Impact of Private Health Insurance on the
Choice of Public versus Private Hospital
Services
Preety Srivastava & Xueyan Zhao
Centre for Health Economics
Monash University
3 June 2008
Background
• Australian Health Care system
– Mix of public and private health services
– One of the highest % of private coverage
across OECD countries
– In 2004-05, 4/10 hospital admissions and 1/4
inpatient days were private (NHS, 2004-05)
– Policy makers recognise the important role of
Background
• PHI reforms after introduction of Medicare to
ensure no ‘crowding out’ of the private sector.
• Severe decline in PHI in the 90s leading to
enormous pressure on public hospitals
• To stem erosion a package of initiatives
introduced in late 90s:
– Tax penalty for high-income individuals without
private cover
– 30% rebate on PHI premiums
– Lifetime health cover
Background
• Reforms strongly criticised by scholars
– Package of initiatives, in particular lifetime cover has
increased PHI coverage but activity in the private
sector not picked up. Not eased much pressure in
public hospitals.
– money could have been better spent if applied directly
toward enhancing capacity of public hospitals to meet
the additional demand (Wilcox, 2001; Duckett and
Jackson, 2000).
– PHI taken purely for financial reasons (Fiebig et al.,
2006) and not necessarily to access private care.
Background
• More recently:
– Income threshold for Medicare surcharge
penalty would be increased both for singles
and families.
• This is also being criticised on the ground
that a number of people are going to drop
their PHI.
Background
• Concerns on equity of care-provision
– In terms of the disproportionate distribution of
tax rebates to high-income earners (Hindle and
McAuley, 2004; Butler, 2002; Wilcox, 2001).
– Subsidy is skewed to the more affluent.
– 80% (20%) of richest (poorest) 10% of
Australians had PHI and nearly 75% (18%)
admitted as private patients in 2005(NHS
2004-05)
Objective of the study
• The objective of this study is to investigate
the determinants of individuals’ choice
between public and private hospital care
and the role of PHI towards this decision.
Motivation and Contribution
• Demand for PHI has received ample attention in
the literature but only a small body of research
has examined its role in public/private health
care utilisation
(Fiebig et al., 2006; Rodriguez and
Stoyanova, 2004; Savage and Wright, 2003; Propper,
2000)
• Also sheds light on the potential substitution
between public and private hospital admissions
in a system where PHI increases the chances of
substitution by providing a duplicate coverage.
Motivation and Contribution
•
Also makes a significant contribution in terms of the
modelling approach.
– In most prior studies the 3 decisions i.e. to seek no care or
private care or public care, has been modelled using a MNL
model.
– In contrast, we model the hospital admission decision in two
parts on the assumption that the decision to seek hospital care
and the decision to get admitted as a public/private patient are
two distinct processes.
• account for selectivity bias in the second stage given that the decision to
get admitted as a public or private patient is only observed for those who
visit a hospital.
– Also unlike prior studies, this study accounts for the endogeneity
of PHI using a system approach instead of a two-step
Prior Studies
•
Relationship between PHI and health service utilisation (Zhang and
Zhao 2007)
–
made no distinction between public and private admissions.
•
Relationship between PHI and hospital admission (Fiebig et al. 2006)
–
focus was more on the impact of insurance type - in terms of reasons for purchasing private
health insurance - on the probability of hospital admission in Australia.
•
Private health insurance participation and the duration of stay in
private hospitals (Savage and Wright 2003)
–
focus was on identifying any moral hazard behaviour and adverse selection in insurance
purchase.
•
Impact of PHI on hospital admission and hospital days (Cameron et al.
1988)
–
made no distinction between public and private admissions.
•
Overseas
– UK (Propper 2000)
–
Spain (Rodriguez and Stoyanova 2004)
–
Harmon and Nolan, 2001 and Holly et al., 1998.
•
Estimation techniques
–
two step estimation to account for endogeneity
–
Accounted for endogeneity using FIML approach – not distinguished between public and
private service utilisation
Economic Framework
• Demand for health care=function of
• the value of benefits of treatment;
• quality of public care vs private care;
• attitude towards quality of care;
• cost of public health care (if any);
• cost of private health care
Economic Framework
•
Value of benefits of treatment:
– Related to medical need which arises from the severity of illness and
importance of good health.
• Importance of good health positively associated with education and
socioeconomic factors and
• Negatively related to lifestyle factors such as drinking and smoking
patterns, and exercise habits
•
Quality of public care vs private care
– Reflected in waiting time, the ability to choose the doctor
•
Attitude towards quality of care
– Quality measures such as waiting time or the inability to choose date
and location of treatment may prove to be inconvenient. Since each
person has his own valuation of time this may cause variations across
people.
