Journal of Gerontology: MEDICAL SCIENCES
Cite journal as: J Gerontol A Biol Sci Med Sci
© The Author 2010. Published by Oxford University Press on behalf of The Gerontological Society of America.
All rights reserved. For permissions, please e-mail: email@example.com.
including those dually eligible (medicare and medicaid)
certified by states as requiring long-term care (LtC). Federal
and state governments are expanding access to community-
based LtC because of this preference and as a result of a
landmark supreme Court ruling (supreme Court 1999
Olmstead v. L.C. decision) that individuals have the right to
live in community settings if possible and desired, rather
than be institutionalized. subsequent lawsuits have com-
pelled states to expand access to community-based LtC, and
provision of federal resources has assisted states in rebalanc-
ing their services from institutional to community care (1).
Furthermore, different forms of community-based care have
evolved in the hope that more people with different LtC
needs can be served without increasing costs (2–4).
two alternative community-based care programs are the
Program of all-inclusive Care for the elderly (PaCe), and
ommunity-based care is preferred to nursing
home (nH) care by most older and disabled americans,
aged and disabled home and community-based care under
1915(c) waiver provisions. PaCe is a prepaid, dually capi-
tated, community-based model in which care for older dis-
abled participants—certified by states as eligible for nH-level
care—is integrated by interdisciplinary teams based in day
centers (5,6). under full financial risk, PaCe provides all
necessary acute, primary, consultative, chronic, and palliative
care, as well as supportive center, home, institutional, trans-
portation, and other services, including meals and caregiver
support, to facilitate participants’ remaining in the commu-
nity. PaCe became a medicare provider and a state medicaid
option under the balanced budget act of 1997. Programs re-
ceive capitated payments from medicare on a diagnosis-
based, frailty-adjusted formula, and from medicaid at fixed,
annually negotiated rates specific for states/localities. as
of september 2009, there were 71 approved, independent
PaCe programs (four pending), in 31 states. in the past year,
the Centers for medicare and medicaid services approved
Five-year survival in a Program of all-inclusive Care For
elderly Compared With alternative institutional and
Home- and Community-based Care
darryl Wieland,1,2 Rebecca boland,2 Judith baskins,2 and bruce Kinosian3,4
1division of Geriatrics, department of medicine, university of south Carolina school of medicine, Columbia. 2division of Geriatrics
services, Palmetto Health Richland, Columbia, south Carolina. 3Center for Health equity Research and Promotion, Philadelphia Va
medical Center, Pennsylvania. 4department of medicine, university of Pennsylvania, Philadelphia.
address correspondence to darryl Wieland, Phd, mPH, 3010 Farrow Road, 300a, Columbia, sC 29203. email: firstname.lastname@example.org
Background. Community-based services are preferred to institutional care for people requiring long-term care (LtC).
states are increasing their medicaid waiver programs, although Program of all-inclusive Care For elderly (PaCe)—
prepaid, community-based comprehensive care—is available in 31 states. despite emerging alternatives, little is known
about their comparative effectiveness.
Methods. For a two-county region of south Carolina, we contrast long-term survival among entrants (n = 2040) to an
aged and disabled waiver program, PaCe, and nursing homes (nHs), stratifying for risk. Participants were followed
for 5 years or until death; those lost to follow-up or surviving less than 5 years as on august 8, 2005 were censored.
analyses included admission descriptive statistics and Kaplan–meier curves. to address cohort risk imbalance, we em-
ployed an established mortality risk index, which showed external validity in waiver, PaCe, and nH cohorts (log-rank
tests = 105.42, 28.72, and 52.23, respectively, all p < .001; c-statistics = .67, .58, .65, p < .001).
Results. Compared with waiver (n = 1,018) and nH (n = 468) admissions, PaCe participants (n = 554) were older,
more cognitively impaired, and had intermediate activities of daily living dependency. PaCe mortality risk (72.6% high-
to-intermediate) was greater than in waiver (58.8%), and similar to nH (71.6%). median nH survival was 2.3 years.
median PaCe survival was 4.2 years versus 3.5 in waiver (unstratified, log rank = .394; p = .53), but accounting for risk,
PaCe’s advantage is significant (log rank = 5.941 (1); p = .015). Compared with waiver, higher risk admissions to PaCe
were most likely to benefit (moderate: PaCe median survival = 4.7 years vs waiver 3.4; high risk: 3.0 vs 2.0).
