Professional Case Management Vol. 19 , No. 6 , 255 - 262.docx

Professional Case Management
Vol. 19 , No. 6 , 255 - 262
Copyright 2014 © Wolters Kluwer Health | Lippincott Williams
& Wilkins
Vol. 19/No. 6 Professional Case Management 255
Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized
reproduction of this article is prohibited.
CE 2.0 ANCCContact Hours
A B S T R A C T
Purpose of Study: In response to the U.S. Affordable Care
Act, the Centers for Medicare & Medicaid Services
proposed a change in reimbursement penalties for hospitals
beginning October 1, 2012. Reducing the
occurrence of unplanned readmissions has become a more
urgently focused topic. As part of the health care
system, care management aligns with physicians to signifi
cantly improve service, fi nancial, and clinical care
outcomes. To address the changing health care climate in 2008,
care management services were restructured
at an academic university medical center located in 1 of the 3
largest counties in California. Changing from
a unit-based to a service-based care management model
partnered care managers and social workers with
physician services. We sought to assess the effect of this change
on surrogates for patient experience and
clinical quality of care.
Primary Practice Setting: Tertiary academic medical center

in southern California.
Methodology and Sample: Retrospective data were
collected from 2 databases for all hospital patient care
services from November 2008 to January 2010 to determine
whether clinical quality of care and experiential
service improvements were realized. Primary outcomes included
all-cause and related readmission rates.
Secondary outcomes were Hospital Consumer Assessment of
Healthcare Providers and Systems (H-CAHPS)
scores. An interrupted time series analysis compared data from
the single institution for the diffusion and
postintervention periods.
Results: Comparing data from the diffusion and
postintervention periods, the rate of disease-related
readmissions
decreased signifi cantly (mean 5.43–4.58, p < .05), and all-
cause readmissions also decreased, although the
difference failed to achieve statistical signifi cance (11.42-
10.49, p = .056). H-CAHPS scores with the patient
response of “recommend this hospital” was unchanged over the
2 time points (mean 78.9%–77.8%, p = .26731).
Data also showed stable care management staffi ng rates
whereas average daily census (ADC) increased over
time (ADC 274–297).
Implications for Case Management Practice: With health
reform driving value-driven care transformation,
partnering care managers and social workers with physician
services has the potential to impact the patient’s
experience as well as fi nancial and clinical care outcomes. Care
managers serve a signifi cant role in improving
the clinical quality of care by reinforcing a consistent and clear
message by the health care team to the patient
during the entire hospitalization, not just at the time of
discharge. At one institution, partnering physicians
with care managers through the acute care continuum (service-
based care management) appeared to reduce

readmissions without compromising patient satisfaction. Both
readmission reduction and effective patient
satisfaction scores impact the Centers for Medicare & Medicaid
Services value-based purchasing reimbursement
calculations.
Key words: organization, quality, readmission,
reimbursement, satisfaction
Reduce Readmissions With
Service-Based Care Management
Alpesh N. Amin , MD, MBA, MACP, SFHM ,
Heather Hofmann , MD , Mary M. Owen , RN, MPA
,
Hai Tran , MPA , Saran Tucker , PhD, MPH ,
and Sherrie H. Kaplan , PhD, MPH
DOI: 10.1097/NCM.0000000000000051
I
n response to the U.S. Affordable Care Act signed
on March 23, 2010, the Centers for Medicare &
Medicaid Services (CMS) sought to reduce Medi-
care payments to hospitals through a Value-Based
Purchasing program ( CMS, 2011 ; McCarthy, John-
son, & Audet, 2013; U.S. Department of Health &
Human Services, 2013 ). This new approach included
a change in reimbursement penalties for hospitals
beginning October 1, 2012. As an opportunity for
improvement in service, quality, and cost, reduc-
ing the occurrence of unplanned readmissions has
become a more urgently focused topic. According

to one study, unplanned all-cause readmissions cost
Medicare $17.4 billion in 2004, with 20% of the
Medicare fee-for-service patients readmitted within
30 days of discharge ( Jencks, Williams, & Coleman,
Address correspondence to Alpesh N. Amin, MD, MBA,
MACP, SFHM, University of California, Irvine Health,
333 City Boulevard West, Suite 400, Orange, CA 92868
( [email protected] ).
The authors report no confl icts of interest.
PCM-D-14-00011_LR 255PCM-D-14-00011_LR 255
19/09/14 7:20 AM19/09/14 7:20 AM
256 Professional Case Management Vol. 19/No. 6
g
2009 ). Whether interventions target all-cause or
diagnosis-specifi c readmissions, this change in reim-
bursement policies and practices makes generating
effective interventions to improve effi ciency, quality
and patient experience of care a systemwide priority
( Hines, Yu, & Randall, 2010 ).
At the national level, the Agency for Healthcare
Research and Quality (AHRQ) supports and offers
information regarding two evidence-based projects to
help prevent avoidable readmissions: Project BOOST
and Project RED (AHRQ, 2012b). Project BOOST,

or Project Better Outcomes for Older adults through
Safe Transitions, reduces readmissions in two ways:
mentor “hospital teams to map current processes and
create and implement action plans for organizational
change,” as well as provide “a suite of evidence-based
clinical interventions that can be easily adapted and
integrated into each unique hospital environment”
( Society of Hospital Medicine, 2012 ). The project
focuses on general medicine patients and recognizes
length of stay (LOS) and Hospital Consumer Assess-
ment of Healthcare Providers and Systems (H-CAHPS)
scores as areas of improvement ( Society of Hospital
Medicine, 2012 ). Project Re-Engineered Discharge
(RED) “is a patient-centered standardized approach
to discharge planning” that prepares patients for
discharge by immediately “designating a Discharge
Advocate to coordinate discharge with the care team
and patient” ( AHRQ, 2012a , 2012b ). Both programs
have associated hard costs for personnel to focus on
the targeted population, but costs vary depending on
the institution size. Care managers can provide the
level of patient advocacy needed to decrease the ten-
sion between evidence-based clinical processes and the
patient’s comprehension and readiness for discharge.
With health reform driving value-driven care
transformation, aligning care management with phy-
sician services can signifi cantly improve the patient’s
experience as well as fi nancial and clinical care
outcomes. Care managers serve a signifi cant role in
reducing readmissions ( Hughes, 2012 ). One meta-
analysis of 12 studies, 7 of which were conducted in
the United States, identifi ed “a 6% decrease in read-
mission rate for patients who received hospital-based
care management interventions” ( Kim & Soeken,
2005 ). However, few studies have examined the rela-

