The impact of non profit health insurance on treatment seeking behavior the case of vietnam(vhlss 2006

By using very rich date set from the Vietnam Household LivingStandards Surveys 2006 VHLSS 2006, the thesis has explicitly estimatedthe effect of health insurance on utilization of all ty

VIETNAM-NEITHERLANDS CENTER FOR DEVELOPMENT ECONOMICS AND PUBLIC POLICY

The impact of non-profit health insurance

on treatment seeking behavior:

The case of vietnam (VHLSS 2006)

A thesis presented by

LE TUAN SANG

In partial Fulfillment of the Requirement for

Obtaining the Degree of Master of Arts in Development Economics

Supervisor: Prof Ardeshir Sepehri, PhD.

HANOI 2011

I hereby certify that this thesis has used materials that has been accepted

or with out copied of any other degrees or diplomas at any other academic institutions By learning and my knowledge, the thesis contains no sources and materials previously published or written by other persons, except the references listed.

Hanoi, June 2011

Le Tuan Sang

Sincere thanks to all my friends for their time and guidance in the course

of my econometrical estimation and interpretation of Probit regressionresults

Finally, I would sincerely express to thanks my family and specially from

my wife who gave me an encouragement and a contritbution of financeduring my study

Le Tuan Sang June 2011

In Vietnam, there are some studies on health care services and it gives outsome significant ruslts of influence on policy makers However, these studiesstill have some limitations due to data as well as information updates

By using very rich date set from the Vietnam Household LivingStandards Surveys 2006 (VHLSS 2006), the thesis has explicitly estimatedthe effect of health insurance on utilization of all types of health care servicesand providing detailed information on both of individual health andhousehold status and the type of provider sought covering all provinces.Beside, the thesis also estimated the impact of health insurance on health careutilization by the poor In addtion, by the stusdy period covered 12 months,the thesis is better capture seasonal effects as well as other time- relateddimensions of health treatment behaviors within a year Findings of the thesishas been suggested that individuals are propensity to visiting or admission tohealth facilitaties, specially for public health facilitation and the insured is noeffect on the frequency of outpatient visit to public providers Deeply, theresult of thesis showed that the groups of health insurance/free health card forthe poor and children under 6 YOs is commune clinics and the group ofcompulsory health insurance mainly visited public polyclinics, districthospital or higher level of public health facilities

TABLE OF CONTENTS

DECLARATION 2

ACKNOWLEDGEMENT 3

ABSTRACT 4

LIST OF ABBREVIATIONS 7

LIST OF TABLES 8

LIST OF FIGURES 9

CHAPTER I 10

1 Introduction: 10

2 Research questions 13

3 Objectives and scope of the study: 13

4 Data and Methodology: 14

5 Structure of the thesis 15

CHAPTER II: LITERATURE REVIEW 16

1 Literature review in general 16

2 Studies of utilization on healthcare in Vietnam 18

CHAPTER III: HEALTH CARE SYSTEM AND HEALTH UTILIZATION IN VIETNAM 21

1 Introduction of healthcare system in Vietnam 21

1.1 Public health care system 21

1.2 Private health care system 22

2 Health care utilization 24

2.1 No treatment and self-medication 24

2.2 Inpatient services 25

2.3 Outpatient services utilization 26

2.4 The schemes of health insurance in Vietnam 27

2.5 Heath care for the poor 28

CHAPTER IV: METHODOLOGY AND EMPIRICAL RESULT 30

1 Methodology and data 30

1.1 Methodology 30

1.1.1 Logit model 30

1.1.2 Count regression models 31

1.2 Model selection 33

1.3 Data 33

2 Variable description 34

3 Descriptive analysis 72

4 Econometric results 76

CHAPTER V: FINDINGS AND POLICY IMPLICATIONS 83

CHAPTER VI: CONCLUSION 84

REFERENCES 87

LIST OF ABBREVIATIONS

CHI : Compulsory Health Insurance

CHS : Commune Health Station

DHO : District Health Office

DPC : District People’s Committee

GSO : General Statistics Office

HCFP : Health Care Fund for the Poor

JHUES : Healthcare Utilization and Expenditure Survey

JHUES : Jordan Healthcare Utilization and Expenditure Survey

MOH : Ministry of Heath

OOP : out-of-pocket

PHD : Provincial Health Department

PPC : Provincial People’s Committee

VHI : Voluntary Health Insurance

VHLSS : Vietnam Household Living Standards Surveys

LIST OF TABLES

Table 1: Health insurance coverage by expenditure quintiles 35

Table 2: The use of outpatient care across providers by the insurance status and the type of insurance 36

Table 3: Decomposition of utilization by health facilities, inpatients 37

Table 4: Number of outpatient contacts per person 38

Table 5: Number of inpatient contacts per person 39

Table 6: Probability of health facilities contacts in general Odd ratio 40

Table 7: The results for contacts and frequency of outpatient contacts at the commune health centres 48

Table 8: Probability of out-patient contacts and Frequencies of contacts: public polyclinics/hospitals 56

Table 9: Probability of out-patient contacts and Frequency of contacts: Private clinics/hospitals 61