Conclusion. Long-term outcomes of LtC alternatives warrant greater research and policy attention.
Key Words: Comparative effectiveness research—Risk stratification—Long-term care—dual eligibles.
Received February 8, 2010; Accepted March 5, 2010
Decision Editor: Luigi Ferrucci, MD, PhD
by guest on March 31, 2010
WIELAND ET AL.
21 new plan applications and eight new state Plan amend-
ments (raising the number of states in which PaCe programs
may develop to 38). the PaCe census in January 2009 was
in contrast to PaCe, 48 states and the district of Colum-
bia have more widely expanded home- and community-
based care under 1915(c) medicaid waivers (1,4,7). these
programs vary across states and local areas, even among
aged and disabled waivers, but the latter generally provide
clients with case managers who receive per diem payments.
Case managers assess client personal care needs and can
authorize in-home personal care and other supportive ser-
vices paid by medicaid. the broader medicaid home- and
community-based population (ie, those receiving services
through various waiver, mandatory home health, and per-
sonal care services programs) has been growing at a rate of
7% a year since 1999 (1). nearly 600,000 americans were
enrolled under aged and disabled waivers in 2005 (8).
although aged and disabled waiver programs are widely
recognized as being limited to providing supportive care
and lacking the comprehensive, integrated, and interdisci-
plinary team services of PaCe, some decision makers view
the waiver option as a lower cost alternative not only to nHs
but also to PaCe for older, disabled people certified by
states for nH eligibility. Less recognized are differences
among community programs in client level-of-care needs
and risk, and outcomes, to justify program input (cost) dif-
ferences. this underrecognition affects government plan-
ners and decision makers, providers, and ultimately patients
and caregivers confronted with selecting among LtC
options, where institutional- and community-based LtC
options coexist in a state or local area.
Within the context of a program of research to compare
the long-term effectiveness of care consequent to alternative
LtC placements, this initial study has three objectives. First,
on a quasiexperimental, intent-to-treat basis, we character-
ize long-term survival for three LtC admission cohorts,
where follow-up has been sufficient to achieve estimates of
median survival and meaningfully assess trajectories (paral-
lel, convergent, and divergent) for the program cohorts.
second, we evaluate external validity of an established mor-
tality risk index in the contrasted program populations.
third, we determine program effects on survival stratified a
priori by level of risk.
in south Carolina (sC), medicaid waiver and nH care is
available statewide, and those programs may admit clients/
residents as young as age 18. during our study, PaCe was
available only in a two-county catchment; here and nation-
ally, PaCe is limited to admitting only participants aged
55 years or more. because we aimed to compare program
entrants and mortality outcomes only in a population eligi-
ble for admission to any of the three main LtC options, we
limit aged and disabled and nH subjects to those living in
the PaCe catchment, age 55 years or more. in sC as else-
where, most PaCe and medicaid nH entrants remain until
they die. in contrast, although many aged and disabled
waiver clients die after long enrollments, others may be dis-
charged if they require higher levels of care (eg, to nH or
PaCe), or no longer meet nH level-of-care criteria on
follow-up. Participants are assigned to program cohorts
according to their initial LtC admission status, as are their
subsequent vital events.
the central south Carolina PaCe—Palmetto senior Care
(PsC)—has operated as many as six day centers in Richland
and Lexington counties, sC, since its establishment as an
original on Lok replication site in 1988. PsC’s average
daily census was in the 350–400 participant range for most
of the study period (1998–2005).
