The Gerontologist
Vol. 48, No. 4, 495–504

Copyright 2008 by The Gerontological Society of America

Postdischarge Environmental and
Socioeconomic Factors and the Likelihood of
Early Hospital Readmission Among
Community-Dwelling Medicare Beneficiaries
Alicia I. Arbaje, MD, MPH,1,2 Jennifer L. Wolff, PhD,2,3 Qilu Yu, PhD,2
Neil R. Powe, MD, MPH, MBA,4 Gerard F. Anderson, PhD,3
and Chad Boult, MD, MPH, MBA3
Purpose: This study attempts to determine the associations between postdischarge environmental (PDE)
and socioeconomic (SES) factors and early readmission to hospitals. Design and Methods: This study
was a cohort study using the 2001 Medicare Current
Beneficiary Survey and Medicare claims for the
period from 2001 to 2002. The participants were
community-dwelling Medicare beneficiaries admitted
to hospitals, discharged home, and surviving at least
1 year after discharge (n = 1,351). The study
measurements were early readmission (within 60
days), PDE factors, and SES factors. PDE factors
consisted of having a usual source of care, requiring
assistance to see the usual source of care, marital
status, living alone, lacking self-management skills,
having unmet functional need, having no helpers with

We thank the Robert Wood Johnson Foundation, Lipitz Center for
Integrated Care at the Johns Hopkins Bloomberg School of Public
Health, and the Division of Geriatric Medicine and Gerontology at the
Johns Hopkins School of Medicine for providing funding for this
research. We also thank the faculty, fellows, and staff of the Robert
Wood Johnson Clinical Scholars Program at Johns Hopkins School of
Medicine, including Marie Diener-West, PhD, and Leon Gordis, MD,
DrPH, who were instrumental to the completion of the study. Robert
Herbert and Wenke Hwang, PhD, provided excellent assistance with
data management. Finally, we thank Eric A. Coleman, MD, MPH; Linda
P. Fried, MD, MPH; Wendy W. Sanders, MA; and Timothy A. Zeffiro,
MPH for their valuable guidance and expertise.
Address correspondence to Alicia I. Arbaje, MD, MPH, Assistant
Professor and Associate Director of Transitional Care Research,
Division of Geriatric Medicine and Gerontology, Department of
Medicine, Johns Hopkins University School of Medicine, 5505 Hopkins
Bayview Circle, Baltimore, MD 21224. E-mail: aarbaje@jhmi.edu
1
Division of Geriatric Medicine and Gerontology, Johns Hopkins
University School of Medicine, Baltimore, MD.
2
Center on Aging and Health, Johns Hopkins University School of
Medicine, Baltimore, MD.
3
Department of Health Policy and Management, Johns Hopkins
University Bloomberg School of Public Health, Baltimore, MD.
4
Division of General Internal Medicine and Welch Center for
Prevention, Epidemiology and Clinical Research, The Johns Hopkins
Medical Institutions, Baltimore, MD.

Vol. 48, No. 4, 2008

495

activities of daily living, number of living children,
and number of levels in the home. SES factors consisted of education, income, and Medicaid enrollment. Results: Of the 1,351 beneficiaries, 202
(15.0%) experienced an early readmission. After
adjustment for demographics, health, and functional
status, the odds of early readmission were increased
by living alone (odds ratio or OR = 1.50, 95%
confidence interval or CI = 1.01–2.24), having unmet
functional need (OR = 1.48, 95% CI = 1.04–2.10),
lacking self-management skills (OR = 1.44, 95% CI =
1.03–2.02), and having limited education (OR =
1.42, 95% CI = 1.01–2.02). Implications: These
findings suggest that PDE and SES factors are associated with early readmission. Considering these
findings may enhance the targeting of pre-discharge
and postdischarge interventions to avert early readmission. Such interventions may include home health
services, patient activation, and comprehensive discharge planning.
Key Words: Care transitions, Cohort study,
Discharge planning, Socioeconomic factors,
Unmet need

The transition from hospital to home is a vulnerable period for hospitalized older patients, and some
return to the hospital soon after discharge (Benioff &
Lazowski, 2004; Coleman, Min, Chomiak, &
Kramer, 2004; Murtaugh & Litke, 2002). Providers,
patients, and caregivers often do not coordinate
services or communicate effectively during this critical transition phase (Anderson, Herbert, Zeffiro, &
Johnson, 2004; Bates et al., 1997; Benioff &
Lazowski), resulting in poor patient satisfaction,

adverse events, and early readmission (Bates et al.;
Benioff & Lazowski; Boockvar et al., 2004; Forster,
Murff, Peterson, Gandhi, & Bates, 2003, 2005).
Although early readmission does not always result
from suboptimal care transitions, observational and
intervention studies demonstrate a strong association
(Benioff & Lazowski; Coleman, Parry, Chalmers, &
Min, 2006; Coleman, Smith, Raha, & Min, 2005;
Moore, Wisnivesky, Williams, & McGinn, 2003;
Phillips et al., 2004; Rich et al., 1995; Van Walraven,
Seth, Austin, & Laupacis, 2002; Vinson, Rich,
Sperry, Shah, & McNamara, 1990). Older adults
are at particular risk for early readmission because
they have more chronic conditions, health care providers, complex medical regimens, hospitalizations,
and transitions in care (Anderson et al.; Soeken,
Prescott, Herron, & J, 1991). Compared with transitions to other facilities, transitions to home are associated with a 9% to 20% increased risk of early
readmission or visit to the emergency department
(Corrigan & Martin, 1992; Lockery, Dunkle, Kart, &
Coulton, 1994; Murtaugh & Litke). This suggests
that factors in the home environment may contribute
to a patient’s early readmission.