Table 10: Probability of in-patient contacts and frequency of contacts: public polyclinics/hospitals 66

LIST OF FIGURES

Firgue 1: Distribution of inpaatient contacts across healthfacilities by ruraland urban population 25Figure 2: Health insurance coverage by expenditure quintiles 72Figure3: Percentage of health contacts of different groups of healthinsurances 73

CHAPTER I

1 Introduction:

In recent years, Vietnam has been encouraging and strengthening theexplicit role of insurance in health system To achieve the “MillenniumDevelopment Goals”, the social care services need to be extended andimproved Health insurance is a kind of socialized project which helps sharethe financial burden of illness between the Government and the population.The insurance coverage was introduced in 1992 in Vietnam, the scheme wasinitially compulsory, then later extended to include a voluntary scheme and afree health-card scheme for the poor was also initiated afterwards TheCompulsory Health Insurance (CHI) scheme includes current and retired civilservants and the employees of state enterprises as well as those in largeprivate enterprises with more than ten employees, employees of foreignowned enterprises and organization, the scheme covers the cost of inpatientand outpatient treatment at hospitals, subject to some ceilings The VoluntaryHealth Insurance (VHI) scheme is in principle open to all those not eligiblefor coverage under the CHI scheme, including the self-employed, employees

of small enterprises, family members of the insured and governmentemployees at or below the district level Currently, the VHI scheme is open

to all community groups provided that minimum community thresholds aremet, including 20 percent of households in a commune or ward and 30percent of students in a given school (Nguyen & Akal, 2003) Although,Giang (2006) gave an evidence that many provinces of Vietnam have beenreluctant to encourage voluntary insurance scheme, partly due to high costresulting from adverse selection and Jowett (2004) reviewed that VietnameseGovernment-organized voluntary schemes may also be crowded by informalrisk-sharing networks Besides, ever since 2002 there has been a healthinsurance scheme implemented for the poor in Vietnam which covers

residents of communes with very difficult socio-economic status and ethnicminorities in disadvantaged provinces of Vietnam The scheme is funded by avariety of charity and donor organizations and the Government through apoverty alleviation program On the other hand, the scheme has been also re-organized and funded more adequately under a national program according towhich provinces and centrally-run cities are instructed to establish HealthCare Funds for the Poor (Sepehri, Simpson & Sarma, 2006)

s In term of theory, health insurance may affect health care utilizationand treatment-seeking behaviors in some manners Firstly, it may reduce theprice of health care at the time of purchasing, so that it helps the insuredindividuals have increasing access to health care services In this aspect thethesis will examine the individuals’ tendencies in using health care servicesthat are propensity to visiting or admission to health facilitaties in general.Secondly, health insurance can probably help the insured reduce health careexpenditure through a risk-sharing and cost-pooling mechanism Since one ofthe insured’s expectations is to reduce costs of health care when beinginsured, health insurance may affect their choices as to the type and place ofhealth care services as compared to when not being insured This isespecially true for the poor when high cost of health care incurred could lead

to their health care catastrophe In this dimension, we examine probability ofvisiting/admission different types of health facilities as well as frequencies ofvisiting/admission of those types Thirdly, besides the influence of healthinsurance on treatment-seeking behaviors of the insured, other socio-economic factors could also affect the insured’s choices to a different extent.For example, education -could partially determine people’s awareness ofhealth status- can be considered an important factor affecting the insured’streatment seeking behavior and therefore this thesis will clarify their impactsand specifically correlation with the factor of being insured under the threenamed insurance schemes

Several studies have examined the influence of health insurance on theutilization of health services including papers written by Jowett (2004),Tridevi (2002) and Giang (2006) However, all those studies remain subject

to some limitations due to lack of comprehensive data or too strongassumptions, which could lead to biases when applying into the recentsituation of the health insurance sector In order that this thesis can fill in theknowledge gap left by the afore-mentioned studies, the thesis examines theimpact of Vietnamese health insurance on treatment seeking behaviors of theinsured, with a further focus on those of the insured under the healthinsurance for the poor scheme, which is a point of high significance andpractical interest Furthermore, by taking advantage of VHLSS 2006 datasetand rich references, the thesis will be extended to study some more newaspects Firstly, the VHLSS 2006 allows us to explicitly estimate the effect ofhealth insurance on utilization of all types of health care services Theanalyses of insurance types allow us to test for their differences, if any, onpattern of utilization and it also allows us to test probability of selection fortreatment seeking behaviors Secondly, the VHLSS 2006 is morecomprehensive providing detailed information on both of individual healthand household status and the type of provider sought covering all provinces

We are also able to estimate the impact of health insurance on health careutilization by the poor Thirdly, the study period cover 12 months especially

it is the first for outpatient contact which are believed to better captureseasonal effects as well as other time- related dimensions of health treatmentbehaviors within a year Findings of the thesis will be given out forresearchers and policy makers who should be interested in finding solutions

to the problems and challenges arisen from the targeted universalization ofcoverage among the whole population by the year of 2015, as set out by theVietnamese Government

3 Objectives and scope of the study:

- Objectives:

