Transcript presentation_apr2005 07

Cash Assistance and Monthly
Cycles in Substance Abuse
Carlos Dobkin and Steve Puller
Policy Concern About Cash Aid – Substance
Abuse Link
• Politicians and health professionals are
concerned that government aid results in
increased substance abuse
• There is also concern that cash aid results in
monthly cycle in drug related
hospitalizations straining already
overloaded emergency rooms.
Politicians have championed Various
Changes in Cash Aid
• Recent changes at the county, state and federal
level include
– Proposition N (San Francisco) “Care not Cash”
• Converted General Relief in San Francisco from cash to in
kind aid
– Welfare Reform Act of 1996
• Ended SSI benefits for people categorized as disabled due to
substance abuse
• ??any provisions for TANF and drug addiction???
– Gramm Amendment
Literature on Cycles in Drug
Consumption
• Monthly pattern in deaths (Phillips, NEJM 1999)
– 14% more substance abuse deaths in first vs. last week of month
• Monthly pattern in psychiatric admissions (Halpern &
Mechem, Am J Med, 2001)
– Psychiatric admissions for substance abuse 14% higher first week
(vs. 6% for non-substance abuse)
• Cocaine use among disabled vets (Shaner, NEJM, 1995)
– 105 male vets on disability with history of schizophrenia &
cocaine use
– Highest cocaine concentration in body during first 3 days of
month, followed by highest number of hospital admissions 3-5
days later
Causes of Drug Cycles is unknown
• “Fat wallets” early in the month could have a
number of causes
– Cash infusions at the beginning of the month due to
monthly pay checks
– Federal transfers (SSI, SSDI)
– State transfers (AFDC/TANF, Food Stamps)
– Local transfers (General Relief)
Contributions of this paper
• Document the monthly cycle in hospital
admissions and see how it varies by substance
• Determine which government programs are
driving the monthly cycle in admissions
• Check if alternate disbursal regimes can smooth
the monthly cycle in admissions
• Test if the programs effect the level of admissions
or just the timing of admissions
Data
• California Hospital Discharge Data 1994-2000
– Census of hospitalizations
– Includes patient demographics, cause of hospitalization
and treatment provided
• Medi-Cal Eligibility Data 1994-2000
– Linked to hospital data
– Includes individuals receiving welfare and
Supplemental Security Income for Aged Blind or
Disabled
– Does not include General Relief or Disability Insurance
Figure 1: Drug Related Hospital Admissions (California 1994-2000)
30,000
14,000
Admissions with a mention of Alcohol
12,000
20,000
10,000
8,000
15,000
6,000
10,000
4,000
Alcohol
Cocaine, Heroin or Amphetamine
5,000
0
2,000
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Day of Month
Admissions with a mention of Cocaine, Heroin or
Amphetamine
25,000
Figure 2: Monthly Cycle in Drug Related Hospital Admissions by Drug Type
(California 1994-2000)
7,000
4,500
4,000
6,000
3,500
5,000
4,000
2,500
2,000
3,000
Amphtamine
Heroin and Cocaine
3,000
1,500
2,000
Cocaine
Heroin
Amphetamine
1,000
1,000
500
0
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Day of Admission
I don’t understand why we exclude 29-31? It’s not a technical
Figure 3: Percent of Patients Leaving Hospital Where the Original Admission Had a Drug Mention
12.50%
12.00%
11.50%
Percent Leaving on Day
11.00%
10.50%
10.00%
9.50%
9.00%
Opioid
Amphetamine
Cocaine
8.50%
8.00%
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Discharge
Is is true that this is roughly flat for non-drug? So we can confidently
Table 1: Demographics by Type of Drug Mentioned on Admission Record
All Admission
Alcohol
Cocaine
Amphetamines
ER admission
42.21%
61.66%
50.84%
45.03%
Length of Stay (first admission)
4.46
6.24
6.75
6.18
Age of Patient
50.69
47.83
37.31
32.50
Insurance
Medicare
33.27%
26.76%
16.83%
12.44%
HMO
22.30%
16.80%
11.03%
14.72%
Medi-Cal
20.24%
18.51%
29.00%
30.48%
PPO
10.31%
7.92%
5.04%
6.58%
Private
4.90%
5.51%
5.43%
6.19%
Injury due to external causes
10.98%
18.13%
14.01%
15.18%
Cash Aid
Welfare
9.07%
4.08%
7.96%
13.42%
SSI Disability
9.26%
19.12%
28.13%
19.64%
Hospital Charges
$17,244
$18,123
$13,790
$13,568
Died first admission
2.26%
2.92%
1.07%
0.86%
Total admissions
18,484,469
773,279
157,150
117,158
Heroin
53.35%
6.75
41.25
17.42%
10.99%
27.37%
6.98%
4.04%
13.13%
5.78%
27.28%
$15,924
1.70%
181,106
Notes: records are included in the tables above if the drug is the primary cause of admission or if it is included as one of the other ICD-9 CM
codes.
