【作者】 侯志远；

【导师】 孟庆跃； Eddy van Doorslaer；

【作者基本信息】 山东大学 ， 社会医学与卫生事业管理， 2012， 博士

【摘要】 研究背景新型农村合作医疗制度(以下简称新农合)自2003年建立以来,保障水平不断提高。到2011年底,各级财政对参合农民每人每年补助标准达到200元,实际住院补偿比达到43.7%。与此同时,农村居民的卫生服务利用增长明显,住院率从2003的3.4%增加到2011年的8.4%；居民疾病经济负担也不断加重,卫生支出占家庭总支出的比例从2003年的12.1%增长至2011年的13.3%。新农合制度在增加农村居民卫生服务利用的同时是否也加重了居民疾病经济负担?保障水平不断提高的新农合制度是否改善了农村居民的福利水平?这是新农合能否实现预期目标的关键。现有新农合研究多数是评价参合与否对医疗服务利用和经济风险保护的影响,较少研究关注不同新农合补偿水平对医疗服务产生的影响,很少研究从家庭一般消费行为角度评价新农合效果,更没有研究结合医疗服务消费和一般消费行为综合评价新农合对居民福利的影响。然而新农合政策效果大小与其补偿水平密切相关,并且作为一种医疗保险制度也可能影响居民一般消费行为。医疗服务价格会影响居民卫生服务利用决策,这两方面又会间接影响居民一般消费行为,所有这些方面都与居民福利水平有关。现有文献多是孤立地研究新农合对某一方面的影响,从而无法综合评价其对居民福利水平的影响。因此,评价新农合不同补偿水平与居民福利之问的关系,保障能力不断增强的新农合对居民福利水平会产生什么影响,是目前亟需解决的问题。本研究从居民整体消费行为入手,构建社会医疗保险的福利效应模型,具有重要理论意义；从居民医疗服务消费和非医疗消费两个途径来系统评价新农合补偿水平对居民福利的总体影响,以便从居民福利最大化角度完善新农合制度设计,具有重要现实意义。研究目的本研究的总目标是通过理论研究和实证研究,探讨新农合补偿水平对农村居民福利的影响,提出通过新农合制度设计实现居民福利效应最大化的相关政策建议。具体研究目的包括：从理论上构建新农合补偿水平的福利效应模型；通过实证研究评价新农合补偿水平对居民医疗服务消费行为和非医疗消费行为的影响,并揭示新农合影响居民总体福利水平的机制；分析新农合补偿水平对居民福利影响的公平性；探讨居民新农合政策知晓程度在新农合福利效应传导过程中的作用。研究设计和研究对象本研究关键政策变量是新农合补偿水平,不受居民特征影响,可视为自然实验。家庭调查采取混合横断面研究设计,资料来自欧洲联盟资助项目《中国和越南建立公平性和可持续性农村健康保障制度研究》2006年基线调查和2008年终末调查。采用多阶段分层随机抽样方法,根据社会经济发展水平和地理位置,分别在山东省和宁夏回族自治区抽取3个样本县,每个样本县再根据经济发展水平随机抽取3个乡镇,共抽取18个乡镇,每个乡镇按照经济发展水平抽取6个村,共抽取108个村。最后根据每个村的户主花名册进行等距抽样,以保证样本的随机性,基线每村抽取60户左右,终末每村抽取30户。每户的所有家庭成员均为调查对象；如被调查对象调查时因故不在家,则请其家人代答。基线调查和终末调查抽取的乡镇和村庄一致,但家庭不一致。研究对象是16岁以上参加新农合的农村居民,共26310人,其中2006年16294人、2008年10016人。研究方法本文以效用函数为基础,通过消费者当期和跨期消费决策整合医疗保险的福利效应理论和预防性储蓄理论,从医疗消费和非医疗消费两条途径构建新农合福利效应模型,通过严谨的推导过程将新农合福利效应分解为卫生服务利用效应、价格效应和一般消费效应,提出研究假设。实证研究中,基于计量经济学中的工具变量方法和流行病学中的标准化方法,构建控制医疗服务消费行为影响的外生性补偿水平综合测量指标,并以新农合补偿水平为关键变量,结合面板数据特征,构建村级固定效应实证模型评价新农合的福利效应。根据分层抽样方法和关键变量是县级层面的新农合补偿水平这两个特征,在计量模型中设定了县级聚类,来消除县级层面某些特征的序列相关。具体计量模型依据变量类型而设定。结合本研究段阶段分层随机抽样方法,二分类因变量的卫生服务利用效应选取条件logit模型；对于价格效应和消费效应,结合价格和消费支出的偏态分布特征选取广义线性模型(GLM)。在以价格和消费支出的绝对值为因变量的模型中,通过Box-Cox和Modified Park检验来确定连接函数和分布。对于家庭卫生支出占总支出比例等率比指标,由于数据分布在0-1之间并且存在大量的0值,计量模型设定为logit连接函数和二项式分布的GLM模型。同时,实证模型中也加入了新农合补偿水平与家庭收入、新农合补偿政策知晓程度的交互作用。研究结果(1)新农合补偿水平的卫生服务利用效应：在控制了其他因素后,新农合住院补偿水平提高1个百分点,居民利用住院服务的概率将增加0.32个百分点。处在20%最低收入组的居民住院服务利用概率要显著低于高收入组3.17个百分点,但是提高住院补偿水平对住院服务利用的积极作用在低收入组中更大,补偿水平提高1个百分点,其改善低收入组住院概率的效应要显著高于高收入组0.11个百分点。新农合住院补偿水平增加住院服务的作用与居民对新农合政策的知晓程度无关。与家庭账户相比,门诊统筹模式显著增加门诊概率,降低住院概率。另外,参加商业医疗保险也显著提高居民住院服务利用概率约3.74个百分点,延长住院1.30天,对门诊服务利用没有显著影响。(2)新农合补偿水平的卫生服务价格效应：在其他条件不变的情况下,新农合住院补偿水平提高1%,次均住院总费用和自付费用分别增加0.72%和0.34%,导致住院费用自付比例下降0.37%。次均住院费用的增加可以分解为：住院补偿水平提高1%,住院天数增加0.46%,日均住院费用增加0.70%。从住院服务价格效应的公平性来讲,低收入组和高收入组之间居民的住院总费用、自付费用、自付比例没有显著差异,新农合补偿水平推高住院服务价格的效应也与收入分组无关。尽管新农合补偿水平知晓者花费的住院总费用和自付费用分别高于不知晓者460元和310元(相当于住院费用均值的10%),但是新农合政策知晓程度会弱化新农合补偿水平推高住院服务价格的效应。与家庭账户相比,门诊统筹增加次均门诊总费用8元,减少次均门诊自付费用22元,进而显著降低门诊费用自付比例6.24个百分点。(3)新农合补偿水平对医疗消费的总体影响：用医疗消费支出综合评价新农合补偿水平的利用效应和价格效应发现,住院补偿水平提高1%,家庭用于医疗服务的支出比例将增加0.73个百分点,因病借债的概率增加0.23个百分点。(4)新农合补偿水平的消费效应：控制居民特征变量后,新农合住院补偿水平提高1%，人均非医疗消费性支出增加0.24%,人均储蓄额减少27元；人均文化教育支出增加12元,在1%显著性水平上有统计学意义。住院补偿水平提高1%,低收入家庭人均消费性支出比高收入家庭显著多增加0.67%,人均文教支出显著多增加10元,人均储蓄显著多减少68元。家庭消费或储蓄在新农合知晓程度分组中没有显著差异,但是住院补偿水平提高1%,知晓家庭人均非医疗消费性支出比不知晓家庭显著多增加0.11%,人均文教支出显著多增加4元,人均储蓄显著多减少30元。结论与政策含义新农合作为覆盖我国农村居民的社会医疗保险,影响居民医疗服务消费行为,是保险制度的客观属性；作为农村居民应对大额医疗支出风险的保障措施,通过降低家庭未来不确定性预期而影响居民当前消费水平,体现为农村居民的主观感受。从医疗服务消费行为来看,提高新农合补偿水平增加卫生服务利用、提高卫生服务价格,进而增加医疗支出,损害居民在医疗消费方面的福利；从非医疗消费行为来看,更高的补偿水平增加非医疗服务消费,改善医疗服务之外的总体福利。总体来看,新农合是否导致居民福利改善,价格效应是关键,其他两个效应在一定程度上依赖于价格效应。从新农合福利效应的公平性来看,提高补偿水平可以更大程度地增加低收入居民卫生服务利用,提高消费水平,进而改善农村居民福利状况的公平性。新农合政策知晓程度可以弱化补偿水平推高价格的效应,显著改善补偿水平刺激消费的效应,是新农合福利效应传导的关键。本研究具有以下政策含义：(1)在评价新农合效果时应综合考虑其对居民医疗服务消费和非医疗消费的影响；(2)新农合适宜补偿水平的确定应以居民福利效应最大化为原则,使新农合的利用效应、价格效应和消费效应达到均衡,在均衡水平上,既能增加卫生服务利用,提高消费水平,又要控制卫生服务价格在合理范围内；(3)价格效应是新农合影响居民福利水平的关键。