The study has three main goals as follows: (i) it will provide somebackground information on health insurance and a discriptive analysis of data(ii) It presents empirical analysis of the impact of health insurance schemes,with focus on the health insurance for the poor, on the treatment seekingbehaviors of the insured over 12-month period which are believed to captureseasonal health status and treatment behavior as well as other seasonal relatedfactors; and (iii) The thesis provide some policy recommendations

- Scope of the Study:

The thesis uses the rich and most updated dataset of Vietnam Householdsand Living Standards Survey (VHLSS 2006) to estimate the impact ofdifferent health insurance schemes over 12-month period Firstly, theprobabilities of outpatient contracts and inpatient admission in general areinvestigated Secondly, it will examine of health treatment seeking behaviorsfor specific types of health facilities- public health facility, governmentpolyclinic/ hospital, and private clinic/hospital

4 Data and Methodology:

a Data:

The study uses the Vietnam Household Living Standards Surveys 2006

(VHLSS 2006) which is a very rich dataset collected from Vietnam in aWorld Bank funded program and executed by the Vietnam General StatisticsOffice (GSO) The dataset includes, among other things, full information ondifferent types of insurance, health care expenditure by the insured, types ofhealth care services as well as types of health care providers sought by theinsured when falling ill, and other socio-economic factors related to theinsured such as sex, age, health status, proximity to health care providers ormarital status

In addition, the study also uses other data from the GSO and the Ministry

of Health as references to enrich the statistical analysis of the sector

b Model:

Appropriate models are selected to answer research questions As probability

of contact or admission could be defined as a binary variable, the logit model

is used to study this theme Meanwhile Count regression models for discretedependent variables are suitable for investigating density of contacts oradmission Justification of model selection will be presented chapter IV-methodology and empirical results

5 Structure of the thesis

The thesis will be presented in the following structure: Chapter I presentsproblem statement, methodology, data, research questions as well as therelevance of the study Chapter II includes some literature reviews, asummary of recent developments of the health care and health insurancesectors of Vietnam Chapter III will be introduced the health care system andhealth utilization in Vietnam Chapter IV presents methodology as well asempirical results of the estimation models A short description of the data andinformation contained in the variables will be presented in this chapter.Chapter V shows the findings and implications while Chapter VI gives outconclusions to the thesis

CHAPTER II LITERATURE REVIEW

This chapter examines the recent studies on the utilization of health care services in Vietnam among the insured patients The focus is on the methodology, data and findings in order to give a comparision with the estimation results and findings which will be addressed in the later parts of the thesis.

1 Literature review in general

There is a rich literature on the effect of health insurance on the pattern

of health service utilization all over the world In developed countries,previous studies show that health insurance has positive impact on the healthservice utilization Cameron et al (1987) used household level data fromAustralia Health Survey (1977-1978) and appropriate econometric models inorder to assess the determinants of health care utilization Firstly, the paperfinds that utilization choices vary considerably across insurance types, butthat health status proves to be a stronger determinant of utilization of carethan for insurance choice Secondly, the paper finds an evidence of both self-selection and moral hazard due to insurance for some health care types(Cameron et al., 1987) (luu y: Anh lam ro diem nay them nhe, kho hieuqua)

As long as the studies of the same topic in middle-income countries areconcerned, Ekman (2007) examined the impact of health insurance onoutpatient utilization and expenditure His study used rigorous quantitativemethods with household level data from Jordan to analyze the effect ofmultiple health insurance availability in terms of utilization of outpatientcare, intensity of use, and individual out-of-pocket (OOP) spending on care.Specifically, the study assesses whether insurance programs differ as to theimpact on utilization and spending, and if the impact vary across income

groups The study used data from the Jordan Healthcare Utilization andExpenditure Survey (JHUES) collected in 2000 The survey contained asample of 8,800 households and obtained information on each individualwithin the household, the head of the household, and detailed health careutilization and expenditure information on a randomly selected householdmember The study applied econometric techniques to a set of specifiedmodels along the two-part model approach to the demand for health care Hisfindings show that about 60 percent of the population is covered by sometypes of insurance However the distribution varies across income groups andresults of estimation showed that the effect of insurance on the outcomeindicators differ substantially across the various programs He gave out theconclusion that insurance is found to increase the intensity of utilization andreduce out of pocket spending However, no general insurance effect on theprobability of use is found (Ekman, 2007).

