Insurance or Savings: I li ti Implications off Individual I di id l H Health lth Accounts in China Guan Gong, Ph.D. Shanghai University of Finance and Economics Hongmei Wang Wang, Ph Ph.D. D University of Nebraska Medical Center Lingli Xu, Ph.D. Shanghai University Academy Health June 2010 Individual Health Accounts (IHAs) Individual s savings account restricted to health care Individual’s expenses ◦ Funds accumulated at the end of life are inheritable inheritable. Earliest implementation is in Singapore in 1984 Implemented in urban China since 1998. Promoted by U.S. policy since 2004 Some discussions in other developed countries Implementation of IHAs An important feature is the integration of IHAs and Catastrophic health insurance. Encourage savings for future medical expenditure and reduce moral hazard Reduce R d risk i k pooling li and d iincrease iinequality lit iin iincome redistribution ◦ A healthy person will accumulate large IHA balances, while the chronically ill may accumulate nothing ◦ The scheme may look like a savings plan to the healthy and selfinsurance to the chronically ill. Research Objectives To explore the implications on equality of a health insurance scheme combining IHAs with catastrophic g data from Kunshan China insurance using Limited Li it d d data t prevented t d effective ff ti examination i ti off the th effects and the evidence remains largely descriptive. Most studies on the effects of IHAs in China draw their conclusions based on cross-sectional analysis. This study uses longitudinal data and examines the effects of IHAs from a lifetime perspective. Kunshan Health Insurance System First carried out in 1997 and had been extended to cover all employees and retirees in 2000. By 2005 2005, covered 560 560,000, 000 82% of the population population. A combination of IHAs, Social Risk Pooling (SRP) fund, and d supplemental l t l catastrophic t t hi risk-pooling i k li (SCRP) ffund d Contributions to IHAs and SRP fund: 10% salary ◦ 2% from employees and 8% from employers ◦ About 4%-6.5% to IHAs and 6%-5.5% to SRP fund (by age) Contributions to SCRP fund: 50 Yuan for an employee and 30 Yuan for a retiree. Kunshan Health Insurance System Payment y depends p on type yp of health services,, working g status, total health expenditure, etc. IHAs used for outpatient expenditure and SRP fund for inpatient expenditure Outpatient Employee Retiree IHA Deductible SRP fund CSRP fund Cap 600 20% / 3000 Inpatient Employee Retiree / / 300 600 500 10% <50,000 <50 000 12%-5% 12% 5% 6%-2 6% 2.5% 5% / 50,000-200,000 10%-5% 10%-5% 200,000+ Other Data Longitudinal health expenditure data in Kunshan city city, Jiangsu Province from 2005 to 2007. Individual health insurance record from the Bureau of Social Insurance of Kunshan. A sample l off 7000 enrollees ll were randomly d l selected, l t d with ith 6355 have all full records for all three years. Sample p Statistics Variable Mean Variance Age(2005) Gender(=Male) 43.7 0.56 Health expenditure from 2005 to 2007 N expenditure No dit 2005 Only 2006 Only 2007 Only 2005 and 2006 2005 and 2007 2006 and 2007 All three years 0.30 0 30 0.06 0.05 0.07 0.15 0.11 0.14 0.12 Variable Mean Variance 10.79 Contributions to IHAs 0.49 2005 497 172 2006 503 305 2007 512 295 Health expenditure 2005 1,521 12,569 0 46 0.46 O t ti t Outpatient 626 1 104 1,104 0.24 Inpatient 895 12,399 0.17 Health expenditure 2006 1,597 9,746 0.30 Outpatient 645 1,534 0.18 Inpatient 953 8,889 0.26 Health expenditure 2007 1,615 11,317 0.46 Outpatient 652 2,011 0.32 Inpatient 963 10,144 • Contributions to IHAs can be used as a proxy for salary income Concentration of Health Expenditure 1 Year 2005 Year 2006 Year 2007 Average of three years 