tionship between the structure of care management
and readmissions.
According to the Case Management Society of
America ( 2012 ), “case management is a collabora-
tive process … to meet an individual’s health needs [in
order] to promote quality cost-effective outcomes.” A
well-established model of outcomes management—
the Vanderbilt model—established the care manager
as part of a triad with the social worker and utilization
management to best coordinate a patient’s hospitaliza-
tion ( Erickson, 1998 ). The arrangement of care man-
agement varies by hospital, with staff often assigned as
either service- or unit-based. Both methods have pros
and cons that must be considered if converting from
one system to another ( Zander & Warren, 2005 ).
S ERVICE -B ASED C ARE M ANAGEMENT
In 2009, care management services were restructured
at the University of California, Irvine (UCI) Medi-
cal Center, a health care setting that renders tertiary
acute and ambulatory services for patients. The
transition from a unit-based to service-based system
was in response to perceived overstaffi ng of the care
management department based on a review by an
independent consulting fi rm in early 2008. By June
2008, a resource management expert reviewed the
care management program structure and functional
responsibilities. Given the national emphasis on effi -
ciency and effectiveness at the bedside, the consul-
tants recommended realignment to a service-based
model. Service-based care management provided uni-
form care coordination, decreased confusion between
care teams and patient families, managed expenses at
the point of service, and mobilized a team leader to

coordinate care from admission to discharge, with a
handoff to the ambulatory setting when necessary. It
was determined that a reduction in care management
staffi ng was not necessary, but rather an enhanced
emphasis on resource management was needed.
A multidisciplinary work team was assembled
and comprised care managers, nonlicensed autho-
rization coordinators, and clinical social workers,
championed by the executive director, UCI Hospi-
talist Program, and other medical staff leaders along
with the Director of Case Management, with the
support of the chief medical offi cer for UCI Medical
Center. The team outlined goals, including a target of
improving effi ciencies in care delivery and reducing
readmissions, while maintaining effective discharge
With health reform driving value-driven care transformation,
aligning care
management with physician services can signifi cantly improve
the patient’s experience
as well as fi nancial and clinical care outcomes .
PCM-D-14-00011_LR 256PCM-D-14-00011_LR 256
19/09/14 7:20 AM19/09/14 7:20 AM
satisfaction scores. The roles and responsibilities of
the disciplines were defi ned, with the understanding

that each service line would function as a team, often
with roles overlapping on the basis of the case (see
Table 1 ).
By August 2008, the staffi ng and resource needs
were drafted utilizing historical case mix index, aver-
age LOS, and volume (the number of discharges and
patient days) for the physician service lines. The payer
mix was also considered an important element due
to insurance resource limitations that often drive the
ability, or inability, to transition care to the next level.
Based on staffi ng guidelines from the Center for Case
Management, it was anticipated that the service-line
model would require 12.7 additional staff, including
3.4 clinical social workers, 4.2 RN care managers,
and one decision support analyst ( Center for Case
Management, 2007 ). The team proceeded in support
of the resource management expert opinion and did
not seek additional staff at that time.
The service-based care management model was
offi cially rolled out in April 2009. We evaluated the
impact of this program on 30-day all-cause and dis-
ease-specifi c readmission rates and patient experience.
M ETHODS
Research Design
Retrospective data for all services from April 2009
to January 2010 were abstracted to assess the impact
of the program on clinical service improvements.
Improvements were a measure of reduced 30-day
readmission rates and increase or maintenance of
patient satisfaction following implementation of the

program. Two databases were available and used:
University Health System Consortium (UHC) and
Offi ce of Statewide Health Planning & Develop-
ment (OSHPD). Primary outcomes include all-cause
and related readmission rates. H-CAHPS scores were
secondary outcomes. The study time period was
evaluated on the basis of 2 defi ned periods of inter-
vention: April 2009 through August 2009 represents
implementation and diffusion of the intervention, and
TABLE 1
Service-Based Care Management Functional Team Practices
Clinical Social Worker Care Manager Care Coordinator
Verify discharge issues
Interact with family and care team
Educate on resources
Initiate border letters
Assess adjustments to illnesses (social)
Evaluate presumptive disability
Initiate disability letters/FMLA
Arrange patient/family conferences
Monitor compliance with advance health care directive
Collaborate with team on alternative care plans
Plan community resource needs
Verify discharge needs
Interact with family and care team
Facilitate resources
Initiate border letters
Assess adjustments to illnesses (social)
Evaluate presumptive disability
Initiate disability letters/FMLA
Arrange patient/family conferences
Assess adjustments to illnesses (clinical)

Collaborate with team on alternative care plans
Plan community resource needs
Conduct retrospective reviews on discharged cases
Monitor utilization management with payer
Evaluate resource utilization opportunities for
organizational improvement
Negotiate with payers to secure resources for safe
patient discharge
Facilitate follow-up care with physicians, anticoagulation
clinic, and education
Arrange follow-up care for unfunded
Collaborate with Pharmacy on medication needs for
safe discharge
Coordinate psychiatric/acute patient transfers
Verify discharge needs
Obtain equipment
Arrange resources
Research border resources
Note . FMLA = Family and Medical Leave Act.
Service-based care management
provided uniform care coordination,
decreased confusion between care
teams and patient families, managed
expenses at the point of service, and
mobilized a team leader to coordinate
care from admission to discharge, with

a handoff to the ambulatory setting
when necessary .
PCM-D-14-00011_LR 257PCM-D-14-00011_LR 257
19/09/14 7:20 AM19/09/14 7:20 AM
September 2009 through January 2010 represents the
postintervention time period.
Data Abstraction
Data were collected for the study time period on
patients discharged from UCI Medical Center and then
they experienced a subsequent readmission within
30 days of the index admission. Data collected from
UHC exclude chemotherapy, radiation therapy, reha-
bilitation, dialysis, newborn deliveries, mental diseases,
and alcohol and drug use patients. Patient satisfaction
scores refl ect patients older than 18 years with a hospi-
tal admission during the defi ned study period.
Intervention Periods
Two time periods were evaluated in this study. The
diffusion period was April 2009 through August
2009 and represents the period of implementa-
tion. Implementation began in April 2009 with full
implementation complete by August 2009. The pos-
tintervention period was September 2009 through

