1 an autoregressive latent trajectory model of resident outcome improvement in nursing homes thomas...
TRANSCRIPT
1
An Autoregressive latent Trajectory An Autoregressive latent Trajectory Model of Resident Outcome Model of Resident Outcome
Improvement in Nursing HomesImprovement in Nursing Homes
Thomas T.H. Wan, Ph.D., M.H.S.Professor and Director
Public Affairs Doctoral ProgramCollege of Health and Public Affairs
University of Central Florida
May 18, 2005
2
IntroductionIntroduction The quality of nursing home care has been a
serious concern. Nursing homes are under increased scrutiny and
regulation due to reports of inadequate or deficient care
Little is known about the trajectories of resident outcomes that are directly related to nurse staffing, nursing care deficiency rating, and rehabilitation.
It is a challenging to develop a theoretically informed framework to guide the longitudinal analysis of nursing homes’ quality performance.
3
Theoretical FrameworkTheoretical Framework
Structure: Nurse Staffing
Process: Rehab Services
Nursing Care Adequacy
Outcome: Improved
Quality
Quality Domains & Their Relationships
4
Research QuestionsResearch Questions
What are the factors associated with the improvement of resident outcomes at the facility level?
Can previous levels influence later levels of quality performance measured by resident outcomes?
Do time-specific measures of nursing related variables influence the improvement of resident outcomes, while the lagged effects of quality measure and influences of contextual factors are simultaneously considered?
5
Purpose of the Study Using 7 waves of data with autoregressive
latent trajectory modeling, we assess the relationships of staffing, nursing care adequacy, and rehabilitative care to each wave of quality improvement, holding constant the contextual factors of nursing homes in the investigation of individual change trajectories.
The hypothesis is that, while controlling for facility and contextual factors, nursing homes with higher nurse staffing, more rehabilitative care, and fewer nursing care deficiencies will show improved resident outcomes.
6
Data & MethodsData & Methods
OSCAR (Online Survey, Certification, and Reporting System) CMS contracts state surveyors to
review and rate each nursing facility annually.
Contains hundreds of variables on every U.S. Medicare- or Medicaid-certified nursing facility.
Data are considered to be accurate reflections of actual deficiencies or citations.
7
MeasurementMeasurement
Panel data (1997-2003): N=11,197 Major outcome variable: Quality index
Change in incidence rate of adverse outcomes (pressure ulcers, physical restraints, and catheters)
Time-varying variables, measured annually Nurse staffing Nursing care deficiencies or citations Rehabilitation (% receiving rehab services)
Eight time-invariant covariates Bed size Private ownership (for profit =1; not for-profit = 0) Chain affiliation (chain = 1; non-chain = 0) Average acuity level % Medicare residents Region ( South = 1; non-South = 0) Urban (urban=1; rural=0) % elders (75+) in the county
8
A Balanced Score Card ApproachA Balanced Score Card Approach
Nursing Home Quality: Resident Outcomes A weighted aggregate measure of quality: Rate
change in the incidents such as developing pressure ulcers, having physical restraints, & on catheters per year
Declining incidents = a positive rate =better resident outcomes
Top 25 percentile of 11,197 facilities rated as high quality nursing homes (N=2,799)
10
Results
Cross-sectional analysis Trend analysis Longitudinal modeling
Cross-lagged model Parallel growth curve model Autoregressive latent trajectory model
11
1. Cross-Sectional Regression Analysis of Resident Outcomes in 1997
B SE Critical value b
QI-97 (Time 1) on Time-Invariant Predictors: (R 2 = .203) Bed size - .070 .002 -45.567 -.393* For-profit 1.161 .254 4.567 .040* Chain .207 .131 1.580 .014 Average acuity -.361 .031 -11.587 -.042* % Medicare 7.753 .792 9.793 .124* South -2.223 .440 - 5.049 -.043* Urban -.454 .267 - 1.670 -.015 % Elders(75+) .000 .000 -.141 -.001 Rehabilitation: Reh_97 .605 .748 .809 .009 Nurse staffing: NS_97 .012 .005 2.458 .020* Nursing care deficiencies : NCD_97 -.703 .115 - 6.137 -.050*
12
The Better Practice is associated with
Homes with a smaller bed size Being for-profit Caring for more Medicare residents Having residents with lower acuity levels Located in the region other than the
