Are Changing Rates of Admission
for Chronic Medical Conditions
Simply a Reflection of Changes in
the Demographics, Health Status and
Geographic Migration Patterns of the
Elderly?
Presented at:
the AcademyHealth 2004 Annual Research
Meeting, San Diego, CA, June 6–8, 2004
Presented by:
Nancy McCall, Sc.D.
P.O. Box 12194 · 3040 Cornwallis Road · Research Triangle Park, NC 27709
Phone: 202-728-1968 · Fax: 202-728-2095 · nmccall@rti.org · www.rti.org
RTI International is a trade name of Research Triangle Institute.
1
Acknowledgements

Lee Mobley, Ph.D.

Sujha Subramanian, Ph.D.

Erica Brody, M.P.H.

Mary Kapp, M. Phil
2
Research Question

What is the influence of beneficiary
sociodemographic and health status
characteristics on the rate of growth of
ACSC admissions?
3
Data

Rates of ACSC admissions and average health
status of the Medicare FFS population
 1992–2000 MQMS Base Analytic Files

Rates of Emergency Room or observation bed
stays
 1992–2000 Outpatient SAFS

Estimates of the proportion of the Medicare
population with specific attributes of interest
 1992–2000 MQMS Base Denominator Files
4
Methods

Full Year Part A and B Medicare FFS,
including deceased

Age 65 and older and residing in U.S.

Approximately 25 million per year

Defined two ACSCs for beneficiaries with
diabetes

Health status measured using the PIP-DCG
predictive expenditure model

Age-sex Adjusted to July 1, 1999 FFS
Population using direct standardization
5
All Cause
Hospitalization Rates
Trend in Age-Sex Adjusted All Cause
Admissions (per thousand) Medicare FFS
Beneficiaries: 1992-2000
1992
1994
1996
1998
2000
% Change
1992–2000
317
316
323
331
336
+ 6.0
6
% Change in Inpatient and Outpatient
Age-Sex Adjusted Rates for Eleven
Selected ACSCs, 1992-2000
Cellulitis
Asthma
COPD
Dehydration
CHF
Acute diabetic events*
Lower Limb PVD
Pneumonia
Septicemia
Stroke
UTI
Inpatient
12.0
-30.0
Outpatient
47.0
14.1
52.2
20.9
0.7
-6.0
-23.4
14.1
11.0
-14.2
12.9
33.6
92.7
19.8
80.7
38.0
46.8
51.4
0.20
25.3
7
Empirical Model
ACSCjt = f (OUTPATjt, SOCIOjt, HEALTHjt, GEO,
YEAR, TIMEt)

ACSCjt = rate of inpatient admissions for the specific
ACSC in year t and region j, where each state is
divided into one MSA and one non-MSA region;

OUTPAT is rate of ER/observation bed stays for the
specific ACSC in year t and region j;

SOCIO = a vector of yearly beneficiary demographic
characteristics aggregated to each region;

HEALTH = a yearly health status measure
aggregated to each region;
8
Empirical Model
ACSCjt = f (OUTPATjt, SOCIOjt, HEALTHjt, GEO,
YEAR, TIMEt)

GEO = a set of census region dummy variables;

YEAR = a set of dummy variables for each year
1993–2000
 Interacted with two variables — median age and
outpatient rates

TIME is a continuous time variable — 1993…2000
9
Empirical Model

Three chronic ACSCs:
 Lower limb peripheral vascular disease (PVD)
 COPD
 CHF

Independent variables in the SOCIO vector are
specified as proportions, except for the median age

Health status is represented as the median of the
PIP-DCG risk score of the population for the year.

For PVD, we add the number of Medicare
beneficiaries with diabetes in the prior year
10
Trend Analysis: Methods

We pool the cross sections for each year
1993–2000

We separate inpatient stays from
ER/observation bed stays

We aggregate ACSC admissions to MSA
and non-MSA regions within states

We tested whether the same relationships
exist in MSA and non-MSA subsets of the
data, and find they are significantly different
11
Trend Analysis: Methods

We first estimated a simple model with a
continuous time variable as the only
regressor and found a significant positive
trend.