and PDE factors that are associated with health care
utilization. In long-term-care settings, factors include
having low SES (Mendoza-Sassi & Beria, 2001),
living alone (Linden, Horgas, Gilberg, & SteinhagenThiessen, 1997; Lledo et al., 1997), receiving limited
social support (Hessel, Gunzelmann, Geyer, &
Brahler, 2000), having unmet functional need (Gaugler,
Kane, Kane, & Newcomer, 2005), and experiencing
increased caregiver stress (Gaugler et al.; Jette,
Tennstedt, & Crawford, 1995; Shyu, Chen, & Lee,
2004; Wolff & Kasper, 2004). Specific to early readmission, one study found that 76% of preventable
admissions for congestive heart failure were related
to inadequacies in discharge planning, follow-up,
social support, or patients’ self-management skills
(Vinson et al., 1990). Other studies have found that
lack of education (Marcantonio et al., 1999), low
SES (Coleman et al., 2004; Weissman, Stern, &
Epstein, 1994), and discharge destination (Kind
et al., 2007) are associated with early readmission.
To our knowledge, prior studies of PDE and SES
factors have not examined their effects on transitional care, early readmission, or the older adult
population.

Previous Studies on Factors Associated
With Early Readmission

Importance of Further Study

Early readmission is important in the Medicare
population because readmissions comprise almost
10% of Medicare inpatient hospital expenditures
(U.S. Government Printing Office, 2006). Previous
studies have identified several demographic and
health-related factors that predispose patients to
early readmission: increasing age, male gender,
diagnosis of congestive heart failure or diabetes,
severe illness, visual impairment, prolonged length
of stay, and high prior utilization (Anderson &
Steinberg, 1984; Boult et al., 1993; Burns & Nichols,
2003; Coleman et al., 2004; Corrigan & Martin,
1992; Gooding & Jette, 1985; Hoskins, WaltonMoss, Clark, Schroeder, & Thiel, 1999; Kind, Smith,
Frytak, & Finch, 2007; Soeken et al., 1991). All but
two of these studies (Anderson & Steinberg; Boult
et al.) were limited to patients from a single institution or region of the country. Five studies were
focused either on patients with a limited number of
diseases (Gooding & Jette; Hoskins et al.; Kind et al.)
or on patients that were not elderly (Corrigan &
Martin; Soeken et al.).
Previous Studies on Postdischarge Environmental
and Socioeconomic Factors
Postdischarge environmental (PDE) factors include characteristics of the home and caregiving
environments of patients after they have been
discharged from hospitals. A growing body of
evidence has identified socioeconomic status (SES)
496

The American Geriatrics Society has called for
research to identify persons at risk for complications
from suboptimal transitional care (Coleman, 2003).
Many studies assessing risk factors for early
readmission have been limited to local or younger
populations with a single disease, or they have
focused on a limited number of PDE factors. It is
important to policymakers, payers, providers, and
patients to have information to better target known
strategies that reduce early readmission, such as
interventions developed by Rich and colleagues
(1995), Naylor and colleagues (1999), and Coleman
and colleagues (2006).
Our objective in this study was to determine the
associations between PDE and SES factors and the
likelihood of Medicare beneficiaries’ early readmission to hospitals. On the basis of the existing
literature, we hypothesized that several factors
would negatively influence patients’ abilities to
remain at home after discharge: (a) having no usual
source of care (USOC; limited access to care); (b)
requiring assistance to see one’s USOC (barrier to
access to care); (c) having a particular marital status
and living alone (less social support or access to
caregivers); (d) lacking self-management skills (limited ability to manage one’s own illnesses and
implement a postdischarge regimen); (e) having
unmet functional need (in terms of activities of daily
living, known as ADLs, or instrumental activities of
daily living, known as IADLs; limited availability of
assistance); (f) lacking ADL helpers or living children
(less access to caregivers); (g) living in multilevel
The Gerontologist

Figure 1. Conceptual model.

homes (physical spaces more difficult to navigate);
(h) having limited education (decreased patient
activation or limited ability to implement a postdischarge regimen); (i) having low income (limited
access to care); or (j) being enrolled in the Medicaid
program (low income but having access to care).
The conceptual model guiding our analyses is
displayed in Figure 1. Hypothesized relationships
between the key constructs of interest, that is,
‘‘socioeconomic factors and postdischarge environment’’ (shown in bold), in relation to care transitions
and early readmission are graphically depicted with
bold arrows. Boxes containing constructs labeled
‘‘health status’’ and ‘‘demographics’’ represent
covariates included in the analyses. These four
constructs collectively are hypothesized as influential
to early readmission during a care transition by
affecting access to care, social support, availability of
assistance, and implementation of the postdischarge
regimen. The arrows emerging from the ‘‘hospital
organizational factors’’ and ‘‘other hospital characteristics’’ boxes represent variables that are unmeasured in this study, but that may also be
important in understanding early readmission in
community-dwelling older adults.
Methods
Study Design and Setting
We conducted a retrospective cohort analysis of
community-dwelling respondents to the Medicare
Current Beneficiary Survey (MCBS) in 2001 who,
over 1-year follow-up, were admitted to U.S. acute
care hospitals during the period from 2001 to 2002.
The MCBS is conducted on a nationally representative sample of Medicare beneficiaries and consists of
Vol. 48, No. 4, 2008

497

quarterly, in-person interviews on a wide range of
sociodemographic and health topics. The response
rate for the MCBS averages 85% to 89% per wave.
Details of the MCBS are described elsewhere
(Centers for Medicare & Medicaid Services, 2006).
We linked Medicare claims for hospital care during
2001 to 2002 with MCBS survey data to ascertain
beneficiaries’ hospitalization experiences, including
nonelective readmissions in the year following initial
admission. Permission to use the Access to Care files
and to publish these results was granted under CMS
Data Use Agreement number 12066. The Institutional
Review Board of the Johns Hopkins Bloomberg
School of Public Health approved this study (protocol number H.30.04.05.10.AX).