在提高新农合补偿水平的同时,应采取措施抑制其所导致的卫生服务价格上涨；(4)提高新农合补偿水平,进而改善低收入居民医疗消费和非医疗消费水平；(5)加大政策宣传力度,使农村居民了解新农合补偿政策并建立信任,形成未来良好预期,发挥新农合更大的福利效应。创新与不足本研究的创新之处：(1)本研究首次融合医疗保险的福利效应理论和预防性储蓄理论,构建了社会医疗保险的福利效应模型,进而将医疗保险福利效应分解为利用效应、价格效应和消费效应；(2)首次将我国农村居民的医疗消费和非医疗消费行为相结合,综合评价新农合补偿水平对居民福利的影响；(3)系统分析了居民新农合政策知晓程度在新农合福利效应传导过程中的作用,为更好地发挥新农合福利效应提供有针对性的证据。本研究的不足之处与研究展望：(1)本研究是新农合补偿水平对居民福利影响的评价研究,进一步的研究应以此为基础,基于居民福利效应最大化确定不同约束条件下新农合的适宜补偿水平；(2)本文研究数据来源于2006年和2008年,这一时期新农合发展较快,便于评价不同新农合政策特征对居民福利的影响。但是2009年深化医改后某些制度环境发生变化,进一步研究需要考虑医改后制度环境改变对新农合福利效应的交互影响,特别是价格效应。更多还原

【Abstract】 BackgroundSocial health insurance schemes are being developed in many low-and middle income countries with the aim of contributing towards the achievement of universal coverage. One of the largest of such initiatives is the implementation, since2003, of a health insurance scheme for the rural population, namely the New Rural Cooperative Medical Scheme (NCMS) by the Chinese government. As a voluntary and heavily subsidized scheme, the NCMS has seen an extremely rapid growth in population coverage in comparison to most new schemes in developing countries, covering3%of the rural population in2003to97.4%in2011. In2011, the government subsidy per enrollee reached200RMB, and the effective reimbursement rate for inpatient care was43.7%.During this period from its inception in2003to2011, health care utilization and economic burden due to diseases increased significantly in rural areas. According to the2003,2008, and2011National Health Services Survey (NHSS), the utilization rates of outpatient services were13.9%,15.2%and15.3%respectively, and hospitalization admission rates were3.4%,6.8%and8.4%. There were13.6%,15.1%and13.8%of households who had catastrophic health expenses in2003,2008, and2011, and the proportions of health expenditure as a share of total household expenditure were12.1%,12.6%and13.3%.The literature to date has mainly used the uneven rollout of NCMS across counties as a way of identifying its effects on health services utilization and financial protection, and less focus on the effect on household consumption. Few studies so far have exploited the heterogeneity in scheme features to identify its effect. However, these effects heavily depend on the NCMS generosity, which also influences consumption behavior of the rural population. The health services utilizationis largely affected by its prices, both of which will indirectly affect the consumption of the rural residents. All these aspects represent the level of residents’ welfare. The existing literatures only evaluate the effect of the NCMS on one of these aspects, which can not be used to estimate its impact on the general welfare of the rural population. Therefore, to evaluate the overall impact of NCMS on the welfare of the residents, and to identify the optimal coinsurance rate are the key problems that need to be addressed in the NCMS research.To formalize a theoretical model on the welfare effects of a social health insurance based on consumer behavior, will be of great value in theory. Evaluating the overall impact of the NCMS on the welfare of the rural population will be helpful to confirm the optimal coinsurance rate according to consumer utility maximization, both of which will make great contributions to NCMS and universal coverage in China.ObjectivesThe overall objective of this study is to conduct both theoretical and empirical study on the impact of the NCMS coverage to the consumer welfare, offering policy recommendations by changing the NCMS design to maximize