January 2010. This was the period following full
implementation of the intervention and represents a
period of equal length of time to the diffusion period.
Description of Measures
All-Cause Readmission
All-cause readmission monthly rate calculations were
based upon patients aged greater than 18 years who
returned to the hospital within 30 days of discharge
from the index admission, regardless of the Medicare
diagnosis-related group (MS-DRG) of either admission.
Related Readmission
Related readmission monthly rates were limited
to patients who returned to the hospital within
30 days with an MS-DRG related to the index
admission.
Analysis
Data Consistency
Trends comparing two data sources for 30-day read-
missions during the study time period were evalu-
ated to confi rm the observed trends for data col-
lected from the primary data source, UHC. Pearson’s
r (see Table 2 ) was calculated to evaluate the correla-
tion of monthly UCI 30-day readmission rates from
the UHC database to the rates from OSHPD data
source.
Interrupted Time Series
An interrupted time series (ITS) analysis (see Table 2 )
was conducted comparing trends between the diffu-
sion and postintervention periods for both all-cause

and related monthly 30-day readmission rates. An
autoregressive integrated moving average model (see
Table 2 ) was used to assess the difference in slope
TABLE 2
Glossary of Statistical Terms
Term Defi nition
Pearson’s r Linear or product
moment correlation
Measures the strength of the linear relationship between 2
variables. The correlation coeffi cient, r , ranges
from − 1 to 1. A value of 0 indicates no association.
ITS Interrupted time series
analysis
A method of statistical analysis to compare time trends before
and after intervention.
ARIMA Autoregressive
integrated moving
average
ARIMA methodology is applied to stationary time series data to
describe movement as a function of
autoregressive (AR) and moving average (MA) parameters.
AR Autoregressive Assessment of how a data set is related to
itself over time.
MA Moving average Used as a form of smoothing, by relating
what happens in a time period to the random error on the
previous time period

From “Interrupted Time-Series Analysis and Its Application to
Behavioral Data,” by D. P. Hartmann, J. M. Gottman, R. R.
Jones, W. Gardner, A. E. Kazdin, and R. S. Vaught,
1980, Journal of Applied Behavior Analysis , 13(4), pp. 543-
559.
…the staffi ng and resource needs
were drafted utilizing historical case
mix index, average length of stay, and
volume (the number of discharges
and patient days) for the physician
service lines. The payer mix was also
considered an important element due
to insurance resource limitations that
often drive the ability, or inability, to
transition care to the next level .
PCM-D-14-00011_LR 258PCM-D-14-00011_LR 258
19/09/14 7:20 AM19/09/14 7:20 AM
between the diffusion and postintervention periods
for each measure.
Overall Mean Comparison
Mean 30-day readmission rates for the diffusion

and postintervention periods were also compared
to assess an overall change in 30-day readmission
rates following implementation and diffusion of the
intervention. Mean monthly hospital satisfaction
scores from the H-CAHPS survey were also assessed
as a secondary outcome to determine impact of the
service-based model on the patient’s satisfaction with
regard to recommending the hospital.
R ESULTS
Readmission Rates
Pearson’s r evaluating the correlation of monthly
UCI 30-day readmission rates from the UHC data-
base to the rates from OSHPD data source showed
a signifi cant correlation verifying consistency of the
observed trends in the primary data source ( r = .9,
p < .001).
The ITS analysis for 30-day readmission rates
for all-cause and related readmissions showed insig-
nifi cant negative slope changes comparing diffusion
with postintervention periods. An overall reduction
in mean readmission rates for all-cause and related
30-day readmissions was noted from diffusion of
intervention to postintervention (see Table 3 ). A
reduction in mean all-cause 30-day readmission from
11.42% to 10.49% readmissions was noted with a
signifi cant reduction in the mean related 30-day read-
mission rate from 5.43% to 4.58% readmissions.
Patient Satisfaction
The ITS analysis performed on monthly patient sat-

isfaction scores for nursing communication showed
insignifi cant slope changes comparing intervention
periods. In addition, there was no signifi cant change
in overall mean satisfaction between diffusion of
intervention and postintervention time periods (see
Table 4 ).
Staffi ng
Service-based care management typically requires
additional staffi ng due to logistical challenges of
services not geographically localized. However,
additional staff was not hired for this realignment.
As the census increased during the diffusion period
of the study due to hospital growth, additional per
diem staff supplemented the team as needed. From
November 2008 through January 2010, 1.67 addi-
tional FTEs were used while the census increased by
TABLE 3
Comparison of Mean Proportion of 30-Day Readmissions for
Diffusion and Postintervention Periods
Readmission Type Phase Mean Rate Mean Difference SE
average n a p
All-cause Diffusion 11.42 0.931 0.429 1474 0.056
Post 10.49 0.301 1444
Related Diffusion 5.43 0.857 0.271 1474 0.02
Post 4.58 0.226 1444
Note . All-cause = patients older than 18 years who returned
to the hospital within 30 days of discharge from index

admission, regardless of Medicare diagnosis-related
group (MS-DRG) of either admission. Related = patients older
than 18 years who returned to the hospital within 30 days of
discharge from index admission, with an MS-
DRG related to the index admission.
Diffusion period = April–August 2009.
Postintervention period = September 2009–January 2010. SE
= standard error.
a Average number of cases per month.
TABLE 4
Comparison of Mean Satisfaction Scores for Diffusion and
Postintervention Periods
Phase Mean Rate Mean Difference SE Average n a p
Diffusion 78.9 1.08 1.43 185 0.267
Post 77.8 0.86 178
Note . Satisfaction scores obtained from H-CAHPS survey
response to “Recommend the hospital.” Diffusion Period =
April–August 2009; Postintervention Period =
September 2009–January 2010. SE = standard error.
a Average number of cases per month.
PCM-D-14-00011_LR 259PCM-D-14-00011_LR 259
19/09/14 7:20 AM19/09/14 7:20 AM

