Multi-Level Social Health Insurance System in the Age of Frequent Employment Change: The Urban Unemployment-Induced Insurance Transition and Healthcare Utilization in China

Job tenure has been significantly shortened with the prevalence of the gig economy around the world. Workers are faced with a new age of frequent employment change. This emerging situation is out of expectation of social health insurance policymakers. As the multi-level social health insurance system in China is closely associated with employment status; urban workers cannot enjoy the urban employee basic medical insurance (UEBMI) during the unemployment period. At this time, unemployed rural-to-urban migrant workers can only rely on the new cooperative medical scheme (NCMS) and unemployed urban residents can only rely on the urban resident basic medical insurance (URBMI). This study provides a preliminary analysis on healthcare utilization change triggered by the unemployment-induced social health insurance transition that has never been investigated. Using the data of a nationwide survey, empirical results show that the unemployment-induced social health insurance transition can significantly deteriorate the healthcare utilization of insurance beneficiaries experiencing the transitions from the UEBMI to the NCMS (or from the UEBMI to the URBMI). Specifically, the outpatient service quality and the conventional physical examination become worse, and the out-of-pocket expenditure increases. Therefore, the multi-level social health insurance system currently in effect can expose workers to a high risk of insufficient health security in the age of frequent employment change.


Introduction
A growing number of enterprises and institutions in China are adapting to the trend of gig economy and tending to hire an increasing number of temporary staff instead of permanent staff. Reports published by LinkedIn, a famous business-oriented social networking service, show that the average job tenure has reduced to 22 months in China [1], and the average tenure for the first job has even reduced to seven months for the post-1995s generation [2]. Workers are faced with a new age of frequent employment change, which they have never experienced before. For policy makers of social health insurance, this emerging situation is out of expectation. It may produce a serious time, becoming a reliable and widely-accepted solution for unemployed people for maintaining a level of healthcare utilization. More importantly, due to being closely associated with employment status, Chinese multi-level social health insurance system can trigger passive health insurance transition in the context of frequent employment change, making people in temporary unemployment status not entitled to benefits of health insurance with wider coverage, but have to turn to others with narrower coverage. Thus, it is particularly necessary to investigate how this issue presents in China.
This paper is structured as follows. In the following sections, this paper first describes the foundation laid by previous research, and the motivation for initiating this study. Then, this paper discusses the status quo of China's social health insurance system, and based on this, analyzes the potential effect of unemployment-induced insurance transition on health utilization in the context of frequent employment change. In the following methods section, this paper uses the data of a nationwide survey (China health and retirement longitudinal study, CHARLS) to empirically examine the effect of health insurance transition from the UEBMI to URBMI or NCMS on health utilization. Meanwhile, this paper makes a detailed description of measures of variables and offers the technical details to clarify the reasonability of the heterogeneously robust linear regression approach applied in this study. Finally, in the discussion and conclusion sections, this study concludes that the current social health insurance system in China has not coped well with the emerging problem of frequent insurance transition triggered by the frequent employment change. The rural-to-urban migrant workers who experience the transition from UEBMI to NCMS are found suffering a far more severe deterioration of healthcare utilization than regular urban residents who experience the transition from UEBMI to URBMI. The current social health insurance schemes thus need to be integrated to meet the new requirements. Future research directions and implications of the present research have also been discussed.

The Motivation of the Theme
Previous research has laid a solid foundation for this study. Some of them focus on one kind of social health insurance to explore its effect on health outcomes (e.g., [20,21]) or to make a comparison between before and after its implementation (e.g., [22]), some focus on the specific provinces or cities, or specific groups to examine the influence they get from the current social health insurance (e.g., [23,24]). Also, a few of research provides an overview of the current social health insurance (e.g., [25]), and compares the difference between the different kinds of it (e.g., [26]), and some others concern the problems generated by this system, such as moral hazards, adverse selection (e.g., [27]), or benefit distribution (e.g., [28]). These devoted efforts help better understand the actual role of the current social health insurance and the design of this system itself. However, with the phenomenon of frequent employment change emerging in Chinese labor market, the findings of previous studies cannot be used to explain the new issue of passive unemployment-induced social health insurance transition and its health outcomes. Thus, it is required to give an empirical investigation on this never explored issue and, in doing so, helps improve the development of China's multi-level social health insurance system.