– A person’s valuation of his time is usually a positive function of income
and type of employment.
Economic Framework
•
Cost of public health care (if any)
– Although public health care is free of user charges, travel and time costs
are also important considerations, in particular for lower socioeconomic
groups.
• Such factors are negatively associated with income.
•
Cost of private health care
– Access costs to private health care depends mainly on price of health
insurance and income
– Copayments can also represent a significant cost to access private
health care, particularly in Australia.
•
Access to health care
– can also vary across the population because of language or cultural
differences. Such differences may result into a lower level of awareness
of health care availability and efficacy or a shyness to use health
Econometric Framework
• System approach with partial observability
Latent form:
Econometric Framework
•
Multivariate Probit (MVP) model.
•
The system approach allows us to account for not only the effect
of the observed variables and but also the effect of unobserved
individual characteristics.
•
This allows us to estimate a whole range of joint and conditional
probabilities.
•
We can also estimate the ‘treatment effect’ of PHI, i.e. the effect
of private insurance participation on the probability of visiting
hospital or on the probability of seeking private care.
Data
• NHS 2004-05 (14 970 Australian adults aged 18+)
– Contains a host of health related information (i.e. SRH, LT health
conditions)
– Health service utilisation
– Other individual characteristics such as gender, marital status,
income, level of education, main activity etc.
– Dependent variables:
•
Y
I: status of individuals who, at the time of the survey, had a private
hospital cover
•
Y
H:
whether an individual had at least one inpatient stay in a hospital
and discharged in the 12 months prior to interview.
Results: Coefficients
YI 2.254 (0.563)** incdech4 0.366 (0.046)** 0.019 (0.051) 0.183 (0.135) age30 0.204 (0.042) ** -0.096 ( 0.045)** 0.265 (0.125)** incdech5 0.393 (0.047)** 0.147 (0.053)** 0.250 (0.131)* age40 0.480 (0.043) ** -0.211 ( 0.047)** 0.104 (0.131) incdech6 0.525 (0.050)** 0.123 (0.057)** 0.193 (0.153) age50 0.792 (0.047) ** -0.238 ( 0.050)** 0.280 (0.161)* incdech7 0.625 (0.050)** 0.126 (0.058)** 0.243 (0.160) age60 1.135 (0.056) ** -0.263 ( 0.056)** 0.533 (0.226)** incdech8 0.793 (0.051)** 0.077 (0.061) 0.172 (0.183) age70+ 1.118 (0.062) ** -0.207 ( 0.059)** 0.690 (0.228)** incdech9 0.991 (0.053)** 0.177 (0.061)** 0.238 (0.200) male -0.067 (0.027) ** -0.133 ( 0.029)** -0.148 (0.077)* incdech10 1.346 (0.056)** 0.187 (0.061)** 0.241 (0.228) married 0.222 (0.028) ** 0.072 ( 0.027)** 0.120 (0.079) concess - 0.482 (0.040)** 0.108 (0.044)** 0.105 (0.123) profeng 0.370 (0.081) ** 0.153 ( 0.090)* 0.195 (0.204) excelh 0.233 (0.063)** -0.741 (0.062)** depkid 0.157 (0.036) ** 0.001 (0.103) vgoodh 0.224 (0.060)** -0.667 (0.056)** sinpar -0.158 (0.064) ** 0.035 (0.167) goodh 0.141 (0.057)** -0.460 (0.052)** majcity 0.120 (0.033) ** -0.004 ( 0.037) -0.056 (0.090) athritis 0.013 (0.032) 0.075 (0.032)** inregn 0.111 (0.038) ** 0.026 ( 0.042) 0.036 (0.098) cancer 0.097 (0.075) 0.478 (0.070)** workft -0.349 (0.065) ** -0.324 ( 0.046)** -0.096 (0.201) heart 0.056 (0.029)* 0.143 (0.030)** workpt -0.294 (0.063) ** -0.141 ( 0.045)** -0.220 (0.205) diabetes - 0.027 (0.054) 0.191 (0.052)** workstud -0.187 (0.112) * -0.144 ( 0.117) -0.011 (0.276) asthm a 0.025 (0.040) 0.073 (0.039)* studyft 0.439 (0.081) ** -0.324 ( 0.089)** -0.433 (0.276) osteo 0.167 (0.056)** 0.048 (0.057) unemp -0.221 (0.102) ** -0.242 ( 0.093)** 0.011 (0.299) smokedly - 0.393 (0.031)** prof 0.448 (0.057) ** 0.187 (0.180) alchirsk - 0.144 (0.062)** trades 0.235 (0.066) ** 0.043 (0.200) overwt - 0.024 (0.025) cler k 0.598 (0.097) ** 0.264 (0.308) noexcise - 0.082 (0.026)** intsales 0.321 (0.061) ** 0.377 (0.202)* copay - 0.006 (0.001)** -0.002 (0.003) prodtran 0.156 (0.072) ** 0.132 (0.220) bed - 0.023 (0.019) -0.025 (0.060) elsales 0.131 (0.071) * 0.082 (0.214) Constant - 1.644 (0.129)** -0.474 (0.119)** -2.614 (1.351)* degree 0.314 (0.042) ** 0.042 ( 0.044) 0.162 (0.108) ΞIH 0.085 (0.018)** tafe 0.096 (0.035) ** 0.106 ( 0.036)** 0.138 (0.092) ΞIP - 0.245 (0.272) year12 0.213 (0.034) ** 0.015 ( 0.036) 0.171 (0.091)* ΞHP 0.392 (0.150)**Standard er rors are given in parentheses. *significant at 10% level; **si gnificant at 5% level.
YH YP
Results: Marginal Effects
YI 0.757 (0.095)** tafe 0.038 (0.014)** 0.026 (0.009)** 0.055 (0.026)** age30 0.081 (0.017)** -0.024 (0.011)** 0.143 (0.033)** year12 0.084 (0.013)** 0.004 (0.009) 0.110 (0.025)** age40 0.189 (0.017)** -0.052 (0.012)** 0.198 (0.033)** incdech4 0.144 (0.018)** 0.005 (0.013) 0.160 (0.036)** age50 0.312 (0.018)** -0.058 (0.012)** 0.345 (0.035)** incdech5 0.155 (0.018)** 0.036 (0.013)** 0.173 (0.034)** age60 0.448 (0.022)** -0.065 (0.014)** 0.521 (0.044)** incdech6 0.207 (0.020)** 0.030 (0.014)** 0.201 (0.038)** age70+ 0.441 (0.024)** -0.051 (0.014)** 0.551 (0.043)** incdech7 0.247 (0.020)** 0.031 (0.014)** 0.245 (0.039)** male -0.026 (0.011)** -0.033 (0.007)** -0.046 (0.021)** incdech8 0.313 (0.020)** 0.019 (0.015) 0.284 (0.042)** married 0.088 (0.011)** 0.018 (0.007)** 0.093 (0.021)** incdech9 0.391 (0.021)** 0.044 (0.015)** 0.352 (0.045)** profeng 0.146 (0.032)** 0.038 (0.022)* 0.151 (0.056)** incdech10 0.531 (0.022)** 0.046 (0.015)** 0.462 (0.051)** depkid 0.062 (0.014)** 0.049 (0.028)* concess -0.190 (0.016)** 0.026 (0.011)** -0.133 (0.034)** sinpar -0.062 (0.025)** -0.040 (0.048) excelh 0.092 (0.025)** -0.182 (0.015)** 0.149 (0.038)** majcity 0.047 (0.013)** -0.001 (0.009) 0.023 (0.023) vgoodh 0.089 (0.023)** -0.164 (0.014)** 0.139 (0.035)** inregn 0.044 (0.015)** 0.006 (0.010) 0.041 (0.028) goodh 0.055 (0.022)** -0.113 (0.013)** 0.091 (0.027)** workft -0.138 (0.026)** -0.079 (0.011)** -0.100 (0.054)* athritis 0.005 (0.013) 0.018 (0.008)** -0.004 (0.011) workpt -0.116 (0.025)** -0.034 (0.011)** -0.135 (0.055)** cancer 0.038 (0.030) 0.117 (0.017)** -0.019 (0.035) workstud -0.074 (0.045)* -0.035 (0.029) -0.046 (0.079) heart 0.022 (0.012)* 0.035 (0.007)** 0.003 (0.012) studyft 0.173 (0.032)** -0.080 (0.022)** 0.056 (0.072) diabetes -0.011 (0.022) 0.047 (0.013)** -0.028 (0.019) unemp -0.087 (0.040)** -0.059 (0.023)** -0.041 (0.082) asthma 0.010 (0.016) 0.018 (0.010)* 0.000 (0.013) prof 0.177 (0.022)** 0.188 (0.048)** osteo 0.066 (0.022)** 0.012 (0.014) 0.047 (0.020)** trades 0.093 (0.026)** 0.084 (0.055) smokedly -0.155 (0.012)** -0.122 (0.022)** clerk 0.236 (0.038)** 0.255 (0.088)** alchirsk -0.057 (0.024)** -0.045 (0.021)** intsales 0.127 (0.024)** 0.199 (0.053)** overwt -0.009 (0.010) -0.007 (0.008) prodtran 0.062 (0.028)** 0.083 (0.061) noexcise -0.032 (0.010)** -0.025 (0.009)** elsales 0.052 (0.028)* 0.062 (0.060) copay -0.003 (0.000)** -0.002 (0.001)** degree 0.124 (0.017)** 0.010 (0.011) 0.136 (0.030)** bed -0.009 (0.007) -0.014 (0.016) P(.|x) 0.441 (0.005)** 0.162 (0.003)** 0.433 (0.033)** YH YP |YH = 1 YI YH YP |YH = 1 YIAge