Participants
For this study, our sample consisted of communitydwelling beneficiaries (all ages) who participated in
the 2001 round of the MCBS, were continuously
enrolled in Medicare throughout the calendar year,
were hospitalized during the period from 2001 to
2002, and were discharged home (N = 1,351). To
identify the study sample, we began with 16,461
beneficiaries interviewed in 2001 and excluded those
respondents who did not meet the selection criteria
(Figure 2): 9,936 individuals (60.4%) either were not
hospitalized or were hospitalized but were not discharged back to the community; 3,391 (20.6%)
lacked at least 1 year of claims data because they
rotated out of the 2002 MCBS; and 1,783 (10.8%)
lacked at least 1 year of claims data because they
were lost to follow-up or died between 2001 and
2002. Not all 1,783 who died or were lost to follow-up
were lost during the study period, and the exact

Figure 2. Sample selection; MCBS = Medicare Current Beneficiary Survey. (Note that the superscript a denotes an estimated value,
based on the reported average response rate of 85% in the MCBS; see the Centers for Medicare & Medicaid Services, 2006.)

number is not obtainable from our data (the Access
to Care MCBS files exclude decedents and do not
distinguish between decedents and persons lost to
follow-up).

except for length of stay (LOS) of the index admission, which we obtained from hospital claims.
Demographics, Health, and Functional Ability

Outcome Measure
The primary outcome, early readmission, was the
occurrence of a nonelective hospital readmission
within 60 days of discharge from the index
admission. To determine the index admission, we
searched Medicare hospital claims for the first
admission occurring after the participant completed
the fall interview of the 2001 MCBS. The search for
an index admission covered a maximum of 305 days
after each participant’s 2001 interview date in order
to allow at least 60 days of postdischarge observation for early readmission. Once we found an index
admission, then we searched Medicare hospital
claims until December 31, 2002 for the first
hospitalization after discharge from the index
admission. If individuals incurred more than one
admission after discharge, then we included only the
first readmission within 60 days in our analysis. We
excluded those admissions classified in the administrative claims data as ‘‘elective.’’
Selection of Variables
For inclusion in the study, we considered PDE and
SES characteristics identified from the literature as
being relevant to early readmission or hypothesized
as negatively influencing a person’s ability to remain
at home after discharge (Table 1). For inclusion as
covariates, we considered demographic, health, and
functional variables shown to be related to readmissions in prior studies. Data for all independent
variables originated from responses to the MCBS,
498

Demographic factors included age (years), gender,
minority status (non-White race or of Hispanic
origin), and living in a metropolitan area (yes or no).
Participants rated their general health status compared with others on a 5-point scale (from 1,
excellent, to 5, poor). We classified participants as
having a disability if they reported having difficulty
with any of six ADLs or six IADLs, or if they did not
perform the activity because of a health problem. We
classified participants as having memory loss if they
reported that memory loss interfered with daily
activity. Number of chronic conditions summarized
respondents’ having been told by a doctor they had
any of 17 chronic conditions. We categorized the
duration of the index admission LOS into tertiles.
PDE Factors
We classified participants as requiring assistance
to see their USOC if, instead of walking or driving
themselves, they either relied on assistance from
another person or used another mode of transportation. Participants indicated their marital status
and the number of people in their household; we
used the latter item to identify those participants
living alone. We classified participants as lacking
self-management skills if they indicated no confidence in their ability to perform each of four tasks:
(a) identifying when medical care was needed; (b)
identifying medication side effects; (c) following
self-care instructions; and (d) changing habits as
recommended. Those with unmet functional need
reported that they had difficulty with an ADL or
The Gerontologist

Table 1. Characteristics of the Study Population
All Patients

Patients Readmitted

Not Readmitted

p

Demographics
Age
Female
Minority
Lives in nonmetropolitan area

72.6 (0.3)
55.8
19.7
28.7

72.5 (0.8)
55.1
19.2
33.5

72.6 (0.3)
55.9
18.6
28.0

.94
.84
.85
.14

Health and functional ability
General health status
Any disability
Presence of memory loss
No. of chronic conditions

3.2 (0.2)
60.2
15.0
3.4 (0.1)

3.4 (0.9)
70.8
14.8
3.8 (0.2)

3.2 (0.4)
58.6
15.1
3.4 (0.1)

.01
.00
.93
.01

76.9
19.3
3.8

67.8
25.8
6.5

78.4
18.2
3.4

3.2
37.1

4.7
43.0

3.0
36.1

Marital status
Married
Widowed
Divorced, separated, or never married

48.7
35.1
16.3

38.6
38.7
27.7

50.2
34.5
15.2

Lives alone
Lacking self-management skills
Any unmet functional need
Has no helpers with ADLs
No. of living children
Lives in multilevel home

32.4
20.2
43.6
37.1
3.0 (0.1)
41.2

39.7
27.2
55.2
34.7
2.9 (0.2)
44.8

31.2
19.1
41.7
37.4
3.1 (0.1)
40.6

.04
.01
.00
.49
.44
.25

37.6
66.0
20.5

46.7
71.7
26.4

36.2
65.0
19.5

.00
.12
.03

Characteristic (%)

Index admission LOS (days)
1–6
7–14
!15
Postdischarge environmental factors
USOC
No USOC
Requires assistance to see USOC

Socioeconomic factors
No high school diploma
Income , $25,000
Medicaid enrollment

.01

.13

.01

Notes: The study population numbers are as follows: all patients, N = 1,351; readmitted, n = 202; not readmitted, n = 1.149.
The p value shows a comparison, using a chi-square or t test, between patients who are readmitted and those who are not.
Numbers shown along with a value in parentheses signify the mean (standard deviation). General health status is listed from
1 (excellent) to 5 (poor). LOS = length of stay; USOC = usual source of care; ADL = activity of daily living. Having no helpers
with ADLs and number of living children are alternative measures of unmet functional need.

