Transcript Information Technology and Economic Growth in the US and Japan

Supplier-Induced Demand in
Japan’ LTC Market
Satoshi Shimizutani
(coauthored with Haruko Noguchi)
Motivation (1)



Rapid speed of aging (65+ exceeds 20%)
and expansion of LTC costs
LTC expenses: 6.18 trillion yen in FY2004
(75.4 %percent increase from FY 2000,
150% increase for at-home care)
How to operate the LTC market efficiently?
How to motivate market participants
behave properly?
20
00
.
20 4
00
.
20 6
00
2 0 .8
00
.
20 10
00
.1
20 2
01
.
20 2
01
.
20 4
01
.
20 6
01
2 0 .8
01
.
20 10
01
.1
20 2
02
.
20 2
02
.
20 4
02
.
20 6
02
2 0 .8
02
.
20 10
02
.1
20 2
03
.
20 2
03
.
20 4
03
.
20 6
03
2 0 .8
03
.
20 10
03
.1
20 2
04
.
20 2
04
.
20 4
04
.
20 6
04
2 0 .8
04
.
20 10
04
.1
20 2
05
.
20 2
05
.4
LTC insurance use
(Billion Yen)
Figure 1 Care Costs through the Long-Term Care Insurance in Japan
600
500
400
300
200
100
0
Institutional services
At-home services
Motivation (2)


Popular hypothesis: realization of
potential demand suppressed before
2000, moral hazard……
Only lower-income households were
eligible to receive LTC provided by the
local government under social welfare
Motivation (3)



Focus on prevention of moral hazard in
LTC providers
Asymmetry of information between
suppliers and demanders
Fixed service prices under public
insurance program
LTC program in Japan


Under the program, once certified, a
insured person are free to choose care
services (at-home care and institutional
care) at a 10 percent co-payment.
More market-oriented policy: allowed
for-profit providers to enter the at-home
care market for the first time
SID-previous research (1)



SID hypothesis: enormous literature
Feldstein (1970): positive correlation bet.
physician incomes & physician density
Fuchs (1974), Evans (1974),Reinhardt (1978)
Several models: Physician takes advantage of
information asymmetry bet. suppliers &
demand (due to skilled knowledge etc.)
SID-previous research (2)



Physician-induced demand exists when the
physician influences a patient’s demand for
care against the physician’s interpretation of
the best interest of the patient (McGuire
(2000)).
Empirical findings are inconclusive.
Identification problem (supplier or demanderinduced): Childbirth & Physician density
SID-previous research (3)
Two phase model (Rossiter & Wilensky
(1984) etc).
 1st phase=probability to use medical service:
Effect of higher accessibility
 2nd phase=medical expenditure per patient :
Effect of physician-induced demand
Escarce (1992) finds the intensity of
physicians affects 1st phase but not 2nd phase.

SID-previous research (4)


Previous studies in Japan
Nishimura (1987): positive correlation bet.
medical expenditure and MD density.
Several studies after the 1990s
SID observed in Yamada (2002) but not in
Suzuki (1998), Kishida (2001)
LTC Case (prefecture data): observed in
Yamauchi (2003) but not in Yuda (2004)
Data





Data: Micro-level data from the “Survey on
Long-term Care Users” in 2002 and 2003,
compiled by ESRI, Gov. of Japan.
Randomly chosen (response rate: 80%).
HH with one un-institutionalized needy
elderly inc. uncertified.
Sample size : around 1,000 in each year.
Matched with density of providers (prefecture
level).
Summary statistics (1)






Share of care receivers: 60%
At-home care exp./month: \12,000-13,000
Female:75%, and Age:84
Care levels 1(20%),2(20%),3(10%)
Brain vein disease, dementia, bone fracture
and frail with aging (>20%)
Frequency to go to hospital: 3 days/month
Summary statistics (2)