Sapelli and Vial (2001) studied the Self Selection and Moral Hazard inChilean Health Insurance The authors used the data from the NationalSocioeconomic Survey (1996) in Chile, According to the data, 60 percent ofthe total population selected the public insurance National Health Fund(FONASA), 25 percent of population selected private the health insuranceinstitution (ISAPRE) beneficiaries, and 11 percent have no insurance Theremaining individuals are distributed among the Armed Forces and lawenforcement at 3 percent, and other groups that have special coverageschemes The model is based on the individual demand for health insuranceand health care services, taking the interdependence between these decisionsinto account, within a framework of intertemporal utility maximization underuncertainty On the theoretical model, it gave the relationship between theindividual’s health insurance decisions and the consumption of health care,the presence of self-selection bias must also be detected as the empiricalanalysis is carried out The empirical model explains the quantity of health

care services consumed The paper showed that self selection with observablerisk variables, adverse selection against public, the probability of purchasinghealth insurance is greater for families with higher income, young children,larger household size, and more education; and when the household head isolder, female, and contributes to a Pension Saving Account A higherincome, younger age, smaller number of dependents, residence in an urbanarea, higher educational level, and employment with a larger companyincrease the probability of choosing private insurance The finding that olderage and more dependents positively affects affiliation to the public Aspect ofmoral hazard, the expected value of moral hazard on physician visits forindependent workers who purchase insurance is positive on average Theestimated value indicates that, on average, insured workers consume morethan twice the quantity consumed by non insured workers If we considerprivate and public insurance beneficiaries separately, we find that moralhazard is larger in the case of public insurance This result is consistent withthe fact that independent workers who purchase public insurance have access

to almost complete coverage in physician visits, but in the private insurancesector copayments are usually different from zero

2 Studies of utilization on healthcare in Vietnam.

In Vietnam, there have been some but increasing studies on the effect ofhealth insurance on health care utilization Jowett (2004) studied the effect ofbeing insured under the voluntary health insurance of Vietnam on pattern oftreatment seeking behavior, as the author used the data surveyed by theInstitute of Sociology in Hanoi which was limited itself to only 3 provinces

of Hai Phong, Ninh Binh and Dong Thap A two-stage multinomial logitmodel is used to tackle the problem of endogeneity, the results of the paperhave given out in both types of provider sought and type of cared in highlysignificant empirical results Because of the collected limitation data in 2004,

he only examined that the relation between the VHI and treatment seeking

behavior, health insurance status in his study is considered as an endogenousand exogenous variable The paper finds that health insurance was crowdedout by the compulsory scheme Thanks to richer data sets, Trivedi (2002)examines the patterns of health care utilization in Vietnam by analysing thedata sets of 1997-1998 VLSS This paper compares utilization patternsamong the insured and the uninsured individuals and finds that insuredindividuals, when ill, tend to seek treatment at commune health centers andgovernmental hospitals rather than private health facilities and pharmacies(Trivedi, 2002) In addition, the paper finds that the average difference ingovernment hospital utilization between the insured and uninsured is alsostatically significant, that for the insured population being higher by a factor

of about 2.5 times, the average difference in the use of private health facilityand drug vendors is significantly higher for the uninsured sample than for theinsured (Trivedi, 2002) However, due to the structure of the data, the paperfails to separate those insured under the voluntary component of the scheme,from those enrolled compulsorily Therefore, insurance status is considered

as exogenous This can lead to biased estimation of the impact healthinsurance on health care utilization if households with expected higher healthexpenditure choose to enroll

Similar to Tridevi (2002) and Jowett (2004), Giang (2006) studied on thehealth insurance and pattern of health care utilization in the case of Vietnam

By using the Vietnam National Health Survey data (VNHS 2001/2002), theinfluence of health insurance under three different types is empiricallyassessed The paper showsthat the public providers, the insured have higherprobability of use and higher frequency of use of inpatient care than theuninsured The result also suggeststhat voluntary health insurance has noeffect on the frequency of outpatient visit to public providers (Giang 2002)

As for private providers, the findings gives out that the health insurancediverts the use of health care services from private health providers to public

providers where the insurance benefits can be accessed Therefore the paperconcludes that both compulsory insurance and health insurance for the poorwere found to have a negative influence on the probability and frequency ofhaving outpatient visits, however, voluntary health insurance has noinfluence on probability and frequency of visits to private providers.However, due to data limitation, Giang could analyze the impact of healthinsurance for the poor or could not address the issue of possible adverseselection and assess the factors influencing to the provider moral hazard.Wagstaff (2007) used the VHLSS data in 2004 to analyze the initialimpact of the Health Care Fund for the Poor (HCFP) The data used for theestimation covers 9,000 households and over 40,000 individuals and theimpact of HCFP is estimated by comparing out–of-pocket payment andutilization between those covered by HCFP and comparable individuals notcovered The paper used the method of propensity score matching whichmeasures the closeness of “treated” and “untreated” individual Wagstaff(2007) finds that there are both bad and good news in the program, the goodnews is that program appears to be increasing the utilization of services quiteconsiderably and reducing the risk of catastrophic out of pocket spending.The bad news is three-fold Firstly, there is no perceptible impact on(average) out-of-pocket spending, and even with HCFP coverage poorhouseholds are left spending a high share of their modest income on out-of-pocket health expenses and at considerable risk of catastrophic spending.Secondly, the utilization impact is far more pronounced for inpatient carethan outpatient care This may not necessarily be the most cost-effective way

of improving the health of poor Vietnamese households, and may leave themfacing including transport costs, informal payments, etc that are higher thannecessary Thirdly, the impacts on utilization are larger among the better off:among the poorest deciles, utilization impacts are rarely significant(Wagstaff, 2007)

CHAPTER III HEALTH CARE SYSTEM AND HEALTH UTILIZATION IN VIETNAM

The Chapter will give out information on health care system in Vietnam, healthcare utilization for self-medicine, inpatient and outpatient services Beside, it is also resumed some different insurance schemes and health care

for the poor under Vietnam health insurance.

