not
included on quarterly assessments. Therefore, information about resident
preferences and discharge potential could be up to one year old. Adding the
Section Q questions to the quarterly assessment would greatly improve
targeting and transition efforts. The ability to capture predicted discharge on a
quarterly basis would be useful for targeting transition candidates, especially at
critical times such as 90 and 180 days. It would also enable researchers to
examine how discharge preferences and predicted discharge change over time.
Assessment Process
The lack of follow up and accountability for Section Q is a significant
weakness of the MDS and inhibits discharge planning. Section Q is one of the
few sections of the MDS whose answers are not included in Resident
Assessment Protocols (RAP). Furthermore, answers from Section Q are not
used in any of the quality indicators developed by CMS, not published in
“Nursing Home Compare,” not used in any of the resource utilization groups
(RUGs) and not included on quarterly assessments.
RAPs are used in conjunction with the MDS. Specific answers to MDS
items can trigger the initiation of a RAP. The purpose of RAPs is to identify

112
when follow-up action is required of the nursing home staff, to assist in
decision-making and to support the development of individualized care plans.
Without a discharge planning RAP, no one is accountable for action based
on residents’ responses in Section Q. A discharge planning RAP would help
facilitate the targeting strategies proposed above. Potential discharge
candidates could be identified and scored based on their answers to questions
such as Q1a (preference), Q1b (support person positive towards discharge),
discharge prediction, legal responsibility, payment source, ADL limitations and
bowel incontinence. The RAP guidelines would require follow-up action by the
nursing home staff. This could include actions such as administering a more in
depth preference and feasibility interview, providing community living skills
training or making a referral to a community transition advocacy organization.
Options counseling and transition planning are complex and time-
consuming tasks. Nursing home staff and advocacy agencies have limited time
and resources to deal with these issues. Therefore, as the transition
assessment process is examined and improved, researchers and advocacy
organizations should try to develop ways to make these tasks more efficient.
This could be as simple as developing an informational DVD for consumers to
watch and/or the provision of a self-assessment workbook to help consumers
appraise their community support network and learn about available
community-based services. These tools could be provided as a low-cost first
step and in-person follow-up could be provided as needed.

113
Conflicting Incentives
As in any real-world process, it is important to acknowledge the different
stakeholders involved and their different incentives. Nursing home personnel
have different incentives than transition advocates and consumers. Nursing
home administrators are focused on occupancy levels and front-line nursing
home staff are likely to prefer an individual with low care needs over someone
with high care needs. These incentives are at odds with the goals of transition
programs. The development of referral protocol, which would require nursing
home personnel to work with community advocacy organizations, could help
minimize some of the issues with misaligned incentives.
Section Q variables are highly predictive of community discharge.
Therefore it is important that residents and their families are informed about
how Section Q was filled out for the resident. The resident or a family member
should be required to sign off on the nursing home’s assessment of the
resident’s discharge potential. If a resident or family member disagrees with the
assessment, they should be referred to an ombudsman or to an advocacy
agency. Furthermore, whether or not a person receives community living skills
training should be decided in conjunction with the individual and his/her family.
Individual should always be given the opportunity to opt into such training
programs.

114
B. Targeted Transition Strategies
Nursing home discharge and transition are complex processes. As
described in Chapters 4 and 5, an array of factors influence transition
outcomes. The results among the predisposing and need variables are in line
with previous studies of nursing home utilization and discharge. Being female,
married, admitted from an acute hospital or having had a recent fracture
increase the likelihood of transition while ADL limitations and bowel
incontinence decrease the likelihood.
The enabling variables, which are often the most susceptible to policy
intervention, prove to be particularly important predictors of transition.
Specifically, SCAN membership was found to significantly increase the
likelihood of community discharge for all sub-populations in this study including
the oldest-old. This finding highlights the importance of community-based
services to support and sustain transition efforts.
Given the importance of the transition-specific enabling variables -
community living preference, presence of support person, predicted discharge
and receipt of community living skills training - these factors provide a good
starting point for developing and refining transition strategies. Specifically, the
presence or absence of these factors can be used to target transition
candidates and to tailor transition interventions.