approximately 25 patients per day (1:15 ratio equiva-
lent; see Figure 1 ).
D ISCUSSION
When faced with limited resources and reductions
in reimbursement for readmissions, changing from a
unit-based to a service-based care management model
assisted with reductions in readmissions without
compromising patients’ experience of the quality of
care. Given the importance of transitions of care, dis-
charge planning is a key focus for reducing readmis-
sions. It is well established that care management is
a fi eld that can effectively assist with discharge plan-
ning ( Maramba, Richards, Myers, & Larrabee, 2004 ;
Naylor, Aiken, Kurtzman, Olds, & Hirschman, 2011 ;
Simmons, 2005 ).
Numerous factors lie at the core of the issue of
readmissions. For one, studies have shown that there
is an important link between clinical quality and the
patient experience, referred to as experiential quality
( Chandrasekaran, Senot, & Boyer, 2012 ; Glickman et
al., 2010 ). Chandrasekaran’s team found that “CMS
process management is positively associated with
clinical quality but negatively associated with expe-
riential quality, suggesting a tension between the two
healthcare outcomes.” When they further explored
the link between readmissions and experiential quality
among 2,942 hospitals, a signifi cant and much stron-
ger correlation was identifi ed than found with clinical
process measures. In addition to experiential qual-
ity, factors such as insurance status and LOS greater
than 2 days can predict a readmission for medicine
service patients (Hasan et al., 2009). Others argue

that there are numerous barriers beyond tangible fac-
tors. According to an online survey of 51 National
Association of Public Hospitals and Health Systems
member facilities, one of four categories of barriers
to reducing readmissions is “hospital quality of care
barriers—systems or processes within a hospital that
When faced with limited resources
and reductions in reimbursement
for readmissions, changing from a
unit-based to a service-based care
management model assisted with
reductions in readmissions without
compromising patients’ experience of
the quality of care .
FIGURE 1
Care management staffi ng versus average daily census for
care managers. Staffi ng, number of care managers
during the given month. Average Daily Census, number of cases
supported by care managers on average per day
during the given month. Preintervention Period, November
2008–March 2009. Diffusion Period, April–August 2009.
Postintervention Period, September 2009–January 2010.
PCM-D-14-00011_LR 260PCM-D-14-00011_LR 260
19/09/14 7:20 AM19/09/14 7:20 AM

directly contribute to poor quality of care for patients
(e.g., poor coordination of care)” ( Siegel, 2011 ).
Addressing these factors is a viable focus for improve-
ment in the quality of care. Employing care manage-
ment in patient satisfaction may reduce readmissions
and LOS ( Patient Satisfaction Planner, 2007 ).
The Affordable Care Act includes provisions for
hospital-level accountability for patients’ experience of
health care quality. As a community safety-net hospital
in southern California, this type of hospital has incurred
greater reimbursement reductions due to unfavorable
readmission rates ( Berenson & Shih, 2012 ; Joynt &
Jha, 2013 ; Rau, 2012 ). Extrapolation of these fi ndings
to other institution types is limited. However, all hospi-
tals should strive for improved patient experience that
service-based care management maintained.
In the age of value-based purchasing, one must
question the most appropriate intervention for care
management structure and readmission management.
Projects RED & BOOST may be less cost-effective
to pursue, compared with changing from a unit- to
service-based structure.
Factors external to the implementation of the ser-
vice-based care management model may have contrib-
uted to the success, given the organizational focus on
improving heart failure, acute myocardial infarction,
and pneumonia readmission rates. With the care man-
agement team aligned by service, the level of expertise
and knowledge of that population’s needs resulted in a
more proactive approach to discharge planning.
Targeted resource utilization analysis for specifi c

diagnostic-related groups (DRG) and cost of care
are currently being evaluated to determine secondary
improvements. A Resource Utilization Council employs
the expertise of the care management experts to help
delineate opportunities for effi ciency and effectiveness
improvements by DRGs within each service line.
There are several limitations to this study. First,
this is a retrospective study, with index cases from
a 5-month period. Data stem from a single institu-
tion, which may limit its application to other popu-
lations. Furthermore, readmissions were considered
if both the index and second admission occurred at
our institution. Failure to capture a readmission at
another institution limits the generalization to larger-
scale readmission reduction efforts. Finally, reliance
on H-CAHPS scores excludes patients who did not
complete the survey and may be an insuffi cient sur-
rogate for all components of patient satisfaction.
C ONCLUSIONS
The presence of care management in patient care is
essential to reducing readmissions. Our service-based
care management model contributed to reducing
related readmissions for all physician service lines
despite decreased staffi ng and without compromising
patient satisfaction.
ACKNOWLEDGMENT
The authors thank Susan Kelleghan, RN, for assis-
tance in H-CAHPS data analysis.
R EFERENCES

Agency for Healthcare Research and Quality . ( 2012a ).
Pre-
venting avoidable readmissions: Improving the hospi-
tal discharge process . Retrieved June 13, 2012, from
http://www.ahrq.gov/qual/impptdis.htm
Agency for Healthcare Research and Quality . ( 2012b ).
Project RED (Re-Engineered Discharge) training pro-
gram . Retrieved June 13, 2012, from http://www.ahrq
.gov/qual/projectred/
Berenson , J. , & Shih , A. ( 2012 ). Higher
readmissions at
safety-net hospitals and potential policy solutions .
New York, NY: The Commonwealth Fund .
Case Management Society of America . ( 2012 ). What is a
case manager? Retrieved June 19, 2012, from http://
www.cmsa.org/Home/CMSA/WhatisaCaseManager/
tabid/224/Default.aspx
Center for Case Management . ( 2007 ). Proposed frame-
work to plan/evaluate/benchmark staffi ng . Wellesley,
MA: Author.
Centers for Medicare & Medicaid Services . ( 2011 ).
Medi-
care program; hospital inpatient prospective payment
systems for acute care hospitals and the …
Managing to improve quality: The
relationship between accreditation
standards, safety practices, and

patient outcomes
Deirdre K. Thornlow
Elizabeth Merwin
Background: Given the trend toward eliminating reimbursement
for ‘‘never events,’’ hospital administrators are
challenged to implement practices designed to prevent their
occurrence. Little evidence exists, however, that
patient safety practices, as evaluated using accreditation
criteria, are related to the achievement of patient safety
outcomes.
Purpose: The aim of this study was to examine the relationship
between patient safety practices, as measured by
accreditation standards, and patient safety outcomes as
measured by hospital rates of infections, decubitus
ulcers, postoperative respiratory failure, and failure to rescue.
Methodology: Secondary data were used to examine
relationships between patient-safety-related accreditation
standards and patient outcomes in U.S. acute care hospitals.
Accreditation performance areas were reduced
into subscores to represent patient safety practices. Outcome
rates were calculated using the Agency for
Healthcare Research and Quality Patient Safety Indicator
software. Multivariate regression was performed to