1 Introduction of healthcare system in Vietnam

1.1 Public health care system

In Vietnam, the health care system is organized by administrative level,the Ministry of Heath (MOH) covers overall responsibility at the centrallevel The MOH also directly controls 70 subordinate institutions in threemajor areas, including: hospitals, preventive medicine and professionalinstitutes and pharmaceutical universities/schools (USAID 2009)

At the provincial level, every province has the Provincial HealthDepartment (PHD), a professional agency under the management of theProvincial People’s Committee (PPC), works to advise the PPC on statemanagement of local people’s health care, protection and promotion, andperforms tasks and obligations as authorized by the PPC and standingregulations According to USAID, the PHD not only works under the control

of the PPC in terms of direction, organizational management, payroll andoperations, but is also under the control of the MOH in terms of technicaldirection, guidance, monitoring and inspection Under this PHD, there are anumber of provincial general hospitals and specialty hospitals, and provincialpreventive health centers

At the district level, each district has the District Health Office (DHO), aprofessional agency under the management of the District People’s

Committee (DPC), works to advise the DPC on state management of localpeople’s health care, protection and promotion, and performs designatedtasks and obligations as authorized by the DPC and the PHD Each districtalso has a district hospital, and a district preventive center, and someintercommunal polyclinics, majorities of these hospitals admit inpatients andprovide emergency care and basic treatment for common diseases (MOH2000).

At the commune level, the Commune Health Station (CHS) is the firstformal point of health care contact in the government health care system, isdesignated to provide primary health care services They carry out earlydetection of epidemics, provide care and treatment for common diseases anddeliveries, mobilize people to use birth control, practice preventive hygiene,and carry out health promotion at the village level The CHS has aresponsibility to the DHO and the Commune People’s Committee for localpeople’s health care, protection and promotion, and receives technicalguidance from the district hospitals (USAID, 2009) It also was added byUSAID, the primary care facilities network is considered a grassroots levelnetwork and covers all districts, and communes By the end of 2006, Vietnamhad 671 districts and 10,876 communes/wards serving with health caresystem

1.2 Private health care system

The private health care system of Vietnam only officially started from

1989 when the government legalised to encourage multi-entities in the healthcare services Basically, by the end of 2006, there were 30,000 privateclinics, 5 semi-public hospitals, 300 private regional general clinics and 87maternity wards throughout the country, this number is nearly increased at 20percent in comparision with 1996 while it was 25.698 private practitioners Inwhole country, there are 49 private hospitals including 36 general hospitalsand 13 specialized hospitals with 4,050 sick beds Although, the contribution

of the private health sector to the provision of health services, especially forinpatient treatment, is very limited In 2003, private health clinics provided

up to 60 percent of outpatient services, but only 4 percent of inpatientservices and about 10 percent of preventive health care services Publichealth facilities provided 40 percent of outpatient services, 96 percent ofinpatient services and 90 percent of preventive health care services (MOH2003)

Currently, the private health sector has two tiers: the first tier consists of asmall but growing number of private hospitals These hospitals are mainlylocated in big cities The second is private clinics which are located both inurban areas and operate under full time or part time physicians (Giang 2006).However, there is a serious imbalance in the distribution of privatepractitioners, with a large concentration in urban areas with higher livingstandards, there are also a large number of private practitioners withoutlicenses, and up to 70 percent of the total number of private clinics are run bydoctors who are also running the public health clinics According to USAID,there are only 26 percent of private clinics participate in primary health careactivities when mobilized Most private consulting rooms violate regulationsthat ban practitioners from selling drugs on their clinic premises Aphenomenon often seen at private clinics is the overuse or inappropriate use

of drugs and advanced technologies, probably for financial benefit of thepractitioners

In despite of some above issues, the private health providers havesignificantly contributed to health services such as more flexible openinghours, more ready access, greater drug supplies and more respectful treatment

of clients Moreover, total cost for treatment in private health services issometimes even less than in public health services due to the prevalence ofinformal “envelope” payment in Vietnam (Giang 2006)

2 Health care utilization

The following section describes the healthcare utilization in Vietnamfrom previous studies, in term of types of services sought, i.e., inpatient andoutpatient, no-treatment and self-medication It also describes the use ofhealthcare services by types of the schemes

2.1 No treatment and self-medication

According to the VNHS 2002, among 66,795 cases who reported illness

in the 4 week reference period, only 4% of people did not receive anytreatment Self-medication is the most popular choice of treatment inVietnam, accounting for 75% of those reported illness in the 4 weekreference period Both VLSS 1998 and VNHS 2002 show consistentevidence of high rate of self-medication among all age groups and levels ofeducation But there is a big difference between ethnic groups: Kinh/Chinesehave the largest rate of self-medication: (73-74%) and the minorities inCentral region and Central high land have the lowest rare of 49%