IADL and they lacked assistance or supervision with
any ADL disability (difficulty walking, eating,
bathing, dressing, transferring from a bed to a chair,
or using the toilet) or IADL disability (difficulty
using the telephone, preparing meals, performing
housework, shopping, or managing finances). We
classified participants as having any helpers with
ADLs (yes or no), and they reported the number of
living children they had. Finally, we categorized
participants as whether they lived in multilevel or
single-level homes.

SES Factors
We categorized education according to whether
or not participants had high school diplomas. We
categorized participants’ annual income on the basis
of a cutoff point of $25,000. We categorized
Vol. 48, No. 4, 2008

499

participants as enrolled in the Medicaid program
or not. We considered both income and Medicaid
enrollment because these variables may have opposite effects on health care utilization; earning a low
income presents a barrier to receiving prompt care,
whereas Medicaid enrollment may facilitate access
to care for persons earning low income.

Analysis
We examined the strength of each variable’s
bivariate relationship with early readmission by
using simple logistic regression (Table 2, first
column). We then entered all variables into a multivariate model of early readmission (Table 2, second
column). To create the final model (Table 2, third
column), we discarded those variables that were
either conceptually similar to other measures within

Table 2. Logistic Model of the Relationship Between Early Readmission and PDE and SES Factors
Characteristic (%)

Unadjusted OR (95% CI)

Adjusted OR (95% CI) Full Model

Final Model

Demographics
Age
Female
Minority
Lives in nonmetropolitan area

1.00
0.97
1.04
1.30

(0.99–1.01)
(0.70–1.34)
(0.69–1.58)
(0.91–1.86)

1.01
0.88
0.92
1.40

(0.99–1.02)
(0.60–1.28)
(0.57–1.49)
(0.95–2.06)

1.00
0.89
0.95
1.27

Health and functional ability
General health status
Any disability
Presence of memory loss
No. of chronic conditions

1.22
1.72
0.98
1.11

(1.06–1.41)
(1.20–2.45)
(0.61–1.57)
(1.03–1.19)

1.11
1.14
0.73
1.07

(0.94–1.31)
(0.67–1.95)
(0.44–1.18)
(0.98–1.16)

1.13 (0.98–1.31)

1.63 (1.12–2.38)
2.20 (1.0–4.59)

1.57 (1.08–2.29)
1.92 (0.89–4.14)

1.60 (1.09–2.36)
1.96 (0.97–3.95)

1.56 (0.72–3.40)
1.38 (0.96–1.98)

1.49 (0.65–3.40)
1.11 (0.73–1.69)

Marital status
Married (reference)
Widowed
Divorced, separated, or never married

1.46 (1.07–1.98)
1.94 (1.22–3.07)

1.21 (0.75–1.95)
1.64 (0.90–3.00)

Lives alone
Lacking self-management skills
Any unmet functional need
Has no helpers with ADLs
No. of living children
Lives in multilevel home

1.45
1.58
1.72
1.15
0.97
1.19

1.17
1.35
1.32
0.79
0.97
1.28

Index admission LOS (days)
1–6 (reference)
7–14
!15
PDE factors
USOC
No USOC
Requires assistance to see USOC

SES factors
No high school diploma
Income , $25,000
Medicaid enrollment

(1.01–2.07)
(1.12–2.22)
(1.27–2.34)
(0.96–1.37)
(0.91–1.04)
(0.89–1.60)

1.54 (1.16–2.05)
1.36 (0.92–2.01)
1.48 (1.04–2.09)

(0.99–1.01)
(0.62–1.28)
(0.60–1.51)
(0.89–1.80)

0.76 (0.47–1.24)

(0.69–1.97)
(0.93–1.98)
(0.87–1.99)
(0.51–1.22)
(0.90–1.05)
(0.94–1.75)

1.50 (1.01–2.24)
1.44 (1.03–2.02)
1.48 (1.04–2.10)

1.52 (1.07–2.17)
0.87 (0.57–1.35)
1.02 (0.62–1.69)

1.43 (1.01–2.02)
0.94 (0.61–1.46)

Notes: The number for the table is n = 1,343. PDE = postdischarge environmental; SES = socioeconomic; OR = odds ratio;
CI = confidence interval; LOS = length of stay; USOC = usual source of care; ADL = activity of daily living. Having no helpers
with ADLs and number of living children are alternative measures of unmet functional need. For demographics, all demographic
variables are deemed as conceptually distinct and kept in the final model. For health and functional ability, we removed disability
and number of chronic conditions: they were conceptually similar to and correlated with general health status by use of chi-square
and t test, respectively (p , .001 for both relationships). For USOC variables, we removed them because they correlated with
unmet functional need by use of chi-square (p , .001). We removed items for marital status, has no helpers with ADLs, and number of living children: they were conceptually similar to and correlated with the lives-alone item by use of chi-square (p , .001 for
all three relationships). We removed the item for lives in multilevel home because it was not significant in bivariate analyses. For
SES factors, we removed Medicaid enrollment: it correlated with low income by use of chi-square (p , .001).