0.9 Perce ent of Expen nditures 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 Percent of Individuals 0.7 0.8 0.9 1 Percent Distribution of 2006 and 2007 Expenditure, b 2005 E by Expenditure dit IInterval t l and db by A Age (5 (5x5 5 ttable) bl ) 2005 Age 16-29 30-39 40-49 50-59 60+ 2005 Expenditure 0 0-300 300-1000 1000-4000 Above 4000 0 0-300 300-1000 1000-4000 Above 4000 0 0-300 300-1000 1000 4000 1000-4000 Above 4000 0 0-300 300-1000 1000-4000 Above 4000 0 0-300 300-1000 1000-4000 Above 4000 0 54 9 (48 54.9 (48.4) 4) 32.2 (24.2) 26.7 (19.9) 19.4 (19.4) 17.8 (17.8) 38.0 (28.1) 35.7 (26.9) 28.0 (27.6) 25.2 (25.8) 15.6 (14.1) 36.2 (19.0) 45.0 (23.5) 34.7 (14.7) 29 4 (14 29.4 (14.6) 6) 15.2 (10.5) 30.7 (15.1) 37.0 (23.4) 26.9 ((19.1)) 14.7 (16.1) 13.0 (13.0) 35.1 (30.6) 40.3 (22.1) 20.1 (10.7) 19.9 (14.5) 14.7 (14.7) 2006 (2007) Expenditure 0-300 300-1000 9 7 (14.9) 9.7 (14 9) 21 0 (21 21.0 (21.4) 4) 31.5 (30.2) 24.8 (17.5) 11.4 (2.5) 33.1 (54.7) 5.4 (8.5) 20.2 (18.6) 13.1 (14.0) 14.0 (13.1) 18.1 (20.3) 29.1 (23.1) 16.2 (12.9) 19.3 (32.1) 7.6 (8.7 ) 40.0 (43.6) 15.4 (18.5) 16.3 (23.5) 8.2 (8.2 ) 18.5 (8.2 ) 18.0 (27.5) 30.7 (27.2) 16.5 (21.8) 29.9 (30.5) 6.0 (31.0) 31.7 (27.0) 6 0 (13.3) 6.0 (13 3) 23 8 (16 23.8 (16.6) 6) 18.6 (18.6) 9.3 (10.5) 28.0 (18.6) 25.8 (46.9) 35.9 (22.3) 20.1 (30.4) 9.0 ((11.7)) 39.3 ((41.8)) 9.5 (3.8 ) 26.7 (10.9) 3.1 (12.4) 21.0 (18.5) 16.2 (8.1) 30.6 (21.6) 23.4 (23.4) 9.1 (9.1) 14.1 (7.4 ) 36.2 (46.3) 7.8 (5.4 ) 12.7 (29.5) 17.3 (17.3) 17.3 (17.3) 1000-4000 14 0 (13 14.0 (13.8) 8) 4.0 (16.8) 16.5 (19.9) 41.1 (37.2) 24.3 (38.3) 11.2 (26.4) 20.0 (23.3) 18.6 (10.2) 37.0 (26.6) 29.6 (51.9) 12.6 (21.1) 5.9 (14.0) 21.0 (24.3) 36 4 (51 36.4 (51.0) 0) 17.4 (46.5) 11.4 (11.7) 2.7 (20.1) 21.9 ((20.1)) 37.4 (59.2) 22.2 (15.4) 8.1 (29.7) 15.6 (27.3) 16.8 (16.8) 40.4 (30.7) 14.7 (34.7) Above 4000 0 4 (1 0.4 (1.5 5) 7.4 (11.4) 12.3 (3.0 ) 14.0 (16.3) 30.8 (16.8) 3.6 (2.2) 8.8 (4.8) 5.8 (9.8) 6.2 (5.6) 28.2 (17.8) 2.4 (5.3) 2.8 (10.4) 6.7 (3.0) 4 6 (4 4.6 (4.6) 6) 39.5 (14.0) 4.0 (7.8) 4.4 (3.8) 3.0 ((7.4)) 12.3 (10.0) 40.7 (40.7) 9.9 (9.9) 11.7 (18.2) 12.8 (18.8) 19.3 (19.9) 36.0 (16.0) Methods Four-part model estimation using data from 2005 to 2007 (Manning et al. 1981,1987 and Duan et al. 1983, 1984). ◦ Model estimates based on demographic and lagged expenditure Non-parametric method for simulation of lifetime distribution of individual health expenditures and payment sources sources. (Eichner et al. 1998) ◦ Simulate individual health expenditures in a life time based on the four part model estimates ◦ Calculate payment from IHAs and end-of-life IHA balances ◦ Calculate expenditure paid by out-of-pocket Four--part Model Specification Four Prob. of health expenditure: Pr MED i 0 x1i 1 , Prob. of inpatient expenditure: PrINPi 0 MEDi 0 x2i 2 , Outpatient expenditure: Inpatient expenditure: logOUTPi MEDi 0 x1i 3 vi , logINPi INPi 0 x2i 4 wi , X1i include Age, g , Gender,, Interaction between age g and g gender,, Annual contribution, and lagged expenditures in year t-1 and t-2 X2i include X1i variables, Outpatient expenditure in year t, and whether h h outpatient i expenditure di reaches h 1 1500 00 iin year t Estimation Results of FourFour-Part Model a a Gender=Male First Part: Probit Probability of health expenditure -0.489 (0.123)*** Second Part: Probit Probability of inpatient expenditure 0.530(0.139)*** Age < 30 in 2007 30<Age<40 in 2007 40<Age<50 in 2007 -0.785 (0.138)*** -0.853 (0.115)*** -0.411 (0.122)*** 0.103 (0.191) 0.159(0.137) 0.491(0.137)*** 50<Age<60 in 2007 -0.284 (0.125)** Variables 0.022(0.151) Third Part: OLS Outpatient expenditure -0.296(0.076) *** *** -0.418(0.094) 0.052 (0.071) *** a Fourth Part: OLS Inpatient