115
Targeting Strategies
There are a variety of ways that the significant predictors found in
Chapters 4 and 5 could be used to create transition criteria and to channel
scarce resources toward appropriate candidates. Some targeting strategies
could be employed right away, while other strategies, such as creating a
discharge planning RAP, will take longer to implement.
Each of the three targeting strategies described below could be used to
classify individuals as having either a low, medium and high likelihood of
transition. Tiered transition interventions based on the different levels are
described in the next section.
Using Discharge Prediction and Accrued Nursing Home Days
The discharge prediction timeframe is a strong predictor of community
discharge. One simple strategy would be to compare accrued length of stay
with discharge prediction. Individuals whose accrued number of days in the
nursing home is less than their predicted discharge timeframe, could be
classified as having a high likelihood of transition. Individuals who have
exceeded their predicted discharge by 1 to 30 days could be classified as
having a medium likelihood of transition. Individuals who have exceeded their
predicted discharge by more than 30 days could be classified as having a low
likelihood of transition. There should be protocol in place to check in with each
individual and his/her family as the resident approaches his/her predicted
discharge time frame. If it appears that the individual is not on track for

116
transition, he/she may need the next level of transition support. This method is
more reactive than proactive, but it is simple and could be an effective targeting
approach for some states. This approach could be used with residents of all
ages.
Using Preference & Discharge Prediction
Combining preference information with the discharge prediction timeframe
is another potential strategy to identify candidates for different levels of
transition assistance. Based on the finding that preference is not a significant
predictor of community discharge for the oldest-old, this targeting strategy is
suggested for use with residents ages 65 to 85. Specific targeting strategies for
the oldest-old are described further below.
Among residents ages 65 to 85, individuals with a preference to return to
the community and a predicted discharge within 30 days, have a high likelihood
of transitioning on their own without assistance. These individuals would
require minimal transition intervention.
Residents with a preference to return to the community who are predicted
to discharge within 30 to 90 days or whose discharge prediction is uncertain are
at higher risk for becoming long-stay residents and would be good potential
candidates for mid-level transition assistance.
Finally, individuals with a community-living preference who are not
predicted to discharge within 90 days have a low likelihood of transition. These
individuals may require the most intensive level of transition assistance.

117
Using Discharge Probability Score
The final approach would be to use regression results to predict each
individual’s probability of being discharged to the community. The findings from
this study, and future studies like it, could be used to determine which
regression coefficients to include in the probability equation. In conjunction
with residents’ informed preference, which should be primary, a discharge
probability score could be used to guide targeting strategies.
The probability score could be used to divide the resident population into
different groups based on the level of transition assistance required. For
example, individuals with a 75% or higher probability of discharge would be
candidates for minimal transition assistance. This level could also include
monitoring and assistance if barriers are encountered, which would increase the
likelihood of successful transition. Individuals who have a 50-75% probability of
transitioning would be candidates for mid-level transition assistance. Finally,
individuals with a less than 50% probability would be candidates for the
intensive transition assistance.
The aforementioned probability cutoff points are arbitrary. Additional
research is needed in order to refine the targeting criteria and to find the optimal
cut off points for the different transition assistance levels.
In the future, a discharge planning RAP could automatically generate each
individuals discharge probability score based on answers to specific MDS
questions. The probability score could be used by transition advocacy

118
organizations to identify transition candidates. The average discharge
probability score could also be used as a quality indicator to compare transition
success across states, counties and nursing homes.
Tiered Transition Interventions
In recognition of the fact that people need different amounts of transition
assistance, states should develop tiered levels of transition support. Potential
transition assistance levels are described below. Targeting methods, like those
described above, can be used to assess the likelihood of transition and
categorize residents into one of the following tiers.
Minimal Assistance
Residents with a high likelihood of transitioning on their own without
transition assistance would be candidates for the lowest level of transition
assistance. This level could be as simple as going over the resident’s predicted
discharge timeframe with the resident and his/her family and making sure they
are aware that transition counselors are available if they feel that the resident is
getting off track for a short-stay, community discharge. Individuals in this
category could also receive basic information about home and community-
based services. If appropriate, these individuals should be offered community
living skills training for themselves and/or their support person.
Mid-Level Assistance
Mid-level transition strategies could be used for individuals with a
moderate probability of community discharge. Strategies for this level could

119
include services such as an additional preference and feasibility assessment,
referral to a community living advocacy organization and information about
community living options. This group should also be offered community living
skills training.
The California Nursing Home Transition Screen (CNFTS) could be used
for the additional preference and feasibility assessment. The CNFTS includes
27 open and closed-ended questions about the resident’s preference and ability
to move to the community. The screen also explores potential living
arrangements and services in the community (Nishita et al., 2008). Using the
CNFTS, Nishita et al (2008) found that community living preference could be
manipulated through education about available home and community-based
services. The CNFTS is being implemented in California and being referenced
as a model by other states.
In order to maximize resource efficiency, transition programs could
consider waiting to administer the time-consuming CNFTS assessment to this
group until around their thirtieth day in the nursing home. This would be long
enough to allow a majority of the short-stay residents to transition on their own,
but not so long that the remaining residents have lost their community housing
or support. Use of this 30-day timeframe is also defensible based on the finding
that the average length of stay of someone who transitioned in this study was
approximately 32 days.