each of the four constructs of interest shown in our
conceptual model in Figure 1, highly correlated using
chi-square or t test, or not significant in bivariate
analyses (see the footnotes to Table 2). We deemed
all demographic variables to be conceptually distinct
and kept them in the final model. We removed
‘‘disability’’ and ‘‘number of chronic conditions’’
because they were conceptually similar to and
correlated with ‘‘general health status’’ by use of
chi-square and t test, respectively (p , .001 for both
relationships). We removed USOC variables because
they were correlated with ‘‘unmet functional need’’
by use of chi-square (p , .001).
We removed ‘‘marital status,’’ ‘‘has no helpers
with ADLs,’’ and ‘‘number of living children’’
500

because they were conceptually similar to and
correlated with ‘‘lives alone’’ by use of chi-square
(p , .001 for all three relationships). We removed
‘‘lives in multilevel home’’ because it was not
significant in bivariate analyses, and we removed
‘‘Medicaid enrollment’’ because it was correlated
with low income using chi-square (p , .001). We
used STATA statistical software, version 8.0, for all
data analyses (StataCorp, College Station, TX). We
used cross-sectional weights for each beneficiary to
reflect the overall selection probability for each
person and to account for the complex sampling
strategy of the MCBS (Centers for Medicare &
Medicaid Services, 2006). The 1,351 persons in the
sample population in this study represent 2,883,726
The Gerontologist

Medicare beneficiaries nationwide who were dwelling in the community, hospitalized, discharged
home, and then survived at least 1 year after
discharge.
Results
Of the analytic sample of 1,351 hospitalized
beneficiaries, 202 (15.0%) were readmitted within
60 days of discharge (Table 1). In bivariate analyses,
readmitted persons reported significantly worse
health status, more disability, more chronic conditions, and longer index admission LOS (Table 2).
Regarding PDE factors, those patients experiencing
early readmission were more likely to be unmarried,
live alone, lack self-management skills, and have
unmet functional needs. Regarding SES factors,
readmitted persons were more likely to have limited
education and to be enrolled in the Medicaid
program.
In the multivariate logistic regression model of
early readmission that adjusted for demographics,
health, and functional status, persons who lived
alone had a 50% increased odds of early readmission
compared with those who did not live alone (Table
2, final model). Those who reported having any
unmet functional need had a 48% increased odds of
early readmission compared with those who did not
report any unmet IADL need. Persons lacking selfmanagement skills had a 44% increased odds of
early readmission compared with those who did not
lack self-management skills. Having limited education was associated with a 42% increased odds of
early readmission. Earning a low income did not
demonstrate a statistically significant association
with early readmission.
Of the covariates, index admission LOS of 7 to
14 days and self-reported general health status were
associated with increased odds of early readmission
(odds ratio or OR = 1.60, 95% confidence interval
or CI = 1.09–2.36 and OR = 1.13, 95% CI = 0.98–
1.31, respectively). Index admission LOS of 15 or
more days showed a trend toward association with
early readmission.
Sensitivity Analyses
We conducted four sensitivity analyses to assess
the possibility that these results were affected by the
definitions of certain independent variables. We
recomputed multivariate analyses by using different
definitions of unmet functional need and health
status. We limited the definition of unmet functional
need to lack of direct assistance from a person,
rather than lack of direct assistance or supervision
from a person, as in the original definition; unmet
functional need remained significantly associated
with early readmission. There was no significant
change in the main findings of the study when we
Vol. 48, No. 4, 2008

501

substituted participants’ Charlson comorbidity indices (Deyo, Cherkin, & Ciol, 1992) for their selfreported general health status.
To investigate the possible effects of multicollinearity, we computed the variance inflation factor
(VIF) for each variable in the final model to provide
an estimate of the number of times each variable’s
variance was increased by multicollinearity. Using
a conservative VIF threshold greater than 2.5 to
determine multicollinearity in logistic regression
models (rather than the usual threshold of 10; see
Multicollinearity in Logistic Regression, 2007), we
found that there did not appear to be significant
multicollinearity in the final model; the mean VIF in
this analysis was 1.11 (range = 1.02–1.24). When
converted to tolerance estimates (tolerance = 1/VIF),
the results conveyed that greater than 80% of each
variable’s variance was independent of all the other
variables in the final model.
To test the assumption that important independent variables remained constant during the year of
follow-up, we compared participants’ responses in
2001 with those in 2002. From 2001 to 2002, there
were no major changes in participants’ live-alone
status or unmet functional need. There was no
inquiry about self-management ability in the 2002
round of the MCBS, so we could not track changes
in this measure.
Finally, to test the possibility that using a composite measure of the total number of chronic conditions as a covariate might have masked relationships
of specific conditions to early readmission, we
examined the binary and multivariate associations
between early readmission and each of the 17 individual chronic conditions in the MCBS. The main
findings of the study did not change with the addition of either an individual chronic condition or
a composite of chronic conditions found to be significant or nearly significant in bivariate analyses.

Discussion
In this study we examined the relevance of PDE
and SES factors to early readmission among a
national sample of community-dwelling Medicare
beneficiaries who were hospitalized, discharged
home, and survived at least 1 year after discharge.
In the study sample, representing almost 3 million
community-dwelling beneficiaries, we found that
living alone, having unmet functional need, lacking
self-management skills, or having limited education
was associated with an increased likelihood of early
readmission. Our study findings support the hypothesized relationships regarding PDE and SES as
influential factors in the quality of transitional care
that is delivered before, during, and after a care
transition (Figure 1).
Our purpose in this study was to identify
screening criteria that institutions, payers, and

providers could consider in identifying hospitalized
older adults at risk for experiencing early readmission. The findings are consistent with previous work
that has demonstrated associations between living
alone, (Linden et al., 1997; Lledo et al., 1997), having
unmet functional need, (Gaugler et al., 2005), having a lower level of education, (Marcantonio et al.,
1999; Mendoza-Sassi & Beria, 2001), lacking selfmanagement skills (Vinson et al., 1990), and general
health care utilization. Our findings contribute
additional information about the importance of
PDE and SES factors specifically to early readmission
in the community-dwelling older adult population
nationwide. In this study we also used the definition
of early readmission within 60, rather than 30, days
to reflect readmission resulting from exacerbations
of conditions with longer recovery times. Although it
is more difficult to attribute early readmission within
60 days to transitional care quality, because further
from discharge other factors may play a greater role
in readmission, Coleman demonstrated that improved transitional care from the hospital to the
home is associated with decreased readmission rates
up to 180 days (Coleman et al., 2006).