Comparing with other previous survey, VLSS (1993 and 1998), there hasbeen an increasing trend in rate of self-medication overtime Using data ofVLSS 1998 Nguyen Thi Hong Ha (2002) found that the rate of self-medication is about 70% compared to 75% in 2001-2002 (VNHS, 2002).While in 1993, average annual services contacts to drugs vendors per capitawas 2.1, the number increased to 6.8 contacts by 1998 Contacts to drugvendors account for two third of the total health contacts (World Bank,2001)

There are many reasons for self-medication, including economicdifficulty, difficult access to health facilities, the use of old prescription, orminor illness According to VNHS, only 1% of the sample population reports

to self-medication owing to difficulty in accessing health facilities Theeconomic difficulty cited as the main reason for self-medication by only 1%

of the richest quintile compared to 17% for the poorest quintile Gertler and

Litvack (1998) conclude that the growth of self-medication is also aconsequence of the low quality of care available for the lower income groups.Further, Trivedi (2002) states that freer availability of drugs and thederegulation of over-the-counter sale of pharmaceuticals with activeingredients have enhanced the practice of self-medication There is a greatconcern about the situation of antibiotic resistant bacteria, caused by the self-medication practice.

2.2 Inpatient services

Firgue 1: Distribution of inpaatient contacts across healthfacilities by rural and urban population

Firgue 1: Distribution of inpatient contacts across

health facilities by rural and urban population

0 10 20 30 40 50 60 CHS/ Regional Polyclinic

District Hospital Provincial Hospital Central Hospital Other public health facility

Source: Vietnam National Health Survey 2002 ( Giang 2006)

Inpatient services are mainly provided by government hospitals.According to VNHS survey, the proportion of people seeking inpatientservices in period of 12 months prior to the survey is 5.4% Figure 1 showsvisits to health facilities for inpatient services in period of 12 months prior toVNHS 2002 survey As the figure shows, the number of patients seeking forinpatient care in private facilities is very low, accounting for 4% However,

although the provision of inpatient care by private clinics/hospitals remainlow, it has been growing rapidly in recent years Data from VLSS 1993 and

1998 show that utilization of inpatient services in private providers wereneglectible (World Bank et al 2001)

Provincial and district hospitals have the highest proportion of patientsseeking inpatient care The regional policlinics and commune health centeraccount for about 4% of inpatient contacts However, it varies across ruraland urban areas The share of inpatient services in central hospitals isrelatively low This may be due to limited accessibility and smaller numbers

of central hospitals as comparing with other health facilities

2.3 Outpatient services utilization

Outpatient care utilization varies across groups Children under tend touse more outpatient services in private health facilities (Nguyen Thi Hong

Ha, 2002) Hospitals with better quality are widely used by the better offhouseholds Healthcare utilization in Vietnam is very unequal and skewedtoward the better-off The affluent and urban people benefit more from thehospital services which receive a disproportionately high share of staterecurrent budget (76% in the period 1996-1998 (Sepehri 2003)) than theother facilities Private hospitals are also located in cities and centers,contributing to even more inequality Moreover, physician and doctorgraduates prefer to practice in cities and urban centers which is turncontributes to the lack of qualified physicians and doctors in rural and remotemountainous are with higher healthcare needs Although legislation of userfee policy in healthcare in 1989 has provided health facilities with anincreasing source of revenue and helped hospitals to improved quality of,users fee and other informal fees have been a very heavy burden for the poor

As a consequence, the poor who are usually the sicker receive less healthcare(Seperhi, 2003) In 2003 the Government established the Healthcare Fund for

the Poor to improve access to healthcare services by the poor and reduce thefinancial burden of illness on the poor.

2.4 The schemes of health insurance in Vietnam.

There are 3 schemes under Vietnam Health Insurance: Compulsory schemes,Voluntary schemes and Schemes for the poor under Healthcare Fund for thePoor

Following the decree No.299 in August of 1992 the Government ofVietnam introduced a non-profit public health insurance in 1993 The schemewas initially compulsory, covering current and retired civil servants and theemployees of state enterprises as well as those with contracts of 3 months orlonger in large private enterprises which have more than 10 employees,employees of foreign owned enterprises and organizations, the disabled,people of merit such as mothers, widows and orphans of veterans, armyinvalids, dependents of military personnel (since 2002) and the elderly aged

90 and over (Knowles et al, 2004) The compulsory scheme does not coverfamily members The Voluntary scheme is in principle open to allVietnamese Many provinces have been reluctant to encourage voluntaryinsurance scheme, partly due to high cost resulting from adverse selection(Knowles et al, 2004) The two main voluntary health insurance initiativeshave been health insurance for school children and Farmer voluntary healthinsurance scheme under which farmers contribute 30% of premium andprovincial government contribute 70%

There is also an insurance scheme for the poor (HIP) covering residents

of communes with very difficult socio-economic circumstances, and ethnicminorities in disadvantaged provinces The scheme is funded by a variety ofcharity and donor organizations as well as by the national Governmentthrough the poverty alleviation program More recently, the scheme has beenreorganized and funded more adequately under a national program according

to which provinces and centrally-run cities are instructed to establish

Healthcare Funds for the Poor Each beneficiaries of the program of healthinsurance for the poor is either provided with free health insurance card withpremium of 50,000 dong or their medical expenses at public hospital are paid

by the program, up to some ceilings No co-payment is required (Sepehri et

al, 2005)