South Having a high level of nurse staffing Certified with lower frequencies of
nursing care deficiencies
13
2. Trends of Four Indicators (1997-2003)
0.525
0.55
0.575
0.6
0.625
0.65
0.675
ND
C
1996 1998 2000 2002 2004
year
3.25
3.5
3.75
4
4.25
4.5
4.75
NS
1996 1998 2000 2002 2004
year
16
17
18
19
20
21
RE
H
1996 1998 2000 2002 2004
year
-11
-10.5
-10
-9.5
-9
-8.5
-8
QI
1996 1998 2000 2002 2004
year
14
3. Autoregressive Model
ns97 ns98 ns99 ns00 ns01 ns02 ns03
ncd97 ncd98 ncd99 ncd00 ncd01 ncd02 ncd03
qi97 qi98 qi99 qi00 qi01 qi02 qi03
e1 e2e3 e4 e5 e6
e7 e8 e9 e10 e11 e12 e13
e14 e15 e16e17 e18 e19 e20
11111
1 1 1 1 1 1
1 1
reh97 reh98 reh99 reh00 reh01 reh02 reh03
1 1 1 1 1 1 1
e21 e22 e23 e24 e25 e26
1 1 1 1 1 1
Goodness of Fit Statistics:X2 = 14,311 with 315 degrees of freedomGFI = .925; AGFI = .903; TLI = .899;RMSEA = .063
15
4. Parallel Process Growth Model
I_qiS _qi
QI97 QI98 QI99 QI00 QI01 QI02 QI03
e1
1
e2
1
e3
1
e4
1
e5
1
e6
1
e7
1
1 11 11 1 1654321
I_ncd S_ncd I_reh S_reh
n cd 9 7
d 1
1
1
n cd 9 8
d 2
1
1
n cd 9 9
d 3
1
1
n cd 0 0
d 4
1
1
n cd 0 1
d 5
1
1
n cd 0 2
d 6
1
1
n cd 0 3
d 7
1
1
1 3 45 6
re h 9 7
d 8
1
1
re h 9 8
d 9
1
1
re h 9 9
d 1 0
1
1
re h 0 0
d 1 1
1
1
re h 0 1
d 1 2
1
1
re h 0 2
d 1 3
1
1
re h 0 3
d 1 4
1
1
1 2 3 45 6
R1 R2
A Latent Grow th Model of Q uality Im provem ent (1997-2003), predic ted by Nurs ing C are D efic ienc ies (NC D ) and R ehabilita tion (R E H)
1 1
GOF statistics:
Chi-square =5,369 (196 DF)
GFU = .954; AGFI = .958;
TLI = .958; RMSEA =.049
16
Findings of the Parallel Process GC Model
The QI growth curve shows a steady improvement in resident outcome.
The Nurse care deficiency growth curve shows a decline in 7 years.
The rehabilitation services use increased after 2000.
The change trajectories in resident outcomes are positively associated the increase in rehabilitation service use and negatively associated with the slope of nursing care deficiencies.
The increase in rehabilitative services enhances improved resident outcomes in 2000-2003, but not earlier years.
18
Predictors B SE
C V b
Intercept (I_qi) on Time-Invariant Predictors: (R 2 = .439)
Bed size -.067 .001 -48.888 -.501*
For-profit 1.076 .226 4.766 .049*
Chain .283 .116 2.446 .026*
Average acuity -.386 .059 -6.564 -.060*
% Medicare 7.648 .626 12.218 .163*
South -2.864 .386 -7.417 -.074*
Urban -.560 .235 -2.386 -.025*
% Elders(75+) .000 .000 -.180 -.002
Slope (S_qi) on Time-Invariant Predictors: (R 2 = .078)
Bed size .002 .000 5.758 .208*
For-profit -.174 .038 -4.559 -.113*
Chain -.032 .019 -1.626 -.042
Average acuity -.025 .011 -2.296 -.056*
% Medicare -.364 .120 -3.033 -.111*
South .132 .066 2.011 .049*
Urban -.002 .040 -.038 -.001
% Elders(75+) .000 .000 .046 .006
Each Period Rate on Time-Specific Predictors:
QI98 on (R2 = .504)
QI97 .240 .010 24.957 .274*
Reh98 .517 .497 1.041 .001
NS98 .006 .004 1.712 .009
NCD98 -.578 .088 -6.561 -.045*
QI99 on (R2 = .500)
QI98 .239 0.10 23.747 .256*
Reh99 .856 .446 1.921 .016
NS99 .004 .004 .790 .006
NCD99 -.286 .081 -3.519 -.024*
QI00 on (R2 = .530)
QI99 .245 .009 28.144 .263*
Reh00 1.910 .389 4.904 .037*
NS00 .007 .005 1.236 .009 NCD00 -.255 .070 -3.642 -.024*
QI01 on (R2 = .468)
QI00 .224 .009 24.463 .226*
Reh01 1.644 .383 4.297 .032*
NS01 .003 .005 .584 .004
NCD01 -.336 .023 -4.577 -.031*
QI02 on (R2 = . 552)
QI01 .234 .011 21.576 .259*
Reh02 1.179 .379 3.111 .025*
NS02 .007 .016 .437 .003
NCD02 -.307 .064 -4.758 -.031*
QI03 on (R2 = .581)
QI02 .227 .017 13.273 .239*
Reh03 1.366 .399 3.425 .031*
NS03 .031 .016 1.978 .014*
NCD03 -.170 .067 -2.539 -.017*
19
Findings of ALT Model The lagged effect of QI is an important
factor that should be statistically controlled in growth curve modeling.
The intercept factor, representing the baseline of quality, was well predicted by eight contextual and facility characteristics variables.
The slope or change trajectory of quality was only weakly predicted by them.
The improved quality in resident outcomes was associated with facilities having fewer nursing care deficiency citations than did their counterparts.
20
ConclusionsConclusions Complimentary results were revealed from
both cross-sectional & longitudinal analyses.
Parallel process growth curve modeling demonstrates its potential utility for policy research.
ALT is a power analytical approach to confirmatory analysis and data mining under a theoretically specified framework.
The best practice in nursing home quality is directly associated with reduced nursing care deficiencies.