determine the significance of the relationships.
Findings: Three of four multivariate models significantly
explained variance in hospital patient safety indicator
rates. Accreditation standards reflecting patient safety practices
were related to some outcomes but not others.
Rates of infections and decubitus ulcers occurred more
frequently in hospitals with poorer performance in
utilizing patient safety practices, but no differences were noted
in rates of postoperative respiratory failure or
failure to rescue.
Practice Implications: Certain adverse events, such as infections
and decubiti, may be reduced by preventive
protocols that are reflected in accreditation standards, whereas
other events, such as failure to rescue and
postoperative respiratory failure, may require multifaceted
strategies that are less easily translated into protocols.
Our approach may have influenced the observed associations yet
represents progress toward assessing whether
safety practices, as measured by accreditation standards, are
related to patient outcomes.
July–September � 2009262
Deirdre K. Thornlow, PhD, RN, CPHQ, is Assistant Professor,
School of Nursing, Duke University, Durham, North Carolina.

E-mail:
[email protected]
Elizabeth Merwin, PhD, RN, FAAN, is Associate Dean,
Research, Madge M. Jones Professor of Nursing, and Director,
Rural Health Care
Research Center, School of Nursing, University of Virginia,
Charlottesville. E-mail: [email protected]
This study was approved by the institutional review board
(UVA No. 2004-0255-00) and supported by Grant No. F31
NR009320-01 from the
National Institute for Nursing Research. The contents of this
article are solely the responsibility of the authors and do not
necessarily represent the
official views of the National Institute for Nursing Research.
Key words: adverse events, hospital accreditation, outcomes,
patient safety, safety practices
Health Care Manage Rev, 2009, 34(3), 262-272
Copyright A 2009 Wolters Kluwer Health | Lippincott Williams
& Wilkins
9Copyright @ 200 Lippincott Williams & Wilkins.
Unauthorized reproduction of this article is prohibited.
H
ospitals should be motivated more than ever to
commit resources and attention to patient safety,
for beginning October 2008, the Centers for
Medicare and Medicaid Services (CMS) eliminated or
reduced payments for pressure ulcers, hospital-acquired
infections, and other ‘‘never events,’’ defined as prevent-
able adverse events that should never occur in health care

(CMS, 2007). This change in reimbursement follows
earlier CMS initiatives that now require hospitals to
submit evidence-based quality measures or suffer reduc-
tions in their Medicare Annual Payment Updates. The
CMS is expanding reimbursement models by reward-
ing hospitals with higher Medicare payment for better
mortality outcomes; poor performing hospitals will be
penalized with reduced payments. Such reimbursement
changes may be justified by research evidence demon-
strating that most medical errors, or adverse events, are
preventable (Lehman, Puopolo, Shaykevich, & Brennan,
2005; Thomas et al., 2000) and hospitalized patients who
experience a medical error remain hospitalized longer
and accrue greater costs when compared with controls
(Nordgren, Johnson, Kirschbaum, & Peterson, 2004;
Rojas, Silver, Llewellyn, & Rances, 2005).
The research approach in this study may inform
hospitals about the influence of organizational character-
istics and processes of care on patient safety outcomes.
Although studies have shown associations between char-
acteristics of hospital systems, such as teaching status,
ownership status, nurse staffing, and patient safety out-
comes (Ayanian, & Weissman, 2002; Devereaux et al.,
2002; Kupersmith, 2005; Stanton, 2004), few studies have
examined how these hospital characteristics influence
utilization of patient safety practices, and even fewer
studies have examined the impact of using patient safety
practices on patient outcomes. In designing patient
safeguards, it is essential to consider how patient safety
practices, defined as types of care processes whose
application reduces the probability of an adverse event
(Shojania, Duncan, McDonald, & Wachter, 2002), con-
tribute to safe care. This is challenging, as little evidence
suggests that safety practices, such as those commonly used
in non-health-care fields, confer benefit in acute care

hospitals, especially on patient outcomes (Shojania et al.,
2002). When evidence does exist regarding efficacy of
safety processes, organizations have made attempts to
translate such evidence into practice, to include incorpo-
rating patient safety standards into the hospital accredi-
tation process (Kizer & Blum 2005; Leape, Berwick, &
Bates, 2002); however, the link between these practices
and outcomes has not yet been clarified.
In this study, secondary data were used to examine
relationships among hospital systems, utilization of patient
safety practices, as measured by accreditation standards,
and patient outcomes in acute care hospitals to determine
whether the use of patient safety practices influences rates
of four patient safety indicators: infections due to medical
care, decubitus ulcers, postoperative respiratory failure,
and failure to rescue. We hypothesized that teaching
hospitals, hospitals with higher nurse staffing levels, and
hospitals using more patient safety practices would ex-
perience lower rates of these patient safety incidents than
would nonteaching hospitals, hospitals with lower levels of
nurse staffing, and hospitals using fewer patient safety
practices. Findings, implications for current practice, and
suggestions for future research designed to improve patient
safety in acute care hospitals will be presented.
Conceptual Framework
The Quality Health Outcomes Model (QHOM; Mitchell,
Ferketich, & Jennings, 1998) served as the conceptual
framework for this study. In 1998, the Expert Panel on
Quality Health Care of the American Academy of
Nursing published the QHOM as a conceptual framework
for quality and outcomes research, most specifically as a
means to test relationships among the elements of struc-

ture, process, and outcomes. The QHOM built on these
three elements from Donabedian’s (1966) seminal work in
assessing the quality of medical care and incorporated
client, or patient, characteristics as a fourth construct. The
QHOM realigned the constructs to capture their dynamic,
rather than linear, relationships. The traditional ‘‘struc-
ture’’ construct was renamed ‘‘system’’ in the QHOM,
whereas the traditional ‘‘process’’ construct was renamed
‘‘intervention.’’ The QHOM posits reciprocal interactions
among the four constructs (system, intervention, outcome,
and client), thus serving as a useful conceptual guide for
health care systems researchers. Several investigators have
used the QHOM model in acute and community care to
organize their choice of variables among the four con-
structs and to build evidence regarding the quality of
health care (Mayberry & Gennarro, 2001; Radwin &
Fawcett, 2002; Sin, Belza, LoGerfo, & Cunningham,
2005). In this study, system variables included hospital
characteristics such as teaching status, ownership status,
size, location, and nurse staffing levels; intervention vari-
ables were defined as utilization of patient safety practices;
outcomes, or patient safety indicators, were defined by the
Agency for Healthcare Research and Quality (AHRQ,
2007); and client characteristics were defined as risk-
adjusted variables including diagnosis, age, and gender.
The client variables were used to flag potentially pre-
ventable complications and to create hospital-level risk-
adjusted patient safety indicator rates (AHRQ, 2007).
Measures for each of the constructs are described below.
Methods
Secondary data were analyzed from a stratified probability
sample of acute care hospitals. Hospital-level data were
Safety Practices and Patient Outcomes 263