120
Intensive Assistance
Intensive transition assistance strategies could be used for individuals
with a low probability of community discharge or a negative discharge prediction
timeframe. These individuals are likely to be the most difficult to transition but
could potentially have the highest payoff in terms of cost savings and quality of
life. In addition to all the mid-level services, individuals in this group could also
be assigned a dedicated transition counselor. The job of a transition counselor
is to locate affordable housing, coordinate community services, and assist with
transition needs (e.g., furnishing the home, transferring prescriptions, stocking
the pantry).
The above services are just an example of potential tiered transition
interventions. Based on resource availability and transition goals, each state
would need to decide which services to provide in each level and on which
level(s) to focus.
Targeting Strategies for the Oldest Old
As discussed in Chapter 4, the oldest-old face unique transition
challenges. Community living preference and community living skills training
are not a significant predictors of community discharge for this population.
However, the presence of a support person who is positive towards discharge is
immensely important for this group. In light of these factors, this group could
benefit from transition strategies customized to meet their specific needs. The
projected growth of this population and their high utilization of long term care

121
services, suggest that tailored transition strategies for this population are worth
pursuing and could have a significant impact.
Targeting strategies tailored specifically for individuals over age 85 are
described below. Based on the unique predictors of transition among this
population, oldest-old individuals can be categorized into one of the transition
assistance levels described above.
Using Discharge Prediction and Support Person
Oldest-old individuals with a predicted discharge within 30 days and a
support person who is positive toward discharge have a high likelihood of
transitioning on their own. This group could be categorized into the minimal
assistance level.
Individuals who are predicted to discharge within 30 days but do not
have a support person should receive mid-level transition assistance. Similarly,
individuals who are predicted to discharge within 30 to 90 days or who have an
uncertain discharge prediction should also receive mid-level assistance. Mid-
level transition assistance could include an additional preference and feasibility
assessment, such as the California Nursing Facility Transition Screen (CNFTS).
The comprehensive CNFTS could be particularly valuable for the oldest-
old, especially since their Q1a preference does not impact their likelihood of
transition. As part of the screen, the individuals social support network is
assessed. Given the significant impact of having a support person who is

122
positive towards discharge, this information is particularly important for this
population.
Furthermore, the transition screen also educates the individuals and their
families about potential living arrangements and community-based services.
This type of education may be particularly valuable for the oldest-old population.
Older adults, people with less than a high school education, low-income
individuals, minorities and individuals with poor health status are the most likely
to experience low health literacy (Institute of Medicine, 2004). Health literacy
refers to an individual’s ability to obtain, process and understand health
information in order to make informed and appropriate health decisions (U.S.
Department of Health and Human Services, 2000). Individuals with low health
literacy may not be aware of community alternatives to nursing homes.
Previous studies have shown that community living preference and perceived
need for services can be changed through education about available programs
and services (Nishita et al., 2008; Borrayo et al., 2002).
The minimal, mid-level and intensive strategies described above each
include the option to receive community living skills training. However, in its
current form, the training program did not have an effect on the transition
outcomes of the oldest-old. Transition programs may choose to develop
training curricula geared specifically toward the unique transition needs of the
oldest-old. As mentioned above, community transition programs might want to

123
explicitly include an education component, such as options counseling, into
training programs for the oldest-old.
Using Discharge Probability Score
A discharge probability score could also be calculated for the oldest-old.
The probability equation could include slightly different targeting criteria based
on the unique predictors of community discharge among this group.
Alternatively, age could be worked in as a factor in the original probability
equation.
C. Conclusion
The difficult challenge facing transition programs is to strike a balance
between the preference of the consumer, the cost of the intervention, the
probability of success and the potential cost savings to Medicaid. Although
nursing home transition is complex, if an individual can have their needs met in
the setting of their choice for a fraction of the cost, it is certainly worthwhile.
Relatively to diversion and delay, nursing home transition has been
largely overlooked. However, it is also important to remember that nursing
home transition is just one of many strategies being employed in order to
rebalance the long-term care system. The success of nursing home transition
programs is intrinsically linked to the success and progress of other rebalancing
efforts such as expansion of community-based services, publicly or waiver-
funded assisted living benefits, caregiver support and nursing home diversion
programs.