a ge30
0.081 (0.01 7)**
-0.024 (0 .01 1)**
0.1 43 (0.033 )**
a ge40
0.189 (0.01 7)**
-0.052 (0 .01 2)**
0.1 98 (0.033 )**
a ge50
0.312 (0.01 8)**
-0.058 (0 .01 2)**
0.3 45 (0.035 )**
a ge60
0.448 (0.02 2)**
-0.065 (0 .01 4)**
0.5 21 (0.044 )**
a ge70 +
0.441 (0.02 4)**
-0.051 (0 .01 4)**
0.5 51 (0.043 )**
Y
IY
HY
P|Y
H=
1
•
Age is a significant in all three equations.
•
The probability of purchase of PHI is found to increase
with age with a slight drop-off for the 70+ age group.
(similar evidence in prior studies)
•
The probability of private care utilisation increases
progressively as individuals get older
•
In contrast, the probability of hospital admission has a
U-shaped distribution with age, with the young and the old
age groups more likely to get admitted.
Employment and Occupation
•
when we control for other factors such as income and
occupations, those who work are less likely to purchase PHI
and use private health care than those NLF (base case)
•
PHI purchase and use of private hospital care is also
associated with individuals' occupations.
– Labourers (base case) have the lowest chances of purchasing
PHI and opting for private hospital care than individuals in any
other occupation.
workft
-0.138 (0.02 6)**
-0.079 (0 .01 1)**
- 0.1 00 (0.054 )*
workp t
-0.116 (0.02 5)**
-0.034 (0 .01 1)**
- 0.1 35 (0.055 )**
workstud
-0.074 (0.04 5)*
-0.035 (0 .02 9)
- 0.0 46 (0.079 )
studyft
0.173 (0.03 2)**
-0.080 (0 .02 2)**
0.0 56 (0.072 )
u nemp
-0.087 (0.04 0)**
-0.059 (0 .02 3)**
- 0.0 41 (0.082 )
p rof
0.177 (0.02 2)**
0.1 88 (0.048 )**
trad es
0.093 (0.02 6)**
0.0 84 (0.055 )
cler k
0.236 (0.03 8)**
0.2 55 (0.088 )**
in tsale s
0.127 (0.02 4)**
0.1 99 (0.053 )**
p rodtran
0.062 (0.02 8)**
0.0 83 (0.061 )
e lsa les
0.052 (0.02 8)*
0.0 62 (0.060 )
Y
IY
HY
P|Y
H=
1
Lifestyle factors, Household
Characteristics
•
Health related lifestyle factors such as heavy smoking,
drinking at high risk levels, lack of exercise and being obese
are all negatively related to insurance decision.
– More than poor health such factors indicate risk attitudes towards health. i.e. A
decision-maker with such characteristics is less likely to indulge in a risk-averse
behaviour such as PHI purchase.
•
The presence of dependant kids is likely to be a significant
stimulus for getting insured from both the risk averseness and
financial point of view.
– The positive and significant coefficient on this indicator supports the hypothesis.
•
On the other hand, single parents are found to be less likely to
purchase PHI. Their decision to purchase insurance may be
potentially constrained by their financial situations.
Education and Income
•
Education is likely to increase individuals’ awareness of health care
services and the benefits of purchasing a private health insurance.
– The insurance decision and private health care utilisation are both found
to be strongly associated with education.
