1. Impact of Hospital Provider Payment Mechanism on Household Health Service Utilization in Vietnam (preliminary results) Sarah Bales Public Policy in Asia, PhD Conference 27 May 2014

2. Vietnam health system context Largely state healthcare system with a district hospital in almost all districts and commune health stations in 99% of all communes. About 50% of outpatient care obtained from private clinics. Only 5% of hospital beds are private, concentrated in urban areas. Nearly 70% of population covered by social health insurance. Compulsory or fully subsidized scheme accounts for majority. About 24% of the population is self-employed group eligible for contributory voluntary health insurance, but adverse selection.

3. Policy problem Medical cost escalation: Total health expenditure increased from 5% of GDP in 1998 to 7% in 2010. State health spending (including social health insurance) rose from 1.6% to 3.1% of GDP. Underlying cause is fee-for-service payments (use of state administered fee schedule to pay for each service prescribed) in combination with: Incentives to generate revenue surplus by increasing quantity of services provided and reducing cost per service; surplus can be used to top up staff remuneration. Profit-sharing arrangements with private investors in medical equipment at public hospitals, allowed to charge higher than official fees. Distorted state fee schedule with higher fee-to-cost ratios for high tech services. Widespread practice of kickbacks to doctors for prescribing more drugs and costlier drugs. Reduced patient financial barriers to obtaining more care because of health insurance.

4. Policy solution: capitation to replace fee-for- service Change basis of contract for health insurance payment of district hospitals from fee-for-service (FFS) to capitation starting 2010 with gradual roll-out to 100% of district facilities by 2015. Capitation is payment of a fixed amount per year per person registered for care at the facility intended to cover 80% of the costs of their care including referrals. Patients responsible for 20% copayment. No additional charges permitted. Crucial intended changes to the incentives include: Carrot: Motivate reduction in costs by allowing hospital to retain surplus between capitation budget and actual costs up to 20% of total capitation budget. Stick: Hospital responsible for financial losses if spend over budget. Expected outcomes: Reductions in quantity of services provided to insured patients through: Reducing number of visits, Decreasing intensity of care per visit (e.g. number of tests, images, surgeries)->reduced copayments per visit. Increasing use of primary and preventive care Reducing referrals to higher level facilities.

5. Research questions Did switching to capitation payments have the expected effects on quantity of care and type of care received by the insured? Did switching to capitation payments have any perverse effects on uninsured patients for whom the hospital is not constrained by the capitation budget? Do effects of switching to capitation differ across groups in the population or across characteristics of district hospitals? Why? 1)To inform policymakers in Vietnam to adjust the policy if not effective 2)To add to the evidence base on impact of provider incentives on health system outcomes

6. Data Nationally representative cross-section data of households in 2008, 2010 and 2012. Information on individual household members: health insurance coverage and type, number of outpatient and inpatient visits by type of facility, household out-of-pocket outpatient and inpatient payments by type of facility. Administrative data of Vietnam Social Security on type of contract (FFS or capitation) for each hospital. Link households and district hospitals through district id code.

7. Methodology District fixed effects model to control for time invariant differences between district hospitals that would bias estimates of impact of capitation. Control for observable time variant factors at the individual and household level (age group, sex, poverty status, employment status, per capita household expenditure, years of education of households in the sample in each district for each time period, health insurance status). Control for observable time variant factors at the hospital reflecting changes in quality of care, capacity to provide services that could affect demand. Assume that any remaining unobservable time varying factors are the same in treated and control districts, e.g. other policy changes affecting all hospitals at the same time: changes to user fees, changes in copayments, changes in penalties for bypassing without a referral.

8. Impact of applying district hospital capitation on insured individuals-district hospital services Outcome OOP spending on district OP services among district OP users OOP spending on district IP services among district IP users Number of OP visits at district hospital given any health service contact Number of IP visits at district hospital given any health service contact Coefficient -0.450-0.275-0.156-0.051 Cluster standard error (0.149)**(0.235)(0.191)(0.196) ModelOLS Negative binomial Only impact observed is on out-of-pocket spending for outpatient services at the district hospital. Suggests reduction in intensity of services per visit. But no reduction in outpatient or inpatient visits.

9. Impact of applying district hospital capitation on insured individuals-higher facility services Outcome OOP spending on higher level OP services among higher level OP users OOP spending on higher level IP services among higher level IP users Number of OP visits at higher level hospital given any health service contact Number of IP visits at higher level hospital given any health service contact Coefficient 0.107-0.191-0.063-0.290 Cluster standard error (0.197)(0.213)(0.205)(0.218) ModelOLS Negative binomial No statistically significant change in spending on or visits to higher level government facilities

10. Impact of applying district hospital capitation on insured individuals-primary care facility services Outcome OOP spending per PHC facility OP contacts Number of OP visits at PHC clinics given any health service contact Coefficient -0.1490.011 Cluster standard error (0.138)(0.202) ModelOLSNegative binomial No statistically significant change in spending on or visits to lower level government facilities

11. Further work Redo the analysis after adding the 2012 survey results (pending data access) Consider lags in policy impact Clean data on hospitals to get more information on changes in quality-related variables. Find a method for analyzing dichotomous variable on probability of referral and probability of bypassing. Examine sub-groups of the population or sub-groups related to availability of competition in the district.