Achieving equitable social health insurance benefits in China: how does domestic 1 migration pose a challenge? 2

17 Background: Universal coverage through social health insurance is promoted by many 18 researchers and policymakers to improve health equity within a country or region. In 19 China, the mass internal migration since 1980s has posed challenges for the social health 20 insurance to ensure equitable benefits for migrant population. This study evaluated the 21 benefit distribution of social health insurance among internal migrants in China. 22 Methods: Using the 2014 China National Internal Migrants Dynamic Monitoring 23 Survey, , by applying a two-part model, we conducted a benefit analysis of social health 24 insurance among a representative sample of migrants in China. 25 Results: The study found that the broader the geographic scope of migration, the lower 26 the probability of receiving reimbursements from the social health insurance and the 27 reimbursement ratio; but among those who received reimbursements, the broader the 28 geographic scope of migration, the larger amounts they were reimbursed for health care 29 use. We attributed this unequal benefit distribution to the current insurance design that 30 replies on localized administration and patients paying services up-front and requesting 31 reimbursement later. 32

benefits across health insurance programs and regions. The unequal benefits in insurance 48 might be amplified among the migrant population for the following reasons [5]. 49 Firstly, the social health insurance system consisted of three separate insurance Chinese government issued policies to allow migrants to enroll in UEBMI or URBMI 63 depending upon their employment status [9], and thus migrants may be enrolled in any of 64 the three insurance programs. This may result in the variation in insurance benefits 65 among migrants. 66 Secondly, the social health insurance programs are administrated and financed by 67 local county or city government [10]. Each county or city designs its own benefit 68 package which mainly covers health services delivered within the county / city, and 69 generally does not reimburse health services delivered outside of the county / city [8]. 70 Even if some counties / cities cover health services provided outside of the county / city, 71 lack of insurance portability across regions could create an additional barrier for 72 insurance enrollees to receive the benefits. Migrants may enroll in health insurance at 73 one place, but move to another place and receive health services at this new place, which 74 thus results in separation between the location of health care use and that of health 75 insurance coverage. For many internal migrants, they need to go back to their place of 76 origin to get insurance reimbursement. residence for more than one month but do not have a "Hukou" of the city (registered 107 resident certificate).

108
In this survey, a stratified multi-stage random sampling method by Probability   In this analysis, benefits of health insurance were measured using three outcomes: the 126 probability of receiving reimbursements from social health insurance, total amounts of 127 8 reimbursement received, and the percentage of reimbursements of total health care 128 expenditures (reimbursement ratio). In the survey, we identified the first outcome by a 129 multiple choice question "where did you receive reimbursement for your last 130 hospitalization in this year", answers including: allowance from NCMS, allowance from 131 UEBMI, the employer, the NCMS office, the local health centers, the commercial 132 insurance, allowance from the Family planning operation, the Family planning operation, 133 and else. We recognized those who answered only "the commercial insurance" or "else" 134 or both "the commercial insurance" and "else" as receiving no reimbursement from 135 social health insurance. We identified the second outcome by the question "how much 136 reimbursement did you receive from social health insurance. We identified the last  widowed, divorced or never married). Socioeconomic status was measured by 150 educational attainment, monthly household income per capita, whether the respondent 151 had a job, whether the respondent had rural "Hukou", and whether the respondent lived 152 in urban areas. Educational attainment was categorized into four subgroups: primary 153 school and below, junior high school, senior high school, and college degree and above.

154
"Hukou" represents the record in the residency registration system in China; people can 155 be registered as having either a rural or urban "Hukou" at birth and cannot be easily categorized into four groups: less than one year; one to five years; five to ten years; and 160 ten years and above. Finally, the facility level for hospitalization included primary care 161 facility, secondary hospital, tertiary hospital, and private hospital.

162
Statistical analysis 163 We first described the general characteristics of our study sample. Chi-square test and 164 one-way variance analysis were used to compare the differences of the probability of 165 receiving reimbursement, the amount and ratio of reimbursement received according to 166 the geographic scope of migration.

167
Since there were many "zero observations" -patients who used inpatient care but 168 received no reimbursement, we used the two-part model to estimate the benefits 169 migrants received from the social health insurance, which can be expressed as follows: level for hospitalization as mentioned above.

186
The above two-part model assumes that the benefit migrants received from the social 187 health insurance is determined by two separate decision making process: equation (1)  the lower the probability that migrants would receive reimbursements; but among those 234 who received reimbursements, those who migrated across cities or across provinces 235 received the larger amounts of reimbursement than those who migrated within a city.

236
There was no significant difference in total expenditure and reimbursement ratio by the 237 13 geographic scope of migration.  According to the association between insurance benefit and migration scope, the 250 geographic scope of migration significantly reduced the probability of receiving 251 reimbursement and the reimbursement ratio, but increased the reimbursement amounts 252 they received. Specifically, the probability of receiving reimbursement for those who 253 migrated across cities and provinces was 14.7% and 26.0%, respectively, lower than 254 those who migrated within a city (P<0.01). However, they received 33.4% and 27.2% 255 higher amount of reimbursement than those who migrated within a city (P<0.01). And 256 those who migrated across provinces had the lowest reimbursement ratio (P<0.10).

257
In addition, there was no significant difference in insurance benefits by age, gender,  (Table 4). The relationships between 271 migration scope and the probability of receiving reimbursement and reimbursement 272 amount did not change, while it differed in its relationship with reimbursement ratio.

273
Those who migrated more broadly had a significantly lower reimbursement ratio among 274 NCMS enrollees, but had a significantly higher ratio among URBMI & UEMBI 275 enrollees.

276
Reasons for not receiving reimbursement from social health insurance 277 We further investigated the reasons why migrants did not receive reimbursement from 278 their social health insurance. Figure 1 showed that the need or plan to go back to   reported that did not get reimbursement because they needed to get reimbursement from 322 their hometowns. In addition, research has shown that many services were not 323 reimbursed and the reimbursement process was much more complex for migrants than 324 for local residents [19]. One study showed that more migrants were treated in a hospital 325 that outside of the NCMS designated network than local residents, and thus their 326 healthcare use were less likely to be covered by NCMS [19].

327
Our findings also showed inequity in insurance benefits among migrants by migration  and DZwho all are aware of its content and approve its submission. This manuscript has 410 not been published, and is not under consideration by another journal.

411
Availability of data and materials 412 The data analyzed during the current study are not publicly available because they 413