gathered from the 2002 American Hospital Association
(AHA) annual survey data and 2002 Joint Commission
(JC) accreditation performance reports. Hospital accred-
itation performance reports were retrieved online from
JC Quality Check (www.jointcommission.org) in 2005.
Patient-level data were gathered from the 2002 Nation-
wide Inpatient Sample (NIS), the largest all-payer
inpatient database in the United States which contains
patient-level clinical and resource data on hospital stays
from states participating in the Healthcare Cost and
Utilization Project and is designed to approximate a 20%
stratified probability sample of U.S. community hospitals
(AHRQ, 2008). Patient characteristics were risk adjusted
by age, gender, diagnoses, and comorbidities using the
AHRQ (2007) comorbidity software algorithm, as de-
scribed in the ‘‘Measures’’ section.
Study Sample
General medical–surgical community hospitals, as clas-
sified by AHA, served as the population for study.
Specialty hospitals, such as children’s, psychiatric, and
rehabilitation hospitals, were excluded because the se-
lected patient safety indicators address incidents that
are more likely to occur in general medical–surgical
adult patients. The 2002 NIS inpatient discharge-level
file, which contains data for 100% of the discharges (n =
7,853,982) from 995 hospitals in 35 participating states,
was used. Of the 35 states in the 2002 NIS sample, 10
states prohibit release of AHA hospital identifiers; there-

fore, these states’ hospitals were excluded. In addition,
one state did not code patient comorbidities, data that
are necessary to risk adjust; therefore, this state and its
hospitals also were excluded from the sample. And fi-
nally, to complete analyses, it was necessary to match the
2002 NIS data to 2002 JC survey data; therefore, only
those hospitals surveyed by JC in 2002 were included in
this study—four states had no surveyed hospitals in the
2002 NIS sample. The study sample was composed of
1,430,981 inpatient discharge records from 115 hospitals
in 20 states. Sample size equaled less than 115 hospitals
for two of four outcomes studied due to coding limitations
in two states that precluded inclusion (e.g., missing ad-
mission type).
Measures
Data were used to construct variables that measured
hospital systems, utilization of patient safety practices, and
risk-adjusted patient outcomes. Hospital system variables
were constructed from NIS and AHA survey data. These
descriptive structural measures included nurse staffing
(ratio of RN full-time equivalents to adjusted average
daily census [A-ADC]), teaching status, hospital location
(urban or rural based on metropolitan statistical area
population standards), hospital size (A-ADC), and owner-
ship. Certain variables were recoded to conserve degrees
of freedom. First, the continuous variable A-ADC was
used to measure hospital size rather than NIS-designated
categories small, medium, and large. Second, hospital
ownership was originally classified by NIS in five cate-
gories: government nonfederal; private, not-for-profit;
private, investor owned; and two additional categories
into which smaller strata of hospitals were collapsed.
Federal hospitals are not sampled in NIS. Due to the

large numbers of hospitals within the two NIS-collapsed
categories, precise ownership information for each of
the 115 sample hospitals was obtained from the 2006
AHA directory so that all hospitals could be accurately
categorized, without the need for collapsed categories.
Ownership was then coded into two levels: for-profit
and other.
Joint Commission Accreditation Performance Reports
were used to construct a measure of each hospital’s
utilization of patient safety practices. During accreditation
surveys, 45 performance areas encompassing nearly 500
standards are evaluated and scored. It should be noted that
accreditation performance areas were scored on a scale of 1
to 5, with 5 being the poorest score; thus, a higher score
indicates poorer performance in using that safety practice.
Because only half of the JC standards related to patient
safety, as noted by the JC (2003), and little variation
existed among our study hospitals in their overall ac-
creditation scores (M = 92.3, SD = 3.68), we sought a
parsimonious method to differentiate patient safety
practices among hospitals. Using a 4-point scale (1 =
‘‘not relevant’’ to 4 = ‘‘very relevant’’), an expert panel
composed of hospital quality improvement directors and
nurse executives, excluding the authors, evaluated the
45 performance areas to determine, in their expert
opinion, which performance areas most embodied patient
safety. The expert panelists unanimously rated 12 per-
formance measures as most relevant to patient safety
(rating of 3 or greater). The content validity index was
measured by percentage reviewer agreement for each item
and for the total 45-item instrument. Across the in-
strument, the content validity index equaled 0.74; a score
of 0.78 or better indicates good content validity (Polit,
Beck, & Owen, 2007). We then conducted principal
components analysis, using an orthogonal rotation, to

determine the underlying structure for the 12 retained
measures; this produced a four-component solution that
was evaluated as adequate using four criteria: eigenvalue
>1, variance, scree plot, and residuals. In this study, factor
loadings ranged from .414 to .798, with an average of .631,
generally considered very good (Comrey & Lee, 1992;
Fleury, 1998). Aggregate factor scores created parsimony
in variables tested; likewise, an additional strength of this
technique is that, because factors are orthogonal, the
factor scores are nearly uncorrelated and can be used in
July–September � 2009264 Health Care Management REVIEW
regression analyses without producing multicollinearity
among the subsets of variables (Tabachnick & Fidell,
2001). The four regression factor scores were used to
represent patient safety practices in this study and were
named ‘‘surveillance capacity,’’ ‘‘assessing patient needs,’’
‘‘care procedures,’’ and ‘‘measuring processes.’’ Detailed
information regarding this factor analysis has been
published elsewhere (Thornlow, 2008).
Hospital rates of occurrence for each of four indicators,
infections due to medical care, decubitus ulcers, failure to
rescue, and postoperative respiratory failure, were calcu-
lated by applying the Patient Safety Indicator (PSI) soft-
ware (version 3.0a) to the NIS data set (http://www.
qualityindicators.ahrq.gov/psi_download.htm). These indi-
cators were selected because prior research has suggested
that these outcomes are potentially attributable to organi-
zational characteristics (AHRQ, 2007; Romano, Geppert,