Interpretation of Findings
As we hypothesized, PDE and SES factors were
related to an increased likelihood of early readmission. Unmet functional need may be associated with
limited availability of assistance, which presents
challenges to implementing a postdischarge regimen,
complicates the care transition, and increases the
risk of early readmission (Figure 1). The findings
demonstrate that having any ADL or IADL need
may be significant for affecting health care utilization. Providing for unmet functional needs (a
modifiable characteristic) may affect the occurrence
of early readmission in community-dwelling older
adults. We found two other studies that showed
a relationship between unmet functional need and
health care utilization; one was limited to a community in Spain (Fernandez-Olano et al., 2006), and the
other was limited to patients with dementia (Gaugler
et al., 2005).
We found that living alone, having limited
education, and lacking self-management skills had
significant associations with early readmission, as we
hypothesized. These associations may represent
limitations in several characteristics that assist older
adults in remaining at home after they have been
discharged from the hospital. For example, living
alone or having limited education in some cases may
be associated with limited social support, restricted
access to caregivers, and limited ability to implement
complicated postdischarge regimens, such as blood
pressure monitoring or anticoagulant administration. Having limited education may also impact the
level of patient activation in medical encounters,
502

which in turn would affect the person’s ability to
manage his or her illness and prevent clinical
deterioration requiring early readmission.
These results should be considered in the context
of several limitations. First, participants’ baseline
PDE and SES characteristics could have changed
during the year of observation. Our comparison of
responses from 2001 with those from 2002 may not
have completely overcome this limitation. Second,
the data set used in these analyses did not include
decedents. Therefore, the study findings are applicable strictly to Medicare beneficiaries who were
hospitalized, discharged home, and survived at least
1 year after discharge. Even if PDE or SES factors
had different relationships to early readmission
among the decedents compared with the survivors,
the absence of data from decedents (n ; 100) is
unlikely to have biased the study’s main findings.
Finally, it is important to interpret the lack of
association between low income and early readmission cautiously, because of statistical power limitations and limited early readmission event rates.
Bivariate power calculations for having low income
revealed that the study sample provided 42% power
to detect differences between those participants who
were readmitted and those who were not. Had the
early readmission rate been higher, then low income
may have had significant associations with early
readmission, and thus could still be important.
Limitations aside, this study has important
strengths. First, the data are nationally representative of community-dwelling Medicare beneficiaries
who have been hospitalized, discharged home, and
survived at least 1 year after discharge. Second, using
Medicare data enables tracking of health care
utilization regardless of the site of health care
delivery, thus capturing most of beneficiaries’ early
readmissions. Third, this study identified new
associations between early readmission and PDE
and SES factors: having unmet functional need,
living alone, having a low educational level, and lacking self-management skills. In contrast to previous
studies (Anderson & Steinberg, 1984; Corrigan &
Martin, 1992; Mendoza-Sassi & Beria, 2001; Soeken
et al., 1991), in this study the measures such as age,
minority status, gender, and income were not significantly associated with readmission. These negative findings may have resulted, in part, from
limitations in statistical power. Alternatively, some
of the measures associated with early readmission in
previous studies may be correlated with some of the
PDE and SES factors included in this analysis.
The proposed study begins to answer the American Geriatrics Society’s call for research on transitional care (Coleman & Boult, 2003) and is of
interest to key stakeholders. Health care organizations and providers may consider PDE and SES
factors in identifying hospitalized patients who
might benefit from more intensive postdischarge
services to reduce early readmission. The findings of
The Gerontologist

this study suggest that PDE and SES factors may be
important for professionals involved in discharge
planning to consider while creating discharge plans
to improve care transitions. Depending on individual
circumstances, such plans may include comprehensive discharge planning (Phillips et al., 2004), patient
activation through transitional care coordinators
(Coleman et al., 2006), self-management programs
(Lorig et al., 1999), caregiver support (Shyu et al.,
2004; Wolff & Kasper, 2004), rehabilitation therapy,
medication reconciliation (Boockvar et al., 2004),
and home health care (Bull, 1994).
Future studies may screen hospitalized older
adults for the PDE and SES factors examined in
this study to select patients to target for intervention
prior to discharge from the hospital. With the use
of PDE and SES screening criteria, specific interventions to be studied may include ensuring caregiver
arrangements or providing home health aides for
patients’ functional disabilities. Other interventions
targeting patients with limited education or lacking
self-management skills may enhance patient activation through home health referrals, participation in
support groups after discharge, or enrollment in
disease-management programs. Interventions to improve the transition of older adults at high risk for
early readmission might also include ensuring direct
communication between inpatient and outpatient
providers to ensure timely follow-up after discharge,
as well as medication reconciliation before and after
discharge.
Institutions and peer review organizations are
interested in measuring and improving the quality of
transitional care, and the Joint Commission on
Accreditation of Healthcare Organizations introduced patient safety goals in 2006 to improve
transitional care (Joint Commission on Accreditation of Healthcare Organizations, 2006). Results of
this study suggest that accommodating PDE and SES
factors may be important in improving quality, that
is, by assisting in difficult caregiving situations,
improving self-management skills, and addressing
health literacy. Future studies of interest to institutions and peer review organizations may include the
development and evaluation of incentive programs
to improve the quality of discharges of older adults,
specifically addressing PDE and SES factors. The
relationships between hospital organizational factors
and early readmission, as shown in Figure 1, also
merit further investigation. Specifically, a high degree of centralization (i.e., vertical and horizontal
integration) within health systems and the provision
of ancillary services important for the care of vulnerable older adults (e.g., case management, palliative care, and transportation) may be associated with
early readmission.
Older adults and their caregivers bear a considerable burden during transitions, and they stand to
benefit from well-targeted, effective interventions.
This study identifies PDE and SES factors that are
Vol. 48, No. 4, 2008