2.5 Heath care for the poor.

The Decision 139 mandated all provincial governments to provide freehealth care to three groups:

Households defined as poor according to official government povertystandards introduced in November 2000; all households regardless of theirown assessed income living in communes covered by a program set up as aresult of another policy known as Decision 135 dating from 1998, whichprovides support and services to especially disadvantaged communes; andethnic minorities living in the province of Thai Nguyen and the sixmountainous provinces designated by Decision 186 as facing specialdifficulties.7 Central government committed VND 52,500 (around $US 3)per beneficiary per year for the province’s health care fund for the poor(HCFP), and required provinces to add a further VND 17,500 though inpractice few have done so to date

Initially provinces were free to decide whether to use the VND 70,000 toenroll HCFP beneficiaries in the government’s SHI program, or to managethe risk themselves and provide direct reimbursement to providers In thislatter option that initially proved more popular with provinces but which isbeing phased out through a 2005 government directive updating Decision

139, HCFP beneficiaries are issued with a free health-care certificate or card.Confusingly, perhaps, some provincial governments that have opted for thedirect reimbursement modality commission the agency that runs the SHIscheme – known as Vietnam social security (VSS) – to issue cards to HCFPbeneficiaries, although VSS does not manage the risk in this modality, and it

is the provincial government’s health fund for the poor not VSS that directlyreimburses providers The VSS card is simply used to provide beneficiarieswith a ‘passport’ to free care.

Whichever modality the province opts for, it is expected to provide thesame benefits to HCFP beneficiaries as enjoyed be those compulsorilyenrolled in the SHI program HCFP beneficiaries are not supposed to payeither deposits or copayments at health facilities The package is, however,focused largely on services delivered by public hospitals and communehealth centers (the coverage at the latter is mostly for drugs on the essentialdrugs list) The scheme does not cover nonprescription drugs bought fromdrug vendors and pharmacies, who are extensely used in Vietnam, and whooften sell drugs that are supposed to be available only with a prescription.The package also excludes services delivered by other private providers,though recently – prompted by another 2005 government directive – VSS hasbegun contracting with private providers, and HCFP beneficiaries will havethe same entitlements vis-a` -vis private providers as other people enrolled inthe SHI program Unsurprisingly, but importantly because of theirpervasiveness in Vietnam, the package excludes informal payments made bypatients to providers In practice, there is a strong bias in reimbursementstoward higher-level facilities and inpatient care: in 2004, 40% of the HCFPbudget went to provincial hospitals, 34% to district hospitals and 20% tocommune health facilities; and nearly 60% of the budget in both 2004 and

2005 was spent covering the costs of inpatient care These figures maskvariations across provinces, which may reflect 6 This section draws heavily

on the excellent reviews of Knowles et al (2005) and Capuno et al (2006) 7These included Cao Bang, Bac Kan, Lao Cai, Ha Giang, Son La and LaiChau (Wagstaff 20…)

CHAPTER IV METHODOLOGY AND EMPIRICAL RESULT

Chapter II and III have given the literature review and resumed some schemes of health insurance in Vietnam as well as knowledge the health care utilization of Vietnam This chapter discusses the econometric models used for examining impacts of different schemes of the health insurance on the use and the frequency of health care services by the type of provider Then after, the empirical results are presented and discussed.

1 Methodology and data

1.1 Methodology

As stated, we have two broad types of dependent variables, (1) whether

an individual sought care or not l (2) the type of health facility used; and (3)frequencies of use This separation will provide detail impacts of differenttypes of health insurances on seeking services provided by different types ofhealth facilities For inpatient admission, only governmentpolyclinics/hospitals are investigated since inpatinet care is primarilyprovided by government health facilities The models are estimated for thewhole sample as well as the urban and rural areas separately to capturedifferences in health treatment seeking behavior of these areas

1.1.1 Logit model

The dependent variable is a dummy variable, taking value of 1 if theindividual i sought care ( inpatient/outpatient) in the 12 month prior tointerview and value of 0 otherwise The logit model has the form (Gujarati,2004):

(1)

Or (2)

Thus, (3)

Dividing (2) to (3), we have: (4)

is called the odd ratio in favor of using healthcare services –the ratio of the probability that the individual is using healthcare services tothe probability of not using healthcare services in 12 month prior tointerview The right hand side of equation (1) represents the cumulativelogistic distribution function

Normally, one often uses - the exponential of slope, in interpreting

of the logit model The exponential of slope measures the change in the oddratio for a unit change in X, that is it tells how much the odd in favor of usinghealthcare services changes as X changes by a unit In case if X is a dummyvariable, for example, health insurance status, taking values of 1 and O, then

measures how much the odds in favor of using healthcare serviceschanges if X takes value of 1 if having health insurance and 0 if not

1.1.2 Count regression models

A/Poisson model:

