A l l l Multilevel Analysis Using 2004 National Nursing

advertisement
Racial
Disparities in Nursing Home Quality of Care: A
t
Multilevel
l l l Analysis
l
Using 2004 Nationall Nursing
Home Survey Data
Yu Kang, PhD
Department
p
of Health Sciences and Administration
University of Michigan, Flint
Nancy A
A. Miller
Miller, PhD
Department of Public Policy
University of Maryland, Baltimore County
Prepared for the AcademyHealth ARM
Boston, MA June 28, 2010
1
Outline
•Introduction
•Theoretical
Th
i l Model
M d l andd Conceptual
C
l Framework
F
k
•Research Methodology
•Findings
•Summary of Findings and Policy Implications
•Study Limitations and Future Research Plans
2
Introduction
•Health care disparities persist; led to federal action.
•AHRQ
Q 7th report
p on national health care disparities
p
(2010).
•Disparities are evident in long term care (LTC).
•Increasing minority populations in U.S.
•Increase in LTC demand and unmet needs.
3
Theoretical Model and Conceptual
p
Framework
•Donabedian (1966, 1988) approach to define quality
measures
– To evaluate the quality of NH care by outcome and
process factors in the model,, while controlling
p
g for
the setting of care.
–O
Outcome approaches:
h physical
h i l conditions,
di i
andd
symptoms experienced in the NH stay.
– Process evaluations: receipt or utilization of certain
programs
g
or care, which include
services, p
hospitalization and daily use of restraints.
4
Research Methodology (1)
•The overall goal: to assess whether there are racial
disparities (Black relative to White residents) in nursing home
quality, using two quality of care measures.
•Hypothesis
yp
1: Compared
p
to White residents,, Black residents
experience poorer quality of care after controlling for
individual socioeconomic and health attributes, and facility
characteristics.
•Hypothesis 2: Compared to residents in facilities without
any Black
Bl k residents,
id t residents
id t in
i facilities
f iliti with
ith either
ith a higher
hi h
percentage (>median) or a lower percentage (<median) of
Blacks experience poorer quality of care after controlling for
individual socioeconomic and health attributes and facility
characteristics.
5
Research Methodology
gy ((2)) Data Set
Data Source: 2004 National Nursing Home Survey (NNHS)
-Current resident survey: current residents’ medical records by a
NH staff.
•A two
two-stage
stage probability design.
design
•Total sample of 14,017 residents.
Variables: health status,
status non
non-MDS
MDS, prescribed medications,
medications
•Variables:
and payment sources.
-Facility File
• NHs recorded by the CMS and private facilities licensed
by states generated 1,174 NH facilities.
– Common variable was acquired through the Research Data
Center at the NCHS (Hyattsville, Maryland).
-http://www.cdc.gov/nchs/nnhs.htm.
http://www cdc gov/nchs/nnhs htm
6
Research Methodology
gy ((3))
•Dependent Variables: Nursing Home Quality of Care
Meas res
Measures
– Daily Use of Restraints (NQF, CMS) (Process measure)
• Trunk,
Trunk limb
limb, or chair prevent rising
– Hospital Use (NQF) (Process measure)
• Hospital use in the past 90 days post admission
•Mechanism of choosingg Dependent
p
Variables:
– Measures used in other studies that are in NNHS.
– Pain management and orders or pain medication are
with sample that are too small.
7
Research Methodology
gy ((4))
Independent Variables
– At the individual level
• residence race (variable of interest), gender, age.
• the
th major
j paymentt source att admission.
d i i
• the primary diagnostic code (ICD-9) at admission and
ADLs.
ADLs
8
Research Methodology (5)
– At the facility level:
• Facility percentage of Black residents,
residents Black_0,
Black 0
Black_h, Black_l (variables of interest).
• Ownership, chain, number of beds, metropolitan
statistical area.
• Facilityy percentage
p
g of Medicaid and Medicare
patients.
• NH staffing.
• Specially trained staff and special services.
• Clusters of beds for special care programs.
• End-of-life programs.