the aged and disabled waiver program in sC (now called
Community Choices [CC]) began in 1983 after a 3-year pi-
lot, and now is one of the several medicaid community-
based waiver programs operated by the state Community
Long-term Care agency. Like PsC, CC is available for
adults qualifying for medicaid and certified as nH eligible
but who prefer to receive services in the community. through
case management and an individualized package of support-
ive services, CC aims to enable clients to remain at home at
a cost to medicaid that is substantially less than the cost of
institutional care (9). statewide, CC case managers (about
5% of CC expenditures) assist clients in selecting among
available services: over three quarters of CC spending is for
personal and attendant care and companion services, and
adult day health care (included skilled nursing at the cen-
ters). Remaining service expenditures were for supplies and
equipment, home delivered meals, home modifications, per-
sonal emergency response systems, and chore services (10).
according to the state, “[nHs provide] nursing, therapy,
and personal care services to individuals who do not require
acute hospital care, but whose mental or physical condition
requires services that are above the level of room and board
and can be made available through licensed, certified and
contracted institutional facilities” (11). during the study pe-
riod, sC maintained a stable medicaid nH bed capacity in
and around Columbia.
Single Point-of-Entry System
all medicaid recipients entering PsC, CC, or nHs must
be certified as meeting criteria for a nH level of care. the
state employs regional teams to conduct comprehensive
preadmission assessments of LtC applicants. through this
process, the teams produce written evaluations of appli-
cants’ medical, psychosocial, functional, environmental,
and support system and service needs, and determinations
of medical necessity for LtC, based upon meeting specific
skilled or intermediate service and/or functional support
by guest on March 31, 2010
FIVE-YEAR SURVIVAL IN LONG-TERM CARE
criteria (12). these “Form 1718” assessments are standard-
ized, provide data for initial care planning in CC and other
community-based programs, and crosswalk into admission
Resident assessment instrument (Rai) fields for those
placed in nHs (13).
Analytic Data Set
We constructed a data set to represent a medicaid LtC
admission cohort entering the two community programs
and institutional care. data describing the medical, psycho-
social, functional, environmental, and social supports of
entrants (see table 1 for selected variables) were derived
from state Form 1718 records. Vital status at follow-up was
determined for each entrant from review of repeat data-
PaCe 1.0 (the PaCe demonstration minimum data set),
and 1718 assessments for those maintaining enrollment
in community programs and review of state vital statistics
Participants, Risk Stratification, and Statistical Analysis
Participants (n = 2,040) were older (≥55) residents of two
counties in south Carolina admitted between 1998 and 2003
to CC (n = 1018), PsC (n = 554) and nHs (n = 468). Par-
ticipants were followed until death or 5 years postadmis-
sion. those lost for the event or surviving with less than
5-year follow-up on august 8, 2005 were right censored.
analyses included contrasts among the three entry cohorts
using descriptive statistics. Cohort survival comparisons—
overall and stratified by mortality risk—were examined us-
ing Kaplan–meier curves and tested with log-rank statistics
(sas Version 9.2; sas institute, inc., Cary, nC).
mortality risk at admission was assessed using the PaCe
Prognostic index (PPi) (14). designed to predict mortality
in community-living frail elderly people, it was developed
(n = 2,232) and validated (n = 1,667) in cohort study of
12 PaCe sites (including PsC) using baseline demographic,
functional, and disease risk factors derived from dataPaCe.
the PPi was adequately calibrated and showed good dis-
crimination (area under the curves = 0.66 and 0.69 for de-
velopment and validation cohorts). scores ranged 0–18,
with higher scores indicating greater risk. PPi risk factors
(and index weights) included male sex (2 points); age 75–84
(2); age 85 years or older (3); dependence in toileting
(1); dependence in dressing, partial (1), and full (3); malig-
nant neoplasm (2); congestive heart failure (CHF) (3);
chronic obstructive pulmonary disease (1); and renal failure/
For risk stratification, we used the PPi cutpoints em-
ployed by Carey and colleagues (14) to designate low (PPi:
0–3), moderate (4–5), and high-risk (≥5) participants. be-
cause the PPi was developed in a PaCe population, we
evaluated its external validity for 5-year mortality in each
program. Calibration and discrimination were evaluated us-
ing stratified Kaplan–meier curves with log-rank tests, as
well as sas PRoC LoGistiC c-statistics, Hosmer–Leme-
show partitions, and tests of goodness-of-fit for survival at
the end of follow-up.