HH income (4-6 bill. yen) and HH assets
(30-50 bill. Yen)
# family members: 4.0
Number of establishments per certified
persons in a prefecture: 0.01 (1
establishments for 100 certified)
Share of for-profits: 25-33 %
Specification
Yi   0   1 X i   2 Intensity i   i
*
Dependent:dummy variable of i th user’s choice to use LTC service
ln( Exp) i   0   1 X i   2 Intensity i   i
Dependent: logarithmic value of i th user’s expenditure for LTC
Results 1 (prob. to use)
FY2002
Variables
marginal
S.E.
FY2002
marginal
S.E.
FY2003
marginal
S.E.
FY2003
marginal
S.E.
5. Intensity (prefecture level)
Number of establishments per approved persons in a prefecture
Share of for-profits
-17.705 (75.930)
-57.509 (92.230) c 21.590 (75.613)
-0.978 (1.280) b
-1.989
-1.142
(79.547)
(0.960) a
Results 2 (Care expenditure)
FY2002
Variables
FY2002
coefficient S.E. coefficient S.E.
FY2003
coefficient
FY2003
S.E. coefficient S.E.
5. Intensity (prefecture level)
Number of establishments per approved persons in a prefecture
Share of for-profits
-63.902 (233.640) -248.605 (286.282) 387.157 (242.662) 229.675 (249.287)
-4.170 (3.734)
-6.947 (2.905) b
Findings


Little evidence on the SID in Japan’s LTC
market. A higher portion of for-profits does
not induce demand.
Consistent with Yuda (2004) at prefectural
data.
SID in AMI treatment
AMI (Acute Myocardinal Infarction)
High-tech treatment: cardiac catheterization
(CATH) and revascularization procedure
PTCA (Percutaneous Transluminal Coronary Angioplasty)
CABG (Coronary-Artery Bypass Graft Surgery)
Low-tech treatment: Acute drug treatments
(aspirin, thrombolytic drugs, beta blocker,
calcium channel blocker etc.)

AMI treatment in Japan
PTCA
Fee schedule for reimbursement
Number of PTCA hospitals
Number of PTCA performed
CABG
Fee schedule for reimbursement
1
2 or more
Number of CABG hospitals
Number of CABG performed
Number of General Hospitals
1993
1996
13,800
381
3,648
15,500
609
5,818
37,100
60,500
397
2,699
8,752
37,100
60,500
453
2,814
8,421
SID in AMI treatment: Data



Data: chart-based microdata from Tokai
Acute Myocardial Infarction Study (called
TAMIS), comparable with Cooperative
Cardiovascular Project (CCP)
2,020 heart attack patients in 14 high-tech
and high-volume medical facilities in the
Tokai area in 1995-1997.
1,047 patients living in 116 municipal areas
matched with regional data.
SID in AMI treatment: Spec.
Ti, s   0   1k Dens k , i, s    2l xl , i, s  i, s
k
l
Dependent:dummy variable of i th patient’s choice of s th treatment, CATH,
PTCA, or low-tech acute drug treatments.
yi, s   0   1k Dens k , i, s    2l xl , i, s   i, s
k
l
Dependent: logarithmic value of i th patient’s expenditure for s th
treatment, CATH, PTCA, or low-tech acute drug treatments.
Results on High-techs
Medical expenditure per
person
Prob to go to hospitals
Explanatory variables
CATH
PTCA
Marginal effect Marginal effect
# beds/100,000
# high-tech hospitals (PTCA available) per
100,000
# low-tech hospitals (no PTCA available)
per 100,000
# high volume hospitals with more than 100
beds per 100,000
# physicians per 100,000
CATH
OTCA
Marginal effect Marginal effect
-0.0001
b
-0.0002
b
-0.002
c
-0.003
c
0.081
a
0.072
a
1.256
a
1.089
a
-0.014
a
-0.010
c
-0.199
a
-0.159
b
0.055
b
0.082
b
0.716
c
0.864
c
0.002
a
0.004
a
0.026
a
0.039
a
Results on Low-techs
Prob. To go to
hospital
Medical
expenditure per
person
Explanatory variables
Marginal effect
Marginal effect
# beds/100,000
0.0001
b
0.001
b
# high-tech hospitals (PTCA available) per
100,000
# low-tech hospitals (no PTCA available)
per 100,000
# high volume hospitals with more than 100
beds per 100,000
# physicians per 100,000
-0.081
a
-0.505
a
0.014
a
0.088
a
-0.055
b
-0.381
b
-0.002
a
-0.013
a
SID in AMI treatment


# of high-tech or high-volumn hospitals and
MDs per person are positively correlated
with medical expenditure in both phases in
PTCA or CABG.
# of low-tech hospitals per persons is
positively correlated with medical
expenditure in both phases in low-tech
treatment.
Conclusions and Discussions



Our findings report SID is not observed
in LTC but in a high-tech treatment.
One explanation is the degree of
information asymmetry
Implications: Maintaining care
manager’s skill, further disclose etc.