124
D. Addendum Regarding MDS 3.0
It is important to note that all of the above results, conclusions and
recommendations are based on MDS version 2.0 and were completed prior to
the release of MDS version 3.0 on May 7, 2009. Full implementation of MDS
3.0, including national data collection and skilled nursing reimbursement based
on MDS 3.0 data, is scheduled to begin October 1, 2010.
As part of a multi-year, national project, CMS has been working with
resident advocates, MDS users and researchers to develop and validate the
new MDS 3.0. The purpose of the project was to “improve the clinical
relevance and accuracy of MDS assessments, increase the voice of residents
in assessments, improve user satisfaction, and increase the efficiency of
reports” (Seliba & Buchanan, 2008. p. ix). In order to streamline the
assessment instrument, the revised focus of the MDS 3.0 is to screen for
common and often-missed geriatric conditions, while follow-up is now
considered the job of the Care Area Triggers (CAT) and the care plan (Seliba &
Buchanan, 2008). CATs in MDS 3.0 are analogous to the Resident
Assessment Protocols (RAPs) in MDS 2.0.
CMS made significant and numerous changes throughout MDS 3.0.
Some of the key changes that impact nursing home transition are discussed
below.

125
One of the most notable transition-oriented changes is the addition of the
Return to Community Referral CAT. This CAT makes discharge-related items
an explicit part of the care planning process, which they are not in MDS 2.0.
The specific triggers of the Return to Community Referral CAT are
scheduled to be published in October 2009. Even if the initial set of triggers are
not sophisticated or the associated care guidelines are weak, the fact that there
is now a discharge-oriented CAT increases the visibility of transition efforts.
The targeting strategies outlined earlier in this chapter could be used to guide
future development of the triggers for the Return to Community Referral CAT.
Furthermore, once sufficient MDS 3.0 data has been collected, studies similar
to those in this dissertation should be re-run and analyzed.
Ambiguity around how Section Q questions are asked in MDS 2.0 has
been a significant concern of transition advocates. One of the major goals of
the MDS revision was to introduce more direct resident interview items in order
to increase the voice of residents. Section Q was one of the specific areas
where additional interview items were added. Questions Q1a, Q1b and Q1c
from MDS 2.0 are not included in MDS 3.0. In their place there is a question
about the resident’s overall goals and a question about whether the resident
wants to talk to someone about the possibility of returning to the community.
Based on the findings in this dissertation, CMS may have been remiss to
remove the questions about the presence of support person and the predicted
time to discharge. Both of these variables were consistently found to be

126
significant in the above studies. These variables are also potentially beneficial
variables to guide targeting and prioritization of transition candidates. Hopefully
these significant variables will be incorporated into the Return to Community
Referral CAT or elsewhere in the care planning process.
One significant risk of MDS 3.0 Section Q is that the referral method may
simply shift the burden away from nursing homes and onto already under-
funded agencies. This potential to overwhelm local agencies and community
advocacy groups should be carefully considered. The targeting strategies
proposed in this dissertation could be used by agencies to help manage and
prioritize the likely influx of referrals.
Finally, although MDS 3.0 includes numerous changes to individual
questions and variables, the various MDS forms will continue be used to keep
track of individuals as they bounce in and out of nursing facilities. Therefore it
is important to note that the episode of care method described in Chapter 2
remains equally valuable for use with MDS 3.0.

127
BIBLIOGRAPHY

Alecxih, L. (2007). Opening plenary: ADRC role in nursing facility transition and
diversion. ADRC Conference: Aging and Disability Resource Center.

Angelelli, J. J., Wilber, K. H., Myrtle, R. (2000). A Comparison of Skilled Nursing
Facility Rehabilitation Treatment and Outcomes Under Medicare
Managed Care and Medicare Fee-for-Service Reimbursement.
Gerontologist 40: 646-653

Andersen, R. (1995). Revising the behavioral model and access to medical
care: Does it matter? Journal of Health and Social Behavior, 36, 1.

Andersen, R., & Newman, J. (1973). Societal and individual determinants of
medical care utilization in the United States. Milbank Memorial Fund
Quarterly, 51, 95-124.

Anderson, W.L., Wiener, J.M., O’Keefe, J. (2006). Money Follows the Person
Initiatives of the Systems Change Grantees: Final Report. Research
Triangle Park, NC: Research Triangle Institute.

Barer, M., Stoddart, G. (1981). Analysis of demand and utilization through
episodes of medical services. In J. van der Gaag and M. Perlman (eds.),
Health, Economics and Health Economics. Amsterdam: North Holland.

Batavia, A. (2002). Consumer direction, consumer choice and future of long-
term care. Journal of Disability Policy Studies, 13(2), 67-74.