PsC admissions compared with CC and nH entrants
(table 1) were older (77.2 ± 0.42 vs 74.5 ± 0.32 and 74.8 ±
0.51), more likely african american (70.6% vs 49.1% and
45.7%), and less educated (high school or more: 27.1% vs
33.5% and 33.1%). as a proportion of admissions, men
comprised less than one quarter of the CC cohort versus
over one-third among PsC and nH entrants. nH admissions
were less likely to be married than CC and PsC admissions.
diagnoses of CHF and diabetes were more prevalent
among CC admissions, although heart disease, renal failure/
insufficiency, cancer, stroke, and dementia were more prev-
alent among PsC admissions. Proportions with activities of
daily living dependencies and incontinence were consistently
lowest in the CC cohort, highest in the nH cohort, with in-
termediate values among PsC entrants. PsC participants
were more likely to manifest behavioral problems.
Five-year unstratified program cohort survival curves are
displayed in Figure 1. the trajectories are significantly dis-
tinct over 5 years (log-rank test = 40.267 (2); p < .001), with
the exception that PsC and CC curves converge at about
4.5 years. median survival of the nH cohort was 2.3 years.
median survival in PsC was 4.2 years versus 3.5 in CC, but
the paired, unstratified trajectories are not significantly dif-
ferent (log rank = 0.394 (1); p = .53).
We evaluated the external validity of the PPi as a mea-
sure of mortality risk in each program. survival curves
showed divergent trajectories over most of the 5-year period
(Figure 2a–C). in logistic regression, the PPi showed ade-
quate discrimination in CC and nH cohorts (c-statistics = .67
and .65, respectively, each p < .001) comparable with the
results of Carey and colleagues (14). the PsC risk-stratum
curves (Figure 2b) show good discrimination through the
fourth follow-up year, but the low- and moderate-risk
curves begin to converge in the fifth year. Here, fit for the
5-year outcome was marginal (Hosmer–Lemeshow c2 =
1.942; p = .164), as the PPi began to overpredict deaths in
the moderate-risk and underpredict in low-risk strata; thus,
discrimination for 5-year PsC survival was lower (c = .58),
but still significant (p = .002).
admission mortality risk is significantly greater in PsC
than in the CC cohort, with mean values in the high- and
moderate-risk range, respectively (table 1; PPi, means ±
SEM: 5.29 ± 0.119 vs 4.29 ± 0.074; p < .001). nearly identi-
cal to the PsC risk index mean was the mean PPi among
nH admissions (5.28 ± 0.118). stratifying the program
cohorts by level of risk, the proportions of moderate to
high mortality risk participants among PsC (72.6%) and
nH admissions (71.6%) are greater than in the CC cohort
by guest on March 31, 2010
WIELAND ET AL.
With risk stratification taken into account, the PsC 5-year
survival advantage over CC is statistically significant (log
rank = 5.941 (2); p = .015). stratum-specific analyses sug-
gest that PsC’s survival advantage relative to CC occurs
among moderate- and high-risk admissions (Figure 3a–C).
median survival among moderate-risk admissions to PsC
was 4.7 years compared with 3.4 years in CC (log rank =
3.08 (1); p =.079). among the high risk, PsC and CC me-
dian survival was 3.0 and 2.0 years, respectively (log rank =
6.53 (1); p = .01). in all-risk strata, CC and PsC survival
curves converge in the fifth year.
at admission into LtC, PsC participants were at signifi-
cantly higher mortality risk than CC clients, as well as bear-
ing a greater overall burden of cognitive impairments and
disabilities in this study of medicaid community LtC pro-
grams in central sC. stratifying for mortality risk, PaCe
participants had a substantial long-term survival advantage
compared with aged and disabled waiver clients into
the fifth year of follow-up. that the benefit seemed most
apparent in moderate- to high-risk admissions suggests the
table 1. Characteristics of Persons admitted to Community Choices (aged and disabled Waiver Program), Palmetto senior Care (PaCe), and
nursing Homes, Richland and Lexington Counties, south Carolina, 1998–2003
admission VariablesCommunity Choices (n = 1,018) Palmetto senior Care (n = 554) nursing Homes (n = 468)
african american (%)
education ≥ high school (%)
Current diseases/conditions (%)
Continent of bladder
Continent of bowel
activities of daily living dependence (%)
behavioral problems (%)
socially inappropriate behavior
74.5 ± 0.32
77.2 ± 0.42
74.8 ± 0.51
4.29 ± 0.074 5.29 ± 0.1195.28 ± 0.118
Notes: CHF = congestive heart failure; CoPd = chronic obstructive pulmonary disease; PaCe = Program of all-inclusive Care For elderly; PPi = PaCe Prog-
* Risk factors included in the PPi (14).
Figure 1. overall survival (Kaplan–meier) trajectories, by program cohort
(Community Choices, Palmetto senior Care, nursing home). Log-rank (mantel–
Cox) test = 40.27 (df = 2); p < .001.
by guest on March 31, 2010
FIVE-YEAR SURVIVAL IN LONG-TERM CARE
Figure 2. (A) Community Choices cohort survival, stratified by mortality
risk. Log-rank test = 105.42 (2); p < .001. (B) Palmetto senior Care cohort
survival, stratified by mortality risk. Log-rank test = 28.72 (2); p < 0.001.
(C) nursing Home cohort survival, stratified by mortality risk. Log-rank test =
52.23 (2); p < .001.
Figure 3. (A) Program cohort survival, low mortality risk (PPi 0–3).
all-program log rank = 7.47 (2); p = .024; CC versus PsC = 0.41 (1); p = .425.
(B) Program cohort survival, moderate mortality risk (PPi 4–5). all-program
log rank = 6.497 (2): p = .039; CC versus PsC = 3.08 (1); p = .079. (C) Program
cohort survival, high mortality risk (PPi >5). all-program log rank = 30.099 (2):
p < .001; CC versus PsC = 6.53 (1); p = .01. Notes: CC = Community Choices;
PPi = Program of all-inclusive Care For elderly Prognostic index; PsC =
Palmetto senior Care.
by guest on March 31, 2010
WIELAND ET AL.
particular importance of an integrated, team-managed med-
ical home for the older, more disabled participants more
commonly admitted to PaCe.
our study has several limitations. First, it employs sec-
ondary analysis of clinicoadministrative information col-
lected for other purposes (15). second, baseline risk
differences between program cohorts in this quasiexperi-
mental study were manifest, raising the challenge of risk
adjustment. Here, we examined the validity of the PPi in
our population and performed simple risk stratification, in-
stead of fitting (and probably overfitting) a multivariable
model of program-related survival using miscellaneous ad-
mission predictors available in the analytic data set. Risk
adjustment issues aside, results concerning comparative
program mortality outcomes reflect specifically local condi-
tions, and may not be observed everywhere. For example,
although nationally PaCe comprises comprehensive, inte-
grated care with a strong medical management component,
the quality and accessibility of primary, consultative, emer-
gency, and acute care for waiver clients may vary consider-
ably. the extent to which medicaid beneficiaries in such
waiver programs enroll in medical homes, special needs
plans, and other programs unavailable in sC at the time of
study could affect outcomes for those groups.
moreover, we were unable to address the important issue
of selection bias (16). much of the large survival advantage
of both CC and PsC over nH (Figure 3a–C) very likely
reflects adverse selection to the latter, as well as probable
ceiling effects of the PPi in nH admissions. thus, we have
not emphasized the survival differences observed between
either of the community cohorts and the nH group in over-
all or stratified analyses. We must also assume there is se-
lection between CC and PsC programs that may influence
their outcomes independent of process and/or quality differ-
ences between PsC and CC care. in future research, we
hope to refine our modeling of long-term outcomes by in-
corporating time-varying covariates as we add annual reas-
sessment information to the data set. this will involve
integration of Rai information for reassessments of nH
residents presently missing. Furthermore, we plan study of
the single point-of-entry process in sC, with a view to iden-
tifying instrumental variables affording us some under-
standing and control of selection bias.
the small literature concerned with the comparative ef-
fectiveness of alternative LtC placements usually limits
follow-up to 2 years or less and suffers from other limita-
tions (4). Few states systematically evaluate expenditures
and outcomes across LtC programs over periods longer
than 1 year. thus, questions linger regarding the longer-
term value of PaCe relative to home- and community-based
waiver or nH placements or among LtC programs gener-
ally. our results suggest that states should make necessary
investments in research and data infrastructure to evaluate
emerging LtC options, and make planning and allocation
decisions based in part on evidence of value for different
levels of need and risk.
the paper was presented at the academy Health annual Research meeting,
Chicago, June 2009.
1. Harrington C, ng t, Kaye sH, newcomer R. Home and Community-
Based Services: Public Policies to Improve Access, Costs, and Qual-
ity. san Francisco, Ca: university of California san Francisco Center
for Personal assistance services; 2009.
2. Grabowski dC. the cost-effectiveness of noninstitutional long-term
care services: review and synthesis of the most recent evidence. Med
Care Res Rev. 2006;63(1):3–28.
3. Kaye Hs, LaPlante mP, Harrington C. do noninstitutional long-term
care services reduce medicaid spending? Health Affairs. 2009;28(1):
4. Golden aG, Roos ba, silverman ma, beers mH. Home and community-
based medicaid options for dependent older Floridians. J Am Geriatr
5. Wieland d. the Program of all-inclusive Care for the elderly [PaCe].
in: R schulz, Ls noelker and K Rockwood (eds), et al. The Encyclo-
pedia of Aging 24th ed. new york: springer; 2006:973–975.
6. national PaCe association(online). Web site. www.npaonline.org.
accessed december 3, 2009.
7. Centers for medicare and medicaid services (online). medicaid state
Waiver Program demonstration Projects—General information.
Web site. www.cms.hhs.gov/medicaidstwaivprogdemopgi/05
_hcbswaivers-section1915(c).asp. accessed december 4, 2009.
8. Houser a, Fox-Grage W, Gibson mJ. Across the States: Profiles in
Long-Term Care and Independent Living, 8th ed. Washington, dC:
aaRP Public Policy institute; 2009.
9. south Carolina department of Health and Human services (online).
bureau of Long term Care and behavioral Health services Waiver
management. Web site. www.scdhhs.gov/insidedhhs/bureaus
october 10, 2009.
10. Kasten J, Coriell J, burwell b, eiken s. Long-Term Care in South
Carolina. Report to the SC Department of Health and Human Ser-
vices. new york: thompson Reuters Healthcare; 2008.
11. south Carolina department of Health and Human services (online). divi-
sion of Community Facilities. Web site. www.scdhhs.gov/divCommFac
.asp. accessed october 8, 2009.
12. south Carolina department of Health and Human services (online). south
Carolina assessment & Level of Care manual for medicaid-sponsored
Long-term Care services. Web site. www.dhhs.state.sc.us/internet/pdf
/manuaLLoC_1.pdf. accessed december 2, 2009.
13. south Carolina department of Health and Human services (online).
South Carolina Long-Term Care Assessment Form SCDHHS
Form 1718 (versions 7/1/98; 10/1/02). Web site. www.scdhhs.gov
accessed december 2, 2009.
14. Carey eC, Covinsky Ke, Lui L-y, eng C, sands LP, Walter LC.
Prediction of mortality in community-living frail elderly people with
long-term care needs. J Am Geriatr Soc. 2008;56: 68–75.
15. stewart dW and Kamins ma (eds). Secondary Research: Information
Sources and Methods, 2nd ed. applied social Research methods
series, no. 4. beverly Hills, Ca: sage Publications; 1992.
16. Pizer sd. an intuitive review of methods for observational studies
of comparative effectiveness. Health Serv Outcomes Res Methodol.
by guest on March 31, 2010