We then stepped in beneficiary
demographics, health status and dummy
variables for 9 Census divisions

We interacted median age with time and
ER/observation bed stays with time to
examine whether there are time varying
associations
12
Trend Analysis: Time Trend
and Variation Explained
Inpatient
PVD Rate
Variable
MSA
Number of Observations
Inpatient
CHF Rate
MSA
NonMSA
MSA
NonMSA
66%
75%
82%
71%
408
400
408
400
–
Continuous Time Variable
Adjusted R2
NonMSA
Inpatient
COPD Rate
40%
408
38%
400
13
Trend Analysis:
Demographics
Inpatient
PVD Rate
Variable
MSA
NonMSA
Inpatient
COPD Rate
MSA
Inpatient
CHF Rate
NonMSA
MSA
NonMSA
+
+
–
Demographics
Proportion Died
Proportion Dual Enrolled in Medicaid
–
+
+
Proportion Men
+
Proportion Black
–
Proportion with ESRD
+
–
+
+
–
+
+
14
Trend Analysis:
Median Age and Time
Inpatient
PVD Rate
MSA
NonMSA
1993 Median Age *1993
–
–
1994 Median Age *1994
–
–
Variable
MSA
NonMSA
Inpatient
COPD Rate
Inpatient
CHF Rate
MSA
NonMSA
Median Age of Medicare FFS Population of each
study year interacted with Time Dummy for the year
+
1995 Median Age *1995
+
–
–
–
+
1996 Median Age *1996
+
–
–
–
+
1997 Median Age *1997
+
–
–
+
1998 Median Age *1998
+
–
–
+
1999 Median Age *1999
+
–
–
+
2000 Median Age *2000
+
–
–
+
15
Trend Analysis:
Health Status
Inpatient
PVD Rate
Variable
MSA
NonMSA
Inpatient
COPD Rate
MSA
NonMSA
Inpatient
CHF Rate
MSA
NonMSA
Health Status
Median PIP-DCG Risk Score
Lag Number of Diabetics
+
+
–
NI
+
NI
NI
NI
16
Trend Analysis:
ER Visit and Time
Inpatient
PVD Rate
MSA
NonMSA
1993 ER Visit Rate *1993
+
+
1994 ER Visit Rate *1994
+
+
1995 ER Visit Rate *1995
+
1996 ER Visit Rate *1996
+
+
Variable
MSA
NonMSA
Inpatient
COPD Rate
Inpatient
CHF Rate
MSA
NonMSA
Year Interacted with ER Visit Rate
+
1997 ER Visit Rate *1997
–
+
+
1998 ER Visit Rate *1998
–
+
+
+
1999 ER Visit Rate *1999
+
+
+
+
2000 ER Visit Rate *2000
–
+
+
+
17
Trend Analysis:
Population Migration
Inpatient
PVD Rate
Variable
MSA
NonMSA
Inpatient
COPD Rate
MSA
NonMSA
Inpatient
CHF Rate
MSA
NonMSA
Population Migration Patterns
Change in Size of Medicare FFS
Beneficiary Population from Current Year
to Prior Year
–
–
18
Trend Analysis:
Census Divisions
Inpatient
PVD Rate
Inpatient
CHF Rate
MSA
NonMSA
MSA
NonMSA
+
+
+
+
+
+
+
+
+
+
–
+
South Atlantic
+
+
East South Central
+
+
Variable
MSA
NonMSA
Inpatient
COPD Rate
Set of Dummy Variables for Each Census Division
(New England is Reference Division)
Middle Atlantic
+
East North Central
West North Central
–
West South Central
+
+
+
+
+
Mountain
–
–
–
Pacific
–
–
–
19
Conclusions: Trend Analysis

Positive trends in the raw rates over time
are substantially explained by demographicspecific factors

Little evidence of substitution of ER for
hospitalization for COPD and CHF; some
evidence of substitution for lower limb PVD

Rural areas that experienced outbound
migration experienced a decline in
admission for COPD and CHF
20
Conclusions: Trend Analysis

Observed variation in direction and strength
of relationship between explanatory factors
and selected chronic conditions suggests
that interventions employed to reduction
hospitalizations may have to be tailored to
the underlying condition

Unexplained geographic variation in
hospitalization rates for all three chronic
conditions remain
21