Table 2: Causes of Admission by Drug
Alcohol
Cocaine
Amphetamine
Cause of Admission
Deliberate Injury
Accident
Drug Dependence
Drug Psychosis
Other Cause
0.041
0.073
0.497
0.194
0.196
0.049
0.026
0.362
0.158
0.405
0.050
0.035
0.300
0.139
0.475
Heroin
0.025
0.026
0.576
0.222
0.151
Cycles in Admissions
• Hospital admissions with a mention of alcohol or
illicit drugs are high in the beginning of the
month.
• The monthly cycle is particularly pronounced for
cocaine and amphetamine
• There is a cycle in people exiting the hospital with
a peak at the beginning of the month and a second
peak on the third of the month
Possible Causes of the Cycle
•
•
Monthly Paychecks
Supplemental Security Income
–
–
–
–
•
Disability Income
–
–
•
Paid bi-weekly
Workers’ Compensation
–
•
County run program for indigent adults (varies by county typically about $250 per month)
Unemployment Insurance
–
•
County administered program for Families
Largest of the programs with 2 million recipients statewide
Benefits about $550/month in 1997
Checks typically arrive on the first but there is variation by county
General Relief
–
•
Replacement rate varies with income
Aid arrives 3rd of month
Welfare
–
–
–
–
•
For the aged. blind or disabled – 5 month waiting period
Approximately 1 Million recipients state-wide, two-thirds disabled
Benefits average about $600/month for individuals ($1100/month couples)
Checks arrive on the 1st (or last previous business day if on weekend)
Some benefits paid bi-weekly
Supply side factors?
Table 3: Hospital Admissions 1994-2000 by Drug and Program
Alcohol
Cocaine
Herion
Counts of Admissions
Welfare
18,476
11,783
9,112
SSI Aged
10,824
186
837
SSI Blind
1,792
409
449
SSI Disabled
147,415
44,084
49,284
Other
592,267
100,185
120,899
Admissions Per Month per 10K Enrollees
Welfare
0.98
0.63
0.48
SSI Aged
3.81
0.07
0.29
SSI Blind
8.49
1.94
2.13
SSI Disabled
24.79
7.41
8.29
Amphetamine
15,233
53
176
22,969
78,437
0.81
0.02
0.83
3.86
Table 4: Timing of Pay Checks
Employees
Weekly
905,562
Biweekly
2,409,063
Semimonthly (Typically 1st and 15th)
720,142
Monthly
252,101
Other
7,704
Total
4,294,572
Note: Based on an American Payroll Association Survey of 872 companies
Percent
0.21
0.56
0.17
0.06
0.00
Figure 4: Hospital Admissions With a Mention of Cocaine, Heroin or Amphetamine by Program
4,500
10,000
4,000
9,000
8,000
3,500
7,000
6,000
2,500
5,000
2,000
Welfare
SSI
Not on SSI or Welfare
1,500
4,000
3,000
1,000
2,000
500
1,000
0
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Day of Admission
Non Welfare non SSI
SSI and Welfare
3,000
Figure 5: Hospital Admissions by Insurance Type for Cocaine, Heroin and Amphetamine for People
Receiving Neither Welfare nor SSI
3,500
1,800
1,600
1,400
2,500
Private Insurance
1,200
2,000
1,000
800
1,500
600
1,000
Private
Self Pay
County Indigent
Medi-Cal
Medicare
500
0
400
200
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Day of Admission
Self Pay, County Indigent, Medi-Cal and Medicare
3,000
Figure 6: Hospital Admissions With a Mention of Alcohol by Program
6,000
25,000
5,000
20,000
15,000
3,000
10,000
2,000
Welfare
SSI
Not on SSI or Welfare
5,000
1,000
0
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Day of Admission
Non Welfare non SSI
SSI and Welfare
4,000
Figure 7: Hospital Admissions by Insurance Type With a Mention of Alcohol for People Receiving
Neither Welfare nor SSI
8,000
2,500
7,000
2,000
5,000
1,500
4,000
1,000
3,000