– degree holders are more likely to get insured and also more likely to use
private health care than someone who has completed less than
secondary education.
•
Higher household income is associated with a higher probability of
purchasing PHI and a higher probability of private health care
utilisation.
– Note that tax incentives can be a significant stimulus for purchasing
private health insurance. A flat Medicare levy with a progressive income
taxation system encourages those on higher incomes to purchase
private insurance irrespective of whether they would use private sector
facilities (Fiebig et al 2006).
Self-Assessed Health
•
Medical need is a potential predictor of health care utilisation. Those who are in
good health are less likely to access health care services.
– The results of the hospital utilisation equation support this hypothesis indicating
that the less healthy individuals are, the more likely they are to get admitted into
hospitals.
– However, we obtain a positive relationship between individuals’ self-assessed
health and the probability of purchasing PHI and the probability of using private
care.
•
counter intuitive to the hypothesis of moral hazard and adverse selection into insurance
•
such finding is not unusual and has been obtained in several previous studies and has
often been associated with risk-related behaviours. i.e. people who are careful about their
health are also more likely to engage into risk averse activities such as purchasing a PHI.
e xcelh
0.092 (0.02 5)**
-0.182 (0 .01 5)**
0.1 49 (0.038 )**
vgoo dh
0.089 (0.02 3)**
-0.164 (0 .01 4)**
0.1 39 (0.035 )**
g oodh
0.055 (0.02 2)**
-0.113 (0 .01 3)**
0.0 91 (0.027 )**
Objective measures of Health
• Some more objective measures of health status in terms
of long-term conditions such as arthritis, cancer, heart
disease, diabetes, asthma and osteoporosis.
– Not related to the choice of private health care or insurance
purchase
– But significantly related with hospital utilisation.
a thr itis
0.005 (0.01 3)
0.018 (0 .00 8)**
- 0.0 04 (0.011 )
can ce r
0.038 (0.03 0)
0.117 (0 .01 7)**
- 0.0 19 (0.035 )
h eart
0.022 (0.01 2)*
0.035 (0 .00 7)**
0.0 03 (0.012 )
d iabetes
-0.011 (0.02 2)
0.047 (0 .01 3)**
- 0.0 28 (0.019 )
a sthma
0.010 (0.01 6)
0.018 (0 .01 0)*
0.0 00 (0.013 )
o steo
0.066 (0.02 2)**
0.012 (0 .01 4)
0.0 47 (0.020 )**
Y
IY
HY
P|Y
H=
1
Cost of insurance and cost of
access to private hospitals
• No data on cost of insurance!
• Average state-level copayments are used as a
measure of the cost of private care.
– Negative effect- the higher the copayments the lower
is the probability of purchasing PHI or the probability
of private care utilisation.
• Those who have concession cards have lower
probability of insurance purchase and private
hospital care.
Quality of health service
•
The effect of the quality of public health care has been
identified as an important determinant of insurance decision in
previous studies
•
A common measure of public hospital care is waiting list and
queuing.
Two different measures of waiting list at state level:
– average waiting time (i.e. days waited at 50th percentile)-insignificant
effect
– the proportion of individuals who waited for more than a year for elective
surgery- positively and significantly related to PHI purchase.
•
Not included in final model - given the Australian waiting list
data at state level is known to be inconsistent with regard to
their collection and presentation (Hopkins and Kidd, 1996;
AIHW, 2007)
Quality of health service
•
Instead we use bed density and full-time equivalent (FTE)
medical practitioners in public hospitals as alternate indicators
of quality of public care.
– Measured at state level and by remoteness- more variation.
– The effects of both these variables are found to be negative with respect
to both insurance purchase and private health care service utilisation
(although mostly insignificant).
Effect of PHI
•
Finally, private insurance is found to be an important
determinant of private health care utilisation.
– In particular, those with private hospital cover are 76% more
likely to seek private health care than use public health services.
Predicted Probabilities and
Treatment Effects
Conclusion
•
This study attempts to provide insights on the role of PHI in the
choice that an individual makes between public and private
health care utilisation
– It uses a system of Probit models (MVP) to allow for potential
endogeneity of private insurance participation.
– It also adjust for selection bias due to partial observability since
we only observe individuals’ choices between P/P if they have
visited a hospital.
•
PHI has certainly been identified as an important determinant of
private hospital care utilisation.
•
However, other factors such as perceived quality of care in the
public sector and cost of access were also found to have an
impact on the use of private hospital care.
•
This system approach allows predictions of a range of joint and
conditional probabilities.