Borrayo, E.A., Salmon, J.R., Polivka, L., Dunlop, B.D. (2002). Utilization Across
the Continuum of Long-Term Care Services. The Gerontologist 42:603-
612 (2002)

Burrows, A., Morris, J., Simon, S., Hirdes, J., Phillips, C. (2000). Development
of a minimum data set-based depression rating scale for use in nursing
homes. Age and Aging, 29, 165-172.

California Pathways, Real Choice Systems Change Grant Preference
Assessment Tool, (2005)

Carpenter, G. I., Hastie, C. L., Morris, J. N., Fries, B. E., Ankri, J. (2006).
Measuring change in activities of daily living in nursing home residents
with moderate to severe cognitive impairment. BMC Geriatrics, 6(7)

128
Centers for Medicare and Medicaid Services. (2002). CMS RAI version 2.0.
Chapter 3: Item-by-item guide to the MDS. Washington, DC: Department
of Health and Human Services. Available at:
http://www.cms.hhs.gov/NursingHomeQualityInits/
Downloads/MDS20rai1202ch3.pdf. Accessed April 24, 2006.

Centers for Medicare and Medicaid Services (2008). MDS Active Resident
Information Report: Fourth Quarter 2008. Washington, DC.: Department
of Health and Human Services. Available at
http://www.cms.hhs.gov/MDSPubQIandResRep/04_activeresreport.asp#
TopOfPage. Accessed April 10, 2008.

Chapin, R., Wilkenson, D. S., Rachlin, R., Levy, M., Lindbloom, R. (1998).
Going home: Community reentry of light care nursing facility residents
age 65 and over. Journal of Health Care Finance, 25(2), 35.

Charlson, M.E., Pompei, P., Ales, K.L. et al (1987). A new method of classifying
prognostic comorbidity in longitudinal studies: Development and
validation. Journal of Chronic Disease, 40, 373-383.

Feder, J., Komisar, H., Friedland, R. (2007). Long-term care financing: Policy
options for the future. Washington, DC: Georgetown University.

Fischer, L. R., Green, C. A., Goodman, M. J., Brody, K. K., Aickin, M., Wei, F.
(2003). Community-based care and the risk of nursing home placement.
Medical Care, 41(12), 1407-1416.

Fischer LR, Hoffman C. (1994) Who cares for the elderly: the dilemma of family
support. In: Lewis M, Miller J, eds. Research in Social Problems and
Public Policy, vol II. Greenwich, CT: JAI Press,169 –215.

Gibson, M. (2003). Beyond 50: A report to the nation on independent living and
disability.

Hing, E., Sekscenski, E., Strahan, G. (1989). The national nursing home
survey: 1985 summary for the United States (DHHS Pub. No. PHS 89-
1758, pp. 1-249). Washington, DC: Government Printing Office.

Hornbrook, M.C., Hurtado, A.V., Johnson, R.E. (1985) Health Care Episodes:
Definition, Measurement and Use. Medical Care Research and Review,
Vol. 42, No. 2, 163-218.

129
Houser, A.N., Gibson, MJ. (2008) Valuing the Invaluable: The Economic Value
of Family Caregiving, 2008 Update. AARP Public Policy Institute.
Washington, DC.

Institute of Medicine (2004). Health Literacy: A Prescription to End Confusion.
Washington, DC: The National Academies Press.

Kane, R. (2008). Providing Structured Opportunities for Nursing Home
Residents to Choose Community Care. Journal of the American
Geriatrics Society. 56(1): 163-165.

Kane, R., Bell, R., Riegler, S., Wilson, A., Keeler, E. (1983). Predicting
outcomes of nursing home patients. The Gerontologist, 23, 200-206.

Kapp, M. (2000). Consumer direction in long-term care: A taxonomy of legal
issues. Generations, 24(3), 16.

Kasper, J., O'Malley, M. (2006). Nursing home transition programs:
Perspectives of state Medicaid officials. Kaiser Commission on Medicaid
and the Uninsured.

Kassner, E., Reinhard, S., Fox-Grage, W., Houser, A., Accius, J. (2008). A
balancing act: State long-term care reform. Washington, DC: AARP
Public Policy Institute (No. 2008-10).

Keeler, E.B., Kane, R.L., Solomon, D.H. (1981). Short- and long-term residents
of nursing homes. Medical Care, 19, 363-370.

Leutz, W.N., Greenberg, J.N., Abrahams, R. et al. (1985). Changing Health
Care for an Aging Society: Planning for the Social Health Maintenance
Organization. Lexington, MA: Lexington Books.