the residents’ welfare. Specifically, our research objectives include:to formalize a model about the welfare effects of NCMS coverage in theory; to conduct empirical research to evaluate the overall impact of the NCMS coverage on health services utilization, prices and household consumption of the rural residents, and reveal how the NCMS affects the consumer welfare in rural China; to analyze how the welfare effects of NCMS differ by household income level; to explore the role of the residents’knowledge on the NCMS policy in the welfare effects.DataThe data derive from the project "Bringing health care to the vulnerable:Developing equitable and sustainable rural health insurance in China and Vietnam (RHINCAV)" that was funded by the European Commission. A repeated cross-section household survey was conducted in Shandong and Ningxia Province in2006and2008. In2006, within each county,18villages spread across3townships were selected to participate in the survey. About60households were randomly selected per village in Shandong, and about50in Ningxia, resulting in a total sample of6,137households (22,636individuals). In2008, a slightly smaller sample of households (30in both provinces) was drawn from the same villages, leading to a sample size of3,288households (13,058individuals). In our analysis, we only use adult individuals aged16or older and enrolled with NCMS, which amounts to a total sample size of26,310individuals (16,294in2006and10,016in2008). Note that the data is a panel at the village level, but not at the individual level.MethodsIn this study, the theoretical model about the comprehensive welfare effects of NCMS is constructed based on the welfare effects of health insurance and precautionary saving theory. Furthermore, it is decomposed into the utilization effect and price effect for health service consumption, and consumption effect for non-medical consumption.Our empirical strategy for identifying the welfare effect of NCMS coverage basically relies on a pooled regression analysis with coinsurance rates as our main variables of interests. Since the data is the panel at village level, a village-level fixed effects model is adopted to evaluate the welfare effects of NCMS. Outcomes related to health service utilization are modeled using the conditional logit model. Generalized Linear Models are used for the continuous outcomes, such as health service price, medical expenses and non-medical consumption expenditure.As the key variable, NCMS coverage degree depends on the copayment, reimbursement rate, and ceiling. All of them vary with the health service providers and total medical expenditure. We use instrumental variables (Ⅳ) and standardization approaches to construct the summary measures of NCMS coverage, combining all three dimensions, which is independent of the utilization behavior of the relevant populations. Besides the NCMS coinsurance rates, the household income groups, knowledge on NCMS policy, and their interaction with coinsurance rates, are also included in these empirical models.Results(1) The effect of NCMS coverage on health care utilization:after controlling for other variables, a one percentage point increase in NCMS coverage raises the probability of using inpatient care by0.32percentage points. The admission probability for residents in the20%low-income group is3.17percentage points lower than the high income group. However, with the NCMS coverage increasing by a percentage point, the increase of the hospitalization probability for low-income group was significantly0.11percentage