For the second type of dependent variable (the number ofadmissions/visits) that takes non negative integer values, count regressionmodels which accounts for discrete nature of the dependent variable areappropriate The Poisson model has been widely used for this kind ofdependent variable Following Green (2003)

The model specifies that discrete dependent Y is drawn from a Poissondistribution with a parameter that is related to Under the Poissonassumption, the probability that Y counts is observed is given by:

for y=0,1,2,3… (5)

is a function of

Normally, there are 3 assumptions under Poisson regression:

(i) Parameter is specified as the log linear function of x and that

(ii) Conditional variance of Yi is equal to conditional mean:

(the equi-dispersion condition)

(iii)/ Yi, xi are independently and identically distributed (maximumlikelihood method)

In practice, the second assumption is rarely held and it is a majorshortcoming of the Poisson model The conditional variance is usually largerthan the conditional mean, which is called over dispersion When overdispersion exists, the estimations of Poisson regression model are stillconsistent but not efficient (Long, 1997 cited in Giang(2006)) The mostcommon alternative model for this case is the negative binomial model B/ Negative binomial model:

The negative binomial regression model imposes a less restrictive assumption

on the equi-dispersion Negative binomial distribution is a compound Poissonwith probability distribution as follows:

(6)Where: is the gamma function and , the parameter measuresdegree of dispersion

In the Poisson regression model, the conditional mean and thevariation in is due to the observed heterogeneity, the conditional mean innegative binomial model includes another random variable: and thevariation model in is due to the unobserved heterogeneity(Long, 1997 cited

in Giang (2006)) The expected value of Y in negative binomial distribution

is as in Poisson distribution, that is However, theconditional variance of negative binomial distribution is larger and exceedsthe conditional mean:

(7) C/ Problem of truncation and models for truncated counts:

As other studies in the filed may face, there are an ‘excessive’ zero in oursample It is not a problem with the binary model However, this is problem

of truncation in count models The normal count regression such as Poisson

or negative binomial models will lead to inconsistent estimation Hence, amodification is needed to correct for this problem These are truncated countmodels, including truncated Poisson and truncated negative binomial Underthese models only people who reported at least one time of outpatient contact

or inpatient admission are included

1.2 Model selection

Truncated Poisson model and truncated negative binomial model arealternatively most appropriate models for studying frequencies of healthutilization as aforementioned Thus, test of overdispersion is the criticalcriterion for selecting consistent and efficient model for empirical estimation.Testing for over dispersion: Because the Poisson regression model is aspecific case of the negative binomial regression model, thus the test is to testHo: This test could be done with likelihood ratio test and the test isavailable when we estimate the negative binomial model under normalsetting, i.e no weight and standard error adjustments included

1.3 Data

Data for this analysis is taken from VHLSS 2006 This is a objective survey that covered different aspects of households includingdemographics, education, income and expenditure, assets…Relating to healthtopic, the survey provide rich information of individuals that were muchextensive to the VHLSS 2002 or VHLSS 2004 The health related section issection 3 The first sub-section- section 3A, health status over 4 weeks and 12months before the survey and health insurance status- involvement in anyhealth insurance schemes- as well as health treatment utilization in generalwere collected The two next sub-sections ask for disability and chronic

multi-diseases Reproduction health and behaviors which potentially affected thehealth were then interviewed Health insurance history and the utilization ofhealth insurance is the next sub-section The section ends with detail healthtreatment utilization There were two times span used for collectinginformation- 4 weeks and 12 months Outpatient contacts to different healthfacilities were surveyed for 4 weeks as well as 12 months Meanwhile,Inpatients contacts were surveyed for 12 months only The survey samplecovered 64 provinces including urban and rural areas for a total of 9190households with 39071 individuals from 3,063 communes/wards Asmentioned above, the current thesis will use 12-month-reference period forboth outpatient contacts and inpatient admissions The study of Giang (2006)only studied outpatient contacts for 4-week reference period as theinformation was not available for 12 months

2 Variable description

As presented above, the thesis have two classes of models- the logit tomodel probability of contact/admission and the count models to investigatethe frequencies of contacts/admission Under the logit model, the dependentvariable is a binary variable that takes 1 if the individual has had sought care

to the health facilities in general or our interested types of service providersand 0 otherwise The dependent variable in the count model number ofcontact/admissions The two classes of models share the same set ofindependent variables that are used commonly is the literature on utilization

of health services The individual-levl variables include age, age-squared,gender, education, income (proxy by expenditure), marital status and healthstatus of the individual The household-level variables includeeducation andethnicity of household head, household income, Education of householdheads could affect health treatment of children as they are decision makersnot the children themselves In addition, living areas (urban or rural) andregions are also included as they are quite conventional in studyingindividual or household’s behaviors in Viet Nam especially using LivingStandard Surveys

Table 1: Health insurance coverage by expenditure quintiles

Unit: percent

Thepoorest

The nearpoor

Themiddle

The nearrich

Therichest

Health insurance / Free health card for the

Table 2: The use of outpatient care across providers by the insurance status and the type of insurance

Communeclinics

Publicpolyclinics/district hospital

100.00Health insurance /Free

health

card for the poor

100.00

100.00

100.00

Communeclinics