Davies, Miller, Elixhauser, & McDonald, 2003), includ-
ing nurse staffing (Aiken, Clarke, Sloane, Sochalski, &
Silber, 2002; Blegen, Goode, & Reed, 1998; Kovner &
Gergen, 1998; Lichting, Knauf, & Milholland, 1999;
Needleman, Buerhaus, Mattke, Stewart, & Zelevinsky,
2002; Unruh, 2003) and care processes (Danks, 2006;
Frantz, 2004; Gastmeier & Geffers, 2006; Kovner &
Gergen, 1998; Lyder, 2003). In calculating rates, the PSI
software generates an algorithm that uses the ICD-9-CM
diagnosis and procedure codes, date of procedure, and pa-
tient characteristics, including age, gender, and diagnosis-
related group, to flag potentially preventable complications.
In running the software, hospital-level risk-adjusted ratios
with smoothing were calculated for the four patient safety
indicators. The smoothing process applies shrinkage fac-
tors to reflect a reliability adjustment unique to each in-
dicator. The less reliable the PSI is over time and across
hospitals, the more the estimate shrinks the PSI toward
the overall mean. The resulting rate appears ‘‘smoother,’’
or more conservative than the raw rate, and random year-
to-year fluctuations in performance are likely to be re-
duced (AHRQ, 2007).
Data Analysis
Univariate and multivariate regression analyses were
conducted at the hospital level; each PSI outcome was
analyzed separately. Moderate to strong statistically
significant correlations existed among the raw, risk-
adjusted, and smoothed rates for each PSI; therefore, all
analyses were conducted on smoothed PSI outcome
rates as smoothed rates had been reported in similar
studies (Miller et. al., 2005; Thornlow & Stukenborg,
2006). Although we tested its association with patient
outcomes, overall accreditation score was not significant
in either univariate or multivariate regression analyses

and was thus not included in further testing. To com-
plete hypothesis testing, final regression models were
built using variables that were hypothesized to be
significant a priori; variables found to be significantly
associated with patient outcomes in preliminary univar-
iate regression analyses ( p < .05) or in preliminary
multivariate regression analysis ( p < .05) were included
in the final multivariate models.
Findings
Most of the 115 hospitals included in the sample were
classified as urban (n = 78), nonteaching (n = 88), and
not-for-profit (n = 76); almost half were considered large
institutions (n = 56) and approximately 43 hospitals
(37%) were located in the South (Table 1). Study
sample hospitals differed from the 2002 NIS sample in
that the study sample had fewer small hospitals, fewer
rural hospitals, fewer government nonfederal hospitals,
and fewer hospitals located in the Midwest than the
national NIS sample did. No differences in teaching
status existed between the 2002 NIS sample and the
study sample (Table 1). The A-ADC of study hospitals
ranged from 10 to 1,397 patients per day (M = 265.59,
SD = 247.68) with an average nurse ratio of 1.18 RN
full-time equivalents (SD = 0.44). Overall accreditation
scores ranged from 83 to 99 on a scale of 100 (M = 92.3,
SD = 3.68). Hospital risk-adjusted (smoothed) rates for
the patient safety indicators ranged from 1.8 cases of
infection per 1,000 discharges (0.0018) to 9.8 cases of
postoperative respiratory failure per 1,000 elective sur-
gical discharges (0.0098) to 21.5 incidences of decubitus
ulcers per 1,000 discharges (0.0215) to 133.9 cases of
failure to rescue, or deaths, per 1,000 discharges among
patients who developed potentially preventable compli-

cations during their hospital stay (0.1339). These rates
are comparable to 2002 national rates for these indi-
cators (http://hcupnet.ahrq.gov/; Table 2).
Overall, three of the four multivariate models attained
significance. Results are shown for the preliminary
multivariate regression models rather than for the final
multivariate regression models because no appreciable
differences were noted in the strength or direction of
relationships or amount of variance explained between the
preliminary and final multivariate models (Table 3).
Hospital system characteristics and patient safety practices
accounted for 21.9% of the adjusted variance in hospital
rates of infection ( p = .000), 13.0% of the adjusted
variance in hospital rates of postoperative respiratory
failure ( p = .011), and 8.9% of the adjusted variance in
hospital rates of decubitus ulcers ( p = .029). None of the
models significantly explained variance in hospital rates of
failure to rescue ( p = .436).
Hospital system characteristics were not consistently
associated with patient outcomes in either univariate or
multivariate regression analyses (Table 3). Although not
hypothesized to be significant a priori, larger hospitals
had higher rates of infections due to medical care and
postoperative respiratory failure than did smaller hos-
pitals, but no differences were noted by hospital size in
either failure to rescue or decubitus ulcers. And when

compared with not-for-profit and government nonfed-
eral hospitals, for-profit hospitals exhibited higher rates
of decubitus ulcers and postoperative respiratory fail-
ure, but no differences were noted by ownership for rates
for failure to rescue or infections due to medical care.
No differences in hospital patient safety indicator rates
were noted by teaching status or levels of nurse staf-
fing (Table 3).
Utilization of patient safety practices, as measured by
accreditation standards, was significantly associated with
two of the four patient safety outcomes in both
univariate and multivariate analyses. Hospitals with
poorer performance using the patient safety practice
‘‘assessing patient needs’’ (Subscore 2) had higher rates
of infection due to medical care than did hospitals with
better performance using that practice, and hospitals
with poorer performance using ‘‘care procedures’’ (Sub-
score 3) had higher rates of decubitus ulcers than did
hospitals with better performance using that patient
safety practice. Utilization of patient safety practices was
not associated with hospital rates of postoperative
respiratory failure or failure to rescue in either univariate
or multivariate analyses.
Discussion
In this study, hospital system characteristics did not
consistently explain patient outcomes, echoing previous
findings in which associations varied depending on the
outcome measured (Baker et al., 2002; Romano et al.,
2003; Thornlow & Stukenborg, 2006). In this study, larger
hospitals demonstrated higher rates of adverse events
than smaller hospitals did, but for only two of the four
indicators analyzed: infections due to medical care and

postoperative respiratory failure. Perhaps, increased con-
tact from a larger number of staff increases the probability
of cross-contamination and infection; likewise, the need
for interdisciplinary communication among the many pro-
viders in a large hospital may predispose such institutions
to higher rates of postoperative respiratory failure than in a
Table 1
Comparison of study sample hospitals to NIS sample hospitals
Study hospitals
(sample, n = 115)
National sample
hospitals (NIS,
n = 995)
Difference between
study and
national samples
Hospital characteristics n % n % �2
Hospital teaching status 2.76
Nonteaching 88 76.5 729 82.8
Teaching 27 23.5 151 17.2
Hospital ownership (five categories) 17.34**
Government or private, collapsed 49 42.6 290 33.0
Government, nonfederal (public) 9 7.8 193 21.9
Private, collapsed 27 23.5 168 19.1
Private, not-for-profit 19 16.5 103 11.7
Private, investor owned 11 9.6 126 14.3
Hospital location 8.68**