expenditure *** 1.175 (0.187) -0.246(0.261) -0.127(0.205) -0.303(0.073) 0.015 (0.074) 1.209 (0.180) *** 1.337 (0.232) *** -1.452(0.405) *** * -1.710(0.304) *** ** -2.176(0.275) *** -1.914(0.280) *** Age < 30 and Gender=male 0.096 (0.163) -0.669(0.273)** 0.116 (0.125) 30<Age<40 and Gender=male 0.466 (0.133)*** -0.821(0.184)*** -0.174(0.090) 40<Age<50 and Gender=male 0.268 (0.139)* -0.858(0.72)*** 50<Age<60 and Gender=male 0.126 (0.145) 0.231(0.179) 0.222 (0.090) -0.139(0.094) DM1:Had expenditure in 2006 only 0.298 (0.155)* -0.707(0.235)*** -1.022(0.103) *** 2.510 (0.407) *** DM2:Had expenditure in 2005 only 0.459 (0.130)*** -1.080(0.455)** -0.623(0.208) *** -1.599(0.554) *** DM3: Had expenditure 2005 & 2006 0.464 (0.184)** -1.343(0.256)*** -2.080(0.122) *** -2.971(0.370) *** Log expenditure in 2006 if DM1=1 0 184 (0.025)*** 0.184 (0 025)*** 0 114(0 033)*** 0.114(0.033)*** 0 220 (0.015) 0.220 (0 015) *** -0.468(0.050) 0 468(0 050) *** Log expenditure in 2005 if DM2=1 -0.049 (0.020)** -0.202(0.066)*** * 0.168 (0.067) ** Log expenditure in 2006 if DM3=1 0.197 (0.025)*** -0.253(0.031)*** 0.363 (0.016) *** 0.241 (0.036) *** Log expenditure in 2005 if DM3=1 -0.068 (0.024)*** -0.046(0.028)* 0.051 (0.015) *** Log outpatient expenditure in 2007 — 0.216(0.041)*** — 0.415 (0.073) *** Outpatient expenditure in 2007 >1500 — -0.074 (0.092) — -0.538(0.137) *** -0.233 (0.049)*** — 0.219 6355 -0.010 (0.091) — 0.127 3902 Log contributions to IHAs in 2007 Variance of random shocks Pseudo R-square/R-square Sample a 0.064 (0.033) *** 0.192 (0.044) 1.224(0.027) 0.231 3902 a 0.041 (0.041) *** 0.911 (0.081) 0.652(0.006) 0.994 264 standard deviations are in parentheses, *** indicates significant at 0.01, ** indicates significant at 0.05, and * indicates significant at 0.10 Simulation Generate a sample of 10,000 ages between 16 to 29, based on actual demographic data in 2005 including age, gender, and annual contribution. Generate health expenditure in 2006 based on actual distribution in 2005 and the 5x5 table. Predict P di 200 2007 expenditure di using i simulated i l dd data iin 200 2005 and d 2006 and repeat the process to get lifetime expenditure till age 80. A Assumptions: ti ◦ a 1% increase in annual contribution to IHAs. (0%, 3%) ◦ Men retire at 60 and women at 50; pension is calculated based on contribution and a fixed formula. Model Fit _ Distribution in 2007 Age < 30 All Age Groups 1 1 Actual expenditure vs. Predicted expenditure distribution in 2007 .9 2007a 2007p 0.9 .8 0.8 Cumulative Probability y 2007a 2007p .7 .6 .5 0.7 0.6 0.5 .4 0.4 .3 .2 0 2 4 6 8 Log of Expenditure 10 12 14 0 2 4 30 <= Age < 40 2007a 2007p .9 2007a 2007p 0.9 .8 0.8 .7 .6 .5 0.7 0.6 0.5 .4 0.4 .3 0.3 .2 0 2 4 6 8 Log of Expe nditure 10 12 0.2 14 0 2 4 50 <= Age < 60 6 8 Log of Expenditure 10 12 10 12 Age >= 60 1 1 2007a 2007p .9 2007a 2007p 0.9 .8 0.8 .7 .6 .5 .4 0.7 0.6 0.5 0.4 .3 0.3 .2 0.2 .1 12 1 Cumulative Probability Fits well for the total sample and at different age categories categories. 10 40 <= Age < 50 1 Cumulative Probability 6 8 Log of Expenditure 0 2 4 6 8 Log of Expe nditure 10 12 14 0.1 0 2 4 6 8 Log of Expenditure Model Fit _ Distribution at Selected Ages 30 <= Age < 40 Age < 30 1 1 Distributions of lifetime predicted expenditure vs. distributions of actual expenditure at selected ages 0.8 .8 .7 .6 .5 0.6 0.5 .4 0.3 0 2 4 6 8 Log of Expenditure 10 12 0.2 14 0 2 4 6 8 Log of Expenditure 40 <= Age < 50 0.8 0.6 0.5 0.4 0.6 0.5 0.4 0.2 0 2 4 6 8 Log of Expenditure 10 12 14 Age >= 60 1 2007a predicted value .8 .7 .6 .5 .4 .3 3 .2 .1 0 0.7 0.3 0 3 0.3 0 2007a predicted value 0.9 0.7 9 .9 12 1 2007a predicted value 0.8 0.2 10 50 <= Age < 60 1 Cumulative Probability Fit well for f all different ff age categories. 