Private
Self Pay
Medicare
County Indigent
Medi-Cal
2,000
1,000
0
500
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Day of Admission
Medi-Cal, County Indigent
Sefl Pay, Medicare, Private
6,000
Figure 8: Proportion of Patients With Cocaine, Heroin or Amphetamine Admission Leaving the
Hospital For Home
0.25
0.12
Proportion of Patients Exiting Hospital
0.08
0.15
0.06
0.1
0.04
Welfare
SSI
Not on SSI or Welfare
0.05
0
0.02
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Discharge
Proportion of Welfare and Non Welfare Non SSI
Admission Exiting Hospital
0.1
0.2
Figure 9: Proportion of Patients Not Receiving Welfare or SSI With Cocaine, Heroin or Amphetamine
Admission Leaving the Hospital For Home
0.16
0.14
Proportion of Patients Exiting Hospital
0.12
0.1
0.08
0.06
Medicare
Medi-Cal
Self Pay
Private
County Indigent
0.04
0.02
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Discharge
Figure 10: Proportion of Patients With Cocaine, Heroin or Amphetamine Admission Leaving the
Hospital Against Medical Advice
Proportion of Patients Exiting Hospital Against Medical Advice
0.025
0.02
0.015
0.01
SSI
Welfare
Not on SSI or Welfare
0.005
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Discharge
Figure 11: Proportion of Patients Not Receiving Welfare or SSI With Cocaine, Heroin or
Amphetamine Admission Leaving the Hospital For Home Against Medical Advice
Proportion of Patients Exiting Hospital Against Medical Advice
0.025
Medicare
Medi-Cal
Self Pay
Private
County Indigent
0.02
0.015
0.01
0.005
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Discharge
Programs and the cycles in
admissions
• Welfare has only a very weak cycle in admissions
• Very strong cycle in admissions for SSI recipients
particularly for illicit drugs
• Indirect evidence of a cycle for people receiving DI
– Can identify likely DI recipients because they are Medicare
recipients under 65.
– Cycle in admissions
– Peak in exits on the third
• The overall cycle appears to be due largely to SSI and DI
• Peoples exit patterns particularly AMA patterns are
consistent with them heading home to pick up their checks.
WILL ALTERNATE DISBURSAL
SCHEMES REDUCE CYCLE?
• Hospital Emergency Rooms in California
are crowded
• Many of the ER resources are fixed
• Cyclical crowding is bad
• Will smoothing the check disbursal smooth
the cycle
(LA Pre vs. Post Analysis -- Notes to
ourselves)
• Keep in mind that this analysis is for a population
(welfare) that doesn’t show much of a cycle
anyway
• We do 3 ways: raw means, regression adjusted,
and testing it statistically
• Basic conclusion:
– Cycle: see pretty convincing shift of cycle for drugs,
but only slightly for alcohol or other non-drug
conditions
– Levels: highly confounded w/ other changes from
welfare reform so we cannot isolate the effect
• Not surprising: Policy does not change “fat wallet” at the
individual level
Empirical Evidence on Alternative Disbursement Regimes
•
Los Angeles county disbursement of
AFDC/TANF
– Pre June 1997: Day 1
– Post June 1997: Staggered Days 1-10 based on case
number (recipients could pick up at issuance outlet
after designated day)
– Effect:
•
•
Individuals still have “fat wallets”, but everybody doesn’t
have them at the same time
DI changed in May 1997
– Post may 1997 new recipients instead of getting on the
third of the month get second third or fourth weds
depending on day of birth
•
Questions
– Does aggregate cycle change?