points more than the high-income group. The residents’knowledge on NCMS policy does not influence the utilization increasing effect of NCMS. Compared with the family saving accounts, social pooling account for outpatient can significantly increase the probability of using outpatient care, but reduce the probability of hospitalization. In addition, the presence of other, mostly private health insurance cover raises the probability of using inpatient care by3.74percentage points, or the equivalent of a15percentage point’s rise in NCMS coverage. It also extends the hospital stay by1.30days, but gives no significant effect on the use of outpatient services.(2) The price effect of NCMS coverage on health services:In the case of other conditions remaining unchanged, with a1%increase of NCMS coverage, the total costs and out-of-pocket expenses per inpatient spell would show an increase of0.72%and0.34%respectively, leading to the share of OOP in total expense decreasing by0.37%. The increase of hospital cost can be decomposed as follows:inpatient reimbursement rate increased by1%, length of stay increased by0.46%, and the average daily cost of hospitalization increased by0.70%. There are no significant differences on the hospital costs, out-of-pocket expenses and its share to total cost per admission between the residents of low-income group and high-income group, and the price increasing effects are also not related to the income groups. Although patients, who are familiar about NCMS reimbursement level, spend about10%of the average hospitalization costs more than those who do not know, these price increasing effect of NCMS coverage becomes weaker to some degree when patient know about the NCMS policy. Social pooling account can increase average total outpatient costs by8RMB, reducing OOP expenses by22RMB, and thus significantly reduce the OOP share by6.24%compared with family saving accounts.(3) The overall impact of NCMS coverage on household medical expenses:Using household medical expenses as an comprehensive measure of the utilization effect and price effect of NCMS coverage, inpatient reimbursement rate increased by1%, the proportions of health expenditure as a share of total household expenditure would increase by0.73percentage points, and the probability of borrowing or selling assets due to illness would increase by0.23percentage points.(4) The consumption effect of NCMS coverage:on average, an increase of1%in NCMS coverage stimulates non-healthcare related consumption by0.24percent and decrease savings by27RMB per capita, though the results are insignificant; and increases cultural and education spending per capita by12RMB, significant at the1%significance level. The NCMS effect on non-healthcare related consumption is much stronger by0.67%among low-income families than high-income families because of a1%increase in scheme coverage, but the consumption effect is insignificant among the high-income families. And a1%increase in scheme coverage will reduce savings per capita by more than68RMB among low-income families than high-income families. Although there was no significant difference of household consumption or savings between residents who know about the NCMS policy and those who do not know, a1%increase in NCMS coverage bring more increase in non-healthcare related consumption among residents who know about the NCMS policy by0.11percent than among those who do not know, but reduce savings per capita less by30RMB.Conclusions and policy inplicationsAs a social health insurance for the rural