0.7 0.4 0.9 2007a predicted value 0.9 Cumulative Probability 2007a predicted value .9 2 4 6 8 Log of Expenditure 10 12 14 0.1 0 2 4 6 8 Log of Expenditure 10 12 Simulation Results_IHA Results IHA Balance Most people retain a zero or low level of end of life IHA balance ◦ 58% end up with nothing in IHAs at the end of life ◦ 78% end up with balances less than 1000 Yuan ◦ 87% end up with balances less than 10% of total contribution Di t ib ti off EOL IHA b Distribution balance l and d It Its Sh Share off T Total t lC Contribution t ib ti IHA balance (%) <1,000 <2,000 , <5,000 <10,000 <50,000 , <80,000 <100,000 78.1 84.3 87.5 88.5 96.7 99.1 99.3 IHA balance as a % of total contribution <0% <10% <20% <50% <70% <90% <100% (%) 58.0 87.1 88.3 89.8 90.2 90.2 100.0 Simulation Results Results_Financial Financial Sources 52% of lifetime health expenditure is from individual financial sources (IHAs and Out-of-pocket). Out-of-pocket expenditure accounts for 59% of the individual financial burden Percent Distribution of sources for expenditure, cumulative by age Age 30 45 5 55 65 80 IHA (I) 38.9 34.6 3 6 29.6 25.0 21.3 OOP (II) 38.5 39.9 39 9 37.5 33.1 30.5 SRP SCRP (III) (IV) 22.3 0.0 24.1 1.4 27.5 5.5 33.5 8.4 39.2 9.1 Individual Burden (I+II) 77.4 74.5 5 67.0 58.1 51.8 Social Risk Pooling (III+IV) 22.6 25.5 55 33.0 41.9 48.3 Simulation Results_OutResults_Out _ -of of--p pocket Expenditure p Large variation in out-of-pocket spending in the form of deductibles and co-insurance Average out-of-pocket expenditures for people with a IHA balance 5,000 or less is about 50 times those for people with a IHA balance greater than 5,000 Average out-of-pocket expenditure by end-of-life IHA balance IHA balance 0 <=1,000 <=2,000 < 5 000 <=5,000 >5,000 Average OOP expenditure 64,900 64 900 60,800 59,700 59 200 59,200 1,200 IHA balance as a percentage of total contribution <=1% <=3% <=5% < 10% <=10% >10% Average OOP expenditure 63,300 63 300 61,600 60,600 59 500 59,500 1,300 Conclusions Relatively equal low level of IHA balance at the end of life suggests that IHAs primarily serve as self-insurance purpose rather than as saving g accounts. The equality of IHA balance accumulation at the end of life comes at the cost of large variation in out-of-pocket out of pocket spending in the form of deductibles and coinsurances. Under Kunshan’s Kunshan s current health insurance system, system most people will deplete their IHA’s and pay a large fraction (30-40%) of lifetime expenditure p out-of-pocket. p Limitations Health expenditure data are only those captured in the system system, expenditure on pharmaceuticals from undesignated stores are not taken account. Only have three years of actual data to estimate the four-part model. model Strong assumptions were made while simulating lifetime expenditure; sensitivity analysis were conducted to provide a robust check. Total T t l contribution t ib ti is i used d as a proxy ffor salary l iincome b butt other th component of income are not included in the model. Analysis was based on data for employees and retirees. Policy Discussion The effect of IHA scheme on equality of income distribution is complicated and has to be examined under specific health insurance schemes. Under current health insurance system in Kunshan, the demand-side demand side cost sharing should be reduced reduced, or subsidies may have to be provided to the low-income to ensure access to care. IHA s integrated with various types of insurance mechanisms IHA’s may lead to different outcomes on equity, efficiency, and quality. Deliberations are needed in various aspects p of the financing g and payment system to reach optimum performance.