9
Research Methodology (6)
– Use a median split (10.06%) to divide facilities with
different share of Black residents.
– Facility percentage of Black residents: trichotomized into
3 ordinal categories.
Variables
Facility% of Black Size
Residents
Black_0
l k
0%
N=7,511
Black_l
0.1-10.06%
N=1,726
Black_h
10.07% and up
N=3,109
10
Research Methodology
gy ((7)) Analytical
y
Approach
pp
Nonlinear Mixed Model (Multilevel Model):
•Dichotomous outcome variable (0,1)
ŋŋijj =log
g [[Pijj / ((1 – Pijj )]
ŋij = β0j + β1j xij
β0j = γ00 + γ01zj + u0j
β1j = γ10
ŋij
ij = γ00
00 + γ01zj
01 j + γ10xij
10 ij + u0j
0j
((1))
(2)
(3)
(4)
(5)
11
Findings (1) Descriptive Analyses Table 1
Total
F
Freq
Black
%
12507
F
Freq
White
%
1258
F
Freq
%
11249
Restraint***
Restraint
722
5.81
82
6.52
640
5.69
Hospadm***
886
7.13
111
8.92
775
6.93
1
755
85.41
89
80.18
666
86.16
2
105
11.88
16
14.41
89
11.51
3
17
1 92
1.92
5
4 50
4.50
12
1 55
1.55
4
4
0.45
0
0.00
4
0.52
6
3
0.34
1
0.90
2
0.26
Hospnum
12
Findings (2) Descriptive Analyses Table 2
Total
Freq
0%
%
12436
Restraint***
Freq
≤
10.06
%
%
7511
Freq
>
10.06
%
%
1726
Freq
%
3109
714
5.82
369
4.93
103
6.00
242
7.86
Hospadm*** 877
7.15
526
7.03
111
6.45
240
7.82
Hospnum
1
747
85.37
455
86.67
96
87.27
196
81.67
2
104
11 89
11.89
58
11 05
11.05
13
11 82
11.82
33
13 75
13.75
3
17
1.94
7
1.33
1
0.91
9
3.75
4
4
0.46
4
0.46
0
0.00
0
0.00
6
3
0.34
1
0.19
0
0.00
2
0.83
p<.05,, ** p<.01,
p
, *** p<.001
p
Notes: p
13
Findings (3) Summary of Descriptive Analyses
•At the individual level, Black residents were more likely to
experience daily restraints and hospitalizations.
•Blacks were more likely to use Medicaid as their primary
payment source and experience severe ADLs.
•Residents in facilities with a higher proportion of Blacks
were more likely
lik l to
t experience
i
restraints
t i t andd hospitalizations.
h it li ti
Individuals in facilities with a higher share of Blacks were
•Individuals
more likely to be in a facility that was chain, for-profit, with
larger bed size, located in a metropolitan area and with a
higher percentage of Medicaid beneficiaries
beneficiaries.