Rural 37 32.2 411 46.7
Urban 78 67.8 469 53.3
Hospital bed size 22.13***
Small 20 17.4 343 39.0
Medium 39 33.9 255 29.0
Large 56 48.7 282 32.1
Hospital region 9.75*
Northeast 24 20.9 112 12.7
Midwest 22 19.1 262 29.8
South 43 37.4 340 38.6
West 26 22.6 166 18.9
Note. NIS = Nationwide Inpatient Sample.
*Significant at p < .05.
**Significant at p < .01.
***Significant at p < .001.
July–September � 2009266 Health Care Management REVIEW
Table 2
Descriptive statistics for study sample hospitals (n = 115)
Variable Operational definition Mean SD
2002 National
PSI rate

Nurse staffing Measured as the ratio of RN full-time
equivalents to A-ADC
1.18 0.44 –
A-ADC Hospital size was measured as A-ADC.
A-ADC reflects the average number
of both inpatients and outpatients
treated at the hospital on a daily basis.
265.59 247.68 –
Overall score This score is derived from an assessment
of an organization’s compliance
with all applicable Joint Commission
standards at the time of the full
triennial accreditation survey.
Score is based on a scale of 0 to
100, with 100 representing the
highest score.
92.30 3.68 –
Decubitus ulcer
(n = 115 hospitals)
Decubitus ulcer development
per 1,000 discharges in lengths
of stay of five or more days.
Excludes patients with paralysis,
diseases of the skin, subcutaneous
tissue, and breast. Excludes
obstetrical admissions and
admissions from long-term care.

0.0215 (21.50) 0.01 (23.63)
Failure to rescue
(n = 99 hospitals)
Deaths per 1,000 discharges among
patients who develop potentially
preventable complications during
their hospital stay. Excludes patients
transferred in or out, patients admitted
from long-term care, neonates, and
patients over 74 years.
0.1339 (133.90) 0.02 (129.37)
Selected infections
due to medical
care (n = 115
hospitals)
Rate per 1,000 discharges of infections
due to medical care, primarily those
related to intravenous lines and catheters.
Defined by including cases based on
secondary diagnosis associated with the
same admission. Excludes patients with
potentially immunocompromised states
(e.g., AIDS and cancer).
0.00181 (1.81) 0.00 (1.53)
Postoperative
respiratory failure
(n = 100 hospitals)
Rates of postoperative respiratory failure

per 1,000 elective surgical discharges.
Limits code to secondary diagnoses
to eliminate respiratory failure that
was present on admission. Excludes
patients who have major respiratory
or circulatory disorders. Limits
the population at risk to elective
surgery patients.
0.0098 (9.80) 0.00 (7.97)
Variable Operational definition Median SD Cronbach’s �
Patient safety
practices/
subscores
Subscore 1 (surveillance capacity)
Reassessment procedures
Implementation of patient safety plans �0.2754 1 .510
Orientation, training staff
Assessing staff competency
(continues)
smaller hospital. Additional research is needed to examine
why larger hospitals demonstrated higher rates for these
patient outcomes.

For-profit hospitals had higher rates of adverse events
than did not-for-profit and nonfederal government
hospitals, but again, for only two of four indicators:
decubitus ulcer and postoperative respiratory failure.
Reasons for these findings are unclear yet support those
of other studies where for-profit hospitals had higher
rates of postoperative pneumonia, pulmonary compro-
mise (Kovner & Gergen, 1998), postoperative respira-
tory failure, and decubitus ulcers (Romano et al., 2003)
than other hospital types did. For-profit and not-for-
profit hospitals may differ in the types of resources used
Table 3
Multivariate regression analysis: relationship of hospital
systems and utilization of
patient safety practices to patient safety outcomes
Standardized � coefficients
Decubitus ulcer
Failure to
rescue
Infection due
to medical care
Postoperative
respiratory failure
Hospital characteristics
Western region .09 .13 .16 .00
Urban location .07 �.01 .04 .11
Hospital size (A-ADC) .17 .17 .30* .36*
Teaching hospital .05 �.11 .08 �.02

For-profit owner .24* .04 .02 .32*
RN staffing (FTE/A-ADC) �.05 �.08 .12 �.16
Patient safety practices
Subscore 1: surveillance capacity .03 .10 �.13 �.09
Subscore 2: assessing patients �.04 .06 .25* .07
Subscore 3: care procedures .27** �.12 �.10 .07
Subscore 4: measuring process .11 .19 �.15 .11
F test 2.12 1.11 4.20 2.48
Model significance .03* .44 .00*** .01*
R2 .17 .10 .29 .22
Adjusted R2 .09 .00 .22 .13
Note. A-ADC = adjusted average daily census; FTE = full-time
equivalent.
*Significant at p < .05.
**Significant at p < .01.
***Significant at p < .001.
Table 2
Continued
Variable Operational definition Median SD Cronbach’s �
Subscore 2 (assessing patient needs)
Initial assessment �0.1212 1 .533
Availability of patient-specific information
Medication use
Subscore 3 (care procedures)
Infection control 0.0611 1 .096

Planning and providing care
Operative procedures
Subscore 4 (measuring …

Professional Case Management Vol. 19 , No. 6 , 255 - 262.docx

Recommended

Recommended

More Related Content

Similar to Professional Case Management Vol. 19 , No. 6 , 255 - 262.docx

Similar to Professional Case Management Vol. 19 , No. 6 , 255 - 262.docx (20)

More from madlynplamondon

More from madlynplamondon (20)

Recently uploaded

Recently uploaded (20)

Professional Case Management Vol. 19 , No. 6 , 255 - 262.docx