population, NCMS affects health services consumption, which is the objective requirements of the health insurance system; as a measure to deal with the risk of large medical expenses, NCMS affects the current consumption level of rural residents by reducing the family’s future uncertainties, reflecting the subjective feelings of rural residents. More generous NCMS increase the utilization and prices of health services, thereby increasing total medical expenses, and it also increases non-healthcare related consumption. Overall, the price effects, which influence both the health care utilization effects and household consumption effects, are the key to the welfare improvement of residents by NCMS. The rising generosity of NCMS will improve the health care utilization and consumption level of low-income residents much more than the high-income group, thereby improving the equity of the rural residents’ welfare level. The knowledge on NCMS policy, weakening the price increasing effect and significantly improving the consumption stimulating effects due to a more generous scheme, plays an important role in the process of the welfare effects.The following policy implications can be drawn and proposed based on this study:(1) The evaluation of NCMS should be done based on both the impact of NCMS on the residents health consumption and non-healthcare related consumption;(2) the optimal level of NCMS compensation should be to maximize residents’ welfare effects, including the effects on health care consumption and non-medical consumption;(3) The price effect is the key to the welfare effects of NCMS. When increasing coverage, the policy should take measures to determine the prices of health care;(4) the coinsurance rate should be declined to improve both the medical consumption and non-medical consumption level of low-income residents;(5) to make sure that rural residents know about the NCMS policy and trust it, residents’ trust in this scheme can be crucial to the welfare improvement effect of the scheme.Innovation and limitationsWe conducted both a theoretical and an empirical study on the welfare effects of NCMS coverage in China for the first time. The research experience and findings are helpful to lay a foundation for further study and provide recommendations for NCMS design.The innovation of this study:(1) This is the first time to model the comprehensive welfare effects of social health insurance, combining the welfare effects with the precautionary saving theory, and further decompose the welfare effects into health care utilization effects, price effects, quality effects and consumption effect;(2) It is the first time to conduct an comprehensive evaluation focused on the effects of China’s NCMS on the welfare level of residents;(3) We analyze how the knowledge and trust in NCMS help to achieve the goal of welfare improvement by NCMS. The limitations of the research:(1) It is only an impact evaluation of NCMS on the resident’s welfare, and the quality effects has not yet been estimated. So further research should be to determine the optimal coinsurance rate, based on the welfare and utility maximum;(2) In this study, data was from2006and2008. However, in2009China unveiled its health-care reform plan, including a reformation of the NCMS payment system and establishing the essential drugs system, which form the new incentive mechanism for residents and health care providers. Further research is needed to measure the welfare effects of this scheme in the new regulatory environment, especially the price effect.更多还原