The Status Quo of Social Health Insurance in China
To briefly summarize the differences between UEBMI, URBMI and NCMS it can help to better understand the problem of unemployment-induced social health insurance transition.
(1) First, the target population is different.
The UEBMI mainly benefits the urban employed [29], while the URBMI mainly serves unemployed people who have urban household registration, and the NCMS is designed for people with rural household registration [30,31].
(2) Second, the payment requirements and funding source are different.
The UEBMI is mandatory for the urban employed and jointly paid by employers and employees [32], while the URBMI and NCMS are both voluntary [33][34][35][36] and enjoy government subsidies on the basis of individual contribution (plus collective contribution for the NCMS) [30]. Proportional to personal wages, the level of payment for UEBMI is generally higher than that for the UEBMI and NCMS [32]. The UEBMI sets a minimum length of time for payment. For those having reached it, there is no need for further payment after retirement [37]. Instead, the URBMI and NCMS need to be paid annually without any predetermined length of time.
(3) Third, the coverage level of treatment is different.
Since the level of funding for URBMI and NCMS are relatively lower, the general level of their medical treatment is shown lower than that of UEBMI's [29,38]. The drug reimbursement list of UEBMI is extensive [39,40], whereas that of NCMS and URBMI is relatively fragmented and varies across different provinces, since many provinces create their own separate and smaller list [41]. For the higher level of benefits, UEBMI beneficiaries are found more likely to take medicines than URBMI and NCMS beneficiaries [42].
(4) Finally, the reimbursement level is different.
The reimbursement rate and reimbursement cap enjoyed by URBMI and NCMS beneficiaries are lower than that for UEBMI [29,32,[43][44][45]. Moreover, the reimbursement rate for NCMS is higher in rural township hospitals than in municipal or above hospitals, and the reimbursement procedure in municipal or above hospitals is more complicated for NCMS beneficiaries [46]. All of this causes a barrier for NCMS beneficiaries seeking better healthcare services [10,47,48]. More details are provided in Table 1.

Target population
Urban employed population, including employed by organizations, self-employed and retirees [29].
Unemployed people with urban household registration, including unemployed elderly residents, low-income people living on basic living allowances, severely disabled people, students and children, etc. [30,31].
People with rural household registration, including rural-to-urban migrant workers, and regular rural residents [30,31].

Funding source
Jointly paid by employers and employees, without government subsidies [30].
With support of modest government subsidies on the basis of personal contribution [30].
Funded by individuals, collectives and the government [30].
With a minimum length of time for payment (25 years for males and 20 years for females), and there is no need for further payment after retirement if the length of time is reached [37].
Without a minimum payment period and paid annually.
Without a minimum payment period and paid annually.
The level of payment is related to personal wages with the employer and employee undertaking 8% and 2% of the wage level, respectively [32].
The level of payment mainly relies on individual contributions, with government subsidies up to 36% [47].
The payment is based on the economic level of provinces and cities; some kinds of common payment level are CNY 100, CNY 200, CNY 300, CNY 400, or CNY 500 per year.
The level of payment for UEBMI is generally higher than that for URBMI and NCMS [32].

Coverage level of treatment
Its drug coverage is always more extensive, with the drug reimbursement list including all the 307 essential medicines and an increasing number of chemical and biologic medicines and traditional Chinese medicines [39,40].
Its drug coverage is comparatively fragmented and varies across different provinces since many provinces create their own separate and smaller list [41].
The drug coverage is comparatively fragmented and varies across different provinces since many provinces create their own separate and smaller list [41].
NCMS's reimbursement rate is higher in rural township hospitals than in municipal hospitals; with 84.04%, 61.25% and 47.71% reimbursement rate respectively at township, county and above medical institutions by the end of 2014 [46]. UEBMI = urban employee basic medical insurance. NCMS = new cooperative medical scheme. URBMI = urban resident basic medical insurance.

Unemployment-Induced Insurance Transition and Health Utilization in the Context of Frequent Employment Change
As mentioned above, the UEBMI is superior in many aspects to the URBMI and the NCMS. For the UEBMI, its stable and reliable funding source can ensure its higher funding level, which in turn can support its high level of treatment coverage for the beneficiaries [43,50]. In term of expenditure on healthcare services provided by municipal or above hospitals, the UEBMI sets a higher reimbursement rate and a higher reimbursement cap line than the other two insurance systems do, which not only alleviates the financial burden of medical expense by the beneficiaries, but also encourages access to the high quality of healthcare services [43,51]. Therefore, the unemployment-induced insurance transition from the UEBMI to the URBMI or NCMS can damage the healthcare utilization (e.g., the decrease in access to the high-quality healthcare services, and the increase in the out-of-pocket expenditure of healthcare services). Further, compared with the transition from the UEBMI to the URBMI, the transition from the UEBMI to the NCMS can further deteriorate the healthcare utilization of people, since the coverage scope is much narrower for NCMS beneficiaries [52][53][54]. Besides, the lower reimbursement rate in municipal hospitals compared to township hospitals can deter the NCMS beneficiaries from seeking better healthcare services. The unmet inpatient need has gone up to 27.9% for NCMS beneficiaries until 2008 [49], and still been great until now.

Study Design
The data of China health and retirement longitudinal study (CHARLS-2015) is used in this study. CHARLS is a widely-used nationwide survey using stratified random sampling, and it collects a high-quality and nationally representative sample of Chinese residents over the age of 45. This survey is jointly initiated by the Center for Social Science Survey in China and the Youth League Committee of Peking University, and is implemented by the National Development Institute of Peking University. CHARLS-2015 is made public since 2017. This survey covers 150 counties, including 450 communities and villages across 28 provinces and municipalities in China, among which about 52.6% are rural areas and 47.4% are urban areas. The sample has covered about 12,400 households in total. This population surveyed is representative of people who are more vulnerable to diseases and more sensitive to the social health insurance transition.

Variables
In this study, three critical dependent variables reflecting healthcare utilization (i.e., outpatient service quality, conventional physical examination and outpatient service cost) are examined. Respondents in this study are those who are the insured of UEBMI in 2013, but turn out to be uninsured  Table 2.
Gender, age and the chronic disease history that may affect healthcare utilization have been controlled in the regression analysis. The chronic disease history includes hypertension, dyslipidemia, diabetes, cancer, chronic lung diseases, liver disease, heart disease, stroke, kidney disease, stomach or other digest disease, emotional or psychiatric problem, memory related disease, arthritis or rheumatism, and asthma. More details about demographic variables and the chronic disease history have been shown in Table 3.

Analytic Strategy
The dataset used in this study is merged from three different data subsets: "healthcare and insurance", "health status and functioning" and "demographic background", according to "individual identity (ID)". The unmatched observations are dropped. The statistical software Stata 13.1 is used in the data analysis. The heterogeneously robust linear regression is applied in this study. Standard errors have been clustered on community level and household level, respectively. The advantage of this method lies in that the estimates can still remain consistent and efficient when individuals could not comply with the independent and identically distributed (I.I.D.) condition. Considering respondents from the same community (or household) are more likely to be homogeneous, the heterogeneously robust linear regression would compute the robust standard errors clustered on community (or household) level instead of the common one to judge the statistical significance. In this way, the estimates are reliable.     Table 4 show that the unemployment-induced social health insurance transition would have negative effect on healthcare utilization. The unemployment-induced transition of UEBMI from active to inactive status would lead to the deterioration of healthcare utilization, which is manifested as the decline of outpatient service quality, the decrease in the number of items of the purchased conventional physical examination, and the increase in out-of-pocket expenditure. What is worth noticing is that such deterioration still exists and is significant, even though people are insured with the NCMS (−1.1027 and −0.8709, outpatient service quality; −3.3827 and −2.1574, conventional physical examination; 0.1596 and 0.1448, out-of-pocket expenditure, all above with p < 0.05 for people in the condition of UEBMI: Yes→No and NCMS: Yes→Yes and UEBMI: Yes→No and NCMS: No→Yes respectively).

Results of
Similar empirical results have been found for the transition between UEBMI and URBMI (see Table 5), although the deterioration of healthcare utilization is slighter than that between UEBMI and NCMS. Specifically, the outpatient service quality and physical examination do not significantly become worse for unemployed urban residents whose URBMI is from inactive to active (UEBMI: Yes→No and URBMI: No→Yes). Besides, the outpatient service quality and out-of-pocket expenditure do not get significantly worse for unemployed urban residents whose URBMI is from inactive to inactive (UEBMI: Yes→No and URBMI: No→No).

Discussion
It is generally known that job tenure has been significantly shortened with the prevalence of the gig economy in China, and a long-termed job has become less available. This new situation leaves the social health insurance system in force lagging behind the status quo of China's economy. The mismatch between the continuing healthcare demand and the health insurance transition during the short unemployed gap period starts to emerge. The current social health insurance system does not cope well with this problem.
In the past, the multi-level social health insurance system is considered to have well covered different types of population. The employed people are covered by UEBMI, and the unemployed people can also enjoy health insurance of URBMI and NCMS. The employment change is not so frequent as before, and the social health insurance system in force does not take the employment change as a matter that frequently happens. Neither the URBMI nor the NCMS is particularly designed for people that experience short unemployed gap periods. Thus, the frequent social health insurance transition resulting from the employment change brings not only the great managing difficulty for beneficiaries, but also huge administrative costs for administrative agencies.
Moreover, the conflict between the continuing demand for healthcare utilization and the passive social health insurance transition can lead to severe consequences. For example, continuing healthcare is particularly in need for chronic disease patients. However, during the short unemployed gap period, the healthcare utilization can be suppressed, and some necessary treatments have to be broken off for saving out-of-pocket expenditure. As such, the patients have to adjust their treatment demand to make it covered by the transitional social health insurance rather than in accordance with their illness status. Even though people can restore their beneficiary status of UEBMI after getting the next job, the interruption or suppression of healthcare utilization within the short unemployed gap period can still have serious impact on chronic disease patients.
Further, the rural-to-urban migrant workers are more vulnerable to the social health insurance transition. Although the healthcare utilization is also damaged for regular urban residents that experience the transition from UEBMI to URBMI, the deterioration of healthcare utilization is much more severe for rural-to-urban migrant workers who experience the transition from UEBMI to NCMS. The difference in benefit package between the two schemes (URBMI and NCMS) can partially lead to this situation. And also, the over-concentration of healthcare resources and the resulting great rural-urban disparity in healthcare access can be the culprit of this asymmetry.

Conclusions
This study is the first attempt to assess the impact of health insurance transition induced by employment change on healthcare utilization. It adds important new knowledge to the existing empirical evaluation of health insurance system. Results further provide empirical support to the idea that the emerging gig economy and the resulting frequent employment change can produce a new challenge to the social health insurance system in China. For policy makers of social health insurance, it is necessary to establish a proper transferring mechanism to integrate UEBMI, URBMI and NCMS. This mechanism needs to link with the original health insurance system, whereby the barrier to receiving NCMS reimbursement for most of rural-to-urban migrant workers could be well dealt with. To successfully build this mechanism, the cohesion between different social health insurance schemes needs to be strengthened through solving some problems like payment age conversion, medical funding compensation for local areas. Moreover, it is helpful to construct an integrated health insurance information platform that implements the centralized management of health information and compensation schemes of various diseases for all kinds of the insured. For those without sufficient conditions for integration of different social health insurance schemes, this practice can serve as the first move towards this integration. Further, the urban-rural dual structure of current social health insurance in China is designed for ease of administration. However, the separate healthcare administration for different social health insurance schemes has now hindered the resolution of emerging issues. Thus, it is essential to unify the administration to facilitate the integration of health insurance, and also to alleviate the financial pressure of operation. Finally, for now, the mitigation measure is to set the buffer period, for example, by 12 months since unemployment, during which the entitlement to UEBMI benefits can be retained. This effort may minimize the damage to health utilization during the period of health insurance transition.
This study is not free of limitations. The nationwide survey used in this study simply documents the information of health insurance transition during the year of 2013-2015, which cannot be used to reveal the dynamics of its effect on health outcomes in a long term. With the subsequent data gradually made public, the effect of a health insurance transition on health outcomes can get the dynamic monitoring. Thus, as more relevant data become available, future research can use new data to provide a comprehensive exploration of this issue.

Conflicts of Interest:
The authors declare no conflict of interest.