Pre vs Post June 1997
Drug Mentions - Means
5.23
4
4.73
3.5
4.23
3
3.73
2.5
3.23
2
2.73
Day 1
Days 1-10
Poly. (Day 1)
Poly. (Days 1-10)
2.23
1.73
1.23
1.5
1
0.5
0
1 2 3 4 5 6 7 8
Note: Vertic al axes are shifted by differenc e in mean
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
Mean Admissions Days
1-10
Mean Admissions Day 1
Admissions with Drug Mention
Welfare Recipients in Los Angeles County
Pre vs Post Shift from Day 1 to Days 1-10 Disbursement
Pre vs Post June 1997
Alcohol Mentions - Means
Admissions with Alcohol Mention
Welfare Recipients in Los Angeles County
Pre vs Post Shift from Day 1 to Days 1-10 Disbursement
Day 1
Days 1-10
Poly. (Day 1)
Poly. (Days 1-10)
3.91
3.41
2.91
4
3.5
3
2.5
2.41
2
1.91
1.5
1.41
1
0.91
0.5
0.41
0
1 2 3 4 5 6 7 8
Note: Vertic al axes are shifted by differenc e in mean
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
Mean Admissions Days
1-10
Mean Admissions Day 1
4.41
Pre vs Post June 1997
Non-drug Admits - Means
14
13.49
12
11.49
10
9.49
8
7.49
Day 1
Days 1-10
Poly. (Day 1)
Poly. (Days 1-10)
5.49
3.49
1.49
6
4
2
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
Other c onditions are pneumonia, bronc hitis, diabetes, asthma and stroke. Note: Vertic al axes are shifted by differenc e in mean
Mean Admissions
Days 1-10
Mean Admissions
Day 1
15.49
Admissions with No Drug Mention
for One or More Other Specified Conditions
Welfare Recipients in Los Angeles County
Pre vs Post Shift from Day 1 to Days 1-10 Disbursement
Statistical Test
Notes to Us
–
“treatment effect of SSI”
•
•
•
there is a 5 month waiting period from time of disability to eligibility to receive benefits
(although this can be waived and we don’t know if an individual get’s it waived – do we
know % that get waived??)
we can see individuals about to go on SSI who were previously on another medical program
(2 largest are non-cash disability for the medically needs and TANF cash assistance) and
those who just went on SSI – arguably these people are similar in the window of time around
the transition (although those just on may be a little worse off)
2 definitions of “Just Went On” and “About To Go On”
1 month window
•
Just went on = SSI this month, non-SSI (but medical) last month
•
About to go on == no SSI this month (but some medical), SSI next month
•
??any worry about the length of stay or timing of exactly when become eligible??
2 month window
»
just went on == recipients who are eligible in the current month and the previous month, but
not 2 months ago
»
about to go on == patients who are not currently on SSI (but are on some other Medical
program in our dataset), will not be on next month, but are eligible for SSI in the 2nd
following month (we do this due to length of stay issues crossing over into next month)
•
•
we test for the effect on both the cycle and level of treating these people with SSI cash aide
Caveat: this is a small sample (estimating level effect off of 2738 admits for the 1 week
window and 2259 admits for the 2 week window)
Recipients Transitioning Onto SSI
Drugs – 1 Month Window
0
.1
.2
.3
.4
.5
Monthly Cycle Among Recipients Transitioning
0 1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031
Day of Month
About To Transition
About To Transition - Mean
Just Transitioned
Just Transitioned - Mean
Recipients Transitioning Onto SSI
Drugs – 2 Month Window
0
.1
.2
.3
.4
.5
Monthly Cycle Among Recipients Transitioning
0 1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031
Day of Month
About To Transition
About To Transition - Mean
Just Transitioned
Just Transitioned - Mean
Simulated Effect on Aggregate Drug Cycle of
Disbursing SSI, SSDI and Welfare
with a Day1-10 Scheme
• Assume
– Cycles for certain subpopulations entirely driven by timing of aide
disbursement
– Consumption pattern independent across groups (no “agglomeration
economies” to consumption)
• Simulate Aggregate cycle
– Take observed cycle (from single day disbursement) & simulate the
cycle if
• 1/10th of recipients = day 1, 1/10th=day 2, …, 1/10th = day 10
–
–
–
–
–
SSI: from Day 1 to Days 1-10
SSDI (proxied by Medicare): from Day 3 to Days 1-10
Welfare outside LA post 97: Day 1 to Days 1-10
Welfare in LA post 97: keep the same observed cycle
All others (employed, UI, others) keep the same
Simulated Monthly Drug Cycle When SSI, SSDI, and Welfare Are Disbursed Days 1-10
250
Drug Admissions
200
150
100
Actual
Day 1-10 For SSI, SSDI, Welfare
50
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Day of Month
18
19
20
21
22
23
24
25
26
27
28
29
30
31
INCLUDE RESULTS FROM
LOOKING AT DI
SUM UP THE GAINS FROM
SWITCHING REGIMES
• Smoother welfare admissions despite the
fact that welfare is pretty flat
• If similar change in disbursement in all
programs occurred for all programs you
would expect a big effect
ALSO
Effect of Disbursement Change on Levels?