Levit, K., Lazenby, H. C., Braden, B. R., Cowan, C. A., McDonnell, P. A.,
Sivarajan, L., et al. (2006). National health expenditures, 1995. Health
Care Financing Review, 18(1), 175-214.

Lewis, M.A., Cretin, S., Kane, R.L. (1985). The natural history of nursing home
patients. Gerontologist, 25, 382-388.

Liu, K. 1994. "A Data Perspective on Long-Term Care." The Gerontologist 34
(4): 476-80.

Liu, K., Manton, K. (1983). The characteristics and utilization pattern of an
admission cohort of nursing home patients. The Gerontologist, 23, 92-98.

130

Liu, K., Palesch, Y. (1981). The nursing home population: Different perspectives
and implications for policy. Health Care Fin Rev, 3, 15-23.

Mehr, D.R., Williams, B.C., Fries, B.E. (1997). Predicting Discharge Outcomes
of VA Nursing Home Residents. Journal of Aging and Health, 9(2), 244-
265.

MetLife Mature Market Institute. (2006). MetLife market survey of nursing home
and home care costs. Westport, CT:

Miller, E. A., Weissert, W. G., (2000) Predicting elderly people's risk for nursing
home placement, hospitalization, functional impairment and mortality: A
comparative review and analysis. Medical Care Research and Review,
57:(3), 259-297

Mor, V. (2004). A comprehensive clinical assessment tool to inform policy and
practice: Applications of the minimum data set. Medical Care, 42(2),
III50.

Mor, V. National Long-Term Care Databases: Clinical Assessment Tools with
Research and Policy Applications, Baltimore, Maryland.

Mor, V., Branco, K., Fleishman, J., Hawes, C., Phillips, C., Morris, J., et al.
(1995). The structure of social engagement among nursing home
residents. Journal of Gerontology: Psychological Sciences, 50(1) B, P1-
P8.

Morris, JN., Fries, BE., Mehr, DR., Hawes, C., Phillips, C., Mor, V., Lipsitz, LA.
(1994). The MDS Cognitive Performance Scale. Journal of Gerontology,
49(4), 174-182.

Morris, J.N., Hawes, C., Fries, B., Phillips, C., Mor, V., Katz, S. (1990).
Designing the national resident assessment instrument for nursing
homes. Gerontologist, 30, 563-564.

Morris, J.N., Nonemaker, S., Murphy, K., Hawes, C., Fries, B.E., Mor, V., et al.
(1997). A commitment to change: Revision of HCFA’s RAI. Journal of the
American Geriatrics Society, 45, 1011-1016.

Morris, JN, Fries, BE, Morris, SA. (1999). Scaling ADLs within the MDS. Journal
of Gerontology: Medical Sciences, 54A(11), M546-M553.

131
Murtaugh, C. M., P. Kemper, B. C. Spillman, and B. L. Carlson. 1997. "The
Amount, Distribution, and Timing of Lifetime Nursing Home Use."
Medical Care 35 (3): 204-18

Narin, P., Rubenstein, L.Z., Wieland, G.D., Rosbrook, B., Strome, L.S.,
Pietruszka, F. & Morley, J.E. (1988). Predictors of immediate and 6-
month outcomes in hospitalized elderly patients: The importance of
functional status. Journal of American Geriatrics Society, 36, 775-783.

Newcomer, R., Harrington, C., Friedlob, A. (1990). Awareness and enrollment
in the Social/HMO. The Gerontologist, 30(1), 86-93.

Nishita, CM., Wilber, KH., Matsumoto, S., and Schnelle, JF. (2008).
Transitioning residents from nursing facilities to community living: Who
wants to leave? Journal of the American Geriatrics Society, 56(1), 1-7.

Phillips, C., Hawes, C., Mor, V., Fries, B., & Morris, J. (1996). Evaluation of the
nursing home resident assessment instrument. Research Triangle
Institute.

Poon, L.W., Jang, Y., Reynolds, S., and McCarthy, E., Profiles of the Oldest Old
in The Cambridge handbook of age and ageing by Johnson, M.L.,
Bengtson, V.L., Coleman, P.G., Kirkwood, T.L.B. Cambridge University
Press, 2005.

Reinhard, S., and Hendrickson, L. (2006). Money Follows the Person: States’
Progress in Using the Minimum Data Set (MDS) to Facilitate Nursing
Home Transition. New Brunswick, New Jersey: Rutgers/NASHP
Community Living Exchange. Retrieved on March, 2007, from
http://www.cshp.rutgers.edu/.

Reinhard, S., and Hendrickson, L. (2007). The minimum data set:
Recommendations to help states better support nursing home residents
who seek community living. New Brunswick, NJ: Rutgers Center for
State Health Policy & National Academy for State Health Policy:
Community Living Exchange.