14
Findings (4) Multivariate Analyses
Variables
V
i bl off
Interest
Individual
Confounders
Restraint
OR 95% CI
Hospadm
OR 95% CI
Bl k
Black
0 689 (0.468,
0.689
(0 468 1.016)
1 016)
1 231 (0.898,
1.231
(0 898 1.696)
1 696)
Black_l
1.149 (0.803, 1.645)
0.960 (0.722, 1.278)
Black_h
1.969*** ((1.439,, 2.693))
1.013 ((0.779,, 1.318))
Age at Admin
1.011 (1.001, 1.021)
0.846*** (0.976, 0.991)
Male
1.018 (0.821, 1.263)
1.321** (1.114, 1.565)
Total ADLs
4.613*** (3.731, 5.704)
1.163*** (1.088, 1.242)
Medicaid
0.948 (0.767, 1.172)
1.059 (0.884, 1.269)
Medicare
0 911 (0
0.911
(0.737,
737 11.125)
125)
1 532*** (1.285,
1.532***
(1 285 11.827)
827)
Blood pressure 1.322 (0.787, 2.220)
0.737 (0.463, 1.172)
CVD
0.868 ((0.679,, 1.109))
1.193 ((0.987,, 1.442))
Stroke
2.191 (0.829, 5.790)
2.053* (1.000, 4.215)
Dementia
1.538** (1.171, 2.020)
0.727* (0.538, 0.983)
Cancer
0.824 (0.375, 1.811)
1.307 (0.764, 2.235)
15
Findings
g ((5)) Multivariate Analyses
y
Restraint
OR 95% CI
Hospadm
OR 95% CI
1.103 (0.842, 1.446)
0.842 (0.682, 1.039)
0.846 (0.656, 1.089)
0.766** (0.628, 0.936)
Metro
0.999 (0.729, 1.368)
0.689** (0.543, 0.875)
Micro
1.022 (0.715, 1.460)
0.765 (0.585, 1.001)
Bed200
1 040 (0.597,
1.040
(0 597 1.812)
1 812)
0 713 (0.435,
0.713
(0 435 1.169)
1 169)
Medicaidp20
1.908* (1.057, 3.443)
1.586* (1.001, 2.513)
Medicaidp60
2.196* (1.206, 3.998)
1.730* (1.081, 2.770)
Mcarep20
0.949 (0.639, 1.409)
0.815 (0.585, 1.135)
RNFTE BC
0.979* (0.959, 0.999)
0.987 (0.972, 1.003)
RN oneyear%
0.996 (0.993, 1.000)
0.996* (0.993, 0.999)
Alzheim Bed
0.900 (0.688, 1.176)
1.047 (0.849, 1.293)
Facility
Ownership
Confounders Chain
N
Notes:
p<.05,
05 ** p<.01,
01 *** p<.001
001
16
Significance of Findings
This Study
Prior Studies
Individuals in facilities with a higher share of
Blacks were associated with greater odds of
restraints, and the effects was significant
Blacks were more likely to be placed in NHs
with more serious deficiencies. Literature on
racial disparities often failed to address
individual and facility racial effects
simultaneously in a single multivariate study.
Medicaid as a payment source was not
associated with any quality measure; whereas
using Medicare predicted greater odds of
hospitalizations.
Medicaid beneficiaries were more likely to
experience quality deficiencies, such as lower
professional nursing staff rate, inappropriate
medication prescriptions or hospitalization.
hospitalization
Being a resident in a facility with higher
ppercentage
g of Medicaid ppatients was
associated with greater odds of experiencing
restraints and hospitalizations.
NH facilities with a larger share of Medicaid
beneficiaries were challenged in terms of
quality of care, especially with fewer nursing
staff, lower occupancy rate and more health
deficiencies.
Residents in a facility with a higher number of RN
FTEs devoted to bedside care or a higher
percentage of RNs that served for more than a year
was associated with a lower odds for restraints or
hospitalizations respectively.
The level of nursing staff, such as RN hours
and CNA hours or percentage of LPN or RN
expense, were positively associated with better
quality of care.
17
Policy Implications
•Racial Concentration at the Facility Level
–R
Racial
i l concentration
i at the
h facility
f ili andd not race at the
h
individual level is associated with poorer quality of care.
– Although
g individual race was associated with increased
use of restraints and hospitalization descriptively, this
was not significant in the multivariate findings.
•Financing Long-Term Care
– The Medicaid Payment: facility level.
– The Medicare Payment: individual level.
•Government Guidance and Regulations
– Nursing Stuffing's positive impact on NH qualities
18
Study Limitations and Future Studies
Limitations
Future Studies
In 2004 NNHS:
-small sample size in ethnic groups;
-lack of discharge information.
Other datasets:
-MDS.
In statistical methods:
-NLMIXED model
Other statistical model:
-GLIMMIX vs. NLMXED.
Limited quality measures.
Other quality measures:
-pressure ulcer.
-flu vaccinations.
The effect of RN FTE to bedside care
was significant but trivial.
Other nursing staff measures used by CMS:
total nursing hours per patient day.
19
Download