• We have seen compelling evidence that Welfare, SSI
and DI cause a cycle in admissions
• People take this as proof that government transfers
also significantly change the number of admissions
• Two approaches
– County level changes in loads
• Abrupt Change in SSI due to Federal Law
• Long term reduction in welfare loads
• Of Unknown origin ie panel approach
– Micro Analysis (note less compelling)
• Look at people going onto welfare
• Look at people going onto SSI
REGRESSSION RESULTS
CROSS-SECTION AND PANEL
REGRESSION RESULTS ON
SPIKE
CASE STUDY OF SSI CHANGE
CHANGE IN SPIKE
19
94
19 -01
94
19 -04
94
19 -07
94
19 -10
95
19 -01
95
19 -04
95
19 -07
95
19 -10
96
19 -01
96
19 -04
96
19 -07
96
19 -10
97
19 -01
97
19 -04
97
19 -07
97
19 -10
98
19 -01
98
19 -04
98
19 -07
98
19 -10
99
19 -01
99
19 -04
99
19 -07
99
20 -10
00
20 -01
00
20 -04
00
20 -07
00
-1
0
SSI Aged, Blind and Disabled
800,000
400,000
1,500,000
300,000
1,000,000
200,000
100,000
SSI Aged
SSI Blind
SSI Disability
Welfare
0
Month
500,000
0
Welfare
Counts of Recipients of SSI and Welfare
3,000,000
700,000
2,500,000
600,000
500,000
2,000,000
POINT OUT HUDE DECLINE IN
WELFARE NO PERCEVABLE
DECLINE IN DRUG
ADMISSIONS RATES
TS Counts of admissions overall and
by SSI, Welfare and NON SSI NON
WELFARE
CHANGE IN LEVELS
MICRO ANALYSIS LOOK AT
PEOPLE TRANSITIONING ONTO
PROGRAM
SUM UP RESULTS DOES IT
CHANCE CYCLES DOES IT
CHANGE LEVELS
CONCLUSIONS
• Which program drive spike
• Can smooth spike
• Levels harder to reduce
OLD STUFF
What we
Daily Average California Hospital Admissions
for Drug Use
Figure 1: Monthly Cycles in Drug Admissions
80
70
Admissions
60
50
40
30
Cocaine
20
Opiates
Amphetamines
10
0
1
8
15
Day of Month
22
29
Cycles in Injuries & Violence?
Cycles in Other Admissions?
Cycles in Leaves Against Medical Advice
Research Design
Data
Drug Admission by Cash Aide Type
Regression Context
Admitst  f ( DayofMonthDumt , DayofWeekDumt , MonthDumt ,YearDumt ,
HolidayDumt , t )
First cut = linear. Later = count data model
Exploiting County-Level Variation in Disbursement
– 1997: LA switched AFDC/CalWORKs
disbursement from Day 1 to staggered over
first 10 days
Individual Level Analysis
•
Ideal experiment:
–
•
Randomly assign aide receipt day 1, day2,
…and compare patterns of admission
Pseudo-experiment:
–
Some counties disburse based on last digit of
case #
Estimate
Pr(admission) = f(days since receipt,
days since first of month,…)
•
Data available?
Policy Implications
• 15-20% of welfare recipients self-report to
drug use, with one-fifth of those dependent
(Pollack et al., 2002)
• Average charge per admit: $xx,xxx
• Possibly smooth cycle or reduce levels of
substance use with:
– Smoothing disbursal (debit cards)
– Substitute in-kind for cash aide
• Even if consumption/admits don’t decrease:
– Smooth hospital caseloads or policing activities
Extra Slides
Basic Demographics of CA in 1996
Source: Urban Institute, 1998
Pattern By Day of Week