503

associated with early readmission. These findings
contribute to a growing body of evidence that may
help guide the development of interventions to
reduce early readmission and other complicated
transitions experienced by older adults.
References
Anderson, G., Herbert, R., Zeffiro, T., & Johnson, N. (2004). Partnership
for solutions: Better lives for people with chronic conditions. Retrieved
May 1, 2007, from http://www.partnershipforsolutions.org
Anderson, G. F., & Steinberg, E. P. (1984). Hospital readmissions in the
Medicare population. New England Journal of Medicine, 311, 1349–
1353.
Bates, D. W., Spell, N., Cullen, D. J., Burdick, E., Laird, N., Petersen, L. A.,
et al. (1997). The costs of adverse drug events in hospitalized patients.
Journal of the American Medical Association, 277, 307–311.
Benioff, M. R., & Lazowski, E. D. (2004). Revolutionizing health care
through information technology: Report to the President. Washington,
DC: National Coordination Office for Information Technology Research
and Development.
Boockvar, K., Fishman, E., Kyriacou, C. K., Monias, A., Gavi, S., & Cortes,
T. (2004). Adverse events due to discontinuations in drug use and dose
changes in patients transferred between acute and long-term care
facilities. Archives of Internal Medicine, 164, 545–550.
Boult, C., Dowd, B., McCaffrey, D., Boult, L., Hernandez, R., & Krulewitch,
H. (1993). Screening elders for risk of hospital admission. Journal of the
American Geriatrics Society, 41, 811–817.
Bull, M. J. (1994). Use of formal community services by elders and their
family caregivers 2 weeks following hospital discharge. Journal of
Advanced Nursing, 19, 503–508.
Burns, R., & Nichols, L. (2003). Factors predicting readmission of older general
medicine patients. Journal of General Internal Medicine, 6, 389–393.
Centers for Medicare & Medicaid Services. (2006). Retrieved June 4, 2007,
from www.cms.hhs.gov
Coleman, E. A. (2003). Falling through the cracks: Challenges and
opportunities for improving transitional care for persons with continuous
complex care needs. Journal of the American Geriatrics Society, 51,
549–555.
Coleman, E. A., & Boult, C. (2003). Improving the quality of transitional care
for persons with complex care needs. Journal of the American Geriatrics
Society, 51, 556–557.
Coleman, E. A., Min, S.-J., Chomiak, A., & Kramer, A. M. (2004).
Posthospital care transitions: Patterns, complications, and risk identification. Health Services Research, 39, 1449–1466.
Coleman, E. A., Parry, C., Chalmers, S., & Min, S.-J. (2006). The Care
Transitions Intervention: Results of a randomized controlled trial.
Archives of Internal Medicine, 166, 1822–1828.
Coleman, E. A., Smith, J. D., Raha, D., & Min, S.-J. (2005). Posthospital
medication discrepancies: Prevalence and contributing factors. Archives
of Internal Medicine, 165, 1842–1847.
Corrigan, J., & Martin, J. (1992). Identification of factors associated with
hospital readmission and development of a predictive model. Health
Services Research, 27, 81–101.
Deyo, R., Cherkin, D., & Ciol, M. (1992). Adapting a clinical comorbidity
index for use with ICD-9-CM administrative databases. Journal of
Clinical Epidemiology, 45, 613–619.
Fernandez-Olano, C., Hidalgo, J. D. L.-T., Cerda-Diaz, R., RequenaGallego, M., Sanchez-Castano, C., Urbistondo-Cascales, L., et al. (2006).
Factors associated with health care utilization by the elderly in a public
health care system. Health Policy, 75, 131–139.
Forster, A. J., Murff, H. J., Peterson, J. F., Gandhi, T. K., & Bates, D. W.
(2003). The incidence and severity of adverse events affecting patients
after discharge from the hospital. Annals of Internal Medicine, 138,
161–167.
Forster, A. J., Murff, H. J., Peterson, J. F., Gandhi, T. K., & Bates, D. W.
(2005). Adverse drug events occurring following hospital discharge.
Journal of General Internal Medicine, 20, 317–323.
Gaugler, J. E., Kane, R. L., Kane, R. A., & Newcomer, R. (2005). Unmet care
needs and key outcomes in dementia. Journal of the American Geriatrics
Society, 53, 2098–2105.
Gooding, J., & Jette, A. M. (1985). Hospital readmissions among the elderly.
Journal of the American Geriatrics Society, 33, 595–601.
Hessel, A., Gunzelmann, T., Geyer, M., & Brahler, E. (2000). Utilization of
medical services and medication intake of patients over 60 in Germany—
Health related, social structure related, socio-demographic and subjective
factors. Zeitschrift Fur Gerontologie Und Geriatrie: Organ Der
Deutschen Gesellschaft Fur Gerontologie Und Geriatrie, 33, 289–299.