Reinhard, S., Hendrickson, L., and Bemis, A. (2005). Using the Minimum Data
set (MDS) to facilitate nursing home transition. New Brunswick, New
Jersey: Rutgers/NASHP Community Living Exchange. Retrieved on
March, 2007, from
http://www.cshp.rutgers.edu/cle/products/MDS_Final_WEB.pdf.

132
Ribar, M. & O’Keefe, J. (2005, June). Systems Change Grant Program Nursing
Facility Transition Initiatives of the Fiscal Year 2001 and 2002 Grantees:
Progress and Challenges Final Report. Center for Medicaid and
Medicare Services, U.S. Department of Health and Human Services.
Retrieved January 9, 2009 from
http://www.hcbs.org/files/74/3656/NFT_final_web.pdf

Seliba, D., Buchanan, J (2008). Development and Validation of a Revised
Nursing Home Assessment Tool: MDS 3.0. Santa Monica, CA: RAND
Health.

Sgadari, A., Morris, J.N., Fries, B.E., Ljunggren, G., Jonsson, P.V., DuPaquier,
J.N., & Schroll, M., (1997). Efforts to establish the reliability of the RAI.
Age and Aging, 26(S2), 27-30.

Shaller, D. (2005). Consumers in health care: The burden of choice. Oakland,
CA: California HealthCare Foundation.

Siebenaler, K., O'Keefe, J., Brown, D., & O'Keefe, C. (2005). Nursing facility
transition initiatives of the fiscal year 2001 and 2002 grantees: Progress
and challenges. Research Triangle Park, NC: RTI International.

Spence, D.A., Wiener, J.M. (1990). Nursing home length of stay patterns:
Results from the 1985 National Nursing Home Study. The Gerontologist,
30 (1), 16-20.

Stone, R. (2000).Consumer direction in long-term care. Generations, 24(3), 5.

Tucker, A., & Decker, F. H. Using long-term care MDS data: Patient days and
case mix. American Health Care Association, Health Services Research
and Evaluation, Washington, DC.

U.S. Census Bureau, Census 2000 special tabulation; 1990 Census of
Population, Nursing Home Population: 1990 (CPH-L-137).

U.S. Department of Health and Human Services (2000). Healthy People 2010.
Washington, DC: U.S. Government Printing Office.

Wayne, S.J., Rhyne, R.L., Thompson, R.E. & Davis, M. (1991). Sampling
issues in nursing home research. Journal of American Geriatrics Society,
39, 308-311.

133
Williams, L. (2000). Long-term care after Olmsted v. L.C.: Will the potential of
the ADA's integration mandate be achieved? Journal of Contemporary
Health Law Policy, 17, 205-239.

Wolinsky F. D. (1990). Health and health behavior among elderly Americans:
An age-stratification. Springer, New York.

Wooldridge J., Brown, R., Foster, L., Hoag, S., Irvin, C., Kane, R.L., Newcomer,
R., Schneider, B., and Smith, K.D. (2001). “Social Health Maintenance
Organizations: Transition into Medicare + Choice.” Mathematica Policy
Research, Inc. Princeton, NJ
134
APPENDIX: SAS CODE FOR EPISODE CALCULATION

data mds1;
set raw_data_file;
format entryd mmddyy10. exitd mmddyy10. doi mmddyy10. datesort 1.
entryyr 4. exityr 4. entrymo 2. exitmo 2. entrydt 2. exitdt 2.;
if aa8a in ("01","06","07","08","09");
if aa8a = "01" AND ab1 < '20001201' AND a4a = '********' then delete;
if aa8a = "01" and a4a ne "********" then entry_doi = a4a;
else if aa8a = "01" and a4a = "********" then entry_doi = ab1;
else if aa8a = "08" then entry_doi = ab1;
else if aa8a = "09" then entry_doi = a4a;
else entry_doi = "";
if entry_doi = "********" then entry_doi = "";
if aa8a = "06" then exit_doi = r4;
else if aa8a = "07" then exit_doi = r4;
else if aa8a = "08" then exit_doi = r4;
else exit_doi = "";
if exit_doi = "********" then exit_doi = "";
entryyr = input(substr(entry_doi,1,4),4.);
entrymo = input(substr(entry_doi,5,2),2.);
entrydt = input(substr(entry_doi,7,2),2.);
entryd = mdy(entrymo,entrydt,entryyr);
exityr = input(substr(exit_doi,1,4),4.);
exitmo = input(substr(exit_doi,5,2),2.);
exitdt = input(substr(exit_doi,7,2),2.);
exitd = mdy(exitmo,exitdt,exityr);
if aa8a = "08" then do;
btwdate = exitd - entryd;
if btwdate gt 14 then entryd = exitd - 14;
end;
if exitd then doi = exitd;
else doi = entryd;
if aa8a in ("06","07") then datesort = 1;
else if aa8a = "08" then datesort = 2;
else if aa8a = "01" then datesort = 3;
else if aa8a = "09" then datesort = 4;
data mds1sort;
set mds1;
proc sort data= mds1sort;
by mdsstid doi datesort;
run;

