Expedited Medicaid Restoration - Academic and Health Policy

Report
Expedited Medicaid Restoration:
Introduction & Overview
Joe Morrissey
University of North Carolina
6th Annual Academic and Health Policy
Conference on Correctional Health
Chicago – March 22, 2013
Expediting Medicaid Benefits
• Focus: Comparative costs of policies restoring
Medicaid benefits prior to release from state
prisons
• Target Group: Persons with severe mental
illness (SMI)
• Study Sites: Washington and Connecticut
2
Acknowledgements
• Funding
• NIMH Research Grant “Community Reentry of Persons with Severe
Mental Illness Released from State Prisons” (MH086232)
• Assistance
• 10 state agencies in WA and CT
• Hsiu-Ju Lin, Connecticut Department of Mental Health & Addiction
•
•
•
•
Services
Shirley Richards and Jennifer Jolley, University of North Carolina at
Chapel Hill
Jeffrey Swanson and Allison Robertson, Duke University
David Mancuso, Division of Research and Data Analysis, Washington
State Department of Social and Health Services
Colleen Gallagher and Dan Bannish, Connecticut Department of
Correction
3
Seminar Presenters*
1.
2.
3.
4.
Joe Morrissey – Introduction & Overview
Linda Frisman – Outcomes in Connecticut
Gary Cuddeback – Outcomes in Washington
Marisa Domino – Cost Findings in Washington
* No conflicts of interest to declare
4
Medicaid and Community Reentry
• Medicaid is the single largest payer of mental health
services for persons w SMI in the US today
– Council of State Government suggests Medicaid is critical to
successful community reentry
– Without sustaining health & welfare benefits justiceinvolved persons with mental illness will be unable to “break
the cycle of recidivism” (Bazelon Center)
• Federal regulations require suspension or termination
of Medicaid benefits after 30 days of incarceration or
hospitalization
• A number of states have introduced policies to
expedite the restoration of Medicaid benefits prior to
release from prison and other institutions
5
What’s the Evidence Base for
Expedited Medicaid?
• Wenzlow et al. (2011) small study (N= 77) of expedited
Medicaid in 3 Oklahoma prisons
– Increased Medicaid enrollment on day of release by 15 percentage
points vs. controls (p=.012)
– Increased Medicaid mental health service use by 16 percentage
points (p=.009) vs. controls at 90 days post release
• Morrissey, Cuddeback et al. (2006, 2007) showed that
persons w SMI in jail with Medicaid at release had:
– Quicker service access & more community service use (p<.001)
– Fewer re-arrests, more days in community (p<.01)
• To date, no comparative study of prison re-entry nationally
or in different states, no studies with large sample sizes
6
Current Study
• Response to NIMH solicitation in 2008 for use
of administrative data to address state-level
behavioral health policies
• NIMH required comparison of at least two
states
• Assembled a research team from UNC, CTDMHAS, and Duke that had prior experiences
using administrative data to address state
policy issues in Washington & Connecticut
7
Expedited Medicaid
• Policy start date
– Connecticut: April 2005
– Washington: January 2006
• Benefits affected by restoration
– Federal Medicaid via SSI/SSDI eligibility
» (i.e., Aged, Blind, Disabled)
– State Medicaid, a less generous benefit often
used as an initial or transitional status pending
Federal Medicaid restoration
» SAGA in Connecticut
» GAU in Washington
8
Policy Implementation
Connecticut in 2005
• DOC only
• Discharge Planners (6) working
for Correctional Managed
Health Care & based in
correctional facilities complete
paperwork to apply for
Medicaid prior to release & fax
to state Medicaid agency
• Entitlement specialists (2) based
at state Medicaid agency
process applications
• Daily e-feed of population list
results in benefits being
“switched on”
Washington in 2006
• DOC, jails, state hospitals
• Two step process
① Referral: DOC staff
identified inmates and
prepared applications
prior to release
② Approval: Following
release, inmate had to
appear at local Community
Service Office to activate
application
• Legislature funded and
distributed 14 FTE Community
Service Officers statewide to
prioritize expedited cases
9
Other Differences b/w States
Connecticut
Washington
• Unitary corrections system:
State DOC operates prisons
and jails
• Much of Medicaid was fee-forservice in the study period
• State Administered General
Assistance (SAGA) covered
services at a similar rate
• Virtually no one is denied
benefits
• Even prisoners not expedited
had 2-4 weeks of Rx at the
time of discharge
• Dual corrections system: State
DOC operates prisons and
County Sheriffs operate jails
• Medicaid managed care with
HMO for medical care and
carve-outs for behavioral
health care
• Federal Medicaid required for
access to specialty mental
health services
• State Medicaid (GAU) covers
meds from primary care MDs
10
Research Design
• Case-control study with data available 3-yrs.
pre and 3-yrs.post start of expedited Medicaid
policy
• Administrative data on inmates with SMI,
service contacts for mental health (inpatient &
outpatient), substance use, arrests &
incarcerations
• Propensity score analysis used to construct
comparison group of inmates with SMI who
were not expedited and released during same
time period
11
Core Analysis for Inmates w SMI
Released from CT & WA Prisons
Usual
Release *
Release w/
Expedited
Medicaid *
12-mo postrelease service
use and costs
12-mo postrelease service
use and costs
12-mo arrest
and
incarceration
12-mo arrest
and
incarceration
* Comparison groups matched on propensity scores created from
demographics, clinical diagnoses, and behavioral health/criminal justice
history for the 3-yrs. prior to an index release
12
Expedited Medicaid Restoration
in Connecticut
Linda Frisman
CT DMHAS and UConn SSW
6th Annual Academic and Health Policy
Conference on Correctional Health
Chicago – March 22, 2013
Methods for CT Data
• Started with DOC discharges w/ 1 year FU
• Experimental group defined by DSS
• Propensity Scoring
– 1,511 Pre-Release Entitlement (PRE) = E
– 1,511 Propensity-matched cases (Non-PRE) = C
• Survival Analysis (Cox Regression)
– Time to event (enrollment, OP Tx, IP, arrest, etc)
• Poisson Models for count data
14
Poisson Models
• Used in a “conditional” situation: e.g., rate of use of EDs & # visits
– First part is about the likelihood of the situation happening
– Second part is the count of the event
• Poisson models involve different assumptions
– Poisson regression:
• assumes equally dispersion (the conditional variance equals
the conditional mean)
– Negative binomial regression (NB):
• allows for over-dispersion
– Zero inflated Poisson model (ZIP):
• allows for excess zeros
– Zero inflated Poisson negative binomial regression (ZINB):
• allows for over-dispersion and excess zeros
• Need to use the model best-suited to the data
15
Propensity-Matched Groups (1)
Prior to matching, all were significantly different
Variable
PRE Group
Non-PRE
Male
1148 (76.0%)
1180 (78.1%)
0.166
Age @ release
37.81 (10.57)
37.41 (11.04)
0.307
White
712 (47.1%)
687 (45.5%)
Black
459 (30.4%)
471 (31.2%)
Hispanic
337 (22.3%)
342 (22.6%)
48 (3.2%)
78 (5.2%)
Released 2006
Released 2007
562 (37.2%)
626 (41.4%)
Released 2008
901 (59.6%)
807 (53.4%)
Prob.
0.157
.000***
16
Propensity-Matched Groups (2)
Prior to match, all were sig. different except crime severity
Variable
PRE Group
Non-PRE
Prob.
Prior SSI
DMHAS history
320 (21.2%)
357 (23.6%)
0.106
1285 (85.0%)
1295 (85.7%)
0.607
Incarceration days
Overall RISK
MH
SA
Medical
Crime severity
Violence
Discipline
Gang
389.9 (586.3)
2.50 (0.97)
2.52 (0.87)
3.02 (1.14)
2.23 (0.89)
2.03 (1.10)
1.71 (0.91)
1.44 (0.95)
1.14 (0.53)
402.52 (807.4)
2.52 (1.01)
2.59 (1.08)
3.05 (1.10)
2.28 (0.93)
2.10 (1.08)
1.74 (0.93)
1.34 (0.85)
1.08 (0.35)
0.623
0.685
0.046*
0.485
0.186
0.074
0.353
0.004**
0.000***
17
Health Outcomes, 12 months
Variable
PRE Group
Non-PRE
On Medicaid
1328 (87.9%)
959 (63.5%)
0.000***
Days to Medicaid
6.11 (30.08)
38.88 (81.13)
0.000***
Any IP admit
518 (34.3%)
475 (31.4%)
0.096
Days IP
50.69 (86.75)
59.72 (100.15)
0.008**
Days to 1st IP
116.49 (104.26)
108.56 (107.11)
0.238
Any OP claims
1269 (84.0%)
971 (64.3%)
0.000***
# OP claims
57.53 (77.61)
41.93 (75.39)
0.000***
Days to 1st OP visits
42.85 (67.61)
71.22 (86.74)
0.000***
Any ED/crisis
265 (17.5%)
202 (13.4%)
0.002**
0.24 (0.63)
130.44 (108.17)
305.41 (90.02)
0.20 (0.71)
132.80 (107.56)
297.22 (98.78)
0.129
0.815
0.017*
# ED/crisis
Days to 1st ED/crisis
Community Days
Prob.
18
Criminal Justice Outcomes, 12 months
Variable
PRE Group
Non-PRE
Prob.
% re-arrested
515 (34.1%)
495 (32.8%)
0.441
# arrests
0.50 (0.84)
0.49 (0.87)
0.701
Days to 1st re-arrest
136.3 (115.7)
137.10 (108.52)
0.079
% re-incarcerated
592 (39.2%)
615 (40.7%)
0.393
Days incarcerated
49.24 (85.49)
51.52 (88.96)
0.474
Time to 1st re-incarceration
157.48 (99.23)
144.86 (102.62)
0.030*
19
Cox Regression
DV: Time to…
IV/Covars
OR (SE)
Prob.
Medicaid Enroll.
PRE
1.63 (0.045)
<0.000***
Inpatient Hosp.
PRE
1.09 (0.064)
0.175
Incarceration days
0.99 (0.00)
0.010*
NH days
1.004 (0.001)
0.001**
Outpatient Service
PRE
Community days
1.89 (0.043)
1.00 (0.00)
<0.000***
0.262
ER/Crisis
PRE
Community days
PRE
Community days
PRE
IP days
NH days
1.37 (0.094)
0.99 (0.000)
1.11 (0.06)
0.99 (0.00)
0.93 (0.06)
0.99 (0.001)
0.99 (0.005)
0.001**
0.027*
0.079
<0.000***
0.190
0.067
0.153
Re-arrest
Re-incarceration
20
Survival to Medicaid Enrollment
21
Survival to First OP Visit
22
Negative Binomial
DV
Inpatient Hosp. Days
Outpatient Services
Incarceration Days
IV/Covars
OR (SE)
Prob.
PRE
-0.52 (0.13)
0.0001***
Incarceration days
-0.003 (0.0007)
0.0001***
NH days
0.005 (0.006)
0.4533
PRE
0.305 (0.065)
0.0001***
Community days
0.001 (0.0004)
0.0073**
PRE
-0.06 (0.12)
0.6005
IP days
-0.01 (0.001)
0.0042**
-0.001 (0.004)
0.0207*
NH days
23
ZINB results for re-arrest
Estimate
SE
t
P value
Count
Intercept
0.784559
0.090457
8.67
<.0001
PRE
0.046296
0.059419
0.78
0.4359
-0.005321
0.000296
-17.98
<.0001
Intercept
1.332494
0.508800
2.62
0.0088
PRE
0.523669
0.994969
0.53
0.5987
-0.104144
0.040395
-2.58
0.0099
0.483689
0.072091
6.71
<.0001
Community Days
Probability
Community Days
Alpha
24
ZINB results for ED/Crisis
Estimate
SE
t
P value
Count
Intercept
-0.365080
0.290971
-1.25
0.2096
PRE
0.063935
0.105793
0.60
0.5456
Community days
-0.003667
0.000845
-4.34
<.0001
Intercept
3.013739
0.751414
4.01
<.0001
PRE
-1.817098
0.791286
-2.30
0.0217
Community Days
-0.025838
0.007342
-3.52
0.0004
Alpha
2.730451
0.331622
8.23
<.0001
Probability
25
Summary of CT Findings
• In Connecticut, the Pre-Release Enrollment
Program resulted in:
– Quicker access to Medicaid
– Quicker use & more use of Outpatient services
– Reduced use of Inpatient Care
– More community days
– NS difference in # of visits to EDs/crisis overall
– BUT people in PRE were more likely to use EDs
– No difference in CJ outcomes
26
Expedited Medicaid Restoration
in Washington State
Gary Cuddeback and Jennifer Jolley
University of North Carolina
6th Annual Academic and Health Policy
Conference on Correctional Health
Chicago – March 22, 2013
Research Question & Methods
• What is the impact of an expedited Medicaid
restoration program for SMI persons released
from prison in Washington State?
• Quasi-experimental design w/PSM
– Individuals with SMI released from prison in 2006
or 2007
• Approved for expedited Medicaid restoration vs. those
who were referred but not approved or eligible but not
referred
– 12-month follow-up after index release in 2006 or
2007
28
Definitions and Data
• Severe mental illness defined as having a
diagnosis of schizophrenia or other psychotic
disorders and/or bipolar disorder (with some
exceptions)
– Dx came from community mental health or DOC
• Linked administrative data available from 2003
to 2010 from Washington State CODB
– Demographics, diagnoses, Medicaid (program type),
inpatient and outpatient mental health service use,
substance use service use, homelessness,
employment, arrests, jail (some), violator facilities,
prisons
– Linked at person-level to create longitudinal file
29
Data Analysis
• Regression and survival models with PSM weights and
robust standard errors
• HB1290 approval = independent variable
• Dependent variables …
– Logistic regression
• Probability of Medicaid
• Probability of outpatient service
• Probability of arrest
– OLS and survival models
• Time to Medicaid
• Time to first service
• Time to arrest
• PSM balanced groups on all observables (more later)
30
Sample Characteristics: Approved, Denied
and Not Referred
Approved
(658)
% (n)
Denied*
(258)
% (n)
Male
White
Age (M(SD))
77 (504)
69 (451)
36 (8.9)
73 (189)
73 (189)
37 (11.9)
Psychotic disorder
SMI/SA
Prior ABD Medicaid
60 (395)
90 (594)
79 (517)
48 (123)
84 (217)
40 (104)
Prior GAU Medicaid
30 (197)
21 (55)
Not Referred
(2538)
% (n)
73 (1823)
73 (1823)
36 (9.5)
40 (1020)
80 (2010)
27 (683)
16 (403)
* Note: Denial reasons included: living arrangement (32%), failed incapacity
requirement (26%), voluntary withdrawal (6%), and other reasons (46%)
31
12-mo. Post-release Outcomes
Homeless
Unemployed
Medicaid (ABD)
Medicaid (GAU)
Any MH outpatient svc
Any AOD outpatient svc
Anti-psychotic meds
Anti-depress meds
Any inpatient service
Arrest
Approved
Denied
Not Referred
% (n)
56 (339)
% (n)
67 (161)
% (n)
60 (1407)
86 (523)
93 (568)
30 (181)
82 (198)
63 (153)
52 (120)
70 (1645)
47 (1102)
48 (1070)
22 (132)
43 (275)
49 (309)
48 (306)
12 (28)
33 (83)
28 (71)
32 (81)
7 (161)
32 (777)
15 (371)
21 (495)
10 (63)
7 (16)
4 (96)
56 (338)
60 (144)
51 (1186)
32
12-mo. Post-release Outcomes*
Outcome
Coeff (SE)
Probability of ABD Medicaid
2.35 (.18) .001 +39%
Probability of GAU Medicaid
-.12 (.11)
Probability of other Medicaid
-.76 (.14) .001 -13%
Probability of outpatient service
1.16(.14) .001 +14%
Probability of inpatient admit
.62 (.18)
ns
Probability of arrest
.11 (.11)
ns
Probability of incarceration
-.05 (.18)
ns
* Logistic regression with PSM weights used
p
+/-
ns
33
12-mo. Post-release Outcomes (cont’d)
Outcome
Coeff (SE)
p
Time to GAU Medicaid
8.9 (7.2)
ns
Time to other Medicaid
38.4 (5.7)
.01
Time to outpatient service
-34.5 (5.3)
.001 -34.5
Time to inpatient admit
-.22 (.58)
ns
Time to arrest
-7.5 (6.6)
ns
Time to incarceration (prison)
7.3 (3.6)
.05
Time to ABD Medicaid
+/days
-115.5 (6.53) .01 -115.5
+38.4
+7.3
34
LR w/o PSM: Probability of Arrest
Indicator
Coeff (SE)
p
OR
Male
.37 (.04)
.001
2.01
Race
.20 (.07)
.001
1.42
Age
-.54 (.07)
.01
.33
Homelessness
.26 (.04)
.001
1.69
Unemployment
.40 (.04)
.001
2.21
Substance use disorder
.19
.001
1.47
Expedited Medicaid
.03
ns
35
Key Findings
• Expedited benefit restoration associated with
– Greater and quicker Medicaid uptake
– Greater and quicker access to outpatient mental health
services
• Restored benefits not associated with lower probability
of criminal justice events
– Some improvement in time in community until prison
• Expedited restoration is working as a health insurance
program but few spill-overs for corrections
• But Medicaid alone is not enough as evidenced by role
of substance use, homelessness and unemployment in
arrest & re-incarceration!
36
Expedited Medicaid Restoration in
Washington State: Cost-Effectiveness
Marisa Domino and Jennifer Jolley
University of North Carolina
6th Annual Academic and Health Policy
Conference on Correctional Health
Chicago – March 22, 2013
Cost Analysis
• We examined the cost of expedited Medicaid
using a government payer perspective
– Costs related to medical and mental health services
use and criminal justice costs were included
– Short-run (12 month) time period examined
• We also analyzed days in the community as a
measure of effectiveness, for a cost-effectiveness
calculation
– Days not incarcerated nor in inpatient settings
38
Cost Methods
• Costs of medical and mental health services used
actual payments by Medicaid, state, and regional
payers for services delivered
• State hospital days costed using per diems
• Services include:
– Outpatient medical and mental health services
– Inpatient services, including state hospitals and local
inpatient hospitals
– Emergency room, crisis treatment, and medications
39
Cost Methods, continued
• Criminal justice costs used Washington State
Institute for Public Policy (WSIPP) calculated
costs, including costs of:
– Arrests
– Jail
– Prison
– Parole
40
Analysis Methods
• Because of concerns over selection bias in that those
receiving expedited Medicaid may differ from those
not receiving expedited Medicaid, we used propensity
score weighting to obtain better balance on baseline
risk factors
• Baseline risk factors include: time in prison and jail,
year of release, race/ethnicity, age, gender, prior
history of mental health service use, homeless prior to
index incarceration, work history, and Medicaid/state
program enrollment prior to index incarceration
• All factors balanced after propensity weighting
<=.25 SD/mean difference in groups
41
Selected Baseline Risk Factors
Weighted mean –
Expedited (n=608)
Weight mean – Controls
(n=2554)
531
513
Minority
41.0%
39.5%
Latino
6.1%
6.3%
Age
35
34
Male
75%
74%
Psychotic disorder
45%
45%
58
58
Homeless prior
51%
52%
Work history
59%
60%
Variable
Time served (days)
Jail days prior
42
Cost Results
Cost type
Weighted difference between Expedited
and Controls
Total cost
$3437**
Antipsychotic medications
$672**
Inpatient/ER- medical
$590 (p=0.051)
Arrests
$559 (p=0.095)
OP medical
$533**
ER
$357**
AOD
$258**
Parole
-$22*
Prison costs (DOC)
-$398 (p=0.067)
**p<0.01; *p<0.05
43
Community Days
• We find approximately 5 more community
days (p=0.076) in 12 months for those on
expedited Medicaid
• 3437/5 ~ $687/community day
44
Conclusions
• Preliminary results indicate that Expediting
Medicaid increases access to services, thus
increasing costs in the short run (1 year)
• While the program results in a greater number
of community days, this amounts to a
relatively high cost per day ($687) in the short
run
45
Conclusions/Limitations
• If investments in health are made in the short
run, the payoff in terms of reductions in
hospitalizations may not be observed until
beyond the 12 month window
– 36 month analysis is pending; will also look at
– 30 and 90 day results re criminal justice outcomes
• Days in the community is a crude measure, which
does not reflect quality of life
– Clinical and person-centered measures are not
available in our data
46
Conclusions
• In summary, expedited Medicaid in
Washington State led to:
– shorter time without insurance coverage
– Better access to services, especially
pharmaceuticals
– Higher costs, in total and on most dimensions
– A few more days in the community post-index
release
– Slight but nonsignificant reduction in DOC costs
47
Comparisons & Implications
• In both states, expedited Medicaid restoration led to
quicker and greater mental health service use
• No strong effects re reduced criminal justice
outcomes & costs
• Many controls went on to obtain Medicaid after
release so further analyses are needed to isolate
overall effects of having vs. not-having Medicaid
• However, our current analyses suggest that Medicaid
alone might not be enough to keep people with SMI
out of criminal justice system
48
Contacts
For additional questions and copies of our
presentation, please contact us:
• Joe Morrissey – [email protected]
• Linda Frisman – [email protected]
• Marisa Domino – [email protected]
• Gary Cuddeback – [email protected]
49

similar documents