Hoskins, L., Walton-Moss, B., Clark, H., Schroeder, M., & Thiel, L. S.
(1999). Predictors of hospital readmission among the elderly with
congestive heart failure. Home Health Care Nurse, 17, 373–381.
Jette, A. M., Tennstedt, S., & Crawford, S. (1995). How does formal and
informal community care affect nursing home use? Journal of
Gerontology: Social Sciences, 50B, S4–S12.
Joint Commission on Accreditation of Healthcare Organizations. (2006).
2007 Hospital/critical access hospital national patient safety goals. Retrieved
June 4, 2007, from http://www.jointcommission.org/PatientSafety/
NationalPatientSafetyGoals/
Kind, A. J. H., Smith, M. A., Frytak, J. R., & Finch, M. D. (2007). Bouncing
back: Patterns and predictors of complicated transitions 30 days after
hospitalization for acute ischemic stroke. Journal of the American
Geriatrics Society, 55, 365–373.
Linden, M., Horgas, A. L., Gilberg, R., & Steinhagen-Thiessen, E. (1997).
Predicting health care utilization in the very old. The role of physical
health, mental health, attitudinal and social factors. Journal of Aging
and Health, 9, 3–27.
Lledo, R., Martin, E., Jimenez, C., Roca, R., Gil, A., Godoy, E., et al. (1997).
Characteristics of elderly inpatients at high risk of needing supportive
social and health care services. European Journal of Epidemiology, 13,
903–907.
Lockery, S., Dunkle, R., Kart, C., & Coulton, C. (1994). Factors contributing
to the early rehospitalization of elderly people. Health and Social Work,
19, 182–191.
Lorig, K. R., Sobel, D. S., Stewart, A. L., William Brown, B., Bandura, A.,
Ritter, P., et al. (1999). Evidence suggesting that a chronic disease selfmanagement program can improve health status while reducing
hospitalization: A randomized trial. Medical Care, 37, 5–14.
Marcantonio, E. R., McKean, S., Goldfinger, M., Kleefield, S., Yurkofsky,
M., & Brennan, T. A. (1999). Factors associated with unplanned hospital
readmission among patients 65 years of age and older in a Medicare
managed care plan. The American Journal of Medicine, 107, 13–17.
Mendoza-Sassi, R., & Beria, J. U. (2001). Health services utilization: A
systematic review of related factors. Cadernos De Saude Publica/
Ministerio Da Saude, Fundacao Oswaldo Cruz, Escola Nacional De
Saude Publica, 17, 819–832.
Moore, C., Wisnivesky, J., Williams, S., & McGinn, T. (2003). Medical
errors related to discontinuity of care from an inpatient to an outpatient
setting. Journal of General Internal Medicine, 18, 646–651.
Multicollinearity in logistic regression. (2007). Retrieved March
17, 2007, from http://www.uky.edu/ComputingCenter/SSTARS/
MulticollinearityinLogisticRegression.htm

504

Murtaugh, C. M., & Litke, A. (2002). Transitions through postacute and
long-term care settings: Patterns of use and outcomes for a national
cohort of elders. Medical Care, 40, 227–236.
Naylor, M. D., Brooten, D., Campbell, R., Jacobsen, B. S., Mezey, M. D.,
Pauly, M. V., et al. (1999). Comprehensive discharge planning and home
follow-up of hospitalized elders: A randomized clinical trial. Journal of
the American Medical Association, 281, 613–620.
Phillips, C. O., Wright, S. M., Kern, D. E., Singa, R. M., Shepperd, S., &
Rubin, H. R. (2004). Comprehensive discharge planning with postdischarge support for older patients with congestive heart failure: A metaanalysis. Journal of the American Medical Association, 291, 1358–1367.
Revolutionizing health care through information technology: Report to the
President. (2004). Washington, DC: President’s Information Technology
Advisory Committee.
Rich, M., Beckham, V., Wittenberg, C., Leven, C., Freedland, K., & Carney,
R. (1995). A multidisciplinary intervention to prevent the readmission of
elderly patients with congestive heart failure. New England Journal of
Medicine, 333, 1190–1195.
Shyu, Y.-I. L., Chen, M.-C., & Lee, H.-C. (2004). Caregiver’s needs as
predictors of hospital readmission for the elderly in Taiwan. Social
Science and Medicine, 58, 1395–1403.
Soeken, K., Prescott, P., Herron, D., & J. C. (1991). Predictors of hospital
readmission. A meta-analysis. Evaluation and the Health Professions,
14, 262–281.
U.S. Government Printing Office. (2006). Older Americans 2006: Key
indicators of well-being. (2006). Washington, DC: Author.
Van Walraven, C., Seth, R., Austin, P. C., & Laupacis, A. (2002). Effect of
discharge summary availability during postdischarge visits on hospital
readmission. Journal of General Internal Medicine, 17, 186–192.
Vinson, J., Rich, M., Sperry, J., Shah, A., & McNamara, T. (1990). Early
readmission of elderly patients with congestive heart failure. Journal of
the American Geriatrics Society, 38, 1290–1295.
Weissman, J., Stern, R., & Epstein, A. (1994). The impact of patient
socioeconomic status and other social factors on readmission: A
prospective study in four Massachusetts hospitals. Inquiry, 31, 163–172.
Wolff, J. L., & Kasper, J. D. (2004). Informal caregiver characteristics and
subsequent hospitalization outcomes among recipients of care. Aging
Clinical and Experimental Research, 16, 307–313.
Received June 28, 2007
Accepted October 10, 2007
Decision Editor: William J. McAuley, PhD

The Gerontologist