data mds2 (keep = mdsstid epi_beg epi_end shortbounce gap);
set mds1sort;
by mdsstid doi datesort;
format epi_beg mmddyy10. epi_end mmddyy10. shortbdate mmddyy10.
retain epi_beg epi_end shortbounce shortbdate gap;
if first.mdsstid then do;
epi_beg = .;
epi_end = .;
shortbdate = .;
shortbounce = 0;
end;

* Scenario 1;
if epi_beg = . AND datesort = 3 AND entryd lt ("31dec2003"d - 120) then
do;
epi_beg = entryd;
shortbdate = entryd;
end;
* Scenario 2;
if epi_beg AND (shortbdate gt epi_end OR epi_end = .) AND datesort = 1
then do;
epi_end = exitd;
end;
* Scenario 3;
if epi_beg AND epi_end AND exitd gt epi_end AND shortbdate le epi_end AND
datesort = 2 then do;

135
gap = max(entryd - epi_end,0);
if gap le 30 then do;
shortbounce = shortbounce + 1;
end;
if gap gt 30 then do;
output;
epi_beg = .;
epi_end = .;
shortbounce = 0;
end;
end;
* Scenario 4;
if epi_beg AND epi_end = . AND datesort = 2 then do;
epi_end = exitd;
end;
* Scenario 5;
if epi_beg AND epi_end AND datesort = 4 then do;
gap = entryd - epi_end;
if shortbdate le epi_end AND gap le 30 then do;
if entryd ne shortbdate then shortbounce = shortbounce + 1;
end;
if shortbdate le epi_end AND gap gt 30 then do;
output;
epi_beg = .;
epi_end = .;
shortbounce = 0;
end;
shortbdate = entryd;
end;
* Scenario 6;
if epi_beg AND epi_end AND datesort = 3 then do;
gap = entryd - epi_end;
if shortbdate le epi_end AND gap le 30 then do;
if entryd ne shortbdate then shortbounce = shortbounce + 1;
end;
if shortbdate le epi_end AND gap gt 30 then do;
output;
epi_end = .;
shortbounce = 0;
if entryd lt ("31dec2003"d - 120) then epi_beg = entryd;
else epi_beg = .;
end;
shortbdate = entryd;
end;
* Scenario 7;
if epi_beg AND last.mdsstid AND datesort in (3,4) then do;
epi_end = "01jan2004"d;
output;
end;
* Scenario 8;
if epi_end AND last.mdsstid AND datesort in (1,2) then output;

run;

Abstract (if available)

Abstract This dissertation contributes to the larger policy objective of rebalancing the long-term care system in the United States. Nursing home transition is one of several strategies that states are using in order to reduce long-term care spending, be responsive to consumer preferences and comply with the Olmstead Decision (Kasper and O’Malley, 2006).

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Residential care in Los Angeles: policy and planning for an aging population

PDF

Forensic markers of physical elder abuse in medical and community contexts: implications for criminal justice interventions

Asset Metadata

Creator Thomas, Kathryn E. (author)

Core Title Nursing home transitions: a new framework for understanding preferences, barriers and outcomes

School Leonard Davis School of Gerontology

Degree Doctor of Philosophy

Degree Program Gerontology

Publication Date 08/07/2009

Defense Date 05/12/2009

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

Tag community discharge,community living preference,long-term care rebalancing,LTC rebalancing,MDS,minimum data set,nursing facility transition,nursing home diversion,nursing home transition,OAI-PMH Harvest,preference,S/HMO,SCAN,SHMO,social health maintenance organization,social HMO

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Wilber, Kathleen H. (committee chair), Enguidanos, Susan M. (committee member), Myrtle, Robert C. (committee member)

Creator Email katythomas@gmail.com,kethomas@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2522

Unique identifier UC1439253

Identifier etd-Thomas-3115 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-180324 (legacy record id),usctheses-m2522 (legacy record id)

Legacy Identifier etd-Thomas-3115.pdf

Dmrecord 180324

Document Type Dissertation

Rights Thomas, Kathryn E.

Type texts

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

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu