Do health insurances reduce catastrophic health expenditure in China? A systematic evidence synthesis

Objective To examine the association of health insurances on catastrophic health expenditure (CHE), and compares that among different health insurances in the last two decades in China. Methods The systematic review was conducted according to the Cochrane Handbook and reported according to PRISMA. We searched English and Chinese literature databases including PubMed, EM base, web of science, CNKI, Wan fang, VIP and CBM (Sino Med) for empirical studies on the association between health insurance and CHE from January 2000 to June 2020. Study selection, data extraction and quality appraisal were conducted by two reviewers. The secular trend of CHE rate and comparisons between population with different health insurances were conducted using meta-analysis, subgroup analysis and meta-regression. Results A total of 4874 citations were obtained, and finally 30 eligible studies with 633917 participants were included. The overall CHE rate was 13.6% (95% CI: 13.1% - 14.0%) from Jan 2000 to June 2020, 12.8% (95% CI: 12.2% - 13.3%) for people with health insurance compared with 16.2% (95% CI:15.4% - 16.9%) for people without health insurance. For types of insurance, the CHE rate was 13.0% (95% CI: 12.4% - 13.6%) for people with new rural cooperative medical scheme (NCMS), 11.9% (95% CI: 9.3% - 14.5%) for urban employees health insurance (UEBMI), 12.0% (95% CI: 8.3% - 15.6%) for urban residents health insurance (URBMI), and 18.0% (95% CI: - 4.5% - 31.5%) for commercial insurance. However, the CHE rate in China has increased in the past 20 years, even adjusted for other factors. The CHE rate of people with NCMS has increased significantly more than people with UEBMI and URBMI. Conclusion In the past 20 years, the basic health insurance plan has reduce the rate of CHE to a certain extent, but due to the rapid increase in medical costs and the release of health needs in recent years, it masks the role of health insurance. More efforts are needed to control unreasonable medical demand and rising costs.

Introduction Every year, about 150 million people in the world spend heavily on health care, and 100 million people are pushed below the poverty line [1]. With rapid population aging in China, the health demands and healthcare expenditure has raising rapidly. However, "it is difficult and expensive to seek medical adviser", has become public concern in the Chinese society [2]. Research found that in 2003, 2008 and 2013, 13.6%,15.1% and 13.8% of households had catastrophic health expenditure (CHE) in China, respectively, of which the poverty rate due to illness was 8.6%. The households with CHE spent an average of 64% of their non-survival expenditure on medical expenses, suggesting the severity of this problem [3].
In order to solve this problem, the Chinese government has made unremitting efforts and made a series of policies involving three basic health insurance schemes, increasing the proportion of reimbursement, making national essential drug lists system, and expanding reimbursement scope of health insurance of new treatments to improve the level of financial protection.
In 1998, the State Council started comprehensive reform of basic health insurances for urban employees (UEBMI) [4]. By 2019, 329.25 million urban workers were involved in the insurance scheme. In rural population, the new rural cooperative medical scheme (NCMS) was piloted in 2002, which was expanded to national wide in 2008, and by 2014, more than 98.9% of rural population was involved (Data from the National Bureau of Statistics). In cities, basic health insurance for urban residents (URBMI) was piloted in 2007. In 2019, 102.51 million urban residents including non-employed or self-employed residents, college students, and children participated in URBMI. Since then, China's basic health insurance system has achieved universal coverage of the population [5].
However, there are some differences among the three basic health insurance systems. First, the financing mechanism is different. UEBMI is jointly paid by employers and individuals. The employers' payment amount is about 6% of employees' wages, while individuals' payment amount is about 2% of employees' wages. In terms of financial level, UEBMI is the highest (2840 yuan in average in 2014), URBMI is in the middle (524 yuan in average in 2014), and NCMS is the lowest (410 yuan in average in 2014). Higher financing level corresponds to higher payment guarantee ability. Second, the compensation level is different. Among them, the out-of-pocket (OOP) payment of UEBMI is the lowest (average 15% in 2015), which compensation ability is the strongest. However, the compensation ability of NCMS (OOP 25% in average in 2015), and URBMI (OOP 30% in average in 2015) was inferior to the former [6].
Third, the reimbursement mechanism is different. UEBMI and URBMI implement a direct settlement method combining personal account and co-ordination fund payment. The health insurance fund regularly pays the insured's personal account for use of outpatient or pharmacy services. The hospitalization expense is paid by the co-ordination fund. The insured only need to pay the OOP when leaving the hospital, and the rest is settled directly by the health insurance fund and the hospital. However, the NCMS adopts the mode that the insured pay in advance when leaving the hospital and being reimbursed after, and the reimbursement process is relatively complex [7]. Fourth, the way of participation is different. For URBMI and NCMS, voluntary participation is adopted, the government will give a certain share of subsidies to encourage more urban residents to participate (S1 Table). All these differences among the three basic insurance schemes may have impact on the ability of preventing CHE for corresponding population.
It has been ten years since the new health reform in China from 2009. Though empirical studies have been conducted to evaluate the effect of health insurance on CHE in China, systematic evaluation of the effect is lacking. Therefore, this study was conducted to systematically examine the effect of health insurance on CHE in China in the last two decades by synthesis of all relevant quantitative empirical evidence.

Methods
A systematic review was conducted according to Cochrane Handbook of Systematic reviews [8] and reported according to the standard of preferred reporting items for systematic reviews and meta-analyses (PRISMA) [9].

Inclusion and exclusion criteria
The inclusion criteria were: (1) participants were residents from mainland China; (2) studies reported the CHE rate and types of health insurance; (3) study design including: randomised controlled trial (RCT), non-RCT, controlled before-after study, cohort study, interrupted time series study and cross-sectional study. Languages were restricted to English and Chinses as they were the main publication languages of studies from mainland China. Duplicate publications were excluded.

Literature search
We systematically searched Chinese literature databases of CNKI, Wan fang and VIP, and English literature databases including PubMed, EMbase, and Web of Science from January 2000 to June 2020. We used terms of "catastrophic health expenditure", "catastrophic medical expenditure", "impoverishing health expenditure", "poverty due to illness" and "returning to poverty due to illness" as keywords (see S2 Table). In addition, reference lists of included studies were scanned for more eligible studies (S2 Table) Study selection Study selection was conducted by one reviewer (YXW) and checked by another (FRL or QQY). First, title and abstract of citations were scanned, obvious irrelevant studies were excluded. Then, full texts of potentially eligible studies were read and selected according to the inclusion and exclusion criteria. Different opinions were decided through discussion or consultation of a senior reviewer (KZ or DDC).

Data extraction and quality evaluation
We used standardized Excel form to extract data from included studies including: study author, publication year, region, participants, sample size, income level, types of health insurance, definition of CHE, CHE rate, and odds ratio (OR). One reviewer (YXW) evaluated the quality of included studies using the AHRQ scale and the NOS scale, which was checked by a second reviewer (FRL or QQY). Different opinions of researchers were resolved by discussion. If there were multiple reports using the same data source, the most comprehensive publication was used in the analysis.

Statistical analysis
The pooled CHE rate and 95% confidence interval (CI) was estimated using meta-analysis. Heterogeneity of studies was examined using χ 2 test and I 2 (I 2 >50% as significant heterogeneity). Random-effect model was used for meta-analysis when there was significant heterogeneity, otherwise fixed-effect model was used.
First, the total CHE rate in China from 2000 to 2020 was estimated. Then, the secular trend of CHE rate was explored. Subgroup analyses were also conducted to investigate the difference of CHE rate between insured and non-insured, and by types of health insurance. Further explore were made on the secular trend of CHE rate among population with different health insurance status. We also performed meta-regression analysis to explore the sources of heterogeneity. The definition of CHE in the study, the year of data collection, whether the study population had special type of disease, whether the study participants were elderly, and types of insurance were included in the meta-regression analysis. Sensitivity analysis was conducted by eliminating the study one by one to examine the influence of each study on the stability of the overall rate. Publication bias was examined using Egger 0 s test. All analyses were performed using STATA15.0.

Study selection
In total, 4874 citations were obtained through systematic literature search, among which 585 citations were included after preliminary screening of titles and abstracts. Finally, 30 studies were included after reading the full texts (Fig 1).

Characteristics of includes studies
The publication year was from 2000 to June 2020, including 22 studies for residents with NCMS, 9 studies for UEBMI, 9 studies for URBMI, 1 study for commercial health insurance, and 30 for non-insured. Six studies were conducted nationwide, 3 were cross-provinces including Hunan, Shandong and Ningxia, and 20 in individual province or metropolises including Shanghai, Shijiazhuang, Beijing, Hangzhou or others. For participants, 12 studies included general population, 2 involved rural population or insurers of NCMS and 13 included hospital patients of major diseases.
Among all, 9 studies used the WHO definition of CHE as 'The proportion of family's medical and health expenses paid in cash to family's consumption expenditure is more than 40%' [9], 4 studies defined CHE as 'The total expenditure of family health accounts for more than 40% of the family's affordability' and 12 studies defined CHE as 'The proportion of self-paid medical and health expenses in the non-food consumption expenditure is more than 40%'. After discussion, we considered that 15 studies may overestimate the rate of CHE compared to the WHO definition of CHE, and one study may underestimate it [10].
For cross-sectional study, there were 6, 6 and 7 studies with NOS quality scores of 3, 4 and 5, respectively, and 9 studies with 6 (full score of 11). The quality score of the 1 cohort study was 5 (full score 9). Overall, the quality of the studies included was common ( Table 1).

Rate of catastrophic health expenditure in the last two decades
There was significant heterogeneity among included studies, thus random-effect model was used. The rate of CHE of Chinese residents was 13.6% (95% CI: 13.1% -14.0%) from 2000 to 2020 (Fig 2).

Catastrophic health expenditure rate and health insurance
Subgroup analysis showed that the rate of CHE in people who with any health insurance (12.8% 95%CI: 12.2% -13.3%) was significantly lower than who without (16.2%, 95%CI: 15.4% -16.9%) (Fig 3). Therefore, health insurance can reduce the rate of CHE (S1 Fig).

Secular trend of rate of health catastrophic expenditure and health insurance
For the people who with health insurances, the rate of CHE was 3.2% ( Since 2010, the rate of CHE in the population with health insurance was significantly lower than those without. Although both of them were on the rise, the rate of CHE in the population without health insurance was increasing faster and the gap between the two became wider (Fig 5).

Secular trend of catastrophic health expenditure by type of insurance
The results showed that the rate of CHE of all subgroups increased with time. The CHE rate of people who did not participate in the health insurance increased more and faster, especially after 2013, followed by NCMS, UEBMI and URBMI. Due to lack of data, the trend of CHE rate of commercial health insurance could not be obtained (Fig 6). 1.34~3.25, p = 0.001) and the rate of CHE, while there was a negative but non-significant association between health insurance (P>0.05, except for commercial health insurance p = 0.028) and the rate of CHE. It is inferred that the heterogeneity between studies is related to the age, diseases, different definitions of CHE and the year, which could explain 43.37% of the heterogeneity between the study results ( Table 2). We further explored the relationship between the growth rate of China's per capita health expenditure and per capita disposable income in recent years, and found that in recent years, China's per capita health expenditure and per capita disposable income have increased to varying degrees, and the growth rate of per capita health expenditure was much higher than that of per capita disposable income. It shown that the excessive growth of health expenses may be an important reason for the occurrence of CHE (S2 Fig).

Publication bias
Egger's test indicated that the studies included had a significant publication bias.

Discussion
There were five main findings of this study. First, health insurances provided financial protection against CHE in China in the last two decades. Second, the financial protective effect against CHE was strongest in UEBMI, followed by URBMI, and NCMS. Third, the rate of CHE for Chinese residents has gradually increased, but the growth rate of people who with health insurance was slower than who without. Fourth, the gap between NCMS and the other two types of health insurances to prevent CHE was increasing. Fifth, major illness and old age were associated with increases risk of CHE.
The rate of CHE is significantly different between those who participate in health insurance and those who do not. However, the rate of CHE has been increasing still, whether participating in or not in health insurances. This was consistent with the finding of Wang and Xu, who  reported that basic health insurance schemes did not reduce the risk of CHE and may even increase it using China Health and Retirement Longitudinal Study data [40,41]. It may because the protective effect of health insurances has been offset by the rapid rise in medical expenditure, caused by developing economics [42], population aging [43,44], increases healthcare needs, the use of new drugs and new materials [45], induced excessive medical services [46,47] which runs counter to the original intention of health insurance [48], service utilization and rising price of medicine, devices and medical technologies [4,49], and lack of health expenditure control measures [45]. The increase in medical expenditure includes reasonable and unreasonable ones.
We also found that the growth rate of per capita health expenditure was much higher than the growth rate of per capita disposable income in the last decades, under this circumstance, any medical insurance plan will face a huge risk of insolvency [50].
Another important reason is that the scope of medical security in my country is currently limited. Outpatient services and some medicines are not included in the reimbursement scope. Because of the limited, households having a chronically ill member may refrain from seeking care until advanced illness sets in. Extending insurance coverage to long-term care for chronically ill patients, outpatient services, routine essential drugs and rehabilitation services should be a priority [19].
For secular trend, the rate of CHE for all three types of health insurances were on the rise, for which NCMS was the most obvious, indicating that the protective effect of NCMS on CHE is the least. The main reason may due to the inconsistency in the system design of the three insurance schemes [51]. The first reason is the compensation methods of NCMS in China [24]. Primary medical institutions are incapable of treating diseases that cause CHE, and must go to high-level medical institutions, but the reimbursement ratio of high-level medical institutions is lower than that of primary medical institutions [52]. Therefore, the risk of CHE is increased.
Second, NCMS insured are rural residents and there has been inequity of access to health services in rural China compared with urban residents. The lack of medical resources and people's poor awareness of disease prevention and control may delay the diagnosis of diseases until very severe or advanced stage, which required higher medical costs than early detected ones [53].
We also found that patients suffering from major diseases and old age were associated with higher rate of CHE. This is consistent with the research finding from Yu Zhen and colleagues in Zhao Tong City, Yunnan Province [54]. Because the elderly was more likely to suffer from chronic diseases, long-term medication and treatment may lead to higher medical expenses.
In order to control and reduce the rate of CHE, policies and measures must be made to control the unreasonable increase of total medical expenses and increase the reimbursement of high-cost diseases. Secondly, measures are needed to unified health insurance systems for serious illnesses, so that severely ill patients may get more economic protection and reduce the risk of CHE.

Strengths and limitations
This study had several strengths. First, as our best knowledge, this was the first systematical review examining the association of health insurance and CHE in China. Second, this study was performed complying to international recognized methodological guideline and standards which provided by far the most comprehensive evidence of health insurance and CHE in China. However, there were several limitations. First, there was large heterogeneity between study included, which could not be solely explained by varied study characteristics. Second, the number of studies for some of the subgroups was small, and with high risk of publication bias, which may affect the reliability of the results.

Conclusion
In the past 20 years, the Chinese government has adopted a series of measures to reduce the rate of CHE. The coverage and reimbursement ratio of health insurance have been improved and show protective role against CHE compared with no insurance, but the rate of CHE have not reduced significantly. This is mainly caused by reasonable and unreasonable increase in medical costs, increased disease burden and limited reimbursement scope. Health policies and measures are warranted to control unreasonable medical costs and further improve the health insurance reimbursement system. In terms of research, more research on the impact of the levels of financial protection on CHE, and the impact of integration of different types of health insurance on CHE is needed.